libwebp/src/dsp/alpha_processing_sse41.c
Pascal Massimino bfc300c7ff SSE4.1 implementation of some alpha-processing functions
DispatchAlpha* functions are hard to speed up, compared to SSE2.
ExtractAlpha sees a ~15% speed-up though.

Change-Id: I8715c2defecbc832f469eed7e6ffd012146b52de
2015-06-19 14:17:39 -07:00

93 lines
3.6 KiB
C

// Copyright 2015 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Utilities for processing transparent channel, SSE4.1 variant.
//
// Author: Skal (pascal.massimino@gmail.com)
#include "./dsp.h"
#if defined(WEBP_USE_SSE41)
#include <smmintrin.h>
//------------------------------------------------------------------------------
static int ExtractAlpha(const uint8_t* argb, int argb_stride,
int width, int height,
uint8_t* alpha, int alpha_stride) {
// alpha_and stores an 'and' operation of all the alpha[] values. The final
// value is not 0xff if any of the alpha[] is not equal to 0xff.
uint32_t alpha_and = 0xff;
int i, j;
const __m128i all_0xff = _mm_set1_epi32(~0u);
__m128i all_alphas = all_0xff;
// We must be able to access 3 extra bytes after the last written byte
// 'src[4 * width - 4]', because we don't know if alpha is the first or the
// last byte of the quadruplet.
const int limit = (width - 1) & ~15;
const __m128i kCstAlpha0 = _mm_set_epi8(-1, -1, -1, -1, -1, -1, -1, -1,
-1, -1, -1, -1, 12, 8, 4, 0);
const __m128i kCstAlpha1 = _mm_set_epi8(-1, -1, -1, -1, -1, -1, -1, -1,
12, 8, 4, 0, -1, -1, -1, -1);
const __m128i kCstAlpha2 = _mm_set_epi8(-1, -1, -1, -1, 12, 8, 4, 0,
-1, -1, -1, -1, -1, -1, -1, -1);
const __m128i kCstAlpha3 = _mm_set_epi8(12, 8, 4, 0, -1, -1, -1, -1,
-1, -1, -1, -1, -1, -1, -1, -1);
for (j = 0; j < height; ++j) {
const __m128i* src = (const __m128i*)argb;
for (i = 0; i < limit; i += 16) {
// load 64 argb bytes
const __m128i a0 = _mm_loadu_si128(src + 0);
const __m128i a1 = _mm_loadu_si128(src + 1);
const __m128i a2 = _mm_loadu_si128(src + 2);
const __m128i a3 = _mm_loadu_si128(src + 3);
const __m128i b0 = _mm_shuffle_epi8(a0, kCstAlpha0);
const __m128i b1 = _mm_shuffle_epi8(a1, kCstAlpha1);
const __m128i b2 = _mm_shuffle_epi8(a2, kCstAlpha2);
const __m128i b3 = _mm_shuffle_epi8(a3, kCstAlpha3);
const __m128i c0 = _mm_or_si128(b0, b1);
const __m128i c1 = _mm_or_si128(b2, b3);
const __m128i d0 = _mm_or_si128(c0, c1);
// store
_mm_storeu_si128((__m128i*)&alpha[i], d0);
// accumulate sixteen alpha 'and' in parallel
all_alphas = _mm_and_si128(all_alphas, d0);
src += 4;
}
for (; i < width; ++i) {
const uint32_t alpha_value = argb[4 * i];
alpha[i] = alpha_value;
alpha_and &= alpha_value;
}
argb += argb_stride;
alpha += alpha_stride;
}
// Combine the sixteen alpha 'and' into an 8-bit mask.
alpha_and |= 0xff00u; // pretend the upper bits [8..15] were tested ok.
alpha_and &= _mm_movemask_epi8(_mm_cmpeq_epi8(all_alphas, all_0xff));
return (alpha_and == 0xffffu);
}
//------------------------------------------------------------------------------
// Entry point
extern void WebPInitAlphaProcessingSSE41(void);
WEBP_TSAN_IGNORE_FUNCTION void WebPInitAlphaProcessingSSE41(void) {
WebPExtractAlpha = ExtractAlpha;
}
#else // !WEBP_USE_SSE41
WEBP_DSP_INIT_STUB(WebPInitAlphaProcessingSSE41)
#endif // WEBP_USE_SSE41