Separate simd implementation by files to maintain them easier.
Now we have avx, c, neon version implementation base,
we can add implementations to them.
source_file = [
'tvgSwCommon.h',
+ 'tvgSwRasterC.h',
+ 'tvgSwRasterAvx.h',
+ 'tvgSwRasterNeon.h',
'tvgSwFill.cpp',
'tvgSwImage.cpp',
'tvgSwMath.cpp',
#include "tvgCommon.h"
#include "tvgRender.h"
-#ifdef THORVG_AVX_VECTOR_SUPPORT
- #include <immintrin.h>
-#endif
-
-#ifdef THORVG_NEON_VECTOR_SUPPORT
- #include <arm_neon.h>
-#endif
-
#if 0
#include <sys/time.h>
static double timeStamp()
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
-#include "tvgSwCommon.h"
-#include "tvgRender.h"
#include <float.h>
#include <math.h>
+#include "tvgSwCommon.h"
+#include "tvgRender.h"
+#include "tvgSwRasterC.h"
+#include "tvgSwRasterAvx.h"
+#include "tvgSwRasterNeon.h"
/************************************************************************/
/* Internal Class Implementation */
/************************************************************************/
-static bool _translucentRle(SwSurface* surface, const SwRleData* rle, uint32_t color)
-{
- auto span = rle->spans;
- uint32_t src;
-
- for (uint32_t i = 0; i < rle->size; ++i) {
- auto dst = &surface->buffer[span->y * surface->stride + span->x];
-
-#if defined(THORVG_NEON_VECTOR_SUPPORT)
- uint8x8_t *vDst = (uint8x8_t*) dst;
-#endif
-
- if (span->coverage < 255) src = ALPHA_BLEND(color, span->coverage);
- else src = color;
- auto ialpha = 255 - surface->blender.alpha(src);
-
-#if defined(THORVG_NEON_VECTOR_SUPPORT)
- uint8x8_t vSrc = (uint8x8_t) vdup_n_u32(src);
- uint8x8_t vIalpha = (uint8x8_t) vdup_n_u32(ialpha);
-
- uint32_t iterations = span->len / 2;
- uint32_t left = span->len % 2;
-
- for (uint32_t x = 0; x < iterations; x+=2) {
- vDst[x] = vadd_u8(vSrc, ALPHA_BLEND_NEON(vDst[x], vIalpha));
- }
-
- if (left) {
- dst[span->len] = src + ALPHA_BLEND(dst[span->len], ialpha);
- }
-#else
-
- for (uint32_t x = 0; x < span->len; ++x) {
- dst[x] = src + ALPHA_BLEND(dst[x], ialpha);
- }
-#endif
- ++span;
- }
- return true;
-}
-
-
static bool _translucentRleAlphaMask(SwSurface* surface, const SwRleData* rle, uint32_t color)
{
TVGLOG("SW_ENGINE", "Rle Alpha Mask Composition");
return _translucentRleInvAlphaMask(surface, rle, color);
}
}
- return _translucentRle(surface, rle, color);
+
+#if defined(THORVG_NEON_VECTOR_SUPPORT)
+ return neonRasterTranslucentRle(surface, rle, color);
+#else
+ return cRasterTranslucentRle(surface, rle, color);
+#endif
}
void rasterRGBA32(uint32_t *dst, uint32_t val, uint32_t offset, int32_t len)
{
#if defined(THORVG_AVX_VECTOR_SUPPORT)
- //1. calculate how many iterations we need to cover length
- uint32_t iterations = len / 8;
- uint32_t avxFilled = iterations * 8;
-
- //2. set beginning of the array
- dst += offset;
- __m256i_u* avxDst = (__m256i_u*) dst;
-
- //3. fill octets
- for (uint32_t i = 0; i < iterations; ++i) {
- *avxDst = _mm256_set1_epi32(val);
- avxDst++;
- }
-
- //4. fill leftovers (in first step we have to set pointer to place where avx job is done)
- int32_t leftovers = len - avxFilled;
- dst += avxFilled;
-
- while (leftovers--) *dst++ = val;
+ avxRasterRGBA32(dst, val, offset, len);
#elif defined(THORVG_NEON_VECTOR_SUPPORT)
- uint32_t iterations = len / 4;
- uint32_t neonFilled = iterations * 4;
-
- dst += offset;
- uint32x4_t vectorVal = {val, val, val, val};
-
- for (uint32_t i = 0; i < iterations; ++i) {
- vst1q_u32(dst, vectorVal);
- dst += 4;
- }
-
- int32_t leftovers = len - neonFilled;
- while (leftovers--) *dst++ = val;
+ neonRasterRGBA32(dst, val, offset, len);
#else
- dst += offset;
- while (len--) *dst++ = val;
+ cRasterRGBA32(dst, val, offset, len);
#endif
}
--- /dev/null
+/*
+ * Copyright (c) 2021 Samsung Electronics Co., Ltd. All rights reserved.
+
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifdef THORVG_AVX_VECTOR_SUPPORT
+
+#include <immintrin.h>
+
+
+static inline void avxRasterRGBA32(uint32_t *dst, uint32_t val, uint32_t offset, int32_t len)
+{
+ //1. calculate how many iterations we need to cover length
+ uint32_t iterations = len / 8;
+ uint32_t avxFilled = iterations * 8;
+
+ //2. set beginning of the array
+ dst += offset;
+ __m256i_u* avxDst = (__m256i_u*) dst;
+
+ //3. fill octets
+ for (uint32_t i = 0; i < iterations; ++i) {
+ *avxDst = _mm256_set1_epi32(val);
+ avxDst++;
+ }
+
+ //4. fill leftovers (in first step we have to set pointer to place where avx job is done)
+ int32_t leftovers = len - avxFilled;
+ dst += avxFilled;
+
+ while (leftovers--) *dst++ = val;
+}
+
+#endif
\ No newline at end of file
--- /dev/null
+/*
+ * Copyright (c) 2021 Samsung Electronics Co., Ltd. All rights reserved.
+
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+
+static inline void cRasterRGBA32(uint32_t *dst, uint32_t val, uint32_t offset, int32_t len)
+{
+ dst += offset;
+ while (len--) *dst++ = val;
+}
+
+
+static inline bool cRasterTranslucentRle(SwSurface* surface, const SwRleData* rle, uint32_t color)
+{
+ auto span = rle->spans;
+ uint32_t src;
+
+ for (uint32_t i = 0; i < rle->size; ++i) {
+ auto dst = &surface->buffer[span->y * surface->stride + span->x];
+
+ if (span->coverage < 255) src = ALPHA_BLEND(color, span->coverage);
+ else src = color;
+
+ auto ialpha = 255 - surface->blender.alpha(src);
+
+ for (uint32_t x = 0; x < span->len; ++x)
+ dst[x] = src + ALPHA_BLEND(dst[x], ialpha);
+
+ ++span;
+ }
+ return true;
+}
\ No newline at end of file
--- /dev/null
+/*
+ * Copyright (c) 2021 Samsung Electronics Co., Ltd. All rights reserved.
+
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifdef THORVG_NEON_VECTOR_SUPPORT
+
+#include <arm_neon.h>
+
+
+static inline void neonRasterRGBA32(uint32_t *dst, uint32_t val, uint32_t offset, int32_t len)
+{
+ uint32_t iterations = len / 4;
+ uint32_t neonFilled = iterations * 4;
+
+ dst += offset;
+ uint32x4_t vectorVal = {val, val, val, val};
+
+ for (uint32_t i = 0; i < iterations; ++i) {
+ vst1q_u32(dst, vectorVal);
+ dst += 4;
+ }
+
+ int32_t leftovers = len - neonFilled;
+ while (leftovers--) *dst++ = val;
+}
+
+
+static inline bool neonRasterTranslucentRle(SwSurface* surface, const SwRleData* rle, uint32_t color)
+{
+ auto span = rle->spans;
+ uint32_t src;
+
+ for (uint32_t i = 0; i < rle->size; ++i) {
+ auto dst = &surface->buffer[span->y * surface->stride + span->x];
+
+ uint8x8_t *vDst = (uint8x8_t*) dst;
+
+ if (span->coverage < 255) src = ALPHA_BLEND(color, span->coverage);
+ else src = color;
+ auto ialpha = 255 - surface->blender.alpha(src);
+
+ uint8x8_t vSrc = (uint8x8_t) vdup_n_u32(src);
+ uint8x8_t vIalpha = (uint8x8_t) vdup_n_u32(ialpha);
+
+ uint32_t iterations = span->len / 2;
+ uint32_t left = span->len % 2;
+
+ for (uint32_t x = 0; x < iterations; x+=2) {
+ vDst[x] = vadd_u8(vSrc, ALPHA_BLEND_NEON(vDst[x], vIalpha));
+ }
+
+ if (left) {
+ dst[span->len] = src + ALPHA_BLEND(dst[span->len], ialpha);
+ }
+ ++span;
+ }
+ return true;
+}
+
+
+#endif
\ No newline at end of file
projects / platform / core / graphics / tizenvg.git / commitdiff