---
Could be folded into the previous commit ..
libavcodec/arm/Makefile | 2 +
libavcodec/arm/dsputil_armv6.S | 55 ----------
libavcodec/arm/dsputil_init_armv6.c | 6 --
libavcodec/arm/mpegvideoencdsp_armv6.S | 76 ++++++++++++++
.../mpegvideoencdsp_init_arm.c} | 34 +++----
libavcodec/bfin/Makefile | 2 +
libavcodec/bfin/dsputil.S | 91 -----------------
libavcodec/bfin/dsputil_init.c | 6 --
libavcodec/bfin/mpegvideoencdsp.S | 113 +++++++++++++++++++++
.../mpegvideoencdsp_init.c} | 38 +++----
libavcodec/dnxhdenc.c | 5 +-
libavcodec/dsputil.c | 79 --------------
libavcodec/dsputil.h | 6 --
libavcodec/motion_est.c | 5 +-
libavcodec/mpegvideo_enc.c | 31 +++---
libavcodec/mpegvideoencdsp.c | 85 ++++++++++++++++
libavcodec/mpegvideoencdsp.h | 12 ++-
libavcodec/ppc/Makefile | 1 +
libavcodec/ppc/dsputil_altivec.c | 60 -----------
libavcodec/ppc/mpegvideoencdsp_altivec.c | 102 +++++++++++++++++++
libavcodec/svq1enc.c | 1 +
libavcodec/x86/Makefile | 4 +-
libavcodec/x86/dsputilenc.asm | 69 -------------
libavcodec/x86/dsputilenc_mmx.c | 4 -
libavcodec/x86/mpegvideoencdsp.asm | 95 +++++++++++++++++
.../{mpegvideoencdsp.c => mpegvideoencdsp_init.c} | 11 +-
26 files changed, 555 insertions(+), 438 deletions(-)
create mode 100644 libavcodec/arm/mpegvideoencdsp_armv6.S
copy libavcodec/{mpegvideoencdsp.h => arm/mpegvideoencdsp_init_arm.c} (54%)
create mode 100644 libavcodec/bfin/mpegvideoencdsp.S
copy libavcodec/{mpegvideoencdsp.h => bfin/mpegvideoencdsp_init.c} (52%)
create mode 100644 libavcodec/ppc/mpegvideoencdsp_altivec.c
create mode 100644 libavcodec/x86/mpegvideoencdsp.asm
rename libavcodec/x86/{mpegvideoencdsp.c => mpegvideoencdsp_init.c} (94%)
diff --git a/libavcodec/arm/Makefile b/libavcodec/arm/Makefile
index eba975b..7d81e99 100644
--- a/libavcodec/arm/Makefile
+++ b/libavcodec/arm/Makefile
@@ -20,6 +20,7 @@ OBJS-$(CONFIG_IDCTDSP) +=
arm/idctdsp_init_arm.o \
arm/simple_idct_arm.o
OBJS-$(CONFIG_MPEGAUDIODSP) += arm/mpegaudiodsp_init_arm.o
OBJS-$(CONFIG_MPEGVIDEO) += arm/mpegvideo_arm.o
+OBJS-$(CONFIG_MPEGVIDEOENC) += arm/mpegvideoencdsp_init_arm.o
OBJS-$(CONFIG_NEON_CLOBBER_TEST) += arm/neontest.o
OBJS-$(CONFIG_VECTORDSP) += arm/vectordsp_init_arm.o
OBJS-$(CONFIG_VIDEODSP) += arm/videodsp_init_arm.o
@@ -56,6 +57,7 @@ ARMV6-OBJS-$(CONFIG_IDCTDSP) +=
arm/idctdsp_init_armv6.o \
arm/idctdsp_armv6.o \
arm/simple_idct_armv6.o
ARMV6-OBJS-$(CONFIG_MPEGAUDIODSP) += arm/mpegaudiodsp_fixed_armv6.o
+ARMV6-OBJS-$(CONFIG_MPEGVIDEOENC) += arm/mpegvideoencdsp_armv6.o
ARMV6-OBJS-$(CONFIG_VP8_DECODER) += arm/vp8_armv6.o \
arm/vp8dsp_init_armv6.o \
diff --git a/libavcodec/arm/dsputil_armv6.S b/libavcodec/arm/dsputil_armv6.S
index b89171f..8ff0e4a 100644
--- a/libavcodec/arm/dsputil_armv6.S
+++ b/libavcodec/arm/dsputil_armv6.S
@@ -297,58 +297,3 @@ function ff_sse16_armv6, export=1
pop {r4-r9, pc}
endfunc
-
-function ff_pix_norm1_armv6, export=1
- push {r4-r6, lr}
- mov r12, #16
- mov lr, #0
-1:
- ldm r0, {r2-r5}
- uxtb16 r6, r2
- uxtb16 r2, r2, ror #8
- smlad lr, r6, r6, lr
- uxtb16 r6, r3
- smlad lr, r2, r2, lr
- uxtb16 r3, r3, ror #8
- smlad lr, r6, r6, lr
- uxtb16 r6, r4
- smlad lr, r3, r3, lr
- uxtb16 r4, r4, ror #8
- smlad lr, r6, r6, lr
- uxtb16 r6, r5
- smlad lr, r4, r4, lr
- uxtb16 r5, r5, ror #8
- smlad lr, r6, r6, lr
- subs r12, r12, #1
- add r0, r0, r1
- smlad lr, r5, r5, lr
- bgt 1b
-
- mov r0, lr
- pop {r4-r6, pc}
-endfunc
-
-function ff_pix_sum_armv6, export=1
- push {r4-r7, lr}
- mov r12, #16
- mov r2, #0
- mov r3, #0
- mov lr, #0
- ldr r4, [r0]
-1:
- subs r12, r12, #1
- ldr r5, [r0, #4]
- usada8 r2, r4, lr, r2
- ldr r6, [r0, #8]
- usada8 r3, r5, lr, r3
- ldr r7, [r0, #12]
- usada8 r2, r6, lr, r2
- beq 2f
- ldr_pre r4, r0, r1
- usada8 r3, r7, lr, r3
- bgt 1b
-2:
- usada8 r3, r7, lr, r3
- add r0, r2, r3
- pop {r4-r7, pc}
-endfunc
diff --git a/libavcodec/arm/dsputil_init_armv6.c
b/libavcodec/arm/dsputil_init_armv6.c
index fab5e0d..85ee8e1 100644
--- a/libavcodec/arm/dsputil_init_armv6.c
+++ b/libavcodec/arm/dsputil_init_armv6.c
@@ -43,9 +43,6 @@ int ff_pix_abs8_armv6(MpegEncContext *s, uint8_t *blk1,
uint8_t *blk2,
int ff_sse16_armv6(MpegEncContext *s, uint8_t *blk1, uint8_t *blk2,
int line_size, int h);
-int ff_pix_norm1_armv6(uint8_t *pix, int line_size);
-int ff_pix_sum_armv6(uint8_t *pix, int line_size);
-
av_cold void ff_dsputil_init_armv6(DSPContext *c, AVCodecContext *avctx,
unsigned high_bit_depth)
{
@@ -63,7 +60,4 @@ av_cold void ff_dsputil_init_armv6(DSPContext *c,
AVCodecContext *avctx,
c->sad[1] = ff_pix_abs8_armv6;
c->sse[0] = ff_sse16_armv6;
-
- c->pix_norm1 = ff_pix_norm1_armv6;
- c->pix_sum = ff_pix_sum_armv6;
}
diff --git a/libavcodec/arm/mpegvideoencdsp_armv6.S
b/libavcodec/arm/mpegvideoencdsp_armv6.S
new file mode 100644
index 0000000..99db501
--- /dev/null
+++ b/libavcodec/arm/mpegvideoencdsp_armv6.S
@@ -0,0 +1,76 @@
+/*
+ * Copyright (c) 2009 Mans Rullgard <[email protected]>
+ *
+ * This file is part of Libav.
+ *
+ * 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.
+ *
+ * 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 Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "libavutil/arm/asm.S"
+
+function ff_pix_norm1_armv6, export=1
+ push {r4-r6, lr}
+ mov r12, #16
+ mov lr, #0
+1:
+ ldm r0, {r2-r5}
+ uxtb16 r6, r2
+ uxtb16 r2, r2, ror #8
+ smlad lr, r6, r6, lr
+ uxtb16 r6, r3
+ smlad lr, r2, r2, lr
+ uxtb16 r3, r3, ror #8
+ smlad lr, r6, r6, lr
+ uxtb16 r6, r4
+ smlad lr, r3, r3, lr
+ uxtb16 r4, r4, ror #8
+ smlad lr, r6, r6, lr
+ uxtb16 r6, r5
+ smlad lr, r4, r4, lr
+ uxtb16 r5, r5, ror #8
+ smlad lr, r6, r6, lr
+ subs r12, r12, #1
+ add r0, r0, r1
+ smlad lr, r5, r5, lr
+ bgt 1b
+
+ mov r0, lr
+ pop {r4-r6, pc}
+endfunc
+
+function ff_pix_sum_armv6, export=1
+ push {r4-r7, lr}
+ mov r12, #16
+ mov r2, #0
+ mov r3, #0
+ mov lr, #0
+ ldr r4, [r0]
+1:
+ subs r12, r12, #1
+ ldr r5, [r0, #4]
+ usada8 r2, r4, lr, r2
+ ldr r6, [r0, #8]
+ usada8 r3, r5, lr, r3
+ ldr r7, [r0, #12]
+ usada8 r2, r6, lr, r2
+ beq 2f
+ ldr_pre r4, r0, r1
+ usada8 r3, r7, lr, r3
+ bgt 1b
+2:
+ usada8 r3, r7, lr, r3
+ add r0, r2, r3
+ pop {r4-r7, pc}
+endfunc
diff --git a/libavcodec/mpegvideoencdsp.h
b/libavcodec/arm/mpegvideoencdsp_init_arm.c
similarity index 54%
copy from libavcodec/mpegvideoencdsp.h
copy to libavcodec/arm/mpegvideoencdsp_init_arm.c
index c321698..ab9ba3e 100644
--- a/libavcodec/mpegvideoencdsp.h
+++ b/libavcodec/arm/mpegvideoencdsp_init_arm.c
@@ -16,27 +16,23 @@
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
-#ifndef AVCODEC_MPEGVIDEOENCDSP_H
-#define AVCODEC_MPEGVIDEOENCDSP_H
-
#include <stdint.h>
-#include "avcodec.h"
-
-#define BASIS_SHIFT 16
-#define RECON_SHIFT 6
-
-typedef struct MpegvideoEncDSPContext {
- int (*try_8x8basis)(int16_t rem[64], int16_t weight[64],
- int16_t basis[64], int scale);
- void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
-
-} MpegvideoEncDSPContext;
+#include "libavutil/cpu.h"
+#include "libavutil/arm/cpu.h"
+#include "libavcodec/avcodec.h"
+#include "libavcodec/mpegvideoencdsp.h"
-void ff_mpegvideoencdsp_init(MpegvideoEncDSPContext *c,
- AVCodecContext *avctx);
-void ff_mpegvideoencdsp_init_x86(MpegvideoEncDSPContext *c,
- AVCodecContext *avctx);
+int ff_pix_norm1_armv6(uint8_t *pix, int line_size);
+int ff_pix_sum_armv6(uint8_t *pix, int line_size);
+av_cold void ff_mpegvideoencdsp_init_arm(MpegvideoEncDSPContext *c,
+ AVCodecContext *avctx)
+{
+ int cpu_flags = av_get_cpu_flags();
-#endif /* AVCODEC_MPEGVIDEOENCDSP_H */
+ if (have_armv6(cpu_flags)) {
+ c->pix_norm1 = ff_pix_norm1_armv6;
+ c->pix_sum = ff_pix_sum_armv6;
+ }
+}
diff --git a/libavcodec/bfin/Makefile b/libavcodec/bfin/Makefile
index 8bfa4dd..daa58d0 100644
--- a/libavcodec/bfin/Makefile
+++ b/libavcodec/bfin/Makefile
@@ -9,5 +9,7 @@ OBJS-$(CONFIG_HPELDSP) += bfin/hpeldsp_init.o
\
OBJS-$(CONFIG_IDCTDSP) += bfin/idctdsp_init.o \
bfin/idctdsp.o \
bfin/pixels_clamped.o
+OBJS-$(CONFIG_MPEGVIDEOENC) += bfin/mpegvideoencdsp_init.o \
+ bfin/mpegvideoencdsp.o
OBJS-$(CONFIG_VP3DSP) += bfin/vp3dsp_init.o \
bfin/vp3dsp.o
diff --git a/libavcodec/bfin/dsputil.S b/libavcodec/bfin/dsputil.S
index b1c3563..d169bad 100644
--- a/libavcodec/bfin/dsputil.S
+++ b/libavcodec/bfin/dsputil.S
@@ -49,53 +49,6 @@ DEFUN(diff_pixels,mL1,
rts;
DEFUN_END(diff_pixels)
-/*
- for (i = 0; i < 16; i++) {
- for (j = 0; j < 16; j++) {
- sum += pix[j];
- }
- pix += line_size;
- }
-*/
-DEFUN(pix_sum,mL1,
- (uint8_t *p, int stride)):
- link 0;
- [--sp] = (r7:4);
- p0=8;
- i0=r0; // s1
- i1=r0;
- m1=r1;
- r1=r1+r1;
- r1+=-16; // stride
- m0=r1;
- i1+=m1;
-
- r6=0;
-
- LSETUP(LS$PS,LE$PS) LC0=P0;
- DISALGNEXCPT || R0 = [I0++] || R2 =[I1++];
-
-LS$PS: DISALGNEXCPT || R1 = [I0++] || R3 = [I1++];
- (R5,R4) = BYTEOP16P (R3:2,R1:0) || R0 = [I0++] || R2 = [I1++];
- r6=r6+|+r5;
- r6=r6+|+r4;
- (R5,R4) = BYTEOP16P (R3:2,R1:0) (R)|| R1 = [I0++] || R3 = [I1++];
- r6=r6+|+r5;
- r6=r6+|+r4;
- (R5,R4) = BYTEOP16P (R3:2,R1:0) || R0 = [I0++m0] || R2 = [I1++m0];
- r6=r6+|+r5;
- r6=r6+|+r4;
- (R5,R4) = BYTEOP16P (R3:2,R1:0) (R)|| R0 = [I0++] || R2 = [I1++];
- r6=r6+|+r5;
-LE$PS: r6=r6+|+r4;
- r0.l=r6.l+r6.h;
- r0.h=0;
-
- (r7:4) = [sp++];
- unlink;
- rts;
-DEFUN_END(pix_sum)
-
DEFUN(get_pixels,mL1,
(int16_t *restrict block, const uint8_t *pixels, int line_size)):
@@ -181,50 +134,6 @@ e$8: DISALGNEXCPT || R1 = [I0++] || R3 =
[I1++];
RTS;
DEFUN_END(z_sad8x8)
-DEFUN(pix_norm1,mL1,
- (uint8_t * pix, int line_size)):
- [--SP]=(R7:4,P5:3);
-
- // Fetch the input arguments.
- P1 = R0; // pix
- P0 = R1; // line_size
- P5 = 16; // loop ctr.
- P0 -= P5;
- M0 = P0; // M0 = line_size-16;
- // Now for the real work.
- A1 = A0 = 0;
- lsetup(_pix_norm1_blkfn_loopStart, _pix_norm1_blkfn_loopEnd) LC1 = P5;
- I0 = P1;
- DISALGNEXCPT || r0 = [i0++];
-
-_pix_norm1_blkfn_loopStart:
- // following unpacks pix1[0..15] pix1+line_size[0..15]
- DISALGNEXCPT || r1 = [i0++];
-
- (r5, r4) = byteunpack r1:0 || r0 = [i0++];
- a1 += r5.h * r5.h, a0 += r5.l * r5.l (is);
- a1 += r4.h * r4.h, a0 += r4.l * r4.l (is);
- (r5, r4) = byteunpack r1:0(r) || r1 = [i0++];
- a1 += r5.h * r5.h, a0 += r5.l * r5.l (is);
- a1 += r4.h * r4.h, a0 += r4.l * r4.l (is);
- (r5, r4) = byteunpack r1:0 || r0 = [i0++M0];
- a1 += r5.h * r5.h, a0 += r5.l * r5.l (is);
- a1 += r4.h * r4.h, a0 += r4.l * r4.l (is);
- (r5, r4) = byteunpack r1:0(r) || r0 = [i0++];
- a1 += r5.h * r5.h, a0 += r5.l * r5.l (is);
-_pix_norm1_blkfn_loopEnd:
- a1 += r4.h * r4.h, a0 += r4.l * r4.l (is);
-
-
-// Clean up at the end:
- R2 = A0, R3 = A1;
- R0 = R2 + R3 (S);
-
- (R7:4,P5:3)=[SP++];
-
- RTS;
-DEFUN_END(pix_norm1)
-
DEFUN(sse4,mL1,
(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)):
link 0;
diff --git a/libavcodec/bfin/dsputil_init.c b/libavcodec/bfin/dsputil_init.c
index 5c33d9c..aa14a81 100644
--- a/libavcodec/bfin/dsputil_init.c
+++ b/libavcodec/bfin/dsputil_init.c
@@ -37,9 +37,6 @@ void ff_bfin_diff_pixels(int16_t *block, const uint8_t *s1,
const uint8_t *s2,
void ff_bfin_get_pixels(int16_t *restrict block, const uint8_t *pixels,
int line_size) attribute_l1_text;
-int ff_bfin_pix_norm1(uint8_t *pix, int line_size) attribute_l1_text;
-int ff_bfin_pix_sum(uint8_t *p, int stride) attribute_l1_text;
-
int ff_bfin_z_sad8x8(uint8_t *blk1, uint8_t *blk2, int dsz,
int line_size, int h) attribute_l1_text;
int ff_bfin_z_sad16x16(uint8_t *blk1, uint8_t *blk2, int dsz,
@@ -123,9 +120,6 @@ av_cold void ff_dsputil_init_bfin(DSPContext *c,
AVCodecContext *avctx,
{
c->diff_pixels = ff_bfin_diff_pixels;
- c->pix_sum = ff_bfin_pix_sum;
- c->pix_norm1 = ff_bfin_pix_norm1;
-
c->sad[0] = bfin_pix_abs16;
c->sad[1] = bfin_pix_abs8;
diff --git a/libavcodec/bfin/mpegvideoencdsp.S
b/libavcodec/bfin/mpegvideoencdsp.S
new file mode 100644
index 0000000..83b1f18
--- /dev/null
+++ b/libavcodec/bfin/mpegvideoencdsp.S
@@ -0,0 +1,113 @@
+/*
+ * Blackfin Pixel Operations
+ * Copyright (C) 2007 Marc Hoffman <[email protected]>
+ *
+ * This file is part of Libav.
+ *
+ * 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.
+ *
+ * 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 Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "libavutil/bfin/asm.h"
+
+/*
+ for (i = 0; i < 16; i++) {
+ for (j = 0; j < 16; j++) {
+ sum += pix[j];
+ }
+ pix += line_size;
+ }
+*/
+DEFUN(pix_sum,mL1,
+ (uint8_t *p, int stride)):
+ link 0;
+ [--sp] = (r7:4);
+ p0=8;
+ i0=r0; // s1
+ i1=r0;
+ m1=r1;
+ r1=r1+r1;
+ r1+=-16; // stride
+ m0=r1;
+ i1+=m1;
+
+ r6=0;
+
+ LSETUP(LS$PS,LE$PS) LC0=P0;
+ DISALGNEXCPT || R0 = [I0++] || R2 =[I1++];
+
+LS$PS: DISALGNEXCPT || R1 = [I0++] || R3 = [I1++];
+ (R5,R4) = BYTEOP16P (R3:2,R1:0) || R0 = [I0++] || R2 = [I1++];
+ r6=r6+|+r5;
+ r6=r6+|+r4;
+ (R5,R4) = BYTEOP16P (R3:2,R1:0) (R)|| R1 = [I0++] || R3 = [I1++];
+ r6=r6+|+r5;
+ r6=r6+|+r4;
+ (R5,R4) = BYTEOP16P (R3:2,R1:0) || R0 = [I0++m0] || R2 = [I1++m0];
+ r6=r6+|+r5;
+ r6=r6+|+r4;
+ (R5,R4) = BYTEOP16P (R3:2,R1:0) (R)|| R0 = [I0++] || R2 = [I1++];
+ r6=r6+|+r5;
+LE$PS: r6=r6+|+r4;
+ r0.l=r6.l+r6.h;
+ r0.h=0;
+
+ (r7:4) = [sp++];
+ unlink;
+ rts;
+DEFUN_END(pix_sum)
+
+DEFUN(pix_norm1,mL1,
+ (uint8_t * pix, int line_size)):
+ [--SP]=(R7:4,P5:3);
+
+ // Fetch the input arguments.
+ P1 = R0; // pix
+ P0 = R1; // line_size
+ P5 = 16; // loop ctr.
+ P0 -= P5;
+ M0 = P0; // M0 = line_size-16;
+ // Now for the real work.
+ A1 = A0 = 0;
+ lsetup(_pix_norm1_blkfn_loopStart, _pix_norm1_blkfn_loopEnd) LC1 = P5;
+ I0 = P1;
+ DISALGNEXCPT || r0 = [i0++];
+
+_pix_norm1_blkfn_loopStart:
+ // following unpacks pix1[0..15] pix1+line_size[0..15]
+ DISALGNEXCPT || r1 = [i0++];
+
+ (r5, r4) = byteunpack r1:0 || r0 = [i0++];
+ a1 += r5.h * r5.h, a0 += r5.l * r5.l (is);
+ a1 += r4.h * r4.h, a0 += r4.l * r4.l (is);
+ (r5, r4) = byteunpack r1:0(r) || r1 = [i0++];
+ a1 += r5.h * r5.h, a0 += r5.l * r5.l (is);
+ a1 += r4.h * r4.h, a0 += r4.l * r4.l (is);
+ (r5, r4) = byteunpack r1:0 || r0 = [i0++M0];
+ a1 += r5.h * r5.h, a0 += r5.l * r5.l (is);
+ a1 += r4.h * r4.h, a0 += r4.l * r4.l (is);
+ (r5, r4) = byteunpack r1:0(r) || r0 = [i0++];
+ a1 += r5.h * r5.h, a0 += r5.l * r5.l (is);
+_pix_norm1_blkfn_loopEnd:
+ a1 += r4.h * r4.h, a0 += r4.l * r4.l (is);
+
+
+// Clean up at the end:
+ R2 = A0, R3 = A1;
+ R0 = R2 + R3 (S);
+
+ (R7:4,P5:3)=[SP++];
+
+ RTS;
+DEFUN_END(pix_norm1)
diff --git a/libavcodec/mpegvideoencdsp.h
b/libavcodec/bfin/mpegvideoencdsp_init.c
similarity index 52%
copy from libavcodec/mpegvideoencdsp.h
copy to libavcodec/bfin/mpegvideoencdsp_init.c
index c321698..d459ada 100644
--- a/libavcodec/mpegvideoencdsp.h
+++ b/libavcodec/bfin/mpegvideoencdsp_init.c
@@ -16,27 +16,17 @@
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
-#ifndef AVCODEC_MPEGVIDEOENCDSP_H
-#define AVCODEC_MPEGVIDEOENCDSP_H
-
-#include <stdint.h>
-
-#include "avcodec.h"
-
-#define BASIS_SHIFT 16
-#define RECON_SHIFT 6
-
-typedef struct MpegvideoEncDSPContext {
- int (*try_8x8basis)(int16_t rem[64], int16_t weight[64],
- int16_t basis[64], int scale);
- void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
-
-} MpegvideoEncDSPContext;
-
-void ff_mpegvideoencdsp_init(MpegvideoEncDSPContext *c,
- AVCodecContext *avctx);
-void ff_mpegvideoencdsp_init_x86(MpegvideoEncDSPContext *c,
- AVCodecContext *avctx);
-
-
-#endif /* AVCODEC_MPEGVIDEOENCDSP_H */
+#include "libavutil/attributes.h"
+#include "libavutil/bfin/attributes.h"
+#include "libavcodec/avcodec.h"
+#include "libavcodec/mpegvideoencdsp.h"
+
+int ff_bfin_pix_norm1(uint8_t *pix, int line_size) attribute_l1_text;
+int ff_bfin_pix_sum(uint8_t *p, int stride) attribute_l1_text;
+
+av_cold void ff_mpegvideoencdsp_init_bfin(MpegvideoEncDSPContext *c,
+ AVCodecContext *avctx)
+{
+ c->pix_norm1 = ff_bfin_pix_norm1;
+ c->pix_sum = ff_bfin_pix_sum;
+}
diff --git a/libavcodec/dnxhdenc.c b/libavcodec/dnxhdenc.c
index b5b7f55..0e46ac0 100644
--- a/libavcodec/dnxhdenc.c
+++ b/libavcodec/dnxhdenc.c
@@ -266,6 +266,7 @@ static av_cold int dnxhd_encode_init(AVCodecContext *avctx)
ff_blockdsp_init(&ctx->bdsp, avctx);
ff_dsputil_init(&ctx->m.dsp, avctx);
ff_idctdsp_init(&ctx->m.idsp, avctx);
+ ff_mpegvideoencdsp_init(&ctx->m.mvedsp, avctx);
ff_dct_common_init(&ctx->m);
if (!ctx->m.dct_quantize)
ctx->m.dct_quantize = ff_dct_quantize_c;
@@ -636,8 +637,8 @@ static int dnxhd_mb_var_thread(AVCodecContext *avctx, void
*arg, int jobnr, int
int varc;
if (!partial_last_row && mb_x * 16 <= avctx->width - 16) {
- sum = ctx->m.dsp.pix_sum(pix, ctx->m.linesize);
- varc = ctx->m.dsp.pix_norm1(pix, ctx->m.linesize);
+ sum = ctx->m.mvedsp.pix_sum(pix, ctx->m.linesize);
+ varc = ctx->m.mvedsp.pix_norm1(pix, ctx->m.linesize);
} else {
int bw = FFMIN(avctx->width - 16 * mb_x, 16);
int bh = FFMIN((avctx->height >> ctx->interlaced) - 16 * mb_y,
16);
diff --git a/libavcodec/dsputil.c b/libavcodec/dsputil.c
index 1fd9013..81dce9f 100644
--- a/libavcodec/dsputil.c
+++ b/libavcodec/dsputil.c
@@ -26,15 +26,12 @@
*/
#include "libavutil/attributes.h"
-#include "libavutil/imgutils.h"
#include "avcodec.h"
#include "copy_block.h"
#include "dct.h"
#include "dsputil.h"
#include "simple_idct.h"
#include "faandct.h"
-#include "imgconvert.h"
-#include "mathops.h"
#include "mpegvideo.h"
#include "config.h"
@@ -47,74 +44,6 @@ uint32_t ff_square_tab[512] = { 0, };
#define BIT_DEPTH 8
#include "dsputilenc_template.c"
-static int pix_sum_c(uint8_t *pix, int line_size)
-{
- int s = 0, i, j;
-
- for (i = 0; i < 16; i++) {
- for (j = 0; j < 16; j += 8) {
- s += pix[0];
- s += pix[1];
- s += pix[2];
- s += pix[3];
- s += pix[4];
- s += pix[5];
- s += pix[6];
- s += pix[7];
- pix += 8;
- }
- pix += line_size - 16;
- }
- return s;
-}
-
-static int pix_norm1_c(uint8_t *pix, int line_size)
-{
- int s = 0, i, j;
- uint32_t *sq = ff_square_tab + 256;
-
- for (i = 0; i < 16; i++) {
- for (j = 0; j < 16; j += 8) {
-#if 0
- s += sq[pix[0]];
- s += sq[pix[1]];
- s += sq[pix[2]];
- s += sq[pix[3]];
- s += sq[pix[4]];
- s += sq[pix[5]];
- s += sq[pix[6]];
- s += sq[pix[7]];
-#else
-#if HAVE_FAST_64BIT
- register uint64_t x = *(uint64_t *) pix;
- s += sq[x & 0xff];
- s += sq[(x >> 8) & 0xff];
- s += sq[(x >> 16) & 0xff];
- s += sq[(x >> 24) & 0xff];
- s += sq[(x >> 32) & 0xff];
- s += sq[(x >> 40) & 0xff];
- s += sq[(x >> 48) & 0xff];
- s += sq[(x >> 56) & 0xff];
-#else
- register uint32_t x = *(uint32_t *) pix;
- s += sq[x & 0xff];
- s += sq[(x >> 8) & 0xff];
- s += sq[(x >> 16) & 0xff];
- s += sq[(x >> 24) & 0xff];
- x = *(uint32_t *) (pix + 4);
- s += sq[x & 0xff];
- s += sq[(x >> 8) & 0xff];
- s += sq[(x >> 16) & 0xff];
- s += sq[(x >> 24) & 0xff];
-#endif
-#endif
- pix += 8;
- }
- pix += line_size - 16;
- }
- return s;
-}
-
static int sse4_c(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,
int line_size, int h)
{
@@ -1061,9 +990,6 @@ av_cold void ff_dsputil_init(DSPContext *c, AVCodecContext
*avctx)
c->sum_abs_dctelem = sum_abs_dctelem_c;
- c->pix_sum = pix_sum_c;
- c->pix_norm1 = pix_norm1_c;
-
/* TODO [0] 16 [1] 8 */
c->pix_abs[0][0] = pix_abs16_c;
c->pix_abs[0][1] = pix_abs16_x2_c;
@@ -1103,11 +1029,6 @@ av_cold void ff_dsputil_init(DSPContext *c,
AVCodecContext *avctx)
c->nsse[0] = nsse16_c;
c->nsse[1] = nsse8_c;
- c->shrink[0] = av_image_copy_plane;
- c->shrink[1] = ff_shrink22;
- c->shrink[2] = ff_shrink44;
- c->shrink[3] = ff_shrink88;
-
c->draw_edges = draw_edges_8_c;
switch (avctx->bits_per_raw_sample) {
diff --git a/libavcodec/dsputil.h b/libavcodec/dsputil.h
index 27a9031..b4f6834 100644
--- a/libavcodec/dsputil.h
+++ b/libavcodec/dsputil.h
@@ -58,9 +58,6 @@ typedef struct DSPContext {
int stride);
int (*sum_abs_dctelem)(int16_t *block /* align 16 */);
- int (*pix_sum)(uint8_t *pix, int line_size);
- int (*pix_norm1)(uint8_t *pix, int line_size);
-
me_cmp_func sad[6]; /* identical to pix_absAxA except additional void * */
me_cmp_func sse[6];
me_cmp_func hadamard8_diff[6];
@@ -92,9 +89,6 @@ typedef struct DSPContext {
#define EDGE_WIDTH 16
#define EDGE_TOP 1
#define EDGE_BOTTOM 2
-
- void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src,
- int src_wrap, int width, int height);
} DSPContext;
void ff_dsputil_init(DSPContext *p, AVCodecContext *avctx);
diff --git a/libavcodec/motion_est.c b/libavcodec/motion_est.c
index 2f4655a..7e6aed1 100644
--- a/libavcodec/motion_est.c
+++ b/libavcodec/motion_est.c
@@ -880,8 +880,9 @@ void ff_estimate_p_frame_motion(MpegEncContext * s,
/* intra / predictive decision */
pix = c->src[0][0];
- sum = s->dsp.pix_sum(pix, s->linesize);
- varc = s->dsp.pix_norm1(pix, s->linesize) - (((unsigned)sum*sum)>>8) + 500;
+ sum = s->mvedsp.pix_sum(pix, s->linesize);
+ varc = s->mvedsp.pix_norm1(pix, s->linesize) -
+ (((unsigned) sum * sum) >> 8) + 500;
pic->mb_mean[s->mb_stride * mb_y + mb_x] = (sum+128)>>8;
pic->mb_var [s->mb_stride * mb_y + mb_x] = (varc+128)>>8;
diff --git a/libavcodec/mpegvideo_enc.c b/libavcodec/mpegvideo_enc.c
index e2e0fc9..4b935be 100644
--- a/libavcodec/mpegvideo_enc.c
+++ b/libavcodec/mpegvideo_enc.c
@@ -869,7 +869,7 @@ static int get_intra_count(MpegEncContext *s, uint8_t *src,
int offset = x + y * stride;
int sad = s->dsp.sad[0](NULL, src + offset, ref + offset, stride,
16);
- int mean = (s->dsp.pix_sum(src + offset, stride) + 128) >> 8;
+ int mean = (s->mvedsp.pix_sum(src + offset, stride) + 128) >> 8;
int sae = get_sae(src + offset, mean, stride);
acc += sae + 500 < sad;
@@ -1113,15 +1113,21 @@ static int estimate_best_b_count(MpegEncContext *s)
pre_input.f.data[2] += INPLACE_OFFSET;
}
- s->dsp.shrink[scale](s->tmp_frames[i]->data[0],
s->tmp_frames[i]->linesize[0],
- pre_input.f.data[0], pre_input.f.linesize[0],
- c->width, c->height);
- s->dsp.shrink[scale](s->tmp_frames[i]->data[1],
s->tmp_frames[i]->linesize[1],
- pre_input.f.data[1], pre_input.f.linesize[1],
- c->width >> 1, c->height >> 1);
- s->dsp.shrink[scale](s->tmp_frames[i]->data[2],
s->tmp_frames[i]->linesize[2],
- pre_input.f.data[2], pre_input.f.linesize[2],
- c->width >> 1, c->height >> 1);
+ s->mvedsp.shrink[scale](s->tmp_frames[i]->data[0],
+ s->tmp_frames[i]->linesize[0],
+ pre_input.f.data[0],
+ pre_input.f.linesize[0],
+ c->width, c->height);
+ s->mvedsp.shrink[scale](s->tmp_frames[i]->data[1],
+ s->tmp_frames[i]->linesize[1],
+ pre_input.f.data[1],
+ pre_input.f.linesize[1],
+ c->width >> 1, c->height >> 1);
+ s->mvedsp.shrink[scale](s->tmp_frames[i]->data[2],
+ s->tmp_frames[i]->linesize[2],
+ pre_input.f.data[2],
+ pre_input.f.linesize[2],
+ c->width >> 1, c->height >> 1);
}
}
@@ -2395,9 +2401,10 @@ static int mb_var_thread(AVCodecContext *c, void *arg){
int yy = mb_y * 16;
uint8_t *pix = s->new_picture.f.data[0] + (yy * s->linesize) + xx;
int varc;
- int sum = s->dsp.pix_sum(pix, s->linesize);
+ int sum = s->mvedsp.pix_sum(pix, s->linesize);
- varc = (s->dsp.pix_norm1(pix, s->linesize) -
(((unsigned)sum*sum)>>8) + 500 + 128)>>8;
+ varc = (s->mvedsp.pix_norm1(pix, s->linesize) -
+ (((unsigned) sum * sum) >> 8) + 500 + 128) >> 8;
s->current_picture.mb_var [s->mb_stride * mb_y + mb_x] = varc;
s->current_picture.mb_mean[s->mb_stride * mb_y + mb_x] =
(sum+128)>>8;
diff --git a/libavcodec/mpegvideoencdsp.c b/libavcodec/mpegvideoencdsp.c
index 16ff1f2..d827186 100644
--- a/libavcodec/mpegvideoencdsp.c
+++ b/libavcodec/mpegvideoencdsp.c
@@ -21,7 +21,10 @@
#include "config.h"
#include "libavutil/attributes.h"
+#include "libavutil/imgutils.h"
#include "avcodec.h"
+#include "dsputil.h"
+#include "imgconvert.h"
#include "mpegvideoencdsp.h"
static int try_8x8basis_c(int16_t rem[64], int16_t weight[64],
@@ -53,12 +56,94 @@ static void add_8x8basis_c(int16_t rem[64], int16_t
basis[64], int scale)
(BASIS_SHIFT - RECON_SHIFT);
}
+static int pix_sum_c(uint8_t *pix, int line_size)
+{
+ int s = 0, i, j;
+
+ for (i = 0; i < 16; i++) {
+ for (j = 0; j < 16; j += 8) {
+ s += pix[0];
+ s += pix[1];
+ s += pix[2];
+ s += pix[3];
+ s += pix[4];
+ s += pix[5];
+ s += pix[6];
+ s += pix[7];
+ pix += 8;
+ }
+ pix += line_size - 16;
+ }
+ return s;
+}
+
+static int pix_norm1_c(uint8_t *pix, int line_size)
+{
+ int s = 0, i, j;
+ uint32_t *sq = ff_square_tab + 256;
+
+ for (i = 0; i < 16; i++) {
+ for (j = 0; j < 16; j += 8) {
+#if 0
+ s += sq[pix[0]];
+ s += sq[pix[1]];
+ s += sq[pix[2]];
+ s += sq[pix[3]];
+ s += sq[pix[4]];
+ s += sq[pix[5]];
+ s += sq[pix[6]];
+ s += sq[pix[7]];
+#else
+#if HAVE_FAST_64BIT
+ register uint64_t x = *(uint64_t *) pix;
+ s += sq[x & 0xff];
+ s += sq[(x >> 8) & 0xff];
+ s += sq[(x >> 16) & 0xff];
+ s += sq[(x >> 24) & 0xff];
+ s += sq[(x >> 32) & 0xff];
+ s += sq[(x >> 40) & 0xff];
+ s += sq[(x >> 48) & 0xff];
+ s += sq[(x >> 56) & 0xff];
+#else
+ register uint32_t x = *(uint32_t *) pix;
+ s += sq[x & 0xff];
+ s += sq[(x >> 8) & 0xff];
+ s += sq[(x >> 16) & 0xff];
+ s += sq[(x >> 24) & 0xff];
+ x = *(uint32_t *) (pix + 4);
+ s += sq[x & 0xff];
+ s += sq[(x >> 8) & 0xff];
+ s += sq[(x >> 16) & 0xff];
+ s += sq[(x >> 24) & 0xff];
+#endif
+#endif
+ pix += 8;
+ }
+ pix += line_size - 16;
+ }
+ return s;
+}
+
av_cold void ff_mpegvideoencdsp_init(MpegvideoEncDSPContext *c,
AVCodecContext *avctx)
{
c->try_8x8basis = try_8x8basis_c;
c->add_8x8basis = add_8x8basis_c;
+ c->shrink[0] = av_image_copy_plane;
+ c->shrink[1] = ff_shrink22;
+ c->shrink[2] = ff_shrink44;
+ c->shrink[3] = ff_shrink88;
+
+ c->pix_sum = pix_sum_c;
+ c->pix_norm1 = pix_norm1_c;
+
+ if (ARCH_ARM)
+ ff_mpegvideoencdsp_init_arm(c, avctx);
+ if (ARCH_BFIN)
+ ff_mpegvideoencdsp_init_bfin(c, avctx);
+ if (ARCH_PPC)
+ ff_mpegvideoencdsp_init_ppc(c, avctx);
if (ARCH_X86)
ff_mpegvideoencdsp_init_x86(c, avctx);
}
diff --git a/libavcodec/mpegvideoencdsp.h b/libavcodec/mpegvideoencdsp.h
index c321698..4e144ed 100644
--- a/libavcodec/mpegvideoencdsp.h
+++ b/libavcodec/mpegvideoencdsp.h
@@ -31,12 +31,22 @@ typedef struct MpegvideoEncDSPContext {
int16_t basis[64], int scale);
void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
+ int (*pix_sum)(uint8_t *pix, int line_size);
+ int (*pix_norm1)(uint8_t *pix, int line_size);
+
+ void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src,
+ int src_wrap, int width, int height);
} MpegvideoEncDSPContext;
void ff_mpegvideoencdsp_init(MpegvideoEncDSPContext *c,
AVCodecContext *avctx);
+void ff_mpegvideoencdsp_init_arm(MpegvideoEncDSPContext *c,
+ AVCodecContext *avctx);
+void ff_mpegvideoencdsp_init_bfin(MpegvideoEncDSPContext *c,
+ AVCodecContext *avctx);
+void ff_mpegvideoencdsp_init_ppc(MpegvideoEncDSPContext *c,
+ AVCodecContext *avctx);
void ff_mpegvideoencdsp_init_x86(MpegvideoEncDSPContext *c,
AVCodecContext *avctx);
-
#endif /* AVCODEC_MPEGVIDEOENCDSP_H */
diff --git a/libavcodec/ppc/Makefile b/libavcodec/ppc/Makefile
index 952d756..607a334 100644
--- a/libavcodec/ppc/Makefile
+++ b/libavcodec/ppc/Makefile
@@ -12,6 +12,7 @@ OBJS-$(CONFIG_IDCTDSP) +=
ppc/idctdsp_altivec.o
OBJS-$(CONFIG_MPEGAUDIODSP) += ppc/mpegaudiodsp_altivec.o
OBJS-$(CONFIG_MPEGVIDEO) += ppc/mpegvideo_altivec.o \
ppc/mpegvideodsp_altivec.o
+OBJS-$(CONFIG_MPEGVIDEOENC) += ppc/mpegvideoencdsp_altivec.o
OBJS-$(CONFIG_VECTORDSP) += ppc/vectordsp_altivec.o
OBJS-$(CONFIG_VIDEODSP) += ppc/videodsp_ppc.o
OBJS-$(CONFIG_VP3DSP) += ppc/vp3dsp_altivec.o
diff --git a/libavcodec/ppc/dsputil_altivec.c b/libavcodec/ppc/dsputil_altivec.c
index c3f90e9..efb4a18 100644
--- a/libavcodec/ppc/dsputil_altivec.c
+++ b/libavcodec/ppc/dsputil_altivec.c
@@ -308,34 +308,6 @@ static int sad8_altivec(MpegEncContext *v, uint8_t *pix1,
uint8_t *pix2,
return s;
}
-static int pix_norm1_altivec(uint8_t *pix, int line_size)
-{
- int i, s = 0;
- const vector unsigned int zero =
- (const vector unsigned int) vec_splat_u32(0);
- vector unsigned char perm = vec_lvsl(0, pix);
- vector unsigned int sv = (vector unsigned int) vec_splat_u32(0);
- vector signed int sum;
-
- for (i = 0; i < 16; i++) {
- /* Read the potentially unaligned pixels. */
- vector unsigned char pixl = vec_ld(0, pix);
- vector unsigned char pixr = vec_ld(15, pix);
- vector unsigned char pixv = vec_perm(pixl, pixr, perm);
-
- /* Square the values, and add them to our sum. */
- sv = vec_msum(pixv, pixv, sv);
-
- pix += line_size;
- }
- /* Sum up the four partial sums, and put the result into s. */
- sum = vec_sums((vector signed int) sv, (vector signed int) zero);
- sum = vec_splat(sum, 3);
- vec_ste(sum, 0, &s);
-
- return s;
-}
-
/* Sum of Squared Errors for an 8x8 block, AltiVec-enhanced.
* It's the sad8_altivec code above w/ squaring added. */
static int sse8_altivec(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,
@@ -430,35 +402,6 @@ static int sse16_altivec(MpegEncContext *v, uint8_t *pix1,
uint8_t *pix2,
return s;
}
-static int pix_sum_altivec(uint8_t *pix, int line_size)
-{
- int i, s;
- const vector unsigned int zero =
- (const vector unsigned int) vec_splat_u32(0);
- vector unsigned char perm = vec_lvsl(0, pix);
- vector unsigned int sad = (vector unsigned int) vec_splat_u32(0);
- vector signed int sumdiffs;
-
- 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 pixr = vec_ld(15, pix);
- vector unsigned char t1 = vec_perm(pixl, pixr, perm);
-
- /* Add each 4 pixel group together and put 4 results into sad. */
- sad = vec_sum4s(t1, sad);
-
- pix += line_size;
- }
-
- /* Sum up the four partial sums, and put the result into s. */
- sumdiffs = vec_sums((vector signed int) sad, (vector signed int) zero);
- sumdiffs = vec_splat(sumdiffs, 3);
- vec_ste(sumdiffs, 0, &s);
-
- return s;
-}
-
static void get_pixels_altivec(int16_t *restrict block, const uint8_t *pixels,
int line_size)
{
@@ -911,9 +854,6 @@ av_cold void ff_dsputil_init_altivec(DSPContext *c,
AVCodecContext *avctx,
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;
if (!high_bit_depth) {
diff --git a/libavcodec/ppc/mpegvideoencdsp_altivec.c
b/libavcodec/ppc/mpegvideoencdsp_altivec.c
new file mode 100644
index 0000000..398f3a9
--- /dev/null
+++ b/libavcodec/ppc/mpegvideoencdsp_altivec.c
@@ -0,0 +1,102 @@
+/*
+ * This file is part of Libav.
+ *
+ * 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.
+ *
+ * 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 Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "config.h"
+#include <stdint.h>
+#if HAVE_ALTIVEC_H
+#include <altivec.h>
+#endif
+
+#include "libavutil/attributes.h"
+#include "libavutil/cpu.h"
+#include "libavutil/ppc/types_altivec.h"
+#include "libavutil/ppc/util_altivec.h"
+#include "libavcodec/mpegvideoencdsp.h"
+
+#if HAVE_ALTIVEC
+
+static int pix_norm1_altivec(uint8_t *pix, int line_size)
+{
+ int i, s = 0;
+ const vector unsigned int zero =
+ (const vector unsigned int) vec_splat_u32(0);
+ vector unsigned char perm = vec_lvsl(0, pix);
+ vector unsigned int sv = (vector unsigned int) vec_splat_u32(0);
+ vector signed int sum;
+
+ for (i = 0; i < 16; i++) {
+ /* Read the potentially unaligned pixels. */
+ vector unsigned char pixl = vec_ld(0, pix);
+ vector unsigned char pixr = vec_ld(15, pix);
+ vector unsigned char pixv = vec_perm(pixl, pixr, perm);
+
+ /* Square the values, and add them to our sum. */
+ sv = vec_msum(pixv, pixv, sv);
+
+ pix += line_size;
+ }
+ /* Sum up the four partial sums, and put the result into s. */
+ sum = vec_sums((vector signed int) sv, (vector signed int) zero);
+ sum = vec_splat(sum, 3);
+ vec_ste(sum, 0, &s);
+
+ return s;
+}
+
+static int pix_sum_altivec(uint8_t *pix, int line_size)
+{
+ int i, s;
+ const vector unsigned int zero =
+ (const vector unsigned int) vec_splat_u32(0);
+ vector unsigned char perm = vec_lvsl(0, pix);
+ vector unsigned int sad = (vector unsigned int) vec_splat_u32(0);
+ vector signed int sumdiffs;
+
+ 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 pixr = vec_ld(15, pix);
+ vector unsigned char t1 = vec_perm(pixl, pixr, perm);
+
+ /* Add each 4 pixel group together and put 4 results into sad. */
+ sad = vec_sum4s(t1, sad);
+
+ pix += line_size;
+ }
+
+ /* Sum up the four partial sums, and put the result into s. */
+ sumdiffs = vec_sums((vector signed int) sad, (vector signed int) zero);
+ sumdiffs = vec_splat(sumdiffs, 3);
+ vec_ste(sumdiffs, 0, &s);
+
+ return s;
+}
+
+#endif /* HAVE_ALTIVEC */
+
+av_cold void ff_mpegvideoencdsp_init_ppc(MpegvideoEncDSPContext *c,
+ AVCodecContext *avctx)
+{
+#if HAVE_ALTIVEC
+ if (!(av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC))
+ return;
+
+ c->pix_norm1 = pix_norm1_altivec;
+ c->pix_sum = pix_sum_altivec;
+#endif /* HAVE_ALTIVEC */
+}
diff --git a/libavcodec/svq1enc.c b/libavcodec/svq1enc.c
index fd1ff27..68e1641 100644
--- a/libavcodec/svq1enc.c
+++ b/libavcodec/svq1enc.c
@@ -506,6 +506,7 @@ static av_cold int svq1_encode_init(AVCodecContext *avctx)
ff_dsputil_init(&s->dsp, avctx);
ff_hpeldsp_init(&s->hdsp, avctx->flags);
+ ff_mpegvideoencdsp_init(&s->m.mvedsp, avctx);
avctx->coded_frame = av_frame_alloc();
s->current_picture = av_frame_alloc();
diff --git a/libavcodec/x86/Makefile b/libavcodec/x86/Makefile
index 56a4e56..54e25d6 100644
--- a/libavcodec/x86/Makefile
+++ b/libavcodec/x86/Makefile
@@ -23,7 +23,7 @@ OBJS-$(CONFIG_MPEGAUDIODSP) += x86/mpegaudiodsp.o
OBJS-$(CONFIG_MPEGVIDEO) += x86/mpegvideo.o \
x86/mpegvideodsp.o
OBJS-$(CONFIG_MPEGVIDEOENC) += x86/mpegvideoenc.o \
- x86/mpegvideoencdsp.o
+ x86/mpegvideoencdsp_init.o
OBJS-$(CONFIG_QPELDSP) += x86/qpeldsp_init.o
OBJS-$(CONFIG_VECTORDSP) += x86/vectordsp_init.o
OBJS-$(CONFIG_VIDEODSP) += x86/videodsp_init.o
@@ -90,6 +90,8 @@ YASM-OBJS-$(CONFIG_HPELDSP) += x86/fpel.o
\
x86/hpeldsp.o
YASM-OBJS-$(CONFIG_HUFFYUVDSP) += x86/huffyuvdsp.o
YASM-OBJS-$(CONFIG_MPEGAUDIODSP) += x86/imdct36.o
+YASM-OBJS-$(CONFIG_MPEGVIDEOENC) += x86/mpegvideoenc.o \
+ x86/mpegvideoencdsp.o
YASM-OBJS-$(CONFIG_QPELDSP) += x86/qpeldsp.o \
x86/fpel.o \
x86/qpel.o
diff --git a/libavcodec/x86/dsputilenc.asm b/libavcodec/x86/dsputilenc.asm
index f676ba6..d996962 100644
--- a/libavcodec/x86/dsputilenc.asm
+++ b/libavcodec/x86/dsputilenc.asm
@@ -418,72 +418,3 @@ cglobal diff_pixels, 4,5
add r4, 16
jne .loop
REP_RET
-
-INIT_MMX mmx
-; int ff_pix_sum16_mmx(uint8_t *pix, int line_size)
-cglobal pix_sum16, 2, 3
- movsxdifnidn r1, r1d
- mov r2, r1
- neg r2
- shl r2, 4
- sub r0, r2
- pxor m7, m7
- pxor m6, m6
-.loop:
- mova m0, [r0+r2+0]
- mova m1, [r0+r2+0]
- mova m2, [r0+r2+8]
- mova m3, [r0+r2+8]
- punpcklbw m0, m7
- punpckhbw m1, m7
- punpcklbw m2, m7
- punpckhbw m3, m7
- paddw m1, m0
- paddw m3, m2
- paddw m3, m1
- paddw m6, m3
- add r2, r1
- js .loop
- mova m5, m6
- psrlq m6, 32
- paddw m6, m5
- mova m5, m6
- psrlq m6, 16
- paddw m6, m5
- movd eax, m6
- and eax, 0xffff
- RET
-
-INIT_MMX mmx
-; int ff_pix_norm1_mmx(uint8_t *pix, int line_size)
-cglobal pix_norm1, 2, 4
- movsxdifnidn r1, r1d
- mov r2, 16
- pxor m0, m0
- pxor m7, m7
-.loop:
- mova m2, [r0+0]
- mova m3, [r0+8]
- mova m1, m2
- punpckhbw m1, m0
- punpcklbw m2, m0
- mova m4, m3
- punpckhbw m3, m0
- punpcklbw m4, m0
- pmaddwd m1, m1
- pmaddwd m2, m2
- pmaddwd m3, m3
- pmaddwd m4, m4
- paddd m2, m1
- paddd m4, m3
- paddd m7, m2
- add r0, r1
- paddd m7, m4
- dec r2
- jne .loop
- mova m1, m7
- psrlq m7, 32
- paddd m1, m7
- movd eax, m1
- RET
-
diff --git a/libavcodec/x86/dsputilenc_mmx.c b/libavcodec/x86/dsputilenc_mmx.c
index 7af0913..563543e 100644
--- a/libavcodec/x86/dsputilenc_mmx.c
+++ b/libavcodec/x86/dsputilenc_mmx.c
@@ -35,8 +35,6 @@ void ff_get_pixels_mmx(int16_t *block, const uint8_t *pixels,
int line_size);
void ff_get_pixels_sse2(int16_t *block, const uint8_t *pixels, int line_size);
void ff_diff_pixels_mmx(int16_t *block, const uint8_t *s1, const uint8_t *s2,
int stride);
-int ff_pix_sum16_mmx(uint8_t *pix, int line_size);
-int ff_pix_norm1_mmx(uint8_t *pix, int line_size);
#if HAVE_INLINE_ASM
@@ -831,8 +829,6 @@ av_cold void ff_dsputilenc_init_mmx(DSPContext *c,
AVCodecContext *avctx,
if (!high_bit_depth)
c->get_pixels = ff_get_pixels_mmx;
c->diff_pixels = ff_diff_pixels_mmx;
- c->pix_sum = ff_pix_sum16_mmx;
- c->pix_norm1 = ff_pix_norm1_mmx;
}
if (EXTERNAL_SSE2(cpu_flags))
diff --git a/libavcodec/x86/mpegvideoencdsp.asm
b/libavcodec/x86/mpegvideoencdsp.asm
new file mode 100644
index 0000000..9326ee7
--- /dev/null
+++ b/libavcodec/x86/mpegvideoencdsp.asm
@@ -0,0 +1,95 @@
+;*****************************************************************************
+;* SIMD-optimized MPEG encoding functions
+;*****************************************************************************
+;* Copyright (c) 2000, 2001 Fabrice Bellard
+;* Copyright (c) 2002-2004 Michael Niedermayer <[email protected]>
+;*
+;* This file is part of Libav.
+;*
+;* 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.
+;*
+;* 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 Libav; if not, write to the Free Software
+;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+;*****************************************************************************
+
+%include "libavutil/x86/x86util.asm"
+
+SECTION .text
+
+INIT_MMX mmx
+; int ff_pix_sum16_mmx(uint8_t *pix, int line_size)
+cglobal pix_sum16, 2, 3
+ movsxdifnidn r1, r1d
+ mov r2, r1
+ neg r2
+ shl r2, 4
+ sub r0, r2
+ pxor m7, m7
+ pxor m6, m6
+.loop:
+ mova m0, [r0+r2+0]
+ mova m1, [r0+r2+0]
+ mova m2, [r0+r2+8]
+ mova m3, [r0+r2+8]
+ punpcklbw m0, m7
+ punpckhbw m1, m7
+ punpcklbw m2, m7
+ punpckhbw m3, m7
+ paddw m1, m0
+ paddw m3, m2
+ paddw m3, m1
+ paddw m6, m3
+ add r2, r1
+ js .loop
+ mova m5, m6
+ psrlq m6, 32
+ paddw m6, m5
+ mova m5, m6
+ psrlq m6, 16
+ paddw m6, m5
+ movd eax, m6
+ and eax, 0xffff
+ RET
+
+INIT_MMX mmx
+; int ff_pix_norm1_mmx(uint8_t *pix, int line_size)
+cglobal pix_norm1, 2, 4
+ movsxdifnidn r1, r1d
+ mov r2, 16
+ pxor m0, m0
+ pxor m7, m7
+.loop:
+ mova m2, [r0+0]
+ mova m3, [r0+8]
+ mova m1, m2
+ punpckhbw m1, m0
+ punpcklbw m2, m0
+ mova m4, m3
+ punpckhbw m3, m0
+ punpcklbw m4, m0
+ pmaddwd m1, m1
+ pmaddwd m2, m2
+ pmaddwd m3, m3
+ pmaddwd m4, m4
+ paddd m2, m1
+ paddd m4, m3
+ paddd m7, m2
+ add r0, r1
+ paddd m7, m4
+ dec r2
+ jne .loop
+ mova m1, m7
+ psrlq m7, 32
+ paddd m1, m7
+ movd eax, m1
+ RET
+
diff --git a/libavcodec/x86/mpegvideoencdsp.c
b/libavcodec/x86/mpegvideoencdsp_init.c
similarity index 94%
rename from libavcodec/x86/mpegvideoencdsp.c
rename to libavcodec/x86/mpegvideoencdsp_init.c
index db2c37f..4ef2f34 100644
--- a/libavcodec/x86/mpegvideoencdsp.c
+++ b/libavcodec/x86/mpegvideoencdsp_init.c
@@ -22,6 +22,9 @@
#include "libavcodec/avcodec.h"
#include "libavcodec/mpegvideoencdsp.h"
+int ff_pix_sum16_mmx(uint8_t *pix, int line_size);
+int ff_pix_norm1_mmx(uint8_t *pix, int line_size);
+
#if HAVE_INLINE_ASM
#define PHADDD(a, t) \
@@ -95,9 +98,15 @@
av_cold void ff_mpegvideoencdsp_init_x86(MpegvideoEncDSPContext *c,
AVCodecContext *avctx)
{
-#if HAVE_INLINE_ASM
int cpu_flags = av_get_cpu_flags();
+ if (EXTERNAL_MMX(cpu_flags)) {
+ c->pix_sum = ff_pix_sum16_mmx;
+ c->pix_norm1 = ff_pix_norm1_mmx;
+ }
+
+#if HAVE_INLINE_ASM
+
if (INLINE_MMX(cpu_flags)) {
if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
c->try_8x8basis = try_8x8basis_mmx;
--
1.8.3.2
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel
