summaryrefslogtreecommitdiff
path: root/contrib/ffmpeg/libavcodec/ppc
diff options
context:
space:
mode:
Diffstat (limited to 'contrib/ffmpeg/libavcodec/ppc')
-rw-r--r--contrib/ffmpeg/libavcodec/ppc/dsputil_altivec.c6
-rw-r--r--contrib/ffmpeg/libavcodec/ppc/dsputil_ppc.c10
-rw-r--r--contrib/ffmpeg/libavcodec/ppc/dsputil_ppc.h6
-rw-r--r--contrib/ffmpeg/libavcodec/ppc/float_altivec.c3
-rw-r--r--contrib/ffmpeg/libavcodec/ppc/h264_altivec.c129
-rw-r--r--contrib/ffmpeg/libavcodec/ppc/idct_altivec.c4
6 files changed, 141 insertions, 17 deletions
diff --git a/contrib/ffmpeg/libavcodec/ppc/dsputil_altivec.c b/contrib/ffmpeg/libavcodec/ppc/dsputil_altivec.c
index 6f48893a4..bbc53d761 100644
--- a/contrib/ffmpeg/libavcodec/ppc/dsputil_altivec.c
+++ b/contrib/ffmpeg/libavcodec/ppc/dsputil_altivec.c
@@ -1107,12 +1107,10 @@ POWERPC_PERF_START_COUNT(altivec_hadamard8_diff8x8_num, 1);
register vector signed short srcV, dstV; \
register vector signed short but0, but1, but2, op1, op2, op3; \
src1 = vec_ld(stride * i, src); \
- if ((((stride * i) + (unsigned long)src) & 0x0000000F) > 8) \
- src2 = vec_ld((stride * i) + 16, src); \
+ src2 = vec_ld((stride * i) + 15, 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); \
+ dst2 = vec_ld((stride * i) + 15, 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 */ \
diff --git a/contrib/ffmpeg/libavcodec/ppc/dsputil_ppc.c b/contrib/ffmpeg/libavcodec/ppc/dsputil_ppc.c
index 9169eaef0..117a7adf1 100644
--- a/contrib/ffmpeg/libavcodec/ppc/dsputil_ppc.c
+++ b/contrib/ffmpeg/libavcodec/ppc/dsputil_ppc.c
@@ -55,7 +55,7 @@ int mm_support(void)
return result;
}
-#ifdef POWERPC_PERFORMANCE_REPORT
+#ifdef CONFIG_POWERPC_PERF
unsigned long long perfdata[POWERPC_NUM_PMC_ENABLED][powerpc_perf_total][powerpc_data_total];
/* list below must match enum in dsputil_ppc.h */
static unsigned char* perfname[] = {
@@ -90,7 +90,7 @@ static unsigned char* perfname[] = {
#include <stdio.h>
#endif
-#ifdef POWERPC_PERFORMANCE_REPORT
+#ifdef CONFIG_POWERPC_PERF
void powerpc_display_perf_report(void)
{
int i, j;
@@ -112,7 +112,7 @@ void powerpc_display_perf_report(void)
}
}
}
-#endif /* POWERPC_PERFORMANCE_REPORT */
+#endif /* CONFIG_POWERPC_PERF */
/* ***** WARNING ***** WARNING ***** WARNING ***** */
/*
@@ -305,7 +305,7 @@ void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx)
}
}
-#ifdef POWERPC_PERFORMANCE_REPORT
+#ifdef CONFIG_POWERPC_PERF
{
int i, j;
for (i = 0 ; i < powerpc_perf_total ; i++)
@@ -319,7 +319,7 @@ void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx)
}
}
}
-#endif /* POWERPC_PERFORMANCE_REPORT */
+#endif /* CONFIG_POWERPC_PERF */
}
#endif /* HAVE_ALTIVEC */
}
diff --git a/contrib/ffmpeg/libavcodec/ppc/dsputil_ppc.h b/contrib/ffmpeg/libavcodec/ppc/dsputil_ppc.h
index ab2b05780..5b25732b2 100644
--- a/contrib/ffmpeg/libavcodec/ppc/dsputil_ppc.h
+++ b/contrib/ffmpeg/libavcodec/ppc/dsputil_ppc.h
@@ -21,7 +21,7 @@
#ifndef _DSPUTIL_PPC_
#define _DSPUTIL_PPC_
-#ifdef POWERPC_PERFORMANCE_REPORT
+#ifdef CONFIG_POWERPC_PERF
void powerpc_display_perf_report(void);
/* the 604* have 2, the G3* have 4, the G4s have 6,
and the G5 are completely different (they MUST use
@@ -145,11 +145,11 @@ extern unsigned long long perfdata[POWERPC_NUM_PMC_ENABLED][powerpc_perf_total][
} \
} \
} while (0)
-#else /* POWERPC_PERFORMANCE_REPORT */
+#else /* CONFIG_POWERPC_PERF */
// those are needed to avoid empty statements.
#define POWERPC_PERF_DECLARE(a, cond) int altivec_placeholder __attribute__ ((unused))
#define POWERPC_PERF_START_COUNT(a, cond) do {} while (0)
#define POWERPC_PERF_STOP_COUNT(a, cond) do {} while (0)
-#endif /* POWERPC_PERFORMANCE_REPORT */
+#endif /* CONFIG_POWERPC_PERF */
#endif /* _DSPUTIL_PPC_ */
diff --git a/contrib/ffmpeg/libavcodec/ppc/float_altivec.c b/contrib/ffmpeg/libavcodec/ppc/float_altivec.c
index c6e43dec2..22c2de61a 100644
--- a/contrib/ffmpeg/libavcodec/ppc/float_altivec.c
+++ b/contrib/ffmpeg/libavcodec/ppc/float_altivec.c
@@ -76,7 +76,6 @@ static void vector_fmul_add_add_altivec(float *dst, const float *src0,
vector unsigned char align = vec_lvsr(0,dst),
mask = vec_lvsl(0, dst);
- t0 = vec_ld(0, dst);
#if 0 //FIXME: there is still something wrong
if (step == 2) {
int y;
@@ -134,6 +133,7 @@ static void vector_fmul_add_add_altivec(float *dst, const float *src0,
#endif
if (step == 1 && src3 == 0)
for (i=0; i<len-3; i+=4) {
+ t0 = vec_ld(0, dst+i);
t1 = vec_ld(15, dst+i);
s0 = vec_ld(0, src0+i);
s1 = vec_ld(0, src1+i);
@@ -144,7 +144,6 @@ static void vector_fmul_add_add_altivec(float *dst, const float *src0,
t0 = vec_perm(edges, d, align);
vec_st(t1, 15, dst+i);
vec_st(t0, 0, dst+i);
- t0 = t1;
}
else
ff_vector_fmul_add_add_c(dst, src0, src1, src2, src3, len, step);
diff --git a/contrib/ffmpeg/libavcodec/ppc/h264_altivec.c b/contrib/ffmpeg/libavcodec/ppc/h264_altivec.c
index 4aa366f97..bac620e82 100644
--- a/contrib/ffmpeg/libavcodec/ppc/h264_altivec.c
+++ b/contrib/ffmpeg/libavcodec/ppc/h264_altivec.c
@@ -23,6 +23,7 @@
#include "gcc_fixes.h"
#include "dsputil_altivec.h"
+#include "types_altivec.h"
#define PUT_OP_U8_ALTIVEC(d, s, dst) d = s
#define AVG_OP_U8_ALTIVEC(d, s, dst) d = vec_avg(dst, s)
@@ -202,7 +203,7 @@ void put_no_rnd_h264_chroma_mc8_altivec(uint8_t * dst, uint8_t * src, int stride
vector signed short vsrc0ssH, vsrc1ssH;
vector unsigned char vsrcCuc, vsrc2uc, vsrc3uc;
vector signed short vsrc2ssH, vsrc3ssH, psum;
- vector unsigned char vdst, ppsum, vfdst, fsum;
+ vector unsigned char vdst, ppsum, fsum;
if (((unsigned long)dst) % 16 == 0) {
fperm = (vector unsigned char)AVV(0x10, 0x11, 0x12, 0x13,
@@ -398,6 +399,131 @@ static inline void avg_pixels16_l2_altivec( uint8_t * dst, const uint8_t * src1,
H264_MC(put_, 16, altivec)
H264_MC(avg_, 16, altivec)
+
+/****************************************************************************
+ * IDCT transform:
+ ****************************************************************************/
+
+#define IDCT8_1D_ALTIVEC(s0, s1, s2, s3, s4, s5, s6, s7, d0, d1, d2, d3, d4, d5, d6, d7) {\
+ /* a0 = SRC(0) + SRC(4); */ \
+ vec_s16_t a0v = vec_add(s0, s4); \
+ /* a2 = SRC(0) - SRC(4); */ \
+ vec_s16_t a2v = vec_sub(s0, s4); \
+ /* a4 = (SRC(2)>>1) - SRC(6); */ \
+ vec_s16_t a4v = vec_sub(vec_sra(s2, onev), s6); \
+ /* a6 = (SRC(6)>>1) + SRC(2); */ \
+ vec_s16_t a6v = vec_add(vec_sra(s6, onev), s2); \
+ /* b0 = a0 + a6; */ \
+ vec_s16_t b0v = vec_add(a0v, a6v); \
+ /* b2 = a2 + a4; */ \
+ vec_s16_t b2v = vec_add(a2v, a4v); \
+ /* b4 = a2 - a4; */ \
+ vec_s16_t b4v = vec_sub(a2v, a4v); \
+ /* b6 = a0 - a6; */ \
+ vec_s16_t b6v = vec_sub(a0v, a6v); \
+ /* a1 = SRC(5) - SRC(3) - SRC(7) - (SRC(7)>>1); */ \
+ /* a1 = (SRC(5)-SRC(3)) - (SRC(7) + (SRC(7)>>1)); */ \
+ vec_s16_t a1v = vec_sub( vec_sub(s5, s3), vec_add(s7, vec_sra(s7, onev)) ); \
+ /* a3 = SRC(7) + SRC(1) - SRC(3) - (SRC(3)>>1); */ \
+ /* a3 = (SRC(7)+SRC(1)) - (SRC(3) + (SRC(3)>>1)); */ \
+ vec_s16_t a3v = vec_sub( vec_add(s7, s1), vec_add(s3, vec_sra(s3, onev)) );\
+ /* a5 = SRC(7) - SRC(1) + SRC(5) + (SRC(5)>>1); */ \
+ /* a5 = (SRC(7)-SRC(1)) + SRC(5) + (SRC(5)>>1); */ \
+ vec_s16_t a5v = vec_add( vec_sub(s7, s1), vec_add(s5, vec_sra(s5, onev)) );\
+ /* a7 = SRC(5)+SRC(3) + SRC(1) + (SRC(1)>>1); */ \
+ vec_s16_t a7v = vec_add( vec_add(s5, s3), vec_add(s1, vec_sra(s1, onev)) );\
+ /* b1 = (a7>>2) + a1; */ \
+ vec_s16_t b1v = vec_add( vec_sra(a7v, twov), a1v); \
+ /* b3 = a3 + (a5>>2); */ \
+ vec_s16_t b3v = vec_add(a3v, vec_sra(a5v, twov)); \
+ /* b5 = (a3>>2) - a5; */ \
+ vec_s16_t b5v = vec_sub( vec_sra(a3v, twov), a5v); \
+ /* b7 = a7 - (a1>>2); */ \
+ vec_s16_t b7v = vec_sub( a7v, vec_sra(a1v, twov)); \
+ /* DST(0, b0 + b7); */ \
+ d0 = vec_add(b0v, b7v); \
+ /* DST(1, b2 + b5); */ \
+ d1 = vec_add(b2v, b5v); \
+ /* DST(2, b4 + b3); */ \
+ d2 = vec_add(b4v, b3v); \
+ /* DST(3, b6 + b1); */ \
+ d3 = vec_add(b6v, b1v); \
+ /* DST(4, b6 - b1); */ \
+ d4 = vec_sub(b6v, b1v); \
+ /* DST(5, b4 - b3); */ \
+ d5 = vec_sub(b4v, b3v); \
+ /* DST(6, b2 - b5); */ \
+ d6 = vec_sub(b2v, b5v); \
+ /* DST(7, b0 - b7); */ \
+ d7 = vec_sub(b0v, b7v); \
+}
+
+#define ALTIVEC_STORE_SUM_CLIP(dest, idctv, perm_ldv, perm_stv, sel) { \
+ /* unaligned load */ \
+ vec_u8_t hv = vec_ld( 0, dest ); \
+ vec_u8_t lv = vec_ld( 7, dest ); \
+ vec_u8_t dstv = vec_perm( hv, lv, (vec_u8_t)perm_ldv ); \
+ vec_s16_t idct_sh6 = vec_sra(idctv, sixv); \
+ vec_u16_t dst16 = (vec_u16_t)vec_mergeh(zero_u8v, dstv); \
+ vec_s16_t idstsum = vec_adds(idct_sh6, (vec_s16_t)dst16); \
+ vec_u8_t idstsum8 = vec_packsu(zero_s16v, idstsum); \
+ vec_u8_t edgehv; \
+ /* unaligned store */ \
+ vec_u8_t bodyv = vec_perm( idstsum8, idstsum8, perm_stv );\
+ vec_u8_t edgelv = vec_perm( sel, zero_u8v, perm_stv ); \
+ lv = vec_sel( lv, bodyv, edgelv ); \
+ vec_st( lv, 7, dest ); \
+ hv = vec_ld( 0, dest ); \
+ edgehv = vec_perm( zero_u8v, sel, perm_stv ); \
+ hv = vec_sel( hv, bodyv, edgehv ); \
+ vec_st( hv, 0, dest ); \
+ }
+
+void ff_h264_idct8_add_altivec( uint8_t *dst, DCTELEM *dct, int stride ) {
+ vec_s16_t s0, s1, s2, s3, s4, s5, s6, s7;
+ vec_s16_t d0, d1, d2, d3, d4, d5, d6, d7;
+ vec_s16_t idct0, idct1, idct2, idct3, idct4, idct5, idct6, idct7;
+
+ vec_u8_t perm_ldv = vec_lvsl(0, dst);
+ vec_u8_t perm_stv = vec_lvsr(8, dst);
+
+ const vec_u16_t onev = vec_splat_u16(1);
+ const vec_u16_t twov = vec_splat_u16(2);
+ const vec_u16_t sixv = vec_splat_u16(6);
+
+ const vec_u8_t sel = (vec_u8_t) AVV(0,0,0,0,0,0,0,0,
+ -1,-1,-1,-1,-1,-1,-1,-1);
+ LOAD_ZERO;
+
+ dct[0] += 32; // rounding for the >>6 at the end
+
+ s0 = vec_ld(0x00, (int16_t*)dct);
+ s1 = vec_ld(0x10, (int16_t*)dct);
+ s2 = vec_ld(0x20, (int16_t*)dct);
+ s3 = vec_ld(0x30, (int16_t*)dct);
+ s4 = vec_ld(0x40, (int16_t*)dct);
+ s5 = vec_ld(0x50, (int16_t*)dct);
+ s6 = vec_ld(0x60, (int16_t*)dct);
+ s7 = vec_ld(0x70, (int16_t*)dct);
+
+ IDCT8_1D_ALTIVEC(s0, s1, s2, s3, s4, s5, s6, s7,
+ d0, d1, d2, d3, d4, d5, d6, d7);
+
+ TRANSPOSE8( d0, d1, d2, d3, d4, d5, d6, d7 );
+
+ IDCT8_1D_ALTIVEC(d0, d1, d2, d3, d4, d5, d6, d7,
+ idct0, idct1, idct2, idct3, idct4, idct5, idct6, idct7);
+
+ ALTIVEC_STORE_SUM_CLIP(&dst[0*stride], idct0, perm_ldv, perm_stv, sel);
+ ALTIVEC_STORE_SUM_CLIP(&dst[1*stride], idct1, perm_ldv, perm_stv, sel);
+ ALTIVEC_STORE_SUM_CLIP(&dst[2*stride], idct2, perm_ldv, perm_stv, sel);
+ ALTIVEC_STORE_SUM_CLIP(&dst[3*stride], idct3, perm_ldv, perm_stv, sel);
+ ALTIVEC_STORE_SUM_CLIP(&dst[4*stride], idct4, perm_ldv, perm_stv, sel);
+ ALTIVEC_STORE_SUM_CLIP(&dst[5*stride], idct5, perm_ldv, perm_stv, sel);
+ ALTIVEC_STORE_SUM_CLIP(&dst[6*stride], idct6, perm_ldv, perm_stv, sel);
+ ALTIVEC_STORE_SUM_CLIP(&dst[7*stride], idct7, perm_ldv, perm_stv, sel);
+}
+
void dsputil_h264_init_ppc(DSPContext* c, AVCodecContext *avctx) {
#ifdef HAVE_ALTIVEC
@@ -405,6 +531,7 @@ void dsputil_h264_init_ppc(DSPContext* c, AVCodecContext *avctx) {
c->put_h264_chroma_pixels_tab[0] = put_h264_chroma_mc8_altivec;
c->put_no_rnd_h264_chroma_pixels_tab[0] = put_no_rnd_h264_chroma_mc8_altivec;
c->avg_h264_chroma_pixels_tab[0] = avg_h264_chroma_mc8_altivec;
+ c->h264_idct8_add = ff_h264_idct8_add_altivec;
#define dspfunc(PFX, IDX, NUM) \
c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_altivec; \
diff --git a/contrib/ffmpeg/libavcodec/ppc/idct_altivec.c b/contrib/ffmpeg/libavcodec/ppc/idct_altivec.c
index cee46fc25..66c8082f7 100644
--- a/contrib/ffmpeg/libavcodec/ppc/idct_altivec.c
+++ b/contrib/ffmpeg/libavcodec/ppc/idct_altivec.c
@@ -171,7 +171,7 @@ void idct_put_altivec(uint8_t* dest, int stride, vector_s16_t* block)
POWERPC_PERF_DECLARE(altivec_idct_put_num, 1);
vector_u8_t tmp;
-#ifdef POWERPC_PERFORMANCE_REPORT
+#ifdef CONFIG_POWERPC_PERF
POWERPC_PERF_START_COUNT(altivec_idct_put_num, 1);
#endif
IDCT
@@ -202,7 +202,7 @@ POWERPC_PERF_DECLARE(altivec_idct_add_num, 1);
vector_u8_t perm1;
vector_u8_t p0, p1, p;
-#ifdef POWERPC_PERFORMANCE_REPORT
+#ifdef CONFIG_POWERPC_PERF
POWERPC_PERF_START_COUNT(altivec_idct_add_num, 1);
#endif
generated by cgit v1.2.3 (git 2.25.1) at 2025-10-17 07:52:52 +0000