AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

SWDEV-446047 - use gfx names

Browse Source 
Change-Id: I43aa93362f461bb3fb912a48cf043cfdff56209c


[ROCm/hip commit: da7313e100]
This commit is contained in:
Jatin Chaudhary
2024-08-23 16:05:33 +01:00
parent 78c9afb098
commit 80cac32a91
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/hip/docs/reference/fp8_numbers.rst
+2 -2
View File
@@ -81,7 +81,7 @@ The same can be done in kernels as well.
return __hip_cvt_float_to_fp8(in, sat, interpret);
}
An important thing to note here is if you use this on MI300 GPU, it will be fnuz number but on any other GPU it will be an OCP number.
An important thing to note here is if you use this on gfx94x GPU, it will be fnuz number but on any other GPU it will be an OCP number.
The following code example does roundtrip FP8 conversions on both the CPU and GPU and compares the results.
@@ -227,4 +227,4 @@ There is support of vector of FP8 types.
- __hip_fp8x2_e5m2_fnuz: holds 2 values of FP8 fnuz e5m2 numbers
- __hip_fp8x4_e5m2_fnuz: holds 4 values of FP8 fnuz e5m2 numbers
FNUZ extensions will be available on MI300 only.
FNUZ extensions will be available on gfx94x only.