Browse code
swscale/arm: add ff_nv{12,21}_to_{argb,rgba,abgr,bgra}_neon_16
Matthieu Bouron authored on 2015/12/07 21:56:25Showing 2 changed files
| ... | ... |
@@ -63,8 +63,8 @@ static int rgbx_to_nv12_neon_16_wrapper(SwsContext *context, const uint8_t *src[ |
| 63 | 63 |
} |
| 64 | 64 |
#endif |
| 65 | 65 |
|
| 66 |
-#define DECLARE_FF_NVX_TO_RGBX_FUNCS(ifmt, ofmt) \ |
|
| 67 |
-int ff_##ifmt##_to_##ofmt##_neon(int w, int h, \ |
|
| 66 |
+#define DECLARE_FF_NVX_TO_RGBX_FUNCS(ifmt, ofmt, precision) \ |
|
| 67 |
+int ff_##ifmt##_to_##ofmt##_neon_##precision(int w, int h, \ |
|
| 68 | 68 |
uint8_t *dst, int linesize, \ |
| 69 | 69 |
const uint8_t *srcY, int linesizeY, \ |
| 70 | 70 |
const uint8_t *srcC, int linesizeC, \ |
| ... | ... |
@@ -72,59 +72,64 @@ int ff_##ifmt##_to_##ofmt##_neon(int w, int h, |
| 72 | 72 |
int y_offset, \ |
| 73 | 73 |
int y_coeff); \ |
| 74 | 74 |
\ |
| 75 |
-static int ifmt##_to_##ofmt##_neon_wrapper(SwsContext *c, const uint8_t *src[], \ |
|
| 75 |
+static int ifmt##_to_##ofmt##_neon_wrapper_##precision(SwsContext *c, const uint8_t *src[], \ |
|
| 76 | 76 |
int srcStride[], int srcSliceY, int srcSliceH, \ |
| 77 | 77 |
uint8_t *dst[], int dstStride[]) { \
|
| 78 | 78 |
const int16_t yuv2rgb_table[] = { \
|
| 79 |
- c->yuv2rgb_v2r_coeff, \ |
|
| 80 |
- c->yuv2rgb_u2g_coeff, \ |
|
| 81 |
- c->yuv2rgb_v2g_coeff, \ |
|
| 82 |
- c->yuv2rgb_u2b_coeff, \ |
|
| 79 |
+ c->yuv2rgb_v2r_coeff / ((precision) == 16 ? 1 << 7 : 1), \ |
|
| 80 |
+ c->yuv2rgb_u2g_coeff / ((precision) == 16 ? 1 << 7 : 1), \ |
|
| 81 |
+ c->yuv2rgb_v2g_coeff / ((precision) == 16 ? 1 << 7 : 1), \ |
|
| 82 |
+ c->yuv2rgb_u2b_coeff / ((precision) == 16 ? 1 << 7 : 1), \ |
|
| 83 | 83 |
}; \ |
| 84 | 84 |
\ |
| 85 |
- ff_##ifmt##_to_##ofmt##_neon(c->srcW, srcSliceH, \ |
|
| 85 |
+ ff_##ifmt##_to_##ofmt##_neon_##precision(c->srcW, srcSliceH, \ |
|
| 86 | 86 |
dst[0] + srcSliceY * dstStride[0], dstStride[0], \ |
| 87 | 87 |
src[0] + srcSliceY * srcStride[0], srcStride[0], \ |
| 88 | 88 |
src[1] + (srcSliceY / 2) * srcStride[1], srcStride[1], \ |
| 89 | 89 |
yuv2rgb_table, \ |
| 90 | 90 |
c->yuv2rgb_y_offset >> 9, \ |
| 91 |
- c->yuv2rgb_y_coeff); \ |
|
| 91 |
+ c->yuv2rgb_y_coeff / ((precision) == 16 ? 1 << 7 : 1)); \ |
|
| 92 | 92 |
\ |
| 93 | 93 |
return 0; \ |
| 94 |
-} |
|
| 94 |
+} \ |
|
| 95 |
+ |
|
| 96 |
+#define DECLARE_FF_NVX_TO_ALL_RGBX_FUNCS(nvx, precision) \ |
|
| 97 |
+DECLARE_FF_NVX_TO_RGBX_FUNCS(nvx, argb, precision) \ |
|
| 98 |
+DECLARE_FF_NVX_TO_RGBX_FUNCS(nvx, rgba, precision) \ |
|
| 99 |
+DECLARE_FF_NVX_TO_RGBX_FUNCS(nvx, abgr, precision) \ |
|
| 100 |
+DECLARE_FF_NVX_TO_RGBX_FUNCS(nvx, bgra, precision) \ |
|
| 95 | 101 |
|
| 96 |
-#define DECLARE_FF_NVX_TO_ALL_RGBX_FUNCS(nvx) \ |
|
| 97 |
-DECLARE_FF_NVX_TO_RGBX_FUNCS(nvx, argb) \ |
|
| 98 |
-DECLARE_FF_NVX_TO_RGBX_FUNCS(nvx, rgba) \ |
|
| 99 |
-DECLARE_FF_NVX_TO_RGBX_FUNCS(nvx, abgr) \ |
|
| 100 |
-DECLARE_FF_NVX_TO_RGBX_FUNCS(nvx, bgra) \ |
|
| 102 |
+#define DECLARE_FF_NVX_TO_ALL_RGBX_ALL_PRECISION_FUNCS(nvx) \ |
|
| 103 |
+DECLARE_FF_NVX_TO_ALL_RGBX_FUNCS(nvx, 16) \ |
|
| 104 |
+DECLARE_FF_NVX_TO_ALL_RGBX_FUNCS(nvx, 32) \ |
|
| 101 | 105 |
|
| 102 |
-DECLARE_FF_NVX_TO_ALL_RGBX_FUNCS(nv12) |
|
| 103 |
-DECLARE_FF_NVX_TO_ALL_RGBX_FUNCS(nv21) |
|
| 106 |
+DECLARE_FF_NVX_TO_ALL_RGBX_ALL_PRECISION_FUNCS(nv12) |
|
| 107 |
+DECLARE_FF_NVX_TO_ALL_RGBX_ALL_PRECISION_FUNCS(nv21) |
|
| 104 | 108 |
|
| 105 | 109 |
/* We need a 16 pixel width alignment. This constraint can easily be removed |
| 106 | 110 |
* for input reading but for the output which is 4-bytes per pixel (RGBA) the |
| 107 | 111 |
* assembly might be writing as much as 4*15=60 extra bytes at the end of the |
| 108 | 112 |
* line, which won't fit the 32-bytes buffer alignment. */ |
| 109 |
-#define SET_FF_NVX_TO_RGBX_FUNC(ifmt, IFMT, ofmt, OFMT) do { \
|
|
| 113 |
+#define SET_FF_NVX_TO_RGBX_FUNC(ifmt, IFMT, ofmt, OFMT, accurate_rnd) do { \
|
|
| 110 | 114 |
if (c->srcFormat == AV_PIX_FMT_##IFMT \ |
| 111 | 115 |
&& c->dstFormat == AV_PIX_FMT_##OFMT \ |
| 112 | 116 |
&& !(c->srcH & 1) \ |
| 113 | 117 |
&& !(c->srcW & 15)) { \
|
| 114 |
- c->swscale = ifmt##_to_##ofmt##_neon_wrapper; \ |
|
| 118 |
+ c->swscale = (accurate_rnd) ? ifmt##_to_##ofmt##_neon_wrapper_32 : \ |
|
| 119 |
+ ifmt##_to_##ofmt##_neon_wrapper_16 ; \ |
|
| 115 | 120 |
} \ |
| 116 | 121 |
} while (0) |
| 117 | 122 |
|
| 118 |
-#define SET_FF_NVX_TO_ALL_RGBX_FUNC(nvx, NVX) do { \
|
|
| 119 |
- SET_FF_NVX_TO_RGBX_FUNC(nvx, NVX, argb, ARGB); \ |
|
| 120 |
- SET_FF_NVX_TO_RGBX_FUNC(nvx, NVX, rgba, RGBA); \ |
|
| 121 |
- SET_FF_NVX_TO_RGBX_FUNC(nvx, NVX, abgr, ABGR); \ |
|
| 122 |
- SET_FF_NVX_TO_RGBX_FUNC(nvx, NVX, bgra, BGRA); \ |
|
| 123 |
+#define SET_FF_NVX_TO_ALL_RGBX_FUNC(nvx, NVX, accurate_rnd) do { \
|
|
| 124 |
+ SET_FF_NVX_TO_RGBX_FUNC(nvx, NVX, argb, ARGB, accurate_rnd); \ |
|
| 125 |
+ SET_FF_NVX_TO_RGBX_FUNC(nvx, NVX, rgba, RGBA, accurate_rnd); \ |
|
| 126 |
+ SET_FF_NVX_TO_RGBX_FUNC(nvx, NVX, abgr, ABGR, accurate_rnd); \ |
|
| 127 |
+ SET_FF_NVX_TO_RGBX_FUNC(nvx, NVX, bgra, BGRA, accurate_rnd); \ |
|
| 123 | 128 |
} while (0) |
| 124 | 129 |
|
| 125 | 130 |
static void get_unscaled_swscale_neon(SwsContext *c) {
|
| 126 |
-#if 0 |
|
| 127 | 131 |
int accurate_rnd = c->flags & SWS_ACCURATE_RND; |
| 132 |
+#if 0 |
|
| 128 | 133 |
if (c->srcFormat == AV_PIX_FMT_RGBA |
| 129 | 134 |
&& c->dstFormat == AV_PIX_FMT_NV12 |
| 130 | 135 |
&& (c->srcW >= 16)) {
|
| ... | ... |
@@ -133,8 +138,8 @@ static void get_unscaled_swscale_neon(SwsContext *c) {
|
| 133 | 133 |
} |
| 134 | 134 |
#endif |
| 135 | 135 |
|
| 136 |
- SET_FF_NVX_TO_ALL_RGBX_FUNC(nv12, NV12); |
|
| 137 |
- SET_FF_NVX_TO_ALL_RGBX_FUNC(nv21, NV21); |
|
| 136 |
+ SET_FF_NVX_TO_ALL_RGBX_FUNC(nv12, NV12, accurate_rnd); |
|
| 137 |
+ SET_FF_NVX_TO_ALL_RGBX_FUNC(nv21, NV21, accurate_rnd); |
|
| 138 | 138 |
} |
| 139 | 139 |
|
| 140 | 140 |
void ff_get_unscaled_swscale_arm(SwsContext *c) |
| ... | ... |
@@ -21,7 +21,35 @@ |
| 21 | 21 |
|
| 22 | 22 |
#include "libavutil/arm/asm.S" |
| 23 | 23 |
|
| 24 |
-.macro compute_premult half_u half_v |
|
| 24 |
+ |
|
| 25 |
+.macro compute_premult_16 half_u1, half_u2, half_v1, half_v2 |
|
| 26 |
+ vmov d2, \half_u1 @ copy left q14 to left q1 |
|
| 27 |
+ vmov d3, \half_u1 @ copy left q14 to right q1 |
|
| 28 |
+ vmov d4, \half_u2 @ copy right q14 to left q2 |
|
| 29 |
+ vmov d5, \half_u2 @ copy right q14 to right q2 |
|
| 30 |
+ |
|
| 31 |
+ vmov d6, \half_v1 @ copy left q15 to left q3 |
|
| 32 |
+ vmov d7, \half_v1 @ copy left q15 to right q3 |
|
| 33 |
+ vmov d8, \half_v2 @ copy right q15 to left q4 |
|
| 34 |
+ vmov d9, \half_v2 @ copy right q15 to right q4 |
|
| 35 |
+ |
|
| 36 |
+ vzip.16 d2, d3 @ U1U1U2U2U3U3U4U4 |
|
| 37 |
+ vzip.16 d4, d5 @ U5U5U6U6U7U7U8U8 |
|
| 38 |
+ |
|
| 39 |
+ vzip.16 d6, d7 @ V1V1V2V2V3V3V4V4 |
|
| 40 |
+ vzip.16 d8, d9 @ V5V5V6V6V7V7V8V8 |
|
| 41 |
+ |
|
| 42 |
+ vmul.s16 q8, q3, d1[0] @ V * v2r (left, red) |
|
| 43 |
+ vmul.s16 q9, q4, d1[0] @ V * v2r (right, red) |
|
| 44 |
+ vmul.s16 q10, q1, d1[1] @ U * u2g |
|
| 45 |
+ vmul.s16 q11, q2, d1[1] @ U * u2g |
|
| 46 |
+ vmla.s16 q10, q3, d1[2] @ U * u2g + V * v2g (left, green) |
|
| 47 |
+ vmla.s16 q11, q4, d1[2] @ U * u2g + V * v2g (right, green) |
|
| 48 |
+ vmul.s16 q12, q1, d1[3] @ U * u2b (left, blue) |
|
| 49 |
+ vmul.s16 q13, q2, d1[3] @ U * u2b (right, blue) |
|
| 50 |
+.endm |
|
| 51 |
+ |
|
| 52 |
+.macro compute_premult_32 half_u half_v |
|
| 25 | 53 |
vmov d2, \half_u @ copy left q14 to left q1 |
| 26 | 54 |
vmov d3, \half_u @ copy left q14 to right q1 |
| 27 | 55 |
vmov d4, \half_v @ copy left q15 to left q2 |
| ... | ... |
@@ -40,7 +68,14 @@ |
| 40 | 40 |
vmull.s16 q13, d3, d1[3] @ U * u2b (right, blue) |
| 41 | 41 |
.endm |
| 42 | 42 |
|
| 43 |
-.macro compute_color dst_comp pre1 pre2 |
|
| 43 |
+.macro compute_color_16 dst_comp1 dst_comp2 pre1 pre2 |
|
| 44 |
+ vadd.s16 q1, q14, \pre1 |
|
| 45 |
+ vadd.s16 q2, q15, \pre2 |
|
| 46 |
+ vqrshrun.s16 \dst_comp1, q1, #6 |
|
| 47 |
+ vqrshrun.s16 \dst_comp2, q2, #6 |
|
| 48 |
+.endm |
|
| 49 |
+ |
|
| 50 |
+.macro compute_color_32 dst_comp pre1 pre2 |
|
| 44 | 51 |
vadd.s32 q3, q1, \pre1 |
| 45 | 52 |
vadd.s32 q4, q2, \pre2 |
| 46 | 53 |
vqrshrun.s32 d10, q3, #13 |
| ... | ... |
@@ -48,14 +83,56 @@ |
| 48 | 48 |
vqmovn.u16 \dst_comp, q5 @ saturate 16bit -> 8bit |
| 49 | 49 |
.endm |
| 50 | 50 |
|
| 51 |
-.macro compute_rgba r g b a |
|
| 52 |
- compute_color \r, q8, q9 |
|
| 53 |
- compute_color \g, q10, q11 |
|
| 54 |
- compute_color \b, q12, q13 |
|
| 51 |
+.macro compute_rgba_16 r1 r2 g1 g2 b1 b2 a1 a2 |
|
| 52 |
+ compute_color_16 \r1, \r2, q8, q9 |
|
| 53 |
+ compute_color_16 \g1, \g2, q10, q11 |
|
| 54 |
+ compute_color_16 \b1, \b2, q12, q13 |
|
| 55 |
+ vmov.u8 \a1, #255 |
|
| 56 |
+ vmov.u8 \a2, #255 |
|
| 57 |
+.endm |
|
| 58 |
+ |
|
| 59 |
+.macro compute_rgba_32 r g b a |
|
| 60 |
+ compute_color_32 \r, q8, q9 |
|
| 61 |
+ compute_color_32 \g, q10, q11 |
|
| 62 |
+ compute_color_32 \b, q12, q13 |
|
| 55 | 63 |
vmov.u8 \a, #255 |
| 56 | 64 |
.endm |
| 57 | 65 |
|
| 58 |
-.macro compute_half_line dst half_y ofmt |
|
| 66 |
+.macro compute_16px_16 dst y0 y1 ofmt |
|
| 67 |
+ vmovl.u8 q14, \y0 @ 8px of y |
|
| 68 |
+ vmovl.u8 q15, \y1 @ 8px of y |
|
| 69 |
+ |
|
| 70 |
+ vdup.16 q5, r9 @ q5 = y_offset |
|
| 71 |
+ vdup.16 q7, r10 @ q7 = y_coeff |
|
| 72 |
+ |
|
| 73 |
+ vsub.s16 q14, q5 |
|
| 74 |
+ vsub.s16 q15, q5 |
|
| 75 |
+ |
|
| 76 |
+ vmul.s16 q14, q7 @ q14 = (srcY - y_offset) * y_coeff (left) |
|
| 77 |
+ vmul.s16 q15, q7 @ q15 = (srcY - y_offset) * y_coeff (right) |
|
| 78 |
+ |
|
| 79 |
+ |
|
| 80 |
+.ifc \ofmt,argb |
|
| 81 |
+ compute_rgba_16 d7, d11, d8, d12, d9, d13, d6, d10 |
|
| 82 |
+.endif |
|
| 83 |
+ |
|
| 84 |
+.ifc \ofmt,rgba |
|
| 85 |
+ compute_rgba_16 d6, d10, d7, d11, d8, d12, d9, d13 |
|
| 86 |
+.endif |
|
| 87 |
+ |
|
| 88 |
+.ifc \ofmt,abgr |
|
| 89 |
+ compute_rgba_16 d9, d13, d8, d12, d7, d11, d6, d10 |
|
| 90 |
+.endif |
|
| 91 |
+ |
|
| 92 |
+.ifc \ofmt,bgra |
|
| 93 |
+ compute_rgba_16 d8, d12, d7, d11, d6, d10, d9, d13 |
|
| 94 |
+.endif |
|
| 95 |
+ vst4.8 {q3, q4}, [\dst,:128]!
|
|
| 96 |
+ vst4.8 {q5, q6}, [\dst,:128]!
|
|
| 97 |
+ |
|
| 98 |
+.endm |
|
| 99 |
+ |
|
| 100 |
+.macro compute_8px_32 dst half_y ofmt |
|
| 59 | 101 |
vmovl.u8 q7, \half_y @ 8px of Y |
| 60 | 102 |
vdup.16 q5, r9 |
| 61 | 103 |
vsub.s16 q7, q5 |
| ... | ... |
@@ -63,26 +140,51 @@ |
| 63 | 63 |
vmull.s16 q2, d15, d0 @ q2 = (srcY - y_offset) * y_coeff (right) |
| 64 | 64 |
|
| 65 | 65 |
.ifc \ofmt,argb |
| 66 |
- compute_rgba d13, d14, d15, d12 |
|
| 66 |
+ compute_rgba_32 d13, d14, d15, d12 |
|
| 67 | 67 |
.endif |
| 68 | 68 |
|
| 69 | 69 |
.ifc \ofmt,rgba |
| 70 |
- compute_rgba d12, d13, d14, d15 |
|
| 70 |
+ compute_rgba_32 d12, d13, d14, d15 |
|
| 71 | 71 |
.endif |
| 72 | 72 |
|
| 73 | 73 |
.ifc \ofmt,abgr |
| 74 |
- compute_rgba d15, d14, d13, d12 |
|
| 74 |
+ compute_rgba_32 d15, d14, d13, d12 |
|
| 75 | 75 |
.endif |
| 76 | 76 |
|
| 77 | 77 |
.ifc \ofmt,bgra |
| 78 |
- compute_rgba d14, d13, d12, d15 |
|
| 78 |
+ compute_rgba_32 d14, d13, d12, d15 |
|
| 79 | 79 |
.endif |
| 80 | 80 |
|
| 81 | 81 |
vst4.8 {q6, q7}, [\dst,:128]!
|
| 82 | 82 |
.endm |
| 83 | 83 |
|
| 84 |
-.macro declare_func ifmt ofmt |
|
| 85 |
-function ff_\ifmt\()_to_\ofmt\()_neon, export=1 |
|
| 84 |
+.macro process_16px_16 ofmt |
|
| 85 |
+ compute_premult_16 d28, d29, d30, d31 |
|
| 86 |
+ |
|
| 87 |
+ vld1.8 {q7}, [r4]! @ first line of luma
|
|
| 88 |
+ compute_16px_16 r2, d14, d15, \ofmt |
|
| 89 |
+ |
|
| 90 |
+ vld1.8 {q7}, [r12]! @ second line of luma
|
|
| 91 |
+ compute_16px_16 r11, d14, d15, \ofmt |
|
| 92 |
+.endm |
|
| 93 |
+ |
|
| 94 |
+.macro process_16px_32 ofmt |
|
| 95 |
+ compute_premult_32 d28, d30 |
|
| 96 |
+ |
|
| 97 |
+ vld1.8 {q7}, [r4]! @ first line of luma
|
|
| 98 |
+ vmov d28, d15 @ save right of the first line of luma for later use |
|
| 99 |
+ compute_8px_32 r2, d14, \ofmt |
|
| 100 |
+ |
|
| 101 |
+ vld1.8 {q7}, [r12]! @ second line of luma
|
|
| 102 |
+ vmov d30, d15 @ save right of the second line of luma for later use |
|
| 103 |
+ compute_8px_32 r11, d14, \ofmt |
|
| 104 |
+ |
|
| 105 |
+ compute_premult_32 d29, d31 |
|
| 106 |
+ compute_8px_32 r2, d28, \ofmt |
|
| 107 |
+ compute_8px_32 r11, d30, \ofmt |
|
| 108 |
+.endm |
|
| 109 |
+ |
|
| 110 |
+.macro load_args |
|
| 86 | 111 |
push {r4-r12, lr}
|
| 87 | 112 |
vpush {q4-q7}
|
| 88 | 113 |
ldr r4, [sp, #104] @ r4 = srcY |
| ... | ... |
@@ -102,6 +204,11 @@ function ff_\ifmt\()_to_\ofmt\()_neon, export=1 |
| 102 | 102 |
sub r3, r3, r8 @ r3 = linesize * 2 - width * 4 (padding) |
| 103 | 103 |
sub r5, r5, r0 @ r5 = linesizeY * 2 - width (paddingY) |
| 104 | 104 |
sub r7, r7, r0 @ r7 = linesizeC - width (paddingC) |
| 105 |
+.endm |
|
| 106 |
+ |
|
| 107 |
+.macro declare_func ifmt ofmt precision |
|
| 108 |
+function ff_\ifmt\()_to_\ofmt\()_neon_\precision\(), export=1 |
|
| 109 |
+ load_args |
|
| 105 | 110 |
1: |
| 106 | 111 |
mov r8, r0 @ r8 = width |
| 107 | 112 |
2: |
| ... | ... |
@@ -119,19 +226,13 @@ function ff_\ifmt\()_to_\ofmt\()_neon, export=1 |
| 119 | 119 |
vsubl.u8 q15, d2, d10 @ q15 = V - 128 |
| 120 | 120 |
.endif |
| 121 | 121 |
|
| 122 |
- compute_premult d28, d30 |
|
| 123 |
- |
|
| 124 |
- vld1.8 {q7}, [r4]! @ first line of luma
|
|
| 125 |
- vmov d28, d15 @ save right of the first line of luma for later use |
|
| 126 |
- compute_half_line r2, d14, \ofmt |
|
| 127 |
- |
|
| 128 |
- vld1.8 {q7}, [r12]! @ second line of luma
|
|
| 129 |
- vmov d30, d15 @ save right of the second line of luma for later use |
|
| 130 |
- compute_half_line r11, d14, \ofmt |
|
| 122 |
+.ifc \precision,16 |
|
| 123 |
+ process_16px_16 \ofmt |
|
| 124 |
+.endif |
|
| 131 | 125 |
|
| 132 |
- compute_premult d29, d31 |
|
| 133 |
- compute_half_line r2, d28, \ofmt |
|
| 134 |
- compute_half_line r11, d30, \ofmt |
|
| 126 |
+.ifc \precision,32 |
|
| 127 |
+ process_16px_32 \ofmt |
|
| 128 |
+.endif |
|
| 135 | 129 |
|
| 136 | 130 |
subs r8, r8, #16 @ width -= 16 |
| 137 | 131 |
bgt 2b |
| ... | ... |
@@ -151,12 +252,14 @@ function ff_\ifmt\()_to_\ofmt\()_neon, export=1 |
| 151 | 151 |
endfunc |
| 152 | 152 |
.endm |
| 153 | 153 |
|
| 154 |
-.macro declare_rgb_funcs ifmt |
|
| 155 |
- declare_func \ifmt, argb |
|
| 156 |
- declare_func \ifmt, rgba |
|
| 157 |
- declare_func \ifmt, abgr |
|
| 158 |
- declare_func \ifmt, bgra |
|
| 154 |
+.macro declare_rgb_funcs ifmt precision |
|
| 155 |
+ declare_func \ifmt, argb, \precision |
|
| 156 |
+ declare_func \ifmt, rgba, \precision |
|
| 157 |
+ declare_func \ifmt, abgr, \precision |
|
| 158 |
+ declare_func \ifmt, bgra, \precision |
|
| 159 | 159 |
.endm |
| 160 | 160 |
|
| 161 |
-declare_rgb_funcs nv12 |
|
| 162 |
-declare_rgb_funcs nv21 |
|
| 161 |
+declare_rgb_funcs nv12, 16 |
|
| 162 |
+declare_rgb_funcs nv21, 16 |
|
| 163 |
+declare_rgb_funcs nv12, 32 |
|
| 164 |
+declare_rgb_funcs nv21, 32 |