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

39
mlx/backend/common/binary.h
View File

@@ -38,8 +38,7 @@ inline void set_binary_op_output_data(
const array& a,
const array& b,
array& out,
BinaryOpType bopt,
bool donate_with_move = false) {
BinaryOpType bopt) {
bool b_donatable = is_donatable(b, out);
bool a_donatable = is_donatable(a, out);
switch (bopt) {
@@ -49,11 +48,7 @@ inline void set_binary_op_output_data(
break;
case BinaryOpType::ScalarVector:
if (b_donatable) {
if (donate_with_move) {
out.move_shared_buffer(b);
} else {
out.copy_shared_buffer(b);
}
out.copy_shared_buffer(b);
} else {
out.set_data(
allocator::malloc_or_wait(b.data_size() * out.itemsize()),
@@ -64,11 +59,7 @@ inline void set_binary_op_output_data(
break;
case BinaryOpType::VectorScalar:
if (a_donatable) {
if (donate_with_move) {
out.move_shared_buffer(a);
} else {
out.copy_shared_buffer(a);
}
out.copy_shared_buffer(a);
} else {
out.set_data(
allocator::malloc_or_wait(a.data_size() * out.itemsize()),
@@ -79,17 +70,9 @@ inline void set_binary_op_output_data(
break;
case BinaryOpType::VectorVector:
if (a_donatable) {
if (donate_with_move) {
out.move_shared_buffer(a);
} else {
out.copy_shared_buffer(a);
}
out.copy_shared_buffer(a);
} else if (b_donatable) {
if (donate_with_move) {
out.move_shared_buffer(b);
} else {
out.copy_shared_buffer(b);
}
out.copy_shared_buffer(b);
} else {
out.set_data(
allocator::malloc_or_wait(a.data_size() * out.itemsize()),
@@ -100,18 +83,10 @@ inline void set_binary_op_output_data(
break;
case BinaryOpType::General:
if (a_donatable && a.flags().row_contiguous && a.size() == out.size()) {
if (donate_with_move) {
out.move_shared_buffer(a);
} else {
out.copy_shared_buffer(a);
}
out.copy_shared_buffer(a);
} else if (
b_donatable && b.flags().row_contiguous && b.size() == out.size()) {
if (donate_with_move) {
out.move_shared_buffer(b);
} else {
out.copy_shared_buffer(b);
}
out.copy_shared_buffer(b);
} else {
out.set_data(allocator::malloc_or_wait(out.nbytes()));
}