git_clone/thorvg
1
0
Fork 0
mirror of https://github.com/thorvg/thorvg.git synced 2025-06-08 13:43:43 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 5

sw common: Added neon translucent rect API implementation.

Browse source 
Changes:
Added neonRasterTranslucentRect implementation. Rendering was tested on
32 lottie fiels. Without neon ~ 18.1 FPS was measured. With neon ~ 20.1
FPS was measured.
This commit is contained in:
Michal Szczecinski 2021-08-12 16:02:40 +02:00 • committed by Hermet Park
parent 52a7a257b2
commit 94a3e92fdd
3 changed files with 60 additions and 12 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

19
src/lib/sw_engine/tvgSwRaster.cpp
View file

@ -120,18 +120,13 @@ static uint32_t _applyBilinearInterpolation(const uint32_t *img, uint32_t w, uin
static bool _translucentRect(SwSurface* surface, const SwBBox& region, uint32_t color) static bool _translucentRect(SwSurface* surface, const SwBBox& region, uint32_t color)
{ {
auto buffer = surface->buffer + (region.min.y * surface->stride) + region.min.x; #if defined(THORVG_AVX_VECTOR_SUPPORT)
auto h = static_cast<uint32_t>(region.max.y - region.min.y); return cRasterTranslucentRect(surface, region, color);
auto w = static_cast<uint32_t>(region.max.x - region.min.x); #elif defined(THORVG_NEON_VECTOR_SUPPORT)
auto ialpha = 255 - surface->blender.alpha(color); return neonRasterTranslucentRect(surface, region, color);
#else
for (uint32_t y = 0; y < h; ++y) { return cRasterTranslucentRect(surface, region, color);
auto dst = &buffer[y * surface->stride]; #endif
for (uint32_t x = 0; x < w; ++x) {
dst[x] = color + ALPHA_BLEND(dst[x], ialpha);
}
}
return true;
} }

16
src/lib/sw_engine/tvgSwRasterC.h
View file

@ -48,3 +48,19 @@ static inline bool cRasterTranslucentRle(SwSurface* surface, const SwRleData* rl
} }
return true; return true;
} }
static inline bool cRasterTranslucentRect(SwSurface* surface, const SwBBox& region, uint32_t color)
{
auto buffer = surface->buffer + (region.min.y * surface->stride) + region.min.x;
auto h = static_cast<uint32_t>(region.max.y - region.min.y);
auto w = static_cast<uint32_t>(region.max.x - region.min.x);
auto ialpha = 255 - surface->blender.alpha(color);
for (uint32_t y = 0; y < h; ++y) {
auto dst = &buffer[y * surface->stride];
for (uint32_t x = 0; x < w; ++x) {
dst[x] = color + ALPHA_BLEND(dst[x], ialpha);
}
}
return true;
}

37
src/lib/sw_engine/tvgSwRasterNeon.h
View file

@ -88,4 +88,41 @@ static inline bool neonRasterTranslucentRle(SwSurface* surface, const SwRleData*
return true; return true;
} }
static inline bool neonRasterTranslucentRect(SwSurface* surface, const SwBBox& region, uint32_t color)
{
uint32_t *buffer = surface->buffer + (region.min.y * surface->stride) + region.min.x;
uint32_t h = (uint32_t)(region.max.y - region.min.y);
uint32_t w = (uint32_t)(region.max.x - region.min.x);
uint32_t ialpha = 255 - surface->blender.alpha(color);
uint8x8_t vColor = (uint8x8_t) vdup_n_u32(color);
uint8x8_t vIalpha = (uint8x8_t) vdup_n_u8((uint8_t) ialpha);
uint8x8_t *vDst = NULL;
for (uint32_t y = 0; y < h; ++y) {
uint32_t align = 0;
uint32_t *dst = &buffer[y * surface->stride];
if ((((uint32_t) dst) & 0x7) != 0) {
vDst = (uint8x8_t*)(dst+1);
align = 1;
} else {
vDst = (uint8x8_t*) dst;
}
uint32_t iterations = w / 2;
uint32_t left = w % 2;
if (align) *dst = color + ALPHA_BLEND(*dst, ialpha);
for (uint32_t x = 0; x < iterations; ++x) vDst[x] = vadd_u8(vColor, ALPHA_BLEND_NEON(vDst[x], vIalpha));
if (left) dst[w] = color + ALPHA_BLEND(dst[w], ialpha);
}
return true;
}
#endif #endif