AuroraMiddleware/qt5base-lts
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Short-cut SSE4 unpremultiply

Browse Source 
Even with SSE4 optimized unpremultiply it is still significantly faster
to skip the calculation on alpha values 0 and 255.

Change-Id: Iafe658fea8eacf35a857f292952b0c1ee056139c
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
This commit is contained in:
Allan Sandfeld Jensen 2015-04-23 15:31:28 +02:00 committed by Allan Sandfeld Jensen
parent ce302a53df
commit bc162382e5
1 changed files with 4 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
src/gui/painting/qdrawingprimitive_sse2_p.h
View File

@ -242,6 +242,8 @@ QT_FUNCTION_TARGET(SSE4_1)
inline QRgb qUnpremultiply_sse4(QRgb p)
{
const uint alpha = qAlpha(p);
if (alpha == 255 || alpha == 0)
return p;
const uint invAlpha = qt_inv_premul_factor[alpha];
const __m128i via = _mm_set1_epi32(invAlpha);
const __m128i vr = _mm_set1_epi32(0x8000);
@ -250,8 +252,8 @@ inline QRgb qUnpremultiply_sse4(QRgb p)
vl = _mm_add_epi32(vl, vr);
vl = _mm_srai_epi32(vl, 16);
vl = _mm_insert_epi32(vl, alpha, 3);
vl = _mm_packus_epi32(vl, _mm_setzero_si128());
vl = _mm_packus_epi16(vl, _mm_setzero_si128());
vl = _mm_packus_epi32(vl, vl);
vl = _mm_packus_epi16(vl, vl);
return _mm_cvtsi128_si32(vl);
}
#endif