Skip to content

F
ffmpeg.wasm-core
Commit 30f3f959 authored by Diego Biurrun's avatar Diego Biurrun
Browse files
Options

ppc: dsputil: K&R formatting cosmetics

parent 82ee14d2
Show whitespace changes
Inline Side-by-side
Showing with 429 additions and 391 deletions
libavcodec/ppc/dsputil_altivec.c
View file @ 30f3f959
......@@ -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;
......@@ -44,13 +47,13 @@ static int sad16_x2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size
vector signed int sumdiffs;
s = 0;
sad = (vector unsigned int)vec_splat_u32(0);
sad = (vector unsigned int) vec_splat_u32(0);
for (i = 0; i < h; i++) {
/* Read unaligned pixels into our vectors. The vectors are as follows:
* pix1v: pix1[0] - pix1[15]
* pix2v: pix2[0] - pix2[15] pix2iv: pix2[1] - pix2[16] */
pix1v = vec_ld( 0, pix1);
pix2l = vec_ld( 0, pix2);
pix1v = vec_ld(0, pix1);
pix2l = vec_ld(0, pix2);
pix2r = vec_ld(16, pix2);
pix2v = vec_perm(pix2l, pix2r, perm1);
pix2iv = vec_perm(pix2l, pix2r, perm2);
......@@ -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;
......@@ -88,7 +93,7 @@ static int sad16_y2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size
uint8_t *pix3 = pix2 + line_size;
s = 0;
sad = (vector unsigned int)vec_splat_u32(0);
sad = (vector unsigned int) vec_splat_u32(0);
/* Due to the fact that pix3 = pix2 + line_size, the pix3 of one
* iteration becomes pix2 in the next iteration. We can use this
......@@ -97,7 +102,7 @@ static int sad16_y2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size
* Read unaligned pixels into our vectors. The vectors are as follows:
* pix2v: pix2[0] - pix2[15]
* Split the pixel vectors into shorts. */
pix2l = vec_ld( 0, pix2);
pix2l = vec_ld(0, pix2);
pix2r = vec_ld(15, pix2);
pix2v = vec_perm(pix2l, pix2r, perm);
......@@ -107,7 +112,7 @@ static int sad16_y2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size
* pix3v: pix3[0] - pix3[15] */
pix1v = vec_ld(0, pix1);
pix2l = vec_ld( 0, pix3);
pix2l = vec_ld(0, pix3);
pix2r = vec_ld(15, pix3);
pix3v = vec_perm(pix2l, pix2r, perm);
......@@ -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));
......@@ -152,7 +159,7 @@ static int sad16_xy2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_siz
vector unsigned int sad;
vector signed int sumdiffs;
sad = (vector unsigned int)vec_splat_u32(0);
sad = (vector unsigned int) vec_splat_u32(0);
s = 0;
......@@ -163,7 +170,7 @@ static int sad16_xy2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_siz
* Read unaligned pixels into our vectors. The vectors are as follows:
* pix2v: pix2[0] - pix2[15] pix2iv: pix2[1] - pix2[16]
* Split the pixel vectors into shorts. */
pix2l = vec_ld( 0, pix2);
pix2l = vec_ld(0, pix2);
pix2r = vec_ld(16, pix2);
pix2v = vec_perm(pix2l, pix2r, perm1);
pix2iv = vec_perm(pix2l, pix2r, perm2);
......@@ -181,7 +188,7 @@ static int sad16_xy2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_siz
* pix3v: pix3[0] - pix3[15] pix3iv: pix3[1] - pix3[16] */
pix1v = vec_ld(0, pix1);
pix2l = vec_ld( 0, pix3);
pix2l = vec_ld(0, pix3);
pix2r = vec_ld(16, pix3);
pix3v = vec_perm(pix2l, pix2r, perm1);
pix3iv = vec_perm(pix2l, pix2r, perm2);
......@@ -228,22 +235,23 @@ 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 char t1, t2, t3, t4, t5;
vector unsigned int sad;
vector signed int sumdiffs;
sad = (vector unsigned int)vec_splat_u32(0);
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);
vector unsigned char pix2l = vec_ld(0, pix2);
vector unsigned char pix2r = vec_ld(15, pix2);
t1 = vec_ld(0, pix1);
t2 = vec_perm(pix2l, pix2r, perm);
......@@ -268,19 +276,23 @@ 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;
vector unsigned char t1, t2, t3, t4, t5;
vector unsigned int sad;
vector signed int sumdiffs;
sad = (vector unsigned int)vec_splat_u32(0);
sad = (vector unsigned int) vec_splat_u32(0);
for (i = 0; i < h; i++) {
/* Read potentially unaligned pixels into t1 and t2.
......@@ -317,18 +329,19 @@ 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;
vector signed int sum;
sv = (vector unsigned int)vec_splat_u32(0);
sv = (vector unsigned int) vec_splat_u32(0);
s = 0;
for (i = 0; i < 16; i++) {
/* Read the potentially unaligned pixels. */
vector unsigned char pixl = vec_ld( 0, pix);
vector unsigned char pixl = vec_ld(0, pix);
vector unsigned char pixr = vec_ld(15, pix);
pixv = vec_perm(pixl, pixr, perm);
......@@ -347,19 +360,23 @@ 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;
vector unsigned char t1, t2, t3, t4, t5;
vector unsigned int sum;
vector signed int sumsqr;
sum = (vector unsigned int)vec_splat_u32(0);
sum = (vector unsigned int) vec_splat_u32(0);
for (i = 0; i < h; i++) {
/* Read potentially unaligned pixels into t1 and t2.
......@@ -397,21 +414,23 @@ 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 char t1, t2, t3, t4, t5;
vector unsigned int sum;
vector signed int sumsqr;
sum = (vector unsigned int)vec_splat_u32(0);
sum = (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);
vector unsigned char pix2l = vec_ld(0, pix2);
vector unsigned char pix2r = vec_ld(15, pix2);
t1 = vec_ld(0, pix1);
t2 = vec_perm(pix2l, pix2r, perm);
......@@ -439,9 +458,10 @@ static int sse16_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, i
return s;
}
static int pix_sum_altivec(uint8_t * pix, int line_size)
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;
......@@ -450,11 +470,11 @@ static int pix_sum_altivec(uint8_t * pix, int line_size)
int i;
int s;
sad = (vector unsigned int)vec_splat_u32(0);
sad = (vector unsigned int) vec_splat_u32(0);
for (i = 0; i < 16; i++) {
/* Read the potentially unaligned 16 pixels into t1. */
vector unsigned char pixl = vec_ld( 0, pix);
vector unsigned char pixl = vec_ld(0, pix);
vector unsigned char pixr = vec_ld(15, pix);
t1 = vec_perm(pixl, pixr, perm);
......@@ -472,30 +492,29 @@ 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);
// Convert the bytes into shorts.
shorts = (vector signed short)vec_mergeh(zero, bytes);
shorts = (vector signed short) vec_mergeh(zero, bytes);
// Save the data to the block, we assume the block is 16-byte aligned.
vec_st(shorts, i*16, (vector signed short*)block);
vec_st(shorts, i * 16, (vector signed short *) block);
pixels += line_size;
}
......@@ -508,33 +527,34 @@ 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++) {
/* Read potentially unaligned pixels.
* We're reading 16 pixels, and actually only want 8,
* but we simply ignore the extras. */
pixl = vec_ld( 0, s1);
pixl = vec_ld(0, s1);
pixr = vec_ld(15, s1);
bytes = vec_perm(pixl, pixr, perm1);
// Convert the bytes into shorts.
shorts1 = (vector signed short)vec_mergeh(zero, bytes);
shorts1 = (vector signed short) vec_mergeh(zero, bytes);
// Do the same for the second block of pixels.
pixl = vec_ld( 0, s2);
pixl = vec_ld(0, s2);
pixr = vec_ld(15, s2);
bytes = vec_perm(pixl, pixr, perm2);
// Convert the bytes into shorts.
shorts2 = (vector signed short)vec_mergeh(zero, bytes);
shorts2 = (vector signed short) vec_mergeh(zero, bytes);
// Do the subtraction.
shorts1 = vec_sub(shorts1, shorts2);
// Save the data to the block, we assume the block is 16-byte aligned.
vec_st(shorts1, 0, (vector signed short*)block);
vec_st(shorts1, 0, (vector signed short *) block);
s1 += stride;
s2 += stride;
......@@ -546,26 +566,26 @@ static void diff_pixels_altivec(int16_t *restrict block, const uint8_t *s1,
/* Read potentially unaligned pixels.
* We're reading 16 pixels, and actually only want 8,
* but we simply ignore the extras. */
pixl = vec_ld( 0, s1);
pixl = vec_ld(0, s1);
pixr = vec_ld(15, s1);
bytes = vec_perm(pixl, pixr, perm1);
// Convert the bytes into shorts.
shorts1 = (vector signed short)vec_mergeh(zero, bytes);
shorts1 = (vector signed short) vec_mergeh(zero, bytes);
// Do the same for the second block of pixels.
pixl = vec_ld( 0, s2);
pixl = vec_ld(0, s2);
pixr = vec_ld(15, s2);
bytes = vec_perm(pixl, pixr, perm2);
// Convert the bytes into shorts.
shorts2 = (vector signed short)vec_mergeh(zero, bytes);
shorts2 = (vector signed short) vec_mergeh(zero, bytes);
// Do the subtraction.
shorts1 = vec_sub(shorts1, shorts2);
// Save the data to the block, we assume the block is 16-byte aligned.
vec_st(shorts1, 0, (vector signed short*)block);
vec_st(shorts1, 0, (vector signed short *) block);
s1 += stride;
s2 += stride;
......@@ -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);
......@@ -586,46 +606,50 @@ 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;
/* dst and src are 16 bytes-aligned (guaranteed). */
for (i = 0 ; (i + 15) < w ; i+=16) {
vdst = vec_ld(i, (unsigned char*)dst);
vsrc = vec_ld(i, (unsigned char*)src);
for (i = 0; (i + 15) < w; i += 16) {
vdst = vec_ld(i, (unsigned char *) dst);
vsrc = vec_ld(i, (unsigned char *) src);
vdst = vec_add(vsrc, vdst);
vec_st(vdst, i, (unsigned char*)dst);
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);
(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)
{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)
{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)
{0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F,
0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07};
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)
{ 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)
{ 0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F,
0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07 };
#define ONEITERBUTTERFLY(i, res) \
{ \
......@@ -641,10 +665,10 @@ static int hadamard8_diff8x8_altivec(/*MpegEncContext*/ void *s, uint8_t *dst, u
dstO = vec_perm(dst1, dst2, vec_lvsl(stride * i, dst)); \
/* Promote the unsigned chars to signed shorts. */ \
/* We're in the 8x8 function, we only care for the first 8. */ \
srcV = (vector signed short)vec_mergeh((vector signed char)vzero, \
(vector signed char)srcO); \
dstV = (vector signed short)vec_mergeh((vector signed char)vzero, \
(vector signed char)dstO); \
srcV = (vector signed short) vec_mergeh((vector signed char) vzero, \
(vector signed char) srcO); \
dstV = (vector signed short) vec_mergeh((vector signed char) vzero, \
(vector signed char) dstO); \
/* subtractions inside the first butterfly */ \
but0 = vec_sub(srcV, dstV); \
op1 = vec_perm(but0, but0, perm1); \
......@@ -701,7 +725,7 @@ static int hadamard8_diff8x8_altivec(/*MpegEncContext*/ void *s, uint8_t *dst, u
vsum = vec_sum4s(vec_abs(line5C), vsum);
vsum = vec_sum4s(vec_abs(line6C), vsum);
vsum = vec_sum4s(vec_abs(line7C), vsum);
vsum = vec_sums(vsum, (vector signed int)vzero);
vsum = vec_sums(vsum, (vector signed int) vzero);
vsum = vec_splat(vsum, 3);
vec_ste(vsum, 0, &sum);
}
......@@ -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"),
......@@ -747,37 +773,44 @@ static int hadamard8_diff16x8_altivec(/*MpegEncContext*/ void *s, uint8_t *dst,
temp6S __asm__ ("v14"),
temp7S __asm__ ("v15");
register const vector unsigned char vzero __asm__ ("v31") =
(const vector unsigned char)vec_splat_u8(0);
(const vector unsigned char) vec_splat_u8(0);
{
register const vector signed short vprod1 __asm__ ("v16") =
(const vector signed short){ 1,-1, 1,-1, 1,-1, 1,-1 };
(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 };
(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 };
(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};
{ 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};
{ 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,
0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07};
{ 0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F,
0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07 };
#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"), \
......@@ -801,14 +834,14 @@ static int hadamard8_diff16x8_altivec(/*MpegEncContext*/ void *s, uint8_t *dst,
dst2 = vec_ld((stride * i) + 16, dst); \
dstO = vec_perm(dst1, dst2, vec_lvsl(stride * i, dst)); \
/* Promote the unsigned chars to signed shorts. */ \
srcV = (vector signed short)vec_mergeh((vector signed char)vzero, \
(vector signed char)srcO); \
dstV = (vector signed short)vec_mergeh((vector signed char)vzero, \
(vector signed char)dstO); \
srcW = (vector signed short)vec_mergel((vector signed char)vzero, \
(vector signed char)srcO); \
dstW = (vector signed short)vec_mergel((vector signed char)vzero, \
(vector signed char)dstO); \
srcV = (vector signed short) vec_mergeh((vector signed char) vzero, \
(vector signed char) srcO); \
dstV = (vector signed short) vec_mergeh((vector signed char) vzero, \
(vector signed char) dstO); \
srcW = (vector signed short) vec_mergel((vector signed char) vzero, \
(vector signed char) srcO); \
dstW = (vector signed short) vec_mergel((vector signed char) vzero, \
(vector signed char) dstO); \
/* subtractions inside the first butterfly */ \
but0 = vec_sub(srcV, dstV); \
but0S = vec_sub(srcW, dstW); \
......@@ -838,10 +871,10 @@ static int hadamard8_diff16x8_altivec(/*MpegEncContext*/ void *s, uint8_t *dst,
{
register vector signed int vsum;
register vector signed short line0S, line1S, line2S, line3S, line4S,
line5S, line6S, line7S, line0BS,line2BS,
line1BS,line3BS,line4BS,line6BS,line5BS,
line7BS,line0CS,line4CS,line1CS,line5CS,
line2CS,line6CS,line3CS,line7CS;
line5S, line6S, line7S, line0BS, line2BS,
line1BS, line3BS, line4BS, line6BS, line5BS,
line7BS, line0CS, line4CS, line1CS, line5CS,
line2CS, line6CS, line3CS, line7CS;
register vector signed short line0 = vec_add(temp0, temp1);
register vector signed short line1 = vec_sub(temp0, temp1);
......@@ -914,19 +947,21 @@ static int hadamard8_diff16x8_altivec(/*MpegEncContext*/ void *s, uint8_t *dst,
vsum = vec_sum4s(vec_abs(line5CS), vsum);
vsum = vec_sum4s(vec_abs(line6CS), vsum);
vsum = vec_sum4s(vec_abs(line7CS), vsum);
vsum = vec_sums(vsum, (vector signed int)vzero);
vsum = vec_sums(vsum, (vector signed int) vzero);
vsum = vec_splat(vsum, 3);
vec_ste(vsum, 0, &sum);
}
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) {
dst += 8*stride;
src += 8*stride;
if (h == 16) {
dst += 8 * stride;
src += 8 * stride;
score += hadamard8_diff16x8_altivec(s, dst, src, stride, 8);
}
return score;
......@@ -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->sad[0] = sad16_altivec;
c->sad[1] = sad8_altivec;
c->sse[0] = sse16_altivec;
c->sse[1] = sse8_altivec;
c->pix_norm1 = pix_norm1_altivec;
c->sse[1]= sse8_altivec;
c->sse[0]= sse16_altivec;
c->pix_sum = pix_sum_altivec;
c->diff_pixels = diff_pixels_altivec;
c->add_bytes= add_bytes_altivec;
c->add_bytes = add_bytes_altivec;
if (!high_bit_depth) {
c->get_pixels = get_pixels_altivec;
c->clear_block = clear_block_altivec;
......
libavcodec/ppc/dsputil_altivec.h
View file @ 30f3f959
......@@ -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,
......@@ -36,7 +38,7 @@ void ff_gmc1_altivec(uint8_t *dst, uint8_t *src, int stride, int h,
void ff_idct_put_altivec(uint8_t *dest, int line_size, int16_t *block);
void ff_idct_add_altivec(uint8_t *dest, int line_size, int16_t *block);
void ff_dsputil_init_altivec(DSPContext* c, AVCodecContext *avctx);
void ff_int_init_altivec(DSPContext* c, AVCodecContext *avctx);
void ff_dsputil_init_altivec(DSPContext *c, AVCodecContext *avctx);
void ff_int_init_altivec(DSPContext *c, AVCodecContext *avctx);
#endif /* AVCODEC_PPC_DSPUTIL_ALTIVEC_H */
libavcodec/ppc/dsputil_ppc.c
View file @ 30f3f959
......@@ -51,23 +51,23 @@
*/
static void clear_blocks_dcbz32_ppc(int16_t *blocks)
{
register int misal = ((unsigned long)blocks & 0x00000010);
register int misal = ((unsigned long) blocks & 0x00000010);
register int i = 0;
if (misal) {
((unsigned long*)blocks)[0] = 0L;
((unsigned long*)blocks)[1] = 0L;
((unsigned long*)blocks)[2] = 0L;
((unsigned long*)blocks)[3] = 0L;
((unsigned long *) blocks)[0] = 0L;
((unsigned long *) blocks)[1] = 0L;
((unsigned long *) blocks)[2] = 0L;
((unsigned long *) blocks)[3] = 0L;
i += 16;
}
for ( ; i < sizeof(int16_t)*6*64-31 ; i += 32) {
__asm__ volatile("dcbz %0,%1" : : "b" (blocks), "r" (i) : "memory");
}
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;
((unsigned long*)blocks)[190] = 0L;
((unsigned long*)blocks)[191] = 0L;
((unsigned long *) blocks)[188] = 0L;
((unsigned long *) blocks)[189] = 0L;
((unsigned long *) blocks)[190] = 0L;
((unsigned long *) blocks)[191] = 0L;
i += 16;
}
}
......@@ -77,23 +77,23 @@ static void clear_blocks_dcbz32_ppc(int16_t *blocks)
#if HAVE_DCBZL
static void clear_blocks_dcbz128_ppc(int16_t *blocks)
{
register int misal = ((unsigned long)blocks & 0x0000007f);
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) {
__asm__ volatile("dcbzl %0,%1" : : "b" (blocks), "r" (i) : "memory");
memset(blocks, 0, sizeof(int16_t) * 6 * 64);
} else {
for (; i < sizeof(int16_t) * 6 * 64; i += 128)
__asm__ volatile ("dcbzl %0,%1" :: "b" (blocks), "r" (i) : "memory");
}
}
#else
static void clear_blocks_dcbz128_ppc(int16_t *blocks)
{
memset(blocks, 0, sizeof(int16_t)*6*64);
memset(blocks, 0, sizeof(int16_t) * 6 * 64);
}
#endif
......@@ -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);
......@@ -120,12 +119,11 @@ static long check_dcbzl_effect(void)
/* Below the constraint "b" seems to mean "address base register"
* in gcc-3.3 / RS/6000 speaks. Seems to avoid using r0, so.... */
__asm__ volatile("dcbzl %0, %1" : : "b" (fakedata_middle), "r" (zero));
__asm__ volatile ("dcbzl %0, %1" :: "b" (fakedata_middle), "r" (zero));
for (i = 0; i < 1024 ; i ++) {
if (fakedata[i] == (char)0)
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;
}
}
}
}
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