diff --git a/README.chromium b/README.chromium index 77317170b..4715c7180 100644 --- a/README.chromium +++ b/README.chromium @@ -1,6 +1,6 @@ Name: libyuv URL: http://code.google.com/p/libyuv/ -Version: 1742 +Version: 1743 License: BSD License File: LICENSE diff --git a/include/libyuv/convert.h b/include/libyuv/convert.h index f571142fa..509f034f2 100644 --- a/include/libyuv/convert.h +++ b/include/libyuv/convert.h @@ -418,6 +418,15 @@ int RGB24ToJ400(const uint8_t* src_rgb24, int width, int height); +// RGB big endian (rgb in memory) to J400. +LIBYUV_API +int RAWToJ400(const uint8_t* src_raw, + int src_stride_raw, + uint8_t* dst_yj, + int dst_stride_yj, + int width, + int height); + #ifdef HAVE_JPEG // src_width/height provided by capture. // dst_width/height for clipping determine final size. diff --git a/include/libyuv/row.h b/include/libyuv/row.h index 46afb0717..620f4dbc1 100644 --- a/include/libyuv/row.h +++ b/include/libyuv/row.h @@ -374,9 +374,11 @@ extern "C" { #define HAS_RAWTORGB24ROW_NEON #define HAS_RAWTOUVROW_NEON #define HAS_RAWTOYROW_NEON +#define HAS_RAWTOYJROW_NEON #define HAS_RGB24TOARGBROW_NEON #define HAS_RGB24TOUVROW_NEON #define HAS_RGB24TOYROW_NEON +#define HAS_RGB24TOYJROW_NEON #define HAS_RGB565TOARGBROW_NEON #define HAS_RGB565TOUVROW_NEON #define HAS_RGB565TOYROW_NEON @@ -1140,7 +1142,9 @@ void BGRAToYRow_NEON(const uint8_t* src_bgra, uint8_t* dst_y, int width); void ABGRToYRow_NEON(const uint8_t* src_abgr, uint8_t* dst_y, int width); void RGBAToYRow_NEON(const uint8_t* src_rgba, uint8_t* dst_y, int width); void RGB24ToYRow_NEON(const uint8_t* src_rgb24, uint8_t* dst_y, int width); +void RGB24ToYJRow_NEON(const uint8_t* src_rgb24, uint8_t* dst_yj, int width); void RAWToYRow_NEON(const uint8_t* src_raw, uint8_t* dst_y, int width); +void RAWToYJRow_NEON(const uint8_t* src_raw, uint8_t* dst_yj, int width); void RGB565ToYRow_NEON(const uint8_t* src_rgb565, uint8_t* dst_y, int width); void ARGB1555ToYRow_NEON(const uint8_t* src_argb1555, uint8_t* dst_y, @@ -1171,7 +1175,9 @@ void BGRAToYRow_C(const uint8_t* src_argb, uint8_t* dst_y, int width); void ABGRToYRow_C(const uint8_t* src_argb, uint8_t* dst_y, int width); void RGBAToYRow_C(const uint8_t* src_argb, uint8_t* dst_y, int width); void RGB24ToYRow_C(const uint8_t* src_argb, uint8_t* dst_y, int width); +void RGB24ToYJRow_C(const uint8_t* src_argb, uint8_t* dst_yj, int width); void RAWToYRow_C(const uint8_t* src_argb, uint8_t* dst_y, int width); +void RAWToYJRow_C(const uint8_t* src_argb, uint8_t* dst_yj, int width); void RGB565ToYRow_C(const uint8_t* src_rgb565, uint8_t* dst_y, int width); void ARGB1555ToYRow_C(const uint8_t* src_argb1555, uint8_t* dst_y, int width); void ARGB4444ToYRow_C(const uint8_t* src_argb4444, uint8_t* dst_y, int width); @@ -1190,7 +1196,9 @@ void BGRAToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width); void ABGRToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width); void RGBAToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width); void RGB24ToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width); +void RGB24ToYJRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width); void RAWToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width); +void RAWToYJRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width); void RGB565ToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width); void ARGB1555ToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, diff --git a/include/libyuv/version.h b/include/libyuv/version.h index 9c3bb5aba..24bf8ebb8 100644 --- a/include/libyuv/version.h +++ b/include/libyuv/version.h @@ -11,6 +11,6 @@ #ifndef INCLUDE_LIBYUV_VERSION_H_ #define INCLUDE_LIBYUV_VERSION_H_ -#define LIBYUV_VERSION 1742 +#define LIBYUV_VERSION 1743 #endif // INCLUDE_LIBYUV_VERSION_H_ diff --git a/source/convert.cc b/source/convert.cc index 614fa4824..cf6ebcdc5 100644 --- a/source/convert.cc +++ b/source/convert.cc @@ -1598,8 +1598,9 @@ int RGB24ToJ420(const uint8_t* src_rgb24, int width, int height) { int y; -#if (defined(HAS_RGB24TOYJROW_NEON) || defined(HAS_RGB24TOYJROW_MSA) || \ - defined(HAS_RGB24TOYJROW_MMI)) +#if (defined(HAS_RGB24TOYJROW_NEON) && defined(HAS_RGB24TOUVJROW_NEON)) || \ + defined(HAS_RGB24TOYJROW_MSA) || \ + defined(HAS_RGB24TOYJROW_MMI) void (*RGB24ToUVJRow)(const uint8_t* src_rgb24, int src_stride_rgb24, uint8_t* dst_u, uint8_t* dst_v, int width) = RGB24ToUVJRow_C; @@ -1625,7 +1626,7 @@ int RGB24ToJ420(const uint8_t* src_rgb24, } // Neon version does direct RGB24 to YUV. -#if defined(HAS_RGB24TOYJROW_NEON) +#if defined(HAS_RGB24TOYJROW_NEON) && defined(HAS_RGB24TOUVJROW_NEON) if (TestCpuFlag(kCpuHasNEON)) { RGB24ToUVJRow = RGB24ToUVJRow_Any_NEON; RGB24ToYJRow = RGB24ToYJRow_Any_NEON; @@ -1689,16 +1690,16 @@ int RGB24ToJ420(const uint8_t* src_rgb24, #endif { -#if !(defined(HAS_RGB24TOYJROW_NEON) || defined(HAS_RGB24TOYJROW_MSA) || \ - defined(HAS_RGB24TOYJROW_MMI)) +#if !((defined(HAS_RGB24TOYJROW_NEON) && defined(HAS_RGB24TOUVJROW_NEON)) || \ + defined(HAS_RGB24TOYJROW_MSA) || defined(HAS_RGB24TOYJROW_MMI)) // Allocate 2 rows of ARGB. const int kRowSize = (width * 4 + 31) & ~31; align_buffer_64(row, kRowSize * 2); #endif for (y = 0; y < height - 1; y += 2) { -#if (defined(HAS_RGB24TOYJROW_NEON) || defined(HAS_RGB24TOYJROW_MSA) || \ - defined(HAS_RGB24TOYJROW_MMI)) +#if ((defined(HAS_RGB24TOYJROW_NEON) && defined(HAS_RGB24TOUVJROW_NEON)) || \ + defined(HAS_RGB24TOYJROW_MSA) || defined(HAS_RGB24TOYJROW_MMI)) RGB24ToUVJRow(src_rgb24, src_stride_rgb24, dst_u, dst_v, width); RGB24ToYJRow(src_rgb24, dst_y, width); RGB24ToYJRow(src_rgb24 + src_stride_rgb24, dst_y + dst_stride_y, width); @@ -1715,8 +1716,8 @@ int RGB24ToJ420(const uint8_t* src_rgb24, dst_v += dst_stride_v; } if (height & 1) { -#if (defined(HAS_RGB24TOYJROW_NEON) || defined(HAS_RGB24TOYJROW_MSA) || \ - defined(HAS_RGB24TOYJROW_MMI)) +#if ((defined(HAS_RGB24TOYJROW_NEON) && defined(HAS_RGB24TOUVJROW_NEON)) || \ + defined(HAS_RGB24TOYJROW_MSA) || defined(HAS_RGB24TOYJROW_MMI)) RGB24ToUVJRow(src_rgb24, 0, dst_u, dst_v, width); RGB24ToYJRow(src_rgb24, dst_y, width); #else @@ -1725,8 +1726,8 @@ int RGB24ToJ420(const uint8_t* src_rgb24, ARGBToYJRow(row, dst_y, width); #endif } -#if !(defined(HAS_RGB24TOYJROW_NEON) || defined(HAS_RGB24TOYJROW_MSA) || \ - defined(HAS_RGB24TOYJROW_MMI)) +#if !((defined(HAS_RGB24TOYJROW_NEON) && defined(HAS_RGB24TOUVJROW_NEON)) || \ + defined(HAS_RGB24TOYJROW_MSA) || defined(HAS_RGB24TOYJROW_MMI)) free_aligned_buffer_64(row); #endif } @@ -1746,8 +1747,9 @@ int RAWToI420(const uint8_t* src_raw, int width, int height) { int y; -#if (defined(HAS_RAWTOYROW_NEON) || defined(HAS_RAWTOYROW_MSA) || \ - defined(HAS_RAWTOYROW_MMI)) +#if (defined(HAS_RAWTOYROW_NEON) && defined(HAS_RAWTOUVROW_NEON)) || \ + defined(HAS_RAWTOYROW_MSA) || \ + defined(HAS_RAWTOYROW_MMI) void (*RAWToUVRow)(const uint8_t* src_raw, int src_stride_raw, uint8_t* dst_u, uint8_t* dst_v, int width) = RAWToUVRow_C; void (*RAWToYRow)(const uint8_t* src_raw, uint8_t* dst_y, int width) = @@ -1772,7 +1774,7 @@ int RAWToI420(const uint8_t* src_raw, } // Neon version does direct RAW to YUV. -#if defined(HAS_RAWTOYROW_NEON) +#if defined(HAS_RAWTOYROW_NEON) && defined(HAS_RAWTOUVROW_NEON) if (TestCpuFlag(kCpuHasNEON)) { RAWToUVRow = RAWToUVRow_Any_NEON; RAWToYRow = RAWToYRow_Any_NEON; @@ -2398,7 +2400,7 @@ int RGB24ToJ400(const uint8_t* src_rgb24, src_stride_rgb24 = -src_stride_rgb24; } -// Neon version does direct RGB24 to YUV. +// Neon version does direct RGB24 to YJ. #if defined(HAS_RGB24TOYJROW_NEON) if (TestCpuFlag(kCpuHasNEON)) { RGB24ToYJRow = RGB24ToYJRow_Any_NEON; @@ -2487,6 +2489,124 @@ int RGB24ToJ400(const uint8_t* src_rgb24, return 0; } +// Convert RAW to J400. +LIBYUV_API +int RAWToJ400(const uint8_t* src_raw, + int src_stride_raw, + uint8_t* dst_yj, + int dst_stride_yj, + int width, + int height) { + int y; +#if (defined(HAS_RAWTOYJROW_NEON) || defined(HAS_RAWTOYJROW_MSA) || \ + defined(HAS_RAWTOYJROW_MMI)) + void (*RAWToYJRow)(const uint8_t* src_raw, uint8_t* dst_yj, int width) = + RAWToYJRow_C; +#else + void (*RAWToARGBRow)(const uint8_t* src_rgb, uint8_t* dst_argb, int width) = + RAWToARGBRow_C; + void (*ARGBToYJRow)(const uint8_t* src_argb, uint8_t* dst_yj, int width) = + ARGBToYJRow_C; +#endif + if (!src_raw || !dst_yj || width <= 0 || height == 0) { + return -1; + } + // Negative height means invert the image. + if (height < 0) { + height = -height; + src_raw = src_raw + (height - 1) * src_stride_raw; + src_stride_raw = -src_stride_raw; + } + +// Neon version does direct RAW to YJ. +#if defined(HAS_RAWTOYJROW_NEON) + if (TestCpuFlag(kCpuHasNEON)) { + RAWToYJRow = RAWToYJRow_Any_NEON; + if (IS_ALIGNED(width, 8)) { + RAWToYJRow = RAWToYJRow_NEON; + } + } +#elif defined(HAS_RAWTOYJROW_MSA) + if (TestCpuFlag(kCpuHasMSA)) { + RAWToYJRow = RAWToYJRow_Any_MSA; + if (IS_ALIGNED(width, 16)) { + RAWToYJRow = RAWToYJRow_MSA; + } + } +#elif defined(HAS_RAWTOYJROW_MMI) + if (TestCpuFlag(kCpuHasMMI)) { + RAWToYJRow = RAWToYJRow_Any_MMI; + if (IS_ALIGNED(width, 8)) { + RAWToYJRow = RAWToYJRow_MMI; + } + } +// Other platforms do intermediate conversion from RAW to ARGB. +#else +#if defined(HAS_RAWTOARGBROW_SSSE3) + if (TestCpuFlag(kCpuHasSSSE3)) { + RAWToARGBRow = RAWToARGBRow_Any_SSSE3; + if (IS_ALIGNED(width, 16)) { + RAWToARGBRow = RAWToARGBRow_SSSE3; + } + } +#endif +#if defined(HAS_ARGBTOYJROW_SSSE3) + if (TestCpuFlag(kCpuHasSSSE3)) { + ARGBToYJRow = ARGBToYJRow_Any_SSSE3; + if (IS_ALIGNED(width, 16)) { + ARGBToYJRow = ARGBToYJRow_SSSE3; + } + } +#endif +#if defined(HAS_ARGBTOYJROW_AVX2) + if (TestCpuFlag(kCpuHasAVX2)) { + ARGBToYJRow = ARGBToYJRow_Any_AVX2; + if (IS_ALIGNED(width, 32)) { + ARGBToYJRow = ARGBToYJRow_AVX2; + } + } +#endif +#endif + + { +#if !(defined(HAS_RAWTOYJROW_NEON) || defined(HAS_RAWTOYJROW_MSA) || \ + defined(HAS_RAWTOYJROW_MMI)) + // Allocate 2 rows of ARGB. + const int kRowSize = (width * 4 + 31) & ~31; + align_buffer_64(row, kRowSize * 2); +#endif + + for (y = 0; y < height - 1; y += 2) { +#if (defined(HAS_RAWTOYJROW_NEON) || defined(HAS_RAWTOYJROW_MSA) || \ + defined(HAS_RAWTOYJROW_MMI)) + RAWToYJRow(src_raw, dst_yj, width); + RAWToYJRow(src_raw + src_stride_raw, dst_yj + dst_stride_yj, width); +#else + RAWToARGBRow(src_raw, row, width); + RAWToARGBRow(src_raw + src_stride_raw, row + kRowSize, width); + ARGBToYJRow(row, dst_yj, width); + ARGBToYJRow(row + kRowSize, dst_yj + dst_stride_yj, width); +#endif + src_raw += src_stride_raw * 2; + dst_yj += dst_stride_yj * 2; + } + if (height & 1) { +#if (defined(HAS_RAWTOYJROW_NEON) || defined(HAS_RAWTOYJROW_MSA) || \ + defined(HAS_RAWTOYJROW_MMI)) + RAWToYJRow(src_raw, dst_yj, width); +#else + RAWToARGBRow(src_raw, row, width); + ARGBToYJRow(row, dst_yj, width); +#endif + } +#if !(defined(HAS_RAWTOYJROW_NEON) || defined(HAS_RAWTOYJROW_MSA) || \ + defined(HAS_RAWTOYJROW_MMI)) + free_aligned_buffer_64(row); +#endif + } + return 0; +} + static void SplitPixels(const uint8_t* src_u, int src_pixel_stride_uv, uint8_t* dst_u, diff --git a/source/row_any.cc b/source/row_any.cc index 3592ffb6c..8f4b06a30 100644 --- a/source/row_any.cc +++ b/source/row_any.cc @@ -695,6 +695,9 @@ ANY11(RGBAToYRow_Any_MMI, RGBAToYRow_MMI, 0, 4, 1, 7) #ifdef HAS_RGB24TOYROW_NEON ANY11(RGB24ToYRow_Any_NEON, RGB24ToYRow_NEON, 0, 3, 1, 7) #endif +#ifdef HAS_RGB24TOYJROW_NEON +ANY11(RGB24ToYJRow_Any_NEON, RGB24ToYJRow_NEON, 0, 3, 1, 7) +#endif #ifdef HAS_RGB24TOYROW_MSA ANY11(RGB24ToYRow_Any_MSA, RGB24ToYRow_MSA, 0, 3, 1, 15) #endif @@ -704,6 +707,9 @@ ANY11(RGB24ToYRow_Any_MMI, RGB24ToYRow_MMI, 0, 3, 1, 7) #ifdef HAS_RAWTOYROW_NEON ANY11(RAWToYRow_Any_NEON, RAWToYRow_NEON, 0, 3, 1, 7) #endif +#ifdef HAS_RAWTOYJROW_NEON +ANY11(RAWToYJRow_Any_NEON, RAWToYJRow_NEON, 0, 3, 1, 7) +#endif #ifdef HAS_RAWTOYROW_MSA ANY11(RAWToYRow_Any_MSA, RAWToYRow_MSA, 0, 3, 1, 15) #endif diff --git a/source/row_common.cc b/source/row_common.cc index 15f00e886..e6a5e80d7 100644 --- a/source/row_common.cc +++ b/source/row_common.cc @@ -565,6 +565,8 @@ static __inline int RGBToVJ(uint8_t r, uint8_t g, uint8_t b) { MAKEROWYJ(ARGB, 2, 1, 0, 4) MAKEROWYJ(RGBA, 3, 2, 1, 4) +MAKEROWYJ(RGB24, 2, 1, 0, 3) +MAKEROWYJ(RAW, 0, 1, 2, 3) #undef MAKEROWYJ void RGB565ToYRow_C(const uint8_t* src_rgb565, uint8_t* dst_y, int width) { diff --git a/source/row_mmi.cc b/source/row_mmi.cc index d7d34e47f..50cfca726 100644 --- a/source/row_mmi.cc +++ b/source/row_mmi.cc @@ -6040,93 +6040,90 @@ void I444ToARGBRow_MMI(const uint8_t* src_y, uint8_t* rgb_buf, const struct YuvConstants* yuvconstants, int width) { - uint64_t y,u,v; - uint64_t b_vec[2],g_vec[2],r_vec[2]; + uint64_t y, u, v; + uint64_t b_vec[2], g_vec[2], r_vec[2]; uint64_t mask = 0xff00ff00ff00ff00ULL; - uint64_t ub,ug,vg,vr,bb,bg,br,yg; - __asm__ volatile ( - "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t"//yg - "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t"//bb - "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t"//ub - "or %[ub], %[ub], %[mask] \n\t"//must sign extension - "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t"//bg - "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t"//ug - "punpcklbh %[ug], %[ug], %[zero] \n\t" - "pshufh %[ug], %[ug], %[zero] \n\t" - "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t"//vg - "punpcklbh %[vg], %[vg], %[zero] \n\t" - "pshufh %[vg], %[vg], %[five] \n\t" - "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t"//br - "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t"//vr - "punpcklbh %[vr], %[vr], %[zero] \n\t" - "pshufh %[vr], %[vr], %[five] \n\t" - "or %[vr], %[vr], %[mask] \n\t"//sign extension + uint64_t ub, ug, vg, vr, bb, bg, br, yg; + __asm__ volatile( + "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" // yg + "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" // bb + "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" // ub + "or %[ub], %[ub], %[mask] \n\t" // must + // sign + // extension + "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" // bg + "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" // ug + "punpcklbh %[ug], %[ug], %[zero] \n\t" + "pshufh %[ug], %[ug], %[zero] \n\t" + "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" // vg + "punpcklbh %[vg], %[vg], %[zero] \n\t" + "pshufh %[vg], %[vg], %[five] \n\t" + "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" // br + "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" // vr + "punpcklbh %[vr], %[vr], %[zero] \n\t" + "pshufh %[vr], %[vr], %[five] \n\t" + "or %[vr], %[vr], %[mask] \n\t" // sign + // extension - "1: \n\t" - "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" - "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" - "gslwlc1 %[u], 0x03(%[u_ptr]) \n\t" - "gslwrc1 %[u], 0x00(%[u_ptr]) \n\t" - "gslwlc1 %[v], 0x03(%[v_ptr]) \n\t" - "gslwrc1 %[v], 0x00(%[v_ptr]) \n\t" + "1: \n\t" + "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" + "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" + "gslwlc1 %[u], 0x03(%[u_ptr]) \n\t" + "gslwrc1 %[u], 0x00(%[u_ptr]) \n\t" + "gslwlc1 %[v], 0x03(%[v_ptr]) \n\t" + "gslwrc1 %[v], 0x00(%[v_ptr]) \n\t" - "punpcklbh %[y], %[y], %[y] \n\t"//y*0x0101 - "pmulhuh %[y], %[y], %[yg] \n\t"//y1 + "punpcklbh %[y], %[y], %[y] \n\t" // y*0x0101 + "pmulhuh %[y], %[y], %[yg] \n\t" // y1 - "punpcklbh %[u], %[u], %[zero] \n\t"//u - "paddsh %[b_vec0], %[y], %[bb] \n\t" - "pmullh %[b_vec1], %[u], %[ub] \n\t" - "psubsh %[b_vec0], %[b_vec0], %[b_vec1] \n\t" - "psrah %[b_vec0], %[b_vec0], %[six] \n\t" + "punpcklbh %[u], %[u], %[zero] \n\t" // u + "paddsh %[b_vec0], %[y], %[bb] \n\t" + "pmullh %[b_vec1], %[u], %[ub] \n\t" + "psubsh %[b_vec0], %[b_vec0], %[b_vec1] \n\t" + "psrah %[b_vec0], %[b_vec0], %[six] \n\t" - "punpcklbh %[v], %[v], %[zero] \n\t"//v - "paddsh %[g_vec0], %[y], %[bg] \n\t" - "pmullh %[g_vec1], %[u], %[ug] \n\t"//u*ug - "psubsh %[g_vec0], %[g_vec0], %[g_vec1] \n\t" - "pmullh %[g_vec1], %[v], %[vg] \n\t"//v*vg - "psubsh %[g_vec0], %[g_vec0], %[g_vec1] \n\t" - "psrah %[g_vec0], %[g_vec0], %[six] \n\t" + "punpcklbh %[v], %[v], %[zero] \n\t" // v + "paddsh %[g_vec0], %[y], %[bg] \n\t" + "pmullh %[g_vec1], %[u], %[ug] \n\t" // u*ug + "psubsh %[g_vec0], %[g_vec0], %[g_vec1] \n\t" + "pmullh %[g_vec1], %[v], %[vg] \n\t" // v*vg + "psubsh %[g_vec0], %[g_vec0], %[g_vec1] \n\t" + "psrah %[g_vec0], %[g_vec0], %[six] \n\t" - "paddsh %[r_vec0], %[y], %[br] \n\t" - "pmullh %[r_vec1], %[v], %[vr] \n\t"//v*vr - "psubsh %[r_vec0], %[r_vec0], %[r_vec1] \n\t" - "psrah %[r_vec0], %[r_vec0], %[six] \n\t" + "paddsh %[r_vec0], %[y], %[br] \n\t" + "pmullh %[r_vec1], %[v], %[vr] \n\t" // v*vr + "psubsh %[r_vec0], %[r_vec0], %[r_vec1] \n\t" + "psrah %[r_vec0], %[r_vec0], %[six] \n\t" - "packushb %[r_vec0], %[b_vec0], %[r_vec0] \n\t"//rrrrbbbb - "packushb %[g_vec0], %[g_vec0], %[alpha] \n\t"//ffffgggg - "punpcklwd %[g_vec0], %[g_vec0], %[alpha] \n\t" - "punpcklbh %[b_vec0], %[r_vec0], %[g_vec0] \n\t"//gbgbgbgb - "punpckhbh %[r_vec0], %[r_vec0], %[g_vec0] \n\t"//frfrfrfr - "punpcklhw %[g_vec0], %[b_vec0], %[r_vec0] \n\t"//frgbfrgb - "punpckhhw %[g_vec1], %[b_vec0], %[r_vec0] \n\t"//frgbfrgb - "gssdlc1 %[g_vec0], 0x07(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[g_vec0], 0x00(%[rgbbuf_ptr]) \n\t" - "gssdlc1 %[g_vec1], 0x0f(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[g_vec1], 0x08(%[rgbbuf_ptr]) \n\t" + "packushb %[r_vec0], %[b_vec0], %[r_vec0] \n\t" // rrrrbbbb + "packushb %[g_vec0], %[g_vec0], %[alpha] \n\t" // ffffgggg + "punpcklwd %[g_vec0], %[g_vec0], %[alpha] \n\t" + "punpcklbh %[b_vec0], %[r_vec0], %[g_vec0] \n\t" // gbgbgbgb + "punpckhbh %[r_vec0], %[r_vec0], %[g_vec0] \n\t" // frfrfrfr + "punpcklhw %[g_vec0], %[b_vec0], %[r_vec0] \n\t" // frgbfrgb + "punpckhhw %[g_vec1], %[b_vec0], %[r_vec0] \n\t" // frgbfrgb + "gssdlc1 %[g_vec0], 0x07(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[g_vec0], 0x00(%[rgbbuf_ptr]) \n\t" + "gssdlc1 %[g_vec1], 0x0f(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[g_vec1], 0x08(%[rgbbuf_ptr]) \n\t" - "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" - "daddiu %[u_ptr], %[u_ptr], 0x04 \n\t" - "daddiu %[v_ptr], %[v_ptr], 0x04 \n\t" - "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x10 \n\t" - "daddi %[width], %[width], -0x04 \n\t" - "bnez %[width], 1b \n\t" - : [y]"=&f"(y), - [u]"=&f"(u), [v]"=&f"(v), - [b_vec0]"=&f"(b_vec[0]), [b_vec1]"=&f"(b_vec[1]), - [g_vec0]"=&f"(g_vec[0]), [g_vec1]"=&f"(g_vec[1]), - [r_vec0]"=&f"(r_vec[0]), [r_vec1]"=&f"(r_vec[1]), - [ub]"=&f"(ub), [ug]"=&f"(ug), - [vg]"=&f"(vg), [vr]"=&f"(vr), - [bb]"=&f"(bb), [bg]"=&f"(bg), - [br]"=&f"(br), [yg]"=&f"(yg) - : [y_ptr]"r"(src_y), [u_ptr]"r"(src_u), - [v_ptr]"r"(src_v), [rgbbuf_ptr]"r"(rgb_buf), - [yuvcons_ptr]"r"(yuvconstants), [width]"r"(width), - [zero]"f"(0x00), [alpha]"f"(-1), - [six]"f"(0x6), [five]"f"(0x55), - [mask]"f"(mask) - : "memory" - ); + "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" + "daddiu %[u_ptr], %[u_ptr], 0x04 \n\t" + "daddiu %[v_ptr], %[v_ptr], 0x04 \n\t" + "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x10 \n\t" + "daddi %[width], %[width], -0x04 \n\t" + "bnez %[width], 1b \n\t" + : [y] "=&f"(y), [u] "=&f"(u), [v] "=&f"(v), [b_vec0] "=&f"(b_vec[0]), + [b_vec1] "=&f"(b_vec[1]), [g_vec0] "=&f"(g_vec[0]), + [g_vec1] "=&f"(g_vec[1]), [r_vec0] "=&f"(r_vec[0]), + [r_vec1] "=&f"(r_vec[1]), [ub] "=&f"(ub), [ug] "=&f"(ug), + [vg] "=&f"(vg), [vr] "=&f"(vr), [bb] "=&f"(bb), [bg] "=&f"(bg), + [br] "=&f"(br), [yg] "=&f"(yg) + : [y_ptr] "r"(src_y), [u_ptr] "r"(src_u), [v_ptr] "r"(src_v), + [rgbbuf_ptr] "r"(rgb_buf), [yuvcons_ptr] "r"(yuvconstants), + [width] "r"(width), [zero] "f"(0x00), [alpha] "f"(-1), [six] "f"(0x6), + [five] "f"(0x55), [mask] "f"(mask) + : "memory"); } // Also used for 420 @@ -6136,99 +6133,96 @@ void I422ToARGBRow_MMI(const uint8_t* src_y, uint8_t* rgb_buf, const struct YuvConstants* yuvconstants, int width) { - uint64_t y,u,v; - uint64_t b_vec[2],g_vec[2],r_vec[2]; + uint64_t y, u, v; + uint64_t b_vec[2], g_vec[2], r_vec[2]; uint64_t mask = 0xff00ff00ff00ff00ULL; - uint64_t ub,ug,vg,vr,bb,bg,br,yg; + uint64_t ub, ug, vg, vr, bb, bg, br, yg; __asm__ volatile( - "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t"//yg - "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t"//bb - "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t"//ub - "or %[ub], %[ub], %[mask] \n\t"//must sign extension - "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t"//bg - "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t"//ug - "punpcklbh %[ug], %[ug], %[zero] \n\t" - "pshufh %[ug], %[ug], %[zero] \n\t" - "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t"//vg - "punpcklbh %[vg], %[vg], %[zero] \n\t" - "pshufh %[vg], %[vg], %[five] \n\t" - "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t"//br - "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t"//vr - "punpcklbh %[vr], %[vr], %[zero] \n\t" - "pshufh %[vr], %[vr], %[five] \n\t" - "or %[vr], %[vr], %[mask] \n\t"//sign extension + "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" // yg + "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" // bb + "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" // ub + "or %[ub], %[ub], %[mask] \n\t" // must + // sign + // extension + "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" // bg + "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" // ug + "punpcklbh %[ug], %[ug], %[zero] \n\t" + "pshufh %[ug], %[ug], %[zero] \n\t" + "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" // vg + "punpcklbh %[vg], %[vg], %[zero] \n\t" + "pshufh %[vg], %[vg], %[five] \n\t" + "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" // br + "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" // vr + "punpcklbh %[vr], %[vr], %[zero] \n\t" + "pshufh %[vr], %[vr], %[five] \n\t" + "or %[vr], %[vr], %[mask] \n\t" // sign + // extension - "1: \n\t" - "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" - "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" - "gslwlc1 %[u], 0x03(%[u_ptr]) \n\t" - "gslwrc1 %[u], 0x00(%[u_ptr]) \n\t" - "gslwlc1 %[v], 0x03(%[v_ptr]) \n\t" - "gslwrc1 %[v], 0x00(%[v_ptr]) \n\t" + "1: \n\t" + "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" + "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" + "gslwlc1 %[u], 0x03(%[u_ptr]) \n\t" + "gslwrc1 %[u], 0x00(%[u_ptr]) \n\t" + "gslwlc1 %[v], 0x03(%[v_ptr]) \n\t" + "gslwrc1 %[v], 0x00(%[v_ptr]) \n\t" - "punpcklbh %[y], %[y], %[y] \n\t"//y*0x0101 - "pmulhuh %[y], %[y], %[yg] \n\t"//y1 + "punpcklbh %[y], %[y], %[y] \n\t" // y*0x0101 + "pmulhuh %[y], %[y], %[yg] \n\t" // y1 - //u3|u2|u1|u0 --> u1|u1|u0|u0 - "punpcklbh %[u], %[u], %[u] \n\t"//u - "punpcklbh %[u], %[u], %[zero] \n\t" - "paddsh %[b_vec0], %[y], %[bb] \n\t" - "pmullh %[b_vec1], %[u], %[ub] \n\t" - "psubsh %[b_vec0], %[b_vec0], %[b_vec1] \n\t" - "psrah %[b_vec0], %[b_vec0], %[six] \n\t" + // u3|u2|u1|u0 --> u1|u1|u0|u0 + "punpcklbh %[u], %[u], %[u] \n\t" // u + "punpcklbh %[u], %[u], %[zero] \n\t" + "paddsh %[b_vec0], %[y], %[bb] \n\t" + "pmullh %[b_vec1], %[u], %[ub] \n\t" + "psubsh %[b_vec0], %[b_vec0], %[b_vec1] \n\t" + "psrah %[b_vec0], %[b_vec0], %[six] \n\t" - //v3|v2|v1|v0 --> v1|v1|v0|v0 - "punpcklbh %[v], %[v], %[v] \n\t"//v - "punpcklbh %[v], %[v], %[zero] \n\t" - "paddsh %[g_vec0], %[y], %[bg] \n\t" - "pmullh %[g_vec1], %[u], %[ug] \n\t"//u*ug - "psubsh %[g_vec0], %[g_vec0], %[g_vec1] \n\t" - "pmullh %[g_vec1], %[v], %[vg] \n\t"//v*vg - "psubsh %[g_vec0], %[g_vec0], %[g_vec1] \n\t" - "psrah %[g_vec0], %[g_vec0], %[six] \n\t" + // v3|v2|v1|v0 --> v1|v1|v0|v0 + "punpcklbh %[v], %[v], %[v] \n\t" // v + "punpcklbh %[v], %[v], %[zero] \n\t" + "paddsh %[g_vec0], %[y], %[bg] \n\t" + "pmullh %[g_vec1], %[u], %[ug] \n\t" // u*ug + "psubsh %[g_vec0], %[g_vec0], %[g_vec1] \n\t" + "pmullh %[g_vec1], %[v], %[vg] \n\t" // v*vg + "psubsh %[g_vec0], %[g_vec0], %[g_vec1] \n\t" + "psrah %[g_vec0], %[g_vec0], %[six] \n\t" - "paddsh %[r_vec0], %[y], %[br] \n\t" - "pmullh %[r_vec1], %[v], %[vr] \n\t"//v*vr - "psubsh %[r_vec0], %[r_vec0], %[r_vec1] \n\t" - "psrah %[r_vec0], %[r_vec0], %[six] \n\t" + "paddsh %[r_vec0], %[y], %[br] \n\t" + "pmullh %[r_vec1], %[v], %[vr] \n\t" // v*vr + "psubsh %[r_vec0], %[r_vec0], %[r_vec1] \n\t" + "psrah %[r_vec0], %[r_vec0], %[six] \n\t" - "packushb %[r_vec0], %[b_vec0], %[r_vec0] \n\t"//rrrrbbbb - "packushb %[g_vec0], %[g_vec0], %[alpha] \n\t"//ffffgggg - "punpcklwd %[g_vec0], %[g_vec0], %[alpha] \n\t" - "punpcklbh %[b_vec0], %[r_vec0], %[g_vec0] \n\t"//gbgbgbgb - "punpckhbh %[r_vec0], %[r_vec0], %[g_vec0] \n\t"//frfrfrfr - "punpcklhw %[g_vec0], %[b_vec0], %[r_vec0] \n\t"//frgbfrgb - "punpckhhw %[g_vec1], %[b_vec0], %[r_vec0] \n\t"//frgbfrgb - "gssdlc1 %[g_vec0], 0x07(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[g_vec0], 0x00(%[rgbbuf_ptr]) \n\t" - "gssdlc1 %[g_vec1], 0x0f(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[g_vec1], 0x08(%[rgbbuf_ptr]) \n\t" + "packushb %[r_vec0], %[b_vec0], %[r_vec0] \n\t" // rrrrbbbb + "packushb %[g_vec0], %[g_vec0], %[alpha] \n\t" // ffffgggg + "punpcklwd %[g_vec0], %[g_vec0], %[alpha] \n\t" + "punpcklbh %[b_vec0], %[r_vec0], %[g_vec0] \n\t" // gbgbgbgb + "punpckhbh %[r_vec0], %[r_vec0], %[g_vec0] \n\t" // frfrfrfr + "punpcklhw %[g_vec0], %[b_vec0], %[r_vec0] \n\t" // frgbfrgb + "punpckhhw %[g_vec1], %[b_vec0], %[r_vec0] \n\t" // frgbfrgb + "gssdlc1 %[g_vec0], 0x07(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[g_vec0], 0x00(%[rgbbuf_ptr]) \n\t" + "gssdlc1 %[g_vec1], 0x0f(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[g_vec1], 0x08(%[rgbbuf_ptr]) \n\t" - "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" - "daddiu %[u_ptr], %[u_ptr], 0x02 \n\t" - "daddiu %[v_ptr], %[v_ptr], 0x02 \n\t" - "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x10 \n\t" - "daddi %[width], %[width], -0x04 \n\t" - "bnez %[width], 1b \n\t" + "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" + "daddiu %[u_ptr], %[u_ptr], 0x02 \n\t" + "daddiu %[v_ptr], %[v_ptr], 0x02 \n\t" + "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x10 \n\t" + "daddi %[width], %[width], -0x04 \n\t" + "bnez %[width], 1b \n\t" - : [y]"=&f"(y), - [u]"=&f"(u), [v]"=&f"(v), - [b_vec0]"=&f"(b_vec[0]), [b_vec1]"=&f"(b_vec[1]), - [g_vec0]"=&f"(g_vec[0]), [g_vec1]"=&f"(g_vec[1]), - [r_vec0]"=&f"(r_vec[0]), [r_vec1]"=&f"(r_vec[1]), - [ub]"=&f"(ub), [ug]"=&f"(ug), - [vg]"=&f"(vg), [vr]"=&f"(vr), - [bb]"=&f"(bb), [bg]"=&f"(bg), - [br]"=&f"(br), [yg]"=&f"(yg) - : [y_ptr]"r"(src_y), [u_ptr]"r"(src_u), - [v_ptr]"r"(src_v), [rgbbuf_ptr]"r"(rgb_buf), - [yuvcons_ptr]"r"(yuvconstants), [width]"r"(width), - [zero]"f"(0x00), [alpha]"f"(-1), - [six]"f"(0x6), [five]"f"(0x55), - [mask]"f"(mask) - : "memory" - ); + : [y] "=&f"(y), [u] "=&f"(u), [v] "=&f"(v), [b_vec0] "=&f"(b_vec[0]), + [b_vec1] "=&f"(b_vec[1]), [g_vec0] "=&f"(g_vec[0]), + [g_vec1] "=&f"(g_vec[1]), [r_vec0] "=&f"(r_vec[0]), + [r_vec1] "=&f"(r_vec[1]), [ub] "=&f"(ub), [ug] "=&f"(ug), + [vg] "=&f"(vg), [vr] "=&f"(vr), [bb] "=&f"(bb), [bg] "=&f"(bg), + [br] "=&f"(br), [yg] "=&f"(yg) + : [y_ptr] "r"(src_y), [u_ptr] "r"(src_u), [v_ptr] "r"(src_v), + [rgbbuf_ptr] "r"(rgb_buf), [yuvcons_ptr] "r"(yuvconstants), + [width] "r"(width), [zero] "f"(0x00), [alpha] "f"(-1), [six] "f"(0x6), + [five] "f"(0x55), [mask] "f"(mask) + : "memory"); } // 10 bit YUV to ARGB @@ -6238,102 +6232,96 @@ void I210ToARGBRow_MMI(const uint16_t* src_y, uint8_t* rgb_buf, const struct YuvConstants* yuvconstants, int width) { - uint64_t y,u,v; - uint64_t b_vec[2],g_vec[2],r_vec[2]; + uint64_t y, u, v; + uint64_t b_vec[2], g_vec[2], r_vec[2]; uint64_t mask = 0xff00ff00ff00ff00ULL; - uint64_t ub,ug,vg,vr,bb,bg,br,yg; + uint64_t ub, ug, vg, vr, bb, bg, br, yg; __asm__ volatile( - "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" - "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" - "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" - "or %[ub], %[ub], %[mask] \n\t" - "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" - "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[ug], %[ug], %[zero] \n\t" - "pshufh %[ug], %[ug], %[zero] \n\t" - "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vg], %[vg], %[zero] \n\t" - "pshufh %[vg], %[vg], %[five] \n\t" - "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" - "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vr], %[vr], %[zero] \n\t" - "pshufh %[vr], %[vr], %[five] \n\t" - "or %[vr], %[vr], %[mask] \n\t" + "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" + "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" + "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" + "or %[ub], %[ub], %[mask] \n\t" + "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" + "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[ug], %[ug], %[zero] \n\t" + "pshufh %[ug], %[ug], %[zero] \n\t" + "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vg], %[vg], %[zero] \n\t" + "pshufh %[vg], %[vg], %[five] \n\t" + "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" + "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vr], %[vr], %[zero] \n\t" + "pshufh %[vr], %[vr], %[five] \n\t" + "or %[vr], %[vr], %[mask] \n\t" - "1: \n\t" - "gsldlc1 %[y], 0x07(%[y_ptr]) \n\t" - "gsldrc1 %[y], 0x00(%[y_ptr]) \n\t" - "gslwlc1 %[u], 0x03(%[u_ptr]) \n\t" - "gslwrc1 %[u], 0x00(%[u_ptr]) \n\t" - "gslwlc1 %[v], 0x03(%[v_ptr]) \n\t" - "gslwrc1 %[v], 0x00(%[v_ptr]) \n\t" + "1: \n\t" + "gsldlc1 %[y], 0x07(%[y_ptr]) \n\t" + "gsldrc1 %[y], 0x00(%[y_ptr]) \n\t" + "gslwlc1 %[u], 0x03(%[u_ptr]) \n\t" + "gslwrc1 %[u], 0x00(%[u_ptr]) \n\t" + "gslwlc1 %[v], 0x03(%[v_ptr]) \n\t" + "gslwrc1 %[v], 0x00(%[v_ptr]) \n\t" - "psllh %[y], %[y], %[six] \n\t" - "pmulhuh %[y], %[y], %[yg] \n\t" + "psllh %[y], %[y], %[six] \n\t" + "pmulhuh %[y], %[y], %[yg] \n\t" - "punpcklhw %[u], %[u], %[u] \n\t" - "psrah %[u], %[u], %[two] \n\t" - "punpcklhw %[v], %[v], %[v] \n\t" - "psrah %[v], %[v], %[two] \n\t" - "pminsh %[u], %[u], %[mask1] \n\t" - "pminsh %[v], %[v], %[mask1] \n\t" + "punpcklhw %[u], %[u], %[u] \n\t" + "psrah %[u], %[u], %[two] \n\t" + "punpcklhw %[v], %[v], %[v] \n\t" + "psrah %[v], %[v], %[two] \n\t" + "pminsh %[u], %[u], %[mask1] \n\t" + "pminsh %[v], %[v], %[mask1] \n\t" - "paddsh %[b_vec0], %[y], %[bb] \n\t" - "pmullh %[b_vec1], %[u], %[ub] \n\t" - "psubsh %[b_vec0], %[b_vec0], %[b_vec1] \n\t" + "paddsh %[b_vec0], %[y], %[bb] \n\t" + "pmullh %[b_vec1], %[u], %[ub] \n\t" + "psubsh %[b_vec0], %[b_vec0], %[b_vec1] \n\t" - "paddsh %[g_vec0], %[y], %[bg] \n\t" - "pmullh %[g_vec1], %[u], %[ug] \n\t" - "psubsh %[g_vec0], %[g_vec0], %[g_vec1] \n\t" - "pmullh %[g_vec1], %[v], %[vg] \n\t" - "psubsh %[g_vec0], %[g_vec0], %[g_vec1] \n\t" + "paddsh %[g_vec0], %[y], %[bg] \n\t" + "pmullh %[g_vec1], %[u], %[ug] \n\t" + "psubsh %[g_vec0], %[g_vec0], %[g_vec1] \n\t" + "pmullh %[g_vec1], %[v], %[vg] \n\t" + "psubsh %[g_vec0], %[g_vec0], %[g_vec1] \n\t" - "paddsh %[r_vec0], %[y], %[br] \n\t" - "pmullh %[r_vec1], %[v], %[vr] \n\t" - "psubsh %[r_vec0], %[r_vec0], %[r_vec1] \n\t" + "paddsh %[r_vec0], %[y], %[br] \n\t" + "pmullh %[r_vec1], %[v], %[vr] \n\t" + "psubsh %[r_vec0], %[r_vec0], %[r_vec1] \n\t" - "psrah %[b_vec0], %[b_vec0], %[six] \n\t" - "psrah %[g_vec0], %[g_vec0], %[six] \n\t" - "psrah %[r_vec0], %[r_vec0], %[six] \n\t" + "psrah %[b_vec0], %[b_vec0], %[six] \n\t" + "psrah %[g_vec0], %[g_vec0], %[six] \n\t" + "psrah %[r_vec0], %[r_vec0], %[six] \n\t" - "packushb %[r_vec0], %[b_vec0], %[r_vec0] \n\t" - "packushb %[g_vec0], %[g_vec0], %[alpha] \n\t" - "punpcklwd %[g_vec0], %[g_vec0], %[alpha] \n\t" - "punpcklbh %[b_vec0], %[r_vec0], %[g_vec0] \n\t" - "punpckhbh %[r_vec0], %[r_vec0], %[g_vec0] \n\t" - "punpcklhw %[g_vec0], %[b_vec0], %[r_vec0] \n\t" - "punpckhhw %[g_vec1], %[b_vec0], %[r_vec0] \n\t" - "gssdlc1 %[g_vec0], 0x07(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[g_vec0], 0x00(%[rgbbuf_ptr]) \n\t" - "gssdlc1 %[g_vec1], 0x0f(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[g_vec1], 0x08(%[rgbbuf_ptr]) \n\t" + "packushb %[r_vec0], %[b_vec0], %[r_vec0] \n\t" + "packushb %[g_vec0], %[g_vec0], %[alpha] \n\t" + "punpcklwd %[g_vec0], %[g_vec0], %[alpha] \n\t" + "punpcklbh %[b_vec0], %[r_vec0], %[g_vec0] \n\t" + "punpckhbh %[r_vec0], %[r_vec0], %[g_vec0] \n\t" + "punpcklhw %[g_vec0], %[b_vec0], %[r_vec0] \n\t" + "punpckhhw %[g_vec1], %[b_vec0], %[r_vec0] \n\t" + "gssdlc1 %[g_vec0], 0x07(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[g_vec0], 0x00(%[rgbbuf_ptr]) \n\t" + "gssdlc1 %[g_vec1], 0x0f(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[g_vec1], 0x08(%[rgbbuf_ptr]) \n\t" - "daddiu %[y_ptr], %[y_ptr], 0x08 \n\t" - "daddiu %[u_ptr], %[u_ptr], 0x04 \n\t" - "daddiu %[v_ptr], %[v_ptr], 0x04 \n\t" - "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x10 \n\t" - "daddi %[width], %[width], -0x04 \n\t" - "bnez %[width], 1b \n\t" + "daddiu %[y_ptr], %[y_ptr], 0x08 \n\t" + "daddiu %[u_ptr], %[u_ptr], 0x04 \n\t" + "daddiu %[v_ptr], %[v_ptr], 0x04 \n\t" + "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x10 \n\t" + "daddi %[width], %[width], -0x04 \n\t" + "bnez %[width], 1b \n\t" - : [y]"=&f"(y), - [u]"=&f"(u), [v]"=&f"(v), - [b_vec0]"=&f"(b_vec[0]), [b_vec1]"=&f"(b_vec[1]), - [g_vec0]"=&f"(g_vec[0]), [g_vec1]"=&f"(g_vec[1]), - [r_vec0]"=&f"(r_vec[0]), [r_vec1]"=&f"(r_vec[1]), - [ub]"=&f"(ub), [ug]"=&f"(ug), - [vg]"=&f"(vg), [vr]"=&f"(vr), - [bb]"=&f"(bb), [bg]"=&f"(bg), - [br]"=&f"(br), [yg]"=&f"(yg) - : [y_ptr]"r"(src_y), [u_ptr]"r"(src_u), - [v_ptr]"r"(src_v), [rgbbuf_ptr]"r"(rgb_buf), - [yuvcons_ptr]"r"(yuvconstants), [width]"r"(width), - [zero]"f"(0x00), [alpha]"f"(-1), - [six]"f"(0x6), [five]"f"(0x55), - [mask]"f"(mask), [two]"f"(0x02), - [mask1]"f"(0x00ff00ff00ff00ff) - : "memory" - ); + : [y] "=&f"(y), [u] "=&f"(u), [v] "=&f"(v), [b_vec0] "=&f"(b_vec[0]), + [b_vec1] "=&f"(b_vec[1]), [g_vec0] "=&f"(g_vec[0]), + [g_vec1] "=&f"(g_vec[1]), [r_vec0] "=&f"(r_vec[0]), + [r_vec1] "=&f"(r_vec[1]), [ub] "=&f"(ub), [ug] "=&f"(ug), + [vg] "=&f"(vg), [vr] "=&f"(vr), [bb] "=&f"(bb), [bg] "=&f"(bg), + [br] "=&f"(br), [yg] "=&f"(yg) + : [y_ptr] "r"(src_y), [u_ptr] "r"(src_u), [v_ptr] "r"(src_v), + [rgbbuf_ptr] "r"(rgb_buf), [yuvcons_ptr] "r"(yuvconstants), + [width] "r"(width), [zero] "f"(0x00), [alpha] "f"(-1), [six] "f"(0x6), + [five] "f"(0x55), [mask] "f"(mask), [two] "f"(0x02), + [mask1] "f"(0x00ff00ff00ff00ff) + : "memory"); } void I422AlphaToARGBRow_MMI(const uint8_t* src_y, @@ -6343,102 +6331,96 @@ void I422AlphaToARGBRow_MMI(const uint8_t* src_y, uint8_t* rgb_buf, const struct YuvConstants* yuvconstants, int width) { - uint64_t y,u,v,a; - uint64_t b_vec[2],g_vec[2],r_vec[2]; + uint64_t y, u, v, a; + uint64_t b_vec[2], g_vec[2], r_vec[2]; uint64_t mask = 0xff00ff00ff00ff00ULL; - uint64_t ub,ug,vg,vr,bb,bg,br,yg; + uint64_t ub, ug, vg, vr, bb, bg, br, yg; __asm__ volatile( - "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" - "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" - "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" - "or %[ub], %[ub], %[mask] \n\t" - "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" - "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[ug], %[ug], %[zero] \n\t" - "pshufh %[ug], %[ug], %[zero] \n\t" - "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vg], %[vg], %[zero] \n\t" - "pshufh %[vg], %[vg], %[five] \n\t" - "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" - "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vr], %[vr], %[zero] \n\t" - "pshufh %[vr], %[vr], %[five] \n\t" - "or %[vr], %[vr], %[mask] \n\t" + "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" + "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" + "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" + "or %[ub], %[ub], %[mask] \n\t" + "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" + "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[ug], %[ug], %[zero] \n\t" + "pshufh %[ug], %[ug], %[zero] \n\t" + "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vg], %[vg], %[zero] \n\t" + "pshufh %[vg], %[vg], %[five] \n\t" + "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" + "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vr], %[vr], %[zero] \n\t" + "pshufh %[vr], %[vr], %[five] \n\t" + "or %[vr], %[vr], %[mask] \n\t" - "1: \n\t" - "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" - "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" - "gslwlc1 %[u], 0x03(%[u_ptr]) \n\t" - "gslwrc1 %[u], 0x00(%[u_ptr]) \n\t" - "gslwlc1 %[v], 0x03(%[v_ptr]) \n\t" - "gslwrc1 %[v], 0x00(%[v_ptr]) \n\t" - "gslwlc1 %[a], 0x03(%[a_ptr]) \n\t" - "gslwrc1 %[a], 0x00(%[a_ptr]) \n\t" + "1: \n\t" + "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" + "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" + "gslwlc1 %[u], 0x03(%[u_ptr]) \n\t" + "gslwrc1 %[u], 0x00(%[u_ptr]) \n\t" + "gslwlc1 %[v], 0x03(%[v_ptr]) \n\t" + "gslwrc1 %[v], 0x00(%[v_ptr]) \n\t" + "gslwlc1 %[a], 0x03(%[a_ptr]) \n\t" + "gslwrc1 %[a], 0x00(%[a_ptr]) \n\t" - "punpcklbh %[y], %[y], %[y] \n\t"//y*0x0101 - "pmulhuh %[y], %[y], %[yg] \n\t"//y1 + "punpcklbh %[y], %[y], %[y] \n\t" // y*0x0101 + "pmulhuh %[y], %[y], %[yg] \n\t" // y1 - //u3|u2|u1|u0 --> u1|u1|u0|u0 - "punpcklbh %[u], %[u], %[u] \n\t"//u - "punpcklbh %[u], %[u], %[zero] \n\t" - "paddsh %[b_vec0], %[y], %[bb] \n\t" - "pmullh %[b_vec1], %[u], %[ub] \n\t" - "psubsh %[b_vec0], %[b_vec0], %[b_vec1] \n\t" - "psrah %[b_vec0], %[b_vec0], %[six] \n\t" + // u3|u2|u1|u0 --> u1|u1|u0|u0 + "punpcklbh %[u], %[u], %[u] \n\t" // u + "punpcklbh %[u], %[u], %[zero] \n\t" + "paddsh %[b_vec0], %[y], %[bb] \n\t" + "pmullh %[b_vec1], %[u], %[ub] \n\t" + "psubsh %[b_vec0], %[b_vec0], %[b_vec1] \n\t" + "psrah %[b_vec0], %[b_vec0], %[six] \n\t" - //v3|v2|v1|v0 --> v1|v1|v0|v0 - "punpcklbh %[v], %[v], %[v] \n\t" - "punpcklbh %[v], %[v], %[zero] \n\t" - "paddsh %[g_vec0], %[y], %[bg] \n\t" - "pmullh %[g_vec1], %[u], %[ug] \n\t" - "psubsh %[g_vec0], %[g_vec0], %[g_vec1] \n\t" - "pmullh %[g_vec1], %[v], %[vg] \n\t" - "psubsh %[g_vec0], %[g_vec0], %[g_vec1] \n\t" - "psrah %[g_vec0], %[g_vec0], %[six] \n\t" + // v3|v2|v1|v0 --> v1|v1|v0|v0 + "punpcklbh %[v], %[v], %[v] \n\t" + "punpcklbh %[v], %[v], %[zero] \n\t" + "paddsh %[g_vec0], %[y], %[bg] \n\t" + "pmullh %[g_vec1], %[u], %[ug] \n\t" + "psubsh %[g_vec0], %[g_vec0], %[g_vec1] \n\t" + "pmullh %[g_vec1], %[v], %[vg] \n\t" + "psubsh %[g_vec0], %[g_vec0], %[g_vec1] \n\t" + "psrah %[g_vec0], %[g_vec0], %[six] \n\t" - "paddsh %[r_vec0], %[y], %[br] \n\t" - "pmullh %[r_vec1], %[v], %[vr] \n\t" - "psubsh %[r_vec0], %[r_vec0], %[r_vec1] \n\t" - "psrah %[r_vec0], %[r_vec0], %[six] \n\t" + "paddsh %[r_vec0], %[y], %[br] \n\t" + "pmullh %[r_vec1], %[v], %[vr] \n\t" + "psubsh %[r_vec0], %[r_vec0], %[r_vec1] \n\t" + "psrah %[r_vec0], %[r_vec0], %[six] \n\t" - "packushb %[r_vec0], %[b_vec0], %[r_vec0] \n\t"//rrrrbbbb - "packushb %[g_vec0], %[g_vec0], %[a] \n\t" - "punpcklwd %[g_vec0], %[g_vec0], %[a] \n\t"//aaaagggg - "punpcklbh %[b_vec0], %[r_vec0], %[g_vec0] \n\t" - "punpckhbh %[r_vec0], %[r_vec0], %[g_vec0] \n\t" - "punpcklhw %[g_vec0], %[b_vec0], %[r_vec0] \n\t" - "punpckhhw %[g_vec1], %[b_vec0], %[r_vec0] \n\t" - "gssdlc1 %[g_vec0], 0x07(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[g_vec0], 0x00(%[rgbbuf_ptr]) \n\t" - "gssdlc1 %[g_vec1], 0x0f(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[g_vec1], 0x08(%[rgbbuf_ptr]) \n\t" + "packushb %[r_vec0], %[b_vec0], %[r_vec0] \n\t" // rrrrbbbb + "packushb %[g_vec0], %[g_vec0], %[a] \n\t" + "punpcklwd %[g_vec0], %[g_vec0], %[a] \n\t" // aaaagggg + "punpcklbh %[b_vec0], %[r_vec0], %[g_vec0] \n\t" + "punpckhbh %[r_vec0], %[r_vec0], %[g_vec0] \n\t" + "punpcklhw %[g_vec0], %[b_vec0], %[r_vec0] \n\t" + "punpckhhw %[g_vec1], %[b_vec0], %[r_vec0] \n\t" + "gssdlc1 %[g_vec0], 0x07(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[g_vec0], 0x00(%[rgbbuf_ptr]) \n\t" + "gssdlc1 %[g_vec1], 0x0f(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[g_vec1], 0x08(%[rgbbuf_ptr]) \n\t" - "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" - "daddiu %[a_ptr], %[a_ptr], 0x04 \n\t" - "daddiu %[u_ptr], %[u_ptr], 0x02 \n\t" - "daddiu %[v_ptr], %[v_ptr], 0x02 \n\t" - "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x10 \n\t" - "daddi %[width], %[width], -0x04 \n\t" - "bnez %[width], 1b \n\t" + "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" + "daddiu %[a_ptr], %[a_ptr], 0x04 \n\t" + "daddiu %[u_ptr], %[u_ptr], 0x02 \n\t" + "daddiu %[v_ptr], %[v_ptr], 0x02 \n\t" + "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x10 \n\t" + "daddi %[width], %[width], -0x04 \n\t" + "bnez %[width], 1b \n\t" - : [y]"=&f"(y), [u]"=&f"(u), - [v]"=&f"(v), [a]"=&f"(a), - [b_vec0]"=&f"(b_vec[0]), [b_vec1]"=&f"(b_vec[1]), - [g_vec0]"=&f"(g_vec[0]), [g_vec1]"=&f"(g_vec[1]), - [r_vec0]"=&f"(r_vec[0]), [r_vec1]"=&f"(r_vec[1]), - [ub]"=&f"(ub), [ug]"=&f"(ug), - [vg]"=&f"(vg), [vr]"=&f"(vr), - [bb]"=&f"(bb), [bg]"=&f"(bg), - [br]"=&f"(br), [yg]"=&f"(yg) - : [y_ptr]"r"(src_y), [u_ptr]"r"(src_u), - [v_ptr]"r"(src_v), [rgbbuf_ptr]"r"(rgb_buf), - [yuvcons_ptr]"r"(yuvconstants), [width]"r"(width), - [a_ptr]"r"(src_a), [zero]"f"(0x00), - [six]"f"(0x6), [five]"f"(0x55), - [mask]"f"(mask) - : "memory" - ); + : [y] "=&f"(y), [u] "=&f"(u), [v] "=&f"(v), [a] "=&f"(a), + [b_vec0] "=&f"(b_vec[0]), [b_vec1] "=&f"(b_vec[1]), + [g_vec0] "=&f"(g_vec[0]), [g_vec1] "=&f"(g_vec[1]), + [r_vec0] "=&f"(r_vec[0]), [r_vec1] "=&f"(r_vec[1]), [ub] "=&f"(ub), + [ug] "=&f"(ug), [vg] "=&f"(vg), [vr] "=&f"(vr), [bb] "=&f"(bb), + [bg] "=&f"(bg), [br] "=&f"(br), [yg] "=&f"(yg) + : [y_ptr] "r"(src_y), [u_ptr] "r"(src_u), [v_ptr] "r"(src_v), + [rgbbuf_ptr] "r"(rgb_buf), [yuvcons_ptr] "r"(yuvconstants), + [width] "r"(width), [a_ptr] "r"(src_a), [zero] "f"(0x00), + [six] "f"(0x6), [five] "f"(0x55), [mask] "f"(mask) + : "memory"); } void I422ToRGB24Row_MMI(const uint8_t* src_y, @@ -6447,113 +6429,105 @@ void I422ToRGB24Row_MMI(const uint8_t* src_y, uint8_t* rgb_buf, const struct YuvConstants* yuvconstants, int width) { - uint64_t y,u,v; - uint64_t b_vec[2],g_vec[2],r_vec[2]; + uint64_t y, u, v; + uint64_t b_vec[2], g_vec[2], r_vec[2]; uint64_t mask = 0xff00ff00ff00ff00ULL; - uint64_t ub,ug,vg,vr,bb,bg,br,yg; + uint64_t ub, ug, vg, vr, bb, bg, br, yg; __asm__ volatile( - "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" - "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" - "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" - "or %[ub], %[ub], %[mask] \n\t" - "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" - "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[ug], %[ug], %[zero] \n\t" - "pshufh %[ug], %[ug], %[zero] \n\t" - "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vg], %[vg], %[zero] \n\t" - "pshufh %[vg], %[vg], %[five] \n\t" - "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" - "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vr], %[vr], %[zero] \n\t" - "pshufh %[vr], %[vr], %[five] \n\t" - "or %[vr], %[vr], %[mask] \n\t" + "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" + "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" + "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" + "or %[ub], %[ub], %[mask] \n\t" + "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" + "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[ug], %[ug], %[zero] \n\t" + "pshufh %[ug], %[ug], %[zero] \n\t" + "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vg], %[vg], %[zero] \n\t" + "pshufh %[vg], %[vg], %[five] \n\t" + "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" + "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vr], %[vr], %[zero] \n\t" + "pshufh %[vr], %[vr], %[five] \n\t" + "or %[vr], %[vr], %[mask] \n\t" - "1: \n\t" - "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" - "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" - "gslwlc1 %[u], 0x03(%[u_ptr]) \n\t" - "gslwrc1 %[u], 0x00(%[u_ptr]) \n\t" - "gslwlc1 %[v], 0x03(%[v_ptr]) \n\t" - "gslwrc1 %[v], 0x00(%[v_ptr]) \n\t" + "1: \n\t" + "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" + "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" + "gslwlc1 %[u], 0x03(%[u_ptr]) \n\t" + "gslwrc1 %[u], 0x00(%[u_ptr]) \n\t" + "gslwlc1 %[v], 0x03(%[v_ptr]) \n\t" + "gslwrc1 %[v], 0x00(%[v_ptr]) \n\t" - "punpcklbh %[y], %[y], %[y] \n\t"//y*0x0101 - "pmulhuh %[y], %[y], %[yg] \n\t"//y1 + "punpcklbh %[y], %[y], %[y] \n\t" // y*0x0101 + "pmulhuh %[y], %[y], %[yg] \n\t" // y1 - //u3|u2|u1|u0 --> u1|u1|u0|u0 - "punpcklbh %[u], %[u], %[u] \n\t"//u - "punpcklbh %[u], %[u], %[zero] \n\t" - "paddsh %[b_vec0], %[y], %[bb] \n\t" - "pmullh %[b_vec1], %[u], %[ub] \n\t" - "psubsh %[b_vec0], %[b_vec0], %[b_vec1] \n\t" - "psrah %[b_vec0], %[b_vec0], %[six] \n\t" + // u3|u2|u1|u0 --> u1|u1|u0|u0 + "punpcklbh %[u], %[u], %[u] \n\t" // u + "punpcklbh %[u], %[u], %[zero] \n\t" + "paddsh %[b_vec0], %[y], %[bb] \n\t" + "pmullh %[b_vec1], %[u], %[ub] \n\t" + "psubsh %[b_vec0], %[b_vec0], %[b_vec1] \n\t" + "psrah %[b_vec0], %[b_vec0], %[six] \n\t" - //v3|v2|v1|v0 --> v1|v1|v0|v0 - "punpcklbh %[v], %[v], %[v] \n\t" - "punpcklbh %[v], %[v], %[zero] \n\t" - "paddsh %[g_vec0], %[y], %[bg] \n\t" - "pmullh %[g_vec1], %[u], %[ug] \n\t" - "psubsh %[g_vec0], %[g_vec0], %[g_vec1] \n\t" - "pmullh %[g_vec1], %[v], %[vg] \n\t" - "psubsh %[g_vec0], %[g_vec0], %[g_vec1] \n\t" - "psrah %[g_vec0], %[g_vec0], %[six] \n\t" + // v3|v2|v1|v0 --> v1|v1|v0|v0 + "punpcklbh %[v], %[v], %[v] \n\t" + "punpcklbh %[v], %[v], %[zero] \n\t" + "paddsh %[g_vec0], %[y], %[bg] \n\t" + "pmullh %[g_vec1], %[u], %[ug] \n\t" + "psubsh %[g_vec0], %[g_vec0], %[g_vec1] \n\t" + "pmullh %[g_vec1], %[v], %[vg] \n\t" + "psubsh %[g_vec0], %[g_vec0], %[g_vec1] \n\t" + "psrah %[g_vec0], %[g_vec0], %[six] \n\t" - "paddsh %[r_vec0], %[y], %[br] \n\t" - "pmullh %[r_vec1], %[v], %[vr] \n\t" - "psubsh %[r_vec0], %[r_vec0], %[r_vec1] \n\t" - "psrah %[r_vec0], %[r_vec0], %[six] \n\t" + "paddsh %[r_vec0], %[y], %[br] \n\t" + "pmullh %[r_vec1], %[v], %[vr] \n\t" + "psubsh %[r_vec0], %[r_vec0], %[r_vec1] \n\t" + "psrah %[r_vec0], %[r_vec0], %[six] \n\t" - "packushb %[r_vec0], %[b_vec0], %[r_vec0] \n\t" - "packushb %[g_vec0], %[g_vec0], %[zero] \n\t" - "punpcklbh %[b_vec0], %[r_vec0], %[g_vec0] \n\t" - "punpckhbh %[r_vec0], %[r_vec0], %[g_vec0] \n\t" - "punpcklhw %[g_vec0], %[b_vec0], %[r_vec0] \n\t" - "punpckhhw %[g_vec1], %[b_vec0], %[r_vec0] \n\t" + "packushb %[r_vec0], %[b_vec0], %[r_vec0] \n\t" + "packushb %[g_vec0], %[g_vec0], %[zero] \n\t" + "punpcklbh %[b_vec0], %[r_vec0], %[g_vec0] \n\t" + "punpckhbh %[r_vec0], %[r_vec0], %[g_vec0] \n\t" + "punpcklhw %[g_vec0], %[b_vec0], %[r_vec0] \n\t" + "punpckhhw %[g_vec1], %[b_vec0], %[r_vec0] \n\t" - "punpckhwd %[r_vec0], %[g_vec0], %[g_vec0] \n\t" - "psllw %[r_vec1], %[r_vec0], %[lmove1] \n\t" - "or %[g_vec0], %[g_vec0], %[r_vec1] \n\t" - "psrlw %[r_vec1], %[r_vec0], %[rmove1] \n\t" - "pextrh %[r_vec1], %[r_vec1], %[zero] \n\t" - "pinsrh_2 %[g_vec0], %[g_vec0], %[r_vec1] \n\t" - "pextrh %[r_vec1], %[g_vec1], %[zero] \n\t" - "pinsrh_3 %[g_vec0], %[g_vec0], %[r_vec1] \n\t" - "pextrh %[r_vec1], %[g_vec1], %[one] \n\t" - "punpckhwd %[g_vec1], %[g_vec1], %[g_vec1] \n\t" - "psllw %[g_vec1], %[g_vec1], %[rmove1] \n\t" - "or %[g_vec1], %[g_vec1], %[r_vec1] \n\t" - "gssdlc1 %[g_vec0], 0x07(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[g_vec0], 0x00(%[rgbbuf_ptr]) \n\t" - "gsswlc1 %[g_vec1], 0x0b(%[rgbbuf_ptr]) \n\t" - "gsswrc1 %[g_vec1], 0x08(%[rgbbuf_ptr]) \n\t" + "punpckhwd %[r_vec0], %[g_vec0], %[g_vec0] \n\t" + "psllw %[r_vec1], %[r_vec0], %[lmove1] \n\t" + "or %[g_vec0], %[g_vec0], %[r_vec1] \n\t" + "psrlw %[r_vec1], %[r_vec0], %[rmove1] \n\t" + "pextrh %[r_vec1], %[r_vec1], %[zero] \n\t" + "pinsrh_2 %[g_vec0], %[g_vec0], %[r_vec1] \n\t" + "pextrh %[r_vec1], %[g_vec1], %[zero] \n\t" + "pinsrh_3 %[g_vec0], %[g_vec0], %[r_vec1] \n\t" + "pextrh %[r_vec1], %[g_vec1], %[one] \n\t" + "punpckhwd %[g_vec1], %[g_vec1], %[g_vec1] \n\t" + "psllw %[g_vec1], %[g_vec1], %[rmove1] \n\t" + "or %[g_vec1], %[g_vec1], %[r_vec1] \n\t" + "gssdlc1 %[g_vec0], 0x07(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[g_vec0], 0x00(%[rgbbuf_ptr]) \n\t" + "gsswlc1 %[g_vec1], 0x0b(%[rgbbuf_ptr]) \n\t" + "gsswrc1 %[g_vec1], 0x08(%[rgbbuf_ptr]) \n\t" + "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" + "daddiu %[u_ptr], %[u_ptr], 0x02 \n\t" + "daddiu %[v_ptr], %[v_ptr], 0x02 \n\t" + "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x0c \n\t" + "daddi %[width], %[width], -0x04 \n\t" + "bnez %[width], 1b \n\t" - "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" - "daddiu %[u_ptr], %[u_ptr], 0x02 \n\t" - "daddiu %[v_ptr], %[v_ptr], 0x02 \n\t" - "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x0c \n\t" - "daddi %[width], %[width], -0x04 \n\t" - "bnez %[width], 1b \n\t" - - : [y]"=&f"(y), [u]"=&f"(u), - [v]"=&f"(v), - [b_vec0]"=&f"(b_vec[0]), [b_vec1]"=&f"(b_vec[1]), - [g_vec0]"=&f"(g_vec[0]), [g_vec1]"=&f"(g_vec[1]), - [r_vec0]"=&f"(r_vec[0]), [r_vec1]"=&f"(r_vec[1]), - [ub]"=&f"(ub), [ug]"=&f"(ug), - [vg]"=&f"(vg), [vr]"=&f"(vr), - [bb]"=&f"(bb), [bg]"=&f"(bg), - [br]"=&f"(br), [yg]"=&f"(yg) - : [y_ptr]"r"(src_y), [u_ptr]"r"(src_u), - [v_ptr]"r"(src_v), [rgbbuf_ptr]"r"(rgb_buf), - [yuvcons_ptr]"r"(yuvconstants), [width]"r"(width), - [zero]"f"(0x00), [five]"f"(0x55), - [six]"f"(0x6), [mask]"f"(mask), - [lmove1]"f"(0x18), [rmove1]"f"(0x8), - [one]"f"(0x1) - : "memory" - ); + : [y] "=&f"(y), [u] "=&f"(u), [v] "=&f"(v), [b_vec0] "=&f"(b_vec[0]), + [b_vec1] "=&f"(b_vec[1]), [g_vec0] "=&f"(g_vec[0]), + [g_vec1] "=&f"(g_vec[1]), [r_vec0] "=&f"(r_vec[0]), + [r_vec1] "=&f"(r_vec[1]), [ub] "=&f"(ub), [ug] "=&f"(ug), + [vg] "=&f"(vg), [vr] "=&f"(vr), [bb] "=&f"(bb), [bg] "=&f"(bg), + [br] "=&f"(br), [yg] "=&f"(yg) + : [y_ptr] "r"(src_y), [u_ptr] "r"(src_u), [v_ptr] "r"(src_v), + [rgbbuf_ptr] "r"(rgb_buf), [yuvcons_ptr] "r"(yuvconstants), + [width] "r"(width), [zero] "f"(0x00), [five] "f"(0x55), [six] "f"(0x6), + [mask] "f"(mask), [lmove1] "f"(0x18), [rmove1] "f"(0x8), [one] "f"(0x1) + : "memory"); } void I422ToARGB4444Row_MMI(const uint8_t* src_y, @@ -6564,110 +6538,103 @@ void I422ToARGB4444Row_MMI(const uint8_t* src_y, int width) { uint64_t y, u, v; uint64_t b_vec, g_vec, r_vec, temp; - uint64_t ub,ug,vg,vr,bb,bg,br,yg; + uint64_t ub, ug, vg, vr, bb, bg, br, yg; __asm__ volatile( - "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" - "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" - "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" - "or %[ub], %[ub], %[mask] \n\t" - "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" - "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[ug], %[ug], %[zero] \n\t" - "pshufh %[ug], %[ug], %[zero] \n\t" - "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vg], %[vg], %[zero] \n\t" - "pshufh %[vg], %[vg], %[five] \n\t" - "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" - "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vr], %[vr], %[zero] \n\t" - "pshufh %[vr], %[vr], %[five] \n\t" - "or %[vr], %[vr], %[mask] \n\t" + "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" + "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" + "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" + "or %[ub], %[ub], %[mask] \n\t" + "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" + "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[ug], %[ug], %[zero] \n\t" + "pshufh %[ug], %[ug], %[zero] \n\t" + "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vg], %[vg], %[zero] \n\t" + "pshufh %[vg], %[vg], %[five] \n\t" + "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" + "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vr], %[vr], %[zero] \n\t" + "pshufh %[vr], %[vr], %[five] \n\t" + "or %[vr], %[vr], %[mask] \n\t" - "1: \n\t" - "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" - "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" - "gslwlc1 %[u], 0x03(%[u_ptr]) \n\t" - "gslwrc1 %[u], 0x00(%[u_ptr]) \n\t" - "gslwlc1 %[v], 0x03(%[v_ptr]) \n\t" - "gslwrc1 %[v], 0x00(%[v_ptr]) \n\t" + "1: \n\t" + "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" + "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" + "gslwlc1 %[u], 0x03(%[u_ptr]) \n\t" + "gslwrc1 %[u], 0x00(%[u_ptr]) \n\t" + "gslwlc1 %[v], 0x03(%[v_ptr]) \n\t" + "gslwrc1 %[v], 0x00(%[v_ptr]) \n\t" - "punpcklbh %[y], %[y], %[y] \n\t"//y*0x0101 - "pmulhuh %[y], %[y], %[yg] \n\t"//y1 + "punpcklbh %[y], %[y], %[y] \n\t" // y*0x0101 + "pmulhuh %[y], %[y], %[yg] \n\t" // y1 - //u3|u2|u1|u0 --> u1|u1|u0|u0 - "punpcklbh %[u], %[u], %[u] \n\t"//u - "punpcklbh %[u], %[u], %[zero] \n\t" - "paddsh %[b_vec], %[y], %[bb] \n\t" - "pmullh %[temp], %[u], %[ub] \n\t" - "psubsh %[b_vec], %[b_vec], %[temp] \n\t" - "psrah %[b_vec], %[b_vec], %[six] \n\t" + // u3|u2|u1|u0 --> u1|u1|u0|u0 + "punpcklbh %[u], %[u], %[u] \n\t" // u + "punpcklbh %[u], %[u], %[zero] \n\t" + "paddsh %[b_vec], %[y], %[bb] \n\t" + "pmullh %[temp], %[u], %[ub] \n\t" + "psubsh %[b_vec], %[b_vec], %[temp] \n\t" + "psrah %[b_vec], %[b_vec], %[six] \n\t" - //v3|v2|v1|v0 --> v1|v1|v0|v0 - "punpcklbh %[v], %[v], %[v] \n\t" - "punpcklbh %[v], %[v], %[zero] \n\t" - "paddsh %[g_vec], %[y], %[bg] \n\t" - "pmullh %[temp], %[u], %[ug] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "pmullh %[temp], %[v], %[vg] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "psrah %[g_vec], %[g_vec], %[six] \n\t" + // v3|v2|v1|v0 --> v1|v1|v0|v0 + "punpcklbh %[v], %[v], %[v] \n\t" + "punpcklbh %[v], %[v], %[zero] \n\t" + "paddsh %[g_vec], %[y], %[bg] \n\t" + "pmullh %[temp], %[u], %[ug] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "pmullh %[temp], %[v], %[vg] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "psrah %[g_vec], %[g_vec], %[six] \n\t" - "paddsh %[r_vec], %[y], %[br] \n\t" - "pmullh %[temp], %[v], %[vr] \n\t" - "psubsh %[r_vec], %[r_vec], %[temp] \n\t" - "psrah %[r_vec], %[r_vec], %[six] \n\t" + "paddsh %[r_vec], %[y], %[br] \n\t" + "pmullh %[temp], %[v], %[vr] \n\t" + "psubsh %[r_vec], %[r_vec], %[temp] \n\t" + "psrah %[r_vec], %[r_vec], %[six] \n\t" - "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" - "packushb %[g_vec], %[g_vec], %[zero] \n\t" - "punpcklwd %[g_vec], %[g_vec], %[alpha] \n\t" - "punpcklbh %[b_vec], %[r_vec], %[g_vec] \n\t" - "punpckhbh %[r_vec], %[r_vec], %[g_vec] \n\t" - "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" - "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[g_vec], %[g_vec], %[zero] \n\t" + "punpcklwd %[g_vec], %[g_vec], %[alpha] \n\t" + "punpcklbh %[b_vec], %[r_vec], %[g_vec] \n\t" + "punpckhbh %[r_vec], %[r_vec], %[g_vec] \n\t" + "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" + "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" - "and %[g_vec], %[g_vec], %[mask1] \n\t" - "psrlw %[g_vec], %[g_vec], %[four] \n\t" - "psrlw %[r_vec], %[g_vec], %[four] \n\t" - "or %[g_vec], %[g_vec], %[r_vec] \n\t" - "punpcklbh %[r_vec], %[alpha], %[zero] \n\t" - "and %[g_vec], %[g_vec], %[r_vec] \n\t" + "and %[g_vec], %[g_vec], %[mask1] \n\t" + "psrlw %[g_vec], %[g_vec], %[four] \n\t" + "psrlw %[r_vec], %[g_vec], %[four] \n\t" + "or %[g_vec], %[g_vec], %[r_vec] \n\t" + "punpcklbh %[r_vec], %[alpha], %[zero] \n\t" + "and %[g_vec], %[g_vec], %[r_vec] \n\t" - "and %[b_vec], %[b_vec], %[mask1] \n\t" - "psrlw %[b_vec], %[b_vec], %[four] \n\t" - "psrlw %[r_vec], %[b_vec], %[four] \n\t" - "or %[b_vec], %[b_vec], %[r_vec] \n\t" - "punpcklbh %[r_vec], %[alpha], %[zero] \n\t" - "and %[b_vec], %[b_vec], %[r_vec] \n\t" - "packushb %[g_vec], %[g_vec], %[b_vec] \n\t" + "and %[b_vec], %[b_vec], %[mask1] \n\t" + "psrlw %[b_vec], %[b_vec], %[four] \n\t" + "psrlw %[r_vec], %[b_vec], %[four] \n\t" + "or %[b_vec], %[b_vec], %[r_vec] \n\t" + "punpcklbh %[r_vec], %[alpha], %[zero] \n\t" + "and %[b_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[g_vec], %[g_vec], %[b_vec] \n\t" - "gssdlc1 %[g_vec], 0x07(%[dst_argb4444]) \n\t" - "gssdrc1 %[g_vec], 0x00(%[dst_argb4444]) \n\t" + "gssdlc1 %[g_vec], 0x07(%[dst_argb4444]) \n\t" + "gssdrc1 %[g_vec], 0x00(%[dst_argb4444]) \n\t" - "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" - "daddiu %[u_ptr], %[u_ptr], 0x02 \n\t" - "daddiu %[v_ptr], %[v_ptr], 0x02 \n\t" - "daddiu %[dst_argb4444], %[dst_argb4444], 0x08 \n\t" - "daddi %[width], %[width], -0x04 \n\t" - "bnez %[width], 1b \n\t" + "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" + "daddiu %[u_ptr], %[u_ptr], 0x02 \n\t" + "daddiu %[v_ptr], %[v_ptr], 0x02 \n\t" + "daddiu %[dst_argb4444], %[dst_argb4444], 0x08 \n\t" + "daddi %[width], %[width], -0x04 \n\t" + "bnez %[width], 1b \n\t" - : [y]"=&f"(y), [u]"=&f"(u), - [v]"=&f"(v), - [b_vec]"=&f"(b_vec), [g_vec]"=&f"(g_vec), - [r_vec]"=&f"(r_vec), [temp]"=&f"(temp), - [ub]"=&f"(ub), [ug]"=&f"(ug), - [vg]"=&f"(vg), [vr]"=&f"(vr), - [bb]"=&f"(bb), [bg]"=&f"(bg), - [br]"=&f"(br), [yg]"=&f"(yg) - : [y_ptr]"r"(src_y), [u_ptr]"r"(src_u), - [v_ptr]"r"(src_v), [dst_argb4444]"r"(dst_argb4444), - [yuvcons_ptr]"r"(yuvconstants), [width]"r"(width), - [zero]"f"(0x00), [five]"f"(0x55), - [six]"f"(0x6), [mask]"f"(0xff00ff00ff00ff00), - [four]"f"(0x4), [mask1]"f"(0xf0f0f0f0f0f0f0f0), - [alpha]"f"(-1) - : "memory" - ); + : [y] "=&f"(y), [u] "=&f"(u), [v] "=&f"(v), [b_vec] "=&f"(b_vec), + [g_vec] "=&f"(g_vec), [r_vec] "=&f"(r_vec), [temp] "=&f"(temp), + [ub] "=&f"(ub), [ug] "=&f"(ug), [vg] "=&f"(vg), [vr] "=&f"(vr), + [bb] "=&f"(bb), [bg] "=&f"(bg), [br] "=&f"(br), [yg] "=&f"(yg) + : [y_ptr] "r"(src_y), [u_ptr] "r"(src_u), [v_ptr] "r"(src_v), + [dst_argb4444] "r"(dst_argb4444), [yuvcons_ptr] "r"(yuvconstants), + [width] "r"(width), [zero] "f"(0x00), [five] "f"(0x55), [six] "f"(0x6), + [mask] "f"(0xff00ff00ff00ff00), [four] "f"(0x4), + [mask1] "f"(0xf0f0f0f0f0f0f0f0), [alpha] "f"(-1) + : "memory"); } void I422ToARGB1555Row_MMI(const uint8_t* src_y, @@ -6678,125 +6645,118 @@ void I422ToARGB1555Row_MMI(const uint8_t* src_y, int width) { uint64_t y, u, v; uint64_t b_vec, g_vec, r_vec, temp; - uint64_t ub,ug,vg,vr,bb,bg,br,yg; + uint64_t ub, ug, vg, vr, bb, bg, br, yg; __asm__ volatile( - "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" - "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" - "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" - "or %[ub], %[ub], %[mask1] \n\t" - "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" - "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[ug], %[ug], %[zero] \n\t" - "pshufh %[ug], %[ug], %[zero] \n\t" - "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vg], %[vg], %[zero] \n\t" - "pshufh %[vg], %[vg], %[five] \n\t" - "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" - "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vr], %[vr], %[zero] \n\t" - "pshufh %[vr], %[vr], %[five] \n\t" - "or %[vr], %[vr], %[mask1] \n\t" + "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" + "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" + "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" + "or %[ub], %[ub], %[mask1] \n\t" + "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" + "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[ug], %[ug], %[zero] \n\t" + "pshufh %[ug], %[ug], %[zero] \n\t" + "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vg], %[vg], %[zero] \n\t" + "pshufh %[vg], %[vg], %[five] \n\t" + "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" + "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vr], %[vr], %[zero] \n\t" + "pshufh %[vr], %[vr], %[five] \n\t" + "or %[vr], %[vr], %[mask1] \n\t" - "1: \n\t" - "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" - "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" - "gslwlc1 %[u], 0x03(%[u_ptr]) \n\t" - "gslwrc1 %[u], 0x00(%[u_ptr]) \n\t" - "gslwlc1 %[v], 0x03(%[v_ptr]) \n\t" - "gslwrc1 %[v], 0x00(%[v_ptr]) \n\t" + "1: \n\t" + "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" + "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" + "gslwlc1 %[u], 0x03(%[u_ptr]) \n\t" + "gslwrc1 %[u], 0x00(%[u_ptr]) \n\t" + "gslwlc1 %[v], 0x03(%[v_ptr]) \n\t" + "gslwrc1 %[v], 0x00(%[v_ptr]) \n\t" - "punpcklbh %[y], %[y], %[y] \n\t" - "pmulhuh %[y], %[y], %[yg] \n\t" + "punpcklbh %[y], %[y], %[y] \n\t" + "pmulhuh %[y], %[y], %[yg] \n\t" - //u3|u2|u1|u0 --> u1|u1|u0|u0 - "punpcklbh %[u], %[u], %[u] \n\t" - "punpcklbh %[u], %[u], %[zero] \n\t" - "paddsh %[b_vec], %[y], %[bb] \n\t" - "pmullh %[temp], %[u], %[ub] \n\t" - "psubsh %[b_vec], %[b_vec], %[temp] \n\t" - "psrah %[b_vec], %[b_vec], %[six] \n\t" + // u3|u2|u1|u0 --> u1|u1|u0|u0 + "punpcklbh %[u], %[u], %[u] \n\t" + "punpcklbh %[u], %[u], %[zero] \n\t" + "paddsh %[b_vec], %[y], %[bb] \n\t" + "pmullh %[temp], %[u], %[ub] \n\t" + "psubsh %[b_vec], %[b_vec], %[temp] \n\t" + "psrah %[b_vec], %[b_vec], %[six] \n\t" - //v3|v2|v1|v0 --> v1|v1|v0|v0 - "punpcklbh %[v], %[v], %[v] \n\t" - "punpcklbh %[v], %[v], %[zero] \n\t" - "paddsh %[g_vec], %[y], %[bg] \n\t" - "pmullh %[temp], %[u], %[ug] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "pmullh %[temp], %[v], %[vg] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "psrah %[g_vec], %[g_vec], %[six] \n\t" + // v3|v2|v1|v0 --> v1|v1|v0|v0 + "punpcklbh %[v], %[v], %[v] \n\t" + "punpcklbh %[v], %[v], %[zero] \n\t" + "paddsh %[g_vec], %[y], %[bg] \n\t" + "pmullh %[temp], %[u], %[ug] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "pmullh %[temp], %[v], %[vg] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "psrah %[g_vec], %[g_vec], %[six] \n\t" - "paddsh %[r_vec], %[y], %[br] \n\t" - "pmullh %[temp], %[v], %[vr] \n\t" - "psubsh %[r_vec], %[r_vec], %[temp] \n\t" - "psrah %[r_vec], %[r_vec], %[six] \n\t" + "paddsh %[r_vec], %[y], %[br] \n\t" + "pmullh %[temp], %[v], %[vr] \n\t" + "psubsh %[r_vec], %[r_vec], %[temp] \n\t" + "psrah %[r_vec], %[r_vec], %[six] \n\t" - "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" - "packushb %[g_vec], %[g_vec], %[zero] \n\t" - "punpcklbh %[b_vec], %[r_vec], %[g_vec] \n\t" - "punpckhbh %[r_vec], %[r_vec], %[g_vec] \n\t" - "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" - "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[g_vec], %[g_vec], %[zero] \n\t" + "punpcklbh %[b_vec], %[r_vec], %[g_vec] \n\t" + "punpckhbh %[r_vec], %[r_vec], %[g_vec] \n\t" + "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" + "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" - "psrlw %[temp], %[g_vec], %[three] \n\t" - "and %[g_vec], %[temp], %[mask2] \n\t" - "psrlw %[temp], %[temp], %[eight] \n\t" - "and %[r_vec], %[temp], %[mask2] \n\t" - "psllw %[r_vec], %[r_vec], %[lmove5] \n\t" - "or %[g_vec], %[g_vec], %[r_vec] \n\t" - "psrlw %[temp], %[temp], %[eight] \n\t" - "and %[r_vec], %[temp], %[mask2] \n\t" - "psllw %[r_vec], %[r_vec], %[lmove5] \n\t" - "psllw %[r_vec], %[r_vec], %[lmove5] \n\t" - "or %[g_vec], %[g_vec], %[r_vec] \n\t" - "or %[g_vec], %[g_vec], %[mask3] \n\t" + "psrlw %[temp], %[g_vec], %[three] \n\t" + "and %[g_vec], %[temp], %[mask2] \n\t" + "psrlw %[temp], %[temp], %[eight] \n\t" + "and %[r_vec], %[temp], %[mask2] \n\t" + "psllw %[r_vec], %[r_vec], %[lmove5] \n\t" + "or %[g_vec], %[g_vec], %[r_vec] \n\t" + "psrlw %[temp], %[temp], %[eight] \n\t" + "and %[r_vec], %[temp], %[mask2] \n\t" + "psllw %[r_vec], %[r_vec], %[lmove5] \n\t" + "psllw %[r_vec], %[r_vec], %[lmove5] \n\t" + "or %[g_vec], %[g_vec], %[r_vec] \n\t" + "or %[g_vec], %[g_vec], %[mask3] \n\t" - "psrlw %[temp], %[b_vec], %[three] \n\t" - "and %[b_vec], %[temp], %[mask2] \n\t" - "psrlw %[temp], %[temp], %[eight] \n\t" - "and %[r_vec], %[temp], %[mask2] \n\t" - "psllw %[r_vec], %[r_vec], %[lmove5] \n\t" - "or %[b_vec], %[b_vec], %[r_vec] \n\t" - "psrlw %[temp], %[temp], %[eight] \n\t" - "and %[r_vec], %[temp], %[mask2] \n\t" - "psllw %[r_vec], %[r_vec], %[lmove5] \n\t" - "psllw %[r_vec], %[r_vec], %[lmove5] \n\t" - "or %[b_vec], %[b_vec], %[r_vec] \n\t" - "or %[b_vec], %[b_vec], %[mask3] \n\t" + "psrlw %[temp], %[b_vec], %[three] \n\t" + "and %[b_vec], %[temp], %[mask2] \n\t" + "psrlw %[temp], %[temp], %[eight] \n\t" + "and %[r_vec], %[temp], %[mask2] \n\t" + "psllw %[r_vec], %[r_vec], %[lmove5] \n\t" + "or %[b_vec], %[b_vec], %[r_vec] \n\t" + "psrlw %[temp], %[temp], %[eight] \n\t" + "and %[r_vec], %[temp], %[mask2] \n\t" + "psllw %[r_vec], %[r_vec], %[lmove5] \n\t" + "psllw %[r_vec], %[r_vec], %[lmove5] \n\t" + "or %[b_vec], %[b_vec], %[r_vec] \n\t" + "or %[b_vec], %[b_vec], %[mask3] \n\t" - "punpcklhw %[r_vec], %[g_vec], %[b_vec] \n\t" - "punpckhhw %[b_vec], %[g_vec], %[b_vec] \n\t" - "punpcklhw %[g_vec], %[r_vec], %[b_vec] \n\t" + "punpcklhw %[r_vec], %[g_vec], %[b_vec] \n\t" + "punpckhhw %[b_vec], %[g_vec], %[b_vec] \n\t" + "punpcklhw %[g_vec], %[r_vec], %[b_vec] \n\t" - "gssdlc1 %[g_vec], 0x07(%[dst_argb1555]) \n\t" - "gssdrc1 %[g_vec], 0x00(%[dst_argb1555]) \n\t" + "gssdlc1 %[g_vec], 0x07(%[dst_argb1555]) \n\t" + "gssdrc1 %[g_vec], 0x00(%[dst_argb1555]) \n\t" - "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" - "daddiu %[u_ptr], %[u_ptr], 0x02 \n\t" - "daddiu %[v_ptr], %[v_ptr], 0x02 \n\t" - "daddiu %[dst_argb1555], %[dst_argb1555], 0x08 \n\t" - "daddi %[width], %[width], -0x04 \n\t" - "bnez %[width], 1b \n\t" + "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" + "daddiu %[u_ptr], %[u_ptr], 0x02 \n\t" + "daddiu %[v_ptr], %[v_ptr], 0x02 \n\t" + "daddiu %[dst_argb1555], %[dst_argb1555], 0x08 \n\t" + "daddi %[width], %[width], -0x04 \n\t" + "bnez %[width], 1b \n\t" - : [y]"=&f"(y), [u]"=&f"(u), - [v]"=&f"(v), - [b_vec]"=&f"(b_vec), [g_vec]"=&f"(g_vec), - [r_vec]"=&f"(r_vec), [temp]"=&f"(temp), - [ub]"=&f"(ub), [ug]"=&f"(ug), - [vg]"=&f"(vg), [vr]"=&f"(vr), - [bb]"=&f"(bb), [bg]"=&f"(bg), - [br]"=&f"(br), [yg]"=&f"(yg) - : [y_ptr]"r"(src_y), [u_ptr]"r"(src_u), - [v_ptr]"r"(src_v), [dst_argb1555]"r"(dst_argb1555), - [yuvcons_ptr]"r"(yuvconstants), [width]"r"(width), - [zero]"f"(0x00), [five]"f"(0x55), - [six]"f"(0x6), [mask1]"f"(0xff00ff00ff00ff00), - [three]"f"(0x3), [mask2]"f"(0x1f0000001f), - [eight]"f"(0x8), [mask3]"f"(0x800000008000), - [lmove5]"f"(0x5) - : "memory" - ); + : [y] "=&f"(y), [u] "=&f"(u), [v] "=&f"(v), [b_vec] "=&f"(b_vec), + [g_vec] "=&f"(g_vec), [r_vec] "=&f"(r_vec), [temp] "=&f"(temp), + [ub] "=&f"(ub), [ug] "=&f"(ug), [vg] "=&f"(vg), [vr] "=&f"(vr), + [bb] "=&f"(bb), [bg] "=&f"(bg), [br] "=&f"(br), [yg] "=&f"(yg) + : [y_ptr] "r"(src_y), [u_ptr] "r"(src_u), [v_ptr] "r"(src_v), + [dst_argb1555] "r"(dst_argb1555), [yuvcons_ptr] "r"(yuvconstants), + [width] "r"(width), [zero] "f"(0x00), [five] "f"(0x55), [six] "f"(0x6), + [mask1] "f"(0xff00ff00ff00ff00), [three] "f"(0x3), + [mask2] "f"(0x1f0000001f), [eight] "f"(0x8), + [mask3] "f"(0x800000008000), [lmove5] "f"(0x5) + : "memory"); } void I422ToRGB565Row_MMI(const uint8_t* src_y, @@ -6807,127 +6767,120 @@ void I422ToRGB565Row_MMI(const uint8_t* src_y, int width) { uint64_t y, u, v; uint64_t b_vec, g_vec, r_vec, temp; - uint64_t ub,ug,vg,vr,bb,bg,br,yg; + uint64_t ub, ug, vg, vr, bb, bg, br, yg; __asm__ volatile( - "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" - "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" - "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" - "or %[ub], %[ub], %[mask1] \n\t" - "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" - "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[ug], %[ug], %[zero] \n\t" - "pshufh %[ug], %[ug], %[zero] \n\t" - "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vg], %[vg], %[zero] \n\t" - "pshufh %[vg], %[vg], %[five] \n\t" - "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" - "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vr], %[vr], %[zero] \n\t" - "pshufh %[vr], %[vr], %[five] \n\t" - "or %[vr], %[vr], %[mask1] \n\t" + "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" + "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" + "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" + "or %[ub], %[ub], %[mask1] \n\t" + "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" + "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[ug], %[ug], %[zero] \n\t" + "pshufh %[ug], %[ug], %[zero] \n\t" + "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vg], %[vg], %[zero] \n\t" + "pshufh %[vg], %[vg], %[five] \n\t" + "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" + "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vr], %[vr], %[zero] \n\t" + "pshufh %[vr], %[vr], %[five] \n\t" + "or %[vr], %[vr], %[mask1] \n\t" - "1: \n\t" - "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" - "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" - "gslwlc1 %[u], 0x03(%[u_ptr]) \n\t" - "gslwrc1 %[u], 0x00(%[u_ptr]) \n\t" - "gslwlc1 %[v], 0x03(%[v_ptr]) \n\t" - "gslwrc1 %[v], 0x00(%[v_ptr]) \n\t" + "1: \n\t" + "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" + "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" + "gslwlc1 %[u], 0x03(%[u_ptr]) \n\t" + "gslwrc1 %[u], 0x00(%[u_ptr]) \n\t" + "gslwlc1 %[v], 0x03(%[v_ptr]) \n\t" + "gslwrc1 %[v], 0x00(%[v_ptr]) \n\t" - "punpcklbh %[y], %[y], %[y] \n\t" - "pmulhuh %[y], %[y], %[yg] \n\t" + "punpcklbh %[y], %[y], %[y] \n\t" + "pmulhuh %[y], %[y], %[yg] \n\t" - //u3|u2|u1|u0 --> u1|u1|u0|u0 - "punpcklbh %[u], %[u], %[u] \n\t" - "punpcklbh %[u], %[u], %[zero] \n\t" - "paddsh %[b_vec], %[y], %[bb] \n\t" - "pmullh %[temp], %[u], %[ub] \n\t" - "psubsh %[b_vec], %[b_vec], %[temp] \n\t" - "psrah %[b_vec], %[b_vec], %[six] \n\t" + // u3|u2|u1|u0 --> u1|u1|u0|u0 + "punpcklbh %[u], %[u], %[u] \n\t" + "punpcklbh %[u], %[u], %[zero] \n\t" + "paddsh %[b_vec], %[y], %[bb] \n\t" + "pmullh %[temp], %[u], %[ub] \n\t" + "psubsh %[b_vec], %[b_vec], %[temp] \n\t" + "psrah %[b_vec], %[b_vec], %[six] \n\t" - //v3|v2|v1|v0 --> v1|v1|v0|v0 - "punpcklbh %[v], %[v], %[v] \n\t" - "punpcklbh %[v], %[v], %[zero] \n\t" - "paddsh %[g_vec], %[y], %[bg] \n\t" - "pmullh %[temp], %[u], %[ug] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "pmullh %[temp], %[v], %[vg] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "psrah %[g_vec], %[g_vec], %[six] \n\t" + // v3|v2|v1|v0 --> v1|v1|v0|v0 + "punpcklbh %[v], %[v], %[v] \n\t" + "punpcklbh %[v], %[v], %[zero] \n\t" + "paddsh %[g_vec], %[y], %[bg] \n\t" + "pmullh %[temp], %[u], %[ug] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "pmullh %[temp], %[v], %[vg] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "psrah %[g_vec], %[g_vec], %[six] \n\t" - "paddsh %[r_vec], %[y], %[br] \n\t" - "pmullh %[temp], %[v], %[vr] \n\t" - "psubsh %[r_vec], %[r_vec], %[temp] \n\t" - "psrah %[r_vec], %[r_vec], %[six] \n\t" + "paddsh %[r_vec], %[y], %[br] \n\t" + "pmullh %[temp], %[v], %[vr] \n\t" + "psubsh %[r_vec], %[r_vec], %[temp] \n\t" + "psrah %[r_vec], %[r_vec], %[six] \n\t" - "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" - "packushb %[g_vec], %[g_vec], %[zero] \n\t" - "punpcklbh %[b_vec], %[r_vec], %[g_vec] \n\t" - "punpckhbh %[r_vec], %[r_vec], %[g_vec] \n\t" - "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" - "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[g_vec], %[g_vec], %[zero] \n\t" + "punpcklbh %[b_vec], %[r_vec], %[g_vec] \n\t" + "punpckhbh %[r_vec], %[r_vec], %[g_vec] \n\t" + "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" + "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" - "psrlh %[temp], %[g_vec], %[three] \n\t" - "and %[g_vec], %[temp], %[mask2] \n\t" - "psrlw %[temp], %[temp], %[seven] \n\t" - "psrlw %[r_vec], %[mask1], %[eight] \n\t" - "and %[r_vec], %[temp], %[r_vec] \n\t" - "psllw %[r_vec], %[r_vec], %[lmove5] \n\t" - "or %[g_vec], %[g_vec], %[r_vec] \n\t" - "paddb %[r_vec], %[three], %[six] \n\t" - "psrlw %[temp], %[temp], %[r_vec] \n\t" - "and %[r_vec], %[temp], %[mask2] \n\t" - "paddb %[temp], %[three], %[eight] \n\t" - "psllw %[r_vec], %[r_vec], %[temp] \n\t" - "or %[g_vec], %[g_vec], %[r_vec] \n\t" + "psrlh %[temp], %[g_vec], %[three] \n\t" + "and %[g_vec], %[temp], %[mask2] \n\t" + "psrlw %[temp], %[temp], %[seven] \n\t" + "psrlw %[r_vec], %[mask1], %[eight] \n\t" + "and %[r_vec], %[temp], %[r_vec] \n\t" + "psllw %[r_vec], %[r_vec], %[lmove5] \n\t" + "or %[g_vec], %[g_vec], %[r_vec] \n\t" + "paddb %[r_vec], %[three], %[six] \n\t" + "psrlw %[temp], %[temp], %[r_vec] \n\t" + "and %[r_vec], %[temp], %[mask2] \n\t" + "paddb %[temp], %[three], %[eight] \n\t" + "psllw %[r_vec], %[r_vec], %[temp] \n\t" + "or %[g_vec], %[g_vec], %[r_vec] \n\t" - "psrlh %[temp], %[b_vec], %[three] \n\t" - "and %[b_vec], %[temp], %[mask2] \n\t" - "psrlw %[temp], %[temp], %[seven] \n\t" - "psrlw %[r_vec], %[mask1], %[eight] \n\t" - "and %[r_vec], %[temp], %[r_vec] \n\t" - "psllw %[r_vec], %[r_vec], %[lmove5] \n\t" - "or %[b_vec], %[b_vec], %[r_vec] \n\t" - "paddb %[r_vec], %[three], %[six] \n\t" - "psrlw %[temp], %[temp], %[r_vec] \n\t" - "and %[r_vec], %[temp], %[mask2] \n\t" - "paddb %[temp], %[three], %[eight] \n\t" - "psllw %[r_vec], %[r_vec], %[temp] \n\t" - "or %[b_vec], %[b_vec], %[r_vec] \n\t" + "psrlh %[temp], %[b_vec], %[three] \n\t" + "and %[b_vec], %[temp], %[mask2] \n\t" + "psrlw %[temp], %[temp], %[seven] \n\t" + "psrlw %[r_vec], %[mask1], %[eight] \n\t" + "and %[r_vec], %[temp], %[r_vec] \n\t" + "psllw %[r_vec], %[r_vec], %[lmove5] \n\t" + "or %[b_vec], %[b_vec], %[r_vec] \n\t" + "paddb %[r_vec], %[three], %[six] \n\t" + "psrlw %[temp], %[temp], %[r_vec] \n\t" + "and %[r_vec], %[temp], %[mask2] \n\t" + "paddb %[temp], %[three], %[eight] \n\t" + "psllw %[r_vec], %[r_vec], %[temp] \n\t" + "or %[b_vec], %[b_vec], %[r_vec] \n\t" - "punpcklhw %[r_vec], %[g_vec], %[b_vec] \n\t" - "punpckhhw %[b_vec], %[g_vec], %[b_vec] \n\t" - "punpcklhw %[g_vec], %[r_vec], %[b_vec] \n\t" + "punpcklhw %[r_vec], %[g_vec], %[b_vec] \n\t" + "punpckhhw %[b_vec], %[g_vec], %[b_vec] \n\t" + "punpcklhw %[g_vec], %[r_vec], %[b_vec] \n\t" - "gssdlc1 %[g_vec], 0x07(%[dst_rgb565]) \n\t" - "gssdrc1 %[g_vec], 0x00(%[dst_rgb565]) \n\t" + "gssdlc1 %[g_vec], 0x07(%[dst_rgb565]) \n\t" + "gssdrc1 %[g_vec], 0x00(%[dst_rgb565]) \n\t" - "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" - "daddiu %[u_ptr], %[u_ptr], 0x02 \n\t" - "daddiu %[v_ptr], %[v_ptr], 0x02 \n\t" - "daddiu %[dst_rgb565], %[dst_rgb565], 0x08 \n\t" - "daddi %[width], %[width], -0x04 \n\t" - "bnez %[width], 1b \n\t" + "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" + "daddiu %[u_ptr], %[u_ptr], 0x02 \n\t" + "daddiu %[v_ptr], %[v_ptr], 0x02 \n\t" + "daddiu %[dst_rgb565], %[dst_rgb565], 0x08 \n\t" + "daddi %[width], %[width], -0x04 \n\t" + "bnez %[width], 1b \n\t" - : [y]"=&f"(y), [u]"=&f"(u), - [v]"=&f"(v), - [b_vec]"=&f"(b_vec), [g_vec]"=&f"(g_vec), - [r_vec]"=&f"(r_vec), [temp]"=&f"(temp), - [ub]"=&f"(ub), [ug]"=&f"(ug), - [vg]"=&f"(vg), [vr]"=&f"(vr), - [bb]"=&f"(bb), [bg]"=&f"(bg), - [br]"=&f"(br), [yg]"=&f"(yg) - : [y_ptr]"r"(src_y), [u_ptr]"r"(src_u), - [v_ptr]"r"(src_v), [dst_rgb565]"r"(dst_rgb565), - [yuvcons_ptr]"r"(yuvconstants), [width]"r"(width), - [zero]"f"(0x00), [five]"f"(0x55), - [six]"f"(0x6), [mask1]"f"(0xff00ff00ff00ff00), - [three]"f"(0x3), [mask2]"f"(0x1f0000001f), - [eight]"f"(0x8), [seven]"f"(0x7), - [lmove5]"f"(0x5) - : "memory" - ); + : [y] "=&f"(y), [u] "=&f"(u), [v] "=&f"(v), [b_vec] "=&f"(b_vec), + [g_vec] "=&f"(g_vec), [r_vec] "=&f"(r_vec), [temp] "=&f"(temp), + [ub] "=&f"(ub), [ug] "=&f"(ug), [vg] "=&f"(vg), [vr] "=&f"(vr), + [bb] "=&f"(bb), [bg] "=&f"(bg), [br] "=&f"(br), [yg] "=&f"(yg) + : [y_ptr] "r"(src_y), [u_ptr] "r"(src_u), [v_ptr] "r"(src_v), + [dst_rgb565] "r"(dst_rgb565), [yuvcons_ptr] "r"(yuvconstants), + [width] "r"(width), [zero] "f"(0x00), [five] "f"(0x55), [six] "f"(0x6), + [mask1] "f"(0xff00ff00ff00ff00), [three] "f"(0x3), + [mask2] "f"(0x1f0000001f), [eight] "f"(0x8), [seven] "f"(0x7), + [lmove5] "f"(0x5) + : "memory"); } void NV12ToARGBRow_MMI(const uint8_t* src_y, @@ -6937,91 +6890,83 @@ void NV12ToARGBRow_MMI(const uint8_t* src_y, int width) { uint64_t y, u, v; uint64_t b_vec, g_vec, r_vec, temp; - uint64_t ub,ug,vg,vr,bb,bg,br,yg; + uint64_t ub, ug, vg, vr, bb, bg, br, yg; __asm__ volatile( - "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" - "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" - "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" - "or %[ub], %[ub], %[mask1] \n\t" - "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" - "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[ug], %[ug], %[zero] \n\t" - "pshufh %[ug], %[ug], %[zero] \n\t" - "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vg], %[vg], %[zero] \n\t" - "pshufh %[vg], %[vg], %[five] \n\t" - "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" - "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vr], %[vr], %[zero] \n\t" - "pshufh %[vr], %[vr], %[five] \n\t" - "or %[vr], %[vr], %[mask1] \n\t" + "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" + "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" + "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" + "or %[ub], %[ub], %[mask1] \n\t" + "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" + "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[ug], %[ug], %[zero] \n\t" + "pshufh %[ug], %[ug], %[zero] \n\t" + "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vg], %[vg], %[zero] \n\t" + "pshufh %[vg], %[vg], %[five] \n\t" + "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" + "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vr], %[vr], %[zero] \n\t" + "pshufh %[vr], %[vr], %[five] \n\t" + "or %[vr], %[vr], %[mask1] \n\t" - "1: \n\t" - "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" - "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" - "gslwlc1 %[u], 0x03(%[uv_ptr]) \n\t" - "gslwrc1 %[u], 0x00(%[uv_ptr]) \n\t" - "punpcklbh %[u], %[u], %[zero] \n\t" - "pshufh %[v], %[u], %[vshu] \n\t" - "pshufh %[u], %[u], %[ushu] \n\t" + "1: \n\t" + "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" + "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" + "gslwlc1 %[u], 0x03(%[uv_ptr]) \n\t" + "gslwrc1 %[u], 0x00(%[uv_ptr]) \n\t" + "punpcklbh %[u], %[u], %[zero] \n\t" + "pshufh %[v], %[u], %[vshu] \n\t" + "pshufh %[u], %[u], %[ushu] \n\t" - "punpcklbh %[y], %[y], %[y] \n\t" - "pmulhuh %[y], %[y], %[yg] \n\t" + "punpcklbh %[y], %[y], %[y] \n\t" + "pmulhuh %[y], %[y], %[yg] \n\t" - "paddsh %[b_vec], %[y], %[bb] \n\t" - "pmullh %[temp], %[u], %[ub] \n\t" - "psubsh %[b_vec], %[b_vec], %[temp] \n\t" - "psrah %[b_vec], %[b_vec], %[six] \n\t" + "paddsh %[b_vec], %[y], %[bb] \n\t" + "pmullh %[temp], %[u], %[ub] \n\t" + "psubsh %[b_vec], %[b_vec], %[temp] \n\t" + "psrah %[b_vec], %[b_vec], %[six] \n\t" - "paddsh %[g_vec], %[y], %[bg] \n\t" - "pmullh %[temp], %[u], %[ug] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "pmullh %[temp], %[v], %[vg] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "psrah %[g_vec], %[g_vec], %[six] \n\t" + "paddsh %[g_vec], %[y], %[bg] \n\t" + "pmullh %[temp], %[u], %[ug] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "pmullh %[temp], %[v], %[vg] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "psrah %[g_vec], %[g_vec], %[six] \n\t" - "paddsh %[r_vec], %[y], %[br] \n\t" - "pmullh %[temp], %[v], %[vr] \n\t" - "psubsh %[r_vec], %[r_vec], %[temp] \n\t" - "psrah %[r_vec], %[r_vec], %[six] \n\t" + "paddsh %[r_vec], %[y], %[br] \n\t" + "pmullh %[temp], %[v], %[vr] \n\t" + "psubsh %[r_vec], %[r_vec], %[temp] \n\t" + "psrah %[r_vec], %[r_vec], %[six] \n\t" - "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" - "packushb %[g_vec], %[g_vec], %[zero] \n\t" - "punpcklwd %[g_vec], %[g_vec], %[alpha] \n\t" - "punpcklbh %[b_vec], %[r_vec], %[g_vec] \n\t" - "punpckhbh %[r_vec], %[r_vec], %[g_vec] \n\t" - "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" - "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[g_vec], %[g_vec], %[zero] \n\t" + "punpcklwd %[g_vec], %[g_vec], %[alpha] \n\t" + "punpcklbh %[b_vec], %[r_vec], %[g_vec] \n\t" + "punpckhbh %[r_vec], %[r_vec], %[g_vec] \n\t" + "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" + "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" - "gssdlc1 %[g_vec], 0x07(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[g_vec], 0x00(%[rgbbuf_ptr]) \n\t" - "gssdlc1 %[b_vec], 0x0f(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[b_vec], 0x08(%[rgbbuf_ptr]) \n\t" + "gssdlc1 %[g_vec], 0x07(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[g_vec], 0x00(%[rgbbuf_ptr]) \n\t" + "gssdlc1 %[b_vec], 0x0f(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[b_vec], 0x08(%[rgbbuf_ptr]) \n\t" - "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" - "daddiu %[uv_ptr], %[uv_ptr], 0x04 \n\t" - "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x10 \n\t" - "daddi %[width], %[width], -0x04 \n\t" - "bnez %[width], 1b \n\t" + "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" + "daddiu %[uv_ptr], %[uv_ptr], 0x04 \n\t" + "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x10 \n\t" + "daddi %[width], %[width], -0x04 \n\t" + "bnez %[width], 1b \n\t" - : [y]"=&f"(y), [u]"=&f"(u), - [v]"=&f"(v), - [b_vec]"=&f"(b_vec), [g_vec]"=&f"(g_vec), - [r_vec]"=&f"(r_vec), [temp]"=&f"(temp), - [ub]"=&f"(ub), [ug]"=&f"(ug), - [vg]"=&f"(vg), [vr]"=&f"(vr), - [bb]"=&f"(bb), [bg]"=&f"(bg), - [br]"=&f"(br), [yg]"=&f"(yg) - : [y_ptr]"r"(src_y), [uv_ptr]"r"(src_uv), - [rgbbuf_ptr]"r"(rgb_buf), - [yuvcons_ptr]"r"(yuvconstants), [width]"r"(width), - [zero]"f"(0x00), [five]"f"(0x55), - [six]"f"(0x6), [mask1]"f"(0xff00ff00ff00ff00), - [ushu]"f"(0xA0), [vshu]"f"(0xf5), - [alpha]"f"(-1) - : "memory" - ); + : [y] "=&f"(y), [u] "=&f"(u), [v] "=&f"(v), [b_vec] "=&f"(b_vec), + [g_vec] "=&f"(g_vec), [r_vec] "=&f"(r_vec), [temp] "=&f"(temp), + [ub] "=&f"(ub), [ug] "=&f"(ug), [vg] "=&f"(vg), [vr] "=&f"(vr), + [bb] "=&f"(bb), [bg] "=&f"(bg), [br] "=&f"(br), [yg] "=&f"(yg) + : [y_ptr] "r"(src_y), [uv_ptr] "r"(src_uv), [rgbbuf_ptr] "r"(rgb_buf), + [yuvcons_ptr] "r"(yuvconstants), [width] "r"(width), [zero] "f"(0x00), + [five] "f"(0x55), [six] "f"(0x6), [mask1] "f"(0xff00ff00ff00ff00), + [ushu] "f"(0xA0), [vshu] "f"(0xf5), [alpha] "f"(-1) + : "memory"); } void NV21ToARGBRow_MMI(const uint8_t* src_y, @@ -7031,91 +6976,83 @@ void NV21ToARGBRow_MMI(const uint8_t* src_y, int width) { uint64_t y, u, v; uint64_t b_vec, g_vec, r_vec, temp; - uint64_t ub,ug,vg,vr,bb,bg,br,yg; + uint64_t ub, ug, vg, vr, bb, bg, br, yg; __asm__ volatile( - "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" - "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" - "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" - "or %[ub], %[ub], %[mask1] \n\t" - "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" - "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[ug], %[ug], %[zero] \n\t" - "pshufh %[ug], %[ug], %[zero] \n\t" - "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vg], %[vg], %[zero] \n\t" - "pshufh %[vg], %[vg], %[five] \n\t" - "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" - "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vr], %[vr], %[zero] \n\t" - "pshufh %[vr], %[vr], %[five] \n\t" - "or %[vr], %[vr], %[mask1] \n\t" + "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" + "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" + "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" + "or %[ub], %[ub], %[mask1] \n\t" + "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" + "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[ug], %[ug], %[zero] \n\t" + "pshufh %[ug], %[ug], %[zero] \n\t" + "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vg], %[vg], %[zero] \n\t" + "pshufh %[vg], %[vg], %[five] \n\t" + "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" + "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vr], %[vr], %[zero] \n\t" + "pshufh %[vr], %[vr], %[five] \n\t" + "or %[vr], %[vr], %[mask1] \n\t" - "1: \n\t" - "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" - "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" - "gslwlc1 %[u], 0x03(%[vu_ptr]) \n\t" - "gslwrc1 %[u], 0x00(%[vu_ptr]) \n\t" - "punpcklbh %[u], %[u], %[zero] \n\t" - "pshufh %[v], %[u], %[ushu] \n\t" - "pshufh %[u], %[u], %[vshu] \n\t" + "1: \n\t" + "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" + "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" + "gslwlc1 %[u], 0x03(%[vu_ptr]) \n\t" + "gslwrc1 %[u], 0x00(%[vu_ptr]) \n\t" + "punpcklbh %[u], %[u], %[zero] \n\t" + "pshufh %[v], %[u], %[ushu] \n\t" + "pshufh %[u], %[u], %[vshu] \n\t" - "punpcklbh %[y], %[y], %[y] \n\t" - "pmulhuh %[y], %[y], %[yg] \n\t" + "punpcklbh %[y], %[y], %[y] \n\t" + "pmulhuh %[y], %[y], %[yg] \n\t" - "paddsh %[b_vec], %[y], %[bb] \n\t" - "pmullh %[temp], %[u], %[ub] \n\t" - "psubsh %[b_vec], %[b_vec], %[temp] \n\t" - "psrah %[b_vec], %[b_vec], %[six] \n\t" + "paddsh %[b_vec], %[y], %[bb] \n\t" + "pmullh %[temp], %[u], %[ub] \n\t" + "psubsh %[b_vec], %[b_vec], %[temp] \n\t" + "psrah %[b_vec], %[b_vec], %[six] \n\t" - "paddsh %[g_vec], %[y], %[bg] \n\t" - "pmullh %[temp], %[u], %[ug] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "pmullh %[temp], %[v], %[vg] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "psrah %[g_vec], %[g_vec], %[six] \n\t" + "paddsh %[g_vec], %[y], %[bg] \n\t" + "pmullh %[temp], %[u], %[ug] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "pmullh %[temp], %[v], %[vg] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "psrah %[g_vec], %[g_vec], %[six] \n\t" - "paddsh %[r_vec], %[y], %[br] \n\t" - "pmullh %[temp], %[v], %[vr] \n\t" - "psubsh %[r_vec], %[r_vec], %[temp] \n\t" - "psrah %[r_vec], %[r_vec], %[six] \n\t" + "paddsh %[r_vec], %[y], %[br] \n\t" + "pmullh %[temp], %[v], %[vr] \n\t" + "psubsh %[r_vec], %[r_vec], %[temp] \n\t" + "psrah %[r_vec], %[r_vec], %[six] \n\t" - "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" - "packushb %[g_vec], %[g_vec], %[zero] \n\t" - "punpcklwd %[g_vec], %[g_vec], %[alpha] \n\t" - "punpcklbh %[b_vec], %[r_vec], %[g_vec] \n\t" - "punpckhbh %[r_vec], %[r_vec], %[g_vec] \n\t" - "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" - "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[g_vec], %[g_vec], %[zero] \n\t" + "punpcklwd %[g_vec], %[g_vec], %[alpha] \n\t" + "punpcklbh %[b_vec], %[r_vec], %[g_vec] \n\t" + "punpckhbh %[r_vec], %[r_vec], %[g_vec] \n\t" + "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" + "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" - "gssdlc1 %[g_vec], 0x07(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[g_vec], 0x00(%[rgbbuf_ptr]) \n\t" - "gssdlc1 %[b_vec], 0x0f(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[b_vec], 0x08(%[rgbbuf_ptr]) \n\t" + "gssdlc1 %[g_vec], 0x07(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[g_vec], 0x00(%[rgbbuf_ptr]) \n\t" + "gssdlc1 %[b_vec], 0x0f(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[b_vec], 0x08(%[rgbbuf_ptr]) \n\t" - "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" - "daddiu %[vu_ptr], %[vu_ptr], 0x04 \n\t" - "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x10 \n\t" - "daddi %[width], %[width], -0x04 \n\t" - "bnez %[width], 1b \n\t" + "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" + "daddiu %[vu_ptr], %[vu_ptr], 0x04 \n\t" + "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x10 \n\t" + "daddi %[width], %[width], -0x04 \n\t" + "bnez %[width], 1b \n\t" - : [y]"=&f"(y), [u]"=&f"(u), - [v]"=&f"(v), - [b_vec]"=&f"(b_vec), [g_vec]"=&f"(g_vec), - [r_vec]"=&f"(r_vec), [temp]"=&f"(temp), - [ub]"=&f"(ub), [ug]"=&f"(ug), - [vg]"=&f"(vg), [vr]"=&f"(vr), - [bb]"=&f"(bb), [bg]"=&f"(bg), - [br]"=&f"(br), [yg]"=&f"(yg) - : [y_ptr]"r"(src_y), [vu_ptr]"r"(src_vu), - [rgbbuf_ptr]"r"(rgb_buf), - [yuvcons_ptr]"r"(yuvconstants), [width]"r"(width), - [zero]"f"(0x00), [five]"f"(0x55), - [six]"f"(0x6), [mask1]"f"(0xff00ff00ff00ff00), - [ushu]"f"(0xA0), [vshu]"f"(0xf5), - [alpha]"f"(-1) - : "memory" - ); + : [y] "=&f"(y), [u] "=&f"(u), [v] "=&f"(v), [b_vec] "=&f"(b_vec), + [g_vec] "=&f"(g_vec), [r_vec] "=&f"(r_vec), [temp] "=&f"(temp), + [ub] "=&f"(ub), [ug] "=&f"(ug), [vg] "=&f"(vg), [vr] "=&f"(vr), + [bb] "=&f"(bb), [bg] "=&f"(bg), [br] "=&f"(br), [yg] "=&f"(yg) + : [y_ptr] "r"(src_y), [vu_ptr] "r"(src_vu), [rgbbuf_ptr] "r"(rgb_buf), + [yuvcons_ptr] "r"(yuvconstants), [width] "r"(width), [zero] "f"(0x00), + [five] "f"(0x55), [six] "f"(0x6), [mask1] "f"(0xff00ff00ff00ff00), + [ushu] "f"(0xA0), [vshu] "f"(0xf5), [alpha] "f"(-1) + : "memory"); } void NV12ToRGB24Row_MMI(const uint8_t* src_y, @@ -7125,103 +7062,95 @@ void NV12ToRGB24Row_MMI(const uint8_t* src_y, int width) { uint64_t y, u, v; uint64_t b_vec, g_vec, r_vec, temp; - uint64_t ub,ug,vg,vr,bb,bg,br,yg; + uint64_t ub, ug, vg, vr, bb, bg, br, yg; __asm__ volatile( - "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" - "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" - "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" - "or %[ub], %[ub], %[mask1] \n\t" - "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" - "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[ug], %[ug], %[zero] \n\t" - "pshufh %[ug], %[ug], %[zero] \n\t" - "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vg], %[vg], %[zero] \n\t" - "pshufh %[vg], %[vg], %[five] \n\t" - "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" - "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vr], %[vr], %[zero] \n\t" - "pshufh %[vr], %[vr], %[five] \n\t" - "or %[vr], %[vr], %[mask1] \n\t" + "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" + "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" + "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" + "or %[ub], %[ub], %[mask1] \n\t" + "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" + "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[ug], %[ug], %[zero] \n\t" + "pshufh %[ug], %[ug], %[zero] \n\t" + "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vg], %[vg], %[zero] \n\t" + "pshufh %[vg], %[vg], %[five] \n\t" + "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" + "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vr], %[vr], %[zero] \n\t" + "pshufh %[vr], %[vr], %[five] \n\t" + "or %[vr], %[vr], %[mask1] \n\t" - "1: \n\t" - "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" - "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" - "gslwlc1 %[u], 0x03(%[uv_ptr]) \n\t" - "gslwrc1 %[u], 0x00(%[uv_ptr]) \n\t" - "punpcklbh %[u], %[u], %[zero] \n\t" - "pshufh %[v], %[u], %[vshu] \n\t" - "pshufh %[u], %[u], %[ushu] \n\t" + "1: \n\t" + "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" + "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" + "gslwlc1 %[u], 0x03(%[uv_ptr]) \n\t" + "gslwrc1 %[u], 0x00(%[uv_ptr]) \n\t" + "punpcklbh %[u], %[u], %[zero] \n\t" + "pshufh %[v], %[u], %[vshu] \n\t" + "pshufh %[u], %[u], %[ushu] \n\t" - "punpcklbh %[y], %[y], %[y] \n\t" - "pmulhuh %[y], %[y], %[yg] \n\t" + "punpcklbh %[y], %[y], %[y] \n\t" + "pmulhuh %[y], %[y], %[yg] \n\t" - "paddsh %[b_vec], %[y], %[bb] \n\t" - "pmullh %[temp], %[u], %[ub] \n\t" - "psubsh %[b_vec], %[b_vec], %[temp] \n\t" - "psrah %[b_vec], %[b_vec], %[six] \n\t" + "paddsh %[b_vec], %[y], %[bb] \n\t" + "pmullh %[temp], %[u], %[ub] \n\t" + "psubsh %[b_vec], %[b_vec], %[temp] \n\t" + "psrah %[b_vec], %[b_vec], %[six] \n\t" - "paddsh %[g_vec], %[y], %[bg] \n\t" - "pmullh %[temp], %[u], %[ug] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "pmullh %[temp], %[v], %[vg] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "psrah %[g_vec], %[g_vec], %[six] \n\t" + "paddsh %[g_vec], %[y], %[bg] \n\t" + "pmullh %[temp], %[u], %[ug] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "pmullh %[temp], %[v], %[vg] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "psrah %[g_vec], %[g_vec], %[six] \n\t" - "paddsh %[r_vec], %[y], %[br] \n\t" - "pmullh %[temp], %[v], %[vr] \n\t" - "psubsh %[r_vec], %[r_vec], %[temp] \n\t" - "psrah %[r_vec], %[r_vec], %[six] \n\t" + "paddsh %[r_vec], %[y], %[br] \n\t" + "pmullh %[temp], %[v], %[vr] \n\t" + "psubsh %[r_vec], %[r_vec], %[temp] \n\t" + "psrah %[r_vec], %[r_vec], %[six] \n\t" - "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" - "packushb %[g_vec], %[g_vec], %[zero] \n\t" - "punpcklbh %[b_vec], %[r_vec], %[g_vec] \n\t" - "punpckhbh %[r_vec], %[r_vec], %[g_vec] \n\t" - "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" - "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[g_vec], %[g_vec], %[zero] \n\t" + "punpcklbh %[b_vec], %[r_vec], %[g_vec] \n\t" + "punpckhbh %[r_vec], %[r_vec], %[g_vec] \n\t" + "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" + "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" - "punpckhwd %[r_vec], %[g_vec], %[g_vec] \n\t" - "psllw %[temp], %[r_vec], %[lmove1] \n\t" - "or %[g_vec], %[g_vec], %[temp] \n\t" - "psrlw %[temp], %[r_vec], %[rmove1] \n\t" - "pextrh %[temp], %[temp], %[zero] \n\t" - "pinsrh_2 %[g_vec], %[g_vec], %[temp] \n\t" - "pextrh %[temp], %[b_vec], %[zero] \n\t" - "pinsrh_3 %[g_vec], %[g_vec], %[temp] \n\t" - "pextrh %[temp], %[b_vec], %[one] \n\t" - "punpckhwd %[b_vec], %[b_vec], %[b_vec] \n\t" - "psllw %[b_vec], %[b_vec], %[rmove1] \n\t" - "or %[b_vec], %[b_vec], %[temp] \n\t" - "gssdlc1 %[g_vec], 0x07(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[g_vec], 0x00(%[rgbbuf_ptr]) \n\t" - "gsswlc1 %[b_vec], 0x0b(%[rgbbuf_ptr]) \n\t" - "gsswrc1 %[b_vec], 0x08(%[rgbbuf_ptr]) \n\t" + "punpckhwd %[r_vec], %[g_vec], %[g_vec] \n\t" + "psllw %[temp], %[r_vec], %[lmove1] \n\t" + "or %[g_vec], %[g_vec], %[temp] \n\t" + "psrlw %[temp], %[r_vec], %[rmove1] \n\t" + "pextrh %[temp], %[temp], %[zero] \n\t" + "pinsrh_2 %[g_vec], %[g_vec], %[temp] \n\t" + "pextrh %[temp], %[b_vec], %[zero] \n\t" + "pinsrh_3 %[g_vec], %[g_vec], %[temp] \n\t" + "pextrh %[temp], %[b_vec], %[one] \n\t" + "punpckhwd %[b_vec], %[b_vec], %[b_vec] \n\t" + "psllw %[b_vec], %[b_vec], %[rmove1] \n\t" + "or %[b_vec], %[b_vec], %[temp] \n\t" + "gssdlc1 %[g_vec], 0x07(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[g_vec], 0x00(%[rgbbuf_ptr]) \n\t" + "gsswlc1 %[b_vec], 0x0b(%[rgbbuf_ptr]) \n\t" + "gsswrc1 %[b_vec], 0x08(%[rgbbuf_ptr]) \n\t" - "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" - "daddiu %[uv_ptr], %[uv_ptr], 0x04 \n\t" - "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x0C \n\t" - "daddi %[width], %[width], -0x04 \n\t" - "bnez %[width], 1b \n\t" + "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" + "daddiu %[uv_ptr], %[uv_ptr], 0x04 \n\t" + "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x0C \n\t" + "daddi %[width], %[width], -0x04 \n\t" + "bnez %[width], 1b \n\t" - : [y]"=&f"(y), [u]"=&f"(u), - [v]"=&f"(v), - [b_vec]"=&f"(b_vec), [g_vec]"=&f"(g_vec), - [r_vec]"=&f"(r_vec), [temp]"=&f"(temp), - [ub]"=&f"(ub), [ug]"=&f"(ug), - [vg]"=&f"(vg), [vr]"=&f"(vr), - [bb]"=&f"(bb), [bg]"=&f"(bg), - [br]"=&f"(br), [yg]"=&f"(yg) - : [y_ptr]"r"(src_y), [uv_ptr]"r"(src_uv), - [rgbbuf_ptr]"r"(rgb_buf), - [yuvcons_ptr]"r"(yuvconstants), [width]"r"(width), - [zero]"f"(0x00), [five]"f"(0x55), - [six]"f"(0x6), [mask1]"f"(0xff00ff00ff00ff00), - [ushu]"f"(0xA0), [vshu]"f"(0xf5), - [alpha]"f"(-1), [lmove1]"f"(0x18), - [one]"f"(0x1), [rmove1]"f"(0x8) - : "memory" - ); + : [y] "=&f"(y), [u] "=&f"(u), [v] "=&f"(v), [b_vec] "=&f"(b_vec), + [g_vec] "=&f"(g_vec), [r_vec] "=&f"(r_vec), [temp] "=&f"(temp), + [ub] "=&f"(ub), [ug] "=&f"(ug), [vg] "=&f"(vg), [vr] "=&f"(vr), + [bb] "=&f"(bb), [bg] "=&f"(bg), [br] "=&f"(br), [yg] "=&f"(yg) + : [y_ptr] "r"(src_y), [uv_ptr] "r"(src_uv), [rgbbuf_ptr] "r"(rgb_buf), + [yuvcons_ptr] "r"(yuvconstants), [width] "r"(width), [zero] "f"(0x00), + [five] "f"(0x55), [six] "f"(0x6), [mask1] "f"(0xff00ff00ff00ff00), + [ushu] "f"(0xA0), [vshu] "f"(0xf5), [alpha] "f"(-1), [lmove1] "f"(0x18), + [one] "f"(0x1), [rmove1] "f"(0x8) + : "memory"); } void NV21ToRGB24Row_MMI(const uint8_t* src_y, @@ -7231,103 +7160,95 @@ void NV21ToRGB24Row_MMI(const uint8_t* src_y, int width) { uint64_t y, u, v; uint64_t b_vec, g_vec, r_vec, temp; - uint64_t ub,ug,vg,vr,bb,bg,br,yg; + uint64_t ub, ug, vg, vr, bb, bg, br, yg; __asm__ volatile( - "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" - "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" - "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" - "or %[ub], %[ub], %[mask1] \n\t" - "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" - "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[ug], %[ug], %[zero] \n\t" - "pshufh %[ug], %[ug], %[zero] \n\t" - "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vg], %[vg], %[zero] \n\t" - "pshufh %[vg], %[vg], %[five] \n\t" - "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" - "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vr], %[vr], %[zero] \n\t" - "pshufh %[vr], %[vr], %[five] \n\t" - "or %[vr], %[vr], %[mask1] \n\t" + "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" + "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" + "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" + "or %[ub], %[ub], %[mask1] \n\t" + "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" + "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[ug], %[ug], %[zero] \n\t" + "pshufh %[ug], %[ug], %[zero] \n\t" + "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vg], %[vg], %[zero] \n\t" + "pshufh %[vg], %[vg], %[five] \n\t" + "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" + "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vr], %[vr], %[zero] \n\t" + "pshufh %[vr], %[vr], %[five] \n\t" + "or %[vr], %[vr], %[mask1] \n\t" - "1: \n\t" - "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" - "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" - "gslwlc1 %[u], 0x03(%[vu_ptr]) \n\t" - "gslwrc1 %[u], 0x00(%[vu_ptr]) \n\t" - "punpcklbh %[u], %[u], %[zero] \n\t" - "pshufh %[v], %[u], %[ushu] \n\t" - "pshufh %[u], %[u], %[vshu] \n\t" + "1: \n\t" + "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" + "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" + "gslwlc1 %[u], 0x03(%[vu_ptr]) \n\t" + "gslwrc1 %[u], 0x00(%[vu_ptr]) \n\t" + "punpcklbh %[u], %[u], %[zero] \n\t" + "pshufh %[v], %[u], %[ushu] \n\t" + "pshufh %[u], %[u], %[vshu] \n\t" - "punpcklbh %[y], %[y], %[y] \n\t" - "pmulhuh %[y], %[y], %[yg] \n\t" + "punpcklbh %[y], %[y], %[y] \n\t" + "pmulhuh %[y], %[y], %[yg] \n\t" - "paddsh %[b_vec], %[y], %[bb] \n\t" - "pmullh %[temp], %[u], %[ub] \n\t" - "psubsh %[b_vec], %[b_vec], %[temp] \n\t" - "psrah %[b_vec], %[b_vec], %[six] \n\t" + "paddsh %[b_vec], %[y], %[bb] \n\t" + "pmullh %[temp], %[u], %[ub] \n\t" + "psubsh %[b_vec], %[b_vec], %[temp] \n\t" + "psrah %[b_vec], %[b_vec], %[six] \n\t" - "paddsh %[g_vec], %[y], %[bg] \n\t" - "pmullh %[temp], %[u], %[ug] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "pmullh %[temp], %[v], %[vg] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "psrah %[g_vec], %[g_vec], %[six] \n\t" + "paddsh %[g_vec], %[y], %[bg] \n\t" + "pmullh %[temp], %[u], %[ug] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "pmullh %[temp], %[v], %[vg] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "psrah %[g_vec], %[g_vec], %[six] \n\t" - "paddsh %[r_vec], %[y], %[br] \n\t" - "pmullh %[temp], %[v], %[vr] \n\t" - "psubsh %[r_vec], %[r_vec], %[temp] \n\t" - "psrah %[r_vec], %[r_vec], %[six] \n\t" + "paddsh %[r_vec], %[y], %[br] \n\t" + "pmullh %[temp], %[v], %[vr] \n\t" + "psubsh %[r_vec], %[r_vec], %[temp] \n\t" + "psrah %[r_vec], %[r_vec], %[six] \n\t" - "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" - "packushb %[g_vec], %[g_vec], %[zero] \n\t" - "punpcklbh %[b_vec], %[r_vec], %[g_vec] \n\t" - "punpckhbh %[r_vec], %[r_vec], %[g_vec] \n\t" - "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" - "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[g_vec], %[g_vec], %[zero] \n\t" + "punpcklbh %[b_vec], %[r_vec], %[g_vec] \n\t" + "punpckhbh %[r_vec], %[r_vec], %[g_vec] \n\t" + "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" + "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" - "punpckhwd %[r_vec], %[g_vec], %[g_vec] \n\t" - "psllw %[temp], %[r_vec], %[lmove1] \n\t" - "or %[g_vec], %[g_vec], %[temp] \n\t" - "psrlw %[temp], %[r_vec], %[rmove1] \n\t" - "pextrh %[temp], %[temp], %[zero] \n\t" - "pinsrh_2 %[g_vec], %[g_vec], %[temp] \n\t" - "pextrh %[temp], %[b_vec], %[zero] \n\t" - "pinsrh_3 %[g_vec], %[g_vec], %[temp] \n\t" - "pextrh %[temp], %[b_vec], %[one] \n\t" - "punpckhwd %[b_vec], %[b_vec], %[b_vec] \n\t" - "psllw %[b_vec], %[b_vec], %[rmove1] \n\t" - "or %[b_vec], %[b_vec], %[temp] \n\t" - "gssdlc1 %[g_vec], 0x07(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[g_vec], 0x00(%[rgbbuf_ptr]) \n\t" - "gsswlc1 %[b_vec], 0x0b(%[rgbbuf_ptr]) \n\t" - "gsswrc1 %[b_vec], 0x08(%[rgbbuf_ptr]) \n\t" + "punpckhwd %[r_vec], %[g_vec], %[g_vec] \n\t" + "psllw %[temp], %[r_vec], %[lmove1] \n\t" + "or %[g_vec], %[g_vec], %[temp] \n\t" + "psrlw %[temp], %[r_vec], %[rmove1] \n\t" + "pextrh %[temp], %[temp], %[zero] \n\t" + "pinsrh_2 %[g_vec], %[g_vec], %[temp] \n\t" + "pextrh %[temp], %[b_vec], %[zero] \n\t" + "pinsrh_3 %[g_vec], %[g_vec], %[temp] \n\t" + "pextrh %[temp], %[b_vec], %[one] \n\t" + "punpckhwd %[b_vec], %[b_vec], %[b_vec] \n\t" + "psllw %[b_vec], %[b_vec], %[rmove1] \n\t" + "or %[b_vec], %[b_vec], %[temp] \n\t" + "gssdlc1 %[g_vec], 0x07(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[g_vec], 0x00(%[rgbbuf_ptr]) \n\t" + "gsswlc1 %[b_vec], 0x0b(%[rgbbuf_ptr]) \n\t" + "gsswrc1 %[b_vec], 0x08(%[rgbbuf_ptr]) \n\t" - "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" - "daddiu %[vu_ptr], %[vu_ptr], 0x04 \n\t" - "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x0C \n\t" - "daddi %[width], %[width], -0x04 \n\t" - "bnez %[width], 1b \n\t" + "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" + "daddiu %[vu_ptr], %[vu_ptr], 0x04 \n\t" + "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x0C \n\t" + "daddi %[width], %[width], -0x04 \n\t" + "bnez %[width], 1b \n\t" - : [y]"=&f"(y), [u]"=&f"(u), - [v]"=&f"(v), - [b_vec]"=&f"(b_vec), [g_vec]"=&f"(g_vec), - [r_vec]"=&f"(r_vec), [temp]"=&f"(temp), - [ub]"=&f"(ub), [ug]"=&f"(ug), - [vg]"=&f"(vg), [vr]"=&f"(vr), - [bb]"=&f"(bb), [bg]"=&f"(bg), - [br]"=&f"(br), [yg]"=&f"(yg) - : [y_ptr]"r"(src_y), [vu_ptr]"r"(src_vu), - [rgbbuf_ptr]"r"(rgb_buf), - [yuvcons_ptr]"r"(yuvconstants), [width]"r"(width), - [zero]"f"(0x00), [five]"f"(0x55), - [six]"f"(0x6), [mask1]"f"(0xff00ff00ff00ff00), - [ushu]"f"(0xA0), [vshu]"f"(0xf5), - [lmove1]"f"(0x18), [rmove1]"f"(0x8), - [one]"f"(0x1) - : "memory" - ); + : [y] "=&f"(y), [u] "=&f"(u), [v] "=&f"(v), [b_vec] "=&f"(b_vec), + [g_vec] "=&f"(g_vec), [r_vec] "=&f"(r_vec), [temp] "=&f"(temp), + [ub] "=&f"(ub), [ug] "=&f"(ug), [vg] "=&f"(vg), [vr] "=&f"(vr), + [bb] "=&f"(bb), [bg] "=&f"(bg), [br] "=&f"(br), [yg] "=&f"(yg) + : [y_ptr] "r"(src_y), [vu_ptr] "r"(src_vu), [rgbbuf_ptr] "r"(rgb_buf), + [yuvcons_ptr] "r"(yuvconstants), [width] "r"(width), [zero] "f"(0x00), + [five] "f"(0x55), [six] "f"(0x6), [mask1] "f"(0xff00ff00ff00ff00), + [ushu] "f"(0xA0), [vshu] "f"(0xf5), [lmove1] "f"(0x18), + [rmove1] "f"(0x8), [one] "f"(0x1) + : "memory"); } void NV12ToRGB565Row_MMI(const uint8_t* src_y, @@ -7337,123 +7258,115 @@ void NV12ToRGB565Row_MMI(const uint8_t* src_y, int width) { uint64_t y, u, v; uint64_t b_vec, g_vec, r_vec, temp; - uint64_t ub,ug,vg,vr,bb,bg,br,yg; + uint64_t ub, ug, vg, vr, bb, bg, br, yg; __asm__ volatile( - "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" - "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" - "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" - "or %[ub], %[ub], %[mask1] \n\t" - "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" - "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[ug], %[ug], %[zero] \n\t" - "pshufh %[ug], %[ug], %[zero] \n\t" - "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vg], %[vg], %[zero] \n\t" - "pshufh %[vg], %[vg], %[five] \n\t" - "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" - "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vr], %[vr], %[zero] \n\t" - "pshufh %[vr], %[vr], %[five] \n\t" - "or %[vr], %[vr], %[mask1] \n\t" + "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" + "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" + "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" + "or %[ub], %[ub], %[mask1] \n\t" + "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" + "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[ug], %[ug], %[zero] \n\t" + "pshufh %[ug], %[ug], %[zero] \n\t" + "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vg], %[vg], %[zero] \n\t" + "pshufh %[vg], %[vg], %[five] \n\t" + "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" + "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vr], %[vr], %[zero] \n\t" + "pshufh %[vr], %[vr], %[five] \n\t" + "or %[vr], %[vr], %[mask1] \n\t" - "1: \n\t" - "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" - "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" - "gslwlc1 %[u], 0x03(%[uv_ptr]) \n\t" - "gslwrc1 %[u], 0x00(%[uv_ptr]) \n\t" - "punpcklbh %[u], %[u], %[zero] \n\t" - "pshufh %[v], %[u], %[vshu] \n\t" - "pshufh %[u], %[u], %[ushu] \n\t" + "1: \n\t" + "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" + "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" + "gslwlc1 %[u], 0x03(%[uv_ptr]) \n\t" + "gslwrc1 %[u], 0x00(%[uv_ptr]) \n\t" + "punpcklbh %[u], %[u], %[zero] \n\t" + "pshufh %[v], %[u], %[vshu] \n\t" + "pshufh %[u], %[u], %[ushu] \n\t" - "punpcklbh %[y], %[y], %[y] \n\t" - "pmulhuh %[y], %[y], %[yg] \n\t" + "punpcklbh %[y], %[y], %[y] \n\t" + "pmulhuh %[y], %[y], %[yg] \n\t" - "paddsh %[b_vec], %[y], %[bb] \n\t" - "pmullh %[temp], %[u], %[ub] \n\t" - "psubsh %[b_vec], %[b_vec], %[temp] \n\t" - "psrah %[b_vec], %[b_vec], %[six] \n\t" + "paddsh %[b_vec], %[y], %[bb] \n\t" + "pmullh %[temp], %[u], %[ub] \n\t" + "psubsh %[b_vec], %[b_vec], %[temp] \n\t" + "psrah %[b_vec], %[b_vec], %[six] \n\t" - "paddsh %[g_vec], %[y], %[bg] \n\t" - "pmullh %[temp], %[u], %[ug] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "pmullh %[temp], %[v], %[vg] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "psrah %[g_vec], %[g_vec], %[six] \n\t" + "paddsh %[g_vec], %[y], %[bg] \n\t" + "pmullh %[temp], %[u], %[ug] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "pmullh %[temp], %[v], %[vg] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "psrah %[g_vec], %[g_vec], %[six] \n\t" - "paddsh %[r_vec], %[y], %[br] \n\t" - "pmullh %[temp], %[v], %[vr] \n\t" - "psubsh %[r_vec], %[r_vec], %[temp] \n\t" - "psrah %[r_vec], %[r_vec], %[six] \n\t" + "paddsh %[r_vec], %[y], %[br] \n\t" + "pmullh %[temp], %[v], %[vr] \n\t" + "psubsh %[r_vec], %[r_vec], %[temp] \n\t" + "psrah %[r_vec], %[r_vec], %[six] \n\t" - "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" - "packushb %[g_vec], %[g_vec], %[zero] \n\t" - "punpcklbh %[b_vec], %[r_vec], %[g_vec] \n\t" - "punpckhbh %[r_vec], %[r_vec], %[g_vec] \n\t" - "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" - "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[g_vec], %[g_vec], %[zero] \n\t" + "punpcklbh %[b_vec], %[r_vec], %[g_vec] \n\t" + "punpckhbh %[r_vec], %[r_vec], %[g_vec] \n\t" + "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" + "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" - "psrlh %[temp], %[g_vec], %[three] \n\t" - "and %[g_vec], %[temp], %[mask2] \n\t" - "psrlw %[temp], %[temp], %[seven] \n\t" - "psrlw %[r_vec], %[mask1], %[eight] \n\t" - "and %[r_vec], %[temp], %[r_vec] \n\t" - "psubb %[y], %[eight], %[three] \n\t"//5 - "psllw %[r_vec], %[r_vec], %[y] \n\t" - "or %[g_vec], %[g_vec], %[r_vec] \n\t" - "paddb %[r_vec], %[three], %[six] \n\t" - "psrlw %[temp], %[temp], %[r_vec] \n\t" - "and %[r_vec], %[temp], %[mask2] \n\t" - "paddb %[temp], %[three], %[eight] \n\t" - "psllw %[r_vec], %[r_vec], %[temp] \n\t" - "or %[g_vec], %[g_vec], %[r_vec] \n\t" + "psrlh %[temp], %[g_vec], %[three] \n\t" + "and %[g_vec], %[temp], %[mask2] \n\t" + "psrlw %[temp], %[temp], %[seven] \n\t" + "psrlw %[r_vec], %[mask1], %[eight] \n\t" + "and %[r_vec], %[temp], %[r_vec] \n\t" + "psubb %[y], %[eight], %[three] \n\t" // 5 + "psllw %[r_vec], %[r_vec], %[y] \n\t" + "or %[g_vec], %[g_vec], %[r_vec] \n\t" + "paddb %[r_vec], %[three], %[six] \n\t" + "psrlw %[temp], %[temp], %[r_vec] \n\t" + "and %[r_vec], %[temp], %[mask2] \n\t" + "paddb %[temp], %[three], %[eight] \n\t" + "psllw %[r_vec], %[r_vec], %[temp] \n\t" + "or %[g_vec], %[g_vec], %[r_vec] \n\t" - "psrlh %[temp], %[b_vec], %[three] \n\t" - "and %[b_vec], %[temp], %[mask2] \n\t" - "psrlw %[temp], %[temp], %[seven] \n\t" - "psrlw %[r_vec], %[mask1], %[eight] \n\t" - "and %[r_vec], %[temp], %[r_vec] \n\t" - "psubb %[y], %[eight], %[three] \n\t"//5 - "psllw %[r_vec], %[r_vec], %[y] \n\t" - "or %[b_vec], %[b_vec], %[r_vec] \n\t" - "paddb %[r_vec], %[three], %[six] \n\t" - "psrlw %[temp], %[temp], %[r_vec] \n\t" - "and %[r_vec], %[temp], %[mask2] \n\t" - "paddb %[temp], %[three], %[eight] \n\t" - "psllw %[r_vec], %[r_vec], %[temp] \n\t" - "or %[b_vec], %[b_vec], %[r_vec] \n\t" + "psrlh %[temp], %[b_vec], %[three] \n\t" + "and %[b_vec], %[temp], %[mask2] \n\t" + "psrlw %[temp], %[temp], %[seven] \n\t" + "psrlw %[r_vec], %[mask1], %[eight] \n\t" + "and %[r_vec], %[temp], %[r_vec] \n\t" + "psubb %[y], %[eight], %[three] \n\t" // 5 + "psllw %[r_vec], %[r_vec], %[y] \n\t" + "or %[b_vec], %[b_vec], %[r_vec] \n\t" + "paddb %[r_vec], %[three], %[six] \n\t" + "psrlw %[temp], %[temp], %[r_vec] \n\t" + "and %[r_vec], %[temp], %[mask2] \n\t" + "paddb %[temp], %[three], %[eight] \n\t" + "psllw %[r_vec], %[r_vec], %[temp] \n\t" + "or %[b_vec], %[b_vec], %[r_vec] \n\t" - "punpcklhw %[r_vec], %[g_vec], %[b_vec] \n\t" - "punpckhhw %[b_vec], %[g_vec], %[b_vec] \n\t" - "punpcklhw %[g_vec], %[r_vec], %[b_vec] \n\t" + "punpcklhw %[r_vec], %[g_vec], %[b_vec] \n\t" + "punpckhhw %[b_vec], %[g_vec], %[b_vec] \n\t" + "punpcklhw %[g_vec], %[r_vec], %[b_vec] \n\t" - "gssdlc1 %[g_vec], 0x07(%[dst_rgb565]) \n\t" - "gssdrc1 %[g_vec], 0x00(%[dst_rgb565]) \n\t" + "gssdlc1 %[g_vec], 0x07(%[dst_rgb565]) \n\t" + "gssdrc1 %[g_vec], 0x00(%[dst_rgb565]) \n\t" - "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" - "daddiu %[uv_ptr], %[uv_ptr], 0x04 \n\t" - "daddiu %[dst_rgb565], %[dst_rgb565], 0x08 \n\t" - "daddi %[width], %[width], -0x04 \n\t" - "bnez %[width], 1b \n\t" + "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" + "daddiu %[uv_ptr], %[uv_ptr], 0x04 \n\t" + "daddiu %[dst_rgb565], %[dst_rgb565], 0x08 \n\t" + "daddi %[width], %[width], -0x04 \n\t" + "bnez %[width], 1b \n\t" - : [y]"=&f"(y), [u]"=&f"(u), - [v]"=&f"(v), - [b_vec]"=&f"(b_vec), [g_vec]"=&f"(g_vec), - [r_vec]"=&f"(r_vec), [temp]"=&f"(temp), - [ub]"=&f"(ub), [ug]"=&f"(ug), - [vg]"=&f"(vg), [vr]"=&f"(vr), - [bb]"=&f"(bb), [bg]"=&f"(bg), - [br]"=&f"(br), [yg]"=&f"(yg) - : [y_ptr]"r"(src_y), [uv_ptr]"r"(src_uv), - [dst_rgb565]"r"(dst_rgb565), - [yuvcons_ptr]"r"(yuvconstants), [width]"r"(width), - [zero]"f"(0x00), [five]"f"(0x55), - [six]"f"(0x6), [mask1]"f"(0xff00ff00ff00ff00), - [ushu]"f"(0xA0), [vshu]"f"(0xf5), - [three]"f"(0x3), [mask2]"f"(0x1f0000001f), - [eight]"f"(0x8), [seven]"f"(0x7) - : "memory" - ); + : [y] "=&f"(y), [u] "=&f"(u), [v] "=&f"(v), [b_vec] "=&f"(b_vec), + [g_vec] "=&f"(g_vec), [r_vec] "=&f"(r_vec), [temp] "=&f"(temp), + [ub] "=&f"(ub), [ug] "=&f"(ug), [vg] "=&f"(vg), [vr] "=&f"(vr), + [bb] "=&f"(bb), [bg] "=&f"(bg), [br] "=&f"(br), [yg] "=&f"(yg) + : [y_ptr] "r"(src_y), [uv_ptr] "r"(src_uv), [dst_rgb565] "r"(dst_rgb565), + [yuvcons_ptr] "r"(yuvconstants), [width] "r"(width), [zero] "f"(0x00), + [five] "f"(0x55), [six] "f"(0x6), [mask1] "f"(0xff00ff00ff00ff00), + [ushu] "f"(0xA0), [vshu] "f"(0xf5), [three] "f"(0x3), + [mask2] "f"(0x1f0000001f), [eight] "f"(0x8), [seven] "f"(0x7) + : "memory"); } void YUY2ToARGBRow_MMI(const uint8_t* src_yuy2, @@ -7462,90 +7375,83 @@ void YUY2ToARGBRow_MMI(const uint8_t* src_yuy2, int width) { uint64_t y, u, v; uint64_t b_vec, g_vec, r_vec, temp; - uint64_t ub,ug,vg,vr,bb,bg,br,yg; + uint64_t ub, ug, vg, vr, bb, bg, br, yg; __asm__ volatile( - "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" - "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" - "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" - "or %[ub], %[ub], %[mask1] \n\t" - "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" - "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[ug], %[ug], %[zero] \n\t" - "pshufh %[ug], %[ug], %[zero] \n\t" - "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vg], %[vg], %[zero] \n\t" - "pshufh %[vg], %[vg], %[five] \n\t" - "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" - "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vr], %[vr], %[zero] \n\t" - "pshufh %[vr], %[vr], %[five] \n\t" - "or %[vr], %[vr], %[mask1] \n\t" + "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" + "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" + "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" + "or %[ub], %[ub], %[mask1] \n\t" + "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" + "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[ug], %[ug], %[zero] \n\t" + "pshufh %[ug], %[ug], %[zero] \n\t" + "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vg], %[vg], %[zero] \n\t" + "pshufh %[vg], %[vg], %[five] \n\t" + "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" + "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vr], %[vr], %[zero] \n\t" + "pshufh %[vr], %[vr], %[five] \n\t" + "or %[vr], %[vr], %[mask1] \n\t" - "1: \n\t" - "gsldlc1 %[y], 0x07(%[yuy2_ptr]) \n\t" - "gsldrc1 %[y], 0x00(%[yuy2_ptr]) \n\t" - "psrlh %[temp], %[y], %[eight] \n\t" - "pshufh %[u], %[temp], %[ushu] \n\t" - "pshufh %[v], %[temp], %[vshu] \n\t" + "1: \n\t" + "gsldlc1 %[y], 0x07(%[yuy2_ptr]) \n\t" + "gsldrc1 %[y], 0x00(%[yuy2_ptr]) \n\t" + "psrlh %[temp], %[y], %[eight] \n\t" + "pshufh %[u], %[temp], %[ushu] \n\t" + "pshufh %[v], %[temp], %[vshu] \n\t" - "psrlh %[temp], %[mask1], %[eight] \n\t" - "and %[y], %[y], %[temp] \n\t" - "psllh %[temp], %[y], %[eight] \n\t" - "or %[y], %[y], %[temp] \n\t" - "pmulhuh %[y], %[y], %[yg] \n\t" + "psrlh %[temp], %[mask1], %[eight] \n\t" + "and %[y], %[y], %[temp] \n\t" + "psllh %[temp], %[y], %[eight] \n\t" + "or %[y], %[y], %[temp] \n\t" + "pmulhuh %[y], %[y], %[yg] \n\t" - "paddsh %[b_vec], %[y], %[bb] \n\t" - "pmullh %[temp], %[u], %[ub] \n\t" - "psubsh %[b_vec], %[b_vec], %[temp] \n\t" - "psrah %[b_vec], %[b_vec], %[six] \n\t" + "paddsh %[b_vec], %[y], %[bb] \n\t" + "pmullh %[temp], %[u], %[ub] \n\t" + "psubsh %[b_vec], %[b_vec], %[temp] \n\t" + "psrah %[b_vec], %[b_vec], %[six] \n\t" - "paddsh %[g_vec], %[y], %[bg] \n\t" - "pmullh %[temp], %[u], %[ug] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "pmullh %[temp], %[v], %[vg] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "psrah %[g_vec], %[g_vec], %[six] \n\t" + "paddsh %[g_vec], %[y], %[bg] \n\t" + "pmullh %[temp], %[u], %[ug] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "pmullh %[temp], %[v], %[vg] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "psrah %[g_vec], %[g_vec], %[six] \n\t" - "paddsh %[r_vec], %[y], %[br] \n\t" - "pmullh %[temp], %[v], %[vr] \n\t" - "psubsh %[r_vec], %[r_vec], %[temp] \n\t" - "psrah %[r_vec], %[r_vec], %[six] \n\t" + "paddsh %[r_vec], %[y], %[br] \n\t" + "pmullh %[temp], %[v], %[vr] \n\t" + "psubsh %[r_vec], %[r_vec], %[temp] \n\t" + "psrah %[r_vec], %[r_vec], %[six] \n\t" - "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" - "packushb %[g_vec], %[g_vec], %[zero] \n\t" - "punpcklwd %[g_vec], %[g_vec], %[alpha] \n\t" - "punpcklbh %[b_vec], %[r_vec], %[g_vec] \n\t" - "punpckhbh %[r_vec], %[r_vec], %[g_vec] \n\t" - "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" - "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[g_vec], %[g_vec], %[zero] \n\t" + "punpcklwd %[g_vec], %[g_vec], %[alpha] \n\t" + "punpcklbh %[b_vec], %[r_vec], %[g_vec] \n\t" + "punpckhbh %[r_vec], %[r_vec], %[g_vec] \n\t" + "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" + "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" - "gssdlc1 %[g_vec], 0x07(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[g_vec], 0x00(%[rgbbuf_ptr]) \n\t" - "gssdlc1 %[b_vec], 0x0f(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[b_vec], 0x08(%[rgbbuf_ptr]) \n\t" + "gssdlc1 %[g_vec], 0x07(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[g_vec], 0x00(%[rgbbuf_ptr]) \n\t" + "gssdlc1 %[b_vec], 0x0f(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[b_vec], 0x08(%[rgbbuf_ptr]) \n\t" - "daddiu %[yuy2_ptr], %[yuy2_ptr], 0x08 \n\t" - "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x10 \n\t" - "daddi %[width], %[width], -0x04 \n\t" - "bnez %[width], 1b \n\t" + "daddiu %[yuy2_ptr], %[yuy2_ptr], 0x08 \n\t" + "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x10 \n\t" + "daddi %[width], %[width], -0x04 \n\t" + "bnez %[width], 1b \n\t" - : [y]"=&f"(y), [u]"=&f"(u), - [v]"=&f"(v), - [b_vec]"=&f"(b_vec), [g_vec]"=&f"(g_vec), - [r_vec]"=&f"(r_vec), [temp]"=&f"(temp), - [ub]"=&f"(ub), [ug]"=&f"(ug), - [vg]"=&f"(vg), [vr]"=&f"(vr), - [bb]"=&f"(bb), [bg]"=&f"(bg), - [br]"=&f"(br), [yg]"=&f"(yg) - : [yuy2_ptr]"r"(src_yuy2), [rgbbuf_ptr]"r"(rgb_buf), - [yuvcons_ptr]"r"(yuvconstants), [width]"r"(width), - [zero]"f"(0x00), [five]"f"(0x55), - [six]"f"(0x6), [mask1]"f"(0xff00ff00ff00ff00), - [ushu]"f"(0xA0), [vshu]"f"(0xf5), - [alpha]"f"(-1), [eight]"f"(0x8) - : "memory" - ); + : [y] "=&f"(y), [u] "=&f"(u), [v] "=&f"(v), [b_vec] "=&f"(b_vec), + [g_vec] "=&f"(g_vec), [r_vec] "=&f"(r_vec), [temp] "=&f"(temp), + [ub] "=&f"(ub), [ug] "=&f"(ug), [vg] "=&f"(vg), [vr] "=&f"(vr), + [bb] "=&f"(bb), [bg] "=&f"(bg), [br] "=&f"(br), [yg] "=&f"(yg) + : [yuy2_ptr] "r"(src_yuy2), [rgbbuf_ptr] "r"(rgb_buf), + [yuvcons_ptr] "r"(yuvconstants), [width] "r"(width), [zero] "f"(0x00), + [five] "f"(0x55), [six] "f"(0x6), [mask1] "f"(0xff00ff00ff00ff00), + [ushu] "f"(0xA0), [vshu] "f"(0xf5), [alpha] "f"(-1), [eight] "f"(0x8) + : "memory"); } void UYVYToARGBRow_MMI(const uint8_t* src_uyvy, @@ -7554,90 +7460,83 @@ void UYVYToARGBRow_MMI(const uint8_t* src_uyvy, int width) { uint64_t y, u, v; uint64_t b_vec, g_vec, r_vec, temp; - uint64_t ub,ug,vg,vr,bb,bg,br,yg; + uint64_t ub, ug, vg, vr, bb, bg, br, yg; __asm__ volatile( - "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" - "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" - "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" - "or %[ub], %[ub], %[mask1] \n\t" - "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" - "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[ug], %[ug], %[zero] \n\t" - "pshufh %[ug], %[ug], %[zero] \n\t" - "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vg], %[vg], %[zero] \n\t" - "pshufh %[vg], %[vg], %[five] \n\t" - "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" - "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vr], %[vr], %[zero] \n\t" - "pshufh %[vr], %[vr], %[five] \n\t" - "or %[vr], %[vr], %[mask1] \n\t" + "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" + "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" + "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" + "or %[ub], %[ub], %[mask1] \n\t" + "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" + "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[ug], %[ug], %[zero] \n\t" + "pshufh %[ug], %[ug], %[zero] \n\t" + "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vg], %[vg], %[zero] \n\t" + "pshufh %[vg], %[vg], %[five] \n\t" + "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" + "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vr], %[vr], %[zero] \n\t" + "pshufh %[vr], %[vr], %[five] \n\t" + "or %[vr], %[vr], %[mask1] \n\t" - "1: \n\t" - "gsldlc1 %[y], 0x07(%[uyvy_ptr]) \n\t" - "gsldrc1 %[y], 0x00(%[uyvy_ptr]) \n\t" - "psrlh %[temp], %[mask1], %[eight] \n\t" - "and %[temp], %[y], %[temp] \n\t" - "pshufh %[u], %[temp], %[ushu] \n\t" - "pshufh %[v], %[temp], %[vshu] \n\t" + "1: \n\t" + "gsldlc1 %[y], 0x07(%[uyvy_ptr]) \n\t" + "gsldrc1 %[y], 0x00(%[uyvy_ptr]) \n\t" + "psrlh %[temp], %[mask1], %[eight] \n\t" + "and %[temp], %[y], %[temp] \n\t" + "pshufh %[u], %[temp], %[ushu] \n\t" + "pshufh %[v], %[temp], %[vshu] \n\t" - "psrlh %[y], %[y], %[eight] \n\t" - "psllh %[temp], %[y], %[eight] \n\t" - "or %[y], %[y], %[temp] \n\t" - "pmulhuh %[y], %[y], %[yg] \n\t" + "psrlh %[y], %[y], %[eight] \n\t" + "psllh %[temp], %[y], %[eight] \n\t" + "or %[y], %[y], %[temp] \n\t" + "pmulhuh %[y], %[y], %[yg] \n\t" - "paddsh %[b_vec], %[y], %[bb] \n\t" - "pmullh %[temp], %[u], %[ub] \n\t" - "psubsh %[b_vec], %[b_vec], %[temp] \n\t" - "psrah %[b_vec], %[b_vec], %[six] \n\t" + "paddsh %[b_vec], %[y], %[bb] \n\t" + "pmullh %[temp], %[u], %[ub] \n\t" + "psubsh %[b_vec], %[b_vec], %[temp] \n\t" + "psrah %[b_vec], %[b_vec], %[six] \n\t" - "paddsh %[g_vec], %[y], %[bg] \n\t" - "pmullh %[temp], %[u], %[ug] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "pmullh %[temp], %[v], %[vg] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "psrah %[g_vec], %[g_vec], %[six] \n\t" + "paddsh %[g_vec], %[y], %[bg] \n\t" + "pmullh %[temp], %[u], %[ug] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "pmullh %[temp], %[v], %[vg] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "psrah %[g_vec], %[g_vec], %[six] \n\t" - "paddsh %[r_vec], %[y], %[br] \n\t" - "pmullh %[temp], %[v], %[vr] \n\t" - "psubsh %[r_vec], %[r_vec], %[temp] \n\t" - "psrah %[r_vec], %[r_vec], %[six] \n\t" + "paddsh %[r_vec], %[y], %[br] \n\t" + "pmullh %[temp], %[v], %[vr] \n\t" + "psubsh %[r_vec], %[r_vec], %[temp] \n\t" + "psrah %[r_vec], %[r_vec], %[six] \n\t" - "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" - "packushb %[g_vec], %[g_vec], %[zero] \n\t" - "punpcklwd %[g_vec], %[g_vec], %[alpha] \n\t" - "punpcklbh %[b_vec], %[r_vec], %[g_vec] \n\t" - "punpckhbh %[r_vec], %[r_vec], %[g_vec] \n\t" - "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" - "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[g_vec], %[g_vec], %[zero] \n\t" + "punpcklwd %[g_vec], %[g_vec], %[alpha] \n\t" + "punpcklbh %[b_vec], %[r_vec], %[g_vec] \n\t" + "punpckhbh %[r_vec], %[r_vec], %[g_vec] \n\t" + "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" + "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" - "gssdlc1 %[g_vec], 0x07(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[g_vec], 0x00(%[rgbbuf_ptr]) \n\t" - "gssdlc1 %[b_vec], 0x0f(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[b_vec], 0x08(%[rgbbuf_ptr]) \n\t" + "gssdlc1 %[g_vec], 0x07(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[g_vec], 0x00(%[rgbbuf_ptr]) \n\t" + "gssdlc1 %[b_vec], 0x0f(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[b_vec], 0x08(%[rgbbuf_ptr]) \n\t" - "daddiu %[uyvy_ptr], %[uyvy_ptr], 0x08 \n\t" - "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x10 \n\t" - "daddi %[width], %[width], -0x04 \n\t" - "bnez %[width], 1b \n\t" + "daddiu %[uyvy_ptr], %[uyvy_ptr], 0x08 \n\t" + "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x10 \n\t" + "daddi %[width], %[width], -0x04 \n\t" + "bnez %[width], 1b \n\t" - : [y]"=&f"(y), [u]"=&f"(u), - [v]"=&f"(v), - [b_vec]"=&f"(b_vec), [g_vec]"=&f"(g_vec), - [r_vec]"=&f"(r_vec), [temp]"=&f"(temp), - [ub]"=&f"(ub), [ug]"=&f"(ug), - [vg]"=&f"(vg), [vr]"=&f"(vr), - [bb]"=&f"(bb), [bg]"=&f"(bg), - [br]"=&f"(br), [yg]"=&f"(yg) - : [uyvy_ptr]"r"(src_uyvy), [rgbbuf_ptr]"r"(rgb_buf), - [yuvcons_ptr]"r"(yuvconstants), [width]"r"(width), - [zero]"f"(0x00), [five]"f"(0x55), - [six]"f"(0x6), [mask1]"f"(0xff00ff00ff00ff00), - [ushu]"f"(0xA0), [vshu]"f"(0xf5), - [alpha]"f"(-1), [eight]"f"(0x8) - : "memory" - ); + : [y] "=&f"(y), [u] "=&f"(u), [v] "=&f"(v), [b_vec] "=&f"(b_vec), + [g_vec] "=&f"(g_vec), [r_vec] "=&f"(r_vec), [temp] "=&f"(temp), + [ub] "=&f"(ub), [ug] "=&f"(ug), [vg] "=&f"(vg), [vr] "=&f"(vr), + [bb] "=&f"(bb), [bg] "=&f"(bg), [br] "=&f"(br), [yg] "=&f"(yg) + : [uyvy_ptr] "r"(src_uyvy), [rgbbuf_ptr] "r"(rgb_buf), + [yuvcons_ptr] "r"(yuvconstants), [width] "r"(width), [zero] "f"(0x00), + [five] "f"(0x55), [six] "f"(0x6), [mask1] "f"(0xff00ff00ff00ff00), + [ushu] "f"(0xA0), [vshu] "f"(0xf5), [alpha] "f"(-1), [eight] "f"(0x8) + : "memory"); } void I422ToRGBARow_MMI(const uint8_t* src_y, @@ -7648,112 +7547,104 @@ void I422ToRGBARow_MMI(const uint8_t* src_y, int width) { uint64_t y, u, v; uint64_t b_vec, g_vec, r_vec, temp; - uint64_t ub,ug,vg,vr,bb,bg,br,yg; + uint64_t ub, ug, vg, vr, bb, bg, br, yg; __asm__ volatile( - "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" - "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" - "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" - "or %[ub], %[ub], %[mask1] \n\t" - "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" - "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[ug], %[ug], %[zero] \n\t" - "pshufh %[ug], %[ug], %[zero] \n\t" - "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vg], %[vg], %[zero] \n\t" - "pshufh %[vg], %[vg], %[five] \n\t" - "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" - "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" - "punpcklbh %[vr], %[vr], %[zero] \n\t" - "pshufh %[vr], %[vr], %[five] \n\t" - "or %[vr], %[vr], %[mask1] \n\t" + "ldc1 %[yg], 0xc0(%[yuvcons_ptr]) \n\t" + "ldc1 %[bb], 0x60(%[yuvcons_ptr]) \n\t" + "ldc1 %[ub], 0x00(%[yuvcons_ptr]) \n\t" + "or %[ub], %[ub], %[mask1] \n\t" + "ldc1 %[bg], 0x80(%[yuvcons_ptr]) \n\t" + "ldc1 %[ug], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[ug], %[ug], %[zero] \n\t" + "pshufh %[ug], %[ug], %[zero] \n\t" + "ldc1 %[vg], 0x20(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vg], %[vg], %[zero] \n\t" + "pshufh %[vg], %[vg], %[five] \n\t" + "ldc1 %[br], 0xa0(%[yuvcons_ptr]) \n\t" + "ldc1 %[vr], 0x40(%[yuvcons_ptr]) \n\t" + "punpcklbh %[vr], %[vr], %[zero] \n\t" + "pshufh %[vr], %[vr], %[five] \n\t" + "or %[vr], %[vr], %[mask1] \n\t" - "1: \n\t" - "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" - "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" - "gslwlc1 %[u], 0x03(%[u_ptr]) \n\t" - "gslwrc1 %[u], 0x00(%[u_ptr]) \n\t" - "gslwlc1 %[v], 0x03(%[v_ptr]) \n\t" - "gslwrc1 %[v], 0x00(%[v_ptr]) \n\t" + "1: \n\t" + "gslwlc1 %[y], 0x03(%[y_ptr]) \n\t" + "gslwrc1 %[y], 0x00(%[y_ptr]) \n\t" + "gslwlc1 %[u], 0x03(%[u_ptr]) \n\t" + "gslwrc1 %[u], 0x00(%[u_ptr]) \n\t" + "gslwlc1 %[v], 0x03(%[v_ptr]) \n\t" + "gslwrc1 %[v], 0x00(%[v_ptr]) \n\t" - "punpcklbh %[y], %[y], %[y] \n\t" - "pmulhuh %[y], %[y], %[yg] \n\t" + "punpcklbh %[y], %[y], %[y] \n\t" + "pmulhuh %[y], %[y], %[yg] \n\t" - "punpcklbh %[u], %[u], %[u] \n\t" - "punpcklbh %[u], %[u], %[zero] \n\t" - "paddsh %[b_vec], %[y], %[bb] \n\t" - "pmullh %[temp], %[u], %[ub] \n\t" - "psubsh %[b_vec], %[b_vec], %[temp] \n\t" - "psrah %[b_vec], %[b_vec], %[six] \n\t" + "punpcklbh %[u], %[u], %[u] \n\t" + "punpcklbh %[u], %[u], %[zero] \n\t" + "paddsh %[b_vec], %[y], %[bb] \n\t" + "pmullh %[temp], %[u], %[ub] \n\t" + "psubsh %[b_vec], %[b_vec], %[temp] \n\t" + "psrah %[b_vec], %[b_vec], %[six] \n\t" - "punpcklbh %[v], %[v], %[v] \n\t" - "punpcklbh %[v], %[v], %[zero] \n\t" - "paddsh %[g_vec], %[y], %[bg] \n\t" - "pmullh %[temp], %[u], %[ug] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "pmullh %[temp], %[v], %[vg] \n\t" - "psubsh %[g_vec], %[g_vec], %[temp] \n\t" - "psrah %[g_vec], %[g_vec], %[six] \n\t" + "punpcklbh %[v], %[v], %[v] \n\t" + "punpcklbh %[v], %[v], %[zero] \n\t" + "paddsh %[g_vec], %[y], %[bg] \n\t" + "pmullh %[temp], %[u], %[ug] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "pmullh %[temp], %[v], %[vg] \n\t" + "psubsh %[g_vec], %[g_vec], %[temp] \n\t" + "psrah %[g_vec], %[g_vec], %[six] \n\t" - "paddsh %[r_vec], %[y], %[br] \n\t" - "pmullh %[temp], %[v], %[vr] \n\t" - "psubsh %[r_vec], %[r_vec], %[temp] \n\t" - "psrah %[r_vec], %[r_vec], %[six] \n\t" + "paddsh %[r_vec], %[y], %[br] \n\t" + "pmullh %[temp], %[v], %[vr] \n\t" + "psubsh %[r_vec], %[r_vec], %[temp] \n\t" + "psrah %[r_vec], %[r_vec], %[six] \n\t" - "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" - "packushb %[g_vec], %[g_vec], %[zero] \n\t" - "punpcklwd %[g_vec], %[alpha], %[g_vec] \n\t" - "punpcklbh %[b_vec], %[g_vec], %[r_vec] \n\t" - "punpckhbh %[r_vec], %[g_vec], %[r_vec] \n\t" - "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" - "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[r_vec], %[b_vec], %[r_vec] \n\t" + "packushb %[g_vec], %[g_vec], %[zero] \n\t" + "punpcklwd %[g_vec], %[alpha], %[g_vec] \n\t" + "punpcklbh %[b_vec], %[g_vec], %[r_vec] \n\t" + "punpckhbh %[r_vec], %[g_vec], %[r_vec] \n\t" + "punpcklhw %[g_vec], %[b_vec], %[r_vec] \n\t" + "punpckhhw %[b_vec], %[b_vec], %[r_vec] \n\t" - "gssdlc1 %[g_vec], 0x07(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[g_vec], 0x00(%[rgbbuf_ptr]) \n\t" - "gssdlc1 %[b_vec], 0x0f(%[rgbbuf_ptr]) \n\t" - "gssdrc1 %[b_vec], 0x08(%[rgbbuf_ptr]) \n\t" + "gssdlc1 %[g_vec], 0x07(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[g_vec], 0x00(%[rgbbuf_ptr]) \n\t" + "gssdlc1 %[b_vec], 0x0f(%[rgbbuf_ptr]) \n\t" + "gssdrc1 %[b_vec], 0x08(%[rgbbuf_ptr]) \n\t" - "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" - "daddiu %[u_ptr], %[u_ptr], 0x02 \n\t" - "daddiu %[v_ptr], %[v_ptr], 0x02 \n\t" - "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x10 \n\t" - "daddi %[width], %[width], -0x04 \n\t" - "bnez %[width], 1b \n\t" + "daddiu %[y_ptr], %[y_ptr], 0x04 \n\t" + "daddiu %[u_ptr], %[u_ptr], 0x02 \n\t" + "daddiu %[v_ptr], %[v_ptr], 0x02 \n\t" + "daddiu %[rgbbuf_ptr], %[rgbbuf_ptr], 0x10 \n\t" + "daddi %[width], %[width], -0x04 \n\t" + "bnez %[width], 1b \n\t" - : [y]"=&f"(y), [u]"=&f"(u), - [v]"=&f"(v), - [b_vec]"=&f"(b_vec), [g_vec]"=&f"(g_vec), - [r_vec]"=&f"(r_vec), [temp]"=&f"(temp), - [ub]"=&f"(ub), [ug]"=&f"(ug), - [vg]"=&f"(vg), [vr]"=&f"(vr), - [bb]"=&f"(bb), [bg]"=&f"(bg), - [br]"=&f"(br), [yg]"=&f"(yg) - : [y_ptr]"r"(src_y), [u_ptr]"r"(src_u), - [v_ptr]"r"(src_v), [rgbbuf_ptr]"r"(rgb_buf), - [yuvcons_ptr]"r"(yuvconstants), [width]"r"(width), - [zero]"f"(0x00), [five]"f"(0x55), - [six]"f"(0x6), [mask1]"f"(0xff00ff00ff00ff00), - [alpha]"f"(-1) - : "memory" - ); + : [y] "=&f"(y), [u] "=&f"(u), [v] "=&f"(v), [b_vec] "=&f"(b_vec), + [g_vec] "=&f"(g_vec), [r_vec] "=&f"(r_vec), [temp] "=&f"(temp), + [ub] "=&f"(ub), [ug] "=&f"(ug), [vg] "=&f"(vg), [vr] "=&f"(vr), + [bb] "=&f"(bb), [bg] "=&f"(bg), [br] "=&f"(br), [yg] "=&f"(yg) + : [y_ptr] "r"(src_y), [u_ptr] "r"(src_u), [v_ptr] "r"(src_v), + [rgbbuf_ptr] "r"(rgb_buf), [yuvcons_ptr] "r"(yuvconstants), + [width] "r"(width), [zero] "f"(0x00), [five] "f"(0x55), [six] "f"(0x6), + [mask1] "f"(0xff00ff00ff00ff00), [alpha] "f"(-1) + : "memory"); } void ARGBSetRow_MMI(uint8_t* dst_argb, uint32_t v32, int width) { - __asm__ volatile ( - "punpcklwd %[v32], %[v32], %[v32] \n\t" - "1: \n\t" - "gssdlc1 %[v32], 0x07(%[dst_ptr]) \n\t" - "gssdrc1 %[v32], 0x00(%[dst_ptr]) \n\t" - "gssdlc1 %[v32], 0x0f(%[dst_ptr]) \n\t" - "gssdrc1 %[v32], 0x08(%[dst_ptr]) \n\t" + __asm__ volatile( + "punpcklwd %[v32], %[v32], %[v32] \n\t" + "1: \n\t" + "gssdlc1 %[v32], 0x07(%[dst_ptr]) \n\t" + "gssdrc1 %[v32], 0x00(%[dst_ptr]) \n\t" + "gssdlc1 %[v32], 0x0f(%[dst_ptr]) \n\t" + "gssdrc1 %[v32], 0x08(%[dst_ptr]) \n\t" - "daddi %[width], %[width], -0x04 \n\t" - "daddiu %[dst_ptr], %[dst_ptr], 0x10 \n\t" - "bnez %[width], 1b \n\t" - : [v32]"+&f"(v32) - : [dst_ptr]"r"(dst_argb), [width]"r"(width) - : "memory" - ); + "daddi %[width], %[width], -0x04 \n\t" + "daddiu %[dst_ptr], %[dst_ptr], 0x10 \n\t" + "bnez %[width], 1b \n\t" + : [v32] "+&f"(v32) + : [dst_ptr] "r"(dst_argb), [width] "r"(width) + : "memory"); } // 10 bit YUV to ARGB diff --git a/source/row_neon.cc b/source/row_neon.cc index 99e7db97f..b379bbf61 100644 --- a/source/row_neon.cc +++ b/source/row_neon.cc @@ -2065,6 +2065,49 @@ void RAWToYRow_NEON(const uint8_t* src_raw, uint8_t* dst_y, int width) { : "cc", "memory", "d0", "d1", "d2", "d3", "d4", "d5", "d6", "d7", "q8"); } + +void RGB24ToYJRow_NEON(const uint8_t* src_rgb24, uint8_t* dst_yj, int width) { + asm volatile( + "vmov.u8 d4, #29 \n" // B * 0.1140 coefficient + "vmov.u8 d5, #150 \n" // G * 0.5870 coefficient + "vmov.u8 d6, #77 \n" // R * 0.2990 coefficient + "1: \n" + "vld3.8 {d0, d1, d2}, [%0]! \n" // load 8 pixels of RGB24. + "subs %2, %2, #8 \n" // 8 processed per loop. + "vmull.u8 q4, d0, d4 \n" // B + "vmlal.u8 q4, d1, d5 \n" // G + "vmlal.u8 q4, d2, d6 \n" // R + "vqrshrn.u16 d0, q4, #8 \n" // 16 bit to 8 bit Y + "vst1.8 {d0}, [%1]! \n" // store 8 pixels Y. + "bgt 1b \n" + : "+r"(src_rgb24), // %0 + "+r"(dst_yj), // %1 + "+r"(width) // %2 + : + : "cc", "memory", "d0", "d1", "d2", "d3", "d4", "d5", "d6", "q4"); +} + +void RAWToYJRow_NEON(const uint8_t* src_raw, uint8_t* dst_yj, int width) { + asm volatile( + "vmov.u8 d6, #29 \n" // B * 0.1140 coefficient + "vmov.u8 d5, #150 \n" // G * 0.5870 coefficient + "vmov.u8 d4, #77 \n" // R * 0.2990 coefficient + "1: \n" + "vld3.8 {d0, d1, d2}, [%0]! \n" // load 8 pixels of RAW. + "subs %2, %2, #8 \n" // 8 processed per loop. + "vmull.u8 q4, d0, d4 \n" // B + "vmlal.u8 q4, d1, d5 \n" // G + "vmlal.u8 q4, d2, d6 \n" // R + "vqrshrn.u16 d0, q4, #8 \n" // 16 bit to 8 bit Y + "vst1.8 {d0}, [%1]! \n" // store 8 pixels Y. + "bgt 1b \n" + : "+r"(src_raw), // %0 + "+r"(dst_yj), // %1 + "+r"(width) // %2 + : + : "cc", "memory", "d0", "d1", "d2", "d3", "d4", "d5", "d6", "q4"); +} + // Bilinear filter 16x2 -> 16x1 void InterpolateRow_NEON(uint8_t* dst_ptr, const uint8_t* src_ptr, diff --git a/source/row_neon64.cc b/source/row_neon64.cc index 8223b4737..197efb2a3 100644 --- a/source/row_neon64.cc +++ b/source/row_neon64.cc @@ -2103,6 +2103,48 @@ void RAWToYRow_NEON(const uint8_t* src_raw, uint8_t* dst_y, int width) { : "cc", "memory", "v0", "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v16"); } +void RGB24ToYJRow_NEON(const uint8_t* src_rgb24, uint8_t* dst_yj, int width) { + asm volatile( + "movi v4.8b, #29 \n" // B * 0.1140 coefficient + "movi v5.8b, #150 \n" // G * 0.5870 coefficient + "movi v6.8b, #77 \n" // R * 0.2990 coefficient + "1: \n" + "ld3 {v0.8b,v1.8b,v2.8b}, [%0], #24 \n" // load 8 pixels. + "subs %w2, %w2, #8 \n" // 8 processed per loop. + "umull v0.8h, v0.8b, v4.8b \n" // B + "umlal v0.8h, v1.8b, v5.8b \n" // G + "umlal v0.8h, v2.8b, v6.8b \n" // R + "uqrshrn v0.8b, v0.8h, #8 \n" // 16 bit to 8 bit Y + "st1 {v0.8b}, [%1], #8 \n" // store 8 pixels Y. + "b.gt 1b \n" + : "+r"(src_rgb24), // %0 + "+r"(dst_yj), // %1 + "+r"(width) // %2 + : + : "cc", "memory", "v0", "v1", "v2", "v3", "v4", "v5", "v6"); +} + +void RAWToYJRow_NEON(const uint8_t* src_raw, uint8_t* dst_yj, int width) { + asm volatile( + "movi v6.8b, #29 \n" // B * 0.1140 coefficient + "movi v5.8b, #150 \n" // G * 0.5870 coefficient + "movi v4.8b, #77 \n" // R * 0.2990 coefficient + "1: \n" + "ld3 {v0.8b,v1.8b,v2.8b}, [%0], #24 \n" // load 8 pixels. + "subs %w2, %w2, #8 \n" // 8 processed per loop. + "umull v0.8h, v0.8b, v4.8b \n" // B + "umlal v0.8h, v1.8b, v5.8b \n" // G + "umlal v0.8h, v2.8b, v6.8b \n" // R + "uqrshrn v0.8b, v0.8h, #8 \n" // 16 bit to 8 bit Y + "st1 {v0.8b}, [%1], #8 \n" // store 8 pixels Y. + "b.gt 1b \n" + : "+r"(src_raw), // %0 + "+r"(dst_yj), // %1 + "+r"(width) // %2 + : + : "cc", "memory", "v0", "v1", "v2", "v3", "v4", "v5", "v6"); +} + // Bilinear filter 16x2 -> 16x1 void InterpolateRow_NEON(uint8_t* dst_ptr, const uint8_t* src_ptr, diff --git a/unit_test/convert_test.cc b/unit_test/convert_test.cc index 698cd9562..af1c94ad6 100644 --- a/unit_test/convert_test.cc +++ b/unit_test/convert_test.cc @@ -1245,6 +1245,7 @@ TESTATOB(RAW, 3, 3, 1, RGBA, 4, 4, 1, 0) TESTATOB(RAW, 3, 3, 1, RGB24, 3, 3, 1, 0) TESTATOB(RGB24, 3, 3, 1, ARGB, 4, 4, 1, 0) TESTATOB(RGB24, 3, 3, 1, J400, 1, 1, 1, 0) +TESTATOB(RAW, 3, 3, 1, J400, 1, 1, 1, 0) #ifdef INTEL_TEST TESTATOB(RGB565, 2, 2, 1, ARGB, 4, 4, 1, 0) #endif