Makefile | 4 +
arm/dsputil_neon.c | 16 ++++
arm/dsputil_neon_s.S | 178 +++++++++++++++++++++++++++++++++++++------------
arm/simple_idct_neon.S | 17 ++++
arm/vp3dsp_neon.S | 94 +++++++++++++++++++++++++
5 files changed, 265 insertions(+), 44 deletions(-)
diff -Nurd ffmpeg.old/libavcodec/arm/dsputil_neon.c ffmpeg-0.5/libavcodec/arm/dsputil_neon.c
--- ffmpeg.old/libavcodec/arm/dsputil_neon.c 2009-01-31 00:13:19.000000000 +0100
+++ ffmpeg-0.5/libavcodec/arm/dsputil_neon.c 2009-05-30 11:27:54.000000000 +0200
@@ -41,6 +41,10 @@
void ff_avg_pixels16_neon(uint8_t *, const uint8_t *, int, int);
+void ff_add_pixels_clamped_neon(const DCTELEM *, uint8_t *, int);
+void ff_put_pixels_clamped_neon(const DCTELEM *, uint8_t *, int);
+void ff_put_signed_pixels_clamped_neon(const DCTELEM *, uint8_t *, int);
+
void ff_put_h264_qpel16_mc00_neon(uint8_t *, uint8_t *, int);
void ff_put_h264_qpel16_mc10_neon(uint8_t *, uint8_t *, int);
void ff_put_h264_qpel16_mc20_neon(uint8_t *, uint8_t *, int);
@@ -146,6 +150,9 @@
DCTELEM *block, int stride,
const uint8_t nnzc[6*8]);
+void ff_vp3_v_loop_filter_neon(uint8_t *, int, int *);
+void ff_vp3_h_loop_filter_neon(uint8_t *, int, int *);
+
void ff_vector_fmul_neon(float *dst, const float *src, int len);
void ff_vector_fmul_window_neon(float *dst, const float *src0,
const float *src1, const float *win,
@@ -176,6 +183,10 @@
c->avg_pixels_tab[0][0] = ff_avg_pixels16_neon;
+ c->add_pixels_clamped = ff_add_pixels_clamped_neon;
+ c->put_pixels_clamped = ff_put_pixels_clamped_neon;
+ c->put_signed_pixels_clamped = ff_put_signed_pixels_clamped_neon;
+
c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_neon;
c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_neon;
@@ -247,6 +258,11 @@
c->h264_idct_add16intra = ff_h264_idct_add16intra_neon;
c->h264_idct_add8 = ff_h264_idct_add8_neon;
+ if (CONFIG_VP3_DECODER || CONFIG_THEORA_DECODER) {
+ c->vp3_v_loop_filter = ff_vp3_v_loop_filter_neon;
+ c->vp3_h_loop_filter = ff_vp3_h_loop_filter_neon;
+ }
+
c->vector_fmul = ff_vector_fmul_neon;
c->vector_fmul_window = ff_vector_fmul_window_neon;
diff -Nurd ffmpeg.old/libavcodec/arm/dsputil_neon_s.S ffmpeg-0.5/libavcodec/arm/dsputil_neon_s.S
--- ffmpeg.old/libavcodec/arm/dsputil_neon_s.S 2009-01-31 00:13:19.000000000 +0100
+++ ffmpeg-0.5/libavcodec/arm/dsputil_neon_s.S 2009-05-30 11:27:54.000000000 +0200
@@ -38,13 +38,13 @@
pld [r1, r2]
pld [r1, r2, lsl #1]
.if \avg
- vld1.64 {d16,d17}, [ip], r2
+ vld1.64 {d16,d17}, [ip,:128], r2
vrhadd.u8 q0, q0, q8
- vld1.64 {d18,d19}, [ip], r2
+ vld1.64 {d18,d19}, [ip,:128], r2
vrhadd.u8 q1, q1, q9
- vld1.64 {d20,d21}, [ip], r2
+ vld1.64 {d20,d21}, [ip,:128], r2
vrhadd.u8 q2, q2, q10
- vld1.64 {d22,d23}, [ip], r2
+ vld1.64 {d22,d23}, [ip,:128], r2
vrhadd.u8 q3, q3, q11
.endif
subs r3, r3, #4
@@ -73,35 +73,29 @@
.endm
.macro pixels16_y2 vhadd=vrhadd.u8
- push {lr}
- add ip, r1, r2
- lsl lr, r2, #1
- vld1.64 {d0, d1}, [r1], lr
- vld1.64 {d2, d3}, [ip], lr
+ vld1.64 {d0, d1}, [r1], r2
+ vld1.64 {d2, d3}, [r1], r2
1: subs r3, r3, #2
\vhadd q2, q0, q1
- vld1.64 {d0, d1}, [r1], lr
+ vld1.64 {d0, d1}, [r1], r2
\vhadd q3, q0, q1
- vld1.64 {d2, d3}, [ip], lr
+ vld1.64 {d2, d3}, [r1], r2
pld [r1]
- pld [ip]
+ pld [r1, r2]
vst1.64 {d4, d5}, [r0,:128], r2
vst1.64 {d6, d7}, [r0,:128], r2
bne 1b
- pop {pc}
+ bx lr
.endm
.macro pixels16_xy2 vshrn=vrshrn.u16 no_rnd=0
- push {lr}
- lsl lr, r2, #1
- add ip, r1, r2
- vld1.64 {d0-d2}, [r1], lr
- vld1.64 {d4-d6}, [ip], lr
+ vld1.64 {d0-d2}, [r1], r2
+ vld1.64 {d4-d6}, [r1], r2
.if \no_rnd
vmov.i16 q13, #1
.endif
pld [r1]
- pld [ip]
+ pld [r1, r2]
vext.8 q1, q0, q1, #1
vext.8 q3, q2, q3, #1
vaddl.u8 q8, d0, d2
@@ -109,7 +103,7 @@
vaddl.u8 q9, d4, d6
vaddl.u8 q11, d5, d7
1: subs r3, r3, #2
- vld1.64 {d0-d2}, [r1], lr
+ vld1.64 {d0-d2}, [r1], r2
vadd.u16 q12, q8, q9
pld [r1]
.if \no_rnd
@@ -123,11 +117,11 @@
.endif
\vshrn d29, q1, #2
vaddl.u8 q8, d0, d30
- vld1.64 {d2-d4}, [ip], lr
+ vld1.64 {d2-d4}, [r1], r2
vaddl.u8 q10, d1, d31
vst1.64 {d28,d29}, [r0,:128], r2
vadd.u16 q12, q8, q9
- pld [ip]
+ pld [r1, r2]
.if \no_rnd
vadd.u16 q12, q12, q13
.endif
@@ -142,7 +136,7 @@
vaddl.u8 q11, d3, d5
vst1.64 {d30,d31}, [r0,:128], r2
bgt 1b
- pop {pc}
+ bx lr
.endm
.macro pixels8
@@ -180,41 +174,35 @@
.endm
.macro pixels8_y2 vhadd=vrhadd.u8
- push {lr}
- add ip, r1, r2
- lsl lr, r2, #1
- vld1.64 {d0}, [r1], lr
- vld1.64 {d1}, [ip], lr
+ vld1.64 {d0}, [r1], r2
+ vld1.64 {d1}, [r1], r2
1: subs r3, r3, #2
\vhadd d4, d0, d1
- vld1.64 {d0}, [r1], lr
+ vld1.64 {d0}, [r1], r2
\vhadd d5, d0, d1
- vld1.64 {d1}, [ip], lr
+ vld1.64 {d1}, [r1], r2
pld [r1]
- pld [ip]
+ pld [r1, r2]
vst1.64 {d4}, [r0,:64], r2
vst1.64 {d5}, [r0,:64], r2
bne 1b
- pop {pc}
+ bx lr
.endm
.macro pixels8_xy2 vshrn=vrshrn.u16 no_rnd=0
- push {lr}
- lsl lr, r2, #1
- add ip, r1, r2
- vld1.64 {d0, d1}, [r1], lr
- vld1.64 {d2, d3}, [ip], lr
+ vld1.64 {d0, d1}, [r1], r2
+ vld1.64 {d2, d3}, [r1], r2
.if \no_rnd
vmov.i16 q11, #1
.endif
pld [r1]
- pld [ip]
+ pld [r1, r2]
vext.8 d4, d0, d1, #1
vext.8 d6, d2, d3, #1
vaddl.u8 q8, d0, d4
vaddl.u8 q9, d2, d6
1: subs r3, r3, #2
- vld1.64 {d0, d1}, [r1], lr
+ vld1.64 {d0, d1}, [r1], r2
pld [r1]
vadd.u16 q10, q8, q9
vext.8 d4, d0, d1, #1
@@ -223,9 +211,9 @@
.endif
vaddl.u8 q8, d0, d4
\vshrn d5, q10, #2
- vld1.64 {d2, d3}, [ip], lr
+ vld1.64 {d2, d3}, [r1], r2
vadd.u16 q10, q8, q9
- pld [ip]
+ pld [r1, r2]
.if \no_rnd
vadd.u16 q10, q10, q11
.endif
@@ -235,7 +223,7 @@
vaddl.u8 q9, d2, d6
vst1.64 {d7}, [r0,:64], r2
bgt 1b
- pop {pc}
+ bx lr
.endm
.macro pixfunc pfx name suf rnd_op args:vararg
@@ -273,6 +261,112 @@
pixfunc2 put_ pixels8_y2, _no_rnd, vhadd.u8
pixfunc2 put_ pixels8_xy2, _no_rnd, vshrn.u16, 1
+function ff_put_pixels_clamped_neon, export=1
+ vld1.64 {d16-d19}, [r0,:128]!
+ vqmovun.s16 d0, q8
+ vld1.64 {d20-d23}, [r0,:128]!
+ vqmovun.s16 d1, q9
+ vld1.64 {d24-d27}, [r0,:128]!
+ vqmovun.s16 d2, q10
+ vld1.64 {d28-d31}, [r0,:128]!
+ vqmovun.s16 d3, q11
+ vst1.64 {d0}, [r1,:64], r2
+ vqmovun.s16 d4, q12
+ vst1.64 {d1}, [r1,:64], r2
+ vqmovun.s16 d5, q13
+ vst1.64 {d2}, [r1,:64], r2
+ vqmovun.s16 d6, q14
+ vst1.64 {d3}, [r1,:64], r2
+ vqmovun.s16 d7, q15
+ vst1.64 {d4}, [r1,:64], r2
+ vst1.64 {d5}, [r1,:64], r2
+ vst1.64 {d6}, [r1,:64], r2
+ vst1.64 {d7}, [r1,:64], r2
+ bx lr
+ .endfunc
+
+function ff_put_signed_pixels_clamped_neon, export=1
+ vmov.u8 d31, #128
+ vld1.64 {d16-d17}, [r0,:128]!
+ vqmovn.s16 d0, q8
+ vld1.64 {d18-d19}, [r0,:128]!
+ vqmovn.s16 d1, q9
+ vld1.64 {d16-d17}, [r0,:128]!
+ vqmovn.s16 d2, q8
+ vld1.64 {d18-d19}, [r0,:128]!
+ vadd.u8 d0, d0, d31
+ vld1.64 {d20-d21}, [r0,:128]!
+ vadd.u8 d1, d1, d31
+ vld1.64 {d22-d23}, [r0,:128]!
+ vadd.u8 d2, d2, d31
+ vst1.64 {d0}, [r1,:64], r2
+ vqmovn.s16 d3, q9
+ vst1.64 {d1}, [r1,:64], r2
+ vqmovn.s16 d4, q10
+ vst1.64 {d2}, [r1,:64], r2
+ vqmovn.s16 d5, q11
+ vld1.64 {d24-d25}, [r0,:128]!
+ vadd.u8 d3, d3, d31
+ vld1.64 {d26-d27}, [r0,:128]!
+ vadd.u8 d4, d4, d31
+ vadd.u8 d5, d5, d31
+ vst1.64 {d3}, [r1,:64], r2
+ vqmovn.s16 d6, q12
+ vst1.64 {d4}, [r1,:64], r2
+ vqmovn.s16 d7, q13
+ vst1.64 {d5}, [r1,:64], r2
+ vadd.u8 d6, d6, d31
+ vadd.u8 d7, d7, d31
+ vst1.64 {d6}, [r1,:64], r2
+ vst1.64 {d7}, [r1,:64], r2
+ bx lr
+ .endfunc
+
+function ff_add_pixels_clamped_neon, export=1
+ mov r3, r1
+ vld1.64 {d16}, [r1,:64], r2
+ vld1.64 {d0-d1}, [r0,:128]!
+ vaddw.u8 q0, q0, d16
+ vld1.64 {d17}, [r1,:64], r2
+ vld1.64 {d2-d3}, [r0,:128]!
+ vqmovun.s16 d0, q0
+ vld1.64 {d18}, [r1,:64], r2
+ vaddw.u8 q1, q1, d17
+ vld1.64 {d4-d5}, [r0,:128]!
+ vaddw.u8 q2, q2, d18
+ vst1.64 {d0}, [r3,:64], r2
+ vqmovun.s16 d2, q1
+ vld1.64 {d19}, [r1,:64], r2
+ vld1.64 {d6-d7}, [r0,:128]!
+ vaddw.u8 q3, q3, d19
+ vqmovun.s16 d4, q2
+ vst1.64 {d2}, [r3,:64], r2
+ vld1.64 {d16}, [r1,:64], r2
+ vqmovun.s16 d6, q3
+ vld1.64 {d0-d1}, [r0,:128]!
+ vaddw.u8 q0, q0, d16
+ vst1.64 {d4}, [r3,:64], r2
+ vld1.64 {d17}, [r1,:64], r2
+ vld1.64 {d2-d3}, [r0,:128]!
+ vaddw.u8 q1, q1, d17
+ vst1.64 {d6}, [r3,:64], r2
+ vqmovun.s16 d0, q0
+ vld1.64 {d18}, [r1,:64], r2
+ vld1.64 {d4-d5}, [r0,:128]!
+ vaddw.u8 q2, q2, d18
+ vst1.64 {d0}, [r3,:64], r2
+ vqmovun.s16 d2, q1
+ vld1.64 {d19}, [r1,:64], r2
+ vqmovun.s16 d4, q2
+ vld1.64 {d6-d7}, [r0,:128]!
+ vaddw.u8 q3, q3, d19
+ vst1.64 {d2}, [r3,:64], r2
+ vqmovun.s16 d6, q3
+ vst1.64 {d4}, [r3,:64], r2
+ vst1.64 {d6}, [r3,:64], r2
+ bx lr
+ .endfunc
+
function ff_float_to_int16_neon, export=1
subs r2, r2, #8
vld1.64 {d0-d1}, [r1,:128]!
diff -Nurd ffmpeg.old/libavcodec/arm/simple_idct_neon.S ffmpeg-0.5/libavcodec/arm/simple_idct_neon.S
--- ffmpeg.old/libavcodec/arm/simple_idct_neon.S 2008-12-30 04:13:52.000000000 +0100
+++ ffmpeg-0.5/libavcodec/arm/simple_idct_neon.S 2009-05-30 11:27:54.000000000 +0200
@@ -68,6 +68,19 @@
.text
.align 6
+function idct_row4_pld_neon
+ pld [r0]
+ add r3, r0, r1, lsl #2
+ pld [r0, r1]
+ pld [r0, r1, lsl #1]
+ pld [r3, -r1]
+ pld [r3]
+ pld [r3, r1]
+ add r3, r3, r1, lsl #1
+ pld [r3]
+ pld [r3, r1]
+ .endfunc
+
function idct_row4_neon
vmov.i32 q15, #(1<<(ROW_SHIFT-1))
vld1.64 {d2-d5}, [r2,:128]!
@@ -252,7 +265,7 @@
function ff_simple_idct_put_neon, export=1
idct_start r2
- bl idct_row4_neon
+ bl idct_row4_pld_neon
bl idct_row4_neon
add r2, r2, #-128
bl idct_col4_neon
@@ -307,7 +320,7 @@
function ff_simple_idct_add_neon, export=1
idct_start r2
- bl idct_row4_neon
+ bl idct_row4_pld_neon
bl idct_row4_neon
add r2, r2, #-128
bl idct_col4_neon
diff -Nurd ffmpeg.old/libavcodec/arm/vp3dsp_neon.S ffmpeg-0.5/libavcodec/arm/vp3dsp_neon.S
--- ffmpeg.old/libavcodec/arm/vp3dsp_neon.S 1970-01-01 01:00:00.000000000 +0100
+++ ffmpeg-0.5/libavcodec/arm/vp3dsp_neon.S 2009-05-30 11:27:54.000000000 +0200
@@ -0,0 +1,94 @@
+/*
+ * Copyright (c) 2009 David Conrad
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg 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,
+ * 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
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "asm.S"
+
+.macro vp3_loop_filter
+ vsubl.u8 q3, d18, d17
+ vsubl.u8 q2, d16, d19
+ vadd.i16 q1, q3, q3
+ vadd.i16 q2, q2, q3
+ vadd.i16 q0, q1, q2
+ vrshr.s16 q0, q0, #3
+ vmovl.u8 q9, d18
+ vdup.u16 q15, r2
+
+ vabs.s16 q1, q0
+ vshr.s16 q0, q0, #15
+ vqsub.u16 q2, q15, q1
+ vqsub.u16 q3, q2, q1
+ vsub.i16 q1, q2, q3
+ veor q1, q1, q0
+ vsub.i16 q0, q1, q0
+
+ vaddw.u8 q2, q0, d17
+ vsub.i16 q3, q9, q0
+ vqmovun.s16 d0, q2
+ vqmovun.s16 d1, q3
+.endm
+
+function ff_vp3_v_loop_filter_neon, export=1
+ sub ip, r0, r1
+ sub r0, r0, r1, lsl #1
+ vld1.64 {d16}, [r0,:64], r1
+ vld1.64 {d17}, [r0,:64], r1
+ vld1.64 {d18}, [r0,:64], r1
+ vld1.64 {d19}, [r0,:64], r1
+ ldrb r2, [r2, #129*4]
+
+ vp3_loop_filter
+
+ vst1.64 {d0}, [ip,:64], r1
+ vst1.64 {d1}, [ip,:64], r1
+ bx lr
+.endfunc
+
+function ff_vp3_h_loop_filter_neon, export=1
+ sub ip, r0, #1
+ sub r0, r0, #2
+ vld1.32 {d16[]}, [r0], r1
+ vld1.32 {d17[]}, [r0], r1
+ vld1.32 {d18[]}, [r0], r1
+ vld1.32 {d19[]}, [r0], r1
+ vld1.32 {d16[1]}, [r0], r1
+ vld1.32 {d17[1]}, [r0], r1
+ vld1.32 {d18[1]}, [r0], r1
+ vld1.32 {d19[1]}, [r0], r1
+ ldrb r2, [r2, #129*4]
+
+ vtrn.8 d16, d17
+ vtrn.8 d18, d19
+ vtrn.16 d16, d18
+ vtrn.16 d17, d19
+
+ vp3_loop_filter
+
+ vtrn.8 d0, d1
+
+ vst1.16 {d0[0]}, [ip], r1
+ vst1.16 {d1[0]}, [ip], r1
+ vst1.16 {d0[1]}, [ip], r1
+ vst1.16 {d1[1]}, [ip], r1
+ vst1.16 {d0[2]}, [ip], r1
+ vst1.16 {d1[2]}, [ip], r1
+ vst1.16 {d0[3]}, [ip], r1
+ vst1.16 {d1[3]}, [ip], r1
+ bx lr
+.endfunc
diff -Nurd ffmpeg.old/libavcodec/Makefile ffmpeg-0.5/libavcodec/Makefile
--- ffmpeg.old/libavcodec/Makefile 2009-02-26 03:29:24.000000000 +0100
+++ ffmpeg-0.5/libavcodec/Makefile 2009-05-30 11:29:51.000000000 +0200
@@ -477,11 +477,15 @@
OBJS-$(HAVE_IWMMXT) += arm/dsputil_iwmmxt.o \
arm/mpegvideo_iwmmxt.o \
+NEON-OBJS-$(CONFIG_THEORA_DECODER) += arm/vp3dsp_neon.o
+NEON-OBJS-$(CONFIG_VP3_DECODER) += arm/vp3dsp_neon.o
+
OBJS-$(HAVE_NEON) += arm/dsputil_neon.o \
arm/dsputil_neon_s.o \
arm/h264dsp_neon.o \
arm/h264idct_neon.o \
arm/simple_idct_neon.o \
+ $(NEON-OBJS-yes)
OBJS-$(ARCH_BFIN) += bfin/dsputil_bfin.o \
bfin/fdct_bfin.o \