zhangyiss/mlx
1
0
Fork 0
mirror of https://github.com/ml-explore/mlx.git synced 2025-12-16 01:49:05 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

redesign for faster cpu/gpu synch (#1869)

Browse Source 
* redesign for faster cpu/gpu synch

* load + more async CPU

* use command encoder API and move more ops to use it

* make fence back-end generic + CPU only fence

* faster build

* fix async eval

* fixes + handle temporaries

* fix / improve cpu conv

* remove unused status, fix siblings

* fix extensions

* fix

* fix no cpu build

* format

* comments

* fix perf regression, remove unecessary abort

* fix events, task limit cpu

* fix waiting

* fix donation / temporaries in normalization
This commit is contained in:
Awni Hannun
2025-03-06 19:23:38 -08:00
committed by GitHub
parent 5245f12a46
commit c4230747a1
103 changed files with 5013 additions and 3873 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
mlx/backend/metal/scan.cpp
View File

@@ -21,7 +21,7 @@ void Scan::eval_gpu(const std::vector<array>& inputs, array& out) {
auto in = inputs[0];
if (in.flags().contiguous && in.strides()[axis_] != 0) {
if (donate && in.itemsize() == out.itemsize()) {
out.move_shared_buffer(in);
out.copy_shared_buffer(in);
} else {
out.set_data(
allocator::malloc_or_wait(in.data_size() * out.itemsize()),
@@ -33,7 +33,7 @@ void Scan::eval_gpu(const std::vector<array>& inputs, array& out) {
array arr_copy(in.shape(), in.dtype(), nullptr, {});
copy_gpu(in, arr_copy, CopyType::General, s);
in = std::move(arr_copy);
out.move_shared_buffer(in);
out.copy_shared_buffer(in);
}
bool contiguous = in.strides()[axis_] == 1;
@@ -68,8 +68,7 @@ void Scan::eval_gpu(const std::vector<array>& inputs, array& out) {
if (contiguous) {
auto& compute_encoder = d.get_command_encoder(s.index);
compute_encoder.set_compute_pipeline_state(kernel);
compute_encoder.set_input_array(
in.data_shared_ptr() == nullptr ? out : in, 0);
compute_encoder.set_input_array(in, 0);
compute_encoder.set_output_array(out, 1);
size_t size = in.shape(axis_);
compute_encoder.set_bytes(size, 2);