Skip to content

[AMDGPU] Fix folding clamp into pseudo scalar instructions #100568

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 3 commits into from
Jul 25, 2024
Merged

[AMDGPU] Fix folding clamp into pseudo scalar instructions #100568

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
578adb3
[AMDGPU] Fix SIFoldOperands when folding clamp into pseudo scalar ins…
mbrkusanin Jul 25, 2024
ed083f0
fix typo
mbrkusanin Jul 25, 2024
e206cce
test update
mbrkusanin Jul 25, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
13 changes: 12 additions & 1 deletion llvm/lib/Target/AMDGPU/SIFoldOperands.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1581,7 +1581,18 @@ bool SIFoldOperands::tryFoldClamp(MachineInstr &MI) {

// Clamp is applied after omod, so it is OK if omod is set.
DefClamp->setImm(1);
MRI->replaceRegWith(MI.getOperand(0).getReg(), Def->getOperand(0).getReg());

Register DefReg = Def->getOperand(0).getReg();
Register MIDstReg = MI.getOperand(0).getReg();
if (TRI->isSGPRReg(*MRI, DefReg)) {
// Pseudo scalar instructions have a SGPR for dst and clamp is a v_max*
// instruction with a VGPR dst.
BuildMI(*MI.getParent(), MI, MI.getDebugLoc(), TII->get(AMDGPU::COPY),
MIDstReg)
.addReg(DefReg);
} else {
MRI->replaceRegWith(MIDstReg, DefReg);
}
MI.eraseFromParent();

// Use of output modifiers forces VOP3 encoding for a VOP2 mac/fmac
Expand Down
26 changes: 26 additions & 0 deletions llvm/test/CodeGen/AMDGPU/si-fold-scalar-clamp.mir
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,26 @@
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 5
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -run-pass=si-fold-operands -verify-machineinstrs -o - %s | FileCheck %s
---
name: test
tracksRegLiveness: true
body: |
bb.0:
liveins: $sgpr0

; CHECK-LABEL: name: test
; CHECK: liveins: $sgpr0
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
; CHECK-NEXT: [[V_S_RSQ_F32_e64_:%[0-9]+]]:sgpr_32 = nofpexcept V_S_RSQ_F32_e64 0, [[COPY]], 1, 0, implicit $mode, implicit $exec
; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[V_S_RSQ_F32_e64_]]
; CHECK-NEXT: [[V_ADD_F32_e32_:%[0-9]+]]:vgpr_32 = nofpexcept V_ADD_F32_e32 [[COPY1]], [[COPY1]], implicit $mode, implicit $exec
; CHECK-NEXT: $vgpr0 = COPY [[V_ADD_F32_e32_]]
; CHECK-NEXT: S_ENDPGM 0
%0:sgpr_32 = COPY $sgpr0
%1:sgpr_32 = nofpexcept V_S_RSQ_F32_e64 0, %0, 0, 0, implicit $mode, implicit $exec
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why does this have a V_ prefix if it's just an S_ instruction?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It is one of new pseudo scalar instructions. It behaves like SOP1 but it is actually a VOP3 and runs on VALU.

%2:vgpr_32 = nofpexcept V_MAX_F32_e64 0, %1, 0, %1, -1, 0, implicit $mode, implicit $exec
%3:vgpr_32 = nofpexcept V_ADD_F32_e32 %2:vgpr_32, %2:vgpr_32, implicit $mode, implicit $exec
$vgpr0 = COPY %3
S_ENDPGM 0

...
Loading