-/**
+/*
* VP8 compatible video decoder
*
* Copyright (C) 2010 David Conrad
*
- * This file is part of FFmpeg.
+ * This file is part of Libav.
*
- * FFmpeg is free software; you can redistribute it and/or
+ * Libav is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
- * FFmpeg is distributed in the hope that it will be useful,
+ * Libav is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with FFmpeg; if not, write to the Free Software
+ * License along with Libav; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
+#include "config.h"
#include "libavutil/cpu.h"
+#include "libavutil/mem.h"
+#include "libavutil/ppc/cpu.h"
+#include "libavutil/ppc/types_altivec.h"
+#include "libavutil/ppc/util_altivec.h"
#include "libavcodec/vp8dsp.h"
-#include "dsputil_altivec.h"
-#include "types_altivec.h"
-#include "util_altivec.h"
+#include "hpeldsp_altivec.h"
+#if HAVE_ALTIVEC && HAVE_BIGENDIAN
#define REPT4(...) { __VA_ARGS__, __VA_ARGS__, __VA_ARGS__, __VA_ARGS__ }
// h subpel filter uses msum to multiply+add 4 pixel taps at once
dstv = vec_sra(dstv, c7)
static av_always_inline
-void put_vp8_epel_h_altivec_core(uint8_t *dst, int dst_stride,
- uint8_t *src, int src_stride,
+void put_vp8_epel_h_altivec_core(uint8_t *dst, ptrdiff_t dst_stride,
+ uint8_t *src, ptrdiff_t src_stride,
int h, int mx, int w, int is6tap)
{
LOAD_H_SUBPEL_FILTER(mx-1);
dstv = vec_sra(dstv, c7)
static av_always_inline
-void put_vp8_epel_v_altivec_core(uint8_t *dst, int dst_stride,
- uint8_t *src, int src_stride,
+void put_vp8_epel_v_altivec_core(uint8_t *dst, ptrdiff_t dst_stride,
+ uint8_t *src, ptrdiff_t src_stride,
int h, int my, int w, int is6tap)
{
LOAD_V_SUBPEL_FILTER(my-1);
#define EPEL_FUNCS(WIDTH, TAPS) \
static av_noinline \
-void put_vp8_epel ## WIDTH ## _h ## TAPS ## _altivec(uint8_t *dst, int dst_stride, uint8_t *src, int src_stride, int h, int mx, int my) \
+void put_vp8_epel ## WIDTH ## _h ## TAPS ## _altivec(uint8_t *dst, ptrdiff_t dst_stride, uint8_t *src, ptrdiff_t src_stride, int h, int mx, int my) \
{ \
put_vp8_epel_h_altivec_core(dst, dst_stride, src, src_stride, h, mx, WIDTH, TAPS == 6); \
} \
\
static av_noinline \
-void put_vp8_epel ## WIDTH ## _v ## TAPS ## _altivec(uint8_t *dst, int dst_stride, uint8_t *src, int src_stride, int h, int mx, int my) \
+void put_vp8_epel ## WIDTH ## _v ## TAPS ## _altivec(uint8_t *dst, ptrdiff_t dst_stride, uint8_t *src, ptrdiff_t src_stride, int h, int mx, int my) \
{ \
put_vp8_epel_v_altivec_core(dst, dst_stride, src, src_stride, h, my, WIDTH, TAPS == 6); \
}
#define EPEL_HV(WIDTH, HTAPS, VTAPS) \
-static void put_vp8_epel ## WIDTH ## _h ## HTAPS ## v ## VTAPS ## _altivec(uint8_t *dst, int stride, uint8_t *src, int s, int h, int mx, int my) \
+static void put_vp8_epel ## WIDTH ## _h ## HTAPS ## v ## VTAPS ## _altivec(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, ptrdiff_t sstride, int h, int mx, int my) \
{ \
DECLARE_ALIGNED(16, uint8_t, tmp)[(2*WIDTH+5)*16]; \
if (VTAPS == 6) { \
- put_vp8_epel ## WIDTH ## _h ## HTAPS ## _altivec(tmp, 16, src-2*stride, stride, h+5, mx, my); \
- put_vp8_epel ## WIDTH ## _v ## VTAPS ## _altivec(dst, stride, tmp+2*16, 16, h, mx, my); \
+ put_vp8_epel ## WIDTH ## _h ## HTAPS ## _altivec(tmp, 16, src-2*sstride, sstride, h+5, mx, my); \
+ put_vp8_epel ## WIDTH ## _v ## VTAPS ## _altivec(dst, dstride, tmp+2*16, 16, h, mx, my); \
} else { \
- put_vp8_epel ## WIDTH ## _h ## HTAPS ## _altivec(tmp, 16, src-stride, stride, h+4, mx, my); \
- put_vp8_epel ## WIDTH ## _v ## VTAPS ## _altivec(dst, stride, tmp+16, 16, h, mx, my); \
+ put_vp8_epel ## WIDTH ## _h ## HTAPS ## _altivec(tmp, 16, src-sstride, sstride, h+4, mx, my); \
+ put_vp8_epel ## WIDTH ## _v ## VTAPS ## _altivec(dst, dstride, tmp+16, 16, h, mx, my); \
} \
}
EPEL_HV(4, 6,4)
EPEL_HV(4, 4,4)
-static void put_vp8_pixels16_altivec(uint8_t *dst, int stride, uint8_t *src, int s, int h, int mx, int my)
+static void put_vp8_pixels16_altivec(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, ptrdiff_t sstride, int h, int mx, int my)
{
- put_pixels16_altivec(dst, src, stride, h);
+ register vector unsigned char pixelsv1, pixelsv2;
+ register vector unsigned char pixelsv1B, pixelsv2B;
+ register vector unsigned char pixelsv1C, pixelsv2C;
+ register vector unsigned char pixelsv1D, pixelsv2D;
+
+ register vector unsigned char perm = vec_lvsl(0, src);
+ int i;
+ register ptrdiff_t dstride2 = dstride << 1, sstride2 = sstride << 1;
+ register ptrdiff_t dstride3 = dstride2 + dstride, sstride3 = sstride + sstride2;
+ register ptrdiff_t dstride4 = dstride << 2, sstride4 = sstride << 2;
+
+// hand-unrolling the loop by 4 gains about 15%
+// mininum execution time goes from 74 to 60 cycles
+// it's faster than -funroll-loops, but using
+// -funroll-loops w/ this is bad - 74 cycles again.
+// all this is on a 7450, tuning for the 7450
+ for (i = 0; i < h; i += 4) {
+ pixelsv1 = vec_ld( 0, src);
+ pixelsv2 = vec_ld(15, src);
+ pixelsv1B = vec_ld(sstride, src);
+ pixelsv2B = vec_ld(15 + sstride, src);
+ pixelsv1C = vec_ld(sstride2, src);
+ pixelsv2C = vec_ld(15 + sstride2, src);
+ pixelsv1D = vec_ld(sstride3, src);
+ pixelsv2D = vec_ld(15 + sstride3, src);
+ vec_st(vec_perm(pixelsv1, pixelsv2, perm),
+ 0, (unsigned char*)dst);
+ vec_st(vec_perm(pixelsv1B, pixelsv2B, perm),
+ dstride, (unsigned char*)dst);
+ vec_st(vec_perm(pixelsv1C, pixelsv2C, perm),
+ dstride2, (unsigned char*)dst);
+ vec_st(vec_perm(pixelsv1D, pixelsv2D, perm),
+ dstride3, (unsigned char*)dst);
+ src += sstride4;
+ dst += dstride4;
+ }
}
-av_cold void ff_vp8dsp_init_altivec(VP8DSPContext *c)
+#endif /* HAVE_ALTIVEC */
+
+
+av_cold void ff_vp78dsp_init_ppc(VP8DSPContext *c)
{
- if (!(av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC))
+#if HAVE_ALTIVEC && HAVE_BIGENDIAN
+ if (!PPC_ALTIVEC(av_get_cpu_flags()))
return;
c->put_vp8_epel_pixels_tab[0][0][0] = put_vp8_pixels16_altivec;
c->put_vp8_epel_pixels_tab[2][1][1] = put_vp8_epel4_h4v4_altivec;
c->put_vp8_epel_pixels_tab[2][1][2] = put_vp8_epel4_h6v4_altivec;
c->put_vp8_epel_pixels_tab[2][2][1] = put_vp8_epel4_h4v6_altivec;
+#endif /* HAVE_ALTIVEC */
}