mirror of
https://github.com/ml-explore/mlx.git
synced 2025-12-16 01:49:05 +08:00
c4230747a1
* 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
68 lines
1.8 KiB
C++
68 lines
1.8 KiB
C++
// Copyright © 2023-2024 Apple Inc.
|
|
#pragma once
|
|
|
|
#include <iomanip>
|
|
#include <sstream>
|
|
#include <unordered_set>
|
|
|
|
#include "mlx/array.h"
|
|
#include "mlx/primitives.h"
|
|
|
|
namespace mlx::core {
|
|
|
|
inline bool is_static_cast(const Primitive& p) {
|
|
return (typeid(p) == typeid(Broadcast) || typeid(p) == typeid(AsType));
|
|
}
|
|
|
|
std::string build_lib_name(
|
|
const std::vector<array>& inputs,
|
|
const std::vector<array>& outputs,
|
|
const std::vector<array>& tape,
|
|
const std::unordered_set<uintptr_t>& constant_ids);
|
|
|
|
std::string get_type_string(Dtype d);
|
|
|
|
template <typename T>
|
|
void print_float_constant(std::ostream& os, const array& x) {
|
|
auto old_precision = os.precision();
|
|
os << std::setprecision(std::numeric_limits<float>::digits10 + 1)
|
|
<< x.item<T>() << std::setprecision(old_precision);
|
|
}
|
|
|
|
template <typename T>
|
|
void print_int_constant(std::ostream& os, const array& x) {
|
|
os << x.item<T>();
|
|
}
|
|
|
|
template <typename T>
|
|
void print_complex_constant(std::ostream& os, const array& x) {
|
|
auto old_precision = os.precision();
|
|
T constant = x.item<T>();
|
|
|
|
os << get_type_string(x.dtype()) << "("
|
|
<< std::setprecision(std::numeric_limits<float>::digits10 + 1)
|
|
<< constant.real() << ", " << constant.imag() << ")"
|
|
<< std::setprecision(old_precision);
|
|
}
|
|
|
|
void print_constant(std::ostream& os, const array& x);
|
|
|
|
inline bool is_scalar(const array& x) {
|
|
return x.ndim() == 0;
|
|
}
|
|
|
|
// Check if we can use a contiguous operation given inputs and the output shape
|
|
bool compiled_check_contiguity(
|
|
const std::vector<array>& inputs,
|
|
const Shape& shape);
|
|
|
|
// Allocate space for the outputs possibly with input donation
|
|
void compiled_allocate_outputs(
|
|
const std::vector<array>& inputs,
|
|
std::vector<array>& outputs,
|
|
const std::vector<array>& inputs_,
|
|
const std::unordered_set<uintptr_t>& constant_ids_,
|
|
bool contiguous);
|
|
|
|
} // namespace mlx::core
|