Chiebot-Mirror
/
FFmpeg
mirror of https://github.com/FFmpeg/FFmpeg.git
8
0
Fork 0
Code Issues Projects Releases Wiki Activity

ppc: dsputil: K&R formatting cosmetics

Browse Source
pull/293/head
Diego Biurrun 11 years ago
parent 82ee14d2ce
commit 30f3f95987
3 changed files with 429 additions and 391 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 137
      libavcodec/ppc/dsputil_altivec.c
  2. 8
      libavcodec/ppc/dsputil_altivec.h
  3. 17
      libavcodec/ppc/dsputil_ppc.c

137
libavcodec/ppc/dsputil_altivec.c
Unescape View File

@ -24,6 +24,7 @@
#if HAVE_ALTIVEC_H
#include <altivec.h>
#endif
#include "libavutil/attributes.h"
#include "libavutil/ppc/types_altivec.h"
#include "libavutil/ppc/util_altivec.h"
@ -31,11 +32,13 @@
#include "libavcodec/dsputil.h"
#include "dsputil_altivec.h"
static int sad16_x2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
static int sad16_x2_altivec(void *v, uint8_t *pix1, uint8_t *pix2,
int line_size, int h)
{
int i;
int s;
const vector unsigned char zero = (const vector unsigned char)vec_splat_u8(0);
const vector unsigned char zero =
(const vector unsigned char) vec_splat_u8(0);
vector unsigned char perm1 = vec_lvsl(0, pix2);
vector unsigned char perm2 = vec_add(perm1, vec_splat_u8(1));
vector unsigned char pix2l, pix2r;
@ -75,11 +78,13 @@ static int sad16_x2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size
return s;
}
static int sad16_y2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
static int sad16_y2_altivec(void *v, uint8_t *pix1, uint8_t *pix2,
int line_size, int h)
{
int i;
int s;
const vector unsigned char zero = (const vector unsigned char)vec_splat_u8(0);
const vector unsigned char zero =
(const vector unsigned char) vec_splat_u8(0);
vector unsigned char perm = vec_lvsl(0, pix2);
vector unsigned char pix2l, pix2r;
vector unsigned char pix1v, pix2v, pix3v, avgv, t5;
@ -123,7 +128,6 @@ static int sad16_y2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size
pix1 += line_size;
pix2v = pix3v;
pix3 += line_size;
}
/* Sum up the four partial sums, and put the result into s. */
@ -133,13 +137,16 @@ static int sad16_y2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size
return s;
}
static int sad16_xy2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
static int sad16_xy2_altivec(void *v, uint8_t *pix1, uint8_t *pix2,
int line_size, int h)
{
int i;
int s;
uint8_t *pix3 = pix2 + line_size;
const vector unsigned char zero = (const vector unsigned char)vec_splat_u8(0);
const vector unsigned short two = (const vector unsigned short)vec_splat_u16(2);
const vector unsigned char zero =
(const vector unsigned char) vec_splat_u8(0);
const vector unsigned short two =
(const vector unsigned short) vec_splat_u16(2);
vector unsigned char avgv, t5;
vector unsigned char perm1 = vec_lvsl(0, pix2);
vector unsigned char perm2 = vec_add(perm1, vec_splat_u8(1));
@ -228,11 +235,13 @@ static int sad16_xy2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_siz
return s;
}
static int sad16_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
static int sad16_altivec(void *v, uint8_t *pix1, uint8_t *pix2,
int line_size, int h)
{
int i;
int s;
const vector unsigned int zero = (const vector unsigned int)vec_splat_u32(0);
const vector unsigned int zero =
(const vector unsigned int) vec_splat_u32(0);
vector unsigned char perm = vec_lvsl(0, pix2);
vector unsigned char t1, t2, t3, t4, t5;
vector unsigned int sad;
@ -240,7 +249,6 @@ static int sad16_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, i
sad = (vector unsigned int) vec_splat_u32(0);
for (i = 0; i < h; i++) {
/* Read potentially unaligned pixels into t1 and t2. */
vector unsigned char pix2l = vec_ld(0, pix2);
@ -268,12 +276,16 @@ static int sad16_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, i
return s;
}
static int sad8_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
static int sad8_altivec(void *v, uint8_t *pix1, uint8_t *pix2,
int line_size, int h)
{
int i;
int s;
const vector unsigned int zero = (const vector unsigned int)vec_splat_u32(0);
const vector unsigned char permclear = (vector unsigned char){255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0};
const vector unsigned int zero =
(const vector unsigned int) vec_splat_u32(0);
const vector unsigned char permclear =
(vector unsigned char)
{ 255, 255, 255, 255, 255, 255, 255, 255, 0, 0, 0, 0, 0, 0, 0, 0 };
vector unsigned char perm1 = vec_lvsl(0, pix1);
vector unsigned char perm2 = vec_lvsl(0, pix2);
vector unsigned char t1, t2, t3, t4, t5;
@ -317,7 +329,8 @@ static int pix_norm1_altivec(uint8_t *pix, int line_size)
{
int i;
int s;
const vector unsigned int zero = (const vector unsigned int)vec_splat_u32(0);
const vector unsigned int zero =
(const vector unsigned int) vec_splat_u32(0);
vector unsigned char perm = vec_lvsl(0, pix);
vector unsigned char pixv;
vector unsigned int sv;
@ -347,12 +360,16 @@ static int pix_norm1_altivec(uint8_t *pix, int line_size)
/* Sum of Squared Errors for an 8x8 block, AltiVec-enhanced.
* It's the sad8_altivec code above w/ squaring added. */
static int sse8_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
static int sse8_altivec(void *v, uint8_t *pix1, uint8_t *pix2,
int line_size, int h)
{
int i;
int s;
const vector unsigned int zero = (const vector unsigned int)vec_splat_u32(0);
const vector unsigned char permclear = (vector unsigned char){255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0};
const vector unsigned int zero =
(const vector unsigned int) vec_splat_u32(0);
const vector unsigned char permclear =
(vector unsigned char)
{ 255, 255, 255, 255, 255, 255, 255, 255, 0, 0, 0, 0, 0, 0, 0, 0 };
vector unsigned char perm1 = vec_lvsl(0, pix1);
vector unsigned char perm2 = vec_lvsl(0, pix2);
vector unsigned char t1, t2, t3, t4, t5;
@ -397,11 +414,13 @@ static int sse8_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, in
/* Sum of Squared Errors for a 16x16 block, AltiVec-enhanced.
* It's the sad16_altivec code above w/ squaring added. */
static int sse16_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
static int sse16_altivec(void *v, uint8_t *pix1, uint8_t *pix2,
int line_size, int h)
{
int i;
int s;
const vector unsigned int zero = (const vector unsigned int)vec_splat_u32(0);
const vector unsigned int zero =
(const vector unsigned int) vec_splat_u32(0);
vector unsigned char perm = vec_lvsl(0, pix2);
vector unsigned char t1, t2, t3, t4, t5;
vector unsigned int sum;
@ -441,7 +460,8 @@ static int sse16_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, i
static int pix_sum_altivec(uint8_t *pix, int line_size)
{
const vector unsigned int zero = (const vector unsigned int)vec_splat_u32(0);
const vector unsigned int zero =
(const vector unsigned int) vec_splat_u32(0);
vector unsigned char perm = vec_lvsl(0, pix);
vector unsigned char t1;
vector unsigned int sad;
@ -472,21 +492,20 @@ static int pix_sum_altivec(uint8_t * pix, int line_size)
return s;
}
static void get_pixels_altivec(int16_t *restrict block, const uint8_t *pixels, int line_size)
static void get_pixels_altivec(int16_t *restrict block, const uint8_t *pixels,
int line_size)
{
int i;
vector unsigned char perm = vec_lvsl(0, pixels);
vector unsigned char bytes;
const vector unsigned char zero = (const vector unsigned char)vec_splat_u8(0);
const vector unsigned char zero =
(const vector unsigned char) vec_splat_u8(0);
vector signed short shorts;
for (i = 0; i < 8; i++) {
/* Read potentially unaligned pixels.
* We're reading 16 pixels, and actually only want 8,
* but we simply ignore the extras. */
// Read potentially unaligned pixels.
// We're reading 16 pixels, and actually only want 8,
// but we simply ignore the extras.
vector unsigned char pixl = vec_ld(0, pixels);
vector unsigned char pixr = vec_ld(7, pixels);
bytes = vec_perm(pixl, pixr, perm);
@ -508,7 +527,8 @@ static void diff_pixels_altivec(int16_t *restrict block, const uint8_t *s1,
vector unsigned char perm1 = vec_lvsl(0, s1);
vector unsigned char perm2 = vec_lvsl(0, s2);
vector unsigned char bytes, pixl, pixr;
const vector unsigned char zero = (const vector unsigned char)vec_splat_u8(0);
const vector unsigned char zero =
(const vector unsigned char) vec_splat_u8(0);
vector signed short shorts1, shorts2;
for (i = 0; i < 4; i++) {
@ -573,8 +593,8 @@ static void diff_pixels_altivec(int16_t *restrict block, const uint8_t *s1,
}
}
static void clear_block_altivec(int16_t *block) {
static void clear_block_altivec(int16_t *block)
{
LOAD_ZERO;
vec_st(zero_s16v, 0, block);
vec_st(zero_s16v, 16, block);
@ -586,8 +606,8 @@ static void clear_block_altivec(int16_t *block) {
vec_st(zero_s16v, 112, block);
}
static void add_bytes_altivec(uint8_t *dst, uint8_t *src, int w) {
static void add_bytes_altivec(uint8_t *dst, uint8_t *src, int w)
{
register int i;
register vector unsigned char vdst, vsrc;
@ -599,31 +619,35 @@ static void add_bytes_altivec(uint8_t *dst, uint8_t *src, int w) {
vec_st(vdst, i, (unsigned char *) dst);
}
/* If w is not a multiple of 16. */
for (; (i < w) ; i++) {
for (; (i < w); i++)
dst[i] = src[i];
}
}
static int hadamard8_diff8x8_altivec(/*MpegEncContext*/ void *s, uint8_t *dst, uint8_t *src, int stride, int h){
static int hadamard8_diff8x8_altivec(/* MpegEncContext */ void *s, uint8_t *dst,
uint8_t *src, int stride, int h)
{
int sum;
register const vector unsigned char vzero =
(const vector unsigned char) vec_splat_u8(0);
register vector signed short temp0, temp1, temp2, temp3, temp4,
temp5, temp6, temp7;
{
register const vector signed short vprod1 =(const vector signed short)
{ 1,-1, 1,-1, 1,-1, 1,-1 };
register const vector signed short vprod2 =(const vector signed short)
{ 1, 1,-1,-1, 1, 1,-1,-1 };
register const vector signed short vprod3 =(const vector signed short)
{ 1, 1, 1, 1,-1,-1,-1,-1 };
register const vector unsigned char perm1 = (const vector unsigned char)
register const vector signed short vprod1 =
(const vector signed short) { 1, -1, 1, -1, 1, -1, 1, -1 };
register const vector signed short vprod2 =
(const vector signed short) { 1, 1, -1, -1, 1, 1, -1, -1 };
register const vector signed short vprod3 =
(const vector signed short) { 1, 1, 1, 1, -1, -1, -1, -1 };
register const vector unsigned char perm1 =
(const vector unsigned char)
{ 0x02, 0x03, 0x00, 0x01, 0x06, 0x07, 0x04, 0x05,
0x0A, 0x0B, 0x08, 0x09, 0x0E, 0x0F, 0x0C, 0x0D };
register const vector unsigned char perm2 = (const vector unsigned char)
register const vector unsigned char perm2 =
(const vector unsigned char)
{ 0x04, 0x05, 0x06, 0x07, 0x00, 0x01, 0x02, 0x03,
0x0C, 0x0D, 0x0E, 0x0F, 0x08, 0x09, 0x0A, 0x0B };
register const vector unsigned char perm3 = (const vector unsigned char)
register const vector unsigned char perm3 =
(const vector unsigned char)
{ 0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F,
0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07 };
@ -726,7 +750,9 @@ static int hadamard8_diff8x8_altivec(/*MpegEncContext*/ void *s, uint8_t *dst, u
* On the 970, the hand-made RA is still a win (around 690 vs. around 780),
* but xlc goes to around 660 on the regular C code...
*/
static int hadamard8_diff16x8_altivec(/*MpegEncContext*/ void *s, uint8_t *dst, uint8_t *src, int stride, int h) {
static int hadamard8_diff16x8_altivec(/* MpegEncContext */ void *s, uint8_t *dst,
uint8_t *src, int stride, int h)
{
int sum;
register vector signed short
temp0 __asm__ ("v0"),
@ -751,18 +777,23 @@ static int hadamard8_diff16x8_altivec(/*MpegEncContext*/ void *s, uint8_t *dst,
{
register const vector signed short vprod1 __asm__ ("v16") =
(const vector signed short) { 1, -1, 1, -1, 1, -1, 1, -1 };
register const vector signed short vprod2 __asm__ ("v17") =
(const vector signed short) { 1, 1, -1, -1, 1, 1, -1, -1 };
register const vector signed short vprod3 __asm__ ("v18") =
(const vector signed short) { 1, 1, 1, 1, -1, -1, -1, -1 };
register const vector unsigned char perm1 __asm__ ("v19") =
(const vector unsigned char)
{ 0x02, 0x03, 0x00, 0x01, 0x06, 0x07, 0x04, 0x05,
0x0A, 0x0B, 0x08, 0x09, 0x0E, 0x0F, 0x0C, 0x0D };
register const vector unsigned char perm2 __asm__ ("v20") =
(const vector unsigned char)
{ 0x04, 0x05, 0x06, 0x07, 0x00, 0x01, 0x02, 0x03,
0x0C, 0x0D, 0x0E, 0x0F, 0x08, 0x09, 0x0A, 0x0B };
register const vector unsigned char perm3 __asm__ ("v21") =
(const vector unsigned char)
{ 0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F,
@ -770,14 +801,16 @@ static int hadamard8_diff16x8_altivec(/*MpegEncContext*/ void *s, uint8_t *dst,
#define ONEITERBUTTERFLY(i, res1, res2) \
{ \
register vector unsigned char src1 __asm__ ("v22"), \
register vector unsigned char \
src1 __asm__ ("v22"), \
src2 __asm__ ("v23"), \
dst1 __asm__ ("v24"), \
dst2 __asm__ ("v25"), \
srcO __asm__ ("v22"), \
dstO __asm__ ("v23"); \
\
register vector signed short srcV __asm__ ("v24"), \
register vector signed short \
srcV __asm__ ("v24"), \
dstV __asm__ ("v25"), \
srcW __asm__ ("v26"), \
dstW __asm__ ("v27"), \
@ -921,7 +954,9 @@ static int hadamard8_diff16x8_altivec(/*MpegEncContext*/ void *s, uint8_t *dst,
return sum;
}
static int hadamard8_diff16_altivec(/*MpegEncContext*/ void *s, uint8_t *dst, uint8_t *src, int stride, int h){
static int hadamard8_diff16_altivec(/* MpegEncContext */ void *s, uint8_t *dst,
uint8_t *src, int stride, int h)
{
int score;
score = hadamard8_diff16x8_altivec(s, dst, src, stride, 8);
if (h == 16) {
@ -941,14 +976,18 @@ av_cold void ff_dsputil_init_altivec(DSPContext *c, AVCodecContext *avctx)
c->pix_abs[0][3] = sad16_xy2_altivec;
c->pix_abs[0][0] = sad16_altivec;
c->pix_abs[1][0] = sad8_altivec;
c->sad[0] = sad16_altivec;
c->sad[1] = sad8_altivec;
c->pix_norm1 = pix_norm1_altivec;
c->sse[1]= sse8_altivec;
c->sse[0] = sse16_altivec;
c->sse[1] = sse8_altivec;
c->pix_norm1 = pix_norm1_altivec;
c->pix_sum = pix_sum_altivec;
c->diff_pixels = diff_pixels_altivec;
c->add_bytes = add_bytes_altivec;
if (!high_bit_depth) {
c->get_pixels = get_pixels_altivec;
c->clear_block = clear_block_altivec;

8
libavcodec/ppc/dsputil_altivec.h
Unescape View File

@ -24,11 +24,13 @@
#define AVCODEC_PPC_DSPUTIL_ALTIVEC_H
#include <stdint.h>
#include "libavcodec/dsputil.h"
void ff_put_pixels16_altivec(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h);
#include "libavcodec/dsputil.h"
void ff_avg_pixels16_altivec(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h);
void ff_avg_pixels16_altivec(uint8_t *block, const uint8_t *pixels,
ptrdiff_t line_size, int h);
void ff_put_pixels16_altivec(uint8_t *block, const uint8_t *pixels,
ptrdiff_t line_size, int h);
void ff_fdct_altivec(int16_t *block);
void ff_gmc1_altivec(uint8_t *dst, uint8_t *src, int stride, int h,

17
libavcodec/ppc/dsputil_ppc.c
Unescape View File

@ -53,6 +53,7 @@ static void clear_blocks_dcbz32_ppc(int16_t *blocks)
{
register int misal = ((unsigned long) blocks & 0x00000010);
register int i = 0;
if (misal) {
((unsigned long *) blocks)[0] = 0L;
((unsigned long *) blocks)[1] = 0L;
@ -60,9 +61,8 @@ static void clear_blocks_dcbz32_ppc(int16_t *blocks)
((unsigned long *) blocks)[3] = 0L;
i += 16;
}
for ( ; i < sizeof(int16_t)*6*64-31 ; i += 32) {
for (; i < sizeof(int16_t) * 6 * 64 - 31; i += 32)
__asm__ volatile ("dcbz %0,%1" :: "b" (blocks), "r" (i) : "memory");
}
if (misal) {
((unsigned long *) blocks)[188] = 0L;
((unsigned long *) blocks)[189] = 0L;
@ -79,14 +79,14 @@ static void clear_blocks_dcbz128_ppc(int16_t *blocks)
{
register int misal = ((unsigned long) blocks & 0x0000007f);
register int i = 0;
if (misal) {
/* We could probably also optimize this case,
* but there's not much point as the machines
* aren't available yet (2003-06-26). */
memset(blocks, 0, sizeof(int16_t) * 6 * 64);
}
else
for ( ; i < sizeof(int16_t)*6*64 ; i += 128) {
} else {
for (; i < sizeof(int16_t) * 6 * 64; i += 128)
__asm__ volatile ("dcbzl %0,%1" :: "b" (blocks), "r" (i) : "memory");
}
}
@ -110,9 +110,8 @@ static long check_dcbzl_effect(void)
register long i = 0;
long count = 0;
if (!fakedata) {
if (!fakedata)
return 0L;
}
fakedata_middle = (fakedata + 512);
@ -122,10 +121,9 @@ static long check_dcbzl_effect(void)
* in gcc-3.3 / RS/6000 speaks. Seems to avoid using r0, so.... */
__asm__ volatile ("dcbzl %0, %1" :: "b" (fakedata_middle), "r" (zero));
for (i = 0; i < 1024 ; i ++) {
for (i = 0; i < 1024; i++)
if (fakedata[i] == (char) 0)
count++;
}
av_free(fakedata);
@ -177,6 +175,5 @@ av_cold void ff_dsputil_init_ppc(DSPContext *c, AVCodecContext *avctx)
c->idct_permutation_type = FF_TRANSPOSE_IDCT_PERM;
}
}
}
}

Write Preview
Loading…
Cancel
Save