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

WIP (cpu)

Browse Source
This commit is contained in:
Ronan Collobert
2025-10-31 09:40:50 -07:00
parent 5a306d3495
commit 981d2fdaf0
2 changed files with 6 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
mlx/backend/cpu/compiled.cpp
View File

@@ -49,7 +49,7 @@ static CompilerCache& cache() {
// GPU compile is always available if the GPU is available and since we are in
// this file CPU compile is also available.
namespace detail {
bool compile_available_for_device(const Device& device) {
bool compile_available_for_device(const Device& /* device */) {
return true;
}
@@ -168,7 +168,7 @@ inline void build_kernel(
// Add the input arguments
int cnt = 0;
int strides_index = 1;
for (size_t i = 0; i < inputs.size(); ++i) {
for (int i = 0; i < std::ssize(inputs); ++i) {
// Skip constants from the input list
if (is_constant(i)) {
continue;
@@ -238,7 +238,7 @@ inline void build_kernel(
} else {
os << x.primitive().name();
os << "()(";
for (int i = 0; i < x.inputs().size() - 1; i++) {
for (int i = 0; i < std::ssize(x.inputs()) - 1; i++) {
os << "tmp_" << namer.get_name(x.inputs()[i]) << ", ";
}
os << "tmp_" << namer.get_name(x.inputs().back()) << ");" << std::endl;

6
mlx/backend/cpu/inverse.cpp
View File

@@ -122,7 +122,7 @@ void inverse_impl(
stream);
const int N = a.shape(-1);
const size_t num_matrices = a.size() / (N * N);
const int64_t num_matrices = a.size() / (N * N);
auto& encoder = cpu::get_command_encoder(stream);
encoder.set_output_array(inv);
@@ -130,13 +130,13 @@ void inverse_impl(
auto inv_ptr = inv.data<T>();
if (tri) {
encoder.dispatch([inv_ptr, N, num_matrices, upper]() {
for (int i = 0; i < num_matrices; i++) {
for (int64_t i = 0; i < num_matrices; i++) {
tri_inv<T>(inv_ptr + N * N * i, N, upper);
}
});
} else {
encoder.dispatch([inv_ptr, N, num_matrices]() {
for (int i = 0; i < num_matrices; i++) {
for (int64_t i = 0; i < num_matrices; i++) {
general_inv<T>(inv_ptr + N * N * i, N);
}
});