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

metal : use shared buffers between CPU and GPU (#1696)

Browse source 
* Use MTLDevice.newBufferWithBytesNoCopy to share buffers between CPU and GPU

* Page-align buffers used by Metal

* Remove trailing whitespace

* Only import unistd.h for Metal builds

* metal : remove unnecessary copies

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
kiltyj 2023-06-05 13:24:04 -07:00 • committed by GitHub
parent efe0507632
commit 9d0693bce3
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
4 changed files with 38 additions and 16 deletions
Download patch file Download diff file Expand all files Collapse all files

13
llama.cpp
View file

@ -53,7 +53,6 @@ enum e_model {
MODEL_65B,
};
static const size_t MB = 1024*1024;
// computed for n_ctx == 2048
@ -1281,12 +1280,6 @@ static bool llama_eval_internal(
ggml_set_name(embd, "embd");
memcpy(embd->data, tokens, N*ggml_element_size(embd));
#ifdef GGML_USE_METAL
if (lctx.ctx_metal && N == 1) {
ggml_metal_set_tensor(lctx.ctx_metal, embd);
}
#endif
struct ggml_tensor * cur;
struct ggml_tensor * inpL = ggml_get_rows(ctx0, model.tok_embeddings, embd);
@ -1484,12 +1477,6 @@ static bool llama_eval_internal(
}
ggml_graph_compute(ctx0, &gf);
if (lctx.ctx_metal) {
// We need to sync the CPU KV cache with the GPU KV cache
ggml_metal_set_tensor(lctx.ctx_metal, kv_self.k);
ggml_metal_set_tensor(lctx.ctx_metal, kv_self.v);
}
}
#else
ggml_graph_compute(ctx0, &gf);