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

6
mlx/backend/metal/scaled_dot_product_attention.cpp
View File

@@ -152,7 +152,7 @@ void sdpa_vector(
compute_encoder.set_compute_pipeline_state(kernel);
// Set its arguments
compute_encoder.set_input_array(q.data_shared_ptr() == nullptr ? out : q, 0);
compute_encoder.set_input_array(q, 0);
compute_encoder.set_input_array(k, 1);
compute_encoder.set_input_array(v, 2);
compute_encoder.set_output_array(out, 3);
@@ -242,7 +242,7 @@ void sdpa_vector_2pass(
compute_encoder.set_compute_pipeline_state(kernel);
// Set its arguments
compute_encoder.set_input_array(q.data_shared_ptr() == nullptr ? out : q, 0);
compute_encoder.set_input_array(q, 0);
compute_encoder.set_input_array(k, 1);
compute_encoder.set_input_array(v, 2);
compute_encoder.set_output_array(intermediate, 3);
@@ -357,7 +357,7 @@ void ScaledDotProductAttention::eval_gpu(
// Donate the query if possible
if (q.is_donatable() && (q.shape(2) == 1 || !q.flags().row_contiguous) &&
q.size() == o.size()) {
o.move_shared_buffer(q);
o.copy_shared_buffer(q);
} else {
if (o.shape(2) == 1) {
o.set_data(allocator::malloc_or_wait(o.nbytes()));