har0ke/FFmpeg
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

dcadsp: add int8x8_fmul_int32 to dsp context

Browse Source 
It is currently declared as a macro who is set to inlinable functions,
among which a Neon and a default C implementations.

Add a DSP parameter to each inline function, unused except by the
default C implementation which calls a function from the DSP context.

On an Arrandale CPU, gain for an inlined SSE2 function vs. a call:
- Win32: 29 to 26 cycles
- Win64: 25 to 23 cycles

Signed-off-by: Janne Grunau <janne-libav@jannau.net>
This commit is contained in:
Christophe Gisquet 2012-05-11 11:17:36 +02:00 committed by Janne Grunau
parent e3fec3f095
commit 2bd44cb705
4 changed files with 16 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
libavcodec/arm/dca.h
View File

@ -83,7 +83,8 @@ static inline int decode_blockcodes(int code1, int code2, int levels,
#if HAVE_NEON_INLINE && HAVE_ASM_MOD_Y
#define int8x8_fmul_int32 int8x8_fmul_int32
static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
static inline void int8x8_fmul_int32(av_unused DCADSPContext *dsp,
float *dst, const int8_t *src, int scale)
{
__asm__ ("vcvt.f32.s32 %2, %2, #4 \n"
"vld1.8 {d0}, [%1,:64] \n"

10
libavcodec/dcadec.c
View File

@ -1086,12 +1086,10 @@ static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
#ifndef int8x8_fmul_int32
static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
static inline void int8x8_fmul_int32(DCADSPContext *dsp, float *dst,
const int8_t *src, int scale)
{
float fscale = scale / 16.0;
int i;
for (i = 0; i < 8; i++)
dst[i] = src[i] * fscale;
dsp->int8x8_fmul_int32(dst, src, scale);
}
#endif
@ -1219,7 +1217,7 @@ static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
s->debug_flag |= 0x01;
}
int8x8_fmul_int32(subband_samples[k][l],
int8x8_fmul_int32(&s->dcadsp, subband_samples[k][l],
&high_freq_vq[hfvq][subsubframe * 8],
s->scale_factor[k][l][0]);
}

9
libavcodec/dcadsp.c
View File

@ -24,6 +24,14 @@
#include "libavutil/intreadwrite.h"
#include "dcadsp.h"
static void int8x8_fmul_int32_c(float *dst, const int8_t *src, int scale)
{
float fscale = scale / 16.0;
int i;
for (i = 0; i < 8; i++)
dst[i] = src[i] * fscale;
}
static void dca_lfe_fir_c(float *out, const float *in, const float *coefs,
int decifactor, float scale)
{
@ -78,5 +86,6 @@ av_cold void ff_dcadsp_init(DCADSPContext *s)
{
s->lfe_fir = dca_lfe_fir_c;
s->qmf_32_subbands = dca_qmf_32_subbands;
s->int8x8_fmul_int32 = int8x8_fmul_int32_c;
if (ARCH_ARM) ff_dcadsp_init_arm(s);
}

1
libavcodec/dcadsp.h
View File

@ -31,6 +31,7 @@ typedef struct DCADSPContext {
int *synth_buf_offset, float synth_buf2[32],
const float window[512], float *samples_out,
float raXin[32], float scale);
void (*int8x8_fmul_int32)(float *dst, const int8_t *src, int scale);
} DCADSPContext;
void ff_dcadsp_init(DCADSPContext *s);