llvm-project
[VectorCombine] Fold scalar selects from bitcast into vector select
#173990
Merged
Merged
[VectorCombine] Fold scalar selects from bitcast into vector select #173990
PankajDwivedi-25
merged
10 commits into
main
from
users/Pankajdwivedi-25/optimize-vgpr-uses-for-ternary-ops
Combine scalarized selects back into vector selects
a5260ed0
added asm test file
f7c2a62a
jmmartinez
commented
on 2025-12-30
incorporate handling of other data type casting
5a2162c0
review: transformation is profitable for other uses
73a77e81
review: avoid making seperate pass over block
966a8255
PankajDwivedi-25
changed the title
review: apply the patch to vectorcombine based on suggestion
d6feb622
add TTI cost check to check profitability
d6ed38c6
review: split nested loop & enhance pattern match
3ae7b94c
jmmartinez
approved these changes
on 2026-01-05
arsenm
commented
on 2026-01-05
review: update run check
79d4c9b1
review:root the combine at def instead of the user
adab81d9
LU-JOHN
approved these changes
on 2026-01-05
PankajDwivedi-25
deleted the users/Pankajdwivedi-25/optimize-vgpr-uses-for-ternary-ops branch 34 days ago
Reviewers
Assignees
No one assigned
Labels
backend:AMDGPU
vectorizers
llvm:transforms
llvm:vectorcombine
Milestone
No milestone
Login to write a write a comment.
Login via GitHub