mirrors/cosmopolitan
1
0
Fork 0
mirror of https://github.com/jart/cosmopolitan.git synced 2025-05-31 17:52:27 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

Get rid of some legacy code

Browse source
This commit is contained in:
Justine Tunney 2024-08-24 17:53:30 -07:00
parent 37ca1badaf
commit 38cc4b3c68
No known key found for this signature in database
GPG key ID: BE714B4575D6E328
27 changed files with 123 additions and 600 deletions
Download patch file Download diff file Expand all files Collapse all files

65
libc/str/tprecode8to16.c
View file

@ -16,34 +16,61 @@
│ TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR │
│ PERFORMANCE OF THIS SOFTWARE. │
╚─────────────────────────────────────────────────────────────────────────────*/
#include "libc/intrin/pcmpgtb.h"
#include "libc/intrin/pmovmskb.h"
#include "libc/intrin/punpckhbw.h"
#include "libc/intrin/punpcklbw.h"
#include <arm_neon.h>
#include <stdint.h>
#include <string.h>
#include "libc/dce.h"
#include "libc/str/str.h"
#include "libc/str/thompike.h"
#include "libc/str/utf16.h"
#include "third_party/aarch64/arm_neon.internal.h"
#include "third_party/intel/emmintrin.internal.h"
#if !IsModeDbg()
#if defined(__x86_64__)
// 34x speedup for ascii
static inline axdx_t tprecode8to16_sse2(char16_t *dst, size_t dstsize,
const char *src, axdx_t r) {
uint8_t v1[16], v2[16], vz[16];
memset(vz, 0, 16);
__m128i v1, v2, vz;
vz = _mm_setzero_si128();
while (r.ax + 16 < dstsize) {
memcpy(v1, src + r.dx, 16);
pcmpgtb((int8_t *)v2, (int8_t *)v1, (int8_t *)vz);
if (pmovmskb(v2) != 0xFFFF)
v1 = _mm_loadu_si128((__m128i *)(src + r.dx));
v2 = _mm_cmpgt_epi8(v1, vz);
if (_mm_movemask_epi8(v2) != 0xFFFF)
break;
punpcklbw(v2, v1, vz);
punpckhbw(v1, v1, vz);
memcpy(dst + r.ax + 0, v2, 16);
memcpy(dst + r.ax + 8, v1, 16);
__m128i lo = _mm_unpacklo_epi8(v1, vz);
__m128i hi = _mm_unpackhi_epi8(v1, vz);
_mm_storeu_si128((__m128i *)(dst + r.ax), lo);
_mm_storeu_si128((__m128i *)(dst + r.ax + 8), hi);
r.ax += 16;
r.dx += 16;
}
return r;
}
#elif defined(__aarch64__)
static inline axdx_t tprecode8to16_neon(char16_t *dst, size_t dstsize,
const char *src, axdx_t r) {
uint8x16_t v1;
while (r.ax + 16 < dstsize) {
v1 = vld1q_u8((const uint8_t *)(src + r.dx));
uint8x16_t cmp = vcgtq_u8(v1, vdupq_n_u8(0));
if (vaddvq_u8(cmp) != 16 * 0xFF)
break;
uint16x8_t lo = vmovl_u8(vget_low_u8(v1));
uint16x8_t hi = vmovl_u8(vget_high_u8(v1));
vst1q_u16((uint16_t *)(dst + r.ax), lo);
vst1q_u16((uint16_t *)(dst + r.ax + 8), hi);
r.ax += 16;
r.dx += 16;
}
return r;
}
#endif
#endif
/**
* Transcodes UTF-8 to UTF-16.
*
@ -64,10 +91,14 @@ axdx_t tprecode8to16(char16_t *dst, size_t dstsize, const char *src) {
r.ax = 0;
r.dx = 0;
for (;;) {
#if defined(__x86_64__) && !IsModeDbg()
if (!((uintptr_t)(src + r.dx) & 15)) {
#if !IsModeDbg()
#if defined(__x86_64__)
if (!((uintptr_t)(src + r.dx) & 15))
r = tprecode8to16_sse2(dst, dstsize, src, r);
}
#elif defined(__aarch64__)
if (!((uintptr_t)(src + r.dx) & 15))
r = tprecode8to16_neon(dst, dstsize, src, r);
#endif
#endif
x = src[r.dx++] & 0377;
if (x >= 0300) {