Skip to content

Conversation

@ravil-mobile
Copy link
Contributor

GFX1250 arch comes with new FP conversion instruction which can convert 8x FP32/FP16/Bf16 to 8x FP8. This PR extends to the AMDGPU backend with the support of the new instructions

};

template <typename InType, typename OutType>
void clampInfInInput(Location loc, ConversionPatternRewriter &rewriter,
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We can drop this part for now.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Done

@antiagainst antiagainst marked this pull request as ready for review November 25, 2025 17:17
@antiagainst antiagainst changed the title [WIP][AMD] Extended FP conversion for gfx1250 arch [AMD] Extended FP conversion for gfx1250 arch Nov 25, 2025
@antiagainst antiagainst merged commit 6feef10 into triton-lang:main Nov 25, 2025
9 checks passed
@ravil-mobile ravil-mobile deleted the ravil/fp-conversion branch November 25, 2025 17:40
tmoreau89 pushed a commit to tmoreau89/triton that referenced this pull request Dec 1, 2025
GFX1250 arch comes with new FP conversion instruction which can convert
8x FP32/FP16/Bf16 to 8x FP8. This PR extends to the AMDGPU backend with
the support of the new instructions
CRobeck pushed a commit to plotfi/triton that referenced this pull request Dec 2, 2025
GFX1250 arch comes with new FP conversion instruction which can convert
8x FP32/FP16/Bf16 to 8x FP8. This PR extends to the AMDGPU backend with
the support of the new instructions
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants