summaryrefslogtreecommitdiffstats
path: root/libavcodec/arm
diff options
context:
space:
mode:
Diffstat (limited to 'libavcodec/arm')
-rw-r--r--libavcodec/arm/Makefile3
-rw-r--r--libavcodec/arm/ac3dsp_init_arm.c39
-rw-r--r--libavcodec/arm/ac3dsp_neon.S80
-rw-r--r--libavcodec/arm/dsputil_init_neon.c5
-rw-r--r--libavcodec/arm/dsputil_neon.S23
5 files changed, 150 insertions, 0 deletions
diff --git a/libavcodec/arm/Makefile b/libavcodec/arm/Makefile
index d223703..08697da 100644
--- a/libavcodec/arm/Makefile
+++ b/libavcodec/arm/Makefile
@@ -1,3 +1,4 @@
+OBJS-$(CONFIG_AC3DSP) += arm/ac3dsp_init_arm.o
OBJS-$(CONFIG_DCA_DECODER) += arm/dcadsp_init_arm.o \
OBJS-$(CONFIG_VP5_DECODER) += arm/vp56dsp_init_arm.o
@@ -46,6 +47,8 @@ NEON-OBJS-$(CONFIG_H264DSP) += arm/h264dsp_neon.o \
NEON-OBJS-$(CONFIG_H264PRED) += arm/h264pred_neon.o \
+NEON-OBJS-$(CONFIG_AC3DSP) += arm/ac3dsp_neon.o
+
NEON-OBJS-$(CONFIG_DCA_DECODER) += arm/dcadsp_neon.o \
arm/synth_filter_neon.o \
diff --git a/libavcodec/arm/ac3dsp_init_arm.c b/libavcodec/arm/ac3dsp_init_arm.c
new file mode 100644
index 0000000..6874546
--- /dev/null
+++ b/libavcodec/arm/ac3dsp_init_arm.c
@@ -0,0 +1,39 @@
+/*
+ * Copyright (c) 2011 Mans Rullgard <mans@mansr.com>
+ *
+ * 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 <stdint.h>
+#include "libavutil/attributes.h"
+#include "libavcodec/ac3dsp.h"
+#include "config.h"
+
+void ff_ac3_exponent_min_neon(uint8_t *exp, int num_reuse_blocks, int nb_coefs);
+int ff_ac3_max_msb_abs_int16_neon(const int16_t *src, int len);
+void ff_ac3_lshift_int16_neon(int16_t *src, unsigned len, unsigned shift);
+void ff_ac3_rshift_int32_neon(int32_t *src, unsigned len, unsigned shift);
+
+av_cold void ff_ac3dsp_init_arm(AC3DSPContext *c, int bit_exact)
+{
+ if (HAVE_NEON) {
+ c->ac3_exponent_min = ff_ac3_exponent_min_neon;
+ c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_neon;
+ c->ac3_lshift_int16 = ff_ac3_lshift_int16_neon;
+ c->ac3_rshift_int32 = ff_ac3_rshift_int32_neon;
+ }
+}
diff --git a/libavcodec/arm/ac3dsp_neon.S b/libavcodec/arm/ac3dsp_neon.S
new file mode 100644
index 0000000..43d4e61
--- /dev/null
+++ b/libavcodec/arm/ac3dsp_neon.S
@@ -0,0 +1,80 @@
+/*
+ * Copyright (c) 2011 Mans Rullgard <mans@mansr.com>
+ *
+ * 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 "asm.S"
+
+function ff_ac3_max_msb_abs_int16_neon, export=1
+ vmov.i16 q0, #0
+ vmov.i16 q2, #0
+1: vld1.16 {q1}, [r0,:128]!
+ vabs.s16 q1, q1
+ vld1.16 {q3}, [r0,:128]!
+ vabs.s16 q3, q3
+ vorr q0, q0, q1
+ vorr q2, q2, q3
+ subs r1, r1, #16
+ bgt 1b
+ vorr q0, q0, q2
+ vorr d0, d0, d1
+ vpmax.u16 d0, d0, d0
+ vpmax.u16 d0, d0, d0
+ vmov.u16 r0, d0[0]
+ bx lr
+endfunc
+
+function ff_ac3_exponent_min_neon, export=1
+ cmp r1, #0
+ bxeq lr
+ push {lr}
+ mov r12, #256
+1:
+ vld1.8 {q0}, [r0,:128]
+ mov lr, r1
+ add r3, r0, #256
+2: vld1.8 {q1}, [r3,:128], r12
+ subs lr, lr, #1
+ vmin.u8 q0, q0, q1
+ bgt 2b
+ subs r2, r2, #16
+ vst1.8 {q0}, [r0,:128]!
+ bgt 1b
+ pop {pc}
+endfunc
+
+function ff_ac3_lshift_int16_neon, export=1
+ vdup.16 q0, r2
+1: vld1.16 {q1}, [r0,:128]
+ vshl.s16 q1, q1, q0
+ vst1.16 {q1}, [r0,:128]!
+ subs r1, r1, #8
+ bgt 1b
+ bx lr
+endfunc
+
+function ff_ac3_rshift_int32_neon, export=1
+ rsb r2, r2, #0
+ vdup.32 q0, r2
+1: vld1.32 {q1}, [r0,:128]
+ vshl.s32 q1, q1, q0
+ vst1.32 {q1}, [r0,:128]!
+ subs r1, r1, #4
+ bgt 1b
+ bx lr
+endfunc
diff --git a/libavcodec/arm/dsputil_init_neon.c b/libavcodec/arm/dsputil_init_neon.c
index 76ae632..ef5b63b 100644
--- a/libavcodec/arm/dsputil_init_neon.c
+++ b/libavcodec/arm/dsputil_init_neon.c
@@ -168,6 +168,9 @@ int32_t ff_scalarproduct_int16_neon(const int16_t *v1, const int16_t *v2, int le
int32_t ff_scalarproduct_and_madd_int16_neon(int16_t *v1, const int16_t *v2,
const int16_t *v3, int len, int mul);
+void ff_apply_window_int16_neon(int16_t *dst, const int16_t *src,
+ const int16_t *window, unsigned n);
+
void ff_dsputil_init_neon(DSPContext *c, AVCodecContext *avctx)
{
if (!avctx->lowres) {
@@ -319,4 +322,6 @@ void ff_dsputil_init_neon(DSPContext *c, AVCodecContext *avctx)
c->scalarproduct_int16 = ff_scalarproduct_int16_neon;
c->scalarproduct_and_madd_int16 = ff_scalarproduct_and_madd_int16_neon;
+
+ c->apply_window_int16 = ff_apply_window_int16_neon;
}
diff --git a/libavcodec/arm/dsputil_neon.S b/libavcodec/arm/dsputil_neon.S
index 05a9115..0dbf5ca 100644
--- a/libavcodec/arm/dsputil_neon.S
+++ b/libavcodec/arm/dsputil_neon.S
@@ -789,3 +789,26 @@ NOVFP ldr r2, [sp]
vst1.f32 {q9},[r0,:128]!
bx lr
endfunc
+
+function ff_apply_window_int16_neon, export=1
+ push {r4,lr}
+ add r4, r1, r3, lsl #1
+ add lr, r0, r3, lsl #1
+ sub r4, r4, #16
+ sub lr, lr, #16
+ mov r12, #-16
+1:
+ vld1.16 {q0}, [r1,:128]!
+ vld1.16 {q2}, [r2,:128]!
+ vld1.16 {q1}, [r4,:128], r12
+ vrev64.16 q3, q2
+ vqrdmulh.s16 q0, q0, q2
+ vqrdmulh.s16 d2, d2, d7
+ vqrdmulh.s16 d3, d3, d6
+ vst1.16 {q0}, [r0,:128]!
+ vst1.16 {q1}, [lr,:128], r12
+ subs r3, r3, #16
+ bgt 1b
+
+ pop {r4,pc}
+endfunc
OpenPOWER on IntegriCloud