#include "avcodec.h"
#include "acelp_filters.h"
-#define FRAC_BITS 13
-#include "mathops.h"
-void ff_acelp_convolve_circ(
- int16_t* fc_out,
- const int16_t* fc_in,
- const int16_t* filter,
- int subframe_size)
+const int16_t ff_acelp_interp_filter[61] =
+{ /* (0.15) */
+ 29443, 28346, 25207, 20449, 14701, 8693,
+ 3143, -1352, -4402, -5865, -5850, -4673,
+ -2783, -672, 1211, 2536, 3130, 2991,
+ 2259, 1170, 0, -1001, -1652, -1868,
+ -1666, -1147, -464, 218, 756, 1060,
+ 1099, 904, 550, 135, -245, -514,
+ -634, -602, -451, -231, 0, 191,
+ 308, 340, 296, 198, 78, -36,
+ -120, -163, -165, -132, -79, -19,
+ 34, 73, 91, 89, 70, 38,
+ 0,
+};
+
+void ff_acelp_interpolate(
+ int16_t* out,
+ const int16_t* in,
+ const int16_t* filter_coeffs,
+ int precision,
+ int frac_pos,
+ int filter_length,
+ int length)
{
- int i, k;
+ int n, i;
- memset(fc_out, 0, subframe_size * sizeof(int16_t));
+ assert(pitch_delay_frac >= 0 && pitch_delay_frac < precision);
- /* Since there are few pulses over entire subframe (i.e. almost all
- fc_in[i] are zero, in case of G.729D the buffer contains two non-zero
- samples before the call to ff_acelp_enhance_harmonics, and (due to
- pitch_delay bounded to [20; 143]) a maximum four non-zero samples
- for a total of 40 after the call to it), it is faster to swap two loops
- and process non-zero samples only. This will reduce the number of
- multiplications from 40*40 to 4*40 for G.729D */
- for(i=0; i<subframe_size; i++)
+ for(n=0; n<length; n++)
{
- if(fc_in[i])
- {
- for(k=0; k<i; k++)
- fc_out[k] += (fc_in[i] * filter[subframe_size + k - i]) >> 15;
+ int idx = 0;
+ int v = 0x4000;
- for(k=i; k<subframe_size; k++)
- fc_out[k] += (fc_in[i] * filter[k - i]) >> 15;
- }
- }
-}
-
-int ff_acelp_lp_synthesis_filter(
- int16_t *out,
- const int16_t* filter_coeffs,
- const int16_t* in,
- int buffer_length,
- int filter_length,
- int stop_on_overflow)
-{
- int i,n;
+ for(i=0; i<filter_length;)
+ {
- for(n=0; n<buffer_length; n++)
- {
- int sum = 0x800;
- for(i=1; i<filter_length; i++)
- sum -= filter_coeffs[i] * out[n-i];
+ /* The reference G.729 and AMR fixed point code performs clipping after
+ each of the two following accumulations.
+ Since clipping affects only the synthetic OVERFLOW test without
+ causing an int type overflow, it was moved outside the loop. */
- sum = (sum >> 12) + in[n];
+ /* R(x):=ac_v[-k+x]
+ v += R(n-i)*ff_acelp_interp_filter(t+6i)
+ v += R(n+i+1)*ff_acelp_interp_filter(6-t+6i) */
- /* Check for overflow */
- if(sum + 0x8000 > 0xFFFFU)
- {
- if(stop_on_overflow)
- return 1;
- sum = (sum >> 31) ^ 32767;
+ v += in[n + i] * filter_coeffs[idx + frac_pos];
+ idx += precision;
+ i++;
+ v += in[n - i] * filter_coeffs[idx - frac_pos];
}
- out[n] = sum;
+ if(av_clip_int16(v>>15) != (v>>15))
+ av_log(NULL, AV_LOG_WARNING, "overflow that would need cliping in ff_acelp_interpolate()\n");
+ out[n] = v >> 15;
}
-
- return 0;
}
-void ff_acelp_weighted_filter(
- int16_t *out,
- const int16_t* in,
- const int16_t *weight_pow,
- int filter_length)
-{
- int n;
- for(n=0; n<filter_length; n++)
- out[n] = (in[n] * weight_pow[n] + 0x4000) >> 15; /* (3.12) = (0.15) * (3.12) with rounding */
-}
void ff_acelp_high_pass_filter(
int16_t* out,
for(i=0; i<length; i++)
{
- tmp = MULL(hpf_f[0], 15836); /* (14.13) = (13.13) * (1.13) */
- tmp += MULL(hpf_f[1], -7667); /* (13.13) = (13.13) * (0.13) */
- tmp += 7699 * (in[i] - 2*in[i-1] + in[i-2]); /* (14.13) = (0.13) * (14.0) */
-
- /* Multiplication by 2 with rounding can cause short type
- overflow, thus clipping is required. */
+ tmp = (hpf_f[0]* 15836LL)>>13;
+ tmp += (hpf_f[1]* -7667LL)>>13;
+ tmp += 7699 * (in[i] - 2*in[i-1] + in[i-2]);
- out[i] = av_clip_int16((tmp + 0x800) >> 12); /* (15.0) = 2 * (13.13) = (14.13) */
+ /* With "+0x800" rounding, clipping is needed
+ for ALGTHM and SPEECH tests. */
+ out[i] = av_clip_int16((tmp + 0x800) >> 12);
hpf_f[1] = hpf_f[0];
hpf_f[0] = tmp;
projects / ffmpeg / blobdiff