cuberite/libdeflate
1
0
Fork 0
mirror of https://github.com/cuberite/libdeflate.git synced 2025-08-03 17:56:17 -04:00
Code Issues Packages Projects Releases Wiki Activity

lz-copy-sse2

Browse Source
This commit is contained in:
Eric Biggers 2018-12-25 15:44:16 -06:00
parent 166084acaa
commit 4abbda6b71
1 changed files with 16 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
lib/decompress_template.h
View File

@ -31,6 +31,8 @@
* target instruction sets. * target instruction sets.
*/ */
#include <emmintrin.h>
static enum libdeflate_result ATTRIBUTES static enum libdeflate_result ATTRIBUTES
FUNCNAME(struct libdeflate_decompressor * restrict d, FUNCNAME(struct libdeflate_decompressor * restrict d,
const void * restrict in, size_t in_nbytes, const void * restrict in, size_t in_nbytes,
@ -347,10 +349,21 @@ have_decode_tables:
out_next += length; out_next += length;
if (UNALIGNED_ACCESS_IS_FAST && if (UNALIGNED_ACCESS_IS_FAST &&
/* max overrun is writing 3 words for a min length match */ /* max overrun is writing 4 words for a min length match */
likely(out_end - out_next >= likely(out_end - out_next >=
3 * WORDBYTES - DEFLATE_MIN_MATCH_LEN)) { 4 * WORDBYTES - DEFLATE_MIN_MATCH_LEN)) {
if (offset >= WORDBYTES) { /* words don't overlap? */ if (offset >= 2 * WORDBYTES) {
_mm_storeu_si128((__m128i*)dst,
_mm_loadu_si128((__m128i*)src));
src += 16;
dst += 16;
do {
_mm_storeu_si128((__m128i*)dst,
_mm_loadu_si128((__m128i*)src));
src += 16;
dst += 16;
} while (dst < out_next);
} else if (offset >= WORDBYTES) { /* words don't overlap? */
copy_word_unaligned(src, dst); copy_word_unaligned(src, dst);
src += WORDBYTES; src += WORDBYTES;
dst += WORDBYTES; dst += WORDBYTES;