Skip to content

[AMDGPU] handling of COPY for wide tuples when v_pk_mov_b32 is used #119966

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

Closed
wants to merge 1 commit into from
Closed

[AMDGPU] handling of COPY for wide tuples when v_pk_mov_b32 is used #119966

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
30 changes: 30 additions & 0 deletions llvm/test/CodeGen/AMDGPU/copy_phys_vgpr1024.mir
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,30 @@
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
# RUN: llc -mtriple=amdgcn -mcpu=gfx90a -run-pass postrapseudos -verify-machineinstrs %s -o - | FileCheck -check-prefix=GFX90A %s

---
name: copy_v1024_to_v1024_fwd
tracksRegLiveness: true
body: |
bb.0:
liveins: $vgpr2_vgpr3_vgpr4_vgpr5, $vgpr32
; GFX90A-LABEL: name: copy_v1024_to_v1024_fwd
; GFX90A: liveins: $vgpr2_vgpr3_vgpr4_vgpr5, $vgpr32
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: $vgpr0_vgpr1 = V_PK_MOV_B32 8, $vgpr32_vgpr33, 12, $vgpr32_vgpr33, 0, 0, 0, 0, 0, implicit $exec, implicit $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39_vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47_vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55_vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31
Copy link
Author

@pravinjagtap pravinjagtap Dec 14, 2024 •
edited
Loading

Choose a reason for hiding this comment

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

While expanding V[0:31] = COPY V[32:63], Src tuple is marked as implicit (for every pair, not just for the first) and Dst tuple is marked as implicit-def (in the first and last).
Everything looks fine here or do we also need to mark Src tuple as implicit-def ?

For other sized tuples, I see the similar thing.

Copy link
Contributor

@arsenm arsenm Dec 16, 2024

Choose a reason for hiding this comment

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

The last instruction is picking up an extra exec use for some reason

Copy link
Author

@pravinjagtap pravinjagtap Dec 16, 2024

Choose a reason for hiding this comment

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

But, we don't have similar bug here right ?

; GFX90A-NEXT: $vgpr2_vgpr3 = V_PK_MOV_B32 8, $vgpr34_vgpr35, 12, $vgpr34_vgpr35, 0, 0, 0, 0, 0, implicit $exec, implicit $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39_vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47_vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55_vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63
; GFX90A-NEXT: $vgpr4_vgpr5 = V_PK_MOV_B32 8, $vgpr36_vgpr37, 12, $vgpr36_vgpr37, 0, 0, 0, 0, 0, implicit $exec, implicit $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39_vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47_vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55_vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63
; GFX90A-NEXT: $vgpr6_vgpr7 = V_PK_MOV_B32 8, $vgpr38_vgpr39, 12, $vgpr38_vgpr39, 0, 0, 0, 0, 0, implicit $exec, implicit $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39_vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47_vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55_vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63
; GFX90A-NEXT: $vgpr8_vgpr9 = V_PK_MOV_B32 8, $vgpr40_vgpr41, 12, $vgpr40_vgpr41, 0, 0, 0, 0, 0, implicit $exec, implicit $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39_vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47_vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55_vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63
; GFX90A-NEXT: $vgpr10_vgpr11 = V_PK_MOV_B32 8, $vgpr42_vgpr43, 12, $vgpr42_vgpr43, 0, 0, 0, 0, 0, implicit $exec, implicit $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39_vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47_vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55_vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63
; GFX90A-NEXT: $vgpr12_vgpr13 = V_PK_MOV_B32 8, $vgpr44_vgpr45, 12, $vgpr44_vgpr45, 0, 0, 0, 0, 0, implicit $exec, implicit $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39_vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47_vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55_vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63
; GFX90A-NEXT: $vgpr14_vgpr15 = V_PK_MOV_B32 8, $vgpr46_vgpr47, 12, $vgpr46_vgpr47, 0, 0, 0, 0, 0, implicit $exec, implicit $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39_vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47_vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55_vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63
; GFX90A-NEXT: $vgpr16_vgpr17 = V_PK_MOV_B32 8, $vgpr48_vgpr49, 12, $vgpr48_vgpr49, 0, 0, 0, 0, 0, implicit $exec, implicit $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39_vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47_vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55_vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63
; GFX90A-NEXT: $vgpr18_vgpr19 = V_PK_MOV_B32 8, $vgpr50_vgpr51, 12, $vgpr50_vgpr51, 0, 0, 0, 0, 0, implicit $exec, implicit $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39_vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47_vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55_vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63
; GFX90A-NEXT: $vgpr20_vgpr21 = V_PK_MOV_B32 8, $vgpr52_vgpr53, 12, $vgpr52_vgpr53, 0, 0, 0, 0, 0, implicit $exec, implicit $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39_vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47_vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55_vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63
; GFX90A-NEXT: $vgpr22_vgpr23 = V_PK_MOV_B32 8, $vgpr54_vgpr55, 12, $vgpr54_vgpr55, 0, 0, 0, 0, 0, implicit $exec, implicit $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39_vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47_vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55_vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63
; GFX90A-NEXT: $vgpr24_vgpr25 = V_PK_MOV_B32 8, $vgpr56_vgpr57, 12, $vgpr56_vgpr57, 0, 0, 0, 0, 0, implicit $exec, implicit $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39_vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47_vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55_vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63
; GFX90A-NEXT: $vgpr26_vgpr27 = V_PK_MOV_B32 8, $vgpr58_vgpr59, 12, $vgpr58_vgpr59, 0, 0, 0, 0, 0, implicit $exec, implicit $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39_vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47_vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55_vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63
; GFX90A-NEXT: $vgpr28_vgpr29 = V_PK_MOV_B32 8, $vgpr60_vgpr61, 12, $vgpr60_vgpr61, 0, 0, 0, 0, 0, implicit $exec, implicit $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39_vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47_vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55_vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63
; GFX90A-NEXT: $vgpr30_vgpr31 = V_PK_MOV_B32 8, $vgpr62_vgpr63, 12, $vgpr62_vgpr63, 0, 0, 0, 0, 0, implicit $exec, implicit killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39_vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47_vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55_vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63, implicit $exec
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 = COPY killed $vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38_vgpr39_vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47_vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54_vgpr55_vgpr56_vgpr57_vgpr58_vgpr59_vgpr60_vgpr61_vgpr62_vgpr63, implicit $exec
...
Loading