add mixed_prec fp16xfp4

This commit is contained in:
Feng Shijie
2025-08-08 20:19:16 +00:00
parent 3dea10a277
commit f788d3d629
9 changed files with 252 additions and 123 deletions

View File

@@ -1450,7 +1450,9 @@ CK_TILE_DEVICE thread_buffer<T, N> amd_buffer_load_impl(int32x4_t src_wave_buffe
(std::is_same<T, bf8_t>::value && (N == 1 || N == 2 || N == 4 || N == 8 || N == 16)) ||
(std::is_same<T, int8_t>::value && (N == 1 || N == 2 || N == 4 || N == 8 || N == 16)) ||
(std::is_same<T, pk_int4_t>::value &&
(N == 1 || N == 2 || N == 4 || N == 8 || N == 16 || N == 32)),
(N == 1 || N == 2 || N == 4 || N == 8 || N == 16 || N == 32) ||
(std::is_same<T, pk_fp4_t>::value &&
(N == 1 || N == 2 || N == 4 || N == 8 || N == 16))),
"wrong! not implemented");
using rtn_type = thread_buffer<T, N>;