ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-24 23:05:54 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge commit 'ec006bb8e008caaf5cb95c1ca5037dc6ac026bde' into develop

Browse Source
This commit is contained in:
assistant-librarian[bot]
2025-09-10 03:20:14 +00:00
parent b2b0389f76
commit 535686e647
27 changed files with 2429 additions and 865 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
include/ck_tile/core/numeric/float8.hpp
View File

@@ -399,9 +399,9 @@ CK_TILE_HOST_DEVICE DstT run_cast_to_f8(SrcT src, unsigned int rng = 0)
}
mantissa += (1u << SrcT_mant); // Add the implicit 1 into mantissa
}
// The value is smaller than min f8 denormal and results in zero (the early exit also prevents
// The value is <= than min f8 denormal/2 and results in zero (the early exit also prevents
// an undefined behavior of bit shifts >= type width).
if(exponent_diff > DstT_mant)
if(exponent_diff > DstT_mant + 1)
{
return is_fnuz ? 0 : (sign << (DstT_exp + DstT_mant));
}