diff --git a/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/alexnet2_cifar10/alexnet2_cifar10.txt b/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/alexnet2_cifar10/alexnet2_cifar10.txt
new file mode 100644
index 0000000000000000000000000000000000000000..6ec4a06d3dbd2e088d6db287d23dd3bd5aad7ddb
--- /dev/null
+++ b/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/alexnet2_cifar10/alexnet2_cifar10.txt
@@ -0,0 +1,419 @@
+1114.3009809999999
++++++
+conf1 1 1 84.98 0.0
+1 gpu conv fp32 11 add fp32 1 tanh fp32 1
+2 gpu conv fp32 11 add fp32 1 tanh fp32 1 pool_max fp32 1
+3 gpu conv fp32 11 add fp32 1 tanh fp32 1
+4 gpu conv fp32 11 add fp32 1 tanh fp32 1 pool_max fp32 1
+5 gpu conv fp32 11 add fp32 1 tanh fp32 1
+6 gpu conv fp32 11 add fp32 1 tanh fp32 1 pool_max fp32 1
+7 gpu mul fp32 11 add fp32 1
+8 gpu softmax fp32 1
+-----
++++++
+conf2 2.4248748377353113 2.0815908534183163 84.5 0.480000000000004
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 157 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf3 2.4055188425519614 2.0586265720811823 84.48 0.5
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 266 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 269 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 157 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf4 2.4156140842962985 2.0617867479342706 84.28 0.7000000000000028
+1 gpu conv perf_fp16 157 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 267 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 163 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf5 2.396416918342732 2.0506214971794585 84.02 0.960000000000008
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 151 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 157 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf6 2.463002582910052 2.1171077568609458 83.84 1.1400000000000006
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 266 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 167 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf7 2.360283215266004 2.0255245321874304 83.78 1.2000000000000028
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 157 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf8 2.4140791541736157 2.0671513522247653 83.74000000000001 1.2399999999999949
+1 gpu conv fp16 11 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 267 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 160 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf9 2.457753689612079 2.1086250651240137 83.7 1.2800000000000011
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 266 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 163 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf10 2.459170454055443 2.1111925341396343 83.7 1.2800000000000011
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 266 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 164 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf11 2.4135986141645764 2.060453960420927 83.62 1.3599999999999994
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 267 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 266 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 157 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf12 2.4631278039012106 2.1092094797926637 83.58 1.4000000000000057
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 266 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 157 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf13 2.535761391794481 2.16998336112692 83.58 1.4000000000000057
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 266 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 157 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf14 2.289006193945062 1.961240158652051 83.54 1.4399999999999977
+1 gpu conv perf_fp16 167 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 155 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 157 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf15 2.4257674844112573 2.0808440756495563 83.5 1.480000000000004
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 161 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf16 2.458122368488622 2.109531159729078 83.48 1.5
+1 gpu conv fp16 11 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 162 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf17 2.281072202152105 1.9539314420536427 83.46000000000001 1.519999999999996
+1 gpu conv fp16 11 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 160 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 160 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf18 2.4572171342078444 2.1088933553775697 83.46000000000001 1.519999999999996
+1 gpu conv fp16 11 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 163 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf19 2.3017607719030058 1.9782265708150768 83.42 1.5600000000000023
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 162 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 261 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 162 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf20 2.379206814483014 2.047909200292713 83.39999999999999 1.5800000000000125
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 266 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 151 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf21 2.4636282705302537 2.1162281156388527 83.39999999999999 1.5800000000000125
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 160 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf22 2.461590101374146 2.1108493881199184 83.22 1.7600000000000051
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 266 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 161 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf23 2.537054645442804 2.167568834938183 83.22 1.7600000000000051
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 266 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 157 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf24 2.4631604723407885 2.1099694757102845 83.17999999999999 1.8000000000000114
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 267 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 157 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf25 2.4636282705302537 2.1162281156388527 83.14 1.8400000000000034
+1 gpu conv fp16 11 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 160 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf26 2.462588899729088 2.109477918791931 83.14 1.8400000000000034
+1 gpu conv fp16 11 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 157 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf27 2.4638085754689025 2.1071960926343603 83.1 1.8800000000000097
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 157 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf28 2.4640079766123635 2.110326453157297 83.08 1.9000000000000057
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 261 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf29 2.459337622764853 2.107249218450713 83.06 1.9200000000000017
+1 gpu conv fp16 11 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 162 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf30 2.538176340059405 2.173287257415721 83.02000000000001 1.9599999999999937
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 164 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf31 2.3905426931959846 2.044333576277581 83.02000000000001 1.9599999999999937
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 160 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 261 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 157 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf32 2.459337622764853 2.107249218450713 83.0 1.980000000000004
+1 gpu conv fp16 11 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 162 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf33 2.458968579288317 2.1063450826631396 82.89999999999999 2.0800000000000125
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 266 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 163 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf34 2.2912974651603877 1.9670210508860688 82.8 2.180000000000007
+1 gpu conv perf_fp16 168 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 155 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 160 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf35 2.4648489763056327 2.113931670664391 82.66 2.3200000000000074
+1 gpu conv fp16 11 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 160 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf36 2.4599076869402854 2.1077397371200193 82.6 2.3800000000000097
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 267 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 162 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf37 2.4636282705302537 2.1162281156388527 82.54 2.4399999999999977
+1 gpu conv fp16 11 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 160 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
++++++
+conf38 2.591814267389778 2.222680944458784 82.26 2.719999999999999
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1
+2 gpu conv perf_fp16 154 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 261 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1
+6 gpu conv perf_fp16 157 add fp16 1 tanh fp16 1 pool_max fp16 1
+7 gpu mul fp16 12 add fp16 1
+8 gpu softmax fp32 1
+-----
diff --git a/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/alexnet_cifar10/alexnet_cifar10.txt b/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/alexnet_cifar10/alexnet_cifar10.txt
new file mode 100644
index 0000000000000000000000000000000000000000..a9ccba6eb63f620c0e3b6f95fd7c50892018f00f
--- /dev/null
+++ b/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/alexnet_cifar10/alexnet_cifar10.txt
@@ -0,0 +1,511 @@
+2592.187221
++++++
+conf1 1 1 79.28 0.0
+1 gpu conv fp32 11 add fp32 1 tanh fp32 1 pool_max fp32 1
+2 gpu conv fp32 11 add fp32 1 tanh fp32 1 pool_max fp32 1
+3 gpu conv fp32 11 add fp32 1 tanh fp32 1
+4 gpu conv fp32 11 add fp32 1 tanh fp32 1
+5 gpu conv fp32 11 add fp32 1 tanh fp32 1 pool_max fp32 1
+6 gpu mul fp32 11 add fp32 1
+7 gpu softmax fp32 1
+-----
++++++
+conf2 1.7593976485873195 1.6193399031642917 79.23 0.04999999999999716
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv fp16 12 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf3 2.092625440752526 1.9139078015388271 78.96 0.3200000000000074
+1 gpu conv samp_fp16 263 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 266 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv fp16 12 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf4 1.8870195448805414 1.7296919053025768 78.8 0.480000000000004
+1 gpu conv samp_fp16 263 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv fp16 12 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf5 2.1184804041774554 1.9598989563949536 78.75999999999999 0.5200000000000102
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 263 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf6 2.1184804041774554 1.9598989563949536 78.75999999999999 0.5200000000000102
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 263 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf7 2.0933825381386364 1.9150743378318535 78.64 0.6400000000000006
+1 gpu conv samp_fp16 263 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv fp16 12 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf8 2.081712090729918 1.9102226906341664 78.5 0.7800000000000011
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv fp16 12 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf9 2.081712090729918 1.9102226906341664 78.5 0.7800000000000011
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv fp16 12 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf10 2.2662606588487595 2.066560750795139 78.48 0.7999999999999972
+1 gpu conv samp_fp16 264 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv fp16 12 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf11 2.121684761285686 1.966318179285323 78.48 0.7999999999999972
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 263 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf12 2.3417491169395532 2.1355030360671465 78.38000000000001 0.8999999999999915
+1 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv fp16 12 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf13 2.2247938983110425 2.060416584958474 78.38000000000001 0.8999999999999915
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf14 2.2247938983110425 2.060416584958474 78.38000000000001 0.8999999999999915
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf15 2.2247938983110425 2.060416584958474 78.38000000000001 0.8999999999999915
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf16 2.2627828537139263 2.065683616898884 78.32000000000001 0.9599999999999937
+1 gpu conv samp_fp16 263 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv fp16 12 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf17 2.2627828537139263 2.065683616898884 78.32000000000001 0.9599999999999937
+1 gpu conv samp_fp16 263 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv fp16 12 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf18 2.2627828537139263 2.065683616898884 78.32000000000001 0.9599999999999937
+1 gpu conv samp_fp16 263 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv fp16 12 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf19 2.146571989407323 1.95711703610764 78.18 1.0999999999999943
+1 gpu conv samp_fp16 264 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 267 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 266 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf20 2.303316973793268 2.1036463961913276 78.10000000000001 1.1799999999999926
+1 gpu conv samp_fp16 263 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv fp16 12 add fp16 1 tanh fp16 1
+5 gpu conv samp_fp16 267 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf21 2.436875653706139 2.2434837737118056 78.08 1.2000000000000028
+1 gpu conv samp_fp16 264 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf22 2.436875653706139 2.2434837737118056 78.08 1.2000000000000028
+1 gpu conv samp_fp16 264 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf23 2.436875653706139 2.2434837737118056 78.08 1.2000000000000028
+1 gpu conv samp_fp16 264 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf24 2.1106508925330925 1.9419233584234938 78.06 1.2199999999999989
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv fp16 12 add fp16 1 tanh fp16 1
+5 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf25 2.3203534290038634 2.116965679235447 78.06 1.2199999999999989
+1 gpu conv samp_fp16 264 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 267 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf26 2.3527290658539215 2.145832257234814 78.03999999999999 1.240000000000009
+1 gpu conv samp_fp16 261 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv fp16 12 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf27 2.3527290658539215 2.145832257234814 78.03999999999999 1.240000000000009
+1 gpu conv samp_fp16 261 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv fp16 12 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf28 2.432854949808342 2.2424500615508003 78.0 1.2800000000000011
+1 gpu conv samp_fp16 263 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf29 2.432854949808342 2.2424500615508003 78.0 1.2800000000000011
+1 gpu conv samp_fp16 263 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf30 2.432854949808342 2.2424500615508003 78.0 1.2800000000000011
+1 gpu conv samp_fp16 263 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf31 2.3137982135449207 2.1281257317083417 77.84 1.4399999999999977
+1 gpu conv samp_fp16 264 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 265 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf32 2.1198074418988333 1.9522214255218437 77.82 1.460000000000008
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf33 2.246924974355375 2.065289762405701 77.8 1.480000000000004
+1 gpu conv samp_fp16 264 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 269 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf34 2.263614734554485 2.090777846534249 77.74 1.5400000000000063
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 267 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf35 2.263614734554485 2.090777846534249 77.74 1.5400000000000063
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 267 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf36 2.263614734554485 2.090777846534249 77.74 1.5400000000000063
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 267 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf37 2.5289288699015304 2.334007588396142 77.72 1.5600000000000023
+1 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf38 2.5289288699015304 2.334007588396142 77.72 1.5600000000000023
+1 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf39 2.3117594882585775 2.1152397180868943 77.56 1.7199999999999989
+1 gpu conv samp_fp16 264 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 266 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf40 2.452732477854469 2.264573687601476 77.56 1.7199999999999989
+1 gpu conv perf_fp16 167 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf41 2.452732477854469 2.264573687601476 77.56 1.7199999999999989
+1 gpu conv perf_fp16 167 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf42 2.382518688546389 2.178614303992064 77.5 1.7800000000000011
+1 gpu conv samp_fp16 261 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 267 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 267 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf43 2.382518688546389 2.178614303992064 77.5 1.7800000000000011
+1 gpu conv samp_fp16 261 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 267 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 267 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf44 2.3900667100485924 2.188128526401265 77.48 1.7999999999999972
+1 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf45 2.3900667100485924 2.188128526401265 77.48 1.7999999999999972
+1 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf46 2.3900667100485924 2.188128526401265 77.48 1.7999999999999972
+1 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf47 2.4835281673276515 2.279527076032239 77.3 1.980000000000004
+1 gpu conv samp_fp16 264 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 267 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf48 2.4835281673276515 2.279527076032239 77.3 1.980000000000004
+1 gpu conv samp_fp16 264 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 267 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf49 2.1553694968551302 1.9959124044028933 77.18 2.0999999999999943
+1 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 265 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 268 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf50 2.5877520959724816 2.3763616521050364 77.03999999999999 2.240000000000009
+1 gpu conv samp_fp16 261 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 267 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
++++++
+conf51 2.5877520959724816 2.3763616521050364 77.03999999999999 2.240000000000009
+1 gpu conv samp_fp16 261 add fp16 1 tanh fp16 1 pool_max fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 267 add fp16 1 tanh fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 tanh fp16 1
+5 gpu conv fp16 12 add fp16 1 tanh fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1
+7 gpu softmax fp32 1
+-----
diff --git a/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/alexnet_imagenet/alexnet_imagenet.txt b/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/alexnet_imagenet/alexnet_imagenet.txt
new file mode 100644
index 0000000000000000000000000000000000000000..b0e42a5aaa5d7b5a06b6422a5c33a0047b6eff8d
--- /dev/null
+++ b/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/alexnet_imagenet/alexnet_imagenet.txt
@@ -0,0 +1,229 @@
+2739.950736
++++++
+conf1 1 1 56.3 0.0
+1 gpu conv fp32 11 add fp32 1 relu fp32 1 pool_max fp32 1
+2 gpu conv fp32 11 add fp32 1 relu fp32 1 pool_max fp32 1
+3 gpu conv fp32 11 add fp32 1 relu fp32 1
+4 gpu conv fp32 11 add fp32 1 relu fp32 1
+5 gpu conv fp32 11 add fp32 1 relu fp32 1 pool_max fp32 1
+6 gpu mul fp32 11 add fp32 1 relu fp32 1
+7 gpu mul fp32 11 add fp32 1 relu fp32 1
+8 gpu mul fp32 11 add fp32 1
+9 gpu softmax fp32 1
+-----
++++++
+conf2 1.802133644103582 1.8186433204507424 55.76 0.5399999999999991
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1
+5 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1 relu fp16 1
+7 gpu mul fp16 12 add fp16 1 relu fp16 1
+8 gpu mul fp16 12 add fp16 1
+9 gpu softmax fp32 1
+-----
++++++
+conf3 1.7574572103878898 1.7673706184460103 55.58 0.7199999999999989
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1
+5 gpu conv samp_fp16 268 add fp16 1 relu fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1 relu fp16 1
+7 gpu mul fp16 12 add fp16 1 relu fp16 1
+8 gpu mul fp16 12 add fp16 1
+9 gpu softmax fp32 1
+-----
++++++
+conf4 2.0227701930718065 2.043112495268932 55.42 0.8799999999999955
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu conv perf_fp16 166 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1
+5 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1 relu fp16 1
+7 gpu mul fp16 12 add fp16 1 relu fp16 1
+8 gpu mul fp16 12 add fp16 1
+9 gpu softmax fp32 1
+-----
++++++
+conf5 1.9872634777043927 2.002789650227035 55.120000000000005 1.1799999999999926
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu conv perf_fp16 164 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1
+5 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1 relu fp16 1
+7 gpu mul fp16 12 add fp16 1 relu fp16 1
+8 gpu mul fp16 12 add fp16 1
+9 gpu softmax fp32 1
+-----
++++++
+conf6 1.8204253918445088 1.843736069756362 54.84 1.4599999999999937
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 164 add fp16 1 relu fp16 1
+5 gpu conv perf_fp16 154 add fp16 1 relu fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1 relu fp16 1
+7 gpu mul fp16 12 add fp16 1 relu fp16 1
+8 gpu mul fp16 12 add fp16 1
+9 gpu softmax fp32 1
+-----
++++++
+conf7 1.9308336510645352 1.934889049414224 54.74 1.5599999999999952
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu conv perf_fp16 168 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 163 add fp16 1 relu fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1 relu fp16 1
+7 gpu mul fp16 12 add fp16 1 relu fp16 1
+8 gpu mul fp16 12 add fp16 1
+9 gpu softmax fp32 1
+-----
++++++
+conf8 2.0146435217865446 2.0367475358800102 54.58 1.7199999999999989
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu conv perf_fp16 162 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1
+5 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1 relu fp16 1
+7 gpu mul fp16 12 add fp16 1 relu fp16 1
+8 gpu mul fp16 12 add fp16 1
+9 gpu softmax fp32 1
+-----
++++++
+conf9 2.0101709494490696 2.0329911158023064 54.400000000000006 1.8999999999999915
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu conv perf_fp16 164 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 163 add fp16 1 relu fp16 1
+5 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1 relu fp16 1
+7 gpu mul fp16 12 add fp16 1 relu fp16 1
+8 gpu mul fp16 12 add fp16 1
+9 gpu softmax fp32 1
+-----
++++++
+conf10 2.0052132441967916 2.0284931705407003 54.300000000000004 1.999999999999993
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu conv perf_fp16 168 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 163 add fp16 1 relu fp16 1
+5 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1 relu fp16 1
+7 gpu mul fp16 12 add fp16 1 relu fp16 1
+8 gpu mul fp16 12 add fp16 1
+9 gpu softmax fp32 1
+-----
++++++
+conf11 2.010827434817262 2.036001862538864 54.2 2.0999999999999943
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu conv perf_fp16 164 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 163 add fp16 1 relu fp16 1
+5 gpu conv perf_fp16 154 add fp16 1 relu fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1 relu fp16 1
+7 gpu mul fp16 12 add fp16 1 relu fp16 1
+8 gpu mul fp16 12 add fp16 1
+9 gpu softmax fp32 1
+-----
++++++
+conf12 2.019868378233057 2.0433540129730265 54.17999999999999 2.1200000000000045
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu conv perf_fp16 162 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 163 add fp16 1 relu fp16 1
+5 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1 relu fp16 1
+7 gpu mul fp16 12 add fp16 1 relu fp16 1
+8 gpu mul fp16 12 add fp16 1
+9 gpu softmax fp32 1
+-----
++++++
+conf13 1.9923471030291253 2.009177323959059 54.120000000000005 2.1799999999999926
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu conv perf_fp16 164 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 163 add fp16 1 relu fp16 1
+5 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1 relu fp16 1
+7 gpu mul fp16 12 add fp16 1 relu fp16 1
+8 gpu mul fp16 12 add fp16 1
+9 gpu softmax fp32 1
+-----
++++++
+conf14 1.9923471030291253 2.009177323959059 54.120000000000005 2.1799999999999926
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu conv perf_fp16 164 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 163 add fp16 1 relu fp16 1
+5 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1 relu fp16 1
+7 gpu mul fp16 12 add fp16 1 relu fp16 1
+8 gpu mul fp16 12 add fp16 1
+9 gpu softmax fp32 1
+-----
++++++
+conf15 2.028037341700216 2.049760395549724 54.0 2.299999999999997
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu conv perf_fp16 166 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 163 add fp16 1 relu fp16 1
+5 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1 relu fp16 1
+7 gpu mul fp16 12 add fp16 1 relu fp16 1
+8 gpu mul fp16 12 add fp16 1
+9 gpu softmax fp32 1
+-----
++++++
+conf16 1.9910730364852436 2.006510848093771 53.54 2.759999999999998
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu conv perf_fp16 164 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 164 add fp16 1 relu fp16 1
+5 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1 relu fp16 1
+7 gpu mul fp16 12 add fp16 1 relu fp16 1
+8 gpu mul fp16 12 add fp16 1
+9 gpu softmax fp32 1
+-----
++++++
+conf17 2.1567475543719614 2.159142310265706 53.300000000000004 2.999999999999993
+1 gpu conv perf_fp16 164 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu conv perf_fp16 166 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 163 add fp16 1 relu fp16 1
+5 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1 relu fp16 1
+7 gpu mul fp16 12 add fp16 1 relu fp16 1
+8 gpu mul fp16 12 add fp16 1
+9 gpu softmax fp32 1
+-----
++++++
+conf18 2.1567475543719614 2.159142310265706 53.300000000000004 2.999999999999993
+1 gpu conv perf_fp16 164 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu conv perf_fp16 166 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 163 add fp16 1 relu fp16 1
+5 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1 relu fp16 1
+7 gpu mul fp16 12 add fp16 1 relu fp16 1
+8 gpu mul fp16 12 add fp16 1
+9 gpu softmax fp32 1
+-----
++++++
+conf19 2.0232690820426464 2.0527698121318476 53.300000000000004 2.999999999999993
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu conv perf_fp16 168 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 11 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 163 add fp16 1 relu fp16 1
+5 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+6 gpu mul fp16 12 add fp16 1 relu fp16 1
+7 gpu mul fp16 12 add fp16 1 relu fp16 1
+8 gpu mul fp16 12 add fp16 1
+9 gpu softmax fp32 1
+-----
diff --git a/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/lenet_keras/lenet_keras.txt b/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/lenet_keras/lenet_keras.txt
new file mode 100644
index 0000000000000000000000000000000000000000..b4e51dff426f4d3c5cb7b9572e6aa5940212acbd
--- /dev/null
+++ b/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/lenet_keras/lenet_keras.txt
@@ -0,0 +1,409 @@
+282.5141369999999
++++++
+conf1 1 1 98.7 0.0
+1 gpu conv fp32 11 add fp32 1 pool_max fp32 1 tanh fp32 1
+2 gpu conv fp32 11 add fp32 1 pool_max fp32 1 tanh fp32 1
+3 gpu mul fp32 11 add fp32 1 tanh fp32 1
+4 gpu mul fp32 11 add fp32 1 tanh fp32 1
+5 gpu softmax fp32 1
+-----
++++++
+conf2 1.828613181003043 2.071721708828981 98.65 0.04999999999999716
+1 gpu conv perf_fp16 156 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf3 1.8936889628815377 2.139779619692146 98.65 0.04999999999999716
+1 gpu conv perf_fp16 152 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf4 1.8936889628815377 2.139779619692146 98.65 0.04999999999999716
+1 gpu conv perf_fp16 152 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf5 1.8936889628815377 2.139779619692146 98.65 0.04999999999999716
+1 gpu conv perf_fp16 152 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf6 1.8247639611533713 2.0227145446958756 98.64 0.060000000000002274
+1 gpu conv fp16 11 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf7 1.8247639611533713 2.0227145446958756 98.64 0.060000000000002274
+1 gpu conv fp16 11 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf8 1.8406161850501603 2.037849502542524 98.64 0.060000000000002274
+1 gpu conv fp16 11 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf9 1.8406161850501603 2.037849502542524 98.64 0.060000000000002274
+1 gpu conv fp16 11 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf10 1.8406161850501603 2.037849502542524 98.64 0.060000000000002274
+1 gpu conv fp16 11 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf11 1.8663357888260776 2.115790921611576 98.64 0.060000000000002274
+1 gpu conv perf_fp16 155 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf12 1.8663357888260776 2.115790921611576 98.64 0.060000000000002274
+1 gpu conv perf_fp16 155 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf13 1.8663357888260776 2.115790921611576 98.64 0.060000000000002274
+1 gpu conv perf_fp16 155 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf14 1.8645645142051612 2.1037012333044935 98.61999999999999 0.0800000000000125
+1 gpu conv perf_fp16 167 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf15 1.8645645142051612 2.1037012333044935 98.61999999999999 0.0800000000000125
+1 gpu conv perf_fp16 167 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf16 1.8645645142051612 2.1037012333044935 98.61999999999999 0.0800000000000125
+1 gpu conv perf_fp16 167 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf17 2.2168527051833635 2.453341076720038 98.61999999999999 0.0800000000000125
+1 gpu conv samp_fp16 264 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf18 2.2168527051833635 2.453341076720038 98.61999999999999 0.0800000000000125
+1 gpu conv samp_fp16 264 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf19 2.2168527051833635 2.453341076720038 98.61999999999999 0.0800000000000125
+1 gpu conv samp_fp16 264 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf20 1.8406161850501603 2.037849502542524 98.6 0.10000000000000853
+1 gpu conv fp16 12 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf21 1.8406161850501603 2.037849502542524 98.6 0.10000000000000853
+1 gpu conv fp16 12 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf22 1.8406161850501603 2.037849502542524 98.6 0.10000000000000853
+1 gpu conv fp16 12 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf23 1.8406161850501603 2.037849502542524 98.6 0.10000000000000853
+1 gpu conv fp16 11 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf24 1.8406161850501603 2.037849502542524 98.6 0.10000000000000853
+1 gpu conv fp16 11 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf25 1.8406161850501603 2.037849502542524 98.6 0.10000000000000853
+1 gpu conv fp16 11 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf26 2.200653361151419 2.425091789360736 98.6 0.10000000000000853
+1 gpu conv samp_fp16 266 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf27 2.200653361151419 2.425091789360736 98.6 0.10000000000000853
+1 gpu conv samp_fp16 266 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf28 1.8406161850501603 2.037849502542524 98.58 0.12000000000000455
+1 gpu conv fp16 11 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf29 1.8406161850501603 2.037849502542524 98.58 0.12000000000000455
+1 gpu conv fp16 11 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf30 1.8406161850501603 2.037849502542524 98.58 0.12000000000000455
+1 gpu conv fp16 11 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf31 1.8445326456180258 2.087601822059355 98.58 0.12000000000000455
+1 gpu conv perf_fp16 156 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf32 1.8445326456180258 2.087601822059355 98.58 0.12000000000000455
+1 gpu conv perf_fp16 156 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf33 1.8445326456180258 2.087601822059355 98.58 0.12000000000000455
+1 gpu conv perf_fp16 156 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf34 1.8916677984300285 2.155437579874673 98.58 0.12000000000000455
+1 gpu conv perf_fp16 158 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf35 1.8916677984300285 2.155437579874673 98.58 0.12000000000000455
+1 gpu conv perf_fp16 158 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf36 1.8916677984300285 2.155437579874673 98.58 0.12000000000000455
+1 gpu conv perf_fp16 158 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf37 1.8649226857257986 2.1076025277601325 98.56 0.14000000000000057
+1 gpu conv perf_fp16 168 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf38 1.8649226857257986 2.1076025277601325 98.56 0.14000000000000057
+1 gpu conv perf_fp16 168 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf39 1.8649226857257986 2.1076025277601325 98.56 0.14000000000000057
+1 gpu conv perf_fp16 168 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf40 1.8463058650555446 2.067271423078985 98.56 0.14000000000000057
+1 gpu conv perf_fp16 157 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf41 1.8463058650555446 2.067271423078985 98.56 0.14000000000000057
+1 gpu conv perf_fp16 157 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf42 1.8463058650555446 2.067271423078985 98.56 0.14000000000000057
+1 gpu conv perf_fp16 157 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf43 1.9234076467497994 2.1864740913112275 98.56 0.14000000000000057
+1 gpu conv perf_fp16 153 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf44 1.9234076467497994 2.1864740913112275 98.56 0.14000000000000057
+1 gpu conv perf_fp16 153 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf45 1.9234076467497994 2.1864740913112275 98.56 0.14000000000000057
+1 gpu conv perf_fp16 153 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf46 1.8698191484268973 2.13979218727595 98.54 0.1599999999999966
+1 gpu conv perf_fp16 159 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf47 1.8698191484268973 2.13979218727595 98.54 0.1599999999999966
+1 gpu conv perf_fp16 159 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf48 1.8575043605938137 2.092057786757256 98.52 0.18000000000000682
+1 gpu conv perf_fp16 165 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf49 1.8575043605938137 2.092057786757256 98.52 0.18000000000000682
+1 gpu conv perf_fp16 165 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf50 1.8575043605938137 2.092057786757256 98.52 0.18000000000000682
+1 gpu conv perf_fp16 165 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
++++++
+conf51 1.8534621507951072 2.1231113105788597 98.44000000000001 0.2599999999999909
+1 gpu conv perf_fp16 159 add fp16 1 pool_max fp16 1 tanh fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 pool_max fp16 1 tanh fp16 1
+3 gpu mul fp16 12 add fp16 1 tanh fp16 1
+4 gpu mul fp16 12 add fp16 1 tanh fp16 1
+5 gpu softmax fp32 1
+-----
diff --git a/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/mobilenet_cifar10/mobilenet_cifar10.txt b/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/mobilenet_cifar10/mobilenet_cifar10.txt
new file mode 100644
index 0000000000000000000000000000000000000000..b4d8bd893c8d9395fce6a3484d75f543f1e72da2
--- /dev/null
+++ b/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/mobilenet_cifar10/mobilenet_cifar10.txt
@@ -0,0 +1,3220 @@
+4077.307063200001
++++++
+conf1 1 1 84.42 0.0
+1 gpu conv fp32 11
+2 gpu batchnorm fp32 11
+3 gpu relu fp32 11
+4 gpu group_conv fp32 11
+5 gpu batchnorm fp32 11
+6 gpu relu fp32 11
+7 gpu conv fp32 11
+8 gpu batchnorm fp32 11
+9 gpu relu fp32 11
+10 gpu group_conv fp32 11
+11 gpu batchnorm fp32 11
+12 gpu relu fp32 11
+13 gpu conv fp32 11
+14 gpu batchnorm fp32 11
+15 gpu relu fp32 11
+16 gpu group_conv fp32 11
+17 gpu batchnorm fp32 11
+18 gpu relu fp32 11
+19 gpu conv fp32 11
+20 gpu batchnorm fp32 11
+21 gpu relu fp32 11
+22 gpu group_conv fp32 11
+23 gpu batchnorm fp32 11
+24 gpu relu fp32 11
+25 gpu conv fp32 11
+26 gpu batchnorm fp32 11
+27 gpu relu fp32 11
+28 gpu group_conv fp32 11
+29 gpu batchnorm fp32 11
+30 gpu relu fp32 11
+31 gpu conv fp32 11
+32 gpu batchnorm fp32 11
+33 gpu relu fp32 11
+34 gpu group_conv fp32 11
+35 gpu batchnorm fp32 11
+36 gpu relu fp32 11
+37 gpu conv fp32 11
+38 gpu batchnorm fp32 11
+39 gpu relu fp32 11
+40 gpu group_conv fp32 11
+41 gpu batchnorm fp32 11
+42 gpu relu fp32 11
+43 gpu conv fp32 11
+44 gpu batchnorm fp32 11
+45 gpu relu fp32 11
+46 gpu group_conv fp32 11
+47 gpu batchnorm fp32 11
+48 gpu relu fp32 11
+49 gpu conv fp32 11
+50 gpu batchnorm fp32 11
+51 gpu relu fp32 11
+52 gpu group_conv fp32 11
+53 gpu batchnorm fp32 11
+54 gpu relu fp32 11
+55 gpu conv fp32 11
+56 gpu batchnorm fp32 11
+57 gpu relu fp32 11
+58 gpu group_conv fp32 11
+59 gpu batchnorm fp32 11
+60 gpu relu fp32 11
+61 gpu conv fp32 11
+62 gpu batchnorm fp32 11
+63 gpu relu fp32 11
+64 gpu group_conv fp32 11
+65 gpu batchnorm fp32 11
+66 gpu relu fp32 11
+67 gpu conv fp32 11
+68 gpu batchnorm fp32 11
+69 gpu relu fp32 11
+70 gpu group_conv fp32 11
+71 gpu batchnorm fp32 11
+72 gpu relu fp32 11
+73 gpu conv fp32 11
+74 gpu batchnorm fp32 11
+75 gpu relu fp32 11
+76 gpu group_conv fp32 11
+77 gpu batchnorm fp32 11
+78 gpu relu fp32 11
+79 gpu conv fp32 11
+80 gpu batchnorm fp32 11
+81 gpu relu fp32 11
+82 gpu pool_mean fp32 11
+83 gpu mul fp32 11 add fp32 1
+84 gpu softmax fp32 1
+-----
++++++
+conf2 1.4930855091460031 1.447990050940341 83.72 0.7000000000000028
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 163
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 151
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv fp16 12
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf3 1.493397883226807 1.449591062426989 83.72 0.7000000000000028
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 163
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 151
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 163
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf4 1.4934429016801338 1.4500582352111675 83.72 0.7000000000000028
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 163
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 151
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 168
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf5 1.4938214813031556 1.450038222978811 83.72 0.7000000000000028
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 163
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 151
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 157
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf6 1.4933879828131855 1.449975636202813 83.72 0.7000000000000028
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 163
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 151
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 160
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf7 1.492663093331302 1.4487067754520524 83.7 0.7199999999999989
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv fp16 12
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 151
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 167
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf8 1.495724395088184 1.4507925552157772 83.56 0.8599999999999994
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv fp16 12
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 151
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 162
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf9 1.496506307637598 1.4521705950285135 83.36 1.0600000000000023
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 163
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 151
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 162
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf10 1.496532672928805 1.4521696542076958 83.36 1.0600000000000023
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 163
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 151
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 156
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf11 1.4988418058849937 1.4555327556053628 83.28 1.1400000000000006
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 168
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv fp16 12
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 164
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 158
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf12 1.4994289979945077 1.4562439330251535 83.28 1.1400000000000006
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 168
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv fp16 12
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 152
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 153
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf13 1.4952028793065038 1.450369851058777 83.14 1.2800000000000011
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 162
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv perf_fp16 161
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 151
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 155
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 156
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 152
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 152
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf14 1.4933978285280285 1.448265686258097 83.12 1.2999999999999972
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 163
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv fp16 12
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 158
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf15 1.491958833559989 1.4459262032919467 83.08 1.3400000000000034
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv fp16 12
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv perf_fp16 161
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 157
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 155
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv fp16 12
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 152
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 152
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf16 1.4937317297990984 1.4498121856525021 83.02000000000001 1.3999999999999915
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 163
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 156
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 158
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf17 1.4963413808686974 1.4522391736954623 82.86 1.5600000000000023
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 163
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 151
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 165
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf18 1.4942172827099065 1.4504631324933321 82.86 1.5600000000000023
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 163
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 157
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 158
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf19 1.4963964073376739 1.4525461321361477 82.86 1.5600000000000023
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 163
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 151
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 158
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf20 1.4932583049858652 1.4472547227714012 82.84 1.5799999999999983
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 163
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 151
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv samp_fp16 266
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf21 1.4964326545281064 1.4526263046333605 82.82000000000001 1.5999999999999943
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 163
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 152
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 158
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf22 1.4966042483929347 1.4527859961226985 82.82000000000001 1.5999999999999943
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 163
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 152
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 153
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf23 1.4966008974318024 1.4527415844509437 82.78 1.6400000000000006
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 163
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 155
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 158
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf24 1.4932738366973777 1.448820445466833 82.64 1.7800000000000011
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 164
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv perf_fp16 161
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 151
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 155
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 156
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 157
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 152
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf25 1.4940402684133964 1.447332235394843 82.48 1.9399999999999977
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv fp16 12
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 151
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv samp_fp16 261
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf26 1.4981764588414919 1.4530714150549078 82.39999999999999 2.0200000000000102
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 152
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv perf_fp16 161
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 151
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 161
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 156
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 152
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 152
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf27 1.5004334658773033 1.4549115105608688 82.3 2.1200000000000045
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 152
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv perf_fp16 161
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 151
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 155
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 156
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 152
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 156
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf28 1.5006808163336343 1.4553824345285296 82.3 2.1200000000000045
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 152
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv perf_fp16 161
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 151
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 155
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 156
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 152
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 152
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf29 1.4999870719460484 1.4571625511374704 82.28 2.1400000000000006
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 168
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 163
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 152
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 165
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf30 1.500042366879961 1.4574715946270216 82.28 2.1400000000000006
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 168
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 163
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 152
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 158
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf31 1.500214789632402 1.4576323532660131 82.28 2.1400000000000006
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 168
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 163
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 164
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 152
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 151
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 153
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf32 1.4927009086066445 1.4484049211953174 82.26 2.1599999999999966
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 164
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv perf_fp16 161
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 151
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 161
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 156
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 152
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 152
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf33 1.5003438014588875 1.4538240352408085 82.22 2.200000000000003
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 152
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv perf_fp16 161
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 151
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 155
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv fp16 12
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 152
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 152
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf34 1.5041587978616728 1.4610492456195174 82.02000000000001 2.3999999999999915
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 168
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 152
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv perf_fp16 161
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 161
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 155
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 152
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 152
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 158
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf35 1.5000040131742656 1.4555601139156464 81.88 2.5400000000000063
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv fp16 12
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv perf_fp16 152
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 12
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv perf_fp16 161
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 151
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 155
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 151
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 152
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 167
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf36 1.4950571524902583 1.451478376045808 81.84 2.5799999999999983
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 164
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv fp16 12
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv perf_fp16 161
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 161
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 155
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 155
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 152
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 153
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
++++++
+conf37 1.4975271575548847 1.4532126224638244 81.44 2.980000000000004
+1 gpu conv fp16 12
+2 gpu batchnorm fp16 12
+3 gpu relu fp16 12
+4 gpu group_conv fp16 12
+5 gpu batchnorm fp16 12
+6 gpu relu fp16 12
+7 gpu conv fp16 12
+8 gpu batchnorm fp16 12
+9 gpu relu fp16 12
+10 gpu group_conv fp16 12
+11 gpu batchnorm fp16 12
+12 gpu relu fp16 12
+13 gpu conv fp16 12
+14 gpu batchnorm fp16 12
+15 gpu relu fp16 12
+16 gpu group_conv fp16 12
+17 gpu batchnorm fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 164
+20 gpu batchnorm fp16 12
+21 gpu relu fp16 12
+22 gpu group_conv fp16 12
+23 gpu batchnorm fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu group_conv fp16 12
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv fp16 12
+32 gpu batchnorm fp16 12
+33 gpu relu fp16 12
+34 gpu group_conv fp16 12
+35 gpu batchnorm fp16 12
+36 gpu relu fp16 12
+37 gpu conv fp16 11
+38 gpu batchnorm fp16 12
+39 gpu relu fp16 12
+40 gpu group_conv fp16 12
+41 gpu batchnorm fp16 12
+42 gpu relu fp16 12
+43 gpu conv fp16 12
+44 gpu batchnorm fp16 12
+45 gpu relu fp16 12
+46 gpu group_conv fp16 12
+47 gpu batchnorm fp16 12
+48 gpu relu fp16 12
+49 gpu conv perf_fp16 155
+50 gpu batchnorm fp16 12
+51 gpu relu fp16 12
+52 gpu group_conv fp16 12
+53 gpu batchnorm fp16 12
+54 gpu relu fp16 12
+55 gpu conv perf_fp16 155
+56 gpu batchnorm fp16 12
+57 gpu relu fp16 12
+58 gpu group_conv fp16 12
+59 gpu batchnorm fp16 12
+60 gpu relu fp16 12
+61 gpu conv perf_fp16 151
+62 gpu batchnorm fp16 12
+63 gpu relu fp16 12
+64 gpu group_conv fp16 12
+65 gpu batchnorm fp16 12
+66 gpu relu fp16 12
+67 gpu conv perf_fp16 155
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu group_conv fp16 12
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv perf_fp16 152
+74 gpu batchnorm fp16 12
+75 gpu relu fp16 12
+76 gpu group_conv fp16 12
+77 gpu batchnorm fp16 12
+78 gpu relu fp16 12
+79 gpu conv perf_fp16 153
+80 gpu batchnorm fp16 12
+81 gpu relu fp16 12
+82 gpu pool_mean fp16 12
+83 gpu mul fp16 12 add fp16 1
+84 gpu softmax fp32 1
+-----
diff --git a/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/resnet18_cifar10/resnet18_cifar10.txt b/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/resnet18_cifar10/resnet18_cifar10.txt
new file mode 100644
index 0000000000000000000000000000000000000000..654cffbf632686dca6310a93ecf56b6521e32039
--- /dev/null
+++ b/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/resnet18_cifar10/resnet18_cifar10.txt
@@ -0,0 +1,2296 @@
+2484.981244
++++++
+conf1 1 1 89.56 0.0
+1 gpu conv fp32 11 add fp32 1 relu fp32 1
+2 gpu conv fp32 11 add fp32 1 relu fp32 1
+3 gpu conv fp32 11 add fp32 1
+4 gpu add fp32 11
+5 gpu relu fp32 11
+6 gpu conv fp32 11 add fp32 1 relu fp32 1
+7 gpu conv fp32 11 add fp32 1
+8 gpu add fp32 11
+9 gpu relu fp32 11
+10 gpu conv fp32 11 add fp32 1 relu fp32 1
+11 gpu conv fp32 11 add fp32 1
+12 gpu add fp32 11
+13 gpu relu fp32 11
+14 gpu conv fp32 11 add fp32 1 relu fp32 1
+15 gpu conv fp32 11 add fp32 1
+16 gpu conv fp32 11 add fp32 1
+17 gpu add fp32 11
+18 gpu relu fp32 11
+19 gpu conv fp32 11 add fp32 1 relu fp32 1
+20 gpu conv fp32 11 add fp32 1
+21 gpu add fp32 11
+22 gpu relu fp32 11
+23 gpu conv fp32 11 add fp32 1 relu fp32 1
+24 gpu conv fp32 11 add fp32 1
+25 gpu add fp32 11
+26 gpu relu fp32 11
+27 gpu conv fp32 11 add fp32 1 relu fp32 1
+28 gpu conv fp32 11 add fp32 1
+29 gpu conv fp32 11 add fp32 1
+30 gpu add fp32 11
+31 gpu relu fp32 11
+32 gpu conv fp32 11 add fp32 1 relu fp32 1
+33 gpu conv fp32 11 add fp32 1
+34 gpu add fp32 11
+35 gpu relu fp32 11
+36 gpu conv fp32 11 add fp32 1 relu fp32 1
+37 gpu conv fp32 11 add fp32 1
+38 gpu add fp32 11
+39 gpu relu fp32 11
+40 gpu pool_mean fp32 11
+41 gpu mul fp32 11 add fp32 1
+42 gpu softmax fp32 1
+-----
++++++
+conf2 1.767527790869615 1.7962938589450996 88.96 0.6000000000000085
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 162 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 167 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 167 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 160 add fp16 1 relu fp16 1
+20 gpu conv perf_fp16 155 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv fp16 12 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 160 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 11 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 155 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf3 1.7676486174436143 1.7967155014984917 88.78 0.7800000000000011
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 162 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 167 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 167 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv perf_fp16 160 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 160 add fp16 1 relu fp16 1
+20 gpu conv perf_fp16 155 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv fp16 12 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 160 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 11 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 155 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf4 1.7674352647250422 1.792910560846682 88.7 0.8599999999999994
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 162 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 167 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 167 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 168 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv fp16 12 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 160 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 11 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 155 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf5 1.8655703338511067 1.8930089896922888 88.53999999999999 1.0200000000000102
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 167 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 158 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv perf_fp16 159 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 165 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 157 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf6 1.9070428103729684 1.9172857853336078 88.38000000000001 1.1799999999999926
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 160 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 157 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 152 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv samp_fp16 266 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+20 gpu conv perf_fp16 152 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv samp_fp16 261 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 11 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 155 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf7 1.769778590701739 1.7956222622694236 88.24 1.3200000000000074
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv fp16 12 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv samp_fp16 268 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf8 1.841404652091802 1.8677947628418006 88.24 1.3200000000000074
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 168 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 162 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf9 1.8679349428783487 1.8995927920729931 88.22 1.3400000000000034
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 159 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 160 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 168 add fp16 1 relu fp16 1
+20 gpu conv perf_fp16 161 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 11 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf10 1.876937310100899 1.9041581451399825 88.1 1.460000000000008
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 158 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 11 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf11 1.842140004857965 1.8673692956620238 88.06 1.5
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 166 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 167 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 11 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf12 1.9070567138857761 1.9165525910492667 88.02 1.5400000000000063
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 160 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 157 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 152 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv samp_fp16 266 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 261 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 152 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 11 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 155 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf13 1.9185835698271805 1.9328202469403 87.98 1.5799999999999983
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 160 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 157 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 152 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv samp_fp16 266 add fp16 1
+16 gpu conv perf_fp16 160 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+20 gpu conv perf_fp16 152 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 152 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 11 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 155 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf14 1.781744853993609 1.8082995958456516 87.92 1.6400000000000006
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 166 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 160 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 168 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv perf_fp16 159 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 165 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv samp_fp16 265 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv samp_fp16 268 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf15 1.9185835698271805 1.9328202469403 87.92 1.6400000000000006
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 160 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 157 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 152 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv samp_fp16 266 add fp16 1
+16 gpu conv perf_fp16 160 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+20 gpu conv perf_fp16 152 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 152 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 12 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 155 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf16 1.875261840315855 1.8986912653657988 87.88 1.6800000000000068
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 159 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 156 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 12 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf17 1.9013559086026153 1.9230901214481015 87.86 1.7000000000000028
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 11 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf18 1.9185835698271805 1.9328202469403 87.83999999999999 1.720000000000013
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 160 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 157 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 152 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv samp_fp16 266 add fp16 1
+16 gpu conv perf_fp16 160 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+20 gpu conv perf_fp16 152 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 152 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 11 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 155 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf19 1.8770503055325798 1.9007923328014182 87.82 1.740000000000009
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 162 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 158 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 151 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf20 1.8774136276932418 1.90365663123621 87.82 1.740000000000009
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 158 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf21 1.943143041264842 1.9591958561422729 87.82 1.740000000000009
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 152 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf22 1.870789918969847 1.8863625217899933 87.8 1.7600000000000051
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 264 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf23 1.7445941809066292 1.7754934270309912 87.78 1.7800000000000011
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 163 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 162 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 167 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 167 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv perf_fp16 160 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 160 add fp16 1 relu fp16 1
+20 gpu conv perf_fp16 155 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv fp16 12 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 160 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 11 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv perf_fp16 166 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 155 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf24 1.9065930313550916 1.928938946228637 87.78 1.7800000000000011
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 167 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 11 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf25 1.9021824494907031 1.9237134505552098 87.78 1.7800000000000011
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 154 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf26 1.9017271009017505 1.9211078231701697 87.78 1.7800000000000011
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 162 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf27 1.8187224917656395 1.820406007609536 87.76 1.7999999999999972
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv samp_fp16 264 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf28 1.9070855899343322 1.9285210655709735 87.76 1.7999999999999972
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv samp_fp16 268 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf29 1.9013559086026153 1.9230901214481015 87.74 1.8200000000000074
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 11 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf30 1.8772990284718367 1.9022146647342513 87.72 1.8400000000000034
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 162 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 158 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf31 1.9013559086026153 1.9230901214481015 87.68 1.8799999999999955
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 11 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf32 1.9020502478364545 1.923319572598976 87.66000000000001 1.8999999999999915
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf33 1.7516394053514481 1.7809034526471939 87.62 1.9399999999999977
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 157 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 162 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 167 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 167 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv perf_fp16 160 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 160 add fp16 1 relu fp16 1
+20 gpu conv perf_fp16 155 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv fp16 12 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 160 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 11 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv perf_fp16 166 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 155 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf34 1.7814953252955337 1.8122658147993431 87.62 1.9399999999999977
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 162 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 167 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 167 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv perf_fp16 160 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 160 add fp16 1 relu fp16 1
+20 gpu conv perf_fp16 155 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv fp16 12 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 160 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 11 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv perf_fp16 166 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 155 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf35 1.887538247557846 1.9103369445911678 87.62 1.9399999999999977
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 158 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 159 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf36 1.9107566783735581 1.9273803227885578 87.6 1.960000000000008
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 157 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf37 1.9013559086026153 1.9230901214481015 87.58 1.980000000000004
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 12 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf38 1.8984089819969947 1.9195632881772446 87.58 1.980000000000004
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 160 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf39 1.9020502478364545 1.923319572598976 87.52 2.0400000000000063
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf40 1.9020502478364545 1.923319572598976 87.52 2.0400000000000063
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf41 1.9013559086026153 1.9230901214481015 87.5 2.0600000000000023
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 11 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf42 1.9013559086026153 1.9230901214481015 87.46000000000001 2.0999999999999943
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv fp16 11 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf43 1.9196179152539186 1.9443459719929068 87.44 2.1200000000000045
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 153 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf44 1.9020502478364545 1.923319572598976 87.4 2.1599999999999966
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf45 1.9152817031040366 1.9357432559063958 87.4 2.1599999999999966
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf46 1.915754791147898 1.9373322475753219 87.4 2.1599999999999966
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf47 1.9130551004051772 1.9409232417921056 87.38 2.180000000000007
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv perf_fp16 153 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf48 1.9421147660673033 1.9584555432766413 87.38 2.180000000000007
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf49 1.9052849920081363 1.9300100333661123 87.32 2.240000000000009
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 153 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf50 1.9154322863033566 1.934908329027621 87.3 2.260000000000005
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv perf_fp16 151 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
++++++
+conf51 1.9079703554020564 1.9287218218306195 86.96000000000001 2.5999999999999943
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1
+3 gpu conv fp16 12 add fp16 1
+4 gpu add fp16 12
+5 gpu relu fp16 12
+6 gpu conv perf_fp16 153 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 161 add fp16 1
+8 gpu add fp16 12
+9 gpu relu fp16 12
+10 gpu conv perf_fp16 154 add fp16 1 relu fp16 1
+11 gpu conv perf_fp16 151 add fp16 1
+12 gpu add fp16 12
+13 gpu relu fp16 12
+14 gpu conv fp16 12 add fp16 1 relu fp16 1
+15 gpu conv fp16 12 add fp16 1
+16 gpu conv fp16 11 add fp16 1
+17 gpu add fp16 12
+18 gpu relu fp16 12
+19 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+20 gpu conv samp_fp16 262 add fp16 1
+21 gpu add fp16 12
+22 gpu relu fp16 12
+23 gpu conv perf_fp16 158 add fp16 1 relu fp16 1
+24 gpu conv perf_fp16 153 add fp16 1
+25 gpu add fp16 12
+26 gpu relu fp16 12
+27 gpu conv fp16 12 add fp16 1 relu fp16 1
+28 gpu conv fp16 12 add fp16 1
+29 gpu conv samp_fp16 261 add fp16 1
+30 gpu add fp16 12
+31 gpu relu fp16 12
+32 gpu conv fp16 12 add fp16 1 relu fp16 1
+33 gpu conv fp16 12 add fp16 1
+34 gpu add fp16 12
+35 gpu relu fp16 12
+36 gpu conv fp16 12 add fp16 1 relu fp16 1
+37 gpu conv perf_fp16 152 add fp16 1
+38 gpu add fp16 12
+39 gpu relu fp16 12
+40 gpu pool_mean fp16 12
+41 gpu mul fp16 12 add fp16 1
+42 gpu softmax fp32 1
+-----
diff --git a/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/resnet50_imagenet.txt b/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/resnet50_imagenet.txt
new file mode 100644
index 0000000000000000000000000000000000000000..094eed413b520f9dd661797b96735438861d1c08
--- /dev/null
+++ b/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/resnet50_imagenet.txt
@@ -0,0 +1,1057 @@
+7161.053769000008
++++++
+conf1 1 1 75.7 0.0
+1 gpu conv fp32 11 add fp32 1 relu fp32 1 pool_max fp32 1
+2 gpu batchnorm fp32 11
+3 gpu conv fp32 11 add fp32 1
+4 gpu batchnorm fp32 11
+5 gpu relu fp32 11
+6 gpu conv fp32 11 add fp32 1
+7 gpu batchnorm fp32 11
+8 gpu relu fp32 11
+9 gpu conv fp32 11 add fp32 1
+10 gpu batchnorm fp32 11
+11 gpu conv fp32 11 add fp32 1
+12 gpu batchnorm fp32 11
+13 gpu add fp32 11
+14 gpu relu fp32 11
+15 gpu conv fp32 11 add fp32 1
+16 gpu batchnorm fp32 11
+17 gpu relu fp32 11
+18 gpu conv fp32 11 add fp32 1
+19 gpu batchnorm fp32 11
+20 gpu relu fp32 11
+21 gpu conv fp32 11 add fp32 1
+22 gpu batchnorm fp32 11
+23 gpu add fp32 11
+24 gpu relu fp32 11
+25 gpu conv fp32 11 add fp32 1
+26 gpu batchnorm fp32 11
+27 gpu relu fp32 11
+28 gpu conv fp32 11 add fp32 1
+29 gpu batchnorm fp32 11
+30 gpu relu fp32 11
+31 gpu conv fp32 11 add fp32 1
+32 gpu batchnorm fp32 11
+33 gpu add fp32 11
+34 gpu relu fp32 11
+35 gpu conv fp32 11 add fp32 1
+36 gpu batchnorm fp32 11
+37 gpu relu fp32 11
+38 gpu conv fp32 11 add fp32 1
+39 gpu batchnorm fp32 11
+40 gpu relu fp32 11
+41 gpu conv fp32 11 add fp32 1
+42 gpu batchnorm fp32 11
+43 gpu conv fp32 11 add fp32 1
+44 gpu batchnorm fp32 11
+45 gpu add fp32 11
+46 gpu relu fp32 11
+47 gpu conv fp32 11 add fp32 1
+48 gpu batchnorm fp32 11
+49 gpu relu fp32 11
+50 gpu conv fp32 11 add fp32 1
+51 gpu batchnorm fp32 11
+52 gpu relu fp32 11
+53 gpu conv fp32 11 add fp32 1
+54 gpu batchnorm fp32 11
+55 gpu add fp32 11
+56 gpu relu fp32 11
+57 gpu conv fp32 11 add fp32 1
+58 gpu batchnorm fp32 11
+59 gpu relu fp32 11
+60 gpu conv fp32 11 add fp32 1
+61 gpu batchnorm fp32 11
+62 gpu relu fp32 11
+63 gpu conv fp32 11 add fp32 1
+64 gpu batchnorm fp32 11
+65 gpu add fp32 11
+66 gpu relu fp32 11
+67 gpu conv fp32 11 add fp32 1
+68 gpu batchnorm fp32 11
+69 gpu relu fp32 11
+70 gpu conv fp32 11 add fp32 1
+71 gpu batchnorm fp32 11
+72 gpu relu fp32 11
+73 gpu conv fp32 11 add fp32 1
+74 gpu batchnorm fp32 11
+75 gpu add fp32 11
+76 gpu relu fp32 11
+77 gpu conv fp32 11 add fp32 1
+78 gpu batchnorm fp32 11
+79 gpu relu fp32 11
+80 gpu conv fp32 11 add fp32 1
+81 gpu batchnorm fp32 11
+82 gpu relu fp32 11
+83 gpu conv fp32 11 add fp32 1
+84 gpu batchnorm fp32 11
+85 gpu conv fp32 11 add fp32 1
+86 gpu batchnorm fp32 11
+87 gpu add fp32 11
+88 gpu relu fp32 11
+89 gpu conv fp32 11 add fp32 1
+90 gpu batchnorm fp32 11
+91 gpu relu fp32 11
+92 gpu conv fp32 11 add fp32 1
+93 gpu batchnorm fp32 11
+94 gpu relu fp32 11
+95 gpu conv fp32 11 add fp32 1
+96 gpu batchnorm fp32 11
+97 gpu add fp32 11
+98 gpu relu fp32 11
+99 gpu conv fp32 11 add fp32 1
+100 gpu batchnorm fp32 11
+101 gpu relu fp32 11
+102 gpu conv fp32 11 add fp32 1
+103 gpu batchnorm fp32 11
+104 gpu relu fp32 11
+105 gpu conv fp32 11 add fp32 1
+106 gpu batchnorm fp32 11
+107 gpu add fp32 11
+108 gpu relu fp32 11
+109 gpu conv fp32 11 add fp32 1
+110 gpu batchnorm fp32 11
+111 gpu relu fp32 11
+112 gpu conv fp32 11 add fp32 1
+113 gpu batchnorm fp32 11
+114 gpu relu fp32 11
+115 gpu conv fp32 11 add fp32 1
+116 gpu batchnorm fp32 11
+117 gpu add fp32 11
+118 gpu relu fp32 11
+119 gpu conv fp32 11 add fp32 1
+120 gpu batchnorm fp32 11
+121 gpu relu fp32 11
+122 gpu conv fp32 11 add fp32 1
+123 gpu batchnorm fp32 11
+124 gpu relu fp32 11
+125 gpu conv fp32 11 add fp32 1
+126 gpu batchnorm fp32 11
+127 gpu add fp32 11
+128 gpu relu fp32 11
+129 gpu conv fp32 11 add fp32 1
+130 gpu batchnorm fp32 11
+131 gpu relu fp32 11
+132 gpu conv fp32 11 add fp32 1
+133 gpu batchnorm fp32 11
+134 gpu relu fp32 11
+135 gpu conv fp32 11 add fp32 1
+136 gpu batchnorm fp32 11
+137 gpu add fp32 11
+138 gpu relu fp32 11
+139 gpu conv fp32 11 add fp32 1
+140 gpu batchnorm fp32 11
+141 gpu relu fp32 11
+142 gpu conv fp32 11 add fp32 1
+143 gpu batchnorm fp32 11
+144 gpu relu fp32 11
+145 gpu conv fp32 11 add fp32 1
+146 gpu batchnorm fp32 11
+147 gpu conv fp32 11 add fp32 1
+148 gpu batchnorm fp32 11
+149 gpu add fp32 11
+150 gpu relu fp32 11
+151 gpu conv fp32 11 add fp32 1
+152 gpu batchnorm fp32 11
+153 gpu relu fp32 11
+154 gpu conv fp32 11 add fp32 1
+155 gpu batchnorm fp32 11
+156 gpu relu fp32 11
+157 gpu conv fp32 11 add fp32 1
+158 gpu batchnorm fp32 11
+159 gpu add fp32 11
+160 gpu relu fp32 11
+161 gpu conv fp32 11 add fp32 1
+162 gpu batchnorm fp32 11
+163 gpu relu fp32 11
+164 gpu conv fp32 11 add fp32 1
+165 gpu batchnorm fp32 11
+166 gpu relu fp32 11
+167 gpu conv fp32 11 add fp32 1
+168 gpu batchnorm fp32 11
+169 gpu add fp32 11
+170 gpu relu fp32 11
+171 gpu pool_max fp32 11
+172 gpu mul fp32 11 add fp32 1
+173 gpu softmax fp32 1
+-----
++++++
+conf2 1.8254789092281507 1.4527803526239977 75.7 0.0
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu batchnorm fp16 12
+3 gpu conv fp16 12 add fp16 1
+4 gpu batchnorm fp16 12
+5 gpu relu fp16 12
+6 gpu conv fp16 12 add fp16 1
+7 gpu batchnorm fp16 12
+8 gpu relu fp16 12
+9 gpu conv fp16 12 add fp16 1
+10 gpu batchnorm fp16 12
+11 gpu conv fp16 12 add fp16 1
+12 gpu batchnorm fp16 12
+13 gpu add fp16 12
+14 gpu relu fp16 12
+15 gpu conv fp16 12 add fp16 1
+16 gpu batchnorm fp16 12
+17 gpu relu fp16 12
+18 gpu conv fp16 12 add fp16 1
+19 gpu batchnorm fp16 12
+20 gpu relu fp16 12
+21 gpu conv fp16 12 add fp16 1
+22 gpu batchnorm fp16 12
+23 gpu add fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12 add fp16 1
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu conv fp16 12 add fp16 1
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv fp16 12 add fp16 1
+32 gpu batchnorm fp16 12
+33 gpu add fp16 12
+34 gpu relu fp16 12
+35 gpu conv fp16 12 add fp16 1
+36 gpu batchnorm fp16 12
+37 gpu relu fp16 12
+38 gpu conv fp16 12 add fp16 1
+39 gpu batchnorm fp16 12
+40 gpu relu fp16 12
+41 gpu conv fp16 12 add fp16 1
+42 gpu batchnorm fp16 12
+43 gpu conv fp16 12 add fp16 1
+44 gpu batchnorm fp16 12
+45 gpu add fp16 12
+46 gpu relu fp16 12
+47 gpu conv fp16 12 add fp16 1
+48 gpu batchnorm fp16 12
+49 gpu relu fp16 12
+50 gpu conv fp16 12 add fp16 1
+51 gpu batchnorm fp16 12
+52 gpu relu fp16 12
+53 gpu conv fp16 12 add fp16 1
+54 gpu batchnorm fp16 12
+55 gpu add fp16 12
+56 gpu relu fp16 12
+57 gpu conv fp16 12 add fp16 1
+58 gpu batchnorm fp16 12
+59 gpu relu fp16 12
+60 gpu conv fp16 12 add fp16 1
+61 gpu batchnorm fp16 12
+62 gpu relu fp16 12
+63 gpu conv fp16 12 add fp16 1
+64 gpu batchnorm fp16 12
+65 gpu add fp16 12
+66 gpu relu fp16 12
+67 gpu conv fp16 12 add fp16 1
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu conv fp16 12 add fp16 1
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv fp16 12 add fp16 1
+74 gpu batchnorm fp16 12
+75 gpu add fp16 12
+76 gpu relu fp16 12
+77 gpu conv fp16 12 add fp16 1
+78 gpu batchnorm fp16 12
+79 gpu relu fp16 12
+80 gpu conv fp16 12 add fp16 1
+81 gpu batchnorm fp16 12
+82 gpu relu fp16 12
+83 gpu conv fp16 12 add fp16 1
+84 gpu batchnorm fp16 12
+85 gpu conv fp16 12 add fp16 1
+86 gpu batchnorm fp16 12
+87 gpu add fp16 12
+88 gpu relu fp16 12
+89 gpu conv fp16 12 add fp16 1
+90 gpu batchnorm fp16 12
+91 gpu relu fp16 12
+92 gpu conv fp16 12 add fp16 1
+93 gpu batchnorm fp16 12
+94 gpu relu fp16 12
+95 gpu conv fp16 12 add fp16 1
+96 gpu batchnorm fp16 12
+97 gpu add fp16 12
+98 gpu relu fp16 12
+99 gpu conv fp16 12 add fp16 1
+100 gpu batchnorm fp16 12
+101 gpu relu fp16 12
+102 gpu conv fp16 12 add fp16 1
+103 gpu batchnorm fp16 12
+104 gpu relu fp16 12
+105 gpu conv fp16 12 add fp16 1
+106 gpu batchnorm fp16 12
+107 gpu add fp16 12
+108 gpu relu fp16 12
+109 gpu conv fp16 12 add fp16 1
+110 gpu batchnorm fp16 12
+111 gpu relu fp16 12
+112 gpu conv fp16 12 add fp16 1
+113 gpu batchnorm fp16 12
+114 gpu relu fp16 12
+115 gpu conv fp16 12 add fp16 1
+116 gpu batchnorm fp16 12
+117 gpu add fp16 12
+118 gpu relu fp16 12
+119 gpu conv fp16 12 add fp16 1
+120 gpu batchnorm fp16 12
+121 gpu relu fp16 12
+122 gpu conv fp16 12 add fp16 1
+123 gpu batchnorm fp16 12
+124 gpu relu fp16 12
+125 gpu conv fp16 12 add fp16 1
+126 gpu batchnorm fp16 12
+127 gpu add fp16 12
+128 gpu relu fp16 12
+129 gpu conv fp16 12 add fp16 1
+130 gpu batchnorm fp16 12
+131 gpu relu fp16 12
+132 gpu conv fp16 12 add fp16 1
+133 gpu batchnorm fp16 12
+134 gpu relu fp16 12
+135 gpu conv fp16 12 add fp16 1
+136 gpu batchnorm fp16 12
+137 gpu add fp16 12
+138 gpu relu fp16 12
+139 gpu conv fp16 12 add fp16 1
+140 gpu batchnorm fp16 12
+141 gpu relu fp16 12
+142 gpu conv fp16 12 add fp16 1
+143 gpu batchnorm fp16 12
+144 gpu relu fp16 12
+145 gpu conv fp16 12 add fp16 1
+146 gpu batchnorm fp16 12
+147 gpu conv fp16 12 add fp16 1
+148 gpu batchnorm fp16 12
+149 gpu add fp16 12
+150 gpu relu fp16 12
+151 gpu conv fp16 12 add fp16 1
+152 gpu batchnorm fp16 12
+153 gpu relu fp16 12
+154 gpu conv fp16 12 add fp16 1
+155 gpu batchnorm fp16 12
+156 gpu relu fp16 12
+157 gpu conv fp16 12 add fp16 1
+158 gpu batchnorm fp16 12
+159 gpu add fp16 12
+160 gpu relu fp16 12
+161 gpu conv fp16 12 add fp16 1
+162 gpu batchnorm fp16 12
+163 gpu relu fp16 12
+164 gpu conv fp16 12 add fp16 1
+165 gpu batchnorm fp16 12
+166 gpu relu fp16 12
+167 gpu conv fp16 12 add fp16 1
+168 gpu batchnorm fp16 12
+169 gpu add fp16 12
+170 gpu relu fp16 12
+171 gpu pool_max fp16 12
+172 gpu mul fp16 12 add fp16 1
+173 gpu softmax fp32 1
+-----
++++++
+conf3 1.8254789092281507 1.4527803526239977 75.7 0.0
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu batchnorm fp16 12
+3 gpu conv fp16 12 add fp16 1
+4 gpu batchnorm fp16 12
+5 gpu relu fp16 12
+6 gpu conv fp16 12 add fp16 1
+7 gpu batchnorm fp16 12
+8 gpu relu fp16 12
+9 gpu conv fp16 12 add fp16 1
+10 gpu batchnorm fp16 12
+11 gpu conv fp16 12 add fp16 1
+12 gpu batchnorm fp16 12
+13 gpu add fp16 12
+14 gpu relu fp16 12
+15 gpu conv fp16 12 add fp16 1
+16 gpu batchnorm fp16 12
+17 gpu relu fp16 12
+18 gpu conv fp16 12 add fp16 1
+19 gpu batchnorm fp16 12
+20 gpu relu fp16 12
+21 gpu conv fp16 12 add fp16 1
+22 gpu batchnorm fp16 12
+23 gpu add fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12 add fp16 1
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu conv fp16 12 add fp16 1
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv fp16 12 add fp16 1
+32 gpu batchnorm fp16 12
+33 gpu add fp16 12
+34 gpu relu fp16 12
+35 gpu conv fp16 12 add fp16 1
+36 gpu batchnorm fp16 12
+37 gpu relu fp16 12
+38 gpu conv fp16 12 add fp16 1
+39 gpu batchnorm fp16 12
+40 gpu relu fp16 12
+41 gpu conv fp16 12 add fp16 1
+42 gpu batchnorm fp16 12
+43 gpu conv fp16 12 add fp16 1
+44 gpu batchnorm fp16 12
+45 gpu add fp16 12
+46 gpu relu fp16 12
+47 gpu conv fp16 12 add fp16 1
+48 gpu batchnorm fp16 12
+49 gpu relu fp16 12
+50 gpu conv fp16 12 add fp16 1
+51 gpu batchnorm fp16 12
+52 gpu relu fp16 12
+53 gpu conv fp16 12 add fp16 1
+54 gpu batchnorm fp16 12
+55 gpu add fp16 12
+56 gpu relu fp16 12
+57 gpu conv fp16 12 add fp16 1
+58 gpu batchnorm fp16 12
+59 gpu relu fp16 12
+60 gpu conv fp16 12 add fp16 1
+61 gpu batchnorm fp16 12
+62 gpu relu fp16 12
+63 gpu conv fp16 12 add fp16 1
+64 gpu batchnorm fp16 12
+65 gpu add fp16 12
+66 gpu relu fp16 12
+67 gpu conv fp16 12 add fp16 1
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu conv fp16 12 add fp16 1
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv fp16 12 add fp16 1
+74 gpu batchnorm fp16 12
+75 gpu add fp16 12
+76 gpu relu fp16 12
+77 gpu conv fp16 12 add fp16 1
+78 gpu batchnorm fp16 12
+79 gpu relu fp16 12
+80 gpu conv fp16 12 add fp16 1
+81 gpu batchnorm fp16 12
+82 gpu relu fp16 12
+83 gpu conv fp16 12 add fp16 1
+84 gpu batchnorm fp16 12
+85 gpu conv fp16 12 add fp16 1
+86 gpu batchnorm fp16 12
+87 gpu add fp16 12
+88 gpu relu fp16 12
+89 gpu conv fp16 12 add fp16 1
+90 gpu batchnorm fp16 12
+91 gpu relu fp16 12
+92 gpu conv fp16 12 add fp16 1
+93 gpu batchnorm fp16 12
+94 gpu relu fp16 12
+95 gpu conv fp16 12 add fp16 1
+96 gpu batchnorm fp16 12
+97 gpu add fp16 12
+98 gpu relu fp16 12
+99 gpu conv fp16 12 add fp16 1
+100 gpu batchnorm fp16 12
+101 gpu relu fp16 12
+102 gpu conv fp16 12 add fp16 1
+103 gpu batchnorm fp16 12
+104 gpu relu fp16 12
+105 gpu conv fp16 12 add fp16 1
+106 gpu batchnorm fp16 12
+107 gpu add fp16 12
+108 gpu relu fp16 12
+109 gpu conv fp16 12 add fp16 1
+110 gpu batchnorm fp16 12
+111 gpu relu fp16 12
+112 gpu conv fp16 12 add fp16 1
+113 gpu batchnorm fp16 12
+114 gpu relu fp16 12
+115 gpu conv fp16 12 add fp16 1
+116 gpu batchnorm fp16 12
+117 gpu add fp16 12
+118 gpu relu fp16 12
+119 gpu conv fp16 12 add fp16 1
+120 gpu batchnorm fp16 12
+121 gpu relu fp16 12
+122 gpu conv fp16 12 add fp16 1
+123 gpu batchnorm fp16 12
+124 gpu relu fp16 12
+125 gpu conv fp16 12 add fp16 1
+126 gpu batchnorm fp16 12
+127 gpu add fp16 12
+128 gpu relu fp16 12
+129 gpu conv fp16 12 add fp16 1
+130 gpu batchnorm fp16 12
+131 gpu relu fp16 12
+132 gpu conv fp16 12 add fp16 1
+133 gpu batchnorm fp16 12
+134 gpu relu fp16 12
+135 gpu conv fp16 12 add fp16 1
+136 gpu batchnorm fp16 12
+137 gpu add fp16 12
+138 gpu relu fp16 12
+139 gpu conv fp16 12 add fp16 1
+140 gpu batchnorm fp16 12
+141 gpu relu fp16 12
+142 gpu conv fp16 12 add fp16 1
+143 gpu batchnorm fp16 12
+144 gpu relu fp16 12
+145 gpu conv fp16 12 add fp16 1
+146 gpu batchnorm fp16 12
+147 gpu conv fp16 12 add fp16 1
+148 gpu batchnorm fp16 12
+149 gpu add fp16 12
+150 gpu relu fp16 12
+151 gpu conv fp16 12 add fp16 1
+152 gpu batchnorm fp16 12
+153 gpu relu fp16 12
+154 gpu conv fp16 12 add fp16 1
+155 gpu batchnorm fp16 12
+156 gpu relu fp16 12
+157 gpu conv fp16 12 add fp16 1
+158 gpu batchnorm fp16 12
+159 gpu add fp16 12
+160 gpu relu fp16 12
+161 gpu conv fp16 12 add fp16 1
+162 gpu batchnorm fp16 12
+163 gpu relu fp16 12
+164 gpu conv fp16 12 add fp16 1
+165 gpu batchnorm fp16 12
+166 gpu relu fp16 12
+167 gpu conv fp16 12 add fp16 1
+168 gpu batchnorm fp16 12
+169 gpu add fp16 12
+170 gpu relu fp16 12
+171 gpu pool_max fp16 12
+172 gpu mul fp16 12 add fp16 1
+173 gpu softmax fp32 1
+-----
++++++
+conf4 1.8254789092281507 1.4527803526239977 75.7 0.0
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu batchnorm fp16 12
+3 gpu conv fp16 12 add fp16 1
+4 gpu batchnorm fp16 12
+5 gpu relu fp16 12
+6 gpu conv fp16 12 add fp16 1
+7 gpu batchnorm fp16 12
+8 gpu relu fp16 12
+9 gpu conv fp16 12 add fp16 1
+10 gpu batchnorm fp16 12
+11 gpu conv fp16 12 add fp16 1
+12 gpu batchnorm fp16 12
+13 gpu add fp16 12
+14 gpu relu fp16 12
+15 gpu conv fp16 12 add fp16 1
+16 gpu batchnorm fp16 12
+17 gpu relu fp16 12
+18 gpu conv fp16 12 add fp16 1
+19 gpu batchnorm fp16 12
+20 gpu relu fp16 12
+21 gpu conv fp16 12 add fp16 1
+22 gpu batchnorm fp16 12
+23 gpu add fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12 add fp16 1
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu conv fp16 12 add fp16 1
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv fp16 12 add fp16 1
+32 gpu batchnorm fp16 12
+33 gpu add fp16 12
+34 gpu relu fp16 12
+35 gpu conv fp16 12 add fp16 1
+36 gpu batchnorm fp16 12
+37 gpu relu fp16 12
+38 gpu conv fp16 12 add fp16 1
+39 gpu batchnorm fp16 12
+40 gpu relu fp16 12
+41 gpu conv fp16 12 add fp16 1
+42 gpu batchnorm fp16 12
+43 gpu conv fp16 12 add fp16 1
+44 gpu batchnorm fp16 12
+45 gpu add fp16 12
+46 gpu relu fp16 12
+47 gpu conv fp16 12 add fp16 1
+48 gpu batchnorm fp16 12
+49 gpu relu fp16 12
+50 gpu conv fp16 12 add fp16 1
+51 gpu batchnorm fp16 12
+52 gpu relu fp16 12
+53 gpu conv fp16 12 add fp16 1
+54 gpu batchnorm fp16 12
+55 gpu add fp16 12
+56 gpu relu fp16 12
+57 gpu conv fp16 12 add fp16 1
+58 gpu batchnorm fp16 12
+59 gpu relu fp16 12
+60 gpu conv fp16 12 add fp16 1
+61 gpu batchnorm fp16 12
+62 gpu relu fp16 12
+63 gpu conv fp16 12 add fp16 1
+64 gpu batchnorm fp16 12
+65 gpu add fp16 12
+66 gpu relu fp16 12
+67 gpu conv fp16 12 add fp16 1
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu conv fp16 12 add fp16 1
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv fp16 12 add fp16 1
+74 gpu batchnorm fp16 12
+75 gpu add fp16 12
+76 gpu relu fp16 12
+77 gpu conv fp16 12 add fp16 1
+78 gpu batchnorm fp16 12
+79 gpu relu fp16 12
+80 gpu conv fp16 12 add fp16 1
+81 gpu batchnorm fp16 12
+82 gpu relu fp16 12
+83 gpu conv fp16 12 add fp16 1
+84 gpu batchnorm fp16 12
+85 gpu conv fp16 12 add fp16 1
+86 gpu batchnorm fp16 12
+87 gpu add fp16 12
+88 gpu relu fp16 12
+89 gpu conv fp16 12 add fp16 1
+90 gpu batchnorm fp16 12
+91 gpu relu fp16 12
+92 gpu conv fp16 12 add fp16 1
+93 gpu batchnorm fp16 12
+94 gpu relu fp16 12
+95 gpu conv fp16 12 add fp16 1
+96 gpu batchnorm fp16 12
+97 gpu add fp16 12
+98 gpu relu fp16 12
+99 gpu conv fp16 12 add fp16 1
+100 gpu batchnorm fp16 12
+101 gpu relu fp16 12
+102 gpu conv fp16 12 add fp16 1
+103 gpu batchnorm fp16 12
+104 gpu relu fp16 12
+105 gpu conv fp16 12 add fp16 1
+106 gpu batchnorm fp16 12
+107 gpu add fp16 12
+108 gpu relu fp16 12
+109 gpu conv fp16 12 add fp16 1
+110 gpu batchnorm fp16 12
+111 gpu relu fp16 12
+112 gpu conv fp16 12 add fp16 1
+113 gpu batchnorm fp16 12
+114 gpu relu fp16 12
+115 gpu conv fp16 12 add fp16 1
+116 gpu batchnorm fp16 12
+117 gpu add fp16 12
+118 gpu relu fp16 12
+119 gpu conv fp16 12 add fp16 1
+120 gpu batchnorm fp16 12
+121 gpu relu fp16 12
+122 gpu conv fp16 12 add fp16 1
+123 gpu batchnorm fp16 12
+124 gpu relu fp16 12
+125 gpu conv fp16 12 add fp16 1
+126 gpu batchnorm fp16 12
+127 gpu add fp16 12
+128 gpu relu fp16 12
+129 gpu conv fp16 12 add fp16 1
+130 gpu batchnorm fp16 12
+131 gpu relu fp16 12
+132 gpu conv fp16 12 add fp16 1
+133 gpu batchnorm fp16 12
+134 gpu relu fp16 12
+135 gpu conv fp16 12 add fp16 1
+136 gpu batchnorm fp16 12
+137 gpu add fp16 12
+138 gpu relu fp16 12
+139 gpu conv fp16 12 add fp16 1
+140 gpu batchnorm fp16 12
+141 gpu relu fp16 12
+142 gpu conv fp16 12 add fp16 1
+143 gpu batchnorm fp16 12
+144 gpu relu fp16 12
+145 gpu conv fp16 12 add fp16 1
+146 gpu batchnorm fp16 12
+147 gpu conv fp16 12 add fp16 1
+148 gpu batchnorm fp16 12
+149 gpu add fp16 12
+150 gpu relu fp16 12
+151 gpu conv fp16 12 add fp16 1
+152 gpu batchnorm fp16 12
+153 gpu relu fp16 12
+154 gpu conv fp16 12 add fp16 1
+155 gpu batchnorm fp16 12
+156 gpu relu fp16 12
+157 gpu conv fp16 12 add fp16 1
+158 gpu batchnorm fp16 12
+159 gpu add fp16 12
+160 gpu relu fp16 12
+161 gpu conv fp16 12 add fp16 1
+162 gpu batchnorm fp16 12
+163 gpu relu fp16 12
+164 gpu conv fp16 12 add fp16 1
+165 gpu batchnorm fp16 12
+166 gpu relu fp16 12
+167 gpu conv fp16 12 add fp16 1
+168 gpu batchnorm fp16 12
+169 gpu add fp16 12
+170 gpu relu fp16 12
+171 gpu pool_max fp16 12
+172 gpu mul fp16 12 add fp16 1
+173 gpu softmax fp32 1
+-----
++++++
+conf5 1.8323072136026506 1.457112696128105 74.76 0.9399999999999977
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu batchnorm fp16 12
+3 gpu conv fp16 12 add fp16 1
+4 gpu batchnorm fp16 12
+5 gpu relu fp16 12
+6 gpu conv fp16 12 add fp16 1
+7 gpu batchnorm fp16 12
+8 gpu relu fp16 12
+9 gpu conv fp16 12 add fp16 1
+10 gpu batchnorm fp16 12
+11 gpu conv fp16 12 add fp16 1
+12 gpu batchnorm fp16 12
+13 gpu add fp16 12
+14 gpu relu fp16 12
+15 gpu conv fp16 12 add fp16 1
+16 gpu batchnorm fp16 12
+17 gpu relu fp16 12
+18 gpu conv fp16 12 add fp16 1
+19 gpu batchnorm fp16 12
+20 gpu relu fp16 12
+21 gpu conv fp16 12 add fp16 1
+22 gpu batchnorm fp16 12
+23 gpu add fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12 add fp16 1
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu conv fp16 12 add fp16 1
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv fp16 12 add fp16 1
+32 gpu batchnorm fp16 12
+33 gpu add fp16 12
+34 gpu relu fp16 12
+35 gpu conv fp16 12 add fp16 1
+36 gpu batchnorm fp16 12
+37 gpu relu fp16 12
+38 gpu conv fp16 12 add fp16 1
+39 gpu batchnorm fp16 12
+40 gpu relu fp16 12
+41 gpu conv fp16 12 add fp16 1
+42 gpu batchnorm fp16 12
+43 gpu conv fp16 12 add fp16 1
+44 gpu batchnorm fp16 12
+45 gpu add fp16 12
+46 gpu relu fp16 12
+47 gpu conv fp16 12 add fp16 1
+48 gpu batchnorm fp16 12
+49 gpu relu fp16 12
+50 gpu conv fp16 12 add fp16 1
+51 gpu batchnorm fp16 12
+52 gpu relu fp16 12
+53 gpu conv fp16 12 add fp16 1
+54 gpu batchnorm fp16 12
+55 gpu add fp16 12
+56 gpu relu fp16 12
+57 gpu conv fp16 12 add fp16 1
+58 gpu batchnorm fp16 12
+59 gpu relu fp16 12
+60 gpu conv fp16 12 add fp16 1
+61 gpu batchnorm fp16 12
+62 gpu relu fp16 12
+63 gpu conv fp16 12 add fp16 1
+64 gpu batchnorm fp16 12
+65 gpu add fp16 12
+66 gpu relu fp16 12
+67 gpu conv fp16 12 add fp16 1
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu conv fp16 12 add fp16 1
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv fp16 12 add fp16 1
+74 gpu batchnorm fp16 12
+75 gpu add fp16 12
+76 gpu relu fp16 12
+77 gpu conv fp16 12 add fp16 1
+78 gpu batchnorm fp16 12
+79 gpu relu fp16 12
+80 gpu conv fp16 12 add fp16 1
+81 gpu batchnorm fp16 12
+82 gpu relu fp16 12
+83 gpu conv fp16 12 add fp16 1
+84 gpu batchnorm fp16 12
+85 gpu conv fp16 12 add fp16 1
+86 gpu batchnorm fp16 12
+87 gpu add fp16 12
+88 gpu relu fp16 12
+89 gpu conv fp16 12 add fp16 1
+90 gpu batchnorm fp16 12
+91 gpu relu fp16 12
+92 gpu conv fp16 12 add fp16 1
+93 gpu batchnorm fp16 12
+94 gpu relu fp16 12
+95 gpu conv fp16 12 add fp16 1
+96 gpu batchnorm fp16 12
+97 gpu add fp16 12
+98 gpu relu fp16 12
+99 gpu conv perf_fp16 157 add fp16 1
+100 gpu batchnorm fp16 12
+101 gpu relu fp16 12
+102 gpu conv fp16 12 add fp16 1
+103 gpu batchnorm fp16 12
+104 gpu relu fp16 12
+105 gpu conv fp16 12 add fp16 1
+106 gpu batchnorm fp16 12
+107 gpu add fp16 12
+108 gpu relu fp16 12
+109 gpu conv fp16 12 add fp16 1
+110 gpu batchnorm fp16 12
+111 gpu relu fp16 12
+112 gpu conv fp16 12 add fp16 1
+113 gpu batchnorm fp16 12
+114 gpu relu fp16 12
+115 gpu conv fp16 12 add fp16 1
+116 gpu batchnorm fp16 12
+117 gpu add fp16 12
+118 gpu relu fp16 12
+119 gpu conv fp16 12 add fp16 1
+120 gpu batchnorm fp16 12
+121 gpu relu fp16 12
+122 gpu conv fp16 12 add fp16 1
+123 gpu batchnorm fp16 12
+124 gpu relu fp16 12
+125 gpu conv fp16 12 add fp16 1
+126 gpu batchnorm fp16 12
+127 gpu add fp16 12
+128 gpu relu fp16 12
+129 gpu conv fp16 12 add fp16 1
+130 gpu batchnorm fp16 12
+131 gpu relu fp16 12
+132 gpu conv fp16 12 add fp16 1
+133 gpu batchnorm fp16 12
+134 gpu relu fp16 12
+135 gpu conv fp16 12 add fp16 1
+136 gpu batchnorm fp16 12
+137 gpu add fp16 12
+138 gpu relu fp16 12
+139 gpu conv fp16 12 add fp16 1
+140 gpu batchnorm fp16 12
+141 gpu relu fp16 12
+142 gpu conv fp16 12 add fp16 1
+143 gpu batchnorm fp16 12
+144 gpu relu fp16 12
+145 gpu conv fp16 12 add fp16 1
+146 gpu batchnorm fp16 12
+147 gpu conv fp16 12 add fp16 1
+148 gpu batchnorm fp16 12
+149 gpu add fp16 12
+150 gpu relu fp16 12
+151 gpu conv fp16 12 add fp16 1
+152 gpu batchnorm fp16 12
+153 gpu relu fp16 12
+154 gpu conv fp16 12 add fp16 1
+155 gpu batchnorm fp16 12
+156 gpu relu fp16 12
+157 gpu conv fp16 12 add fp16 1
+158 gpu batchnorm fp16 12
+159 gpu add fp16 12
+160 gpu relu fp16 12
+161 gpu conv fp16 12 add fp16 1
+162 gpu batchnorm fp16 12
+163 gpu relu fp16 12
+164 gpu conv perf_fp16 152 add fp16 1
+165 gpu batchnorm fp16 12
+166 gpu relu fp16 12
+167 gpu conv fp16 12 add fp16 1
+168 gpu batchnorm fp16 12
+169 gpu add fp16 12
+170 gpu relu fp16 12
+171 gpu pool_max fp16 12
+172 gpu mul fp16 12 add fp16 1
+173 gpu softmax fp32 1
+-----
++++++
+conf6 1.8333922701839533 1.4589203187717397 74.53999999999999 1.1600000000000108
+1 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+2 gpu batchnorm fp16 12
+3 gpu conv fp16 12 add fp16 1
+4 gpu batchnorm fp16 12
+5 gpu relu fp16 12
+6 gpu conv fp16 12 add fp16 1
+7 gpu batchnorm fp16 12
+8 gpu relu fp16 12
+9 gpu conv fp16 12 add fp16 1
+10 gpu batchnorm fp16 12
+11 gpu conv fp16 12 add fp16 1
+12 gpu batchnorm fp16 12
+13 gpu add fp16 12
+14 gpu relu fp16 12
+15 gpu conv fp16 12 add fp16 1
+16 gpu batchnorm fp16 12
+17 gpu relu fp16 12
+18 gpu conv fp16 12 add fp16 1
+19 gpu batchnorm fp16 12
+20 gpu relu fp16 12
+21 gpu conv fp16 12 add fp16 1
+22 gpu batchnorm fp16 12
+23 gpu add fp16 12
+24 gpu relu fp16 12
+25 gpu conv fp16 12 add fp16 1
+26 gpu batchnorm fp16 12
+27 gpu relu fp16 12
+28 gpu conv fp16 12 add fp16 1
+29 gpu batchnorm fp16 12
+30 gpu relu fp16 12
+31 gpu conv fp16 12 add fp16 1
+32 gpu batchnorm fp16 12
+33 gpu add fp16 12
+34 gpu relu fp16 12
+35 gpu conv fp16 12 add fp16 1
+36 gpu batchnorm fp16 12
+37 gpu relu fp16 12
+38 gpu conv fp16 12 add fp16 1
+39 gpu batchnorm fp16 12
+40 gpu relu fp16 12
+41 gpu conv fp16 12 add fp16 1
+42 gpu batchnorm fp16 12
+43 gpu conv fp16 12 add fp16 1
+44 gpu batchnorm fp16 12
+45 gpu add fp16 12
+46 gpu relu fp16 12
+47 gpu conv fp16 12 add fp16 1
+48 gpu batchnorm fp16 12
+49 gpu relu fp16 12
+50 gpu conv fp16 12 add fp16 1
+51 gpu batchnorm fp16 12
+52 gpu relu fp16 12
+53 gpu conv fp16 12 add fp16 1
+54 gpu batchnorm fp16 12
+55 gpu add fp16 12
+56 gpu relu fp16 12
+57 gpu conv fp16 12 add fp16 1
+58 gpu batchnorm fp16 12
+59 gpu relu fp16 12
+60 gpu conv fp16 12 add fp16 1
+61 gpu batchnorm fp16 12
+62 gpu relu fp16 12
+63 gpu conv fp16 12 add fp16 1
+64 gpu batchnorm fp16 12
+65 gpu add fp16 12
+66 gpu relu fp16 12
+67 gpu conv fp16 12 add fp16 1
+68 gpu batchnorm fp16 12
+69 gpu relu fp16 12
+70 gpu conv fp16 12 add fp16 1
+71 gpu batchnorm fp16 12
+72 gpu relu fp16 12
+73 gpu conv fp16 12 add fp16 1
+74 gpu batchnorm fp16 12
+75 gpu add fp16 12
+76 gpu relu fp16 12
+77 gpu conv fp16 12 add fp16 1
+78 gpu batchnorm fp16 12
+79 gpu relu fp16 12
+80 gpu conv fp16 12 add fp16 1
+81 gpu batchnorm fp16 12
+82 gpu relu fp16 12
+83 gpu conv fp16 12 add fp16 1
+84 gpu batchnorm fp16 12
+85 gpu conv fp16 12 add fp16 1
+86 gpu batchnorm fp16 12
+87 gpu add fp16 12
+88 gpu relu fp16 12
+89 gpu conv fp16 12 add fp16 1
+90 gpu batchnorm fp16 12
+91 gpu relu fp16 12
+92 gpu conv fp16 12 add fp16 1
+93 gpu batchnorm fp16 12
+94 gpu relu fp16 12
+95 gpu conv fp16 12 add fp16 1
+96 gpu batchnorm fp16 12
+97 gpu add fp16 12
+98 gpu relu fp16 12
+99 gpu conv perf_fp16 157 add fp16 1
+100 gpu batchnorm fp16 12
+101 gpu relu fp16 12
+102 gpu conv samp_fp16 267 add fp16 1
+103 gpu batchnorm fp16 12
+104 gpu relu fp16 12
+105 gpu conv fp16 12 add fp16 1
+106 gpu batchnorm fp16 12
+107 gpu add fp16 12
+108 gpu relu fp16 12
+109 gpu conv fp16 12 add fp16 1
+110 gpu batchnorm fp16 12
+111 gpu relu fp16 12
+112 gpu conv fp16 12 add fp16 1
+113 gpu batchnorm fp16 12
+114 gpu relu fp16 12
+115 gpu conv fp16 12 add fp16 1
+116 gpu batchnorm fp16 12
+117 gpu add fp16 12
+118 gpu relu fp16 12
+119 gpu conv fp16 12 add fp16 1
+120 gpu batchnorm fp16 12
+121 gpu relu fp16 12
+122 gpu conv fp16 12 add fp16 1
+123 gpu batchnorm fp16 12
+124 gpu relu fp16 12
+125 gpu conv fp16 12 add fp16 1
+126 gpu batchnorm fp16 12
+127 gpu add fp16 12
+128 gpu relu fp16 12
+129 gpu conv fp16 12 add fp16 1
+130 gpu batchnorm fp16 12
+131 gpu relu fp16 12
+132 gpu conv fp16 12 add fp16 1
+133 gpu batchnorm fp16 12
+134 gpu relu fp16 12
+135 gpu conv fp16 12 add fp16 1
+136 gpu batchnorm fp16 12
+137 gpu add fp16 12
+138 gpu relu fp16 12
+139 gpu conv fp16 12 add fp16 1
+140 gpu batchnorm fp16 12
+141 gpu relu fp16 12
+142 gpu conv fp16 12 add fp16 1
+143 gpu batchnorm fp16 12
+144 gpu relu fp16 12
+145 gpu conv fp16 12 add fp16 1
+146 gpu batchnorm fp16 12
+147 gpu conv fp16 12 add fp16 1
+148 gpu batchnorm fp16 12
+149 gpu add fp16 12
+150 gpu relu fp16 12
+151 gpu conv fp16 12 add fp16 1
+152 gpu batchnorm fp16 12
+153 gpu relu fp16 12
+154 gpu conv fp16 12 add fp16 1
+155 gpu batchnorm fp16 12
+156 gpu relu fp16 12
+157 gpu conv fp16 12 add fp16 1
+158 gpu batchnorm fp16 12
+159 gpu add fp16 12
+160 gpu relu fp16 12
+161 gpu conv fp16 12 add fp16 1
+162 gpu batchnorm fp16 12
+163 gpu relu fp16 12
+164 gpu conv perf_fp16 152 add fp16 1
+165 gpu batchnorm fp16 12
+166 gpu relu fp16 12
+167 gpu conv fp16 12 add fp16 1
+168 gpu batchnorm fp16 12
+169 gpu add fp16 12
+170 gpu relu fp16 12
+171 gpu pool_max fp16 12
+172 gpu mul fp16 12 add fp16 1
+173 gpu softmax fp32 1
+-----
diff --git a/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/vgg16_cifar10/vgg16_cifar10.txt b/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/vgg16_cifar10/vgg16_cifar10.txt
new file mode 100644
index 0000000000000000000000000000000000000000..2b325a9fe2d122e74cdd2b80e2768e68591313bf
--- /dev/null
+++ b/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/vgg16_cifar10/vgg16_cifar10.txt
@@ -0,0 +1,913 @@
+3776.508929999999
++++++
+conf1 1 1 89.96 0.0
+1 gpu conv fp32 11 add fp32 1 relu fp32 1
+2 gpu conv fp32 11 add fp32 1 relu fp32 1 pool_max fp32 1
+3 gpu conv fp32 11 add fp32 1 relu fp32 1
+4 gpu conv fp32 11 add fp32 1 relu fp32 1 pool_max fp32 1
+5 gpu conv fp32 11 add fp32 1 relu fp32 1
+6 gpu conv fp32 11 add fp32 1 relu fp32 1
+7 gpu conv fp32 11 add fp32 1 relu fp32 1 pool_max fp32 1
+8 gpu conv fp32 11 add fp32 1 relu fp32 1
+9 gpu conv fp32 11 add fp32 1 relu fp32 1
+10 gpu conv fp32 11 add fp32 1 relu fp32 1 pool_max fp32 1
+11 gpu conv fp32 11 add fp32 1 relu fp32 1
+12 gpu conv fp32 11 add fp32 1 relu fp32 1
+13 gpu conv fp32 11 add fp32 1 relu fp32 1 pool_max fp32 1
+14 gpu mul fp32 11 add fp32 1 relu fp32 1
+15 gpu mul fp32 11 add fp32 1
+16 gpu softmax fp32 1
+-----
++++++
+conf2 2.1225958306417145 1.9771056444390926 89.91 0.04999999999999716
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 167 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 155 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 164 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 267 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 161 add fp16 1 relu fp16 1
+12 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf3 2.090180991844805 1.9532689756636086 89.82 0.14000000000000057
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 167 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 269 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 155 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 164 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 161 add fp16 1 relu fp16 1
+12 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf4 2.169931036393396 2.0048851858669283 89.53999999999999 0.4200000000000017
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 155 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv perf_fp16 162 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 264 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 269 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf5 2.1012179398201756 1.9325098819632314 89.42 0.539999999999992
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 155 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 160 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 264 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv fp16 11 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 269 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf6 2.2313002482945326 2.069581185407626 89.38000000000001 0.5799999999999841
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 158 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 269 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv fp16 12 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 269 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 264 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf7 2.143061101834193 1.9675759235961738 89.3 0.6599999999999966
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 155 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv fp16 12 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 265 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 264 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 269 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf8 2.199379444387758 2.0314348091429677 89.2 0.7599999999999909
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 155 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 264 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf9 2.3236298452294624 2.156907976575644 89.03999999999999 0.9200000000000017
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 159 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv fp16 11 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 160 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf10 2.3224369486241603 2.1560351277882046 89.03999999999999 0.9200000000000017
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 159 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv fp16 11 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 160 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf11 2.358467412507993 2.1904290636262784 89.02 0.9399999999999977
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 159 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 266 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 264 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 160 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf12 2.3633503986583126 2.1980949050120437 88.88000000000001 1.079999999999984
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 159 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 264 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 160 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf13 2.4903388172036043 2.3063593441573564 88.82 1.1400000000000006
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 160 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf14 2.508156996742662 2.3204109539869595 88.78 1.1799999999999926
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 156 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf15 2.4818531813049622 2.2910866330696744 88.75999999999999 1.2000000000000028
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 263 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 160 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf16 2.4591564896606 2.272664410995804 88.74 1.2199999999999989
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 263 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 160 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf17 2.5370582721089496 2.3464665753522405 88.72 1.2399999999999949
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf18 2.438100014978735 2.257620696759345 88.7 1.259999999999991
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 263 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 160 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf19 2.4776935382337006 2.2949598026093168 88.7 1.259999999999991
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 160 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf20 2.4380041604279596 2.254330054479329 88.68 1.279999999999987
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 157 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 160 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf21 2.4745444350223327 2.2883888475386525 88.64 1.3199999999999932
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 160 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf22 2.4136652022060625 2.2360545757445407 88.52 1.4399999999999977
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv perf_fp16 160 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 160 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf23 2.510093966915115 2.316437144001897 88.52 1.4399999999999977
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 266 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf24 2.475990790728594 2.28127562431577 88.5 1.4599999999999937
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv fp16 11 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf25 2.4761929121466926 2.290365501363375 88.5 1.4599999999999937
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf26 2.4763575559033875 2.291312348847263 88.5 1.4599999999999937
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 160 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf27 2.600249602991055 2.4123747341424644 88.06 1.8999999999999915
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv perf_fp16 165 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf28 2.596077615026303 2.4115375655840245 88.02 1.9399999999999977
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv perf_fp16 166 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf29 2.580888020555937 2.3840829703999833 87.88 2.0799999999999983
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv samp_fp16 269 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf30 2.556352783745439 2.3641413704751537 87.8 2.1599999999999966
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv samp_fp16 269 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf31 2.5559756082494527 2.3677471703724575 87.78 2.1799999999999926
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 11 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf32 2.597413373332546 2.4091972878097585 87.76 2.1999999999999886
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv perf_fp16 164 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf33 2.4797467027434656 2.2874608793842612 87.74 2.219999999999999
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf34 2.593675604602072 2.400513932866452 87.7 2.259999999999991
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv perf_fp16 156 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 264 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf35 2.6300759173431336 2.432687374579977 87.62 2.339999999999989
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 266 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf36 2.5907083037103864 2.4042762580264356 87.6 2.3599999999999994
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv perf_fp16 156 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf37 2.6143261650366187 2.423427684623993 87.6 2.3599999999999994
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv perf_fp16 156 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf38 2.6144436259117203 2.4231961521843344 87.6 2.3599999999999994
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv perf_fp16 156 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf39 2.662088796913144 2.4660859696742032 87.6 2.3599999999999994
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf40 2.6210428708834517 2.423389791646294 87.58 2.3799999999999955
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv perf_fp16 156 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 265 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf41 2.6399924349243533 2.4443864221157914 87.58 2.3799999999999955
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv perf_fp16 156 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf42 2.616443708384916 2.4217582570150697 87.58 2.3799999999999955
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf43 2.6883473596205225 2.5036952786284137 87.5 2.4599999999999937
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv perf_fp16 166 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv perf_fp16 156 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf44 2.6117356623585875 2.420771216556161 87.48 2.4799999999999898
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv perf_fp16 156 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf45 2.6359174040106708 2.444231592562593 87.48 2.4799999999999898
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf46 2.56504192294198 2.371871906722655 87.44 2.519999999999996
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv fp16 11 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf47 2.5652588453899727 2.3816996471861174 87.44 2.519999999999996
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf48 2.68806951500876 2.5007647690311425 87.14 2.819999999999993
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv perf_fp16 166 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv perf_fp16 156 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
diff --git a/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/vgg16_cifar100/vgg16_cifar100.txt b/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/vgg16_cifar100/vgg16_cifar100.txt
new file mode 100644
index 0000000000000000000000000000000000000000..2c29bedd096aec2c7f66afbe729353e372fac403
--- /dev/null
+++ b/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/vgg16_cifar100/vgg16_cifar100.txt
@@ -0,0 +1,970 @@
+3768.819777999999
++++++
+conf1 1 1 66.5 0.0
+1 gpu conv fp32 11 add fp32 1 relu fp32 1
+2 gpu conv fp32 11 add fp32 1 relu fp32 1 pool_max fp32 1
+3 gpu conv fp32 11 add fp32 1 relu fp32 1
+4 gpu conv fp32 11 add fp32 1 relu fp32 1 pool_max fp32 1
+5 gpu conv fp32 11 add fp32 1 relu fp32 1
+6 gpu conv fp32 11 add fp32 1 relu fp32 1
+7 gpu conv fp32 11 add fp32 1 relu fp32 1 pool_max fp32 1
+8 gpu conv fp32 11 add fp32 1 relu fp32 1
+9 gpu conv fp32 11 add fp32 1 relu fp32 1
+10 gpu conv fp32 11 add fp32 1 relu fp32 1 pool_max fp32 1
+11 gpu conv fp32 11 add fp32 1 relu fp32 1
+12 gpu conv fp32 11 add fp32 1 relu fp32 1
+13 gpu conv fp32 11 add fp32 1 relu fp32 1 pool_max fp32 1
+14 gpu mul fp32 11 add fp32 1 relu fp32 1
+15 gpu mul fp32 11 add fp32 1
+16 gpu softmax fp32 1
+-----
++++++
+conf2 2.2877724452131787 2.08025704453875 66.45 0.04999999999999716
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 153 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 266 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 269 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 268 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 162 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 266 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf3 2.5314658805383816 2.30737681453141 66.45 0.04999999999999716
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf4 2.044123178914057 1.8616966918258782 66.32000000000001 0.1799999999999926
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 168 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 268 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 11 add fp16 1 relu fp16 1
+13 gpu conv fp16 11 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf5 2.231179358259141 2.0317825813373864 66.18 0.3199999999999932
+1 gpu conv fp16 11 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 161 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv fp16 12 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 269 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 265 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf6 2.2474834421641057 2.0338639876373272 65.88000000000001 0.6199999999999903
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 266 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 268 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 267 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 265 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 267 add fp16 1 relu fp16 1
+13 gpu conv fp16 11 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf7 2.22281439516094 2.0205460706906377 65.88000000000001 0.6199999999999903
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 155 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 268 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 161 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 167 add fp16 1 relu fp16 1
+12 gpu conv perf_fp16 161 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf8 2.1625085012968484 1.94560449637282 65.88000000000001 0.6199999999999903
+1 gpu conv fp16 11 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 266 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv fp16 11 add fp16 1 relu fp16 1
+10 gpu conv fp16 11 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 264 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 263 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf9 2.639337323402163 2.3960416499256825 65.8 0.7000000000000028
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 269 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf10 2.672718090670276 2.4276905528801507 65.68 0.8199999999999932
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf11 2.699089631751789 2.446114054498494 65.68 0.8199999999999932
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf12 2.6003752638648767 2.3553067802112344 65.64 0.8599999999999994
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv fp16 11 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf13 2.638763904718665 2.395072565223988 65.64 0.8599999999999994
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 268 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf14 2.6003752638648767 2.3553067802112344 65.64 0.8599999999999994
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv fp16 11 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf15 2.6003752638648767 2.3553067802112344 65.64 0.8599999999999994
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv fp16 11 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf16 2.6732183804279006 2.4287517162140326 65.62 0.8799999999999955
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf17 2.6728394017929027 2.428768169588016 65.60000000000001 0.8999999999999915
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf18 2.4549989178389238 2.2406620346549433 65.56 0.9399999999999977
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 156 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 266 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf19 2.673556689244081 2.429092581627209 65.52 0.980000000000004
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf20 2.6525635304451756 2.406830663552284 65.5 1.0
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 263 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf21 2.6692288605087553 2.423462800937785 65.5 1.0
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf22 2.583650505571873 2.3471533059252194 65.48 1.019999999999996
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 160 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 263 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf23 2.6474572655420125 2.400471260394867 65.48 1.019999999999996
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 265 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf24 2.4710116424304736 2.2555966923178996 65.46 1.0400000000000063
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 161 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 266 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf25 2.557911102074785 2.3292661683311526 65.46 1.0400000000000063
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 156 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf26 2.6032957018479532 2.367574146141511 65.44 1.0600000000000023
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 163 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf27 2.6029968728098916 2.3672068592437223 65.44 1.0600000000000023
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 164 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf28 2.602540311129756 2.3691028781436954 65.44 1.0600000000000023
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 167 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf29 2.602756708588441 2.3708111025211718 65.44 1.0600000000000023
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 168 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf30 2.603240857443844 2.3662875785790183 65.44 1.0600000000000023
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 157 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf31 2.602882717372841 2.368011704225619 65.44 1.0600000000000023
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 160 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf32 2.67999343314603 2.4305182001043826 65.4 1.0999999999999943
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf33 2.670314990364046 2.4275308713267485 65.38000000000001 1.1199999999999903
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf34 2.650982630033638 2.405821467700663 65.36 1.1400000000000006
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 263 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf35 2.6507266317871756 2.405938171802741 65.36 1.1400000000000006
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 265 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf36 2.6523068534836174 2.406695716686769 65.34 1.1599999999999966
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 264 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf37 2.6533198495191073 2.4077689394073865 65.34 1.1599999999999966
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 264 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf38 2.64630900155657 2.4073892305914986 65.32 1.1800000000000068
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 152 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf39 2.6725522534379413 2.42903505877629 65.32 1.1800000000000068
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf40 2.6435249267602225 2.403536258709464 65.3 1.2000000000000028
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 161 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf41 2.6442059720503557 2.4037376163252024 65.3 1.2000000000000028
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf42 2.6536933126724027 2.4077527693156053 65.3 1.2000000000000028
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 264 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf43 2.6442798101298948 2.4056031584129225 65.3 1.2000000000000028
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf44 2.603921271336049 2.3665955131107683 65.28 1.2199999999999989
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 157 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf45 2.4967248028856828 2.2748997625822716 65.25999999999999 1.240000000000009
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 266 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 157 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf46 2.4963953691980665 2.2764932409573166 65.25999999999999 1.240000000000009
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 266 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 160 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf47 2.678944927989822 2.4251978482969956 65.24 1.2600000000000051
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 264 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf48 2.6727135417173904 2.428897140422096 65.22 1.2800000000000011
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf49 2.600256135586627 2.355428067042657 65.16 1.3400000000000034
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 151 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+12 gpu conv fp16 11 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf50 2.264460006128871 2.058037581586567 64.9 1.5999999999999943
+1 gpu conv fp16 11 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 155 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv perf_fp16 165 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv fp16 12 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 269 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 164 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 263 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 265 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
++++++
+conf51 2.2817447204106736 2.0758846029697513 64.84 1.6599999999999966
+1 gpu conv fp16 11 add fp16 1 relu fp16 1
+2 gpu conv perf_fp16 155 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv perf_fp16 165 add fp16 1 relu fp16 1
+4 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv fp16 12 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 267 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+9 gpu conv perf_fp16 155 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv perf_fp16 160 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 265 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1
+16 gpu softmax fp32 1
+-----
diff --git a/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/vgg16_imagenet/vgg16_imagenet.txt b/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/vgg16_imagenet/vgg16_imagenet.txt
new file mode 100644
index 0000000000000000000000000000000000000000..108a101c810f4ebe488e6f2029be4d970d7869a2
--- /dev/null
+++ b/llvm/projects/hpvm-tensor-rt/PPoPP_results/soc_sim_results/dev_time/vgg16_imagenet/vgg16_imagenet.txt
@@ -0,0 +1,561 @@
+19194.623482
++++++
+conf1 1 1 72.84 0.0
+1 gpu conv fp32 11 add fp32 1 relu fp32 1
+2 gpu conv fp32 11 add fp32 1 relu fp32 1 pool_max fp32 1
+3 gpu conv fp32 11 add fp32 1 relu fp32 1
+4 gpu conv fp32 11 add fp32 1 relu fp32 1 pool_max fp32 1
+5 gpu conv fp32 11 add fp32 1 relu fp32 1
+6 gpu conv fp32 11 add fp32 1 relu fp32 1
+7 gpu conv fp32 11 add fp32 1 relu fp32 1 pool_max fp32 1
+8 gpu conv fp32 11 add fp32 1 relu fp32 1
+9 gpu conv fp32 11 add fp32 1 relu fp32 1
+10 gpu conv fp32 11 add fp32 1 relu fp32 1 pool_max fp32 1
+11 gpu conv fp32 11 add fp32 1 relu fp32 1
+12 gpu conv fp32 11 add fp32 1 relu fp32 1
+13 gpu conv fp32 11 add fp32 1 relu fp32 1 pool_max fp32 1
+14 gpu mul fp32 11 add fp32 1 relu fp32 1
+15 gpu mul fp32 11 add fp32 1 relu fp32 1
+16 gpu mul fp32 11 add fp32 1
+17 gpu softmax fp32 1
+-----
++++++
+conf2 2.0787477568568082 1.7725701909562666 72.76 0.0799999999999983
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 267 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv fp16 12 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv fp16 12 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf3 2.2877881266029436 1.9268677640464096 72.04 0.7999999999999972
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv fp16 12 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf4 2.493698381711785 2.0336802939709626 72.02 0.8200000000000074
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 162 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv fp16 12 add fp16 1 relu fp16 1
+9 gpu conv fp16 12 add fp16 1 relu fp16 1
+10 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf5 2.164723960411776 1.8442442134020163 71.94 0.9000000000000057
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 267 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf6 2.53794461743687 2.069640641367895 71.67999999999999 1.1600000000000108
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 162 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv fp16 12 add fp16 1 relu fp16 1
+9 gpu conv fp16 12 add fp16 1 relu fp16 1
+10 gpu conv perf_fp16 156 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf7 1.7943268128686711 1.6103705347377417 71.58 1.2600000000000051
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv fp16 12 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv fp16 12 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 266 add fp16 1 relu fp16 1
+10 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv fp16 11 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf8 1.8143284638396158 1.6288620764171362 71.5 1.3400000000000034
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv fp16 12 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv fp16 12 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 266 add fp16 1 relu fp16 1
+10 gpu conv perf_fp16 162 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv fp16 11 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf9 2.5462742331906263 2.076061630349781 71.48 1.3599999999999994
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 167 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv fp16 12 add fp16 1 relu fp16 1
+9 gpu conv fp16 12 add fp16 1 relu fp16 1
+10 gpu conv perf_fp16 156 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf10 2.526515422129153 2.063839193109964 71.39999999999999 1.440000000000012
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 162 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+9 gpu conv fp16 12 add fp16 1 relu fp16 1
+10 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv fp16 11 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf11 2.1596661517243856 1.8351710968407349 71.34 1.5
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 267 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 268 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv perf_fp16 156 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf12 2.3444383477958337 1.981259839350623 71.22 1.6200000000000045
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+9 gpu conv fp16 12 add fp16 1 relu fp16 1
+10 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf13 1.8402020049200172 1.652343405000522 71.2 1.6400000000000006
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv fp16 12 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 266 add fp16 1 relu fp16 1
+10 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+13 gpu conv fp16 11 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf14 2.6420417968257306 2.167425635999969 71.12 1.7199999999999989
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 167 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+9 gpu conv fp16 12 add fp16 1 relu fp16 1
+10 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 155 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf15 2.543198098440602 2.0805826545876145 71.1 1.740000000000009
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 162 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+9 gpu conv fp16 12 add fp16 1 relu fp16 1
+10 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf16 2.6224991911009328 2.1476958232678807 70.89999999999999 1.940000000000012
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 167 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+9 gpu conv fp16 12 add fp16 1 relu fp16 1
+10 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf17 2.5978010917593752 2.131515210392801 70.8 2.0400000000000063
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 162 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+9 gpu conv fp16 12 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 157 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf18 2.623210258119482 2.156636511928761 70.76 2.0799999999999983
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 160 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv fp16 12 add fp16 1 relu fp16 1
+9 gpu conv fp16 12 add fp16 1 relu fp16 1
+10 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf19 2.598187894495609 2.1322228990374104 70.76 2.0799999999999983
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 162 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+9 gpu conv fp16 12 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 160 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf20 2.640464221374653 2.1682626030871295 70.76 2.0799999999999983
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 167 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+9 gpu conv fp16 12 add fp16 1 relu fp16 1
+10 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf21 2.659563405662692 2.1881035849678936 70.54 2.299999999999997
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 160 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+9 gpu conv fp16 12 add fp16 1 relu fp16 1
+10 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf22 2.636584103560761 2.1652496021557557 70.39999999999999 2.440000000000012
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 165 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+9 gpu conv fp16 12 add fp16 1 relu fp16 1
+10 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf23 2.6315080449303547 2.161259580137757 70.38 2.460000000000008
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 162 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+9 gpu conv fp16 12 add fp16 1 relu fp16 1
+10 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf24 2.7367939789033153 2.263326406058847 70.34 2.5
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 160 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf25 2.712182817327382 2.2404693918737233 70.24000000000001 2.5999999999999943
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 168 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf26 2.660510795888948 2.187299344706456 70.22 2.6200000000000045
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 159 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+9 gpu conv fp16 12 add fp16 1 relu fp16 1
+10 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf27 2.457573203839654 2.0936930776435383 70.1 2.740000000000009
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv fp16 12 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv fp16 12 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+10 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv samp_fp16 262 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv samp_fp16 261 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----
++++++
+conf28 2.7452293174567757 2.2593302388139347 69.92 2.9200000000000017
+1 gpu conv fp16 12 add fp16 1 relu fp16 1
+2 gpu conv samp_fp16 262 add fp16 1 relu fp16 1 pool_max fp16 1
+3 gpu conv fp16 12 add fp16 1 relu fp16 1
+4 gpu conv perf_fp16 159 add fp16 1 relu fp16 1 pool_max fp16 1
+5 gpu conv fp16 12 add fp16 1 relu fp16 1
+6 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+7 gpu conv samp_fp16 266 add fp16 1 relu fp16 1 pool_max fp16 1
+8 gpu conv fp16 12 add fp16 1 relu fp16 1
+9 gpu conv samp_fp16 261 add fp16 1 relu fp16 1
+10 gpu conv perf_fp16 152 add fp16 1 relu fp16 1 pool_max fp16 1
+11 gpu conv fp16 12 add fp16 1 relu fp16 1
+12 gpu conv fp16 12 add fp16 1 relu fp16 1
+13 gpu conv perf_fp16 151 add fp16 1 relu fp16 1 pool_max fp16 1
+14 gpu mul fp16 12 add fp16 1 relu fp16 1
+15 gpu mul fp16 12 add fp16 1 relu fp16 1
+16 gpu mul fp16 12 add fp16 1
+17 gpu softmax fp32 1
+-----