devs/foxbox
1
0
Fork 0
forked from len0rd/rockbox
Code Pull requests Activity

opus: allocate mdct f2 buffer in iram

Browse source 
Speeds up decoding of 64kbps test file by 6.3MHz on h300 (cf)
and 1.2MHz on c200 (pp).

Signed-off-by: Nils Wallménius <nils@rockbox.org>

Change-Id: I08c2c332153abcbef9447c81986777fd2fcc73fe
This commit is contained in:
Andree Buschmann 2012-10-01 22:07:44 +02:00 • committed by Nils Wallménius
parent b6bcb1338e
commit d7799aaf33
1 changed files with 17 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

22
lib/rbcodec/codecs/libopus/celt/mdct.c
View file

@ -208,22 +208,34 @@ void clt_mdct_forward(const mdct_lookup *l, kiss_fft_scalar *in, kiss_fft_scalar
} }
#endif #endif
#define S_F_BUF_SIZE (1920>>1) /* N = 1920 for static modes */
static kiss_fft_scalar s_f2[S_F_BUF_SIZE] IBSS_ATTR MEM_ALIGN_ATTR;
void clt_mdct_backward(const mdct_lookup *l, kiss_fft_scalar *in, kiss_fft_scalar * OPUS_RESTRICT out, void clt_mdct_backward(const mdct_lookup *l, kiss_fft_scalar *in, kiss_fft_scalar * OPUS_RESTRICT out,
const opus_val16 * OPUS_RESTRICT window, int overlap, int shift, int stride) const opus_val16 * OPUS_RESTRICT window, int overlap, int shift, int stride)
{ {
int i; int i;
int N, N2, N4; int N, N2, N4;
kiss_twiddle_scalar sine; kiss_twiddle_scalar sine;
/* VARDECL(kiss_fft_scalar, f); */ VARDECL(kiss_fft_scalar, f);
VARDECL(kiss_fft_scalar, f2); VARDECL(kiss_fft_scalar, f2);
SAVE_STACK; SAVE_STACK;
N = l->n; /* static modes => N = 1920 */ N = l->n;
N >>= shift; N >>= shift;
N2 = N>>1; N2 = N>>1;
N4 = N>>2; N4 = N>>2;
/* ALLOC(f, N2, kiss_fft_scalar); */ kiss_fft_scalar s_f[S_F_BUF_SIZE];
kiss_fft_scalar f[N2]; /* worst case 3840b */
ALLOC(f2, N2, kiss_fft_scalar); if (S_F_BUF_SIZE >= N2)
{
f = s_f;
f2 = s_f2;
}
else
{
ALLOC(f , N2, kiss_fft_scalar);
ALLOC(f2, N2, kiss_fft_scalar);
}
/* sin(x) ~= x here */ /* sin(x) ~= x here */
#ifdef FIXED_POINT #ifdef FIXED_POINT
sine = TRIG_UPSCALE*(QCONST16(0.7853981f, 15)+N2)/N; sine = TRIG_UPSCALE*(QCONST16(0.7853981f, 15)+N2)/N;