summaryrefslogtreecommitdiff
path: root/libavcodec/ppc
diff options
context:
space:
mode:
Diffstat (limited to 'libavcodec/ppc')
-rw-r--r--libavcodec/ppc/dsputil_altivec.c128
-rw-r--r--libavcodec/ppc/dsputil_h264_altivec.c48
-rw-r--r--libavcodec/ppc/dsputil_ppc.c56
-rw-r--r--libavcodec/ppc/dsputil_ppc.h12
-rw-r--r--libavcodec/ppc/fft_altivec.c8
-rw-r--r--libavcodec/ppc/gcc_fixes.h18
-rw-r--r--libavcodec/ppc/idct_altivec.c246
-rw-r--r--libavcodec/ppc/mpegvideo_altivec.c12
8 files changed, 264 insertions, 264 deletions
diff --git a/libavcodec/ppc/dsputil_altivec.c b/libavcodec/ppc/dsputil_altivec.c
index 20ee382f29..da9b3dd15a 100644
--- a/libavcodec/ppc/dsputil_altivec.c
+++ b/libavcodec/ppc/dsputil_altivec.c
@@ -67,7 +67,7 @@ int sad16_x2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h
/*
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]
+ pix2v: pix2[0]-pix2[15] pix2iv: pix2[1]-pix2[16]
*/
tv = (vector unsigned char *) pix1;
pix1v = vec_perm(tv[0], tv[1], vec_lvsl(0, pix1));
@@ -184,7 +184,7 @@ int sad16_xy2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int
fact to avoid a potentially expensive unaligned read, as well
as some splitting, and vector addition each time around the loop.
Read unaligned pixels into our vectors. The vectors are as follows:
- pix2v: pix2[0]-pix2[15] pix2iv: pix2[1]-pix2[16]
+ pix2v: pix2[0]-pix2[15] pix2iv: pix2[1]-pix2[16]
Split the pixel vectors into shorts
*/
tv = (vector unsigned char *) &pix2[0];
@@ -204,7 +204,7 @@ int sad16_xy2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int
/*
Read unaligned pixels into our vectors. The vectors are as follows:
pix1v: pix1[0]-pix1[15]
- pix3v: pix3[0]-pix3[15] pix3iv: pix3[1]-pix3[16]
+ pix3v: pix3[0]-pix3[15] pix3iv: pix3[1]-pix3[16]
*/
tv = (vector unsigned char *) pix1;
pix1v = vec_perm(tv[0], tv[1], vec_lvsl(0, pix1));
@@ -273,7 +273,7 @@ int sad16_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
for(i=0;i<h;i++) {
- /* Read potentially unaligned pixels into t1 and t2 */
+ /* Read potentially unaligned pixels into t1 and t2 */
perm1 = vec_lvsl(0, pix1);
pix1v = (vector unsigned char *) pix1;
perm2 = vec_lvsl(0, pix2);
@@ -281,12 +281,12 @@ int sad16_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
t1 = vec_perm(pix1v[0], pix1v[1], perm1);
t2 = vec_perm(pix2v[0], pix2v[1], perm2);
- /* Calculate a sum of abs differences vector */
+ /* Calculate a sum of abs differences vector */
t3 = vec_max(t1, t2);
t4 = vec_min(t1, t2);
t5 = vec_sub(t3, t4);
- /* Add each 4 pixel group together and put 4 results into sad */
+ /* Add each 4 pixel group together and put 4 results into sad */
sad = vec_sum4s(t5, sad);
pix1 += line_size;
@@ -316,9 +316,9 @@ int sad8_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
permclear = (vector unsigned char)AVV(255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0);
for(i=0;i<h;i++) {
- /* Read potentially unaligned pixels into t1 and t2
- Since we're reading 16 pixels, and actually only want 8,
- mask out the last 8 pixels. The 0s don't change the sum. */
+ /* Read potentially unaligned pixels into t1 and t2
+ Since we're reading 16 pixels, and actually only want 8,
+ mask out the last 8 pixels. The 0s don't change the sum. */
perm1 = vec_lvsl(0, pix1);
pix1v = (vector unsigned char *) pix1;
perm2 = vec_lvsl(0, pix2);
@@ -326,12 +326,12 @@ int sad8_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
t1 = vec_and(vec_perm(pix1v[0], pix1v[1], perm1), permclear);
t2 = vec_and(vec_perm(pix2v[0], pix2v[1], perm2), permclear);
- /* Calculate a sum of abs differences vector */
+ /* Calculate a sum of abs differences vector */
t3 = vec_max(t1, t2);
t4 = vec_min(t1, t2);
t5 = vec_sub(t3, t4);
- /* Add each 4 pixel group together and put 4 results into sad */
+ /* Add each 4 pixel group together and put 4 results into sad */
sad = vec_sum4s(t5, sad);
pix1 += line_size;
@@ -398,9 +398,9 @@ int sse8_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
for(i=0;i<h;i++) {
- /* Read potentially unaligned pixels into t1 and t2
- Since we're reading 16 pixels, and actually only want 8,
- mask out the last 8 pixels. The 0s don't change the sum. */
+ /* Read potentially unaligned pixels into t1 and t2
+ Since we're reading 16 pixels, and actually only want 8,
+ mask out the last 8 pixels. The 0s don't change the sum. */
perm1 = vec_lvsl(0, pix1);
pix1v = (vector unsigned char *) pix1;
perm2 = vec_lvsl(0, pix2);
@@ -413,7 +413,7 @@ int sse8_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
of the fact that abs(a-b)^2 = (a-b)^2.
*/
- /* Calculate abs differences vector */
+ /* Calculate abs differences vector */
t3 = vec_max(t1, t2);
t4 = vec_min(t1, t2);
t5 = vec_sub(t3, t4);
@@ -451,7 +451,7 @@ int sse16_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
sum = (vector unsigned int)vec_splat_u32(0);
for(i=0;i<h;i++) {
- /* Read potentially unaligned pixels into t1 and t2 */
+ /* Read potentially unaligned pixels into t1 and t2 */
perm1 = vec_lvsl(0, pix1);
pix1v = (vector unsigned char *) pix1;
perm2 = vec_lvsl(0, pix2);
@@ -464,7 +464,7 @@ int sse16_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
of the fact that abs(a-b)^2 = (a-b)^2.
*/
- /* Calculate abs differences vector */
+ /* Calculate abs differences vector */
t3 = vec_max(t1, t2);
t4 = vec_min(t1, t2);
t5 = vec_sub(t3, t4);
@@ -498,12 +498,12 @@ int pix_sum_altivec(uint8_t * pix, int line_size)
sad = (vector unsigned int)vec_splat_u32(0);
for (i = 0; i < 16; i++) {
- /* Read the potentially unaligned 16 pixels into t1 */
+ /* Read the potentially unaligned 16 pixels into t1 */
perm = vec_lvsl(0, pix);
pixv = (vector unsigned char *) pix;
t1 = vec_perm(pixv[0], pixv[1], perm);
- /* Add each 4 pixel group together and put 4 results into sad */
+ /* Add each 4 pixel group together and put 4 results into sad */
sad = vec_sum4s(t1, sad);
pix += line_size;
@@ -1335,32 +1335,32 @@ POWERPC_PERF_START_COUNT(altivec_hadamard8_diff8x8_num, 1);
0x00, 0x01, 0x02, 0x03,
0x04, 0x05, 0x06, 0x07);
-#define ONEITERBUTTERFLY(i, res) \
- { \
- register vector unsigned char src1, src2, srcO; \
- register vector unsigned char dst1, dst2, dstO; \
- src1 = vec_ld(stride * i, src); \
- if ((((stride * i) + (unsigned long)src) & 0x0000000F) > 8) \
- src2 = vec_ld((stride * i) + 16, src); \
- srcO = vec_perm(src1, src2, vec_lvsl(stride * i, src)); \
- dst1 = vec_ld(stride * i, dst); \
- if ((((stride * i) + (unsigned long)dst) & 0x0000000F) > 8) \
- dst2 = vec_ld((stride * i) + 16, dst); \
- 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 */ \
- register vector signed short srcV = \
- (vector signed short)vec_mergeh((vector signed char)vzero, (vector signed char)srcO); \
- register vector signed short dstV = \
- (vector signed short)vec_mergeh((vector signed char)vzero, (vector signed char)dstO); \
- /* substractions inside the first butterfly */ \
- register vector signed short but0 = vec_sub(srcV, dstV); \
- register vector signed short op1 = vec_perm(but0, but0, perm1); \
- register vector signed short but1 = vec_mladd(but0, vprod1, op1); \
- register vector signed short op2 = vec_perm(but1, but1, perm2); \
- register vector signed short but2 = vec_mladd(but1, vprod2, op2); \
- register vector signed short op3 = vec_perm(but2, but2, perm3); \
- res = vec_mladd(but2, vprod3, op3); \
+#define ONEITERBUTTERFLY(i, res) \
+ { \
+ register vector unsigned char src1, src2, srcO; \
+ register vector unsigned char dst1, dst2, dstO; \
+ src1 = vec_ld(stride * i, src); \
+ if ((((stride * i) + (unsigned long)src) & 0x0000000F) > 8) \
+ src2 = vec_ld((stride * i) + 16, src); \
+ srcO = vec_perm(src1, src2, vec_lvsl(stride * i, src)); \
+ dst1 = vec_ld(stride * i, dst); \
+ if ((((stride * i) + (unsigned long)dst) & 0x0000000F) > 8) \
+ dst2 = vec_ld((stride * i) + 16, dst); \
+ 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 */ \
+ register vector signed short srcV = \
+ (vector signed short)vec_mergeh((vector signed char)vzero, (vector signed char)srcO); \
+ register vector signed short dstV = \
+ (vector signed short)vec_mergeh((vector signed char)vzero, (vector signed char)dstO); \
+ /* substractions inside the first butterfly */ \
+ register vector signed short but0 = vec_sub(srcV, dstV); \
+ register vector signed short op1 = vec_perm(but0, but0, perm1); \
+ register vector signed short but1 = vec_mladd(but0, vprod1, op1); \
+ register vector signed short op2 = vec_perm(but1, but1, perm2); \
+ register vector signed short but2 = vec_mladd(but1, vprod2, op2); \
+ register vector signed short op3 = vec_perm(but2, but2, perm3); \
+ res = vec_mladd(but2, vprod3, op3); \
}
ONEITERBUTTERFLY(0, temp0);
ONEITERBUTTERFLY(1, temp1);
@@ -1480,26 +1480,26 @@ static int hadamard8_diff16x8_altivec(/*MpegEncContext*/ void *s, uint8_t *dst,
0x00, 0x01, 0x02, 0x03,
0x04, 0x05, 0x06, 0x07);
-#define ONEITERBUTTERFLY(i, res1, res2) \
- { \
+#define ONEITERBUTTERFLY(i, res1, res2) \
+ { \
register vector unsigned char src1 asm ("v22"), src2 asm ("v23"); \
register vector unsigned char dst1 asm ("v24"), dst2 asm ("v25"); \
- src1 = vec_ld(stride * i, src); \
- src2 = vec_ld((stride * i) + 16, src); \
+ src1 = vec_ld(stride * i, src); \
+ src2 = vec_ld((stride * i) + 16, src); \
register vector unsigned char srcO asm ("v22") = vec_perm(src1, src2, vec_lvsl(stride * i, src)); \
- dst1 = vec_ld(stride * i, dst); \
- dst2 = vec_ld((stride * i) + 16, dst); \
+ dst1 = vec_ld(stride * i, dst); \
+ dst2 = vec_ld((stride * i) + 16, dst); \
register vector unsigned char dstO asm ("v23") = vec_perm(dst1, dst2, vec_lvsl(stride * i, dst)); \
- /* promote the unsigned chars to signed shorts */ \
+ /* promote the unsigned chars to signed shorts */ \
register vector signed short srcV asm ("v24") = \
- (vector signed short)vec_mergeh((vector signed char)vzero, (vector signed char)srcO); \
+ (vector signed short)vec_mergeh((vector signed char)vzero, (vector signed char)srcO); \
register vector signed short dstV asm ("v25") = \
- (vector signed short)vec_mergeh((vector signed char)vzero, (vector signed char)dstO); \
+ (vector signed short)vec_mergeh((vector signed char)vzero, (vector signed char)dstO); \
register vector signed short srcW asm ("v26") = \
- (vector signed short)vec_mergel((vector signed char)vzero, (vector signed char)srcO); \
+ (vector signed short)vec_mergel((vector signed char)vzero, (vector signed char)srcO); \
register vector signed short dstW asm ("v27") = \
- (vector signed short)vec_mergel((vector signed char)vzero, (vector signed char)dstO); \
- /* substractions inside the first butterfly */ \
+ (vector signed short)vec_mergel((vector signed char)vzero, (vector signed char)dstO); \
+ /* substractions inside the first butterfly */ \
register vector signed short but0 asm ("v28") = vec_sub(srcV, dstV); \
register vector signed short but0S asm ("v29") = vec_sub(srcW, dstW); \
register vector signed short op1 asm ("v30") = vec_perm(but0, but0, perm1); \
@@ -1511,9 +1511,9 @@ static int hadamard8_diff16x8_altivec(/*MpegEncContext*/ void *s, uint8_t *dst,
register vector signed short op2S asm ("v27") = vec_perm(but1S, but1S, perm2); \
register vector signed short but2S asm ("v28") = vec_mladd(but1S, vprod2, op2S); \
register vector signed short op3 asm ("v29") = vec_perm(but2, but2, perm3); \
- res1 = vec_mladd(but2, vprod3, op3); \
+ res1 = vec_mladd(but2, vprod3, op3); \
register vector signed short op3S asm ("v30") = vec_perm(but2S, but2S, perm3); \
- res2 = vec_mladd(but2S, vprod3, op3S); \
+ res2 = vec_mladd(but2S, vprod3, op3S); \
}
ONEITERBUTTERFLY(0, temp0, temp0S);
ONEITERBUTTERFLY(1, temp1, temp1S);
@@ -1623,12 +1623,12 @@ POWERPC_PERF_STOP_COUNT(altivec_hadamard8_diff16_num, 1);
int has_altivec(void)
{
#ifdef __AMIGAOS4__
- ULONG result = 0;
- extern struct ExecIFace *IExec;
+ ULONG result = 0;
+ extern struct ExecIFace *IExec;
- IExec->GetCPUInfoTags(GCIT_VectorUnit, &result, TAG_DONE);
- if (result == VECTORTYPE_ALTIVEC) return 1;
- return 0;
+ IExec->GetCPUInfoTags(GCIT_VectorUnit, &result, TAG_DONE);
+ if (result == VECTORTYPE_ALTIVEC) return 1;
+ return 0;
#else /* __AMIGAOS4__ */
#ifdef CONFIG_DARWIN
diff --git a/libavcodec/ppc/dsputil_h264_altivec.c b/libavcodec/ppc/dsputil_h264_altivec.c
index 7dd9dcf0a1..0efb169d8b 100644
--- a/libavcodec/ppc/dsputil_h264_altivec.c
+++ b/libavcodec/ppc/dsputil_h264_altivec.c
@@ -191,33 +191,33 @@ static void OPNAME ## h264_qpel ## SIZE ## _mc32_ ## CODETYPE(uint8_t *dst, uint
/* from dsputil.c */
static inline void put_pixels8_l2(uint8_t * dst, const uint8_t * src1, const uint8_t * src2, int dst_stride, int src_stride1, int src_stride2, int h) {
- int i;
- for (i = 0; i < h; i++) {
- uint32_t a, b;
- a = (((const struct unaligned_32 *) (&src1[i * src_stride1]))->l);
- b = (((const struct unaligned_32 *) (&src2[i * src_stride2]))->l);
- *((uint32_t *) & dst[i * dst_stride]) = rnd_avg32(a, b);
- a = (((const struct unaligned_32 *) (&src1[i * src_stride1 + 4]))->l);
- b = (((const struct unaligned_32 *) (&src2[i * src_stride2 + 4]))->l);
- *((uint32_t *) & dst[i * dst_stride + 4]) = rnd_avg32(a, b);
- }
+ int i;
+ for (i = 0; i < h; i++) {
+ uint32_t a, b;
+ a = (((const struct unaligned_32 *) (&src1[i * src_stride1]))->l);
+ b = (((const struct unaligned_32 *) (&src2[i * src_stride2]))->l);
+ *((uint32_t *) & dst[i * dst_stride]) = rnd_avg32(a, b);
+ a = (((const struct unaligned_32 *) (&src1[i * src_stride1 + 4]))->l);
+ b = (((const struct unaligned_32 *) (&src2[i * src_stride2 + 4]))->l);
+ *((uint32_t *) & dst[i * dst_stride + 4]) = rnd_avg32(a, b);
+ }
} static inline void avg_pixels8_l2(uint8_t * dst, const uint8_t * src1, const uint8_t * src2, int dst_stride, int src_stride1, int src_stride2, int h) {
- int i;
- for (i = 0; i < h; i++) {
- uint32_t a, b;
- a = (((const struct unaligned_32 *) (&src1[i * src_stride1]))->l);
- b = (((const struct unaligned_32 *) (&src2[i * src_stride2]))->l);
- *((uint32_t *) & dst[i * dst_stride]) = rnd_avg32(*((uint32_t *) & dst[i * dst_stride]), rnd_avg32(a, b));
- a = (((const struct unaligned_32 *) (&src1[i * src_stride1 + 4]))->l);
- b = (((const struct unaligned_32 *) (&src2[i * src_stride2 + 4]))->l);
- *((uint32_t *) & dst[i * dst_stride + 4]) = rnd_avg32(*((uint32_t *) & dst[i * dst_stride + 4]), rnd_avg32(a, b));
- }
+ int i;
+ for (i = 0; i < h; i++) {
+ uint32_t a, b;
+ a = (((const struct unaligned_32 *) (&src1[i * src_stride1]))->l);
+ b = (((const struct unaligned_32 *) (&src2[i * src_stride2]))->l);
+ *((uint32_t *) & dst[i * dst_stride]) = rnd_avg32(*((uint32_t *) & dst[i * dst_stride]), rnd_avg32(a, b));
+ a = (((const struct unaligned_32 *) (&src1[i * src_stride1 + 4]))->l);
+ b = (((const struct unaligned_32 *) (&src2[i * src_stride2 + 4]))->l);
+ *((uint32_t *) & dst[i * dst_stride + 4]) = rnd_avg32(*((uint32_t *) & dst[i * dst_stride + 4]), rnd_avg32(a, b));
+ }
} static inline void put_pixels16_l2(uint8_t * dst, const uint8_t * src1, const uint8_t * src2, int dst_stride, int src_stride1, int src_stride2, int h) {
- put_pixels8_l2(dst, src1, src2, dst_stride, src_stride1, src_stride2, h);
- put_pixels8_l2(dst + 8, src1 + 8, src2 + 8, dst_stride, src_stride1, src_stride2, h);
+ put_pixels8_l2(dst, src1, src2, dst_stride, src_stride1, src_stride2, h);
+ put_pixels8_l2(dst + 8, src1 + 8, src2 + 8, dst_stride, src_stride1, src_stride2, h);
} static inline void avg_pixels16_l2(uint8_t * dst, const uint8_t * src1, const uint8_t * src2, int dst_stride, int src_stride1, int src_stride2, int h) {
- avg_pixels8_l2(dst, src1, src2, dst_stride, src_stride1, src_stride2, h);
- avg_pixels8_l2(dst + 8, src1 + 8, src2 + 8, dst_stride, src_stride1, src_stride2, h);
+ avg_pixels8_l2(dst, src1, src2, dst_stride, src_stride1, src_stride2, h);
+ avg_pixels8_l2(dst + 8, src1 + 8, src2 + 8, dst_stride, src_stride1, src_stride2, h);
}
/* UNIMPLEMENTED YET !! */
diff --git a/libavcodec/ppc/dsputil_ppc.c b/libavcodec/ppc/dsputil_ppc.c
index f8c7f94d3a..b7bec99b19 100644
--- a/libavcodec/ppc/dsputil_ppc.c
+++ b/libavcodec/ppc/dsputil_ppc.c
@@ -87,16 +87,16 @@ void powerpc_display_perf_report(void)
{
for (j = 0; j < POWERPC_NUM_PMC_ENABLED ; j++)
{
- if (perfdata[j][i][powerpc_data_num] != (unsigned long long)0)
- av_log(NULL, AV_LOG_INFO,
- " Function \"%s\" (pmc%d):\n\tmin: %llu\n\tmax: %llu\n\tavg: %1.2lf (%llu)\n",
- perfname[i],
- j+1,
- perfdata[j][i][powerpc_data_min],
- perfdata[j][i][powerpc_data_max],
- (double)perfdata[j][i][powerpc_data_sum] /
- (double)perfdata[j][i][powerpc_data_num],
- perfdata[j][i][powerpc_data_num]);
+ if (perfdata[j][i][powerpc_data_num] != (unsigned long long)0)
+ av_log(NULL, AV_LOG_INFO,
+ " Function \"%s\" (pmc%d):\n\tmin: %llu\n\tmax: %llu\n\tavg: %1.2lf (%llu)\n",
+ perfname[i],
+ j+1,
+ perfdata[j][i][powerpc_data_min],
+ perfdata[j][i][powerpc_data_max],
+ (double)perfdata[j][i][powerpc_data_sum] /
+ (double)perfdata[j][i][powerpc_data_num],
+ perfdata[j][i][powerpc_data_num]);
}
}
}
@@ -179,7 +179,7 @@ POWERPC_PERF_START_COUNT(powerpc_clear_blocks_dcbz128, 1);
}
else
for ( ; i < sizeof(DCTELEM)*6*64 ; i += 128) {
- asm volatile("dcbzl %0,%1" : : "b" (blocks), "r" (i) : "memory");
+ asm volatile("dcbzl %0,%1" : : "b" (blocks), "r" (i) : "memory");
}
#else
memset(blocks, 0, sizeof(DCTELEM)*6*64);
@@ -284,25 +284,25 @@ void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx)
c->put_no_rnd_pixels_tab[0][0] = put_pixels16_altivec;
c->avg_pixels_tab[0][0] = avg_pixels16_altivec;
c->avg_pixels_tab[1][0] = avg_pixels8_altivec;
- c->avg_pixels_tab[1][3] = avg_pixels8_xy2_altivec;
+ c->avg_pixels_tab[1][3] = avg_pixels8_xy2_altivec;
c->put_pixels_tab[1][3] = put_pixels8_xy2_altivec;
c->put_no_rnd_pixels_tab[1][3] = put_no_rnd_pixels8_xy2_altivec;
c->put_pixels_tab[0][3] = put_pixels16_xy2_altivec;
c->put_no_rnd_pixels_tab[0][3] = put_no_rnd_pixels16_xy2_altivec;
- c->gmc1 = gmc1_altivec;
+ c->gmc1 = gmc1_altivec;
#ifdef CONFIG_DARWIN // ATM gcc-3.3 and gcc-3.4 fail to compile these in linux...
- c->hadamard8_diff[0] = hadamard8_diff16_altivec;
- c->hadamard8_diff[1] = hadamard8_diff8x8_altivec;
+ c->hadamard8_diff[0] = hadamard8_diff16_altivec;
+ c->hadamard8_diff[1] = hadamard8_diff8x8_altivec;
#endif
#ifdef CONFIG_ENCODERS
- if (avctx->dct_algo == FF_DCT_AUTO ||
- avctx->dct_algo == FF_DCT_ALTIVEC)
- {
- c->fdct = fdct_altivec;
- }
+ if (avctx->dct_algo == FF_DCT_AUTO ||
+ avctx->dct_algo == FF_DCT_ALTIVEC)
+ {
+ c->fdct = fdct_altivec;
+ }
#endif //CONFIG_ENCODERS
if (avctx->lowres==0)
@@ -325,14 +325,14 @@ void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx)
int i, j;
for (i = 0 ; i < powerpc_perf_total ; i++)
{
- for (j = 0; j < POWERPC_NUM_PMC_ENABLED ; j++)
- {
- perfdata[j][i][powerpc_data_min] = 0xFFFFFFFFFFFFFFFFULL;
- perfdata[j][i][powerpc_data_max] = 0x0000000000000000ULL;
- perfdata[j][i][powerpc_data_sum] = 0x0000000000000000ULL;
- perfdata[j][i][powerpc_data_num] = 0x0000000000000000ULL;
- }
- }
+ for (j = 0; j < POWERPC_NUM_PMC_ENABLED ; j++)
+ {
+ perfdata[j][i][powerpc_data_min] = 0xFFFFFFFFFFFFFFFFULL;
+ perfdata[j][i][powerpc_data_max] = 0x0000000000000000ULL;
+ perfdata[j][i][powerpc_data_sum] = 0x0000000000000000ULL;
+ perfdata[j][i][powerpc_data_num] = 0x0000000000000000ULL;
+ }
+ }
}
#endif /* POWERPC_PERFORMANCE_REPORT */
} else
diff --git a/libavcodec/ppc/dsputil_ppc.h b/libavcodec/ppc/dsputil_ppc.h
index 7e01677f19..17b4c7d38f 100644
--- a/libavcodec/ppc/dsputil_ppc.h
+++ b/libavcodec/ppc/dsputil_ppc.h
@@ -114,10 +114,10 @@ extern unsigned long long perfdata[POWERPC_NUM_PMC_ENABLED][powerpc_perf_total][
#define POWERPC_GET_PMC6(a) do {} while (0)
#endif
#endif /* POWERPC_MODE_64BITS */
-#define POWERPC_PERF_DECLARE(a, cond) \
- POWERP_PMC_DATATYPE \
- pmc_start[POWERPC_NUM_PMC_ENABLED], \
- pmc_stop[POWERPC_NUM_PMC_ENABLED], \
+#define POWERPC_PERF_DECLARE(a, cond) \
+ POWERP_PMC_DATATYPE \
+ pmc_start[POWERPC_NUM_PMC_ENABLED], \
+ pmc_stop[POWERPC_NUM_PMC_ENABLED], \
pmc_loop_index;
#define POWERPC_PERF_START_COUNT(a, cond) do { \
POWERPC_GET_PMC6(pmc_start[5]); \
@@ -141,8 +141,8 @@ extern unsigned long long perfdata[POWERPC_NUM_PMC_ENABLED][powerpc_perf_total][
pmc_loop_index++) \
{ \
if (pmc_stop[pmc_loop_index] >= pmc_start[pmc_loop_index]) \
- { \
- POWERP_PMC_DATATYPE diff = \
+ { \
+ POWERP_PMC_DATATYPE diff = \
pmc_stop[pmc_loop_index] - pmc_start[pmc_loop_index]; \
if (diff < perfdata[pmc_loop_index][a][powerpc_data_min]) \
perfdata[pmc_loop_index][a][powerpc_data_min] = diff; \
diff --git a/libavcodec/ppc/fft_altivec.c b/libavcodec/ppc/fft_altivec.c
index 52aecc163f..1b27579ebd 100644
--- a/libavcodec/ppc/fft_altivec.c
+++ b/libavcodec/ppc/fft_altivec.c
@@ -65,8 +65,8 @@ void ff_fft_calc_altivec(FFTContext *s, FFTComplex *z)
POWERPC_PERF_DECLARE(altivec_fft_num, s->nbits >= 6);
#ifdef ALTIVEC_USE_REFERENCE_C_CODE
int ln = s->nbits;
- int j, np, np2;
- int nblocks, nloops;
+ int j, np, np2;
+ int nblocks, nloops;
register FFTComplex *p, *q;
FFTComplex *exptab = s->exptab;
int l;
@@ -147,8 +147,8 @@ POWERPC_PERF_STOP_COUNT(altivec_fft_num, s->nbits >= 6);
#endif
int ln = s->nbits;
- int j, np, np2;
- int nblocks, nloops;
+ int j, np, np2;
+ int nblocks, nloops;
register FFTComplex *p, *q;
FFTComplex *cptr, *cptr1;
int k;
diff --git a/libavcodec/ppc/gcc_fixes.h b/libavcodec/ppc/gcc_fixes.h
index 194a3f8be4..288fdf834f 100644
--- a/libavcodec/ppc/gcc_fixes.h
+++ b/libavcodec/ppc/gcc_fixes.h
@@ -30,31 +30,31 @@
*/
static inline vector signed char ff_vmrglb (vector signed char const A,
- vector signed char const B)
+ vector signed char const B)
{
static const vector unsigned char lowbyte = {
- 0x08, 0x18, 0x09, 0x19, 0x0a, 0x1a, 0x0b, 0x1b,
- 0x0c, 0x1c, 0x0d, 0x1d, 0x0e, 0x1e, 0x0f, 0x1f
+ 0x08, 0x18, 0x09, 0x19, 0x0a, 0x1a, 0x0b, 0x1b,
+ 0x0c, 0x1c, 0x0d, 0x1d, 0x0e, 0x1e, 0x0f, 0x1f
};
return vec_perm (A, B, lowbyte);
}
static inline vector signed short ff_vmrglh (vector signed short const A,
- vector signed short const B)
+ vector signed short const B)
{
static const vector unsigned char lowhalf = {
- 0x08, 0x09, 0x18, 0x19, 0x0a, 0x0b, 0x1a, 0x1b,
- 0x0c, 0x0d, 0x1c, 0x1d, 0x0e, 0x0f, 0x1e, 0x1f
+ 0x08, 0x09, 0x18, 0x19, 0x0a, 0x0b, 0x1a, 0x1b,
+ 0x0c, 0x0d, 0x1c, 0x1d, 0x0e, 0x0f, 0x1e, 0x1f
};
return vec_perm (A, B, lowhalf);
}
static inline vector signed int ff_vmrglw (vector signed int const A,
- vector signed int const B)
+ vector signed int const B)
{
static const vector unsigned char lowword = {
- 0x08, 0x09, 0x0a, 0x0b, 0x18, 0x19, 0x1a, 0x1b,
- 0x0c, 0x0d, 0x0e, 0x0f, 0x1c, 0x1d, 0x1e, 0x1f
+ 0x08, 0x09, 0x0a, 0x0b, 0x18, 0x19, 0x1a, 0x1b,
+ 0x0c, 0x0d, 0x0e, 0x0f, 0x1c, 0x1d, 0x1e, 0x1f
};
return vec_perm (A, B, lowword);
}
diff --git a/libavcodec/ppc/idct_altivec.c b/libavcodec/ppc/idct_altivec.c
index ac91199ea5..85bc57baef 100644
--- a/libavcodec/ppc/idct_altivec.c
+++ b/libavcodec/ppc/idct_altivec.c
@@ -51,108 +51,108 @@
#define vector_s32_t vector signed int
#define vector_u32_t vector unsigned int
-#define IDCT_HALF \
- /* 1st stage */ \
- t1 = vec_mradds (a1, vx7, vx1 ); \
- t8 = vec_mradds (a1, vx1, vec_subs (zero, vx7)); \
- t7 = vec_mradds (a2, vx5, vx3); \
- t3 = vec_mradds (ma2, vx3, vx5); \
- \
- /* 2nd stage */ \
- t5 = vec_adds (vx0, vx4); \
- t0 = vec_subs (vx0, vx4); \
- t2 = vec_mradds (a0, vx6, vx2); \
- t4 = vec_mradds (a0, vx2, vec_subs (zero, vx6)); \
- t6 = vec_adds (t8, t3); \
- t3 = vec_subs (t8, t3); \
- t8 = vec_subs (t1, t7); \
- t1 = vec_adds (t1, t7); \
- \
- /* 3rd stage */ \
- t7 = vec_adds (t5, t2); \
- t2 = vec_subs (t5, t2); \
- t5 = vec_adds (t0, t4); \
- t0 = vec_subs (t0, t4); \
- t4 = vec_subs (t8, t3); \
- t3 = vec_adds (t8, t3); \
- \
- /* 4th stage */ \
- vy0 = vec_adds (t7, t1); \
- vy7 = vec_subs (t7, t1); \
- vy1 = vec_mradds (c4, t3, t5); \
- vy6 = vec_mradds (mc4, t3, t5); \
- vy2 = vec_mradds (c4, t4, t0); \
- vy5 = vec_mradds (mc4, t4, t0); \
- vy3 = vec_adds (t2, t6); \
+#define IDCT_HALF \
+ /* 1st stage */ \
+ t1 = vec_mradds (a1, vx7, vx1 ); \
+ t8 = vec_mradds (a1, vx1, vec_subs (zero, vx7)); \
+ t7 = vec_mradds (a2, vx5, vx3); \
+ t3 = vec_mradds (ma2, vx3, vx5); \
+ \
+ /* 2nd stage */ \
+ t5 = vec_adds (vx0, vx4); \
+ t0 = vec_subs (vx0, vx4); \
+ t2 = vec_mradds (a0, vx6, vx2); \
+ t4 = vec_mradds (a0, vx2, vec_subs (zero, vx6)); \
+ t6 = vec_adds (t8, t3); \
+ t3 = vec_subs (t8, t3); \
+ t8 = vec_subs (t1, t7); \
+ t1 = vec_adds (t1, t7); \
+ \
+ /* 3rd stage */ \
+ t7 = vec_adds (t5, t2); \
+ t2 = vec_subs (t5, t2); \
+ t5 = vec_adds (t0, t4); \
+ t0 = vec_subs (t0, t4); \
+ t4 = vec_subs (t8, t3); \
+ t3 = vec_adds (t8, t3); \
+ \
+ /* 4th stage */ \
+ vy0 = vec_adds (t7, t1); \
+ vy7 = vec_subs (t7, t1); \
+ vy1 = vec_mradds (c4, t3, t5); \
+ vy6 = vec_mradds (mc4, t3, t5); \
+ vy2 = vec_mradds (c4, t4, t0); \
+ vy5 = vec_mradds (mc4, t4, t0); \
+ vy3 = vec_adds (t2, t6); \
vy4 = vec_subs (t2, t6);
-#define IDCT \
- vector_s16_t vx0, vx1, vx2, vx3, vx4, vx5, vx6, vx7; \
- vector_s16_t vy0, vy1, vy2, vy3, vy4, vy5, vy6, vy7; \
- vector_s16_t a0, a1, a2, ma2, c4, mc4, zero, bias; \
- vector_s16_t t0, t1, t2, t3, t4, t5, t6, t7, t8; \
- vector_u16_t shift; \
- \
- c4 = vec_splat (constants[0], 0); \
- a0 = vec_splat (constants[0], 1); \
- a1 = vec_splat (constants[0], 2); \
- a2 = vec_splat (constants[0], 3); \
- mc4 = vec_splat (constants[0], 4); \
- ma2 = vec_splat (constants[0], 5); \
- bias = (vector_s16_t)vec_splat ((vector_s32_t)constants[0], 3); \
- \
- zero = vec_splat_s16 (0); \
- shift = vec_splat_u16 (4); \
- \
- vx0 = vec_mradds (vec_sl (block[0], shift), constants[1], zero); \
- vx1 = vec_mradds (vec_sl (block[1], shift), constants[2], zero); \
- vx2 = vec_mradds (vec_sl (block[2], shift), constants[3], zero); \
- vx3 = vec_mradds (vec_sl (block[3], shift), constants[4], zero); \
- vx4 = vec_mradds (vec_sl (block[4], shift), constants[1], zero); \
- vx5 = vec_mradds (vec_sl (block[5], shift), constants[4], zero); \
- vx6 = vec_mradds (vec_sl (block[6], shift), constants[3], zero); \
- vx7 = vec_mradds (vec_sl (block[7], shift), constants[2], zero); \
- \
- IDCT_HALF \
- \
- vx0 = vec_mergeh (vy0, vy4); \
- vx1 = vec_mergel (vy0, vy4); \
- vx2 = vec_mergeh (vy1, vy5); \
- vx3 = vec_mergel (vy1, vy5); \
- vx4 = vec_mergeh (vy2, vy6); \
- vx5 = vec_mergel (vy2, vy6); \
- vx6 = vec_mergeh (vy3, vy7); \
- vx7 = vec_mergel (vy3, vy7); \
- \
- vy0 = vec_mergeh (vx0, vx4); \
- vy1 = vec_mergel (vx0, vx4); \
- vy2 = vec_mergeh (vx1, vx5); \
- vy3 = vec_mergel (vx1, vx5); \
- vy4 = vec_mergeh (vx2, vx6); \
- vy5 = vec_mergel (vx2, vx6); \
- vy6 = vec_mergeh (vx3, vx7); \
- vy7 = vec_mergel (vx3, vx7); \
- \
- vx0 = vec_adds (vec_mergeh (vy0, vy4), bias); \
- vx1 = vec_mergel (vy0, vy4); \
- vx2 = vec_mergeh (vy1, vy5); \
- vx3 = vec_mergel (vy1, vy5); \
- vx4 = vec_mergeh (vy2, vy6); \
- vx5 = vec_mergel (vy2, vy6); \
- vx6 = vec_mergeh (vy3, vy7); \
- vx7 = vec_mergel (vy3, vy7); \
- \
- IDCT_HALF \
- \
- shift = vec_splat_u16 (6); \
- vx0 = vec_sra (vy0, shift); \
- vx1 = vec_sra (vy1, shift); \
- vx2 = vec_sra (vy2, shift); \
- vx3 = vec_sra (vy3, shift); \
- vx4 = vec_sra (vy4, shift); \
- vx5 = vec_sra (vy5, shift); \
- vx6 = vec_sra (vy6, shift); \
+#define IDCT \
+ vector_s16_t vx0, vx1, vx2, vx3, vx4, vx5, vx6, vx7; \
+ vector_s16_t vy0, vy1, vy2, vy3, vy4, vy5, vy6, vy7; \
+ vector_s16_t a0, a1, a2, ma2, c4, mc4, zero, bias; \
+ vector_s16_t t0, t1, t2, t3, t4, t5, t6, t7, t8; \
+ vector_u16_t shift; \
+ \
+ c4 = vec_splat (constants[0], 0); \
+ a0 = vec_splat (constants[0], 1); \
+ a1 = vec_splat (constants[0], 2); \
+ a2 = vec_splat (constants[0], 3); \
+ mc4 = vec_splat (constants[0], 4); \
+ ma2 = vec_splat (constants[0], 5); \
+ bias = (vector_s16_t)vec_splat ((vector_s32_t)constants[0], 3); \
+ \
+ zero = vec_splat_s16 (0); \
+ shift = vec_splat_u16 (4); \
+ \
+ vx0 = vec_mradds (vec_sl (block[0], shift), constants[1], zero); \
+ vx1 = vec_mradds (vec_sl (block[1], shift), constants[2], zero); \
+ vx2 = vec_mradds (vec_sl (block[2], shift), constants[3], zero); \
+ vx3 = vec_mradds (vec_sl (block[3], shift), constants[4], zero); \
+ vx4 = vec_mradds (vec_sl (block[4], shift), constants[1], zero); \
+ vx5 = vec_mradds (vec_sl (block[5], shift), constants[4], zero); \
+ vx6 = vec_mradds (vec_sl (block[6], shift), constants[3], zero); \
+ vx7 = vec_mradds (vec_sl (block[7], shift), constants[2], zero); \
+ \
+ IDCT_HALF \
+ \
+ vx0 = vec_mergeh (vy0, vy4); \
+ vx1 = vec_mergel (vy0, vy4); \
+ vx2 = vec_mergeh (vy1, vy5); \
+ vx3 = vec_mergel (vy1, vy5); \
+ vx4 = vec_mergeh (vy2, vy6); \
+ vx5 = vec_mergel (vy2, vy6); \
+ vx6 = vec_mergeh (vy3, vy7); \
+ vx7 = vec_mergel (vy3, vy7); \
+ \
+ vy0 = vec_mergeh (vx0, vx4); \
+ vy1 = vec_mergel (vx0, vx4); \
+ vy2 = vec_mergeh (vx1, vx5); \
+ vy3 = vec_mergel (vx1, vx5); \
+ vy4 = vec_mergeh (vx2, vx6); \
+ vy5 = vec_mergel (vx2, vx6); \
+ vy6 = vec_mergeh (vx3, vx7); \
+ vy7 = vec_mergel (vx3, vx7); \
+ \
+ vx0 = vec_adds (vec_mergeh (vy0, vy4), bias); \
+ vx1 = vec_mergel (vy0, vy4); \
+ vx2 = vec_mergeh (vy1, vy5); \
+ vx3 = vec_mergel (vy1, vy5); \
+ vx4 = vec_mergeh (vy2, vy6); \
+ vx5 = vec_mergel (vy2, vy6); \
+ vx6 = vec_mergeh (vy3, vy7); \
+ vx7 = vec_mergel (vy3, vy7); \
+ \
+ IDCT_HALF \
+ \
+ shift = vec_splat_u16 (6); \
+ vx0 = vec_sra (vy0, shift); \
+ vx1 = vec_sra (vy1, shift); \
+ vx2 = vec_sra (vy2, shift); \
+ vx3 = vec_sra (vy3, shift); \
+ vx4 = vec_sra (vy4, shift); \
+ vx5 = vec_sra (vy5, shift); \
+ vx6 = vec_sra (vy6, shift); \
vx7 = vec_sra (vy7, shift);
@@ -180,18 +180,18 @@ POWERPC_PERF_START_COUNT(altivec_idct_put_num, 1);
#endif
IDCT
-#define COPY(dest,src) \
- tmp = vec_packsu (src, src); \
- vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); \
+#define COPY(dest,src) \
+ tmp = vec_packsu (src, src); \
+ vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); \
vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest);
- COPY (dest, vx0) dest += stride;
- COPY (dest, vx1) dest += stride;
- COPY (dest, vx2) dest += stride;
- COPY (dest, vx3) dest += stride;
- COPY (dest, vx4) dest += stride;
- COPY (dest, vx5) dest += stride;
- COPY (dest, vx6) dest += stride;
+ COPY (dest, vx0) dest += stride;
+ COPY (dest, vx1) dest += stride;
+ COPY (dest, vx2) dest += stride;
+ COPY (dest, vx3) dest += stride;
+ COPY (dest, vx4) dest += stride;
+ COPY (dest, vx5) dest += stride;
+ COPY (dest, vx6) dest += stride;
COPY (dest, vx7)
POWERPC_PERF_STOP_COUNT(altivec_idct_put_num, 1);
@@ -225,22 +225,22 @@ POWERPC_PERF_START_COUNT(altivec_idct_add_num, 1);
perm0 = vec_mergeh (p, p0);
perm1 = vec_mergeh (p, p1);
-#define ADD(dest,src,perm) \
- /* *(uint64_t *)&tmp = *(uint64_t *)dest; */ \
- tmp = vec_ld (0, dest); \
- tmp2 = (vector_s16_t)vec_perm (tmp, (vector_u8_t)zero, perm); \
- tmp3 = vec_adds (tmp2, src); \
- tmp = vec_packsu (tmp3, tmp3); \
- vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); \
+#define ADD(dest,src,perm) \
+ /* *(uint64_t *)&tmp = *(uint64_t *)dest; */ \
+ tmp = vec_ld (0, dest); \
+ tmp2 = (vector_s16_t)vec_perm (tmp, (vector_u8_t)zero, perm); \
+ tmp3 = vec_adds (tmp2, src); \
+ tmp = vec_packsu (tmp3, tmp3); \
+ vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); \
vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest);
- ADD (dest, vx0, perm0) dest += stride;
- ADD (dest, vx1, perm1) dest += stride;
- ADD (dest, vx2, perm0) dest += stride;
- ADD (dest, vx3, perm1) dest += stride;
- ADD (dest, vx4, perm0) dest += stride;
- ADD (dest, vx5, perm1) dest += stride;
- ADD (dest, vx6, perm0) dest += stride;
+ ADD (dest, vx0, perm0) dest += stride;
+ ADD (dest, vx1, perm1) dest += stride;
+ ADD (dest, vx2, perm0) dest += stride;
+ ADD (dest, vx3, perm1) dest += stride;
+ ADD (dest, vx4, perm0) dest += stride;
+ ADD (dest, vx5, perm1) dest += stride;
+ ADD (dest, vx6, perm0) dest += stride;
ADD (dest, vx7, perm1)
POWERPC_PERF_STOP_COUNT(altivec_idct_add_num, 1);
diff --git a/libavcodec/ppc/mpegvideo_altivec.c b/libavcodec/ppc/mpegvideo_altivec.c
index 4477d3ffa8..647221d83f 100644
--- a/libavcodec/ppc/mpegvideo_altivec.c
+++ b/libavcodec/ppc/mpegvideo_altivec.c
@@ -152,9 +152,9 @@ int dct_quantize_altivec(MpegEncContext* s,
}
// The following block could exist as a separate an altivec dct
- // function. However, if we put it inline, the DCT data can remain
- // in the vector local variables, as floats, which we'll use during the
- // quantize step...
+ // function. However, if we put it inline, the DCT data can remain
+ // in the vector local variables, as floats, which we'll use during the
+ // quantize step...
{
const vector float vec_0_298631336 = (vector float)FOUROF(0.298631336f);
const vector float vec_0_390180644 = (vector float)FOUROF(-0.390180644f);
@@ -206,11 +206,11 @@ int dct_quantize_altivec(MpegEncContext* s,
z1 = vec_madd(vec_add(tmp12, tmp13), vec_0_541196100, (vector float)zero);
// dataptr[2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
- // CONST_BITS-PASS1_BITS);
+ // CONST_BITS-PASS1_BITS);
row2 = vec_madd(tmp13, vec_0_765366865, z1);
// dataptr[6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
- // CONST_BITS-PASS1_BITS);
+ // CONST_BITS-PASS1_BITS);
row6 = vec_madd(tmp12, vec_1_847759065, z1);
z1 = vec_add(tmp4, tmp7); // z1 = tmp4 + tmp7;
@@ -315,7 +315,7 @@ int dct_quantize_altivec(MpegEncContext* s,
}
// Load the bias vector (We add 0.5 to the bias so that we're
- // rounding when we convert to int, instead of flooring.)
+ // rounding when we convert to int, instead of flooring.)
{
vector signed int biasInt;
const vector float negOneFloat = (vector float)FOUROF(-1.0f);
generated by cgit v1.2.3 (git 2.39.1) at 2024-07-11 02:22:45 +0000