gfx/angle/src/libGLESv2/renderer/ImageSSE2.cpp@129ffea94266 (annotated)
gfx/angle/src/libGLESv2/renderer/ImageSSE2.cpp
Sat, 03 Jan 2015 20:18:00 +0100
- author
- Michael Schloh von Bennewitz <michael@schloh.com>
- date
- Sat, 03 Jan 2015 20:18:00 +0100
- branch
- TOR_BUG_3246
- changeset 7
- 129ffea94266
- permissions
- -rw-r--r--
Conditionally enable double key logic according to:
private browsing mode or privacy.thirdparty.isolate preference and
implement in GetCookieStringCommon and FindCookie where it counts...
With some reservations of how to convince FindCookie users to test
condition and pass a nullptr when disabling double key logic.
| michael@0 | 1 | #include "precompiled.h" |
| michael@0 | 2 | // |
| michael@0 | 3 | // Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. |
| michael@0 | 4 | // Use of this source code is governed by a BSD-style license that can be |
| michael@0 | 5 | // found in the LICENSE file. |
| michael@0 | 6 | // |
| michael@0 | 7 | |
| michael@0 | 8 | // ImageSSE2.cpp: Implements SSE2-based functions of rx::Image class. It's |
| michael@0 | 9 | // in a separated file for GCC, which can enable SSE usage only per-file, |
| michael@0 | 10 | // not for code blocks that use SSE2 explicitly. |
| michael@0 | 11 | |
| michael@0 | 12 | #include "libGLESv2/Texture.h" |
| michael@0 | 13 | #include "libGLESv2/renderer/Image.h" |
| michael@0 | 14 | |
| michael@0 | 15 | namespace rx |
| michael@0 | 16 | { |
| michael@0 | 17 | |
| michael@0 | 18 | void Image::loadRGBAUByteDataToBGRASSE2(GLsizei width, GLsizei height, |
| michael@0 | 19 | int inputPitch, const void *input, size_t outputPitch, void *output) |
| michael@0 | 20 | { |
| michael@0 | 21 | const unsigned int *source = NULL; |
| michael@0 | 22 | unsigned int *dest = NULL; |
| michael@0 | 23 | __m128i brMask = _mm_set1_epi32(0x00ff00ff); |
| michael@0 | 24 | |
| michael@0 | 25 | for (int y = 0; y < height; y++) |
| michael@0 | 26 | { |
| michael@0 | 27 | source = reinterpret_cast<const unsigned int*>(static_cast<const unsigned char*>(input) + y * inputPitch); |
| michael@0 | 28 | dest = reinterpret_cast<unsigned int*>(static_cast<unsigned char*>(output) + y * outputPitch); |
| michael@0 | 29 | int x = 0; |
| michael@0 | 30 | |
| michael@0 | 31 | // Make output writes aligned |
| michael@0 | 32 | for (x = 0; ((reinterpret_cast<intptr_t>(&dest[x]) & 15) != 0) && x < width; x++) |
| michael@0 | 33 | { |
| michael@0 | 34 | unsigned int rgba = source[x]; |
| michael@0 | 35 | dest[x] = (_rotl(rgba, 16) & 0x00ff00ff) | (rgba & 0xff00ff00); |
| michael@0 | 36 | } |
| michael@0 | 37 | |
| michael@0 | 38 | for (; x + 3 < width; x += 4) |
| michael@0 | 39 | { |
| michael@0 | 40 | __m128i sourceData = _mm_loadu_si128(reinterpret_cast<const __m128i*>(&source[x])); |
| michael@0 | 41 | // Mask out g and a, which don't change |
| michael@0 | 42 | __m128i gaComponents = _mm_andnot_si128(brMask, sourceData); |
| michael@0 | 43 | // Mask out b and r |
| michael@0 | 44 | __m128i brComponents = _mm_and_si128(sourceData, brMask); |
| michael@0 | 45 | // Swap b and r |
| michael@0 | 46 | __m128i brSwapped = _mm_shufflehi_epi16(_mm_shufflelo_epi16(brComponents, _MM_SHUFFLE(2, 3, 0, 1)), _MM_SHUFFLE(2, 3, 0, 1)); |
| michael@0 | 47 | __m128i result = _mm_or_si128(gaComponents, brSwapped); |
| michael@0 | 48 | _mm_store_si128(reinterpret_cast<__m128i*>(&dest[x]), result); |
| michael@0 | 49 | } |
| michael@0 | 50 | |
| michael@0 | 51 | // Perform leftover writes |
| michael@0 | 52 | for (; x < width; x++) |
| michael@0 | 53 | { |
| michael@0 | 54 | unsigned int rgba = source[x]; |
| michael@0 | 55 | dest[x] = (_rotl(rgba, 16) & 0x00ff00ff) | (rgba & 0xff00ff00); |
| michael@0 | 56 | } |
| michael@0 | 57 | } |
| michael@0 | 58 | } |
| michael@0 | 59 | |
| michael@0 | 60 | void Image::loadAlphaDataToBGRASSE2(GLsizei width, GLsizei height, |
| michael@0 | 61 | int inputPitch, const void *input, size_t outputPitch, void *output) |
| michael@0 | 62 | { |
| michael@0 | 63 | const unsigned char *source = NULL; |
| michael@0 | 64 | unsigned int *dest = NULL; |
| michael@0 | 65 | __m128i zeroWide = _mm_setzero_si128(); |
| michael@0 | 66 | |
| michael@0 | 67 | for (int y = 0; y < height; y++) |
| michael@0 | 68 | { |
| michael@0 | 69 | source = static_cast<const unsigned char*>(input) + y * inputPitch; |
| michael@0 | 70 | dest = reinterpret_cast<unsigned int*>(static_cast<unsigned char*>(output) + y * outputPitch); |
| michael@0 | 71 | |
| michael@0 | 72 | int x; |
| michael@0 | 73 | // Make output writes aligned |
| michael@0 | 74 | for (x = 0; ((reinterpret_cast<intptr_t>(&dest[x]) & 0xF) != 0 && x < width); x++) |
| michael@0 | 75 | { |
| michael@0 | 76 | dest[x] = static_cast<unsigned int>(source[x]) << 24; |
| michael@0 | 77 | } |
| michael@0 | 78 | |
| michael@0 | 79 | for (; x + 7 < width; x += 8) |
| michael@0 | 80 | { |
| michael@0 | 81 | __m128i sourceData = _mm_loadl_epi64(reinterpret_cast<const __m128i*>(&source[x])); |
| michael@0 | 82 | // Interleave each byte to 16bit, make the lower byte to zero |
| michael@0 | 83 | sourceData = _mm_unpacklo_epi8(zeroWide, sourceData); |
| michael@0 | 84 | // Interleave each 16bit to 32bit, make the lower 16bit to zero |
| michael@0 | 85 | __m128i lo = _mm_unpacklo_epi16(zeroWide, sourceData); |
| michael@0 | 86 | __m128i hi = _mm_unpackhi_epi16(zeroWide, sourceData); |
| michael@0 | 87 | |
| michael@0 | 88 | _mm_store_si128(reinterpret_cast<__m128i*>(&dest[x]), lo); |
| michael@0 | 89 | _mm_store_si128(reinterpret_cast<__m128i*>(&dest[x + 4]), hi); |
| michael@0 | 90 | } |
| michael@0 | 91 | |
| michael@0 | 92 | // Handle the remainder |
| michael@0 | 93 | for (; x < width; x++) |
| michael@0 | 94 | { |
| michael@0 | 95 | dest[x] = static_cast<unsigned int>(source[x]) << 24; |
| michael@0 | 96 | } |
| michael@0 | 97 | } |
| michael@0 | 98 | } |
| michael@0 | 99 | |
| michael@0 | 100 | } |
