Skip to content
Snippets Groups Projects

Repository graph

You can move around the graph by using the arrow keys.
Select Git revision
  • hpvm-release-epochs
  • hpvm-release-epochs0
  • main default
  • v2.0
  • v2.0rc
  • v1.0
  • v0.5
7 results
Created with Raphaël 2.2.05Jul230Jun29282625242219181715141311109830May27262523211816543230Apr292827262521191613121198765432131Mar3029262524232223222120192019181716151413121312111092127Feb262423221918171614121110987543Update README.mdUpdate README.mdUpdate README.mdupdating instructionsAdding instructions for compiling mini-era benchmark for NVDLAAdding script to invoke hpvm-clang with mini-era compilationUpdating LeNet source to be NVDLA consistent (reads fp16 weights)testing images for MNISTAdding working FP16 quantization and Tranpose Dense weights scripts (tested on simulator)Adding weight manipulation scripts (FP16 + Transpose)Backing up Shubham's new scalesAdding check for file exists calib.txtAdding FP16 weight quantization script - NVDLA needs FP16 weightsCompiling AlexNet with INT8 -- pushing calib values shared by ShubhamAdding few test images for mini-era CNNMoving location of calib table fileHPVM NVDLA backend INT8 mode working with mini-era CNNfix ordering-related crash in Fusesupport for node fusion in fpga benchmark makefilesMerge branch 'hpvm-hypermapper' into hpvm-hypermapper-nodefusionUpdate README.mdPorting Mini-era CNN to HPVM-9 -- compiles with ported NVDLA passModifying hpvm-clang to skip CPU backend and link stepsReadTrainedWeights filename extraction fixed -- still not working Pass (close)Modifying readTrainedWeights calls in Alexnet to read global constantshpvm-clang: Modifying clang to use -O0 and opt to use -mem2regCorrectly handling readInputBatch -- arg ordering changedfixing tensorutils definitions to return void*Generalize input buffering function a littleFix ordering of hpvm2fpga passes in pipeline MAkefileproper memory alignment in pipeline-fpga benchesFixing tensorUtils definitions to avoid undefs in LLVM IRMake the trip count evenadd missing analyses for instcombine in hpvm2fpgasupport outer unrolling in fpga pipelin benchTweaking the passes applied in some placesadd __restrict__ to seq tokenstopgap for memory errors in DFGraphFix memory errors in ClearDFGset num frames in non-serial fpga pipeline bench