ver4a/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

ggml : upgrade init_tensor API to return a ggml_status (#11854)

Browse source 
* Upgrade init_tensor API to return a ggml_status

To prepare for an 'abort-free' ggml
(ggml not to abort on OOMs but return a OOM status),
as agreeed with Diego in the ggml repo,
upgrade the init_tensor() and view_init() APIs
to return a ggml_status.

* misc fixes

---------

Co-authored-by: slaren <slarengh@gmail.com>
This commit is contained in:
William Tambellini 2025-02-28 05:41:47 -08:00 • committed by GitHub
parent c43a3e7996
commit 70680c48e5
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
16 changed files with 136 additions and 97 deletions
Download patch file Download diff file Expand all files Collapse all files

8
ggml/src/ggml-sycl/ggml-sycl.cpp
View file

@ -323,14 +323,14 @@ static void * ggml_backend_sycl_buffer_get_base(ggml_backend_buffer_t buffer) {
return ctx->dev_ptr;
}
static void
static enum ggml_status
ggml_backend_sycl_buffer_init_tensor(ggml_backend_buffer_t buffer,
ggml_tensor *tensor) try {
ggml_backend_sycl_buffer_context * ctx = (ggml_backend_sycl_buffer_context *)buffer->context;
if (tensor->view_src != NULL) {
assert(tensor->view_src->buffer->buft == buffer->buft);
return;
return GGML_STATUS_SUCCESS;
}
ggml_tensor_extra_gpu * extra = new ggml_tensor_extra_gpu{};
@ -348,6 +348,7 @@ ggml_backend_sycl_buffer_init_tensor(ggml_backend_buffer_t buffer,
padded_size - original_size).wait()));
}
}
return GGML_STATUS_SUCCESS;
}
catch (sycl::exception const &exc) {
std::cerr << exc.what() << "Exception caught at file:" << __FILE__
@ -729,7 +730,7 @@ static void * ggml_backend_sycl_split_buffer_get_base(ggml_backend_buffer_t buff
GGML_UNUSED(buffer);
}
static void
static enum ggml_status
ggml_backend_sycl_split_buffer_init_tensor(ggml_backend_buffer_t buffer,
ggml_tensor *tensor) try {
GGML_ASSERT(tensor->view_src == nullptr); // views of split tensors are not supported
@ -804,6 +805,7 @@ ggml_backend_sycl_split_buffer_init_tensor(ggml_backend_buffer_t buffer,
}
}
tensor->extra = extra;
return GGML_STATUS_SUCCESS;
}
catch (sycl::exception const &exc) {
std::cerr << exc.what() << "Exception caught at file:" << __FILE__