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

metal : improve F32, F16 and BF16 mat-vec multiplication (#16057)

Browse Source 
* metal : improve F32, F16 and BF16 mat-vec multiplication

ggml-ci

* metal : make the NSG a function constant in mul_mv kernels

ggml-ci
This commit is contained in:
Georgi Gerganov
2025-09-18 12:33:45 +03:00
committed by GitHub
parent e00f3fd8ff
commit b213fce89b
6 changed files with 355 additions and 288 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
ggml/src/ggml-metal/ggml-metal-device.m
View File

@@ -51,6 +51,10 @@ void ggml_metal_cv_free(ggml_metal_cv_t cv) {
free(cv);
}
void ggml_metal_cv_set_int16(ggml_metal_cv_t cv, int16_t value, int32_t idx) {
[cv->obj setConstantValue:&value type:MTLDataTypeShort atIndex:idx];
}
void ggml_metal_cv_set_int32(ggml_metal_cv_t cv, int32_t value, int32_t idx) {
[cv->obj setConstantValue:&value type:MTLDataTypeInt atIndex:idx];
}