AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

P4 to Git Change 1792217 by asalmanp@asalmanp-ocl-stg on 2019/06/05 12:26:00

Browse Source 
SWDEV-132899 - [OCL][GFX10] increase the numScratchWavesPerCu in Wave32 mode and use the actual num of CUs not the total num of WGPs when calculating the scratch buffer size

	ReviewBoardURL = http://ocltc.amd.com/reviews/r/17474/

Affected files ...

... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.cpp#139 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palsettings.cpp#81 edit


[ROCm/clr commit: de52451edd]
This commit is contained in:
foreman
2019-06-05 12:35:41 -04:00
parent 00ac955a2f
commit 170f42c656
2 changed files with 4 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/clr/rocclr/runtime/device/pal/paldevice.cpp
+1 -1
View File
@@ -2003,7 +2003,7 @@ bool Device::allocScratch(uint regNum, const VirtualGPU* vgpu) {
if (scratchBuf->regNum_ > 0) {
scratchBuf->destroyMemory();
// Calculate the size of the scratch buffer for a queue
uint32_t numTotalCUs = info().maxComputeUnits_;
uint32_t numTotalCUs = properties().gfxipProperties.shaderCore.numAvailableCus;
uint32_t numMaxWaves = settings().numScratchWavesPerCu_ * numTotalCUs;
scratchBuf->size_ = static_cast<uint64_t>(info().wavefrontWidth_) * scratchBuf->regNum_ *
numMaxWaves * sizeof(uint32_t);