Use of long double can be error-prone since it could be one of 80-bit
extended precision float, IEEE 128-bit float, or IBM 128-bit float.
Instead use an explicit xf_float typedef for the remaining cases where
long double is being used in the implementation. This patch does not
touch
the PPC specializations which still use long double.
GCC provides these functions (e.g. __addtf3, etc.) in libgcc on x86_64.
Since Clang supports float128, we can also enable the existing code by
using float128 for fp_t if either __FLOAT128__ or __SIZEOF_FLOAT128__ is
defined instead of only supporting these builtins for platforms with
128-bit IEEE long doubles.
This commit defines a new tf_float typedef that matches a float with
attribute((mode(TF)) on each given architecture.
There are more tests that could be enabled for x86, but to keep the diff
smaller, I restricted test changes to ones that started failing as part
of this refactoring.
This change has been tested on x86 (natively) and
aarch64,powerpc64,riscv64 and sparc64 via qemu-user.
This supersedes https://reviews.llvm.org/D98261 and should also cover
the changes from https://github.com/llvm/llvm-project/pull/68041.
This patch implements __extendxftf2 (long double -> f128) and
__trunctfxf2 (f128 -> long double) on x86_64.
This is a preparation to unblock https://reviews.llvm.org/D53608,
We intentionally do not modify compiler-rt/lib/builtins/fp_lib.h in this
PR
(in particular, to limit the scope and avoid exposing other functions on
X86_64 in this PR).
Instead, TODOs were added to use fp_lib.h once it is available.
Test plan:
1. ninja check-compiler-rt (verified on X86_64 and on Aarch64)
In particular, new tests (extendxftf2_test.c and trunctfxf2_test.c) were
added.
2. compared the results of conversions with what other compilers (gcc)
produce.
This reverts commit 0729d00135ccbe1514a49168bbf26986823235c2.
Buildbot https://lab.llvm.org/buildbot/#/builders/19/builds/11989 failed
with "error: ISO C requires a translation unit to contain at least one declaration [-Werror,-Wempty-translation-unit]"
#56854 shows a backwards compatibility problem when builtins of compiler-rt don't follow ABI. We need to prevent to fall into the trap again for BF16.
Reviewed By: bkramer
Differential Revision: https://reviews.llvm.org/D131147
Instead of detecting `_Float16` support at CMake configuration time,
detect it at compile time by checking for the predefined (by the
compiler) macro `__FLT16_MAX__` instead.
This solves the issue where compiler-rt is built simultaneously for both
x86_64 and i386 targets, and the CMake configuration uses x86_64
compilation to detect `_Float16` support, while it may not be supported
by the i386 target (if it does not have SSE2).
While here, rename `COMPILERT_RT_HAS_FLOAT16` to `CRT_HAS_FLOAT16`, to
conform more to the naming style used in `int_lib.h` and `int_types.h`.
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D130718
This is modeled after the half-precision fp support. Two new nodes are
introduced for casting from and to bf16. Since casting from bf16 is a
simple operation I opted to always directly lower it to integer
arithmetic. The other way round is more complicated if you want to
preserve IEEE semantics, so it's handled by a new __truncsfbf2
compiler-rt builtin.
This is of course very bare bones, but sufficient to get a semi-softened
fadd on x86.
Possible future improvements:
- Targets with bf16 conversion instructions can now make fp_to_bf16 legal
- The software conversion to bf16 can be replaced by a trivial
implementation under fast math.
Differential Revision: https://reviews.llvm.org/D126953
On AArch64 it allows use the native FP16 ABI (although libcalls are
not emitted for fptrunc/fpext lowering), while on other architectures
the expected current semantic is preserved (arm for instance).
For testing the _Float16 usage is enabled by architecture base,
currently only for arm, aarch64, and arm64.
This re-enabled revert done by https://reviews.llvm.org/rGb534beabeed3ba1777cd0ff9ce552d077e496726
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D92241
Revert "[compiler-rt] [builtins] Support conversion between fp16 and fp128" & dependency
Revert "[compiler-rt] [builtins] Use _Float16 on extendhfsf2, truncdfhf2 __truncsfhf2 if available"
This reverts commit 7a948298813c913841a36ed0b460db0856fe7082.
This reverts commit 1fb91fcf9cfe849c7e9996597c491306e34e7abc.
On AArch64 it allows use the native FP16 ABI (although libcalls are
not emitted for fptrunc/fpext lowering), while on other architectures
the expected current semantic is preserved (arm for instance).
Differential Revision: https://reviews.llvm.org/D91733
Update formatting to use the LLVM style.
This is part of the cleanup proposed in "[RFC] compiler-rt builtins
cleanup and refactoring".
Differential Revision: https://reviews.llvm.org/D60351
llvm-svn: 359410
to reflect the new license. These used slightly different spellings that
defeated my regular expressions.
We understand that people may be surprised that we're moving the header
entirely to discuss the new license. We checked this carefully with the
Foundation's lawyer and we believe this is the correct approach.
Essentially, all code in the project is now made available by the LLVM
project under our new license, so you will see that the license headers
include that license only. Some of our contributors have contributed
code under our old license, and accordingly, we have retained a copy of
our old license notice in the top-level files in each project and
repository.
llvm-svn: 351648
__inline is a vendor specific spelling for inline. clang and gcc treat it the
same as inline, and is available in MSVC 2013 which does not implement C99
(VS2015 supports the inline keyword though). This will allow us to build the
builtins using MSVC.
llvm-svn: 249953
Mostly uninteresting, except:
- in __extendXfYf2, when checking if the number is normal, the old
code relied on the unsignedness of src_rep_t, which is a problem
when sizeof(src_rep_t) < sizeof(int): the result gets promoted to
int, the signedness of which breaks the comparison.
I added an explicit cast; it shouldn't affect other types.
- we can't pass __fp16, so src_t and src_rep_t are the same.
- the gnu_*_ieee symbols are simply duplicated definitions, as aliases
are problematic on mach-o (where only weak aliases are supported;
that's not what we want).
Differential Revision: http://reviews.llvm.org/D9693
llvm-svn: 237161
