vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

kompute : initial attempt at ggml-backend v2 support

Browse source
This commit is contained in:
Jared Van Bortel 2024-01-09 16:24:10 -05:00
parent 1eb8804c18
commit d5670d6e46
3 changed files with 203 additions and 142 deletions
Download patch file Download diff file Expand all files Collapse all files

16
ggml-kompute.h
View file

@ -1,5 +1,7 @@
#pragma once
#include "ggml-backend.h"
#include <cstddef>
#include <vector>
#include <string>
@ -55,3 +57,17 @@ void ggml_vk_d2h_all(struct ggml_kompute_context * ctx);
void ggml_vk_h2d_tensor(struct ggml_kompute_context * ctx, struct ggml_tensor * t);
void ggml_vk_d2h_tensor(struct ggml_kompute_context * ctx, struct ggml_tensor * t);
void ggml_vk_graph_compute(struct ggml_kompute_context * ctx, struct ggml_cgraph * gf);
//
// backend API
// user-code should use only these functions
//
// forward declaration
typedef struct ggml_backend * ggml_backend_t;
GGML_API ggml_backend_t ggml_backend_kompute_init(void);
GGML_API bool ggml_backend_is_kompute(ggml_backend_t backend);
GGML_API ggml_backend_buffer_type_t ggml_backend_kompute_buffer_type(void);