[SLP]Fix PR91467: Look through scalar cast, when trying to cast to another type.

Need to look through the SExt/ZExt scalars to be gathered, when trying
to reduce their width after minbitwidth analys

[SLP]Fix PR91467: Look through scalar cast, when trying to cast to another type.

Need to look through the SExt/ZExt scalars to be gathered, when trying
to reduce their width after minbitwidth analysis to prevent permanent
attempts to revectorize such gathered instructions.

show more ...

Revert "[SLP]Fix PR91467: Look through scalar cast, when trying to cast to another type."

This reverts commit 2475efa91d8b4fa8f1a2d16052cb6d14be7d5dc6.

Causes crashes, see comments on https://githu

Revert "[SLP]Fix PR91467: Look through scalar cast, when trying to cast to another type."

This reverts commit 2475efa91d8b4fa8f1a2d16052cb6d14be7d5dc6.

Causes crashes, see comments on https://github.com/llvm/llvm-project/commit/2475efa91d8b4fa8f1a2d16052cb6d14be7d5dc6.

show more ...

[SLP]Fix PR91467: Look through scalar cast, when trying to cast to another type.

Need to look through the SExt/ZExt scalars to be gathered, when trying
to reduce their width after minbitwidth analys

[SLP]Fix PR91467: Look through scalar cast, when trying to cast to another type.

Need to look through the SExt/ZExt scalars to be gathered, when trying
to reduce their width after minbitwidth analysis to prevent permanent
attempts to revectorize such gathered instructions.

show more ...

[SLP]Support minbitwidth analisys for buildvector nodes.

Metric: size..text

Program

[SLP]Support minbitwidth analisys for buildvector nodes.

Metric: size..text

Program size..text
exp ref diff
test-suite :: MultiSource/Benchmarks/mediabench/gsm/toast/toast.test 42906.00 42986.00 0.2%
test-suite :: MultiSource/Benchmarks/MiBench/telecomm-gsm/telecomm-gsm.test 42909.00 42989.00 0.2%
test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test 664581.00 664661.00 0.0%
test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test 664581.00 664661.00 0.0%

Less is better.

Replaces `buildvector <p x in> + trunc <p x in> to <p x im>` sequences to
`buildvector <p x im> of { trunc in to im }` scalars, which is free in
most cases, results in better code.

Reviewers: RKSimon

Reviewed By: RKSimon

Pull Request: https://github.com/llvm/llvm-project/pull/88504

show more ...

[SLP]Fix PR86763: do not truncate reductions to the demanded bits size.

Need to adjust ReductionBitWIdth after minbitwidth analysis, if the
demanded bits analysis sjows tht its size is less than the

[SLP]Fix PR86763: do not truncate reductions to the demanded bits size.

Need to adjust ReductionBitWIdth after minbitwidth analysis, if the
demanded bits analysis sjows tht its size is less than the size of the
vectorized value. It prevents incorrect sign-zero extension
transformation after.

show more ...

[SLP]Fix PR80027: include initial trunc nodes to the demoted values.

Need to include initial sext/zext/trunc nodes to the list of the demoted
root values to correctly calculate the cost and handle t

[SLP]Fix PR80027: include initial trunc nodes to the demoted values.

Need to include initial sext/zext/trunc nodes to the list of the demoted
root values to correctly calculate the cost and handle the
vectorization.

show more ...