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

* 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
+4
View File
@@ -197,6 +197,10 @@ static size_t ggml_backend_metal_buffer_type_get_alloc_size(ggml_backend_buffer_
res += ggml_metal_op_flash_attn_ext_extra_blk(tensor);
res += ggml_metal_op_flash_attn_ext_extra_tmp(tensor);
} break;
case GGML_OP_ARGSORT:
{
res *= 2;
} break;
default:
break;
}