From c8cf71f1a046b4758e20d6398a654777c3fb5193 Mon Sep 17 00:00:00 2001 From: Alexandru Dutu Date: Tue, 4 Oct 2016 13:03:52 -0400 Subject: [PATCH] gpu-compute: Added method to compute the actual workgroup size This patch adds a method to the Wavefront class to compute the actual workgroup size. This can be different from the maximum workgroup size specified when launching the kernel through the NDRange object. Current solution is still not optimal, as we are computing these for each wavefront and the dispatcher also needs to have this information and can't actually call Wavefront::computeActuallWgSz before the wavefronts are being created. A long term solution would be to have a Workgroup class that deals with all these details. --- src/gpu-compute/compute_unit.cc | 50 +++++++++++++-------------------- src/gpu-compute/compute_unit.hh | 6 ++-- src/gpu-compute/wavefront.cc | 11 ++++++++ src/gpu-compute/wavefront.hh | 6 ++++ 4 files changed, 39 insertions(+), 34 deletions(-) diff --git a/src/gpu-compute/compute_unit.cc b/src/gpu-compute/compute_unit.cc index b937584eb..97e018713 100644 --- a/src/gpu-compute/compute_unit.cc +++ b/src/gpu-compute/compute_unit.cc @@ -174,7 +174,7 @@ ComputeUnit::~ComputeUnit() } void -ComputeUnit::FillKernelState(Wavefront *w, NDRange *ndr) +ComputeUnit::fillKernelState(Wavefront *w, NDRange *ndr) { w->resizeRegFiles(ndr->q.cRegCount, ndr->q.sRegCount, ndr->q.dRegCount); @@ -190,6 +190,7 @@ ComputeUnit::FillKernelState(Wavefront *w, NDRange *ndr) w->spillSizePerItem = ndr->q.spillMemPerItem; w->roBase = ndr->q.roMemStart; w->roSize = ndr->q.roMemTotal; + w->computeActualWgSz(ndr); } void @@ -220,19 +221,16 @@ ComputeUnit::updateEvents() { void -ComputeUnit::StartWF(Wavefront *w, int trueWgSize[], int trueWgSizeTotal, - int waveId, LdsChunk *ldsChunk, NDRange *ndr) +ComputeUnit::startWavefront(Wavefront *w, int waveId, LdsChunk *ldsChunk, + NDRange *ndr) { static int _n_wave = 0; - // Fill in Kernel state - FillKernelState(w, ndr); - VectorMask init_mask; init_mask.reset(); for (int k = 0; k < wfSize(); ++k) { - if (k + waveId * wfSize() < trueWgSizeTotal) + if (k + waveId * wfSize() < w->actualWgSzTotal) init_mask[k] = 1; } @@ -241,18 +239,18 @@ ComputeUnit::StartWF(Wavefront *w, int trueWgSize[], int trueWgSizeTotal, w->initMask = init_mask.to_ullong(); for (int k = 0; k < wfSize(); ++k) { - w->workItemId[0][k] = (k + waveId * wfSize()) % trueWgSize[0]; - w->workItemId[1][k] = - ((k + waveId * wfSize()) / trueWgSize[0]) % trueWgSize[1]; - w->workItemId[2][k] = - (k + waveId * wfSize()) / (trueWgSize[0] * trueWgSize[1]); - - w->workItemFlatId[k] = w->workItemId[2][k] * trueWgSize[0] * - trueWgSize[1] + w->workItemId[1][k] * trueWgSize[0] + + w->workItemId[0][k] = (k + waveId * wfSize()) % w->actualWgSz[0]; + w->workItemId[1][k] = ((k + waveId * wfSize()) / w->actualWgSz[0]) % + w->actualWgSz[1]; + w->workItemId[2][k] = (k + waveId * wfSize()) / + (w->actualWgSz[0] * w->actualWgSz[1]); + + w->workItemFlatId[k] = w->workItemId[2][k] * w->actualWgSz[0] * + w->actualWgSz[1] + w->workItemId[1][k] * w->actualWgSz[0] + w->workItemId[0][k]; } - w->barrierSlots = divCeil(trueWgSizeTotal, wfSize()); + w->barrierSlots = divCeil(w->actualWgSzTotal, wfSize()); w->barCnt.resize(wfSize(), 0); @@ -294,8 +292,8 @@ ComputeUnit::StartWF(Wavefront *w, int trueWgSize[], int trueWgSizeTotal, // is this the last wavefront in the workgroup // if set the spillWidth to be the remaining work-items // so that the vector access is correct - if ((waveId + 1) * wfSize() >= trueWgSizeTotal) { - w->spillWidth = trueWgSizeTotal - (waveId * wfSize()); + if ((waveId + 1) * wfSize() >= w->actualWgSzTotal) { + w->spillWidth = w->actualWgSzTotal - (waveId * wfSize()); } else { w->spillWidth = wfSize(); } @@ -328,17 +326,6 @@ ComputeUnit::StartWorkgroup(NDRange *ndr) injectGlobalMemFence(gpuDynInst, true); } - // Get true size of workgroup (after clamping to grid size) - int trueWgSize[3]; - int trueWgSizeTotal = 1; - - for (int d = 0; d < 3; ++d) { - trueWgSize[d] = std::min(ndr->q.wgSize[d], ndr->q.gdSize[d] - - ndr->wgId[d] * ndr->q.wgSize[d]); - - trueWgSizeTotal *= trueWgSize[d]; - } - // calculate the number of 32-bit vector registers required by wavefront int vregDemand = ndr->q.sRegCount + (2 * ndr->q.dRegCount); int wave_id = 0; @@ -350,9 +337,10 @@ ComputeUnit::StartWorkgroup(NDRange *ndr) // It must be stopped and not waiting // for a release to complete S_RETURNING if (w->status == Wavefront::S_STOPPED) { + fillKernelState(w, ndr); // if we have scheduled all work items then stop // scheduling wavefronts - if (wave_id * wfSize() >= trueWgSizeTotal) + if (wave_id * wfSize() >= w->actualWgSzTotal) break; // reserve vector registers for the scheduled wavefront @@ -365,7 +353,7 @@ ComputeUnit::StartWorkgroup(NDRange *ndr) w->reservedVectorRegs = normSize; vectorRegsReserved[m % numSIMDs] += w->reservedVectorRegs; - StartWF(w, trueWgSize, trueWgSizeTotal, wave_id, ldsChunk, ndr); + startWavefront(w, wave_id, ldsChunk, ndr); ++wave_id; } } diff --git a/src/gpu-compute/compute_unit.hh b/src/gpu-compute/compute_unit.hh index 34b710cd6..a3547402a 100644 --- a/src/gpu-compute/compute_unit.hh +++ b/src/gpu-compute/compute_unit.hh @@ -254,10 +254,10 @@ class ComputeUnit : public MemObject void exec(); void initiateFetch(Wavefront *wavefront); void fetch(PacketPtr pkt, Wavefront *wavefront); - void FillKernelState(Wavefront *w, NDRange *ndr); + void fillKernelState(Wavefront *w, NDRange *ndr); - void StartWF(Wavefront *w, int trueWgSize[], int trueWgSizeTotal, - int cnt, LdsChunk *ldsChunk, NDRange *ndr); + void startWavefront(Wavefront *w, int waveId, LdsChunk *ldsChunk, + NDRange *ndr); void StartWorkgroup(NDRange *ndr); int ReadyWorkgroup(NDRange *ndr); diff --git a/src/gpu-compute/wavefront.cc b/src/gpu-compute/wavefront.cc index 42739a7b0..c677cbe41 100644 --- a/src/gpu-compute/wavefront.cc +++ b/src/gpu-compute/wavefront.cc @@ -1066,3 +1066,14 @@ Wavefront::setContext(const void *in) ldsChunk->write(i, val); } } + +void +Wavefront::computeActualWgSz(NDRange *ndr) +{ + actualWgSzTotal = 1; + for (int d = 0; d < 3; ++d) { + actualWgSz[d] = std::min(workGroupSz[d], + gridSz[d] - ndr->wgId[d] * workGroupSz[d]); + actualWgSzTotal *= actualWgSz[d]; + } +} diff --git a/src/gpu-compute/wavefront.hh b/src/gpu-compute/wavefront.hh index ef8c80989..0df8a6c82 100644 --- a/src/gpu-compute/wavefront.hh +++ b/src/gpu-compute/wavefront.hh @@ -47,6 +47,7 @@ #include "gpu-compute/condition_register_state.hh" #include "gpu-compute/lds_state.hh" #include "gpu-compute/misc.hh" +#include "gpu-compute/ndrange.hh" #include "params/Wavefront.hh" #include "sim/sim_object.hh" @@ -189,11 +190,16 @@ class Wavefront : public SimObject std::vector lastAddr; std::vector workItemId[3]; std::vector workItemFlatId; + /* kernel launch parameters */ uint32_t workGroupId[3]; uint32_t workGroupSz[3]; uint32_t gridSz[3]; uint32_t wgId; uint32_t wgSz; + /* the actual WG size can differ than the maximum size */ + uint32_t actualWgSz[3]; + uint32_t actualWgSzTotal; + void computeActualWgSz(NDRange *ndr); // wavefront id within a workgroup uint32_t wfId; uint32_t maxDynWaveId; -- 2.30.2