diff --git a/libavfilter/af_sofalizer.c b/libavfilter/af_sofalizer.c index 877f009709..33937d8221 100644 --- a/libavfilter/af_sofalizer.c +++ b/libavfilter/af_sofalizer.c @@ -269,7 +269,7 @@ static int load_sofa(AVFilterContext *ctx, char *filename, int *samplingrate) sp_r = s->sofa.sp_r = av_malloc_array(m_dim, sizeof(float)); /* delay and IR values required for each ear and measurement position: */ data_delay = s->sofa.data_delay = av_calloc(m_dim, 2 * sizeof(int)); - data_ir = s->sofa.data_ir = av_malloc_array(m_dim * n_samples, sizeof(float) * 2); + data_ir = s->sofa.data_ir = av_calloc(m_dim * FFALIGN(n_samples, 16), sizeof(float) * 2); if (!data_delay || !sp_a || !sp_e || !sp_r || !data_ir) { /* if memory could not be allocated */ @@ -352,6 +352,8 @@ static int load_sofa(AVFilterContext *ctx, char *filename, int *samplingrate) s->sofa.ncid = ncid; /* netCDF ID of SOFA file */ nc_close(ncid); /* close SOFA file */ + av_log(ctx, AV_LOG_DEBUG, "m_dim: %d n_samples %d\n", m_dim, n_samples); + return 0; error: @@ -554,7 +556,7 @@ static int sofalizer_convolute(AVFilterContext *ctx, void *arg, int jobnr, int n /* LFE is an input channel but requires no convolution */ /* apply gain to LFE signal and add to output buffer */ *dst += *(buffer[s->lfe_channel] + wr) * s->gain_lfe; - temp_ir += n_samples; + temp_ir += FFALIGN(n_samples, 16); continue; } @@ -574,7 +576,7 @@ static int sofalizer_convolute(AVFilterContext *ctx, void *arg, int jobnr, int n /* multiply signal and IR, and add up the results */ dst[0] += s->fdsp->scalarproduct_float(temp_ir, temp_src, n_samples); - temp_ir += n_samples; + temp_ir += FFALIGN(n_samples, 16); } /* clippings counter */ @@ -812,8 +814,8 @@ static int load_data(AVFilterContext *ctx, int azim, int elev, float radius) s->temp_src[1] = av_calloc(FFALIGN(n_samples, 16), sizeof(float)); /* get temporary IR for L and R channel */ - data_ir_l = av_malloc_array(n_conv * n_samples, sizeof(*data_ir_l)); - data_ir_r = av_malloc_array(n_conv * n_samples, sizeof(*data_ir_r)); + data_ir_l = av_calloc(n_conv * FFALIGN(n_samples, 16), sizeof(*data_ir_l)); + data_ir_r = av_calloc(n_conv * FFALIGN(n_samples, 16), sizeof(*data_ir_r)); if (!data_ir_r || !data_ir_l || !s->temp_src[0] || !s->temp_src[1]) { av_free(data_ir_l); av_free(data_ir_r); @@ -842,7 +844,7 @@ static int load_data(AVFilterContext *ctx, int azim, int elev, float radius) delay_r[i] = *(s->sofa.data_delay + 2 * m[i] + 1); if (s->type == TIME_DOMAIN) { - offset = i * n_samples; /* no. samples already written */ + offset = i * FFALIGN(n_samples, 16); /* no. samples already written */ for (j = 0; j < n_samples; j++) { /* load reversed IRs of the specified source position * sample-by-sample for left and right ear; and apply gain */ @@ -889,8 +891,8 @@ static int load_data(AVFilterContext *ctx, int azim, int elev, float radius) if (s->type == TIME_DOMAIN) { /* copy IRs and delays to allocated memory in the SOFAlizerContext struct: */ - memcpy(s->data_ir[0], data_ir_l, sizeof(float) * n_conv * n_samples); - memcpy(s->data_ir[1], data_ir_r, sizeof(float) * n_conv * n_samples); + memcpy(s->data_ir[0], data_ir_l, sizeof(float) * n_conv * FFALIGN(n_samples, 16)); + memcpy(s->data_ir[1], data_ir_r, sizeof(float) * n_conv * FFALIGN(n_samples, 16)); av_freep(&data_ir_l); /* free temporary IR memory */ av_freep(&data_ir_r); @@ -1006,8 +1008,8 @@ static int config_input(AVFilterLink *inlink) /* Allocate memory for the impulse responses, delays and the ringbuffers */ /* size: (longest IR) * (number of channels to convolute) */ - s->data_ir[0] = av_malloc_array(n_max_ir, sizeof(float) * s->n_conv); - s->data_ir[1] = av_malloc_array(n_max_ir, sizeof(float) * s->n_conv); + s->data_ir[0] = av_calloc(FFALIGN(n_max_ir, 16), sizeof(float) * s->n_conv); + s->data_ir[1] = av_calloc(FFALIGN(n_max_ir, 16), sizeof(float) * s->n_conv); /* length: number of channels to convolute */ s->delay[0] = av_malloc_array(s->n_conv, sizeof(float)); s->delay[1] = av_malloc_array(s->n_conv, sizeof(float));