Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

metal : add top-k radix implementation (#28073)

Browse Source 
Assisted-by: DeepSeek-v4-Flash-0731
This commit is contained in:
Georgi Gerganov
2026-08-31 21:31:53 +03:00
committed by GitHub
parent 2d8d612e4c
commit 2a74817f93
5 changed files with 206 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-metal/ggml-metal-device.cpp
+18 -1
View File
@@ -1336,7 +1336,7 @@ ggml_metal_pipeline_with_params ggml_metal_library_get_pipeline_fwht(ggml_metal_
return res;
}
// note: reuse the argsort kernel for top_k
// note: reuse the argsort kernel for the bitonic top_k fallback
ggml_metal_pipeline_with_params ggml_metal_library_get_pipeline_top_k(ggml_metal_library_t lib, const ggml_tensor * op) {
assert(op->op == GGML_OP_TOP_K);
@@ -1364,6 +1364,23 @@ ggml_metal_pipeline_with_params ggml_metal_library_get_pipeline_top_k(ggml_metal
return res;
}
ggml_metal_pipeline_with_params ggml_metal_library_get_pipeline_top_k_radix(ggml_metal_library_t lib, const ggml_tensor * op) {
assert(op->op == GGML_OP_TOP_K);
char base[256];
char name[256];
snprintf(base, 256, "kernel_top_k_%s_%s", ggml_type_name(op->src[0]->type), ggml_type_name(op->type));
snprintf(name, 256, "%s", base);
ggml_metal_pipeline_with_params res = ggml_metal_library_get_pipeline(lib, name);
if (!res.pipeline) {
res = ggml_metal_library_compile_pipeline(lib, base, name, nullptr);
}
return res;
}
ggml_metal_pipeline_with_params ggml_metal_library_get_pipeline_top_k_merge(ggml_metal_library_t lib, const ggml_tensor * op) {
assert(op->op == GGML_OP_TOP_K);