Repository navigation
[AVX10.2]Optimize codegen for x86 using the new saturating conversion instructions when possible. #109080
Description
AVX10.2 introduces a set of saturating conversion instructions. These will allow us to further optimize the various floating point-> integer conversions. Specifically cases where we are currently handling edge cases with additional instructions. With these new instructions, we can accomplish this with just 1 instruction
Eg. double -> ulong
This currently generates the following 2 instructions
vfixupimmsd
vcvttsd2usi
The new Avx10.2 instructions are the following
Activity
[-][AVX10.2]Optimize codegen for x86 using saturating conversions when possible[/-][+][AVX10.2]Optimize codegen for x86 using the new saturating conversion instructions when possible.[/+]dotnet-policy-service commentedon Oct 21, 2024
Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.
JulieLeeMSFT commentedon Feb 25, 2025
@khushal1996, can we close this issue becuase @111775 is merged now?
khushal1996 commentedon Feb 26, 2025
@JulieLeeMSFT yes. we can close this now and the work related to this issue has been merged.