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

metal : support argsort for ne00 > 1024 (#17247)

Browse Source 
* metal : refactor argsort

* cont : sort chunks

* cont : merge sorted buckets

* cont : cleanup
This commit is contained in:
Georgi Gerganov
2025-11-14 09:36:06 +02:00
committed by GitHub
parent 2606b0adab
commit 45c6ef7307
8 changed files with 266 additions and 45 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-metal/ggml-metal-impl.h
+20 -2
View File
@@ -793,10 +793,28 @@ typedef struct {
} ggml_metal_kargs_leaky_relu;
typedef struct {
int64_t ncols;
int64_t ncols_pad;
int64_t ne00;
int64_t ne01;
int64_t ne02;
int64_t ne03;
uint64_t nb00;
uint64_t nb01;
uint64_t nb02;
uint64_t nb03;
} ggml_metal_kargs_argsort;
typedef struct {
int64_t ne00;
int64_t ne01;
int64_t ne02;
int64_t ne03;
uint64_t nb00;
uint64_t nb01;
uint64_t nb02;
uint64_t nb03;
int32_t len;
} ggml_metal_kargs_argsort_merge;
typedef struct {
int64_t ne0;
float start;