FP16 causes cross-compilation errors for x86-64 targets on macOS.
Tinuv-Dev opened this issue · 4 comments
Tinuv-Dev commented
I attempted to cross-compile XNNPACK today using the latest code from the master branch, but the compilation failed with an error. I am quite puzzled by this because I was able to successfully cross-compile XNNPACK not long ago.
After analyzing the error, I found that it is likely caused by the use of FP16. It seems that a recent commit (5443c56, commit message: “Add a typedef for float16 to uint16”) added #include <fp16/fp16.h> to libXNNPACK-source-master/src/xnnpack/math.h. This inclusion causes a compilation error when cross-compiling for the x86-64 target (including macOS x86-64 physical devices and iOS x86-64 simulators). It appears that FP16 is a library used for optimization. Could you add an option to the CMakeLists.txt file that allows me to disable FP16 optimization? Alternatively, if there is another way to do this, I would appreciate any guidance that would allow me to successfully cross-compile.
Additional Information:
• macOS Version: macOS Sequoia 15.0 beta 7
• Xcode Version: Version 16.0 beta 6 (16A5230g)
Log Information:
[ 0%] Built target pthreadpool
[ 0%] Built target cpuinfo_internals
[ 1%] Built target cpuinfo
[ 1%] Built target microparams-init
[ 1%] Built target normalization
[ 1%] Built target indirection
[ 1%] Built target logging
[ 1%] Built target microkernel-utils
[ 1%] Built target allocator
[ 1%] Built target hardware-config
[ 1%] Built target cache
[ 1%] Built target mutex
[ 1%] Built target memory
[ 1%] Built target packing
[ 1%] Built target operator-utils
[ 1%] Built target operator-run
[ 1%] Built target operators
[ 1%] Building C object CMakeFiles/subgraph.dir/src/runtime.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/memory-planner.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/argmax-pooling-2d.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/abs.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/add2.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/average-pooling-2d.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/bankers-rounding.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/batch-matrix-multiply.c.o
[ 0%] Building C object CMakeFiles/microkernels-prod.dir/src/f32-avgpool/f32-avgpool-9p8x-minmax-sse-c4.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/ceiling.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/clamp.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/concatenate.c.o
In file included from /Users/tinuv/Downloads/build1/build/libXNNPACK-source-master/src/f32-avgpool/f32-avgpool-9p8x-minmax-sse-c4.c:12:
In file included from /Users/tinuv/Downloads/build1/build/libXNNPACK-source-master/src/xnnpack/avgpool.h:15:
In file included from /Users/tinuv/Downloads/build1/build/libXNNPACK-source-master/src/xnnpack/microparams.h:12:
In file included from /Users/tinuv/Downloads/build1/build/libXNNPACK-source-master/src/xnnpack/math.h:21:
In file included from /Users/tinuv/Downloads/build1/build/libXNNPACK-build/macos/scratch/x86_64/FP16-source/include/fp16/fp16.h:10:
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:614:27: error: _Float16 is not supported on this target
614 | extern _Float16 __fabsf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:614:8: error: _Float16 is not supported on this target
614 | extern _Float16 __fabsf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:615:28: error: _Float16 is not supported on this target
615 | extern _Float16 __hypotf16(_Float16, _Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:615:38: error: _Float16 is not supported on this target
615 | extern _Float16 __hypotf16(_Float16, _Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:615:8: error: _Float16 is not supported on this target
615 | extern _Float16 __hypotf16(_Float16, _Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:616:27: error: _Float16 is not supported on this target
616 | extern _Float16 __sqrtf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:616:8: error: _Float16 is not supported on this target
616 | extern _Float16 __sqrtf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:617:27: error: _Float16 is not supported on this target
617 | extern _Float16 __ceilf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:617:8: error: _Float16 is not supported on this target
617 | extern _Float16 __ceilf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:618:28: error: _Float16 is not supported on this target
618 | extern _Float16 __floorf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:618:8: error: _Float16 is not supported on this target
618 | extern _Float16 __floorf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:619:27: error: _Float16 is not supported on this target
619 | extern _Float16 __rintf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), wat[ 1%] Built target pthreadpool
chos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:619:8: error: _Float16 is not supported on this target
619 | extern _Float16 __rintf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:620:28: error: _Float16 is not supported on this target
620 | extern _Float16 __roundf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:620:8: error: _Float16 is not supported on this target
620 | extern _Float16 __roundf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:621:28: error: _Float16 is not supported on this target
621 | extern _Float16 __truncf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:621:8: error: _Float16 is not supported on this target
621 | extern _Float16 __truncf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:622:31: error: _Float16 is not supported on this target
622 | extern _Float16 __copysignf16(_Float16, _Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:622:41: error: _Float16 is not supported on this target
622 | extern _Float16 __copysignf16(_Float16, _Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
fatal error: too many errors emitted, stopping now [-ferror-limit=]
20 errors generated.
make[2]: *** [CMakeFiles/microkernels-prod.dir/src/f32-avgpool/f32-avgpool-9p8x-minmax-sse-c4.c.o] Error 1
make[1]: *** [CMakeFiles/microkernels-prod.dir/all] Error 2
make[1]: *** Waiting for unfinished jobs....
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/convert.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/convolution-2d.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/copy.c.o
[ 1%] Built target logging
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/deconvolution-2d.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/copysign.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/depth-to-space-2d.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/divide.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/depthwise-convolution-2d.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/even-split.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/elu.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/exp.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/floor.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/fully-connected-sparse.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/fully-connected.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/gelu.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/global-sum-pooling.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/global-average-pooling.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/hardswish.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/leaky-relu.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/log.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/max-pooling-2d.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/maximum2.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/minimum2.c.o
[ 1%] Building C object CMakeFiles/microkernels-prod.dir/src/f32-avgpool/f32-avgpool-9p8x-minmax-sse-c4.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/multiply2.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/negate.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/reciprocal-square-root.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/prelu.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/reshape-helpers.c.o
In file included from /Users/tinuv/Downloads/build1/build/libXNNPACK-source-master/src/f32-avgpool/f32-avgpool-9p8x-minmax-sse-c4.c:12:
In file included from /Users/tinuv/Downloads/build1/build/libXNNPACK-source-master/src/xnnpack/avgpool.h:15:
In file included from /Users/tinuv/Downloads/build1/build/libXNNPACK-source-master/src/xnnpack/microparams.h:12:
In file included from /Users/tinuv/Downloads/build1/build/libXNNPACK-source-master/src/xnnpack/math.h:21:
In file included from /Users/tinuv/Downloads/build1/build/libXNNPACK-build/macos/scratch/x86_64/FP16-source/include/fp16/fp16.h:10:
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:614:27: error: _Float16 is not supported on this target
614 | extern _Float16 __fabsf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:614:8: error: _Float16 is not supported on this target
614 | extern _Float16 __fabsf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:615:28: error: _Float16 is not supported on this target
615 | extern _Float16 __hypotf16(_Float16, _Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:615:38: error: _Float16 is not supported on this target
615 | extern _Float16 __hypotf16(_Float16, _Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:615:8: error: _Float16 is not supported on this target
615 | extern _Float16 __hypotf16(_Float16, _Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:616:27: error: _Float16 is not supported on this target
616 | extern _Float16 __sqrtf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:616:8: error: _Float16 is not supported on this target
616 | extern _Float16 __sqrtf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:617:27: error: _Float16 is not supported on this target
617 | extern _Float16 __ceilf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:617:8: error: _Float16 is not supported on this target
617 | extern _Float16 __ceilf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:618:28: error: _Float16 is not supported on this target
618 | extern _Float16 __floorf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:618:8: error: _Float16 is not supported on this target
618 | extern _Float16 __floorf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:619:27: error: _Float16 is not supported on this target
619 | extern _Float16 __rintf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:619:8: error: _Float16 is not supported on this target
619 | extern _Float16 __rintf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:620:28: error: _Float16 is not supported on this target
620 | extern _Float16 __roundf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:620:8: error: _Float16 is not supported on this target
620 | extern _Float16 __roundf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:621:28: error: _Float16 is not supported on this target
621 | extern _Float16 __truncf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:621:8: error: _Float16 is not supported on this target
621 | extern _Float16 __truncf16(_Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:622:31: error: _Float16 is not supported on this target
622 | extern _Float16 __copysignf16(_Float16, _Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
/Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX15.0.sdk/usr/include/math.h:622:41: error: _Float16 is not supported on this target
622 | extern _Float16 __copysignf16(_Float16, _Float16) __API_AVAILABLE(macos(15.0), ios(18.0), watchos(11.0), tvos(18.0));
| ^
fatal error: too many errors emitted, stopping now [-ferror-limit=]
20 errors generated.
make[2]: *** [CMakeFiles/microkernels-prod.dir/src/f32-avgpool/f32-avgpool-9p8x-minmax-sse-c4.c.o] Error 1
make[1]: *** [CMakeFiles/microkernels-prod.dir/all] Error 2
make: *** [all] Error 2
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/scaled-dot-product-attention.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/sigmoid.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/softmax.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/space-to-depth-2d.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/square-root.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/square.c.o
(base) ---------------------------------------------------------------------------------------------
~/Downloads/build1/build/libXNNPACK-build/macos/scratch/x86_64 » [ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/squared-difference.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/static-constant-pad.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/static-mean.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/static-resize-bilinear-2d.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/static-slice.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/static-transpose.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/subtract.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/tanh.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/unpooling-2d.c.o
[ 1%] Building C object CMakeFiles/subgraph.dir/src/subgraph/validation.c.o
[ 2%] Building C object CMakeFiles/subgraph.dir/src/tensor.c.o
[ 2%] Built target subgraph
make: *** [all] Error 2
dsharlet commented
I think FP16 really is a necessary dependency, it's not optional for optimization. The thing that would need to be optional is support for float16 ops, which AFAIK is not and cannot easily be made so.
Maratyszcza/FP16#32 is a workaround for this that might be easy for you to apply.
Can you share how you are attempting to cross-compile XNNPACK? I'm not sure if this is an issue with the cross compilation setup (which could be looking for the wrong math.h, i.e. using an x86 math.h when it should be using an ARM one), or an issue with XNNPACK's build.
tinuv commented
https://github.com/Tinuv-Dev/TensorflowLiteBuilder
I'm happy to share my build script, even though it's a bit rough. You just need to enable libXNNPACK in the TensorflowLiteBuilder and comment out the other libraries.
extension TensorflowLiteBuilder {
func initLibrayList() -> [Library] {
var libraryList: [Library] = []
// libraryList.append(.libAbseil)
// libraryList.append(.libeigen)
// libraryList.append(.libflatbuffers)
// libraryList.append(.libNEON_2_SSE)
// libraryList.append(.libcpuinfo)
// libraryList.append(.libruy)
// libraryList.append(.libpthreadpool)
libraryList.append(.libXNNPACK)
// libraryList.append(.libtensorflow)
return libraryList
}
}After that, modify your workDirectory and other configurations as needed.
static let workDirectory = "file:///your/work/path"
static let buildDirectory = workDirectory + "/build"
static let distDirectory = workDirectory + "/dist"
static let patchDirector = workDirectory + "/patch"
static let platforms = PlatformType.allCases.filter { ![.watchos, .watchsimulator, .android].contains($0) }then run the main file to build with XNNPACK. You can see in the Builder file that I checked out the source code for XNNPACK to a specific commit ID to successfully compile it. If you comment out those statements, you'll be able to see the errors I described.
func obtainSource() {
if !FileManager.default.fileExists(atPath: lib.libSourceDirectory.path()) {
var arguments = ["clone", "--recurse-submodules"]
arguments.append(contentsOf: ["--branch", lib.libVersion, lib.libRepoURL, lib.libSourceDirectory.path()])
try! Utility.launch(path: "/usr/bin/git", arguments: arguments)
//if lib == .libXNNPACK {
// try! Utility.launch(path: "/usr/bin/git", arguments: ["checkout", "e716e05befe59f2512f82432fb161ba1fea9969a"],currentDirectoryURL: lib.libSourceDirectory)
//}
}
}tinuv commented
Some compilation options can be found in the LibXNNPACKBuilder file.
dsharlet commented
I think this should be fixed now.