Matt Sinclair has submitted this change. (
https://gem5-review.googlesource.com/c/public/gem5/+/45346 )
Change subject: arch-gcn3,gpu-compute: Set gpuDynInst exec_mask before use
......................................................................
arch-gcn3,gpu-compute: Set gpuDynInst exec_mask before use
vector_register_file uses the exec_mask of a memory instruction in
order to determine if it should mark a register as in-use or not.
Previously, the exec_mask of memory instructions was only set on
execution of that instruction, which occurs after the code in
vector_register_file. This led to the code reading potentially garbage
data, leading to a scenario where a register would be marked used when
it shouldn't be.
This fix sets the exec_mask of memory instructions in schedule_stage,
which works because the only time the wavefront execMask() is updated is
on a instruction executing, and we know the previous instruction will
have executed by the time schedule_stage executes, due to the order the
pipeline is executed in.
This also undoes part of a patch from last year (62ec973) which treated
the symptom of accidental register allocation, without preventing the
registers from being allocated in the first place.
This patch also removes now redundant code that sets the exec_mask in
instructions.cc for memory instructions
Change-Id: Idabd35020000764fb06133ac2458606c1aaf6f04
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/45346
Reviewed-by: Matt Sinclair <mattdsincl...@gmail.com>
Reviewed-by: Matthew Poremba <matthew.pore...@amd.com>
Maintainer: Matthew Poremba <matthew.pore...@amd.com>
Tested-by: kokoro <noreply+kok...@google.com>
---
M src/arch/amdgpu/gcn3/insts/instructions.cc
M src/gpu-compute/schedule_stage.cc
M src/gpu-compute/vector_register_file.cc
3 files changed, 30 insertions(+), 156 deletions(-)
Approvals:
Matthew Poremba: Looks good to me, approved; Looks good to me, approved
Matt Sinclair: Looks good to me, but someone else must approve
kokoro: Regressions pass
diff --git a/src/arch/amdgpu/gcn3/insts/instructions.cc
b/src/arch/amdgpu/gcn3/insts/instructions.cc
index 8c77b8c..bc66ebe 100644
--- a/src/arch/amdgpu/gcn3/insts/instructions.cc
+++ b/src/arch/amdgpu/gcn3/insts/instructions.cc
@@ -31243,7 +31243,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(
gpuDynInst->computeUnit()->cyclesToTicks(Cycles(24)));
@@ -31304,7 +31303,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(
gpuDynInst->computeUnit()->cyclesToTicks(Cycles(24)));
@@ -31368,7 +31366,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(
gpuDynInst->computeUnit()->cyclesToTicks(Cycles(24)));
@@ -31548,7 +31545,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(
gpuDynInst->computeUnit()->cyclesToTicks(Cycles(24)));
@@ -31608,7 +31604,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(
gpuDynInst->computeUnit()->cyclesToTicks(Cycles(24)));
@@ -32073,7 +32068,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(
gpuDynInst->computeUnit()->cyclesToTicks(Cycles(24)));
@@ -32135,7 +32129,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(
gpuDynInst->computeUnit()->cyclesToTicks(Cycles(24)));
@@ -32200,7 +32193,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(
gpuDynInst->computeUnit()->cyclesToTicks(Cycles(24)));
@@ -32284,7 +32276,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(
gpuDynInst->computeUnit()->cyclesToTicks(Cycles(24)));
@@ -32365,7 +32356,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(
gpuDynInst->computeUnit()->cyclesToTicks(Cycles(24)));
@@ -32547,7 +32537,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()
->cyclesToTicks(Cycles(24)));
@@ -32619,7 +32608,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()
->cyclesToTicks(Cycles(24)));
@@ -32924,7 +32912,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(
gpuDynInst->computeUnit()->cyclesToTicks(Cycles(24)));
@@ -32985,7 +32972,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(
gpuDynInst->computeUnit()->cyclesToTicks(Cycles(24)));
@@ -33521,7 +33507,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(
gpuDynInst->computeUnit()->cyclesToTicks(Cycles(24)));
@@ -33583,7 +33568,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(
gpuDynInst->computeUnit()->cyclesToTicks(Cycles(24)));
@@ -33648,7 +33632,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(
gpuDynInst->computeUnit()->cyclesToTicks(Cycles(24)));
@@ -35049,7 +35032,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -35180,7 +35162,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -35311,7 +35292,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -35411,7 +35391,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -35516,7 +35495,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -35626,7 +35604,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -35741,7 +35718,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -35836,7 +35812,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -35931,7 +35906,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -36026,7 +36000,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -36125,7 +36098,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -36228,7 +36200,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -36351,7 +36322,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -36409,7 +36379,6 @@
{
Wavefront *wf = gpuDynInst->wavefront();
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -39425,19 +39394,15 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->rdGmReqsInPipe--;
wf->rdLmReqsInPipe--;
- gpuDynInst->exec_mask = wf->execMask();
- wf->computeUnit->vrf[wf->simdId]->
- scheduleWriteOperandsFromLoad(wf, gpuDynInst);
return;
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = gpuDynInst->wavefront()->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -39498,19 +39463,15 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->rdGmReqsInPipe--;
wf->rdLmReqsInPipe--;
- gpuDynInst->exec_mask = wf->execMask();
- wf->computeUnit->vrf[wf->simdId]->
- scheduleWriteOperandsFromLoad(wf, gpuDynInst);
return;
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = gpuDynInst->wavefront()->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -39570,19 +39531,15 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->rdGmReqsInPipe--;
wf->rdLmReqsInPipe--;
- gpuDynInst->exec_mask = wf->execMask();
- wf->computeUnit->vrf[wf->simdId]->
- scheduleWriteOperandsFromLoad(wf, gpuDynInst);
return;
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = gpuDynInst->wavefront()->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -39671,19 +39628,15 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->rdGmReqsInPipe--;
wf->rdLmReqsInPipe--;
- gpuDynInst->exec_mask = wf->execMask();
- wf->computeUnit->vrf[wf->simdId]->
- scheduleWriteOperandsFromLoad(wf, gpuDynInst);
return;
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = gpuDynInst->wavefront()->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -39744,19 +39697,15 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->rdGmReqsInPipe--;
wf->rdLmReqsInPipe--;
- gpuDynInst->exec_mask = wf->execMask();
- wf->computeUnit->vrf[wf->simdId]->
- scheduleWriteOperandsFromLoad(wf, gpuDynInst);
return;
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = gpuDynInst->wavefront()->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -39817,19 +39766,15 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->rdGmReqsInPipe--;
wf->rdLmReqsInPipe--;
- gpuDynInst->exec_mask = wf->execMask();
- wf->computeUnit->vrf[wf->simdId]->
- scheduleWriteOperandsFromLoad(wf, gpuDynInst);
return;
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -39899,19 +39844,15 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->rdGmReqsInPipe--;
wf->rdLmReqsInPipe--;
- gpuDynInst->exec_mask = wf->execMask();
- wf->computeUnit->vrf[wf->simdId]->
- scheduleWriteOperandsFromLoad(wf, gpuDynInst);
return;
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -39984,7 +39925,7 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->wrGmReqsInPipe--;
@@ -39993,7 +39934,6 @@
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -40053,7 +39993,7 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->wrGmReqsInPipe--;
@@ -40062,7 +40002,6 @@
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -40122,7 +40061,7 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->wrGmReqsInPipe--;
@@ -40131,7 +40070,6 @@
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -40192,7 +40130,7 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->wrGmReqsInPipe--;
@@ -40201,7 +40139,6 @@
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -40262,7 +40199,7 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->wrGmReqsInPipe--;
@@ -40271,7 +40208,6 @@
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -40340,7 +40276,7 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->wrGmReqsInPipe--;
@@ -40349,7 +40285,6 @@
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -40428,23 +40363,17 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->wrGmReqsInPipe--;
wf->rdGmReqsInPipe--;
wf->wrLmReqsInPipe--;
wf->rdLmReqsInPipe--;
- if (instData.GLC) {
- gpuDynInst->exec_mask = wf->execMask();
- wf->computeUnit->vrf[wf->simdId]->
- scheduleWriteOperandsFromLoad(wf, gpuDynInst);
- }
return;
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -40537,23 +40466,17 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->wrGmReqsInPipe--;
wf->rdGmReqsInPipe--;
wf->wrLmReqsInPipe--;
wf->rdLmReqsInPipe--;
- if (instData.GLC) {
- gpuDynInst->exec_mask = wf->execMask();
- wf->computeUnit->vrf[wf->simdId]->
- scheduleWriteOperandsFromLoad(wf, gpuDynInst);
- }
return;
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -40647,23 +40570,17 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->wrGmReqsInPipe--;
wf->rdGmReqsInPipe--;
wf->wrLmReqsInPipe--;
wf->rdLmReqsInPipe--;
- if (instData.GLC) {
- gpuDynInst->exec_mask = wf->execMask();
- wf->computeUnit->vrf[wf->simdId]->
- scheduleWriteOperandsFromLoad(wf, gpuDynInst);
- }
return;
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -40744,23 +40661,17 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->wrGmReqsInPipe--;
wf->rdGmReqsInPipe--;
wf->wrLmReqsInPipe--;
wf->rdLmReqsInPipe--;
- if (instData.GLC) {
- gpuDynInst->exec_mask = wf->execMask();
- wf->computeUnit->vrf[wf->simdId]->
- scheduleWriteOperandsFromLoad(wf, gpuDynInst);
- }
return;
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -41015,23 +40926,17 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->wrGmReqsInPipe--;
wf->rdGmReqsInPipe--;
wf->wrLmReqsInPipe--;
wf->rdLmReqsInPipe--;
- if (instData.GLC) {
- gpuDynInst->exec_mask = wf->execMask();
- wf->computeUnit->vrf[wf->simdId]->
- scheduleWriteOperandsFromLoad(wf, gpuDynInst);
- }
return;
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -41112,23 +41017,17 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->wrGmReqsInPipe--;
wf->rdGmReqsInPipe--;
wf->wrLmReqsInPipe--;
wf->rdLmReqsInPipe--;
- if (instData.GLC) {
- gpuDynInst->exec_mask = wf->execMask();
- wf->computeUnit->vrf[wf->simdId]->
- scheduleWriteOperandsFromLoad(wf, gpuDynInst);
- }
return;
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -41238,23 +41137,17 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->wrGmReqsInPipe--;
wf->rdGmReqsInPipe--;
wf->wrLmReqsInPipe--;
wf->rdLmReqsInPipe--;
- if (instData.GLC) {
- gpuDynInst->exec_mask = wf->execMask();
- wf->computeUnit->vrf[wf->simdId]->
- scheduleWriteOperandsFromLoad(wf, gpuDynInst);
- }
return;
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -41349,23 +41242,17 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->wrGmReqsInPipe--;
wf->rdGmReqsInPipe--;
wf->wrLmReqsInPipe--;
wf->rdLmReqsInPipe--;
- if (instData.GLC) {
- gpuDynInst->exec_mask = wf->execMask();
- wf->computeUnit->vrf[wf->simdId]->
- scheduleWriteOperandsFromLoad(wf, gpuDynInst);
- }
return;
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -41448,23 +41335,17 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->wrGmReqsInPipe--;
wf->rdGmReqsInPipe--;
wf->wrLmReqsInPipe--;
wf->rdLmReqsInPipe--;
- if (instData.GLC) {
- gpuDynInst->exec_mask = wf->execMask();
- wf->computeUnit->vrf[wf->simdId]->
- scheduleWriteOperandsFromLoad(wf, gpuDynInst);
- }
return;
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -41729,23 +41610,17 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->wrGmReqsInPipe--;
wf->rdGmReqsInPipe--;
wf->wrLmReqsInPipe--;
wf->rdLmReqsInPipe--;
- if (instData.GLC) {
- gpuDynInst->exec_mask = wf->execMask();
- wf->computeUnit->vrf[wf->simdId]->
- scheduleWriteOperandsFromLoad(wf, gpuDynInst);
- }
return;
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
@@ -41829,23 +41704,17 @@
{
Wavefront *wf = gpuDynInst->wavefront();
- if (wf->execMask().none()) {
+ if (gpuDynInst->exec_mask.none()) {
wf->decVMemInstsIssued();
wf->decLGKMInstsIssued();
wf->wrGmReqsInPipe--;
wf->rdGmReqsInPipe--;
wf->wrLmReqsInPipe--;
wf->rdLmReqsInPipe--;
- if (instData.GLC) {
- gpuDynInst->exec_mask = wf->execMask();
- wf->computeUnit->vrf[wf->simdId]->
- scheduleWriteOperandsFromLoad(wf, gpuDynInst);
- }
return;
}
gpuDynInst->execUnitId = wf->execUnitId;
- gpuDynInst->exec_mask = wf->execMask();
gpuDynInst->latency.init(gpuDynInst->computeUnit());
gpuDynInst->latency.set(gpuDynInst->computeUnit()->clockPeriod());
diff --git a/src/gpu-compute/schedule_stage.cc
b/src/gpu-compute/schedule_stage.cc
index b3c91a8..88a758f 100644
--- a/src/gpu-compute/schedule_stage.cc
+++ b/src/gpu-compute/schedule_stage.cc
@@ -585,6 +585,11 @@
computeUnit.globalMemoryPipe.acqCoalescerToken(mp);
}
+ // Set instruction's exec_mask if it's a mem operation
+ if (mp->isMemRef()) {
+ mp->exec_mask = mp->wavefront()->execMask();
+ }
+
doDispatchListTransition(j, EXREADY, schIter->first);
DPRINTF(GPUSched, "dispatchList[%d]:
fillDispatchList: "
"EMPTY->EXREADY\n", j);
diff --git a/src/gpu-compute/vector_register_file.cc
b/src/gpu-compute/vector_register_file.cc
index de2d992..dc5434d 100644
--- a/src/gpu-compute/vector_register_file.cc
+++ b/src/gpu-compute/vector_register_file.cc
@@ -90,7 +90,7 @@
* send/recv any packets, and therefore it will never free
its
* dst reg(s)
*/
- if (!ii->isLoad() || (ii->isLoad() &&
ii->exec_mask.any())) {
+ if (ii->exec_mask.any()) {
markReg(physIdx, true);
}
}
--
To view, visit https://gem5-review.googlesource.com/c/public/gem5/+/45346
To unsubscribe, or for help writing mail filters, visit
https://gem5-review.googlesource.com/settings
Gerrit-Project: public/gem5
Gerrit-Branch: develop
Gerrit-Change-Id: Idabd35020000764fb06133ac2458606c1aaf6f04
Gerrit-Change-Number: 45346
Gerrit-PatchSet: 7
Gerrit-Owner: Kyle Roarty <kyleroarty1...@gmail.com>
Gerrit-Reviewer: Alex Dutu <alexandru.d...@amd.com>
Gerrit-Reviewer: Matt Sinclair <mattdsincl...@gmail.com>
Gerrit-Reviewer: Matthew Poremba <matthew.pore...@amd.com>
Gerrit-Reviewer: kokoro <noreply+kok...@google.com>
Gerrit-MessageType: merged
_______________________________________________
gem5-dev mailing list -- gem5-dev@gem5.org
To unsubscribe send an email to gem5-dev-le...@gem5.org
%(web_page_url)slistinfo%(cgiext)s/%(_internal_name)s
