[libclc] Add new function implementations in math.

[GYDmedwin]

In the math package, there are only atan.cl, cbrt.cl, cos.cl, sin.cl, and sqrt.cl files initially. The other files in the math package I submitted this time are all newly implemented. We are using the SYCL-CTS test set for testing, one of which is the math_builtin_api test. Only when these files are newly added can it be compiled and passed.

[GYDmedwin]

@bader Excuse me to interrupt, can you please see if the new request meets the requirements?

[GYDmedwin]

@jchlanda Excuse me to interrupt, can you please see if the new request meets the requirements?

[GYDmedwin]

@t4c1 Excuse me to interrupt, can you please see if the new request meets the requirements?

[t4c1]

LGTM.

[bader]

The description doesn't really help reviewers.

Made some changes based on the last suggestion.

Could you clarify what suggestion do you mean here?

Add some new function implementations in the math package, so that the operator can run normally

Which function implementations were added and what is "the operator"? Could you provide more context, please?

[GYDmedwin]

@bader Sorry to trouble you, I mean the suggestions made by you and other reviewers in the last request, including removing redundant comment codes, adding open source information in the code, and modifying the newly added names in the SOURCE file Arrange lexicographically, etc.

[GYDmedwin]

@bader In the math package, there are only atan.cl, cbrt.cl, cos.cl, sin.cl, and sqrt.cl files initially. The other files in the math package I submitted this time are all newly implemented. We are using the SYCL-CTS test set for testing, one of which is the math_builtin_api test. Only when these files are newly added can it be compiled and passed. I'm sorry for not being able to describe clearly before, thank you for your patience to advise me.

[bader]

the last request

Do you refer to #4849?

[bader]

@bader In the math package, there are only atan.cl, cbrt.cl, cos.cl, sin.cl, and sqrt.cl files initially. The other files in the math package I submitted this time are all newly implemented. We are using the SYCL-CTS test set for testing, one of which is the math_builtin_api test. Only when these files are newly added can it be compiled and passed. I'm sorry for not being able to describe clearly before, thank you for your patience to advise me.

Thanks. Please, add this information to the pull request description.

[bader]

@GYDmedwin, could you fix formatting, please?

[bader]

Interesting... clang-format doesn't complain here - all code except the first comment block is shifted by 2 spaces.

[bader]

/summary:run

[zjin-lcf]

I listed the source files that are not found.

CMake Error at /path/to/llvm/libclc/cmake/modules/AddLibclc.cmake:54 (add_library):
Cannot find source file:

/path/to/llvm/libclc/amdgcn-amdhsa/libspirv/math/fma.cl
/path/to/llvm/libclc/amdgcn-amdhsa/libspirv/math/ilogb.cl
/path/to/llvm/libclc/amdgcn-amdhsa/libspirv/math/sinh.cl

Tried extensions .c .C .c++ .cc .cpp .cxx .cu .m .M .mm .h .hh .h++ .hm
.hpp .hxx .in .txx
Call Stack (most recent call first):
/path/to/llvm/libclc/CMakeLists.txt:368 (add_libclc_builtin_set)

[GYDmedwin]

@zjin-lcf Sorry, this is an error, you can delete the file name directly in the SOURCE file.

[zjin-lcf]

I appreciate very much your effort of adding many math functions for the HIP support!

[GYDmedwin]

@zjin-lcf Thank you for your appreciation, and I am also very grateful to you and other reviewers for your suggestions to me, which allowed me to learn a lot of new knowledge.

[zjin-lcf]

@GYDmedwin I have a question. Are 'fma', 'ilogb', and 'sinh' the last three un-implemented math functions ?

[GYDmedwin]

@zjin-lcf In my test，‘ilogb’, ‘log2’, ‘remainder’ the last three un-implemented math functions. I am trying to implement them, but there are still some problems. I did not encounter ‘fma’, it may be my handwriting wrong. The ‘sinh’ source file was dropped, I will upload it now. I am currently repairing ‘ilogb’ and have been testing it. I forgot to delete its name in the SOURCE file. Thank you for your careful inspection, I will correct it immediately.

[zjin-lcf]

@GYDmedwin No problem. I am building the repository with the Python script and observed some errors that abort the compilation.
Thanks for your great work.

In file included from /path/to/llvm/libclc/amdgcn-amdhsa/libspirv/math/ldexp.cl:10:
/path/to/llvm/libclc/generic/include/config.h:23:1: error: unknown type name '_CLC_DECL'
_CLC_DECL bool __clc_subnormals_disabled();
^
/path/to/llvm/libclc/generic/include/config.h:24:1: error: unknown type name '_CLC_DECL'
_CLC_DECL bool __clc_fp16_subnormals_supported();
^
/path/to/llvm/libclc/generic/include/config.h:25:1: error: unknown type name '_CLC_DECL'
_CLC_DECL bool __clc_fp32_subnormals_supported();
^
/path/to/llvm/libclc/generic/include/config.h:26:1: error: unknown type name '_CLC_DECL'
_CLC_DECL bool __clc_fp64_subnormals_supported();
^
In file included from /path/to/llvm/libclc/amdgcn-amdhsa/libspirv/math/ldexp.cl:11:
In file included from /path/to/llvm/libclc/generic/include/math/math.h:13:
/path/to/llvm/libclc/generic/include/config.h:23:1: error: unknown type name '_CLC_DECL'
_CLC_DECL bool __clc_subnormals_disabled();
^
/path/to/llvm/libclc/generic/include/config.h:24:1: error: unknown type name '_CLC_DECL'
_CLC_DECL bool __clc_fp16_subnormals_supported();
^
/path/to/llvm/libclc/generic/include/config.h:25:1: error: unknown type name '_CLC_DECL'
_CLC_DECL bool __clc_fp32_subnormals_supported();
^
/path/to/llvm/libclc/generic/include/config.h:26:1: error: unknown type name '_CLC_DECL'
_CLC_DECL bool __clc_fp64_subnormals_supported();
^
In file included from /path/to/llvm/libclc/amdgcn-amdhsa/libspirv/math/ldexp.cl:10:
/path/to/llvm/libclc/generic/include/config.h:23:1: error: unknown type name '_CLC_DECL'
_CLC_DECL bool __clc_subnormals_disabled();
^
/path/to/llvm/libclc/generic/include/config.h:24:1: error: unknown type name '_CLC_DECL'
_CLC_DECL bool __clc_fp16_subnormals_supported();
^
/path/to/llvm/libclc/generic/include/config.h:25:1: error: unknown type name '_CLC_DECL'
_CLC_DECL bool __clc_fp32_subnormals_supported();
^
/path/to/llvm/libclc/generic/include/config.h:26:1: error: unknown type name '_CLC_DECL'
_CLC_DECL bool __clc_fp64_subnormals_supported();
^
In file included from /path/to/llvm/libclc/amdgcn-amdhsa/libspirv/math/ldexp.cl:11:
In file included from /path/to/llvm/libclc/generic/include/math/math.h:13:
/path/to/llvm/libclc/generic/include/config.h:23:1: error: unknown type name '_CLC_DECL'
_CLC_DECL bool __clc_subnormals_disabled();
^
/path/to/llvm/libclc/generic/include/config.h:24:1: error: unknown type name '_CLC_DECL'
_CLC_DECL bool __clc_fp16_subnormals_supported();
^
/path/to/sycl-space/llvm/libclc/generic/include/config.h:25:1: error: unknown type name '_CLC_DECL'
_CLC_DECL bool __clc_fp32_subnormals_supported();
^
/path/to/llvm/libclc/generic/include/config.h:26:1: error: unknown type name '_CLC_DECL'
_CLC_DECL bool __clc_fp64_subnormals_supported();
^

[GYDmedwin]

@zjin-lcf Sorry, I did not encounter this problem when compiling, I will take time to look at this problem later, thank you.

[zjin-lcf]

I will try again. Thanks.

[zjin-lcf]

I needed to comment the lines in the ldexp.cl so that the compile script in Python would not report compile error shown above.
#include <config.h>
#include <math/math.h>

[zjin-lcf]

@GYDmedwin

I tried an example and got the following error. Is it reproducible ? If not, I will build again. Thanks.

lld: error: undefined hidden symbol: __spirv_ocl_frexp(float, int*)

[zjin-lcf]

The example is https://github.com/zjin-lcf/HeCBench/tree/master/rfs-sycl

[FMarno]

@GYDmedwin I've been able to reproduce the error that @zjin-lcf got so I will look into it.

[FMarno]

@zjin-lcf I've put up a PR here that should fix frexp #5377.
There will likely be similar problems with modf and sincos since they both also use pointers.

[bader]

@FMarno, can we fix modf and sincos in #5377 too?

[FMarno]

@bader Sure, I just need to get my head around the macros first!