void qt_blend_argb32_on_argb32_arm_simd(uchar *destPixels, int dbpl,
const uchar *srcPixels, int sbpl,
int w, int h,
int const_alpha)
{
const uint *src = (const uint *) srcPixels;
uint *dst = (uint *) destPixels;
if (const_alpha == 256) {
for (int y=0; y<h; ++y) {
for (int x=0; x<w; ++x) {
uint s = src[x];
if (s >= 0xff000000)
dst[x] = s;
else if (s != 0)
dst[x] = s + BYTE_MUL(dst[x], qAlpha(~s));
}
dst = (quint32 *)(((uchar *) dst) + dbpl);
src = (const quint32 *)(((const uchar *) src) + sbpl);
}
} else if (const_alpha != 0) {
const_alpha = (const_alpha * 255) >> 8;
for (int y=0; y<h; ++y) {
for (int x=0; x<w; ++x) {
uint s = BYTE_MUL(src[x], const_alpha);
dst[x] = s + BYTE_MUL(dst[x], qAlpha(~s));
}
dst = (quint32 *)(((uchar *) dst) + dbpl);
src = (const quint32 *)(((const uchar *) src) + sbpl);
}
}
QT_BEGIN_NAMESPACE
#define toF26Dot6(x) ((int)((x)*64.))
static inline uint sourceOver(uint d, uint color)
{
return color + BYTE_MUL(d, qAlpha(~color));
}
QT_BEGIN_NAMESPACE
inline static void blend_pixel(quint32 &dst, const quint32 src)
{
if (src >= 0xff000000)
dst = src;
else if (src != 0)
dst = src + BYTE_MUL(dst, qAlpha(~src));
}
inline void drawPixelARGB32(QCosmeticStroker *stroker, int x, int y, int coverage)
{
const QRect &cl = stroker->clip;
if (x < cl.x() || x > cl.right() || y < cl.y() || y > cl.bottom())
return;
int offset = x + stroker->ppl*y;
uint c = BYTE_MUL(stroker->color, coverage);
stroker->pixels[offset] = sourceOver(stroker->pixels[offset], c);
}
QT_BEGIN_NAMESPACE
void qt_blend_argb32_on_argb32_mips_dsp(uchar *destPixels, int dbpl,
const uchar *srcPixels, int sbpl,
int w, int h,
int const_alpha)
{
#ifdef QT_DEBUG_DRAW
fprintf(stdout,
"qt_blend_argb32_on_argb32: dst=(%p, %d), src=(%p, %d), dim=(%d, %d) alpha=%d\n",
destPixels, dbpl, srcPixels, sbpl, w, h, const_alpha);
fflush(stdout);
#endif
const uint *src = (const uint *) srcPixels;
uint *dst = (uint *) destPixels;
if (const_alpha == 256) {
for (int y=0; y<h; ++y) {
qt_blend_argb32_on_argb32_const_alpha_256_mips_dsp_asm(dst, src, w);
dst = (quint32 *)(((uchar *) dst) + dbpl);
src = (const quint32 *)(((const uchar *) src) + sbpl);
}
} else if (const_alpha != 0) {
const_alpha = (const_alpha * 255) >> 8;
for (int y=0; y<h; ++y) {
if (h%2 > 0) {
uint s = BYTE_MUL(src[0], const_alpha);
dst[0] = s + BYTE_MUL(dst[0], qAlpha(~s));
h--;
dst++;
src++;
}
qt_blend_argb32_on_argb32_mips_dsp_asm_x2(dst, src, h, const_alpha);
dst = (quint32 *)(((uchar *) dst) + dbpl);
src = (const quint32 *)(((const uchar *) src) + sbpl);
}
}
static inline uint sourceOver(uint d, uint color)
{
return color + BYTE_MUL(d, qAlpha(~color));
}
void qt_blend_argb32_on_argb32_neon(uchar *destPixels, int dbpl,
const uchar *srcPixels, int sbpl,
int w, int h,
int const_alpha)
{
const uint *src = (const uint *) srcPixels;
uint *dst = (uint *) destPixels;
int16x8_t half = vdupq_n_s16(0x80);
int16x8_t full = vdupq_n_s16(0xff);
if (const_alpha == 256) {
for (int y = 0; y < h; ++y) {
int x = 0;
for (; x < w-3; x += 4) {
int32x4_t src32 = vld1q_s32((int32_t *)&src[x]);
if ((src[x] & src[x+1] & src[x+2] & src[x+3]) >= 0xff000000) {
// all opaque
vst1q_s32((int32_t *)&dst[x], src32);
} else if (src[x] | src[x+1] | src[x+2] | src[x+3]) {
int32x4_t dst32 = vld1q_s32((int32_t *)&dst[x]);
const uint8x16_t src8 = vreinterpretq_u8_s32(src32);
const uint8x16_t dst8 = vreinterpretq_u8_s32(dst32);
const uint8x8_t src8_low = vget_low_u8(src8);
const uint8x8_t dst8_low = vget_low_u8(dst8);
const uint8x8_t src8_high = vget_high_u8(src8);
const uint8x8_t dst8_high = vget_high_u8(dst8);
const int16x8_t src16_low = vreinterpretq_s16_u16(vmovl_u8(src8_low));
const int16x8_t dst16_low = vreinterpretq_s16_u16(vmovl_u8(dst8_low));
const int16x8_t src16_high = vreinterpretq_s16_u16(vmovl_u8(src8_high));
const int16x8_t dst16_high = vreinterpretq_s16_u16(vmovl_u8(dst8_high));
const int16x8_t result16_low = qvsource_over_s16(src16_low, dst16_low, half, full);
const int16x8_t result16_high = qvsource_over_s16(src16_high, dst16_high, half, full);
const int32x2_t result32_low = vreinterpret_s32_s8(vmovn_s16(result16_low));
const int32x2_t result32_high = vreinterpret_s32_s8(vmovn_s16(result16_high));
vst1q_s32((int32_t *)&dst[x], vcombine_s32(result32_low, result32_high));
}
}
for (; x<w; ++x) {
uint s = src[x];
if (s >= 0xff000000)
dst[x] = s;
else if (s != 0)
dst[x] = s + BYTE_MUL(dst[x], qAlpha(~s));
}
dst = (quint32 *)(((uchar *) dst) + dbpl);
src = (const quint32 *)(((const uchar *) src) + sbpl);
}
} else if (const_alpha != 0) {
const_alpha = (const_alpha * 255) >> 8;
int16x8_t const_alpha16 = vdupq_n_s16(const_alpha);
for (int y = 0; y < h; ++y) {
int x = 0;
for (; x < w-3; x += 4) {
if (src[x] | src[x+1] | src[x+2] | src[x+3]) {
int32x4_t src32 = vld1q_s32((int32_t *)&src[x]);
int32x4_t dst32 = vld1q_s32((int32_t *)&dst[x]);
const uint8x16_t src8 = vreinterpretq_u8_s32(src32);
const uint8x16_t dst8 = vreinterpretq_u8_s32(dst32);
const uint8x8_t src8_low = vget_low_u8(src8);
const uint8x8_t dst8_low = vget_low_u8(dst8);
const uint8x8_t src8_high = vget_high_u8(src8);
const uint8x8_t dst8_high = vget_high_u8(dst8);
const int16x8_t src16_low = vreinterpretq_s16_u16(vmovl_u8(src8_low));
const int16x8_t dst16_low = vreinterpretq_s16_u16(vmovl_u8(dst8_low));
const int16x8_t src16_high = vreinterpretq_s16_u16(vmovl_u8(src8_high));
const int16x8_t dst16_high = vreinterpretq_s16_u16(vmovl_u8(dst8_high));
const int16x8_t srcalpha16_low = qvbyte_mul_s16(src16_low, const_alpha16, half);
const int16x8_t srcalpha16_high = qvbyte_mul_s16(src16_high, const_alpha16, half);
const int16x8_t result16_low = qvsource_over_s16(srcalpha16_low, dst16_low, half, full);
const int16x8_t result16_high = qvsource_over_s16(srcalpha16_high, dst16_high, half, full);
const int32x2_t result32_low = vreinterpret_s32_s8(vmovn_s16(result16_low));
const int32x2_t result32_high = vreinterpret_s32_s8(vmovn_s16(result16_high));
vst1q_s32((int32_t *)&dst[x], vcombine_s32(result32_low, result32_high));
}
}
for (; x<w; ++x) {
uint s = src[x];
if (s != 0) {
s = BYTE_MUL(s, const_alpha);
dst[x] = s + BYTE_MUL(dst[x], qAlpha(~s));
}
}
dst = (quint32 *)(((uchar *) dst) + dbpl);
src = (const quint32 *)(((const uchar *) src) + sbpl);
}
}
