[BUG] arm64 CPU backend fails JIT compile when using Float16 #3080
Description
Describe the bug
When you configure MLX (through MLX-Swift in my case) to run on a arm64 Linux machine, JIT compilation fails when it involves a float16.
To Reproduce
I have a draft PR for MLX-swift-LM that demonstrates this in a Dockerfile:
ml-explore/mlx-swift-lm#87
Expected behavior
CPU backend, although slow, successfully JIT compiles the code and runs some inference.
Desktop (please complete the following information):
FROM --platform=linux/arm64 swift:6.2.3-jammy AS base
RUN apt-get update && apt-get install -y \
libblas-dev \
liblapack-dev \
liblapacke-dev \
libopenblas-dev \
gfortran \
g++ \
&& rm -rf /var/lib/apt/lists/*
WORKDIR /app
FROM base AS builder
COPY . .
RUN swift build --product WebExample --static-swift-stdlib -Xlinker -s -v
# Final image
FROM base
# Copy executable from SwiftPM build directory
COPY --from=builder /app/.build/debug/WebExample /app/WebExample
EXPOSE 8080
CMD ["./WebExample"]Additional context
MLX/ErrorHandler.swift:343: Fatal error: [Compile::eval_cpu] Failed to compile function
Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous: Failed to execute
command with return code 1: "g++ -std=c++17 -O3 -Wall -fPIC -shared
"/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp"
-o "/tmp/mlx/0.24.2/cpu/libBf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous
.so" 2>&1", the output is: /tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188
972423790_contiguous.cpp:10:9: error: 'float16_t' has not been declared in '::'
10 | using ::float16_t;
| ^~~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:74
:14: error: declaration of 'operator+' as non-function
74 | inline float operator+(float16_t lhs, bfloat16_t rhs) { return
static_cast<float>(lhs) + static_cast<float>(rhs); } inline float operator+(bfloat16_t
lhs, float16_t rhs) { return static_cast<float>(lhs) + static_cast<float>(rhs); }
| ^~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:74
:24: error: 'float16_t' was not declared in this scope; did you mean 'bfloat16_t'?
74 | inline float operator+(float16_t lhs, bfloat16_t rhs) { return
static_cast<float>(lhs) + static_cast<float>(rhs); } inline float operator+(bfloat16_t
lhs, float16_t rhs) { return static_cast<float>(lhs) + static_cast<float>(rhs); }
| ^~~~~~~~~
| bfloat16_t
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:74
:50: error: expected primary-expression before 'rhs'
74 | inline float operator+(float16_t lhs, bfloat16_t rhs) { return
static_cast<float>(lhs) + static_cast<float>(rhs); } inline float operator+(bfloat16_t
lhs, float16_t rhs) { return static_cast<float>(lhs) + static_cast<float>(rhs); }
| ^~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:74
:156: error: 'float16_t' has not been declared
74 | inline float operator+(float16_t lhs, bfloat16_t rhs) { return
static_cast<float>(lhs) + static_cast<float>(rhs); } inline float operator+(bfloat16_t
lhs, float16_t rhs) { return static_cast<float>(lhs) + static_cast<float>(rhs); }
|
^~~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:74
:130: error: ambiguating new declaration of 'float
mlx::core::operator+(mlx::core::bfloat16_t, int)'
74 | inline float operator+(float16_t lhs, bfloat16_t rhs) { return
static_cast<float>(lhs) + static_cast<float>(rhs); } inline float operator+(bfloat16_t
lhs, float16_t rhs) { return static_cast<float>(lhs) + static_cast<float>(rhs); }
|
^~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:46
:869: note: old declaration 'mlx::core::_MLX_BFloat16
mlx::core::operator+(mlx::core::_MLX_BFloat16, int32_t)'
46 | inline _MLX_BFloat16 operator+(_MLX_BFloat16 lhs, _MLX_BFloat16 rhs) { return
static_cast<float>(lhs) + static_cast<float>(rhs); }; inline float
operator+(_MLX_BFloat16 lhs, float rhs) { return static_cast<float>(lhs) +
static_cast<float>(rhs); } inline float operator+(float lhs, _MLX_BFloat16 rhs) {
return static_cast<float>(lhs) + static_cast<float>(rhs); }; inline double
operator+(_MLX_BFloat16 lhs, double rhs) { return static_cast<double>(lhs) +
static_cast<double>(rhs); } inline double operator+(double lhs, _MLX_BFloat16 rhs) {
return static_cast<double>(lhs) + static_cast<double>(rhs); }; inline _MLX_BFloat16
operator+(_MLX_BFloat16 lhs, bool rhs) { return static_cast<float>(lhs) +
static_cast<float>(rhs); } inline _MLX_BFloat16 operator+(bool lhs, _MLX_BFloat16 rhs)
{ return static_cast<float>(lhs) + static_cast<float>(rhs); }; inline _MLX_BFloat16
operator+(_MLX_BFloat16 lhs, int32_t rhs) { return static_cast<float>(lhs) +
static_cast<float>(rhs); } inline _MLX_BFloat16 operator+(int32_t lhs, _MLX_BFloat16
rhs) { return static_cast<float>(lhs) + static_cast<float>(rhs); }; inline
_MLX_BFloat16 operator+(_MLX_BFloat16 lhs, uint32_t rhs) { return
static_cast<float>(lhs) + static_cast<float>(rhs); } inline _MLX_BFloat16
operator+(uint32_t lhs, _MLX_BFloat16 rhs) { return static_cast<float>(lhs) +
static_cast<float>(rhs); }; inline _MLX_BFloat16 operator+(_MLX_BFloat16 lhs, int64_t
rhs) { return static_cast<float>(lhs) + static_cast<float>(rhs); } inline _MLX_BFloat16
operator+(int64_t lhs, _MLX_BFloat16 rhs) { return static_cast<float>(lhs) +
static_cast<float>(rhs); }; inline _MLX_BFloat16 operator+(_MLX_BFloat16 lhs, uint64_t
rhs) { return static_cast<float>(lhs) + static_cast<float>(rhs); } inline _MLX_BFloat16
operator+(uint64_t lhs, _MLX_BFloat16 rhs) { return static_cast<float>(lhs) +
static_cast<float>(rhs); };;
|
^~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:75
:14: error: declaration of 'operator-' as non-function
75 | inline float operator-(float16_t lhs, bfloat16_t rhs) { return
static_cast<float>(lhs) - static_cast<float>(rhs); } inline float operator-(bfloat16_t
lhs, float16_t rhs) { return static_cast<float>(lhs) - static_cast<float>(rhs); }
| ^~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:75
:24: error: 'float16_t' was not declared in this scope; did you mean 'bfloat16_t'?
75 | inline float operator-(float16_t lhs, bfloat16_t rhs) { return
static_cast<float>(lhs) - static_cast<float>(rhs); } inline float operator-(bfloat16_t
lhs, float16_t rhs) { return static_cast<float>(lhs) - static_cast<float>(rhs); }
| ^~~~~~~~~
| bfloat16_t
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:75
:50: error: expected primary-expression before 'rhs'
75 | inline float operator-(float16_t lhs, bfloat16_t rhs) { return
static_cast<float>(lhs) - static_cast<float>(rhs); } inline float operator-(bfloat16_t
lhs, float16_t rhs) { return static_cast<float>(lhs) - static_cast<float>(rhs); }
| ^~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:75
:156: error: 'float16_t' has not been declared
75 | inline float operator-(float16_t lhs, bfloat16_t rhs) { return
static_cast<float>(lhs) - static_cast<float>(rhs); } inline float operator-(bfloat16_t
lhs, float16_t rhs) { return static_cast<float>(lhs) - static_cast<float>(rhs); }
|
^~~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:75
:130: error: ambiguating new declaration of 'float
mlx::core::operator-(mlx::core::bfloat16_t, int)'
75 | inline float operator-(float16_t lhs, bfloat16_t rhs) { return
static_cast<float>(lhs) - static_cast<float>(rhs); } inline float operator-(bfloat16_t
lhs, float16_t rhs) { return static_cast<float>(lhs) - static_cast<float>(rhs); }
|
^~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:47
:869: note: old declaration 'mlx::core::_MLX_BFloat16
mlx::core::operator-(mlx::core::_MLX_BFloat16, int32_t)'
47 | inline _MLX_BFloat16 operator-(_MLX_BFloat16 lhs, _MLX_BFloat16 rhs) { return
static_cast<float>(lhs) - static_cast<float>(rhs); }; inline float
operator-(_MLX_BFloat16 lhs, float rhs) { return static_cast<float>(lhs) -
static_cast<float>(rhs); } inline float operator-(float lhs, _MLX_BFloat16 rhs) {
return static_cast<float>(lhs) - static_cast<float>(rhs); }; inline double
operator-(_MLX_BFloat16 lhs, double rhs) { return static_cast<double>(lhs) -
static_cast<double>(rhs); } inline double operator-(double lhs, _MLX_BFloat16 rhs) {
return static_cast<double>(lhs) - static_cast<double>(rhs); }; inline _MLX_BFloat16
operator-(_MLX_BFloat16 lhs, bool rhs) { return static_cast<float>(lhs) -
static_cast<float>(rhs); } inline _MLX_BFloat16 operator-(bool lhs, _MLX_BFloat16 rhs)
{ return static_cast<float>(lhs) - static_cast<float>(rhs); }; inline _MLX_BFloat16
operator-(_MLX_BFloat16 lhs, int32_t rhs) { return static_cast<float>(lhs) -
static_cast<float>(rhs); } inline _MLX_BFloat16 operator-(int32_t lhs, _MLX_BFloat16
rhs) { return static_cast<float>(lhs) - static_cast<float>(rhs); }; inline
_MLX_BFloat16 operator-(_MLX_BFloat16 lhs, uint32_t rhs) { return
static_cast<float>(lhs) - static_cast<float>(rhs); } inline _MLX_BFloat16
operator-(uint32_t lhs, _MLX_BFloat16 rhs) { return static_cast<float>(lhs) -
static_cast<float>(rhs); }; inline _MLX_BFloat16 operator-(_MLX_BFloat16 lhs, int64_t
rhs) { return static_cast<float>(lhs) - static_cast<float>(rhs); } inline _MLX_BFloat16
operator-(int64_t lhs, _MLX_BFloat16 rhs) { return static_cast<float>(lhs) -
static_cast<float>(rhs); }; inline _MLX_BFloat16 operator-(_MLX_BFloat16 lhs, uint64_t
rhs) { return static_cast<float>(lhs) - static_cast<float>(rhs); } inline _MLX_BFloat16
operator-(uint64_t lhs, _MLX_BFloat16 rhs) { return static_cast<float>(lhs) -
static_cast<float>(rhs); };;
|
^~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:76
:14: error: declaration of 'operator*' as non-function
76 | inline float operator*(float16_t lhs, bfloat16_t rhs) { return
static_cast<float>(lhs) * static_cast<float>(rhs); } inline float operator*(bfloat16_t
lhs, float16_t rhs) { return static_cast<float>(lhs) * static_cast<float>(rhs); }
| ^~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:76
:24: error: 'float16_t' was not declared in this scope; did you mean 'bfloat16_t'?
76 | inline float operator*(float16_t lhs, bfloat16_t rhs) { return
static_cast<float>(lhs) * static_cast<float>(rhs); } inline float operator*(bfloat16_t
lhs, float16_t rhs) { return static_cast<float>(lhs) * static_cast<float>(rhs); }
| ^~~~~~~~~
| bfloat16_t
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:76
:50: error: expected primary-expression before 'rhs'
76 | inline float operator*(float16_t lhs, bfloat16_t rhs) { return
static_cast<float>(lhs) * static_cast<float>(rhs); } inline float operator*(bfloat16_t
lhs, float16_t rhs) { return static_cast<float>(lhs) * static_cast<float>(rhs); }
| ^~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:76
:156: error: 'float16_t' has not been declared
76 | inline float operator*(float16_t lhs, bfloat16_t rhs) { return
static_cast<float>(lhs) * static_cast<float>(rhs); } inline float operator*(bfloat16_t
lhs, float16_t rhs) { return static_cast<float>(lhs) * static_cast<float>(rhs); }
|
^~~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:76
:130: error: ambiguating new declaration of 'float
mlx::core::operator*(mlx::core::bfloat16_t, int)'
76 | inline float operator*(float16_t lhs, bfloat16_t rhs) { return
static_cast<float>(lhs) * static_cast<float>(rhs); } inline float operator*(bfloat16_t
lhs, float16_t rhs) { return static_cast<float>(lhs) * static_cast<float>(rhs); }
|
^~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:48
:869: note: old declaration 'mlx::core::_MLX_BFloat16
mlx::core::operator*(mlx::core::_MLX_BFloat16, int32_t)'
48 | inline _MLX_BFloat16 operator*(_MLX_BFloat16 lhs, _MLX_BFloat16 rhs) { return
static_cast<float>(lhs) * static_cast<float>(rhs); }; inline float
operator*(_MLX_BFloat16 lhs, float rhs) { return static_cast<float>(lhs) *
static_cast<float>(rhs); } inline float operator*(float lhs, _MLX_BFloat16 rhs) {
return static_cast<float>(lhs) * static_cast<float>(rhs); }; inline double
operator*(_MLX_BFloat16 lhs, double rhs) { return static_cast<double>(lhs) *
static_cast<double>(rhs); } inline double operator*(double lhs, _MLX_BFloat16 rhs) {
return static_cast<double>(lhs) * static_cast<double>(rhs); }; inline _MLX_BFloat16
operator*(_MLX_BFloat16 lhs, bool rhs) { return static_cast<float>(lhs) *
static_cast<float>(rhs); } inline _MLX_BFloat16 operator*(bool lhs, _MLX_BFloat16 rhs)
{ return static_cast<float>(lhs) * static_cast<float>(rhs); }; inline _MLX_BFloat16
operator*(_MLX_BFloat16 lhs, int32_t rhs) { return static_cast<float>(lhs) *
static_cast<float>(rhs); } inline _MLX_BFloat16 operator*(int32_t lhs, _MLX_BFloat16
rhs) { return static_cast<float>(lhs) * static_cast<float>(rhs); }; inline
_MLX_BFloat16 operator*(_MLX_BFloat16 lhs, uint32_t rhs) { return
static_cast<float>(lhs) * static_cast<float>(rhs); } inline _MLX_BFloat16
operator*(uint32_t lhs, _MLX_BFloat16 rhs) { return static_cast<float>(lhs) *
static_cast<float>(rhs); }; inline _MLX_BFloat16 operator*(_MLX_BFloat16 lhs, int64_t
rhs) { return static_cast<float>(lhs) * static_cast<float>(rhs); } inline _MLX_BFloat16
operator*(int64_t lhs, _MLX_BFloat16 rhs) { return static_cast<float>(lhs) *
static_cast<float>(rhs); }; inline _MLX_BFloat16 operator*(_MLX_BFloat16 lhs, uint64_t
rhs) { return static_cast<float>(lhs) * static_cast<float>(rhs); } inline _MLX_BFloat16
operator*(uint64_t lhs, _MLX_BFloat16 rhs) { return static_cast<float>(lhs) *
static_cast<float>(rhs); };;
|
^~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:77
:14: error: declaration of 'operator/' as non-function
77 | inline float operator/(float16_t lhs, bfloat16_t rhs) { return
static_cast<float>(lhs) / static_cast<float>(rhs); } inline float operator/(bfloat16_t
lhs, float16_t rhs) { return static_cast<float>(lhs) / static_cast<float>(rhs); }
| ^~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:77
:24: error: 'float16_t' was not declared in this scope; did you mean 'bfloat16_t'?
77 | inline float operator/(float16_t lhs, bfloat16_t rhs) { return
static_cast<float>(lhs) / static_cast<float>(rhs); } inline float operator/(bfloat16_t
lhs, float16_t rhs) { return static_cast<float>(lhs) / static_cast<float>(rhs); }
| ^~~~~~~~~
| bfloat16_t
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:77
:50: error: expected primary-expression before 'rhs'
77 | inline float operator/(float16_t lhs, bfloat16_t rhs) { return
static_cast<float>(lhs) / static_cast<float>(rhs); } inline float operator/(bfloat16_t
lhs, float16_t rhs) { return static_cast<float>(lhs) / static_cast<float>(rhs); }
| ^~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:77
:156: error: 'float16_t' has not been declared
77 | inline float operator/(float16_t lhs, bfloat16_t rhs) { return
static_cast<float>(lhs) / static_cast<float>(rhs); } inline float operator/(bfloat16_t
lhs, float16_t rhs) { return static_cast<float>(lhs) / static_cast<float>(rhs); }
|
^~~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:77
:130: error: ambiguating new declaration of 'float
mlx::core::operator/(mlx::core::bfloat16_t, int)'
77 | inline float operator/(float16_t lhs, bfloat16_t rhs) { return
static_cast<float>(lhs) / static_cast<float>(rhs); } inline float operator/(bfloat16_t
lhs, float16_t rhs) { return static_cast<float>(lhs) / static_cast<float>(rhs); }
|
^~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:49
:869: note: old declaration 'mlx::core::_MLX_BFloat16
mlx::core::operator/(mlx::core::_MLX_BFloat16, int32_t)'
49 | inline _MLX_BFloat16 operator/(_MLX_BFloat16 lhs, _MLX_BFloat16 rhs) { return
static_cast<float>(lhs) / static_cast<float>(rhs); }; inline float
operator/(_MLX_BFloat16 lhs, float rhs) { return static_cast<float>(lhs) /
static_cast<float>(rhs); } inline float operator/(float lhs, _MLX_BFloat16 rhs) {
return static_cast<float>(lhs) / static_cast<float>(rhs); }; inline double
operator/(_MLX_BFloat16 lhs, double rhs) { return static_cast<double>(lhs) /
static_cast<double>(rhs); } inline double operator/(double lhs, _MLX_BFloat16 rhs) {
return static_cast<double>(lhs) / static_cast<double>(rhs); }; inline _MLX_BFloat16
operator/(_MLX_BFloat16 lhs, bool rhs) { return static_cast<float>(lhs) /
static_cast<float>(rhs); } inline _MLX_BFloat16 operator/(bool lhs, _MLX_BFloat16 rhs)
{ return static_cast<float>(lhs) / static_cast<float>(rhs); }; inline _MLX_BFloat16
operator/(_MLX_BFloat16 lhs, int32_t rhs) { return static_cast<float>(lhs) /
static_cast<float>(rhs); } inline _MLX_BFloat16 operator/(int32_t lhs, _MLX_BFloat16
rhs) { return static_cast<float>(lhs) / static_cast<float>(rhs); }; inline
_MLX_BFloat16 operator/(_MLX_BFloat16 lhs, uint32_t rhs) { return
static_cast<float>(lhs) / static_cast<float>(rhs); } inline _MLX_BFloat16
operator/(uint32_t lhs, _MLX_BFloat16 rhs) { return static_cast<float>(lhs) /
static_cast<float>(rhs); }; inline _MLX_BFloat16 operator/(_MLX_BFloat16 lhs, int64_t
rhs) { return static_cast<float>(lhs) / static_cast<float>(rhs); } inline _MLX_BFloat16
operator/(int64_t lhs, _MLX_BFloat16 rhs) { return static_cast<float>(lhs) /
static_cast<float>(rhs); }; inline _MLX_BFloat16 operator/(_MLX_BFloat16 lhs, uint64_t
rhs) { return static_cast<float>(lhs) / static_cast<float>(rhs); } inline _MLX_BFloat16
operator/(uint64_t lhs, _MLX_BFloat16 rhs) { return static_cast<float>(lhs) /
static_cast<float>(rhs); };;
|
^~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:13
7:1503: error: declaration of 'operator+' as non-function
137 | inline complex64_t operator+(const std::complex<float>& x, const complex64_t&
y) { return x + static_cast<std::complex<float>>(y); } inline complex64_t
operator+(const complex64_t& x, const std::complex<float>& y) { return
static_cast<std::complex<float>>(x) + y; } inline complex64_t operator+(const
complex64_t& x, const complex64_t& y) { return static_cast<std::complex<float>>(x) +
static_cast<std::complex<float>>(y); } inline complex64_t operator+(bool x, const
complex64_t& y) { return static_cast<complex64_t>(x) + y; } inline complex64_t
operator+(const complex64_t& x, bool y) { return x + static_cast<complex64_t>(y); }
inline complex64_t operator+(uint32_t x, const complex64_t& y) { return
static_cast<complex64_t>(x) + y; } inline complex64_t operator+(const complex64_t& x,
uint32_t y) { return x + static_cast<complex64_t>(y); } inline complex64_t
operator+(uint64_t x, const complex64_t& y) { return static_cast<complex64_t>(x) + y; }
inline complex64_t operator+(const complex64_t& x, uint64_t y) { return x +
static_cast<complex64_t>(y); } inline complex64_t operator+(int32_t x, const
complex64_t& y) { return static_cast<complex64_t>(x) + y; } inline complex64_t
operator+(const complex64_t& x, int32_t y) { return x + static_cast<complex64_t>(y); }
inline complex64_t operator+(int64_t x, const complex64_t& y) { return
static_cast<complex64_t>(x) + y; } inline complex64_t operator+(const complex64_t& x,
int64_t y) { return x + static_cast<complex64_t>(y); } inline complex64_t
operator+(float16_t x, const complex64_t& y) { return static_cast<complex64_t>(x) + y;
} inline complex64_t operator+(const complex64_t& x, float16_t y) { return x +
static_cast<complex64_t>(y); } inline complex64_t operator+(bfloat16_t x, const
complex64_t& y) { return static_cast<complex64_t>(x) + y; } inline complex64_t
operator+(const complex64_t& x, bfloat16_t y) { return x + static_cast<complex64_t>(y);
} inline complex64_t operator+(float x, const complex64_t& y) { return
static_cast<complex64_t>(x) + y; } inline complex64_t operator+(const complex64_t& x,
float y) { return x + static_cast<complex64_t>(y); }
|
^~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:13
7:1513: error: 'float16_t' was not declared in this scope; did you mean 'bfloat16_t'?
137 | inline complex64_t operator+(const std::complex<float>& x, const complex64_t&
y) { return x + static_cast<std::complex<float>>(y); } inline complex64_t
operator+(const complex64_t& x, const std::complex<float>& y) { return
static_cast<std::complex<float>>(x) + y; } inline complex64_t operator+(const
complex64_t& x, const complex64_t& y) { return static_cast<std::complex<float>>(x) +
static_cast<std::complex<float>>(y); } inline complex64_t operator+(bool x, const
complex64_t& y) { return static_cast<complex64_t>(x) + y; } inline complex64_t
operator+(const complex64_t& x, bool y) { return x + static_cast<complex64_t>(y); }
inline complex64_t operator+(uint32_t x, const complex64_t& y) { return
static_cast<complex64_t>(x) + y; } inline complex64_t operator+(const complex64_t& x,
uint32_t y) { return x + static_cast<complex64_t>(y); } inline complex64_t
operator+(uint64_t x, const complex64_t& y) { return static_cast<complex64_t>(x) + y; }
inline complex64_t operator+(const complex64_t& x, uint64_t y) { return x +
static_cast<complex64_t>(y); } inline complex64_t operator+(int32_t x, const
complex64_t& y) { return static_cast<complex64_t>(x) + y; } inline complex64_t
operator+(const complex64_t& x, int32_t y) { return x + static_cast<complex64_t>(y); }
inline complex64_t operator+(int64_t x, const complex64_t& y) { return
static_cast<complex64_t>(x) + y; } inline complex64_t operator+(const complex64_t& x,
int64_t y) { return x + static_cast<complex64_t>(y); } inline complex64_t
operator+(float16_t x, const complex64_t& y) { return static_cast<complex64_t>(x) + y;
} inline complex64_t operator+(const complex64_t& x, float16_t y) { return x +
static_cast<complex64_t>(y); } inline complex64_t operator+(bfloat16_t x, const
complex64_t& y) { return static_cast<complex64_t>(x) + y; } inline complex64_t
operator+(const complex64_t& x, bfloat16_t y) { return x + static_cast<complex64_t>(y);
} inline complex64_t operator+(float x, const complex64_t& y) { return
static_cast<complex64_t>(x) + y; } inline complex64_t operator+(const complex64_t& x,
float y) { return x + static_cast<complex64_t>(y); }
|
^~~~~~~~~
|
bfloat16_t
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:13
7:1526: error: expected primary-expression before 'const'
137 | inline complex64_t operator+(const std::complex<float>& x, const complex64_t&
y) { return x + static_cast<std::complex<float>>(y); } inline complex64_t
operator+(const complex64_t& x, const std::complex<float>& y) { return
static_cast<std::complex<float>>(x) + y; } inline complex64_t operator+(const
complex64_t& x, const complex64_t& y) { return static_cast<std::complex<float>>(x) +
static_cast<std::complex<float>>(y); } inline complex64_t operator+(bool x, const
complex64_t& y) { return static_cast<complex64_t>(x) + y; } inline complex64_t
operator+(const complex64_t& x, bool y) { return x + static_cast<complex64_t>(y); }
inline complex64_t operator+(uint32_t x, const complex64_t& y) { return
static_cast<complex64_t>(x) + y; } inline complex64_t operator+(const complex64_t& x,
uint32_t y) { return x + static_cast<complex64_t>(y); } inline complex64_t
operator+(uint64_t x, const complex64_t& y) { return static_cast<complex64_t>(x) + y; }
inline complex64_t operator+(const complex64_t& x, uint64_t y) { return x +
static_cast<complex64_t>(y); } inline complex64_t operator+(int32_t x, const
complex64_t& y) { return static_cast<complex64_t>(x) + y; } inline complex64_t
operator+(const complex64_t& x, int32_t y) { return x + static_cast<complex64_t>(y); }
inline complex64_t operator+(int64_t x, const complex64_t& y) { return
static_cast<complex64_t>(x) + y; } inline complex64_t operator+(const complex64_t& x,
int64_t y) { return x + static_cast<complex64_t>(y); } inline complex64_t
operator+(float16_t x, const complex64_t& y) { return static_cast<complex64_t>(x) + y;
} inline complex64_t operator+(const complex64_t& x, float16_t y) { return x +
static_cast<complex64_t>(y); } inline complex64_t operator+(bfloat16_t x, const
complex64_t& y) { return static_cast<complex64_t>(x) + y; } inline complex64_t
operator+(const complex64_t& x, bfloat16_t y) { return x + static_cast<complex64_t>(y);
} inline complex64_t operator+(float x, const complex64_t& y) { return
static_cast<complex64_t>(x) + y; } inline complex64_t operator+(const complex64_t& x,
float y) { return x + static_cast<complex64_t>(y); }
|
^~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:13
7:1643: error: 'float16_t' has not been declared
137 | inline complex64_t operator+(const std::complex<float>& x, const complex64_t&
y) { return x + static_cast<std::complex<float>>(y); } inline complex64_t
operator+(const complex64_t& x, const std::complex<float>& y) { return
static_cast<std::complex<float>>(x) + y; } inline complex64_t operator+(const
complex64_t& x, const complex64_t& y) { return static_cast<std::complex<float>>(x) +
static_cast<std::complex<float>>(y); } inline complex64_t operator+(bool x, const
complex64_t& y) { return static_cast<complex64_t>(x) + y; } inline complex64_t
operator+(const complex64_t& x, bool y) { return x + static_cast<complex64_t>(y); }
inline complex64_t operator+(uint32_t x, const complex64_t& y) { return
static_cast<complex64_t>(x) + y; } inline complex64_t operator+(const complex64_t& x,
uint32_t y) { return x + static_cast<complex64_t>(y); } inline complex64_t
operator+(uint64_t x, const complex64_t& y) { return static_cast<complex64_t>(x) + y; }
inline complex64_t operator+(const complex64_t& x, uint64_t y) { return x +
static_cast<complex64_t>(y); } inline complex64_t operator+(int32_t x, const
complex64_t& y) { return static_cast<complex64_t>(x) + y; } inline complex64_t
operator+(const complex64_t& x, int32_t y) { return x + static_cast<complex64_t>(y); }
inline complex64_t operator+(int64_t x, const complex64_t& y) { return
static_cast<complex64_t>(x) + y; } inline complex64_t operator+(const complex64_t& x,
int64_t y) { return x + static_cast<complex64_t>(y); } inline complex64_t
operator+(float16_t x, const complex64_t& y) { return static_cast<complex64_t>(x) + y;
} inline complex64_t operator+(const complex64_t& x, float16_t y) { return x +
static_cast<complex64_t>(y); } inline complex64_t operator+(bfloat16_t x, const
complex64_t& y) { return static_cast<complex64_t>(x) + y; } inline complex64_t
operator+(const complex64_t& x, bfloat16_t y) { return x + static_cast<complex64_t>(y);
} inline complex64_t operator+(float x, const complex64_t& y) { return
static_cast<complex64_t>(x) + y; } inline complex64_t operator+(const complex64_t& x,
float y) { return x + static_cast<complex64_t>(y); }
|
^~~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:13
7:1611: error: redefinition of 'mlx::core::complex64_t mlx::core::operator+(const
mlx::core::complex64_t&, int)'
137 | inline complex64_t operator+(const std::complex<float>& x, const complex64_t&
y) { return x + static_cast<std::complex<float>>(y); } inline complex64_t
operator+(const complex64_t& x, const std::complex<float>& y) { return
static_cast<std::complex<float>>(x) + y; } inline complex64_t operator+(const
complex64_t& x, const complex64_t& y) { return static_cast<std::complex<float>>(x) +
static_cast<std::complex<float>>(y); } inline complex64_t operator+(bool x, const
complex64_t& y) { return static_cast<complex64_t>(x) + y; } inline complex64_t
operator+(const complex64_t& x, bool y) { return x + static_cast<complex64_t>(y); }
inline complex64_t operator+(uint32_t x, const complex64_t& y) { return
static_cast<complex64_t>(x) + y; } inline complex64_t operator+(const complex64_t& x,
uint32_t y) { return x + static_cast<complex64_t>(y); } inline complex64_t
operator+(uint64_t x, const complex64_t& y) { return static_cast<complex64_t>(x) + y; }
inline complex64_t operator+(const complex64_t& x, uint64_t y) { return x +
static_cast<complex64_t>(y); } inline complex64_t operator+(int32_t x, const
complex64_t& y) { return static_cast<complex64_t>(x) + y; } inline complex64_t
operator+(const complex64_t& x, int32_t y) { return x + static_cast<complex64_t>(y); }
inline complex64_t operator+(int64_t x, const complex64_t& y) { return
static_cast<complex64_t>(x) + y; } inline complex64_t operator+(const complex64_t& x,
int64_t y) { return x + static_cast<complex64_t>(y); } inline complex64_t
operator+(float16_t x, const complex64_t& y) { return static_cast<complex64_t>(x) + y;
} inline complex64_t operator+(const complex64_t& x, float16_t y) { return x +
static_cast<complex64_t>(y); } inline complex64_t operator+(bfloat16_t x, const
complex64_t& y) { return static_cast<complex64_t>(x) + y; } inline complex64_t
operator+(const complex64_t& x, bfloat16_t y) { return x + static_cast<complex64_t>(y);
} inline complex64_t operator+(float x, const complex64_t& y) { return
static_cast<complex64_t>(x) + y; } inline complex64_t operator+(const complex64_t& x,
float y) { return x + static_cast<complex64_t>(y); }
|
^~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:13
7:1185: note: 'mlx::core::complex64_t mlx::core::operator+(const
mlx::core::complex64_t&, int32_t)' previously defined here
137 | inline complex64_t operator+(const std::complex<float>& x, const complex64_t&
y) { return x + static_cast<std::complex<float>>(y); } inline complex64_t
operator+(const complex64_t& x, const std::complex<float>& y) { return
static_cast<std::complex<float>>(x) + y; } inline complex64_t operator+(const
complex64_t& x, const complex64_t& y) { return static_cast<std::complex<float>>(x) +
static_cast<std::complex<float>>(y); } inline complex64_t operator+(bool x, const
complex64_t& y) { return static_cast<complex64_t>(x) + y; } inline complex64_t
operator+(const complex64_t& x, bool y) { return x + static_cast<complex64_t>(y); }
inline complex64_t operator+(uint32_t x, const complex64_t& y) { return
static_cast<complex64_t>(x) + y; } inline complex64_t operator+(const complex64_t& x,
uint32_t y) { return x + static_cast<complex64_t>(y); } inline complex64_t
operator+(uint64_t x, const complex64_t& y) { return static_cast<complex64_t>(x) + y; }
inline complex64_t operator+(const complex64_t& x, uint64_t y) { return x +
static_cast<complex64_t>(y); } inline complex64_t operator+(int32_t x, const
complex64_t& y) { return static_cast<complex64_t>(x) + y; } inline complex64_t
operator+(const complex64_t& x, int32_t y) { return x + static_cast<complex64_t>(y); }
inline complex64_t operator+(int64_t x, const complex64_t& y) { return
static_cast<complex64_t>(x) + y; } inline complex64_t operator+(const complex64_t& x,
int64_t y) { return x + static_cast<complex64_t>(y); } inline complex64_t
operator+(float16_t x, const complex64_t& y) { return static_cast<complex64_t>(x) + y;
} inline complex64_t operator+(const complex64_t& x, float16_t y) { return x +
static_cast<complex64_t>(y); } inline complex64_t operator+(bfloat16_t x, const
complex64_t& y) { return static_cast<complex64_t>(x) + y; } inline complex64_t
operator+(const complex64_t& x, bfloat16_t y) { return x + static_cast<complex64_t>(y);
} inline complex64_t operator+(float x, const complex64_t& y) { return
static_cast<complex64_t>(x) + y; } inline complex64_t operator+(const complex64_t& x,
float y) { return x + static_cast<complex64_t>(y); }
|
^~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:
In function 'mlx::core::simd::Simd<T, 1>
mlx::core::simd::clamp(mlx::core::simd::Simd<T, 1>, mlx::core::simd::Simd<T, 1>,
mlx::core::simd::Simd<T, 1>)':
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:34
2:15: error: 'clamp' is not a member of 'std'; did you mean 'mlx::core::simd::clamp'?
342 | return std::clamp(v.value, min.value, max.value);
| ^~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:34
1:12: note: 'mlx::core::simd::clamp' declared here
341 | Simd<T, 1> clamp(Simd<T, 1> v, Simd<T, 1> min, Simd<T, 1> max) {
| ^~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:
In instantiation of 'mlx::core::simd::Simd<float, N>
mlx::core::detail::fp32_from_bits(mlx::core::simd::Simd<unsigned int, N>) [with int N =
1]':
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:62
0:26: required from 'mlx::core::simd::Simd<float, N>
mlx::core::detail::FromFP8::operator()(mlx::core::simd::Simd<unsigned char, N>) [with
int N = 1]'
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:62
3:19: required from here
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:56
8:11: warning: dereferencing type-punned pointer will break strict-aliasing rules
[-Wstrict-aliasing]
568 | return *(Simd<float, N>*)(&x);
| ^~~~~~~~~~~~~~~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:56
8:11: warning: dereferencing type-punned pointer will break strict-aliasing rules
[-Wstrict-aliasing]
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:
In instantiation of 'mlx::core::simd::Simd<T, N>
mlx::core::simd::exp(mlx::core::simd::Simd<T, N>) [with T = float; int N = 1]':
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:53
8:38: required from 'mlx::core::simd::Simd<T, N>
mlx::core::detail::Sigmoid::operator()(mlx::core::simd::Simd<T, N>) [with int N = 1; T
= float]'
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:54
1:59: required from 'T mlx::core::detail::Sigmoid::operator()(T) [with T = float]'
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:69
5:26: required from here
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:37
6:51: warning: dereferencing type-punned pointer will break strict-aliasing rules
[-Wstrict-aliasing]
376 | auto result = select(isnan(x_init), x_init, (*(Simd<float, N>*)&epart) *
x);
| ^~~~~~~~~~~~~~~~~~~~~~~
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:37
6:51: warning: dereferencing type-punned pointer will break strict-aliasing rules
[-Wstrict-aliasing]
/tmp/mlx/0.24.2/cpu/Bf4ISigmoidACf4OMultiplyAB_V__6142509188972423790_contiguous.cpp:25
6:65: note: initializing argument 1 of 'mlx::core::simd::Simd<decltype ((a.value *
b.value)), 1> mlx::core::simd::operator*(mlx::core::simd::Simd<T, 1>,
mlx::core::simd::Simd<T2, 1>) [with T1 = float; T2 = float; decltype ((a.value *
b.value)) = float]'
256 | template <typename T1, typename T2> auto operator *(Simd<T1, 1> a, Simd<T2, 1>
b) ->Simd<decltype(a.value * b.value), 1> { return a.value * b.value; } template
<typename T1, typename T2> auto operator *(T1 a, Simd<T2, 1> b)->Simd<decltype(a *
b.value), 1> { return a * b.value; } template <typename T1, typename T2> auto operator
*(Simd<T1, 1> a, T2 b)->Simd<decltype(a.value * b), 1> { return a.value * b; }
| ~~~~~~~~~~~~^ at
/app/.build/checkouts/mlx-swift/Source/Cmlx/mlx-c/mlx/c/transforms.cpp:15