mirror of
https://github.com/jart/cosmopolitan.git
synced 2025-02-25 07:19:02 +00:00
377 lines
9.1 KiB
C
377 lines
9.1 KiB
C
#ifndef _AARCH64_FP16_H_
|
|
#define _AARCH64_FP16_H_
|
|
#ifdef __aarch64__
|
|
#include "libc/inttypes.h"
|
|
#include "libc/limits.h"
|
|
#include "libc/literal.h"
|
|
|
|
#pragma GCC push_options
|
|
#pragma GCC target("arch=armv8.2-a+fp16")
|
|
|
|
#define FUNC \
|
|
__extension__ extern __inline \
|
|
__attribute__((__always_inline__, __gnu_inline__, __artificial__))
|
|
|
|
typedef __fp16 float16_t;
|
|
|
|
FUNC float16_t vabsh_f16(float16_t __a) {
|
|
return __builtin_aarch64_abshf(__a);
|
|
}
|
|
|
|
FUNC uint16_t vceqzh_f16(float16_t __a) {
|
|
return __builtin_aarch64_cmeqhf_uss(__a, 0.0f);
|
|
}
|
|
|
|
FUNC uint16_t vcgezh_f16(float16_t __a) {
|
|
return __builtin_aarch64_cmgehf_uss(__a, 0.0f);
|
|
}
|
|
|
|
FUNC uint16_t vcgtzh_f16(float16_t __a) {
|
|
return __builtin_aarch64_cmgthf_uss(__a, 0.0f);
|
|
}
|
|
|
|
FUNC uint16_t vclezh_f16(float16_t __a) {
|
|
return __builtin_aarch64_cmlehf_uss(__a, 0.0f);
|
|
}
|
|
|
|
FUNC uint16_t vcltzh_f16(float16_t __a) {
|
|
return __builtin_aarch64_cmlthf_uss(__a, 0.0f);
|
|
}
|
|
|
|
FUNC float16_t vcvth_f16_s16(int16_t __a) {
|
|
return __builtin_aarch64_floathihf(__a);
|
|
}
|
|
|
|
FUNC float16_t vcvth_f16_s32(int32_t __a) {
|
|
return __builtin_aarch64_floatsihf(__a);
|
|
}
|
|
|
|
FUNC float16_t vcvth_f16_s64(int64_t __a) {
|
|
return __builtin_aarch64_floatdihf(__a);
|
|
}
|
|
|
|
FUNC float16_t vcvth_f16_u16(uint16_t __a) {
|
|
return __builtin_aarch64_floatunshihf_us(__a);
|
|
}
|
|
|
|
FUNC float16_t vcvth_f16_u32(uint32_t __a) {
|
|
return __builtin_aarch64_floatunssihf_us(__a);
|
|
}
|
|
|
|
FUNC float16_t vcvth_f16_u64(uint64_t __a) {
|
|
return __builtin_aarch64_floatunsdihf_us(__a);
|
|
}
|
|
|
|
FUNC int16_t vcvth_s16_f16(float16_t __a) {
|
|
return __builtin_aarch64_fix_trunchfhi(__a);
|
|
}
|
|
|
|
FUNC int32_t vcvth_s32_f16(float16_t __a) {
|
|
return __builtin_aarch64_fix_trunchfsi(__a);
|
|
}
|
|
|
|
FUNC int64_t vcvth_s64_f16(float16_t __a) {
|
|
return __builtin_aarch64_fix_trunchfdi(__a);
|
|
}
|
|
|
|
FUNC uint16_t vcvth_u16_f16(float16_t __a) {
|
|
return __builtin_aarch64_fixuns_trunchfhi_us(__a);
|
|
}
|
|
|
|
FUNC uint32_t vcvth_u32_f16(float16_t __a) {
|
|
return __builtin_aarch64_fixuns_trunchfsi_us(__a);
|
|
}
|
|
|
|
FUNC uint64_t vcvth_u64_f16(float16_t __a) {
|
|
return __builtin_aarch64_fixuns_trunchfdi_us(__a);
|
|
}
|
|
|
|
FUNC int16_t vcvtah_s16_f16(float16_t __a) {
|
|
return __builtin_aarch64_lroundhfhi(__a);
|
|
}
|
|
|
|
FUNC int32_t vcvtah_s32_f16(float16_t __a) {
|
|
return __builtin_aarch64_lroundhfsi(__a);
|
|
}
|
|
|
|
FUNC int64_t vcvtah_s64_f16(float16_t __a) {
|
|
return __builtin_aarch64_lroundhfdi(__a);
|
|
}
|
|
|
|
FUNC uint16_t vcvtah_u16_f16(float16_t __a) {
|
|
return __builtin_aarch64_lrounduhfhi_us(__a);
|
|
}
|
|
|
|
FUNC uint32_t vcvtah_u32_f16(float16_t __a) {
|
|
return __builtin_aarch64_lrounduhfsi_us(__a);
|
|
}
|
|
|
|
FUNC uint64_t vcvtah_u64_f16(float16_t __a) {
|
|
return __builtin_aarch64_lrounduhfdi_us(__a);
|
|
}
|
|
|
|
FUNC int16_t vcvtmh_s16_f16(float16_t __a) {
|
|
return __builtin_aarch64_lfloorhfhi(__a);
|
|
}
|
|
|
|
FUNC int32_t vcvtmh_s32_f16(float16_t __a) {
|
|
return __builtin_aarch64_lfloorhfsi(__a);
|
|
}
|
|
|
|
FUNC int64_t vcvtmh_s64_f16(float16_t __a) {
|
|
return __builtin_aarch64_lfloorhfdi(__a);
|
|
}
|
|
|
|
FUNC uint16_t vcvtmh_u16_f16(float16_t __a) {
|
|
return __builtin_aarch64_lflooruhfhi_us(__a);
|
|
}
|
|
|
|
FUNC uint32_t vcvtmh_u32_f16(float16_t __a) {
|
|
return __builtin_aarch64_lflooruhfsi_us(__a);
|
|
}
|
|
|
|
FUNC uint64_t vcvtmh_u64_f16(float16_t __a) {
|
|
return __builtin_aarch64_lflooruhfdi_us(__a);
|
|
}
|
|
|
|
FUNC int16_t vcvtnh_s16_f16(float16_t __a) {
|
|
return __builtin_aarch64_lfrintnhfhi(__a);
|
|
}
|
|
|
|
FUNC int32_t vcvtnh_s32_f16(float16_t __a) {
|
|
return __builtin_aarch64_lfrintnhfsi(__a);
|
|
}
|
|
|
|
FUNC int64_t vcvtnh_s64_f16(float16_t __a) {
|
|
return __builtin_aarch64_lfrintnhfdi(__a);
|
|
}
|
|
|
|
FUNC uint16_t vcvtnh_u16_f16(float16_t __a) {
|
|
return __builtin_aarch64_lfrintnuhfhi_us(__a);
|
|
}
|
|
|
|
FUNC uint32_t vcvtnh_u32_f16(float16_t __a) {
|
|
return __builtin_aarch64_lfrintnuhfsi_us(__a);
|
|
}
|
|
|
|
FUNC uint64_t vcvtnh_u64_f16(float16_t __a) {
|
|
return __builtin_aarch64_lfrintnuhfdi_us(__a);
|
|
}
|
|
|
|
FUNC int16_t vcvtph_s16_f16(float16_t __a) {
|
|
return __builtin_aarch64_lceilhfhi(__a);
|
|
}
|
|
|
|
FUNC int32_t vcvtph_s32_f16(float16_t __a) {
|
|
return __builtin_aarch64_lceilhfsi(__a);
|
|
}
|
|
|
|
FUNC int64_t vcvtph_s64_f16(float16_t __a) {
|
|
return __builtin_aarch64_lceilhfdi(__a);
|
|
}
|
|
|
|
FUNC uint16_t vcvtph_u16_f16(float16_t __a) {
|
|
return __builtin_aarch64_lceiluhfhi_us(__a);
|
|
}
|
|
|
|
FUNC uint32_t vcvtph_u32_f16(float16_t __a) {
|
|
return __builtin_aarch64_lceiluhfsi_us(__a);
|
|
}
|
|
|
|
FUNC uint64_t vcvtph_u64_f16(float16_t __a) {
|
|
return __builtin_aarch64_lceiluhfdi_us(__a);
|
|
}
|
|
|
|
FUNC float16_t vnegh_f16(float16_t __a) {
|
|
return __builtin_aarch64_neghf(__a);
|
|
}
|
|
|
|
FUNC float16_t vrecpeh_f16(float16_t __a) {
|
|
return __builtin_aarch64_frecpehf(__a);
|
|
}
|
|
|
|
FUNC float16_t vrecpxh_f16(float16_t __a) {
|
|
return __builtin_aarch64_frecpxhf(__a);
|
|
}
|
|
|
|
FUNC float16_t vrndh_f16(float16_t __a) {
|
|
return __builtin_aarch64_btrunchf(__a);
|
|
}
|
|
|
|
FUNC float16_t vrndah_f16(float16_t __a) {
|
|
return __builtin_aarch64_roundhf(__a);
|
|
}
|
|
|
|
FUNC float16_t vrndih_f16(float16_t __a) {
|
|
return __builtin_aarch64_nearbyinthf(__a);
|
|
}
|
|
|
|
FUNC float16_t vrndmh_f16(float16_t __a) {
|
|
return __builtin_aarch64_floorhf(__a);
|
|
}
|
|
|
|
FUNC float16_t vrndnh_f16(float16_t __a) {
|
|
return __builtin_aarch64_frintnhf(__a);
|
|
}
|
|
|
|
FUNC float16_t vrndph_f16(float16_t __a) {
|
|
return __builtin_aarch64_ceilhf(__a);
|
|
}
|
|
|
|
FUNC float16_t vrndxh_f16(float16_t __a) {
|
|
return __builtin_aarch64_rinthf(__a);
|
|
}
|
|
|
|
FUNC float16_t vrsqrteh_f16(float16_t __a) {
|
|
return __builtin_aarch64_rsqrtehf(__a);
|
|
}
|
|
|
|
FUNC float16_t vsqrth_f16(float16_t __a) {
|
|
return __builtin_aarch64_sqrthf(__a);
|
|
}
|
|
|
|
FUNC float16_t vaddh_f16(float16_t __a, float16_t __b) {
|
|
return __a + __b;
|
|
}
|
|
|
|
FUNC float16_t vabdh_f16(float16_t __a, float16_t __b) {
|
|
return __builtin_aarch64_fabdhf(__a, __b);
|
|
}
|
|
|
|
FUNC uint16_t vcageh_f16(float16_t __a, float16_t __b) {
|
|
return __builtin_aarch64_facgehf_uss(__a, __b);
|
|
}
|
|
|
|
FUNC uint16_t vcagth_f16(float16_t __a, float16_t __b) {
|
|
return __builtin_aarch64_facgthf_uss(__a, __b);
|
|
}
|
|
|
|
FUNC uint16_t vcaleh_f16(float16_t __a, float16_t __b) {
|
|
return __builtin_aarch64_faclehf_uss(__a, __b);
|
|
}
|
|
|
|
FUNC uint16_t vcalth_f16(float16_t __a, float16_t __b) {
|
|
return __builtin_aarch64_faclthf_uss(__a, __b);
|
|
}
|
|
|
|
FUNC uint16_t vceqh_f16(float16_t __a, float16_t __b) {
|
|
return __builtin_aarch64_cmeqhf_uss(__a, __b);
|
|
}
|
|
|
|
FUNC uint16_t vcgeh_f16(float16_t __a, float16_t __b) {
|
|
return __builtin_aarch64_cmgehf_uss(__a, __b);
|
|
}
|
|
|
|
FUNC uint16_t vcgth_f16(float16_t __a, float16_t __b) {
|
|
return __builtin_aarch64_cmgthf_uss(__a, __b);
|
|
}
|
|
|
|
FUNC uint16_t vcleh_f16(float16_t __a, float16_t __b) {
|
|
return __builtin_aarch64_cmlehf_uss(__a, __b);
|
|
}
|
|
|
|
FUNC uint16_t vclth_f16(float16_t __a, float16_t __b) {
|
|
return __builtin_aarch64_cmlthf_uss(__a, __b);
|
|
}
|
|
|
|
FUNC float16_t vcvth_n_f16_s16(int16_t __a, const int __b) {
|
|
return __builtin_aarch64_scvtfhi(__a, __b);
|
|
}
|
|
|
|
FUNC float16_t vcvth_n_f16_s32(int32_t __a, const int __b) {
|
|
return __builtin_aarch64_scvtfsihf(__a, __b);
|
|
}
|
|
|
|
FUNC float16_t vcvth_n_f16_s64(int64_t __a, const int __b) {
|
|
return __builtin_aarch64_scvtfdihf(__a, __b);
|
|
}
|
|
|
|
FUNC float16_t vcvth_n_f16_u16(uint16_t __a, const int __b) {
|
|
return __builtin_aarch64_ucvtfhi_sus(__a, __b);
|
|
}
|
|
|
|
FUNC float16_t vcvth_n_f16_u32(uint32_t __a, const int __b) {
|
|
return __builtin_aarch64_ucvtfsihf_sus(__a, __b);
|
|
}
|
|
|
|
FUNC float16_t vcvth_n_f16_u64(uint64_t __a, const int __b) {
|
|
return __builtin_aarch64_ucvtfdihf_sus(__a, __b);
|
|
}
|
|
|
|
FUNC int16_t vcvth_n_s16_f16(float16_t __a, const int __b) {
|
|
return __builtin_aarch64_fcvtzshf(__a, __b);
|
|
}
|
|
|
|
FUNC int32_t vcvth_n_s32_f16(float16_t __a, const int __b) {
|
|
return __builtin_aarch64_fcvtzshfsi(__a, __b);
|
|
}
|
|
|
|
FUNC int64_t vcvth_n_s64_f16(float16_t __a, const int __b) {
|
|
return __builtin_aarch64_fcvtzshfdi(__a, __b);
|
|
}
|
|
|
|
FUNC uint16_t vcvth_n_u16_f16(float16_t __a, const int __b) {
|
|
return __builtin_aarch64_fcvtzuhf_uss(__a, __b);
|
|
}
|
|
|
|
FUNC uint32_t vcvth_n_u32_f16(float16_t __a, const int __b) {
|
|
return __builtin_aarch64_fcvtzuhfsi_uss(__a, __b);
|
|
}
|
|
|
|
FUNC uint64_t vcvth_n_u64_f16(float16_t __a, const int __b) {
|
|
return __builtin_aarch64_fcvtzuhfdi_uss(__a, __b);
|
|
}
|
|
|
|
FUNC float16_t vdivh_f16(float16_t __a, float16_t __b) {
|
|
return __a / __b;
|
|
}
|
|
|
|
FUNC float16_t vmaxh_f16(float16_t __a, float16_t __b) {
|
|
return __builtin_aarch64_fmaxhf(__a, __b);
|
|
}
|
|
|
|
FUNC float16_t vmaxnmh_f16(float16_t __a, float16_t __b) {
|
|
return __builtin_aarch64_fmaxhf(__a, __b);
|
|
}
|
|
|
|
FUNC float16_t vminh_f16(float16_t __a, float16_t __b) {
|
|
return __builtin_aarch64_fminhf(__a, __b);
|
|
}
|
|
|
|
FUNC float16_t vminnmh_f16(float16_t __a, float16_t __b) {
|
|
return __builtin_aarch64_fminhf(__a, __b);
|
|
}
|
|
|
|
FUNC float16_t vmulh_f16(float16_t __a, float16_t __b) {
|
|
return __a * __b;
|
|
}
|
|
|
|
FUNC float16_t vmulxh_f16(float16_t __a, float16_t __b) {
|
|
return __builtin_aarch64_fmulxhf(__a, __b);
|
|
}
|
|
|
|
FUNC float16_t vrecpsh_f16(float16_t __a, float16_t __b) {
|
|
return __builtin_aarch64_frecpshf(__a, __b);
|
|
}
|
|
|
|
FUNC float16_t vrsqrtsh_f16(float16_t __a, float16_t __b) {
|
|
return __builtin_aarch64_rsqrtshf(__a, __b);
|
|
}
|
|
|
|
FUNC float16_t vsubh_f16(float16_t __a, float16_t __b) {
|
|
return __a - __b;
|
|
}
|
|
|
|
FUNC float16_t vfmah_f16(float16_t __a, float16_t __b, float16_t __c) {
|
|
return __builtin_aarch64_fmahf(__b, __c, __a);
|
|
}
|
|
|
|
FUNC float16_t vfmsh_f16(float16_t __a, float16_t __b, float16_t __c) {
|
|
return __builtin_aarch64_fnmahf(__b, __c, __a);
|
|
}
|
|
|
|
#pragma GCC pop_options
|
|
|
|
#undef FUNC
|
|
#endif /* __aarch64__ */
|
|
#endif /* _AARCH64_FP16_H_ */
|