llvm-project
[AMDGPU] Waterfall loop codegen improvement in SIInstrInfo
#192415
Merged

[AMDGPU] Waterfall loop codegen improvement in SIInstrInfo #192415

gretay-amd
github-actions
gretay-amd gretay-amd marked this pull request as ready for review 82 days ago
llvmorg-github-actions llvmorg-github-actions added backend:AMDGPU
llvmorg-github-actions
jayfoad
jayfoad commented on 2026-04-30
arsenm
arsenm commented on 2026-04-30
arsenm
arsenm commented on 2026-04-30
jayfoad
gretay-amd gretay-amd force pushed from 1ff946c7 to 4a82036b 82 days ago
arsenm
arsenm commented on 2026-04-30
gretay-amd
gretay-amd commented on 2026-05-11
arsenm
gretay-amd gretay-amd force pushed from a19da49e to 1620a05d 62 days ago
github-actions
gretay-amd
gretay-amd [AMDGPU] Waterfall loop codegen improvement
30534954
gretay-amd [AMDGPU] NFC: Rename AndN2WRExecOpc to AndN2WrExecOpc and add availab…
acda1ad8
gretay-amd [AMDGPU] NFC: Move createVirtualRegister calls inside the if block
ad93d423
gretay-amd [AMDGPU] Always use VScalarOpUndef flag and factor out condition
36edaf4e
gretay-amd [AMDGPU] Add _term variants of v_cmpx instructions
6362f068
gretay-amd [AMDGPU] Use v_cmpx _term in waterfall loops
772e4ba8
gretay-amd Revert "[AMDGPU] Use v_cmpx _term in waterfall loops"
b76e28d4
gretay-amd [AMDGPU] Defer v_cmpx_term after all v_readfirstlane
f24e1f8f
gretay-amd [AMDGPU] NFC: Add comment about latency vs register pressure trade-off
d134f9a4
gretay-amd [AMDGPU] NFC: Add comment about why saveexec/wrexec cannot use _term
59e1c50d
gretay-amd Format
7feb5780
gretay-amd gretay-amd force pushed from fd6b17c4 to 7feb5780 61 days ago
arsenm
gretay-amd
gretay-amd
gretay-amd
jayfoad
jayfoad commented on 2026-06-04
jayfoad
gretay-amd Bug fix: cmpx does not depend on wave size.
e0068137
gretay-amd Cleanup: remove unused _term definitions introduced early in this PR
cc3bbcd4
gretay-amd
gretay-amd Format
78c021fb
jayfoad
jayfoad
jayfoad commented on 2026-06-05
gretay-amd
jayfoad
jayfoad commented on 2026-06-19
gretay-amd [AMDGPU] Accurately detect s_andn2_wrexec availability in waterfall loop
7bd56f70
gretay-amd [AMDGPU] Fix pre-existing formatting: add missing braces on the "if" …
f31629df
gretay-amd
gretay-amd commented on 2026-06-19
jayfoad
jayfoad approved these changes on 2026-06-22
gretay-amd gretay-amd merged 4661c024 into main 29 days ago
github-actions
llvm-ci

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone