Simplify and optimize i8/i16 interleaving SIMD ops for x64 #567
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Replace several instructions for combining second 128-bit lane from `lo` and `hi` with a single permute. The generated code is the same because LLVM was already doing this transformation itself, but this makes it obvious in the source that the operation can be done with one instruction.
This results in some code duplication but it makes debugging easier.
Instead of interleaving the low and high 256 halves separately and stiching them together, use the AVX-512 versions of the unpacklo / unpackhi instructions to interleave from the low/high halves of all 128-bit lanes in one go, then use `_mm512_permutex2var_epi32` to combine interleaved values from the different 128b lanes.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
zip_hi_i8andzip_hi_i16for AVX2. The new code generates the same instructions as before, because LLVM was smart about optimizing the old code, but it is now simpler.zip_lo_i8andzip_hi_i8and the same for the i16-variants. This makes debugging easier.zip_*methods for AVX-512.As a side note, these changes were made with the help of OpenAI o3-mini-high model. This seems like a notable step up from o1-mini.