#include "src/dsp/dsp.h"
#if defined(WEBP_USE_MSA)
#include "src/dsp/lossless.h"
#include "src/dsp/msa_macro.h"
#define CONVERT16_BGRA_XXX …
#define CONVERT12_BGRA_XXX …
#define CONVERT8_BGRA_XXX …
#define CONVERT4_BGRA_XXX …
#define CONVERT1_BGRA_BGR …
#define CONVERT1_BGRA_RGB …
#define TRANSFORM_COLOR_INVERSE_8 …
#define TRANSFORM_COLOR_INVERSE_4 …
static void ConvertBGRAToRGBA_MSA(const uint32_t* src,
int num_pixels, uint8_t* dst) {
int i;
const uint8_t* ptemp_src = (const uint8_t*)src;
uint8_t* ptemp_dst = (uint8_t*)dst;
v16u8 src0, dst0;
const v16u8 mask = { 2, 1, 0, 3, 6, 5, 4, 7, 10, 9, 8, 11, 14, 13, 12, 15 };
while (num_pixels >= 8) {
v16u8 src1, dst1;
LD_UB2(ptemp_src, 16, src0, src1);
VSHF_B2_UB(src0, src0, src1, src1, mask, mask, dst0, dst1);
ST_UB2(dst0, dst1, ptemp_dst, 16);
ptemp_src += 32;
ptemp_dst += 32;
num_pixels -= 8;
}
if (num_pixels > 0) {
if (num_pixels >= 4) {
src0 = LD_UB(ptemp_src);
dst0 = VSHF_UB(src0, src0, mask);
ST_UB(dst0, ptemp_dst);
ptemp_src += 16;
ptemp_dst += 16;
num_pixels -= 4;
}
for (i = 0; i < num_pixels; i++) {
const uint8_t b = ptemp_src[2];
const uint8_t g = ptemp_src[1];
const uint8_t r = ptemp_src[0];
const uint8_t a = ptemp_src[3];
ptemp_dst[0] = b;
ptemp_dst[1] = g;
ptemp_dst[2] = r;
ptemp_dst[3] = a;
ptemp_src += 4;
ptemp_dst += 4;
}
}
}
static void ConvertBGRAToBGR_MSA(const uint32_t* src,
int num_pixels, uint8_t* dst) {
const uint8_t* ptemp_src = (const uint8_t*)src;
uint8_t* ptemp_dst = (uint8_t*)dst;
const v16u8 mask0 = { 0, 1, 2, 4, 5, 6, 8, 9, 10, 12, 13, 14,
16, 17, 18, 20 };
const v16u8 mask1 = { 5, 6, 8, 9, 10, 12, 13, 14, 16, 17, 18, 20,
21, 22, 24, 25 };
const v16u8 mask2 = { 10, 12, 13, 14, 16, 17, 18, 20, 21, 22, 24, 25,
26, 28, 29, 30 };
while (num_pixels >= 16) {
CONVERT16_BGRA_XXX(ptemp_src, ptemp_dst, mask0, mask1, mask2);
ptemp_src += 64;
ptemp_dst += 48;
num_pixels -= 16;
}
if (num_pixels > 0) {
if (num_pixels >= 12) {
CONVERT12_BGRA_XXX(ptemp_src, ptemp_dst, mask0, mask1, mask2);
ptemp_src += 48;
ptemp_dst += 36;
num_pixels -= 12;
} else if (num_pixels >= 8) {
CONVERT8_BGRA_XXX(ptemp_src, ptemp_dst, mask0, mask1);
ptemp_src += 32;
ptemp_dst += 24;
num_pixels -= 8;
} else if (num_pixels >= 4) {
CONVERT4_BGRA_XXX(ptemp_src, ptemp_dst, mask0);
ptemp_src += 16;
ptemp_dst += 12;
num_pixels -= 4;
}
if (num_pixels == 3) {
CONVERT1_BGRA_BGR(ptemp_src + 0, ptemp_dst + 0);
CONVERT1_BGRA_BGR(ptemp_src + 4, ptemp_dst + 3);
CONVERT1_BGRA_BGR(ptemp_src + 8, ptemp_dst + 6);
} else if (num_pixels == 2) {
CONVERT1_BGRA_BGR(ptemp_src + 0, ptemp_dst + 0);
CONVERT1_BGRA_BGR(ptemp_src + 4, ptemp_dst + 3);
} else if (num_pixels == 1) {
CONVERT1_BGRA_BGR(ptemp_src, ptemp_dst);
}
}
}
static void ConvertBGRAToRGB_MSA(const uint32_t* src,
int num_pixels, uint8_t* dst) {
const uint8_t* ptemp_src = (const uint8_t*)src;
uint8_t* ptemp_dst = (uint8_t*)dst;
const v16u8 mask0 = { 2, 1, 0, 6, 5, 4, 10, 9, 8, 14, 13, 12,
18, 17, 16, 22 };
const v16u8 mask1 = { 5, 4, 10, 9, 8, 14, 13, 12, 18, 17, 16, 22,
21, 20, 26, 25 };
const v16u8 mask2 = { 8, 14, 13, 12, 18, 17, 16, 22, 21, 20, 26, 25,
24, 30, 29, 28 };
while (num_pixels >= 16) {
CONVERT16_BGRA_XXX(ptemp_src, ptemp_dst, mask0, mask1, mask2);
ptemp_src += 64;
ptemp_dst += 48;
num_pixels -= 16;
}
if (num_pixels) {
if (num_pixels >= 12) {
CONVERT12_BGRA_XXX(ptemp_src, ptemp_dst, mask0, mask1, mask2);
ptemp_src += 48;
ptemp_dst += 36;
num_pixels -= 12;
} else if (num_pixels >= 8) {
CONVERT8_BGRA_XXX(ptemp_src, ptemp_dst, mask0, mask1);
ptemp_src += 32;
ptemp_dst += 24;
num_pixels -= 8;
} else if (num_pixels >= 4) {
CONVERT4_BGRA_XXX(ptemp_src, ptemp_dst, mask0);
ptemp_src += 16;
ptemp_dst += 12;
num_pixels -= 4;
}
if (num_pixels == 3) {
CONVERT1_BGRA_RGB(ptemp_src + 0, ptemp_dst + 0);
CONVERT1_BGRA_RGB(ptemp_src + 4, ptemp_dst + 3);
CONVERT1_BGRA_RGB(ptemp_src + 8, ptemp_dst + 6);
} else if (num_pixels == 2) {
CONVERT1_BGRA_RGB(ptemp_src + 0, ptemp_dst + 0);
CONVERT1_BGRA_RGB(ptemp_src + 4, ptemp_dst + 3);
} else if (num_pixels == 1) {
CONVERT1_BGRA_RGB(ptemp_src, ptemp_dst);
}
}
}
static void AddGreenToBlueAndRed_MSA(const uint32_t* const src, int num_pixels,
uint32_t* dst) {
int i;
const uint8_t* in = (const uint8_t*)src;
uint8_t* out = (uint8_t*)dst;
v16u8 src0, dst0, tmp0;
const v16u8 mask = { 1, 255, 1, 255, 5, 255, 5, 255, 9, 255, 9, 255,
13, 255, 13, 255 };
while (num_pixels >= 8) {
v16u8 src1, dst1, tmp1;
LD_UB2(in, 16, src0, src1);
VSHF_B2_UB(src0, src1, src1, src0, mask, mask, tmp0, tmp1);
ADD2(src0, tmp0, src1, tmp1, dst0, dst1);
ST_UB2(dst0, dst1, out, 16);
in += 32;
out += 32;
num_pixels -= 8;
}
if (num_pixels > 0) {
if (num_pixels >= 4) {
src0 = LD_UB(in);
tmp0 = VSHF_UB(src0, src0, mask);
dst0 = src0 + tmp0;
ST_UB(dst0, out);
in += 16;
out += 16;
num_pixels -= 4;
}
for (i = 0; i < num_pixels; i++) {
const uint8_t b = in[0];
const uint8_t g = in[1];
const uint8_t r = in[2];
out[0] = (b + g) & 0xff;
out[1] = g;
out[2] = (r + g) & 0xff;
out[4] = in[4];
out += 4;
}
}
}
static void TransformColorInverse_MSA(const VP8LMultipliers* const m,
const uint32_t* src, int num_pixels,
uint32_t* dst) {
v16u8 src0, dst0;
const v16i8 g2br = (v16i8)__msa_fill_w(m->green_to_blue_ |
(m->green_to_red_ << 16));
const v16i8 r2b = (v16i8)__msa_fill_w(m->red_to_blue_);
const v16u8 mask0 = { 1, 255, 1, 255, 5, 255, 5, 255, 9, 255, 9, 255,
13, 255, 13, 255 };
const v16u8 mask1 = { 16, 1, 18, 3, 20, 5, 22, 7, 24, 9, 26, 11,
28, 13, 30, 15 };
while (num_pixels >= 8) {
v16u8 src1, dst1;
LD_UB2(src, 4, src0, src1);
TRANSFORM_COLOR_INVERSE_8(src0, src1, dst0, dst1, g2br, r2b, mask0, mask1);
ST_UB2(dst0, dst1, dst, 4);
src += 8;
dst += 8;
num_pixels -= 8;
}
if (num_pixels > 0) {
if (num_pixels >= 4) {
src0 = LD_UB(src);
TRANSFORM_COLOR_INVERSE_4(src0, dst0, g2br, r2b, mask0, mask1);
ST_UB(dst0, dst);
src += 4;
dst += 4;
num_pixels -= 4;
}
if (num_pixels > 0) {
src0 = LD_UB(src);
TRANSFORM_COLOR_INVERSE_4(src0, dst0, g2br, r2b, mask0, mask1);
if (num_pixels == 3) {
const uint64_t pix_d = __msa_copy_s_d((v2i64)dst0, 0);
const uint32_t pix_w = __msa_copy_s_w((v4i32)dst0, 2);
SD(pix_d, dst + 0);
SW(pix_w, dst + 2);
} else if (num_pixels == 2) {
const uint64_t pix_d = __msa_copy_s_d((v2i64)dst0, 0);
SD(pix_d, dst);
} else {
const uint32_t pix_w = __msa_copy_s_w((v4i32)dst0, 0);
SW(pix_w, dst);
}
}
}
}
extern void VP8LDspInitMSA(void);
WEBP_TSAN_IGNORE_FUNCTION void VP8LDspInitMSA(void) {
VP8LConvertBGRAToRGBA = ConvertBGRAToRGBA_MSA;
VP8LConvertBGRAToBGR = ConvertBGRAToBGR_MSA;
VP8LConvertBGRAToRGB = ConvertBGRAToRGB_MSA;
VP8LAddGreenToBlueAndRed = AddGreenToBlueAndRed_MSA;
VP8LTransformColorInverse = TransformColorInverse_MSA;
}
#else
WEBP_DSP_INIT_STUB(VP8LDspInitMSA)
#endif