Added predictive tuner

llvm/projects/hpvm-tensor-rt/autotuner/data/alexnet2_canny_hpvm/layers.txt

+conv  add  tanh
+conv  add  tanh  pool
+conv  add  tanh
+conv  add  tanh  pool
+conv  add  tanh
+conv  add  tanh  pool
+dense  add
+reduce
+conv
+conv
+conv
+reduce
+reduce

llvm/projects/hpvm-tensor-rt/autotuner/data/alexnet2_canny_hpvm/op_cost.txt

+468.076
+947.434
+255.422
+348.769
+256.658
+1.05427
+1.05427
+107.5062
+666.888
+432.622
+252.458
+11.51922
+2.01168

llvm/projects/hpvm-tensor-rt/autotuner/data/alexnet_imagenet/layer_composition.txt

+conv  add  activation  pool  
+conv  add  activation  pool  
+conv  add  activation  
+conv  add  activation  
+conv  add  activation  pool  
+dense  add  activation  
+dense  add  activation  
+dense  add  

llvm/projects/hpvm-tensor-rt/autotuner/data/alexnet_imagenet/op_cost.txt

+1457111.000000
+4478976.000000
+2242805.750000
+2990407.750000
+1993605.125000
+754974.750000
+335544.312500
+81920.000000

llvm/projects/hpvm-tensor-rt/autotuner/data/global_knobs.txt

@@ -18,6 +18,24 @@ perf,135	4,1,0	1.33	tensorConvolution	tensorConvApprox	dev	conv
 perf,136	4,1,1	1.33	tensorConvolution	tensorConvApprox	dev	conv
 perf,137	4,1,2	1.33	tensorConvolution	tensorConvApprox	dev	conv
 perf,138	4,1,3	1.33	tensorConvolution	tensorConvApprox	dev	conv
+perf_fp16,151	1,2,0	3.0	tensorConvolution	tensorConvApprox	install	conv
+perf_fp16,152	1,2,1	3.0	tensorConvolution	tensorConvApprox	install	conv
+perf_fp16,153	2,1,0	3.0	tensorConvolution	tensorConvApprox	install	conv
+perf_fp16,154	2,1,1	3.0	tensorConvolution	tensorConvApprox	install	conv
+perf_fp16,155	1,3,0	2.25	tensorConvolution	tensorConvApprox	install	conv
+perf_fp16,156	1,3,1	2.25	tensorConvolution	tensorConvApprox	install	conv
+perf_fp16,157	1,3,2	2.25	tensorConvolution	tensorConvApprox	install	conv
+perf_fp16,158	3,1,0	2.25	tensorConvolution	tensorConvApprox	install	conv
+perf_fp16,159	3,1,1	2.25	tensorConvolution	tensorConvApprox	install	conv
+perf_fp16,160	3,1,2	2.25	tensorConvolution	tensorConvApprox	install	conv
+perf_fp16,161	1,4,0	2.0	tensorConvolution	tensorConvApprox	install	conv
+perf_fp16,162	1,4,1	2.0	tensorConvolution	tensorConvApprox	install	conv
+perf_fp16,163	1,4,2	2.0	tensorConvolution	tensorConvApprox	install	conv
+perf_fp16,164	1,4,3	2.0	tensorConvolution	tensorConvApprox	install	conv
+perf_fp16,165	4,1,0	2.0	tensorConvolution	tensorConvApprox	install	conv
+perf_fp16,166	4,1,1	2.0	tensorConvolution	tensorConvApprox	install	conv
+perf_fp16,167	4,1,2	2.0	tensorConvolution	tensorConvApprox	install	conv
+perf_fp16,168	4,1,3	2.0	tensorConvolution	tensorConvApprox	install	conv
 samp,231	2,0,1	2.0	tensorConvolution	tensorConvApprox	dev	conv
 samp,232	2,1,1	2.0	tensorConvolution	tensorConvApprox	dev	conv
 samp,233	3,0,1	1.5	tensorConvolution	tensorConvApprox	dev	conv
@@ -27,6 +45,15 @@ samp,236	4,0,1	1.33	tensorConvolution	tensorConvApprox	dev	conv
 samp,237	4,1,1	1.33	tensorConvolution	tensorConvApprox	dev	conv
 samp,238	4,2,1	1.33	tensorConvolution	tensorConvApprox	dev	conv
 samp,239	4,3,1	1.33	tensorConvolution	tensorConvApprox	dev	conv
+samp_fp16,261	2,0,1	3.0	tensorConvolution	tensorConvApprox	install	conv
+samp_fp16,262	2,1,1	3.0	tensorConvolution	tensorConvApprox	install	conv
+samp_fp16,263	3,0,1	2.25	tensorConvolution	tensorConvApprox	install	conv
+samp_fp16,264	3,1,1	2.25	tensorConvolution	tensorConvApprox	install	conv
+samp_fp16,265	3,2,1	2.25	tensorConvolution	tensorConvApprox	install	conv
+samp_fp16,266	4,0,1	2.0	tensorConvolution	tensorConvApprox	install	conv
+samp_fp16,267	4,1,1	2.0	tensorConvolution	tensorConvApprox	install	conv
+samp_fp16,268	4,2,1	2.0	tensorConvolution	tensorConvApprox	install	conv
+samp_fp16,269	4,3,1	2.0	tensorConvolution	tensorConvApprox	install	conv
 red_samp,41	1	1.5	tensorReduction		tensorReduction		dev	red
 red_samp,42	1	2.25	tensorReduction		tensorReduction		dev	red
 red_samp,43	1	1.4	tensorReduction		tensorReduction		dev	red

llvm/projects/hpvm-tensor-rt/autotuner/data/mobilenet_torch/layers.txt

+conv
+batchnorm
+depthwise_conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+depthwise_conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+depthwise_conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+depthwise_conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+depthwise_conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+depthwise_conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+depthwise_conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+depthwise_conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+depthwise_conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+depthwise_conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+depthwise_conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+depthwise_conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+depthwise_conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+dense  add

llvm/projects/hpvm-tensor-rt/autotuner/data/mobilenet_torch/op_cost.txt

+44236.80078
+104857.6019
+104857.6019
+209715.2037
+104857.6019
+209715.2037
+104857.6019
+209715.2037
+209715.2037
+209715.2037
+209715.2037
+209715.2037
+104857.6019
+209715.2037
+256.000000

llvm/projects/hpvm-tensor-rt/autotuner/data/resnet18_torch/op_cost.txt

+88473.60156
+1887436.833
+1887436.833
+1887436.833
+1887436.833
+3774873.667
+1887436.833
+26214.40046
+1887436.833
+1887436.833
+3774873.667
+1887436.833
+13107.20023
+1887436.833
+1887436.833
+3774873.667
+1887436.833
+6553.600116
+1887436.833
+1887436.833
+64.0000000

llvm/projects/hpvm-tensor-rt/autotuner/data/resnet18_torch/resnet_layers.txt

+conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+conv
+batchnorm
+conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+conv
+batchnorm
+conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+conv
+batchnorm
+conv
+batchnorm
+conv
+batchnorm
+activation
+conv
+batchnorm
+activation
+activation
+pool_mean
+dense  add
\ No newline at end of file

llvm/projects/hpvm-tensor-rt/autotuner/data/vgg16_cifar10_torch/layers.txt

+conv  add
+batchnorm
+activation
+conv  add
+batchnorm
+activation
+pool
+conv  add
+batchnorm
+activation
+conv  add
+batchnorm
+activation
+pool
+conv  add
+batchnorm
+activation
+conv  add
+batchnorm
+activation
+conv  add
+batchnorm
+activation
+pool
+conv  add
+batchnorm
+activation
+conv  add
+batchnorm
+activation
+conv  add
+batchnorm
+activation
+pool
+conv  add
+batchnorm
+activation
+conv  add
+batchnorm
+activation
+conv  add
+batchnorm
+activation
+pool
+pool_mean
+dense  add

llvm/projects/hpvm-tensor-rt/autotuner/data/vgg16_cifar10_torch/op_cost.txt

+88473.60156
+1887436.833
+943718.4167
+1887436.833
+943718.4167
+1887436.833
+1887436.833
+943718.4167
+1887436.833
+1887436.833
+471859.2083
+471859.2083
+471859.2083
+13107.200195
+256.000000

llvm/projects/hpvm-tensor-rt/model_params/alexnet2_cifar10/quant_ranges.txt

+-1.8816435 2.0934134 -0.5421946 0.3710851 -0.06697306 0.040868897 -0.775027394891 0.779944300652 
+-0.775027394891 0.779944300652 -0.42474225 0.31460348 -0.3557253 -0.17281663 -0.808667064309 0.983953297734 
+-0.808667064309 0.983953297734 -0.44134507 0.79587924 -0.80424446 0.75330096 -0.995678424835 0.998566448689 
+-0.995678424835 0.998566448689 -0.2883836 0.31025785 -0.6353164 0.29015934 -0.993219196796 0.992379009724 
+-0.993219196796 0.992379009724 -0.2792431 0.37689754 -1.1379756 1.2391574 -0.999901354313 0.999910891056 
+-0.999901354313 0.999910891056 -0.27078503 0.27942517 -0.503003 0.12762362 -0.991036117375 0.971404970288 
+-0.991036117375 0.971404970288 -0.24273404 0.5845544 -0.53745 0.558251 -119.27973732 -25.2262819576

llvm/projects/hpvm-tensor-rt/model_params/alexnet_cifar10/quant_ranges.txt

+-1.88164262419 2.09340954985 -0.33087718 0.3323643 -0.7782218 0.6020472 -0.978641152382 0.998945295811 
+-0.978641152382 0.998945295811 -0.2095158 0.33543423 -0.45020863 0.30596754 -0.999703943729 0.999930202961 
+-0.999703943729 0.999930202961 -0.1715614 0.17037082 -0.6519161 0.5939945 -0.999933600426 0.999940037727 
+-0.999933600426 0.999940037727 -0.15575546 0.14456555 -0.55873865 0.4704539 -0.99999910593 0.999999344349 
+-0.99999910593 0.999999344349 -0.16108225 0.16864482 -0.22135437 0.10401678 -0.999434411526 0.999634206295 
+-0.999434411526 0.999634206295 -0.18183032 0.19018902 -0.07189204 0.106005594 -15.0765653801 19.4225852203 

llvm/projects/hpvm-tensor-rt/model_params/lenet_mnist/quant_ranges.txt

+0 1 -1 1 -1 1 -1 1
+-1 1 -1 1 -1 1 -1 1
+-1 1 -1 1 -1 1 -1 1
+-1 1 -1 1 -1 1 -1 1

llvm/projects/hpvm-tensor-rt/model_params/mobilenet/quant_ranges.txt

+-1.9892114 2.126797 -2.19630692005 1.34758170414  0.0  0.0  -60.892750473 51.9925691605 
+0.0 5.71354155397 -0.931772116065 1.07742589378   0.0  0.0 -6.51858950329 6.81084251881 
+0.0 4.93213940287 -0.531654466152 0.57537904036   0.0  0.0  -4.48263123512 3.96730119753 
+0.0 4.10326339769 -0.362340988219 0.407691390038   0.0  0.0  -4.04261828327 3.8867793293 
+0.0 5.38322130251 -0.313120054901 0.293576799393   0.0  0.0  -5.92146921539 4.33867932415 
+0.0 4.31673815441 -0.232992478013 0.258029025793   0.0  0.0  -4.20778994751 3.93243697071 
+0.0 5.8304081068 -0.202337772191 0.189983081758   0.0  0.0  -6.29828691578 4.84813511753 
+0.0 4.44641780996 -0.174427356511 0.176958308667  0.0  0.0   -4.34791088581 3.61443646955 
+0.0 4.5180956049 -0.145467961878 0.15256431669   0.0  0.0   -3.02877027559 2.94873657799 
+0.0 6.34857563496 -0.130258745223 0.135582433432   0.0  0.0  -4.22931008053 3.53150463724 
+0.0 5.22100311041 -0.119001727596 0.125363747835   0.0  0.0  -4.03820378017 4.00400940704 
+0.0 5.73249834776 -0.108397216856 0.116256686077    0.0  0.0  -3.31110151148 4.46293323326 
+0.0 7.24049821186 -0.0862374496162 0.0885944995135   0.0  0.0  -4.17543139458 6.2043294754 
+0.0 7.81395883465 -0.0681302513927 0.0700202777982    0.0  0.0  -10.9205664234 2.64429125786 
+0.0 2.86920666504 -0.223010196954 0.14426593782 -0.1654396 0.23336112 -12.2459499588 23.8053251343

llvm/projects/hpvm-tensor-rt/model_params/resnet18_cifar10/quant_ranges.txt

+-0.5500815 0.60786617 -1.0248864 1.2929907 -0.36291853 0.2533059 0.0 0.753551840782 
+0.0 0.753551840782 -0.69884616 0.71849966 -0.2781147 0.45571187 0.0 1.01057458043 
+0.0 1.01057458043 -0.59568167 0.7714691 -0.8602873 0.19743633 -1.84771883726 1.87930787086 
+0.0 2.33981014252 -0.41976976 0.43748936 -0.7021962 0.3033103 0.0 1.04317724705 
+0.0 1.04317724705 -0.46757826 0.4635873 -0.20662616 0.1778044 -0.829483509064 0.786805033684 
+0.0 2.49733686686 -0.64404047 0.45383143 -0.819547 0.38550296 0.0 0.897360802293 
+0.0 0.897360802293 -0.41986948 0.33654243 -0.3563013 0.22371122 -0.957150224447 0.54919362247 
+0.0 2.37362146616 -0.4805263 0.50655717 -0.296758 0.7742441 0.0 3.01592136621 
+0.0 3.01592136621 -0.52083415 0.45517674 -0.20242067 0.8236838 -5.2759475708 5.79733039856 
+0.0 2.37362146616 -0.5338656 1.3395424 -0.20242067 0.8236838 -0.738995380998 2.33600783587 
+0.0 7.07933432579 -0.34429058 0.43629733 -1.0744808 0.056708273 0.0 1.58645607233 
+0.0 1.58645607233 -0.30342352 0.39493486 -0.44630566 0.6492069 -1.49672914267 1.29970229745 
+0.0 7.11914063454 -0.38351893 0.45775774 -1.4733055 -0.014426912 0.0 1.52876508832 
+0.0 1.52876508832 -0.25695276 0.45372736 -0.5259744 0.26591402 -1.59576894164 1.08074297309 
+0.0 6.94405080318 -0.55299705 0.5443531 -0.71790683 1.2730768 0.0 10.3651468277 
+0.0 10.3651468277 -0.4203967 0.48641303 -0.90653443 1.3546854 -22.372925148 17.2033731079 
+0.0 6.94405080318 -0.4365755 0.84913826 -0.90653443 1.3546851 -3.66810325861 4.87814051151 
+0.0 18.8401451111 -0.38657624 0.5228989 -1.2083547 0.76361173 0.0 19.1229192352 
+0.0 19.1229192352 -0.40857902 0.575035 -1.8731614 1.0960501 -31.3229312897 14.8234729958 
+0.0 23.7382488823 -0.33079496 0.5893278 -1.0234511 1.0016295 0.0 19.5892774963 
+0.0 19.5892774963 -0.27897888 0.38280907 -2.2086356 1.0066502 -34.4416886902 20.9890329933 
+0.0 10.8541981602 -1.5092047 1.0279838 -0.49379802 0.61032647 -40.9121678543 25.7082381058

llvm/projects/hpvm-tensor-rt/model_params/vgg16_cifar10/quant_ranges.txt

+-1.8816367 2.0934217 -0.53275156 0.49437004 -0.6403629 0.2490165 0.0 1.35908746719 
+0.0 1.35908746719 -0.2688396 0.20639156 -0.7745511 0.82006615 0.0 2.52123117924 
+0.0 2.52123117924 -0.16776876 0.14878987 -0.35283303 0.5154362 0.0 1.20119857848 
+0.0 1.20119857848 -0.088948585 0.114222586 -0.30250227 0.36856708 0.0 1.03598809302 
+0.0 1.03598809302 -0.07739562 0.10973293 -0.15568458 0.17634983 0.0 0.300495595038 
+0.0 0.300495595038 -0.051649556 0.05435231 -0.07395447 0.07996062 0.0 0.11490475405 
+0.0 0.11490475405 -0.043513633 0.07577866 -0.06921874 0.02660573 0.0 0.16232508488 
+0.0 0.16232508488 -0.033842053 0.045218028 -0.022827804 0.023845317 0.0 0.124249965735 
+0.0 0.124249965735 -0.02211613 0.032084666 -0.02699063 0.03773564 0.0 0.174634486511 
+0.0 0.174634486511 -0.01979376 0.034854397 -0.036107242 0.07056531 0.0 0.575175762177 
+0.0 0.575175762177 -0.03452098 0.046055835 -0.051925894 0.07039055 0.0 0.771875114441 
+0.0 0.771875114441 -0.025946895 0.040090334 -0.06049362 0.12658806 0.0 1.17285169065 
+0.0 1.17285169065 -0.021766115 0.03315237 -0.20705001 0.117947325 0.0 2.00157693863 
+0.0 2.00157693863 -0.042597745 0.046707444 -0.21937433 0.2545502 0.0 2.00236111879 
+0.0 2.00236111879 -0.32550547 0.30829763 -1.1787822 1.2378151 -18.2514705467 24.1736344528

llvm/projects/hpvm-tensor-rt/model_params/vgg16_cifar100/quant_ranges.txt

+-1.7829767 1.9456929 -0.7450515 0.71249133 -1.5885142 0.275554 0.0 8.190712 
+0.0 8.190712 -0.30790088 0.43504623 -1.4242363 1.2602744 0.0 19.023172 
+0.0 19.023172 -0.29189092 0.26958522 -1.0527138 0.9075671 0.0 14.428051 
+0.0 14.428051 -0.15521508 0.1829038 -0.845419 1.9358484 0.0 23.065294 
+0.0 23.065294 -0.13149762 0.14811686 -0.7162557 1.0370971 0.0 15.165984 
+0.0 15.165984 -0.06236292 0.08321518 -0.9067523 0.9922458 0.0 13.664733 
+0.0 13.664733 -0.06471479 0.1024472 -0.15943134 0.7988499 0.0 19.025272 
+0.0 19.025272 -0.06320205 0.08291938 -0.32540628 0.5203079 0.0 6.727217 
+0.0 6.727217 -0.037707984 0.051601283 -0.25622904 0.11251946 0.0 3.2003012 
+0.0 3.2003012 -0.056007143 0.09549151 -0.11591503 0.06267536 0.0 4.321189 
+0.0 4.321189 -0.060094673 0.10868926 -0.105962686 0.09584572 0.0 2.936297 
+0.0 2.936297 -0.034618977 0.05792674 -0.4237576 0.11035452 0.0 4.87262 
+0.0 4.87262 -0.035480656 0.058295887 -0.21477045 0.14263579 0.0 10.32133 
+0.0 10.32133 -0.08929961 0.11301676 -0.20798548 0.47405547 0.0 13.91 
+0.0 13.91 -0.6627122 0.35539475 -1.0631907 0.9830786 -70.45701 87.34367 

llvm/projects/pred_tuner/.gitignore

+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Opentuner
+opentuner.db/
+opentuner.log
+
+# Custom
+.idea/
+.vscode/
+/data/
+results/
+tuner_results
+tuner_results/
+*.sh
+*.ipynb
+logistics/
+autotuner/

llvm/projects/pred_tuner/LICENSE

+MIT License
+
+Copyright (c) 2017 liukuang
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.