maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/compiler-rt/lib/builtins/truncsfhf2.c

33 lines
965 B
C
Raw Normal View History

[Builtins] Implement half-precision conversions. 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
2015-05-13 02:33:42 +08:00
//===-- lib/truncsfhf2.c - single -> half conversion --------------*- C -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file is dual licensed under the MIT and the University of Illinois Open
// Source Licenses. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
#define SRC_SINGLE
#define DST_HALF
#include "fp_trunc_impl.inc"
[Builtins] Implement f2h/h2f by jumping to trunc/extend. Follow-up to r237161; seems like we can't use aliases, but we can do better than duplicating the bodies, especially when that body, after inlining, isn't as small as it looks. Better approaches welcome. Perhaps the best thing is just to have an #ifndef __APPLE__ over the GNUEABI names, since they're not used there. llvm-svn: 237323
2015-05-14 08:50:28 +08:00
// Use a forwarding definition and noinline to implement a poor man's alias,
// as there isn't a good cross-platform way of defining one.
builtins: Use MSVC-equivalents of attributes This allows us to build the builtins using MSVC. NFC. Patch by Tee Hao Wei! llvm-svn: 249375
2015-10-06 12:33:05 +08:00
COMPILER_RT_ABI NOINLINE uint16_t __truncsfhf2(float a) {
[Builtins] Implement half-precision conversions. 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
2015-05-13 02:33:42 +08:00
return __truncXfYf2__(a);
}
COMPILER_RT_ABI uint16_t __gnu_f2h_ieee(float a) {
[Builtins] Implement f2h/h2f by jumping to trunc/extend. Follow-up to r237161; seems like we can't use aliases, but we can do better than duplicating the bodies, especially when that body, after inlining, isn't as small as it looks. Better approaches welcome. Perhaps the best thing is just to have an #ifndef __APPLE__ over the GNUEABI names, since they're not used there. llvm-svn: 237323
2015-05-14 08:50:28 +08:00
return __truncsfhf2(a);
[Builtins] Implement half-precision conversions. 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
2015-05-13 02:33:42 +08:00
}
builtins: expand out the AEABI function stubs These actually may change calling conventions. We cannot simply provide function aliases as the aliased function may have a different calling convention. Provide a forwarding function instead to permit the compiler to synthesize the calling convention adjustment thunk. Remove the `ARM_EABI_FNALIAS` macro as that is not safe to use. Resolves PR33030! llvm-svn: 303188
2017-05-17 00:41:37 +08:00
#if defined(__ARM_EABI__)
[compiler-rt] Add back ARM EABI aliases where legal. r303188 removed all the uses of aliases for EABI functions from compiler-rt, because some of them had mismatched calling conventions. Obviously, we can't use aliases for functions which don't have the same calling convention, but that's only an issue for floating-point functions with the hardfloat ABI. In other cases, the stubs increase size and reduce performance for no benefit. This patch adds back the aliases, with appropriate checks to make sure they're only used in cases where the calling convention matches. llvm-svn: 314851
2017-10-04 05:25:07 +08:00
#if defined(COMPILER_RT_ARMHF_TARGET)
builtins: expand out the AEABI function stubs These actually may change calling conventions. We cannot simply provide function aliases as the aliased function may have a different calling convention. Provide a forwarding function instead to permit the compiler to synthesize the calling convention adjustment thunk. Remove the `ARM_EABI_FNALIAS` macro as that is not safe to use. Resolves PR33030! llvm-svn: 303188
2017-05-17 00:41:37 +08:00
AEABI_RTABI uint16_t __aeabi_f2h(float a) {
return __truncsfhf2(a);
}
[compiler-rt] Add back ARM EABI aliases where legal. r303188 removed all the uses of aliases for EABI functions from compiler-rt, because some of them had mismatched calling conventions. Obviously, we can't use aliases for functions which don't have the same calling convention, but that's only an issue for floating-point functions with the hardfloat ABI. In other cases, the stubs increase size and reduce performance for no benefit. This patch adds back the aliases, with appropriate checks to make sure they're only used in cases where the calling convention matches. llvm-svn: 314851
2017-10-04 05:25:07 +08:00
#else
AEABI_RTABI uint16_t __aeabi_f2h(float a) COMPILER_RT_ALIAS(__truncsfhf2);
#endif
builtins: expand out the AEABI function stubs These actually may change calling conventions. We cannot simply provide function aliases as the aliased function may have a different calling convention. Provide a forwarding function instead to permit the compiler to synthesize the calling convention adjustment thunk. Remove the `ARM_EABI_FNALIAS` macro as that is not safe to use. Resolves PR33030! llvm-svn: 303188
2017-05-17 00:41:37 +08:00
#endif