derek/gem5
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

arch-vega: Implementing global_atomic_smax

Browse Source 
Change-Id: Id4053424c98eec1e98eb555bb35b48f0b5d2407b
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/64513
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
This commit is contained in:
Alexandru Dutu
2022-10-07 17:33:50 -07:00
committed by Matthew Poremba
parent 8375058e73
commit bb8f370e4d
2 changed files with 54 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

53
src/arch/amdgpu/vega/insts/instructions.cc
View File

@@ -45079,8 +45079,59 @@ namespace VegaISA
void
Inst_FLAT__FLAT_ATOMIC_SMAX::execute(GPUDynInstPtr gpuDynInst)
{
panicUnimplemented();
Wavefront *wf = gpuDynInst->wavefront();
if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
return;
}
gpuDynInst->execUnitId = wf->execUnitId;
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
ConstVecOperandU64 addr(gpuDynInst, extData.ADDR);
ConstVecOperandU32 data(gpuDynInst, extData.DATA);
addr.read();
data.read();
calcAddr(gpuDynInst, addr, extData.SADDR, instData.OFFSET);
for (int lane = 0; lane < NumVecElemPerVecReg; ++lane) {
if (gpuDynInst->exec_mask[lane]) {
(reinterpret_cast<VecElemU32*>(gpuDynInst->a_data))[lane]
= data[lane];
}
}
gpuDynInst->computeUnit()->globalMemoryPipe.
issueRequest(gpuDynInst);
} // execute
void
Inst_FLAT__FLAT_ATOMIC_SMAX::initiateAcc(GPUDynInstPtr gpuDynInst)
{
initAtomicAccess<VecElemU32>(gpuDynInst);
} // initiateAcc
void
Inst_FLAT__FLAT_ATOMIC_SMAX::completeAcc(GPUDynInstPtr gpuDynInst)
{
if (isAtomicRet()) {
VecOperandU32 vdst(gpuDynInst, extData.VDST);
for (int lane = 0; lane < NumVecElemPerVecReg; ++lane) {
if (gpuDynInst->exec_mask[lane]) {
vdst[lane] = (reinterpret_cast<VecElemU32*>(
gpuDynInst->d_data))[lane];
}
}
vdst.write();
}
} // completeAcc
// --- Inst_FLAT__FLAT_ATOMIC_UMAX class methods ---
Inst_FLAT__FLAT_ATOMIC_UMAX::Inst_FLAT__FLAT_ATOMIC_UMAX(InFmt_FLAT *iFmt)

2
src/arch/amdgpu/vega/insts/instructions.hh
View File

@@ -42691,6 +42691,8 @@ namespace VegaISA
} // getOperandSize
void execute(GPUDynInstPtr) override;
void initiateAcc(GPUDynInstPtr) override;
void completeAcc(GPUDynInstPtr) override;
}; // Inst_FLAT__FLAT_ATOMIC_SMAX
class Inst_FLAT__FLAT_ATOMIC_UMAX : public Inst_FLAT