virtualenv/FFmpeg
1
0
Fork 0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2025-08-10 06:10:52 +02:00
Code Issues Releases Activity

swscale: K&R formatting cosmetics for PowerPC code (part II/II)

Browse Source
This commit is contained in:
Diego Biurrun
2012-02-21 19:59:38 +01:00
parent c05e2be9a2
commit 36697be187
1 changed files with 622 additions and 592 deletions
Download Patch File Download Diff File Expand all files Collapse all files

502
libswscale/ppc/yuv2rgb_altivec.c
View File

@@ -21,68 +21,68 @@
*/ */
/* /*
Convert I420 YV12 to RGB in various formats, * Convert I420 YV12 to RGB in various formats,
it rejects images that are not in 420 formats, * it rejects images that are not in 420 formats,
it rejects images that don't have widths of multiples of 16, * it rejects images that don't have widths of multiples of 16,
it rejects images that don't have heights of multiples of 2. * it rejects images that don't have heights of multiples of 2.
Reject defers to C simulation code. * Reject defers to C simulation code.
*
Lots of optimizations to be done here. * Lots of optimizations to be done here.
*
1. Need to fix saturation code. I just couldn't get it to fly with packs * 1. Need to fix saturation code. I just couldn't get it to fly with packs
and adds, so we currently use max/min to clip. * and adds, so we currently use max/min to clip.
*
2. The inefficient use of chroma loading needs a bit of brushing up. * 2. The inefficient use of chroma loading needs a bit of brushing up.
*
3. Analysis of pipeline stalls needs to be done. Use shark to identify * 3. Analysis of pipeline stalls needs to be done. Use shark to identify
pipeline stalls. * pipeline stalls.
*
*
MODIFIED to calculate coeffs from currently selected color space. * MODIFIED to calculate coeffs from currently selected color space.
MODIFIED core to be a macro where you specify the output format. * MODIFIED core to be a macro where you specify the output format.
ADDED UYVY conversion which is never called due to some thing in swscale. * ADDED UYVY conversion which is never called due to some thing in swscale.
CORRECTED algorithim selection to be strict on input formats. * CORRECTED algorithim selection to be strict on input formats.
ADDED runtime detection of AltiVec. * ADDED runtime detection of AltiVec.
*
ADDED altivec_yuv2packedX vertical scl + RGB converter * ADDED altivec_yuv2packedX vertical scl + RGB converter
*
March 27,2004 * March 27,2004
PERFORMANCE ANALYSIS * PERFORMANCE ANALYSIS
*
The C version uses 25% of the processor or ~250Mips for D1 video rawvideo * The C version uses 25% of the processor or ~250Mips for D1 video rawvideo
used as test. * used as test.
The AltiVec version uses 10% of the processor or ~100Mips for D1 video * The AltiVec version uses 10% of the processor or ~100Mips for D1 video
same sequence. * same sequence.
*
720 * 480 * 30 ~10MPS * 720 * 480 * 30 ~10MPS
*
so we have roughly 10 clocks per pixel. This is too high, something has * so we have roughly 10 clocks per pixel. This is too high, something has
to be wrong. * to be wrong.
*
OPTIMIZED clip codes to utilize vec_max and vec_packs removing the * OPTIMIZED clip codes to utilize vec_max and vec_packs removing the
need for vec_min. * need for vec_min.
*
OPTIMIZED DST OUTPUT cache/DMA controls. We are pretty much guaranteed to have * OPTIMIZED DST OUTPUT cache/DMA controls. We are pretty much guaranteed to
the input video frame, it was just decompressed so it probably resides in L1 * have the input video frame, it was just decompressed so it probably resides
caches. However, we are creating the output video stream. This needs to use the * in L1 caches. However, we are creating the output video stream. This needs
DSTST instruction to optimize for the cache. We couple this with the fact that * to use the DSTST instruction to optimize for the cache. We couple this with
we are not going to be visiting the input buffer again so we mark it Least * the fact that we are not going to be visiting the input buffer again so we
Recently Used. This shaves 25% of the processor cycles off. * mark it Least Recently Used. This shaves 25% of the processor cycles off.
*
Now memcpy is the largest mips consumer in the system, probably due * Now memcpy is the largest mips consumer in the system, probably due
to the inefficient X11 stuff. * to the inefficient X11 stuff.
*
GL libraries seem to be very slow on this machine 1.33Ghz PB running * GL libraries seem to be very slow on this machine 1.33Ghz PB running
Jaguar, this is not the case for my 1Ghz PB. I thought it might be * Jaguar, this is not the case for my 1Ghz PB. I thought it might be
a versioning issue, however I have libGL.1.2.dylib for both * a versioning issue, however I have libGL.1.2.dylib for both
machines. (We need to figure this out now.) * machines. (We need to figure this out now.)
*
GL2 libraries work now with patch for RGB32. * GL2 libraries work now with patch for RGB32.
*
NOTE: quartz vo driver ARGB32_to_RGB24 consumes 30% of the processor. * NOTE: quartz vo driver ARGB32_to_RGB24 consumes 30% of the processor.
*
Integrated luma prescaling adjustment for saturation/contrast/brightness * Integrated luma prescaling adjustment for saturation/contrast/brightness
adjustment. * adjustment.
*/ */
#include <stdio.h> #include <stdio.h>
@@ -90,6 +90,7 @@ adjustment.
#include <string.h> #include <string.h>
#include <inttypes.h> #include <inttypes.h>
#include <assert.h> #include <assert.h>
#include "config.h" #include "config.h"
#include "libswscale/rgb2rgb.h" #include "libswscale/rgb2rgb.h"
#include "libswscale/swscale.h" #include "libswscale/swscale.h"
@@ -103,45 +104,43 @@ adjustment.
typedef unsigned char ubyte; typedef unsigned char ubyte;
typedef signed char sbyte; typedef signed char sbyte;
/* RGB interleaver, 16 planar pels 8-bit samples per channel in /* RGB interleaver, 16 planar pels 8-bit samples per channel in
homogeneous vector registers x0,x1,x2 are interleaved with the * homogeneous vector registers x0,x1,x2 are interleaved with the
following technique: * following technique:
*
o0 = vec_mergeh (x0,x1); * o0 = vec_mergeh(x0, x1);
o1 = vec_perm (o0, x2, perm_rgb_0); * o1 = vec_perm(o0, x2, perm_rgb_0);
o2 = vec_perm (o0, x2, perm_rgb_1); * o2 = vec_perm(o0, x2, perm_rgb_1);
o3 = vec_mergel (x0,x1); * o3 = vec_mergel(x0, x1);
o4 = vec_perm (o3,o2,perm_rgb_2); * o4 = vec_perm(o3, o2, perm_rgb_2);
o5 = vec_perm (o3,o2,perm_rgb_3); * o5 = vec_perm(o3, o2, perm_rgb_3);
*
perm_rgb_0: o0(RG).h v1(B) --> o1* * perm_rgb_0: o0(RG).h v1(B) --> o1*
0 1 2 3 4 * 0 1 2 3 4
rgbr|gbrg|brgb|rgbr * rgbr|gbrg|brgb|rgbr
0010 0100 1001 0010 * 0010 0100 1001 0010
0102 3145 2673 894A * 0102 3145 2673 894A
*
perm_rgb_1: o0(RG).h v1(B) --> o2 * perm_rgb_1: o0(RG).h v1(B) --> o2
0 1 2 3 4 * 0 1 2 3 4
gbrg|brgb|bbbb|bbbb * gbrg|brgb|bbbb|bbbb
0100 1001 1111 1111 * 0100 1001 1111 1111
B5CD 6EF7 89AB CDEF * B5CD 6EF7 89AB CDEF
*
perm_rgb_2: o3(RG).l o2(rgbB.l) --> o4* * perm_rgb_2: o3(RG).l o2(rgbB.l) --> o4*
0 1 2 3 4 * 0 1 2 3 4
gbrg|brgb|rgbr|gbrg * gbrg|brgb|rgbr|gbrg
1111 1111 0010 0100 * 1111 1111 0010 0100
89AB CDEF 0182 3945 * 89AB CDEF 0182 3945
*
perm_rgb_2: o3(RG).l o2(rgbB.l) ---> o5* * perm_rgb_2: o3(RG).l o2(rgbB.l) ---> o5*
0 1 2 3 4 * 0 1 2 3 4
brgb|rgbr|gbrg|brgb * brgb|rgbr|gbrg|brgb
1001 0010 0100 1001 * 1001 0010 0100 1001
a67b 89cA BdCD eEFf * a67b 89cA BdCD eEFf
*
*/ */
static static const vector unsigned char
const vector unsigned char
perm_rgb_0 = { 0x00, 0x01, 0x10, 0x02, 0x03, 0x11, 0x04, 0x05, perm_rgb_0 = { 0x00, 0x01, 0x10, 0x02, 0x03, 0x11, 0x04, 0x05,
0x12, 0x06, 0x07, 0x13, 0x08, 0x09, 0x14, 0x0a }, 0x12, 0x06, 0x07, 0x13, 0x08, 0x09, 0x14, 0x0a },
perm_rgb_1 = { 0x0b, 0x15, 0x0c, 0x0d, 0x16, 0x0e, 0x0f, 0x17, perm_rgb_1 = { 0x0b, 0x15, 0x0c, 0x0d, 0x16, 0x0e, 0x0f, 0x17,
@@ -181,70 +180,76 @@ do { \
} while (0) } while (0)
/* pack the pixels in rgb0 format /* pack the pixels in rgb0 format
msb R * msb R
lsb 0 * lsb 0
*/ */
#define vec_mstrgb32(T, x0, x1, x2, x3, ptr) \ #define vec_mstrgb32(T, x0, x1, x2, x3, ptr) \
do { \ do { \
T _0, _1, _2, _3; \ T _0, _1, _2, _3; \
_0 = vec_mergeh(x0, x1); \ _0 = vec_mergeh(x0, x1); \
_1 = vec_mergeh(x2, x3); \ _1 = vec_mergeh(x2, x3); \
_2 = (T)vec_mergeh ((vector unsigned short)_0,(vector unsigned short)_1); \ _2 = (T) vec_mergeh((vector unsigned short) _0, \
_3 = (T)vec_mergel ((vector unsigned short)_0,(vector unsigned short)_1); \ (vector unsigned short) _1); \
_3 = (T) vec_mergel((vector unsigned short) _0, \
(vector unsigned short) _1); \
vec_st(_2, 0 * 16, (T *) ptr); \ vec_st(_2, 0 * 16, (T *) ptr); \
vec_st(_3, 1 * 16, (T *) ptr); \ vec_st(_3, 1 * 16, (T *) ptr); \
_0 = vec_mergel(x0, x1); \ _0 = vec_mergel(x0, x1); \
_1 = vec_mergel(x2, x3); \ _1 = vec_mergel(x2, x3); \
_2 = (T)vec_mergeh ((vector unsigned short)_0,(vector unsigned short)_1); \ _2 = (T) vec_mergeh((vector unsigned short) _0, \
_3 = (T)vec_mergel ((vector unsigned short)_0,(vector unsigned short)_1); \ (vector unsigned short) _1); \
_3 = (T) vec_mergel((vector unsigned short) _0, \
(vector unsigned short) _1); \
vec_st(_2, 2 * 16, (T *) ptr); \ vec_st(_2, 2 * 16, (T *) ptr); \
vec_st(_3, 3 * 16, (T *) ptr); \ vec_st(_3, 3 * 16, (T *) ptr); \
ptr += 4; \ ptr += 4; \
} while (0) } while (0)
/* /*
* 1 0 1.4021 | | Y |
| 1 0 1.4021 | | Y | * 1 -0.3441 -0.7142 |x| Cb|
| 1 -0.3441 -0.7142 |x| Cb| * 1 1.7718 0 | | Cr|
| 1 1.7718 0 | | Cr| *
*
* Y: [-128 127]
Y: [-128 127] * Cb/Cr : [-128 127]
Cb/Cr : [-128 127] *
* typical YUV conversion works on Y: 0-255 this version has been
typical yuv conversion work on Y: 0-255 this version has been optimized for jpeg decode. * optimized for JPEG decoding.
*/ */
#define vec_unh(x) \ #define vec_unh(x) \
(vector signed short) \ (vector signed short) \
vec_perm(x, (__typeof__(x)) { 0 }, \ vec_perm(x, (__typeof__(x)) { 0 }, \
((vector unsigned char){0x10,0x00,0x10,0x01,0x10,0x02,0x10,0x03,\ ((vector unsigned char) { \
0x10, 0x00, 0x10, 0x01, 0x10, 0x02, 0x10, 0x03, \
0x10, 0x04, 0x10, 0x05, 0x10, 0x06, 0x10, 0x07 })) 0x10, 0x04, 0x10, 0x05, 0x10, 0x06, 0x10, 0x07 }))
#define vec_unl(x) \ #define vec_unl(x) \
(vector signed short) \ (vector signed short) \
vec_perm(x, (__typeof__(x)) { 0 }, \ vec_perm(x, (__typeof__(x)) { 0 }, \
((vector unsigned char){0x10,0x08,0x10,0x09,0x10,0x0A,0x10,0x0B,\ ((vector unsigned char) { \
0x10, 0x08, 0x10, 0x09, 0x10, 0x0A, 0x10, 0x0B, \
0x10, 0x0C, 0x10, 0x0D, 0x10, 0x0E, 0x10, 0x0F })) 0x10, 0x0C, 0x10, 0x0D, 0x10, 0x0E, 0x10, 0x0F }))
#define vec_clip_s16(x) \ #define vec_clip_s16(x) \
vec_max (vec_min (x, ((vector signed short){235,235,235,235,235,235,235,235})), \ vec_max(vec_min(x, ((vector signed short) { \
235, 235, 235, 235, 235, 235, 235, 235 })), \
((vector signed short) { 16, 16, 16, 16, 16, 16, 16, 16 })) ((vector signed short) { 16, 16, 16, 16, 16, 16, 16, 16 }))
#define vec_packclp(x, y) \ #define vec_packclp(x, y) \
(vector unsigned char)vec_packs \ (vector unsigned char) \
((vector unsigned short)vec_max (x,((vector signed short) {0})), \ vec_packs((vector unsigned short) \
(vector unsigned short)vec_max (y,((vector signed short) {0}))) vec_max(x, ((vector signed short) { 0 })), \
(vector unsigned short) \
vec_max(y, ((vector signed short) { 0 })))
//#define out_pixels(a, b, c, ptr) vec_mstrgb32(__typeof__(a), ((__typeof__(a)) { 255 }), a, a, a, ptr) //#define out_pixels(a, b, c, ptr) vec_mstrgb32(__typeof__(a), ((__typeof__(a)) { 255 }), a, a, a, ptr)
static inline void cvtyuvtoRGB(SwsContext *c, vector signed short Y,
static inline void cvtyuvtoRGB (SwsContext *c, vector signed short U, vector signed short V,
vector signed short Y, vector signed short U, vector signed short V, vector signed short *R, vector signed short *G,
vector signed short *R, vector signed short *G, vector signed short *B) vector signed short *B)
{ {
vector signed short vx, ux, uvx; vector signed short vx, ux, uvx;
@@ -267,18 +272,15 @@ static inline void cvtyuvtoRGB (SwsContext *c,
*G = vec_mradds(V, c->CGV, uvx); *G = vec_mradds(V, c->CGV, uvx);
} }
/* /*
------------------------------------------------------------------------------ * ------------------------------------------------------------------------------
CS converters * CS converters
------------------------------------------------------------------------------ * ------------------------------------------------------------------------------
*/ */
#define DEFCSP420_CVT(name, out_pixels) \ #define DEFCSP420_CVT(name, out_pixels) \
static int altivec_##name (SwsContext *c, \ static int altivec_ ## name(SwsContext *c, const unsigned char **in, \
const unsigned char **in, int *instrides, \ int *instrides, int srcSliceY, int srcSliceH, \
int srcSliceY, int srcSliceH, \
unsigned char **oplanes, int *outstrides) \ unsigned char **oplanes, int *outstrides) \
{ \ { \
int w = c->srcW; \ int w = c->srcW; \
@@ -301,14 +303,12 @@ static int altivec_##name (SwsContext *c, \
vector unsigned char *y1ivP, *y2ivP, *uivP, *vivP; \ vector unsigned char *y1ivP, *y2ivP, *uivP, *vivP; \
vector unsigned char align_perm; \ vector unsigned char align_perm; \
\ \
vector signed short \ vector signed short lCY = c->CY; \
lCY = c->CY, \ vector signed short lOY = c->OY; \
lOY = c->OY, \ vector signed short lCRV = c->CRV; \
lCRV = c->CRV, \ vector signed short lCBU = c->CBU; \
lCBU = c->CBU, \ vector signed short lCGU = c->CGU; \
lCGU = c->CGU, \ vector signed short lCGV = c->CGV; \
lCGV = c->CGV; \
\
vector unsigned short lCSHIFT = c->CSHIFT; \ vector unsigned short lCSHIFT = c->CSHIFT; \
\ \
const ubyte *y1i = in[0]; \ const ubyte *y1i = in[0]; \
@@ -316,25 +316,25 @@ static int altivec_##name (SwsContext *c, \
const ubyte *ui = in[1]; \ const ubyte *ui = in[1]; \
const ubyte *vi = in[2]; \ const ubyte *vi = in[2]; \
\ \
vector unsigned char *oute \ vector unsigned char *oute = \
= (vector unsigned char *) \ (vector unsigned char *) \
(oplanes[0] + srcSliceY * outstrides[0]); \ (oplanes[0] + srcSliceY * outstrides[0]); \
vector unsigned char *outo \ vector unsigned char *outo = \
= (vector unsigned char *) \ (vector unsigned char *) \
(oplanes[0] + srcSliceY * outstrides[0] + outstrides[0]); \ (oplanes[0] + srcSliceY * outstrides[0] + outstrides[0]); \
\ \
\ /* loop moves y{1, 2}i by w */ \
instrides_scl[0] = instrides[0]*2-w; /* the loop moves y{1,2}i by w */ \ instrides_scl[0] = instrides[0] * 2 - w; \
instrides_scl[1] = instrides[1]-w/2; /* the loop moves ui by w/2 */ \ /* loop moves ui by w / 2 */ \
instrides_scl[2] = instrides[2]-w/2; /* the loop moves vi by w/2 */ \ instrides_scl[1] = instrides[1] - w / 2; \
\ /* loop moves vi by w / 2 */ \
instrides_scl[2] = instrides[2] - w / 2; \
\ \
for (i = 0; i < h / 2; i++) { \ for (i = 0; i < h / 2; i++) { \
vec_dstst(outo, (0x02000002 | (((w * 3 + 32) / 32) << 16)), 0); \ vec_dstst(outo, (0x02000002 | (((w * 3 + 32) / 32) << 16)), 0); \
vec_dstst(oute, (0x02000002 | (((w * 3 + 32) / 32) << 16)), 1); \ vec_dstst(oute, (0x02000002 | (((w * 3 + 32) / 32) << 16)), 1); \
\ \
for (j = 0; j < w / 16; j++) { \ for (j = 0; j < w / 16; j++) { \
\
y1ivP = (vector unsigned char *) y1i; \ y1ivP = (vector unsigned char *) y1i; \
y2ivP = (vector unsigned char *) y2i; \ y2ivP = (vector unsigned char *) y2i; \
uivP = (vector unsigned char *) ui; \ uivP = (vector unsigned char *) ui; \
@@ -357,16 +357,17 @@ static int altivec_##name (SwsContext *c, \
vec_perm(vivP[0], vivP[1], align_perm); \ vec_perm(vivP[0], vivP[1], align_perm); \
\ \
u = (vector signed char) \ u = (vector signed char) \
vec_sub (u,(vector signed char) \ vec_sub(u, \
(vector signed char) \
vec_splat((vector signed char) { 128 }, 0)); \ vec_splat((vector signed char) { 128 }, 0)); \
v = (vector signed char) \ v = (vector signed char) \
vec_sub (v,(vector signed char) \ vec_sub(v, \
(vector signed char) \
vec_splat((vector signed char) { 128 }, 0)); \ vec_splat((vector signed char) { 128 }, 0)); \
\ \
U = vec_unpackh(u); \ U = vec_unpackh(u); \
V = vec_unpackh(v); \ V = vec_unpackh(v); \
\ \
\
Y0 = vec_unh(y0); \ Y0 = vec_unh(y0); \
Y1 = vec_unl(y0); \ Y1 = vec_unl(y0); \
Y2 = vec_unh(y1); \ Y2 = vec_unh(y1); \
@@ -425,7 +426,6 @@ static int altivec_##name (SwsContext *c, \
y2i += 16; \ y2i += 16; \
ui += 8; \ ui += 8; \
vi += 8; \ vi += 8; \
\
} \ } \
\ \
outo += (outstrides[0]) >> 4; \ outo += (outstrides[0]) >> 4; \
@@ -439,11 +439,14 @@ static int altivec_##name (SwsContext *c, \
return srcSliceH; \ return srcSliceH; \
} }
#define out_abgr(a, b, c, ptr) \
#define out_abgr(a,b,c,ptr) vec_mstrgb32(__typeof__(a),((__typeof__ (a)){255}),c,b,a,ptr) vec_mstrgb32(__typeof__(a), ((__typeof__(a)) { 255 }), c, b, a, ptr)
#define out_bgra(a,b,c,ptr) vec_mstrgb32(__typeof__(a),c,b,a,((__typeof__ (a)){255}),ptr) #define out_bgra(a, b, c, ptr) \
#define out_rgba(a,b,c,ptr) vec_mstrgb32(__typeof__(a),a,b,c,((__typeof__ (a)){255}),ptr) vec_mstrgb32(__typeof__(a), c, b, a, ((__typeof__(a)) { 255 }), ptr)
#define out_argb(a,b,c,ptr) vec_mstrgb32(__typeof__(a),((__typeof__ (a)){255}),a,b,c,ptr) #define out_rgba(a, b, c, ptr) \
vec_mstrgb32(__typeof__(a), a, b, c, ((__typeof__(a)) { 255 }), ptr)
#define out_argb(a, b, c, ptr) \
vec_mstrgb32(__typeof__(a), ((__typeof__(a)) { 255 }), a, b, c, ptr)
#define out_rgb24(a, b, c, ptr) vec_mstrgb24(a, b, c, ptr) #define out_rgb24(a, b, c, ptr) vec_mstrgb24(a, b, c, ptr)
#define out_bgr24(a, b, c, ptr) vec_mstbgr24(a, b, c, ptr) #define out_bgr24(a, b, c, ptr) vec_mstbgr24(a, b, c, ptr)
@@ -454,11 +457,9 @@ DEFCSP420_CVT (yuv2_argb, out_argb)
DEFCSP420_CVT(yuv2_rgb24, out_rgb24) DEFCSP420_CVT(yuv2_rgb24, out_rgb24)
DEFCSP420_CVT(yuv2_bgr24, out_bgr24) DEFCSP420_CVT(yuv2_bgr24, out_bgr24)
// uyvy|uyvy|uyvy|uyvy // uyvy|uyvy|uyvy|uyvy
// 0123 4567 89ab cdef // 0123 4567 89ab cdef
static static const vector unsigned char
const vector unsigned char
demux_u = { 0x10, 0x00, 0x10, 0x00, demux_u = { 0x10, 0x00, 0x10, 0x00,
0x10, 0x04, 0x10, 0x04, 0x10, 0x04, 0x10, 0x04,
0x10, 0x08, 0x10, 0x08, 0x10, 0x08, 0x10, 0x08,
@@ -473,11 +474,10 @@ const vector unsigned char
0x10, 0x0D, 0x10, 0x0F }; 0x10, 0x0D, 0x10, 0x0F };
/* /*
this is so I can play live CCIR raw video * this is so I can play live CCIR raw video
*/ */
static int altivec_uyvy_rgb32 (SwsContext *c, static int altivec_uyvy_rgb32(SwsContext *c, const unsigned char **in,
const unsigned char **in, int *instrides, int *instrides, int srcSliceY, int srcSliceH,
int srcSliceY, int srcSliceH,
unsigned char **oplanes, int *outstrides) unsigned char **oplanes, int *outstrides)
{ {
int w = c->srcW; int w = c->srcW;
@@ -493,27 +493,25 @@ static int altivec_uyvy_rgb32 (SwsContext *c,
img = in[0]; img = in[0];
out = (vector unsigned char *) (oplanes[0] + srcSliceY * outstrides[0]); out = (vector unsigned char *) (oplanes[0] + srcSliceY * outstrides[0]);
for (i=0;i<h;i++) { for (i = 0; i < h; i++)
for (j = 0; j < w / 16; j++) { for (j = 0; j < w / 16; j++) {
uyvy = vec_ld(0, img); uyvy = vec_ld(0, img);
U = (vector signed short) U = (vector signed short)
vec_perm(uyvy, (vector unsigned char) { 0 }, demux_u); vec_perm(uyvy, (vector unsigned char) { 0 }, demux_u);
V = (vector signed short) V = (vector signed short)
vec_perm(uyvy, (vector unsigned char) { 0 }, demux_v); vec_perm(uyvy, (vector unsigned char) { 0 }, demux_v);
Y = (vector signed short) Y = (vector signed short)
vec_perm(uyvy, (vector unsigned char) { 0 }, demux_y); vec_perm(uyvy, (vector unsigned char) { 0 }, demux_y);
cvtyuvtoRGB(c, Y, U, V, &R0, &G0, &B0); cvtyuvtoRGB(c, Y, U, V, &R0, &G0, &B0);
uyvy = vec_ld(16, img); uyvy = vec_ld(16, img);
U = (vector signed short) U = (vector signed short)
vec_perm(uyvy, (vector unsigned char) { 0 }, demux_u); vec_perm(uyvy, (vector unsigned char) { 0 }, demux_u);
V = (vector signed short) V = (vector signed short)
vec_perm(uyvy, (vector unsigned char) { 0 }, demux_v); vec_perm(uyvy, (vector unsigned char) { 0 }, demux_v);
Y = (vector signed short) Y = (vector signed short)
vec_perm(uyvy, (vector unsigned char) { 0 }, demux_y); vec_perm(uyvy, (vector unsigned char) { 0 }, demux_y);
@@ -528,17 +526,14 @@ static int altivec_uyvy_rgb32 (SwsContext *c,
img += 32; img += 32;
} }
}
return srcSliceH; return srcSliceH;
} }
/* Ok currently the acceleration routine only supports /* Ok currently the acceleration routine only supports
inputs of widths a multiple of 16 * inputs of widths a multiple of 16
and heights a multiple 2 * and heights a multiple 2
*
So we just fall back to the C codes for this. * So we just fall back to the C codes for this.
*/ */
SwsFunc ff_yuv2rgb_init_altivec(SwsContext *c) SwsFunc ff_yuv2rgb_init_altivec(SwsContext *c)
{ {
@@ -546,13 +541,14 @@ SwsFunc ff_yuv2rgb_init_altivec(SwsContext *c)
return NULL; return NULL;
/* /*
and this seems not to matter too much I tried a bunch of * and this seems not to matter too much I tried a bunch of
videos with abnormal widths and MPlayer crashes elsewhere. * videos with abnormal widths and MPlayer crashes elsewhere.
mplayer -vo x11 -rawvideo on:w=350:h=240 raw-350x240.eyuv * mplayer -vo x11 -rawvideo on:w=350:h=240 raw-350x240.eyuv
boom with X11 bad match. * boom with X11 bad match.
*
*/ */
if ((c->srcW & 0xf) != 0) return NULL; if ((c->srcW & 0xf) != 0)
return NULL;
switch (c->srcFormat) { switch (c->srcFormat) {
case PIX_FMT_YUV410P: case PIX_FMT_YUV410P:
@@ -595,12 +591,13 @@ SwsFunc ff_yuv2rgb_init_altivec(SwsContext *c)
default: return NULL; default: return NULL;
} }
break; break;
} }
return NULL; return NULL;
} }
void ff_yuv2rgb_init_tables_altivec(SwsContext *c, const int inv_table[4], int brightness, int contrast, int saturation) void ff_yuv2rgb_init_tables_altivec(SwsContext *c, const int inv_table[4],
int brightness, int contrast,
int saturation)
{ {
union { union {
DECLARE_ALIGNED(16, signed short, tmp)[8]; DECLARE_ALIGNED(16, signed short, tmp)[8];
@@ -614,7 +611,6 @@ void ff_yuv2rgb_init_tables_altivec(SwsContext *c, const int inv_table[4], int b
buf.tmp[4] = -((inv_table[2] >> 1) * (contrast >> 16) * (saturation >> 16)); // cgu buf.tmp[4] = -((inv_table[2] >> 1) * (contrast >> 16) * (saturation >> 16)); // cgu
buf.tmp[5] = -((inv_table[3] >> 1) * (contrast >> 16) * (saturation >> 16)); // cgv buf.tmp[5] = -((inv_table[3] >> 1) * (contrast >> 16) * (saturation >> 16)); // cgv
c->CSHIFT = (vector unsigned short) vec_splat_u16(2); c->CSHIFT = (vector unsigned short) vec_splat_u16(2);
c->CY = vec_splat((vector signed short) buf.vec, 0); c->CY = vec_splat((vector signed short) buf.vec, 0);
c->OY = vec_splat((vector signed short) buf.vec, 1); c->OY = vec_splat((vector signed short) buf.vec, 1);
@@ -625,14 +621,18 @@ void ff_yuv2rgb_init_tables_altivec(SwsContext *c, const int inv_table[4], int b
return; return;
} }
static av_always_inline void ff_yuv2packedX_altivec(SwsContext *c,
static av_always_inline void const int16_t *lumFilter,
ff_yuv2packedX_altivec(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc,
const int16_t **lumSrc, int lumFilterSize, int lumFilterSize,
const int16_t *chrFilter, const int16_t **chrUSrc, const int16_t *chrFilter,
const int16_t **chrVSrc, int chrFilterSize, const int16_t **chrUSrc,
const int16_t **alpSrc, uint8_t *dest, const int16_t **chrVSrc,
int dstW, int dstY, enum PixelFormat target) int chrFilterSize,
const int16_t **alpSrc,
uint8_t *dest,
int dstW, int dstY,
enum PixelFormat target)
{ {
int i, j; int i, j;
vector signed short X, X0, X1, Y0, U0, V0, Y1, U1, V1, U, V; vector signed short X, X0, X1, Y0, U0, V0, Y1, U1, V1, U, V;
@@ -685,12 +685,12 @@ ff_yuv2packedX_altivec(SwsContext *c, const int16_t *lumFilter,
V = vec_clip_s16(V); V = vec_clip_s16(V);
/* now we have /* now we have
Y0= y0 y1 y2 y3 y4 y5 y6 y7 Y1= y8 y9 y10 y11 y12 y13 y14 y15 * Y0 = y0 y1 y2 y3 y4 y5 y6 y7 Y1 = y8 y9 y10 y11 y12 y13 y14 y15
U= u0 u1 u2 u3 u4 u5 u6 u7 V= v0 v1 v2 v3 v4 v5 v6 v7 * U = u0 u1 u2 u3 u4 u5 u6 u7 V = v0 v1 v2 v3 v4 v5 v6 v7
*
Y0= y0 y1 y2 y3 y4 y5 y6 y7 Y1= y8 y9 y10 y11 y12 y13 y14 y15 * Y0 = y0 y1 y2 y3 y4 y5 y6 y7 Y1 = y8 y9 y10 y11 y12 y13 y14 y15
U0= u0 u0 u1 u1 u2 u2 u3 u3 U1= u4 u4 u5 u5 u6 u6 u7 u7 * U0 = u0 u0 u1 u1 u2 u2 u3 u3 U1 = u4 u4 u5 u5 u6 u6 u7 u7
V0= v0 v0 v1 v1 v2 v2 v3 v3 V1= v4 v4 v5 v5 v6 v6 v7 v7 * V0 = v0 v0 v1 v1 v2 v2 v3 v3 V1 = v4 v4 v5 v5 v6 v6 v7 v7
*/ */
U0 = vec_mergeh(U, U); U0 = vec_mergeh(U, U);
@@ -707,19 +707,32 @@ ff_yuv2packedX_altivec(SwsContext *c, const int16_t *lumFilter,
B = vec_packclp(B0, B1); B = vec_packclp(B0, B1);
switch (target) { switch (target) {
case PIX_FMT_ABGR: out_abgr (R,G,B,out); break; case PIX_FMT_ABGR:
case PIX_FMT_BGRA: out_bgra (R,G,B,out); break; out_abgr(R, G, B, out);
case PIX_FMT_RGBA: out_rgba (R,G,B,out); break; break;
case PIX_FMT_ARGB: out_argb (R,G,B,out); break; case PIX_FMT_BGRA:
case PIX_FMT_RGB24: out_rgb24 (R,G,B,out); break; out_bgra(R, G, B, out);
case PIX_FMT_BGR24: out_bgr24 (R,G,B,out); break; break;
case PIX_FMT_RGBA:
out_rgba(R, G, B, out);
break;
case PIX_FMT_ARGB:
out_argb(R, G, B, out);
break;
case PIX_FMT_RGB24:
out_rgb24(R, G, B, out);
break;
case PIX_FMT_BGR24:
out_bgr24(R, G, B, out);
break;
default: default:
{ {
/* If this is reached, the caller should have called yuv2packedXinC /* If this is reached, the caller should have called yuv2packedXinC
instead. */ * instead. */
static int printed_error_message; static int printed_error_message;
if (!printed_error_message) { if (!printed_error_message) {
av_log(c, AV_LOG_ERROR, "altivec_yuv2packedX doesn't support %s output\n", av_log(c, AV_LOG_ERROR,
"altivec_yuv2packedX doesn't support %s output\n",
sws_format_name(c->dstFormat)); sws_format_name(c->dstFormat));
printed_error_message = 1; printed_error_message = 1;
} }
@@ -763,12 +776,12 @@ ff_yuv2packedX_altivec(SwsContext *c, const int16_t *lumFilter,
V = vec_clip_s16(V); V = vec_clip_s16(V);
/* now we have /* now we have
Y0= y0 y1 y2 y3 y4 y5 y6 y7 Y1= y8 y9 y10 y11 y12 y13 y14 y15 * Y0 = y0 y1 y2 y3 y4 y5 y6 y7 Y1 = y8 y9 y10 y11 y12 y13 y14 y15
U = u0 u1 u2 u3 u4 u5 u6 u7 V = v0 v1 v2 v3 v4 v5 v6 v7 * U = u0 u1 u2 u3 u4 u5 u6 u7 V = v0 v1 v2 v3 v4 v5 v6 v7
*
Y0= y0 y1 y2 y3 y4 y5 y6 y7 Y1= y8 y9 y10 y11 y12 y13 y14 y15 * Y0 = y0 y1 y2 y3 y4 y5 y6 y7 Y1 = y8 y9 y10 y11 y12 y13 y14 y15
U0= u0 u0 u1 u1 u2 u2 u3 u3 U1= u4 u4 u5 u5 u6 u6 u7 u7 * U0 = u0 u0 u1 u1 u2 u2 u3 u3 U1 = u4 u4 u5 u5 u6 u6 u7 u7
V0= v0 v0 v1 v1 v2 v2 v3 v3 V1= v4 v4 v5 v5 v6 v6 v7 v7 * V0 = v0 v0 v1 v1 v2 v2 v3 v3 V1 = v4 v4 v5 v5 v6 v6 v7 v7
*/ */
U0 = vec_mergeh(U, U); U0 = vec_mergeh(U, U);
@@ -786,35 +799,52 @@ ff_yuv2packedX_altivec(SwsContext *c, const int16_t *lumFilter,
nout = (vector unsigned char *) scratch; nout = (vector unsigned char *) scratch;
switch (target) { switch (target) {
case PIX_FMT_ABGR: out_abgr (R,G,B,nout); break; case PIX_FMT_ABGR:
case PIX_FMT_BGRA: out_bgra (R,G,B,nout); break; out_abgr(R, G, B, nout);
case PIX_FMT_RGBA: out_rgba (R,G,B,nout); break; break;
case PIX_FMT_ARGB: out_argb (R,G,B,nout); break; case PIX_FMT_BGRA:
case PIX_FMT_RGB24: out_rgb24 (R,G,B,nout); break; out_bgra(R, G, B, nout);
case PIX_FMT_BGR24: out_bgr24 (R,G,B,nout); break; break;
case PIX_FMT_RGBA:
out_rgba(R, G, B, nout);
break;
case PIX_FMT_ARGB:
out_argb(R, G, B, nout);
break;
case PIX_FMT_RGB24:
out_rgb24(R, G, B, nout);
break;
case PIX_FMT_BGR24:
out_bgr24(R, G, B, nout);
break;
default: default:
/* Unreachable, I think. */ /* Unreachable, I think. */
av_log(c, AV_LOG_ERROR, "altivec_yuv2packedX doesn't support %s output\n", av_log(c, AV_LOG_ERROR,
"altivec_yuv2packedX doesn't support %s output\n",
sws_format_name(c->dstFormat)); sws_format_name(c->dstFormat));
return; return;
} }
memcpy(&((uint32_t *) dest)[i], scratch, (dstW - i) / 4); memcpy(&((uint32_t *) dest)[i], scratch, (dstW - i) / 4);
} }
} }
#define YUV2PACKEDX_WRAPPER(suffix, pixfmt) \ #define YUV2PACKEDX_WRAPPER(suffix, pixfmt) \
void ff_yuv2 ## suffix ## _X_altivec(SwsContext *c, const int16_t *lumFilter, \ void ff_yuv2 ## suffix ## _X_altivec(SwsContext *c, \
const int16_t **lumSrc, int lumFilterSize, \ const int16_t *lumFilter, \
const int16_t *chrFilter, const int16_t **chrUSrc, \ const int16_t **lumSrc, \
const int16_t **chrVSrc, int chrFilterSize, \ int lumFilterSize, \
const int16_t **alpSrc, uint8_t *dest, \ const int16_t *chrFilter, \
int dstW, int dstY) \ const int16_t **chrUSrc, \
const int16_t **chrVSrc, \
int chrFilterSize, \
const int16_t **alpSrc, \
uint8_t *dest, int dstW, int dstY) \
{ \ { \
ff_yuv2packedX_altivec(c, lumFilter, lumSrc, lumFilterSize, \ ff_yuv2packedX_altivec(c, lumFilter, lumSrc, lumFilterSize, \
chrFilter, chrUSrc, chrVSrc, chrFilterSize, \ chrFilter, chrUSrc, chrVSrc, \
alpSrc, dest, dstW, dstY, pixfmt); \ chrFilterSize, alpSrc, \
dest, dstW, dstY, pixfmt); \
} }
YUV2PACKEDX_WRAPPER(abgr, PIX_FMT_ABGR); YUV2PACKEDX_WRAPPER(abgr, PIX_FMT_ABGR);