1berry/qtbase
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

QtPrivate::isLatin1: fix SSE2 non-SSE4.1 code

Browse Source 
The implementation was broken. The "high" in PUNPCKHBW's "unpack high
data" means the high 64-bit of the 128-bit, not the high byte of a 16-
bit word. This never worked. It always passed for me because I don't
build non-SSE4.2 code (too old, no longer relevant).

So just use the working version of simdTestMask.

Change-Id: I35a1b4d0a19a43149daefffd16284542f0de3fa3
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
(cherry picked from commit 9422b5ebc3592f4687650a84131e736219308b9f)
This commit is contained in:
Thiago Macieira 2020-08-04 21:37:01 -07:00
parent 11740acbca
commit 46214c96af
1 changed files with 1 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
src/corelib/tools/qstring.cpp
View File

@ -560,31 +560,12 @@ bool QtPrivate::isLatin1(QStringView s) Q_DECL_NOTHROW
const QChar *ptr = s.begin();
const QChar *end = s.end();
#if defined(__SSE4_1__)
#ifdef __SSE2__
const char *ptr8 = reinterpret_cast<const char *>(ptr);
const char *end8 = reinterpret_cast<const char *>(end);
if (!simdTestMask(ptr8, end8, 0xff00ff00))
return false;
ptr = reinterpret_cast<const QChar *>(ptr8);
#elif defined(__SSE2__)
// Testing if every other byte is non-zero can be done efficiently by
// using PUNPCKHBW (unpack high order bytes) and comparing that to zero.
while (ptr + 32 < end) {
__m128i data1 = _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr));
__m128i data2 = _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr + 16));
__m128i high = _mm_unpackhi_epi8(data1, data2);
__m128i comparison = _mm_cmpeq_epi16(high, _mm_setzero_si128());
if (_mm_movemask_epi8(comparison))
return false;
ptr += 16;
}
if (ptr + 16 < end) {
__m128i data1 = _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr));
__m128i high = _mm_unpackhi_epi8(data1, data1);
__m128i comparison = _mm_cmpeq_epi16(high, _mm_setzero_si128());
if (_mm_movemask_epi8(comparison))
return false;
}
#endif
while (ptr != end) {