mirror of
https://github.com/jart/cosmopolitan.git
synced 2025-02-07 15:03:34 +00:00
c9fecf3a55
- You can now run `make -j8 toolchain` on Windows - You can now run `make -j` on MacOS ARM64 and BSD OSes - You can now use our Emacs dev environment on MacOS/Windows - Fix bug where the x16 register was being corrupted by --ftrace - The programs under build/bootstrap/ are updated as fat binaries - The Makefile now explains how to download cosmocc-0.0.12 toolchain - The build scripts under bin/ now support "cosmo" branded toolchains - stat() now goes faster on Windows (shaves 100ms off `make` latency) - Code cleanup and added review on the Windows signal checking code - posix_spawnattr_setrlimit() now works around MacOS ARM64 bugs - Landlock Make now favors posix_spawn() on non-Linux/OpenBSD - posix_spawn() now has better --strace logging on Windows - fstatat() can now avoid EACCES in more cases on Windows - fchmod() can now change the readonly bit on Windows
84 lines
4.6 KiB
C
84 lines
4.6 KiB
C
#ifndef COSMOPOLITAN_LIBC_INTRIN_MACROS_H_
|
|
#define COSMOPOLITAN_LIBC_INTRIN_MACROS_H_
|
|
#include "libc/dce.h"
|
|
#include "libc/nexgen32e/x86feature.h"
|
|
#if !(__ASSEMBLER__ + __LINKER__ + 0)
|
|
|
|
#define INTRIN_COMMUTATIVE "%"
|
|
#define INTRIN_NONCOMMUTATIVE
|
|
|
|
#if defined(__x86_64__) && !defined(__STRICT_ANSI__)
|
|
|
|
typedef char __intrin_xmm_t
|
|
__attribute__((__vector_size__(16), __aligned__(16), __may_alias__));
|
|
|
|
#define INTRIN_SSEVEX_X_X_X_(PURE, ISA, OP, FLAGS, A, B, C) \
|
|
do { \
|
|
if (X86_HAVE(ISA)) { \
|
|
__intrin_xmm_t *Xmm0 = (void *)(A); \
|
|
const __intrin_xmm_t *Xmm1 = (const __intrin_xmm_t *)(B); \
|
|
const __intrin_xmm_t *Xmm2 = (const __intrin_xmm_t *)(C); \
|
|
if (!X86_NEED(AVX)) { \
|
|
asm(OP "\t%1,%0" : "=x"(*Xmm0) : FLAGS "x"(*Xmm2), "0"(*Xmm1)); \
|
|
} else { \
|
|
asm("v" OP "\t%2,%1,%0" : "=x"(*Xmm0) : FLAGS "x"(*Xmm1), "x"(*Xmm2)); \
|
|
} \
|
|
} else { \
|
|
PURE(A, B, C); \
|
|
} \
|
|
} while (0)
|
|
|
|
#define INTRIN_SSEVEX_X_X_I_(PURE, ISA, OP, A, B, I) \
|
|
do { \
|
|
if (X86_HAVE(ISA)) { \
|
|
__intrin_xmm_t *Xmm0 = (void *)(A); \
|
|
const __intrin_xmm_t *Xmm1 = (const __intrin_xmm_t *)(B); \
|
|
if (!X86_NEED(AVX)) { \
|
|
asm(OP "\t%2,%1,%0" : "=x"(*Xmm0) : "x"(*Xmm1), "i"(I)); \
|
|
} else { \
|
|
asm("v" OP "\t%2,%1,%0" : "=x"(*Xmm0) : "x"(*Xmm1), "i"(I)); \
|
|
} \
|
|
} else { \
|
|
PURE(A, B, I); \
|
|
} \
|
|
} while (0)
|
|
|
|
#define INTRIN_SSEVEX_X_X_(PURE, ISA, OP, A, B) \
|
|
do { \
|
|
if (X86_HAVE(ISA)) { \
|
|
__intrin_xmm_t *Xmm0 = (void *)(A); \
|
|
const __intrin_xmm_t *Xmm1 = (const __intrin_xmm_t *)(B); \
|
|
if (!X86_NEED(AVX)) { \
|
|
asm(OP "\t%1,%0" : "=x"(*Xmm0) : "0"(*Xmm1)); \
|
|
} else { \
|
|
asm("v" OP "\t%1,%0" : "=x"(*Xmm0) : "x"(*Xmm1)); \
|
|
} \
|
|
} else { \
|
|
PURE(A, B); \
|
|
} \
|
|
} while (0)
|
|
|
|
#define INTRIN_SSEVEX_X_I_(PURE, ISA, OP, A, B, I) \
|
|
do { \
|
|
if (!IsModeDbg() && X86_HAVE(ISA)) { \
|
|
__intrin_xmm_t *Xmm0 = (void *)(A); \
|
|
const __intrin_xmm_t *Xmm1 = (const __intrin_xmm_t *)(B); \
|
|
if (!X86_NEED(AVX)) { \
|
|
asm(OP "\t%1,%0" : "=x"(*Xmm0) : "i"(I), "0"(*Xmm1)); \
|
|
} else { \
|
|
asm("v" OP "\t%2,%1,%0" : "=x"(*Xmm0) : "x"(*Xmm1), "i"(I)); \
|
|
} \
|
|
} else { \
|
|
PURE(A, B, I); \
|
|
} \
|
|
} while (0)
|
|
|
|
#else
|
|
#define INTRIN_SSEVEX_X_X_X_(PURE, ISA, OP, FLAGS, A, B, C) PURE(A, B, C)
|
|
#define INTRIN_SSEVEX_X_X_I_(PURE, ISA, OP, A, B, I) PURE(A, B, I)
|
|
#define INTRIN_SSEVEX_X_I_(PURE, ISA, OP, A, B, I) PURE(A, B, I)
|
|
#define INTRIN_SSEVEX_X_X_(PURE, ISA, OP, A, B) PURE(A, B)
|
|
#endif /* X86 && !ANSI */
|
|
|
|
#endif /* !(__ASSEMBLER__ + __LINKER__ + 0) */
|
|
#endif /* COSMOPOLITAN_LIBC_INTRIN_MACROS_H_ */
|