Skip to content

F
ffmpeg.wasm-core
Commit 04217de4 authored by Diego Biurrun's avatar Diego Biurrun
Browse files
Options

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

parent 3f486e0d
Show whitespace changes
Inline Side-by-side
Showing with 314 additions and 303 deletions
libswscale/ppc/swscale_altivec.c
View file @ 04217de4
......@@ -22,6 +22,7 @@
*/
#include <inttypes.h>
#include "config.h"
#include "libswscale/swscale.h"
#include "libswscale/swscale_internal.h"
......@@ -30,8 +31,8 @@
#define vzero vec_splat_s32(0)
static inline void
altivec_packIntArrayToCharArray(int *val, uint8_t* dest, int dstW)
static inline void altivec_packIntArrayToCharArray(int *val, uint8_t *dest,
int dstW)
{
register int i;
vector unsigned int altivec_vectorShiftInt19 =
......@@ -41,14 +42,14 @@ altivec_packIntArrayToCharArray(int *val, uint8_t* dest, int dstW)
/* and will handle load misalignment on val w/ vec_perm */
vector unsigned char perm1;
vector signed int v1;
for (i = 0 ; (i < dstW) &&
(((uintptr_t)dest + i) % 16) ; i++) {
for (i = 0; (i < dstW) &&
(((uintptr_t)dest + i) % 16); i++) {
int t = val[i] >> 19;
dest[i] = (t < 0) ? 0 : ((t > 255) ? 255 : t);
}
perm1 = vec_lvsl(i << 2, val);
v1 = vec_ld(i << 2, val);
for ( ; i < (dstW - 15); i+=16) {
for (; i < (dstW - 15); i += 16) {
int offset = i << 2;
vector signed int v2 = vec_ld(offset + 16, val);
vector signed int v3 = vec_ld(offset + 32, val);
......@@ -70,7 +71,7 @@ altivec_packIntArrayToCharArray(int *val, uint8_t* dest, int dstW)
v1 = v5;
}
} else { // dest is properly aligned, great
for (i = 0; i < (dstW - 15); i+=16) {
for (i = 0; i < (dstW - 15); i += 16) {
int offset = i << 2;
vector signed int v1 = vec_ld(offset, val);
vector signed int v2 = vec_ld(offset + 16, val);
......@@ -86,22 +87,21 @@ altivec_packIntArrayToCharArray(int *val, uint8_t* dest, int dstW)
vec_st(vf, i, dest);
}
}
for ( ; i < dstW ; i++) {
for (; i < dstW; i++) {
int t = val[i] >> 19;
dest[i] = (t < 0) ? 0 : ((t > 255) ? 255 : t);
}
}
//FIXME remove the usage of scratch buffers.
static void
yuv2planeX_altivec(const int16_t *filter, int filterSize,
// FIXME remove the usage of scratch buffers.
static void yuv2planeX_altivec(const int16_t *filter, int filterSize,
const int16_t **src, uint8_t *dest, int dstW,
const uint8_t *dither, int offset)
{
register int i, j;
DECLARE_ALIGNED(16, int, val)[dstW];
for (i=0; i<dstW; i++)
for (i = 0; i < dstW; i++)
val[i] = dither[(i + offset) & 7] << 12;
for (j = 0; j < filterSize; j++) {
......@@ -113,7 +113,7 @@ yuv2planeX_altivec(const int16_t *filter, int filterSize,
perm = vec_lvsl(0, src[j]);
l1 = vec_ld(0, src[j]);
for (i = 0; i < (dstW - 7); i+=8) {
for (i = 0; i < (dstW - 7); i += 8) {
int offset = i << 2;
vector signed short l2 = vec_ld((i << 1) + 16, src[j]);
......@@ -136,10 +136,9 @@ yuv2planeX_altivec(const int16_t *filter, int filterSize,
l1 = l2;
}
for ( ; i < dstW; i++) {
for (; i < dstW; i++)
val[i] += src[j][i] * filter[j];
}
}
altivec_packIntArrayToCharArray(val, dest, dstW);
}
......@@ -151,20 +150,18 @@ static void hScale_altivec_real(SwsContext *c, int16_t *dst, int dstW,
DECLARE_ALIGNED(16, int, tempo)[4];
if (filterSize % 4) {
for (i=0; i<dstW; i++) {
for (i = 0; i < dstW; i++) {
register int j;
register int srcPos = filterPos[i];
register int val = 0;
for (j=0; j<filterSize; j++) {
val += ((int)src[srcPos + j])*filter[filterSize*i + j];
}
dst[i] = FFMIN(val>>7, (1<<15)-1);
for (j = 0; j < filterSize; j++)
val += ((int)src[srcPos + j]) * filter[filterSize * i + j];
dst[i] = FFMIN(val >> 7, (1 << 15) - 1);
}
}
else
} else
switch (filterSize) {
case 4:
for (i=0; i<dstW; i++) {
for (i = 0; i < dstW; i++) {
register int srcPos = filterPos[i];
vector unsigned char src_v0 = vec_ld(srcPos, src);
......@@ -196,12 +193,12 @@ static void hScale_altivec_real(SwsContext *c, int16_t *dst, int dstW,
val_vEven = vec_mule(src_v, filter_v);
val_s = vec_sums(val_vEven, vzero);
vec_st(val_s, 0, tempo);
dst[i] = FFMIN(tempo[3]>>7, (1<<15)-1);
dst[i] = FFMIN(tempo[3] >> 7, (1 << 15) - 1);
}
break;
case 8:
for (i=0; i<dstW; i++) {
for (i = 0; i < dstW; i++) {
register int srcPos = filterPos[i];
vector unsigned char src_v0 = vec_ld(srcPos, src);
......@@ -221,12 +218,12 @@ static void hScale_altivec_real(SwsContext *c, int16_t *dst, int dstW,
val_v = vec_msums(src_v, filter_v, (vector signed int)vzero);
val_s = vec_sums(val_v, vzero);
vec_st(val_s, 0, tempo);
dst[i] = FFMIN(tempo[3]>>7, (1<<15)-1);
dst[i] = FFMIN(tempo[3] >> 7, (1 << 15) - 1);
}
break;
case 16:
for (i=0; i<dstW; i++) {
for (i = 0; i < dstW; i++) {
register int srcPos = filterPos[i];
vector unsigned char src_v0 = vec_ld(srcPos, src);
......@@ -248,12 +245,12 @@ static void hScale_altivec_real(SwsContext *c, int16_t *dst, int dstW,
vector signed int val_s = vec_sums(val_v, vzero);
vec_st(val_s, 0, tempo);
dst[i] = FFMIN(tempo[3]>>7, (1<<15)-1);
dst[i] = FFMIN(tempo[3] >> 7, (1 << 15) - 1);
}
break;
default:
for (i=0; i<dstW; i++) {
for (i = 0; i < dstW; i++) {
register int j;
register int srcPos = filterPos[i];
......@@ -264,7 +261,7 @@ static void hScale_altivec_real(SwsContext *c, int16_t *dst, int dstW,
vector unsigned char src_v0 = vec_ld(srcPos, src);
vector unsigned char permS = vec_lvsl(srcPos, src);
for (j = 0 ; j < filterSize - 15; j += 16) {
for (j = 0; j < filterSize - 15; j += 16) {
vector unsigned char src_v1 = vec_ld(srcPos + j + 16, src);
vector unsigned char src_vF = vec_perm(src_v0, src_v1, permS);
......@@ -285,9 +282,9 @@ static void hScale_altivec_real(SwsContext *c, int16_t *dst, int dstW,
src_v0 = src_v1;
}
if (j < filterSize-7) {
if (j < filterSize - 7) {
// loading src_v0 is useless, it's already done above
//vector unsigned char src_v0 = vec_ld(srcPos + j, src);
// vector unsigned char src_v0 = vec_ld(srcPos + j, src);
vector unsigned char src_v1, src_vF;
vector signed short src_v, filter_v1R, filter_v;
if ((((uintptr_t)src + srcPos) % 16) > 8) {
......@@ -298,7 +295,7 @@ static void hScale_altivec_real(SwsContext *c, int16_t *dst, int dstW,
src_v = // vec_unpackh sign-extends...
(vector signed short)(vec_mergeh((vector unsigned char)vzero, src_vF));
// loading filter_v0R is useless, it's already done above
//vector signed short filter_v0R = vec_ld((i * 2 * filterSize) + j, filter);
// vector signed short filter_v0R = vec_ld((i * 2 * filterSize) + j, filter);
filter_v1R = vec_ld((i * 2 * filterSize) + (j * 2) + 16, filter);
filter_v = vec_perm(filter_v0R, filter_v1R, permF);
......@@ -308,9 +305,8 @@ static void hScale_altivec_real(SwsContext *c, int16_t *dst, int dstW,
val_s = vec_sums(val_v, vzero);
vec_st(val_s, 0, tempo);
dst[i] = FFMIN(tempo[3]>>7, (1<<15)-1);
dst[i] = FFMIN(tempo[3] >> 7, (1 << 15) - 1);
}
}
}
......@@ -334,12 +330,24 @@ void ff_sws_init_swScale_altivec(SwsContext *c)
* match what's found in the body of ff_yuv2packedX_altivec() */
if (!(c->flags & (SWS_BITEXACT | SWS_FULL_CHR_H_INT)) && !c->alpPixBuf) {
switch (c->dstFormat) {
case PIX_FMT_ABGR: c->yuv2packedX = ff_yuv2abgr_X_altivec; break;
case PIX_FMT_BGRA: c->yuv2packedX = ff_yuv2bgra_X_altivec; break;
case PIX_FMT_ARGB: c->yuv2packedX = ff_yuv2argb_X_altivec; break;
case PIX_FMT_RGBA: c->yuv2packedX = ff_yuv2rgba_X_altivec; break;
case PIX_FMT_BGR24: c->yuv2packedX = ff_yuv2bgr24_X_altivec; break;
case PIX_FMT_RGB24: c->yuv2packedX = ff_yuv2rgb24_X_altivec; break;
case PIX_FMT_ABGR:
c->yuv2packedX = ff_yuv2abgr_X_altivec;
break;
case PIX_FMT_BGRA:
c->yuv2packedX = ff_yuv2bgra_X_altivec;
break;
case PIX_FMT_ARGB:
c->yuv2packedX = ff_yuv2argb_X_altivec;
break;
case PIX_FMT_RGBA:
c->yuv2packedX = ff_yuv2rgba_X_altivec;
break;
case PIX_FMT_BGR24:
c->yuv2packedX = ff_yuv2bgr24_X_altivec;
break;
case PIX_FMT_RGB24:
c->yuv2packedX = ff_yuv2rgb24_X_altivec;
break;
}
}
}
libswscale/ppc/yuv2yuv_altivec.c
View file @ 04217de4
......@@ -22,18 +22,20 @@
*/
#include <inttypes.h>
#include "config.h"
#include "libswscale/swscale.h"
#include "libswscale/swscale_internal.h"
#include "libavutil/cpu.h"
static int yv12toyuy2_unscaled_altivec(SwsContext *c, const uint8_t* src[],
static int yv12toyuy2_unscaled_altivec(SwsContext *c, const uint8_t *src[],
int srcStride[], int srcSliceY,
int srcSliceH, uint8_t* dstParam[],
int srcSliceH, uint8_t *dstParam[],
int dstStride_a[])
{
uint8_t *dst=dstParam[0] + dstStride_a[0]*srcSliceY;
// yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
uint8_t *dst = dstParam[0] + dstStride_a[0] * srcSliceY;
// yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH,
// srcStride[0], srcStride[1], dstStride[0]);
const uint8_t *ysrc = src[0];
const uint8_t *usrc = src[1];
const uint8_t *vsrc = src[2];
......@@ -47,16 +49,16 @@ static int yv12toyuy2_unscaled_altivec(SwsContext *c, const uint8_t* src[],
register unsigned int y;
/* This code assumes:
1) dst is 16 bytes-aligned
2) dstStride is a multiple of 16
3) width is a multiple of 16
4) lum & chrom stride are multiples of 8
*
* 1) dst is 16 bytes-aligned
* 2) dstStride is a multiple of 16
* 3) width is a multiple of 16
* 4) lum & chrom stride are multiples of 8
*/
for (y=0; y<height; y++) {
for (y = 0; y < height; y++) {
int i;
for (i = 0; i < width - 31; i+= 32) {
for (i = 0; i < width - 31; i += 32) {
const unsigned int j = i >> 1;
vector unsigned char v_yA = vec_ld(i, ysrc);
vector unsigned char v_yB = vec_ld(i + 16, ysrc);
......@@ -91,7 +93,7 @@ static int yv12toyuy2_unscaled_altivec(SwsContext *c, const uint8_t* src[],
vec_st(v_yuy2_0, (i << 1), dst);
vec_st(v_yuy2_1, (i << 1) + 16, dst);
}
if ((y&(vertLumPerChroma-1)) == vertLumPerChroma-1) {
if ((y & (vertLumPerChroma - 1)) == vertLumPerChroma - 1) {
usrc += chromStride;
vsrc += chromStride;
}
......@@ -102,13 +104,14 @@ static int yv12toyuy2_unscaled_altivec(SwsContext *c, const uint8_t* src[],
return srcSliceH;
}
static int yv12touyvy_unscaled_altivec(SwsContext *c, const uint8_t* src[],
static int yv12touyvy_unscaled_altivec(SwsContext *c, const uint8_t *src[],
int srcStride[], int srcSliceY,
int srcSliceH, uint8_t* dstParam[],
int srcSliceH, uint8_t *dstParam[],
int dstStride_a[])
{
uint8_t *dst=dstParam[0] + dstStride_a[0]*srcSliceY;
// yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
uint8_t *dst = dstParam[0] + dstStride_a[0] * srcSliceY;
// yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH,
// srcStride[0], srcStride[1], dstStride[0]);
const uint8_t *ysrc = src[0];
const uint8_t *usrc = src[1];
const uint8_t *vsrc = src[2];
......@@ -122,16 +125,16 @@ static int yv12touyvy_unscaled_altivec(SwsContext *c, const uint8_t* src[],
register unsigned int y;
/* This code assumes:
1) dst is 16 bytes-aligned
2) dstStride is a multiple of 16
3) width is a multiple of 16
4) lum & chrom stride are multiples of 8
*
* 1) dst is 16 bytes-aligned
* 2) dstStride is a multiple of 16
* 3) width is a multiple of 16
* 4) lum & chrom stride are multiples of 8
*/
for (y=0; y<height; y++) {
for (y = 0; y < height; y++) {
int i;
for (i = 0; i < width - 31; i+= 32) {
for (i = 0; i < width - 31; i += 32) {
const unsigned int j = i >> 1;
vector unsigned char v_yA = vec_ld(i, ysrc);
vector unsigned char v_yB = vec_ld(i + 16, ysrc);
......@@ -166,7 +169,7 @@ static int yv12touyvy_unscaled_altivec(SwsContext *c, const uint8_t* src[],
vec_st(v_uyvy_0, (i << 1), dst);
vec_st(v_uyvy_1, (i << 1) + 16, dst);
}
if ((y&(vertLumPerChroma-1)) == vertLumPerChroma-1) {
if ((y & (vertLumPerChroma - 1)) == vertLumPerChroma - 1) {
usrc += chromStride;
vsrc += chromStride;
}
......@@ -184,8 +187,8 @@ void ff_swscale_get_unscaled_altivec(SwsContext *c)
// unscaled YV12 -> packed YUV, we want speed
if (dstFormat == PIX_FMT_YUYV422)
c->swScale= yv12toyuy2_unscaled_altivec;
c->swScale = yv12toyuy2_unscaled_altivec;
else if (dstFormat == PIX_FMT_UYVY422)
c->swScale= yv12touyvy_unscaled_altivec;
c->swScale = yv12touyvy_unscaled_altivec;
}
}
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment