EngineX-Ascend/enginex-ascend-910-llama.cpp
10
0
Fork 0
Code Issues Pull Requests Actions 4 Projects Releases Wiki Activity

metal : Add template specialization for mul_mm_id w/ ne20 == 10 (#15799)

Browse Source 
Branch: GGMLMetalNE20

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
This commit is contained in:
Gabe Goodhart
2025-09-04 09:53:22 -06:00
committed by GitHub
parent d1e2adba65
commit 856ed0947f
2 changed files with 4 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
ggml/src/ggml-metal/ggml-metal.metal
View File

@@ -7618,6 +7618,7 @@ template [[host_name("kernel_mul_mm_id_map0_f16_ne20_2" )]] kernel kernel_mul_mm
template [[host_name("kernel_mul_mm_id_map0_f16_ne20_4" )]] kernel kernel_mul_mm_id_map0_t kernel_mul_mm_id_map0<4>;
template [[host_name("kernel_mul_mm_id_map0_f16_ne20_6" )]] kernel kernel_mul_mm_id_map0_t kernel_mul_mm_id_map0<6>;
template [[host_name("kernel_mul_mm_id_map0_f16_ne20_8" )]] kernel kernel_mul_mm_id_map0_t kernel_mul_mm_id_map0<8>;
template [[host_name("kernel_mul_mm_id_map0_f16_ne20_10")]] kernel kernel_mul_mm_id_map0_t kernel_mul_mm_id_map0<10>;
template [[host_name("kernel_mul_mm_id_map0_f16_ne20_16")]] kernel kernel_mul_mm_id_map0_t kernel_mul_mm_id_map0<16>;
template<typename T, typename T4x4, typename simdgroup_T8x8, typename block_q, short nl, void (*dequantize_func)(device const block_q *, short, thread T4x4 &)>