add pk_fp4_t and e8m0_t support for amd_buffer_load_impl

This commit is contained in:
Feng Shijie
2025-08-20 06:40:03 +00:00
parent c27eb0771a
commit 81899bd920

View File

@@ -1308,15 +1308,20 @@ CK_TILE_DEVICE thread_buffer<T, N> amd_buffer_load_impl(int32x4_t src_wave_buffe
static_assert(
(std::is_same<T, double>::value && (N == 1 || N == 2 || N == 4 || N == 8)) ||
(std::is_same<T, float>::value && (N == 1 || N == 2 || N == 4 || N == 8 || N == 16)) ||
(std::is_same<T, fp16_t>::value && (N == 1 || N == 2 || N == 4 || N == 8)) ||
(std::is_same<T, bf16_t>::value && (N == 1 || N == 2 || N == 4 || N == 8)) ||
(std::is_same<T, fp16_t>::value &&
(N == 1 || N == 2 || N == 4 || N == 8 || N == 16 || N == 32)) ||
(std::is_same<T, bf16_t>::value &&
(N == 1 || N == 2 || N == 4 || N == 8 || N == 16 || N == 32)) ||
(std::is_same<T, int32_t>::value &&
(N == 1 || N == 2 || N == 4 || N == 8 || N == 16)) ||
(std::is_same<T, fp8_t>::value && (N == 1 || N == 2 || N == 4 || N == 8 || N == 16)) ||
(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, int8_t>::value || std::is_same<T, e8m0_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>;