diff options
Diffstat (limited to 'src/opts/SkPMFloat_SSE2.h')
-rw-r--r-- | src/opts/SkPMFloat_SSE2.h | 35 |
1 files changed, 24 insertions, 11 deletions
diff --git a/src/opts/SkPMFloat_SSE2.h b/src/opts/SkPMFloat_SSE2.h index 2a85b1a74f..156c0c9897 100644 --- a/src/opts/SkPMFloat_SSE2.h +++ b/src/opts/SkPMFloat_SSE2.h @@ -1,3 +1,10 @@ +/* + * Copyright 2015 Google Inc. + * + * Use of this source code is governed by a BSD-style license that can be + * found in the LICENSE file. + */ + inline SkPMFloat& SkPMFloat::operator=(const SkPMFloat& that) { fColors = that.fColors; return *this; @@ -34,25 +41,31 @@ inline SkPMColor SkPMFloat::clamped() const { return c; } -inline void SkPMFloat::From4PMColors(SkPMFloat floats[4], const SkPMColor colors[4]) { +inline void SkPMFloat::From4PMColors(const SkPMColor colors[4], + SkPMFloat* a, SkPMFloat* b, SkPMFloat* c, SkPMFloat* d) { // Haven't beaten this yet. - for (int i = 0; i < 4; i++) { floats[i] = FromPMColor(colors[i]); } + *a = FromPMColor(colors[0]); + *b = FromPMColor(colors[1]); + *c = FromPMColor(colors[2]); + *d = FromPMColor(colors[3]); } -inline void SkPMFloat::To4PMColors(SkPMColor colors[4], const SkPMFloat floats[4]) { - SkASSERT(floats[0].isValid() && floats[1].isValid() - && floats[2].isValid() && floats[3].isValid()); +inline void SkPMFloat::To4PMColors( + const SkPMFloat& a, const SkPMFloat& b, const SkPMFloat&c, const SkPMFloat& d, + SkPMColor colors[4]) { // Haven't beaten this yet. - ClampTo4PMColors(colors, floats); + ClampTo4PMColors(a,b,c,d, colors); } -inline void SkPMFloat::ClampTo4PMColors(SkPMColor colors[4], const SkPMFloat floats[4]) { +inline void SkPMFloat::ClampTo4PMColors( + const SkPMFloat& a, const SkPMFloat& b, const SkPMFloat&c, const SkPMFloat& d, + SkPMColor colors[4]) { // Same as _SSSE3.h's. We use 3 _mm_packus_epi16() where the naive loop uses 8. // We don't use _mm_cvtps_epi32, because we want precise control over how 0.5 rounds (up). - __m128i c0 = _mm_cvttps_epi32(_mm_add_ps(_mm_set1_ps(0.5f), floats[0].fColors)), - c1 = _mm_cvttps_epi32(_mm_add_ps(_mm_set1_ps(0.5f), floats[1].fColors)), - c2 = _mm_cvttps_epi32(_mm_add_ps(_mm_set1_ps(0.5f), floats[2].fColors)), - c3 = _mm_cvttps_epi32(_mm_add_ps(_mm_set1_ps(0.5f), floats[3].fColors)); + __m128i c0 = _mm_cvttps_epi32(_mm_add_ps(_mm_set1_ps(0.5f), a.fColors)), + c1 = _mm_cvttps_epi32(_mm_add_ps(_mm_set1_ps(0.5f), b.fColors)), + c2 = _mm_cvttps_epi32(_mm_add_ps(_mm_set1_ps(0.5f), c.fColors)), + c3 = _mm_cvttps_epi32(_mm_add_ps(_mm_set1_ps(0.5f), d.fColors)); __m128i c3210 = _mm_packus_epi16(_mm_packus_epi16(c0, c1), _mm_packus_epi16(c2, c3)); _mm_storeu_si128((__m128i*)colors, c3210); |