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

Use MTLDevice.newBufferWithBytesNoCopy to share buffers between CPU and GPU

Browse source
This commit is contained in:
Kilty McGowan 2023-06-04 21:30:54 -07:00
parent 827f5eda91
commit bc04508666
1 changed files with 14 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

17
ggml-metal.m
View file

@ -195,14 +195,25 @@ bool ggml_metal_add_buffer(
} }
} }
size_t page_size = getpagesize();
size_t aligned_size = size;
if ((aligned_size % page_size) != 0) {
aligned_size += (page_size - (aligned_size % page_size));
}
ctx->buffers[ctx->n_buffers].name = name; ctx->buffers[ctx->n_buffers].name = name;
ctx->buffers[ctx->n_buffers].data = data; ctx->buffers[ctx->n_buffers].data = data;
ctx->buffers[ctx->n_buffers].size = size; ctx->buffers[ctx->n_buffers].size = size;
ctx->buffers[ctx->n_buffers].metal = [ctx->device newBufferWithBytes:data length:size options:MTLResourceStorageModeShared]; ctx->buffers[ctx->n_buffers].metal = [ctx->device newBufferWithBytesNoCopy:data length:aligned_size options:MTLResourceStorageModeShared deallocator:nil];
if (ctx->buffers[ctx->n_buffers].metal == nil) {
fprintf(stderr, "%s: failed to allocate '%-16s' buffer, size = %8.2f MB\n", __func__, name, aligned_size / 1024.0 / 1024.0);
return false;
} else {
fprintf(stderr, "%s: allocated '%-16s' buffer, size = %8.2f MB\n", __func__, name, aligned_size / 1024.0 / 1024.0);
}
++ctx->n_buffers; ++ctx->n_buffers;
fprintf(stderr, "%s: allocated '%-16s' buffer, size = %8.2f MB\n", __func__, name, size / 1024.0 / 1024.0);
} }
return true; return true;