Skip to content

Audio: ASRC: HiFi5 optimizations #9875

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
lgirdwood merged 6 commits into from
Mar 11, 2025
Merged

Audio: ASRC: HiFi5 optimizations #9875

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
9938d52
Audio: ASRC: Duplicate HiFi5 version from HiFi3 version
singalsu Mar 6, 2025
0460b3c
Audio: ASRC: Add coefficients files for HiFi5
singalsu Mar 6, 2025
c3dce12
Audio: ASRC: Optimize for HiFi5 impulse response calculations
singalsu Mar 6, 2025
f309250
Audio: ASRC: Reverse the FIR delay write direction
singalsu Mar 7, 2025
d2942f3
Audio: ASRC: Optimize for HiFi5 function asrc_fir_filter32()
singalsu Mar 7, 2025
407710b
Audio: ASRC: Optimize for HiFi5 function asrc_fir_filter16()
singalsu Mar 10, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion src/audio/asrc/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
# SPDX-License-Identifier: BSD-3-Clause

add_local_sources(sof asrc.c asrc_farrow.c asrc_farrow_generic.c
asrc_farrow_hifi3.c)
asrc_farrow_hifi3.c asrc_farrow_hifi5.c)

if(CONFIG_IPC_MAJOR_3)
add_local_sources(sof asrc_ipc3.c)
Expand Down
5 changes: 3 additions & 2 deletions src/audio/asrc/asrc.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -251,10 +251,11 @@ static int asrc_initialize_buffers(struct asrc_farrow *src_obj)

/* Set buffer_length to filter_length * 2 to compensate for
* missing element wise wrap around while loading but allowing
* aligned loads.
* aligned loads. FIR delay line write is initialized to last
* position of first copy block for reverse direction write.
*/
src_obj->buffer_length = src_obj->filter_length * 2;
src_obj->buffer_write_position = src_obj->filter_length;
src_obj->buffer_write_position = src_obj->filter_length - 1;

if (src_obj->bit_depth == 32) {
buffer_size = src_obj->buffer_length * sizeof(int32_t);
Expand Down
62 changes: 54 additions & 8 deletions src/audio/asrc/asrc_farrow.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -145,6 +145,51 @@ static const struct asrc_filter_params c_filter_params[CR_NUM] = {
* coefficients will be attached to the _Src_farrow struct via the
* initialise_filter function.
*/

#if SOF_USE_MIN_HIFI(5, ASRC)
#include "coef/asrc_farrow_coeff_4x_44100Hz_to_48000Hz.h"
#include "coef/asrc_farrow_coeff_4x_48000Hz_to_48000Hz.h"

#if (CONFIG_ASRC_SUPPORT_CONVERSION_24000_TO_08000)
#include "coef/asrc_farrow_coeff_4x_24000Hz_to_08000Hz.h"
#endif

#if (CONFIG_ASRC_SUPPORT_CONVERSION_24000_TO_16000)
#include "coef/asrc_farrow_coeff_4x_24000Hz_to_16000Hz.h"
#endif

#if (CONFIG_ASRC_SUPPORT_CONVERSION_48000_TO_08000)
#include "coef/asrc_farrow_coeff_4x_48000Hz_to_08000Hz.h"
#endif

#if (CONFIG_ASRC_SUPPORT_CONVERSION_48000_TO_11025)
#include "coef/asrc_farrow_coeff_4x_48000Hz_to_11025Hz.h"
#endif

#if (CONFIG_ASRC_SUPPORT_CONVERSION_48000_TO_12000)
#include "coef/asrc_farrow_coeff_4x_48000Hz_to_12000Hz.h"
#endif

#if (CONFIG_ASRC_SUPPORT_CONVERSION_48000_TO_16000)
#include "coef/asrc_farrow_coeff_4x_48000Hz_to_16000Hz.h"
#endif

#if (CONFIG_ASRC_SUPPORT_CONVERSION_48000_TO_22050)
#include "coef/asrc_farrow_coeff_4x_48000Hz_to_22050Hz.h"
#endif

#if (CONFIG_ASRC_SUPPORT_CONVERSION_48000_TO_24000)
#include "coef/asrc_farrow_coeff_4x_48000Hz_to_24000Hz.h"
#endif

#if (CONFIG_ASRC_SUPPORT_CONVERSION_48000_TO_32000)
#include "coef/asrc_farrow_coeff_4x_48000Hz_to_32000Hz.h"
#endif

#if (CONFIG_ASRC_SUPPORT_CONVERSION_48000_TO_44100)
#include "coef/asrc_farrow_coeff_4x_48000Hz_to_44100Hz.h"
#endif
#else
#include "coef/asrc_farrow_coeff_44100Hz_to_48000Hz.h"

#include "coef/asrc_farrow_coeff_48000Hz_to_48000Hz.h"
Expand Down Expand Up @@ -188,6 +233,7 @@ static const struct asrc_filter_params c_filter_params[CR_NUM] = {
#if (CONFIG_ASRC_SUPPORT_CONVERSION_48000_TO_44100)
#include "coef/asrc_farrow_coeff_48000Hz_to_44100Hz.h"
#endif
#endif

/*
* FUNCTION DECLARATIONS
Expand Down Expand Up @@ -836,11 +882,11 @@ void asrc_write_to_ring_buffer16(struct asrc_farrow *src_obj,
int m;

/* update the buffer_write_position */
(src_obj->buffer_write_position)++;
(src_obj->buffer_write_position)--;

/* since it's a ring buffer we need a wrap around */
if (src_obj->buffer_write_position >= src_obj->buffer_length)
src_obj->buffer_write_position -= (src_obj->buffer_length >> 1);
if (src_obj->buffer_write_position < 0)
src_obj->buffer_write_position += (src_obj->buffer_length >> 1);

/* handle input format */
if (src_obj->input_format == ASRC_IOF_INTERLEAVED)
Expand All @@ -850,7 +896,7 @@ void asrc_write_to_ring_buffer16(struct asrc_farrow *src_obj,

/* write data to each channel */
j = src_obj->buffer_write_position;
k = j - (src_obj->buffer_length >> 1);
k = j + (src_obj->buffer_length >> 1);
for (ch = 0; ch < src_obj->num_channels; ch++) {
/*
* Since we want the filter function to load 64 bit of
Expand Down Expand Up @@ -878,11 +924,11 @@ void asrc_write_to_ring_buffer32(struct asrc_farrow *src_obj,
int m;

/* update the buffer_write_position */
(src_obj->buffer_write_position)++;
(src_obj->buffer_write_position)--;

/* since it's a ring buffer we need a wrap around */
if (src_obj->buffer_write_position >= src_obj->buffer_length)
src_obj->buffer_write_position -= (src_obj->buffer_length >> 1);
if (src_obj->buffer_write_position < 0)
src_obj->buffer_write_position += (src_obj->buffer_length >> 1);

/* handle input format */
if (src_obj->input_format == ASRC_IOF_INTERLEAVED)
Expand All @@ -892,7 +938,7 @@ void asrc_write_to_ring_buffer32(struct asrc_farrow *src_obj,

/* write data to each channel */
j = src_obj->buffer_write_position;
k = j - (src_obj->buffer_length >> 1);
k = j + (src_obj->buffer_length >> 1);
for (ch = 0; ch < src_obj->num_channels; ch++) {
/*
* Since we want the filter function to load 64 bit of
Expand Down
4 changes: 2 additions & 2 deletions src/audio/asrc/asrc_farrow_generic.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,7 +44,7 @@ void asrc_fir_filter16(struct asrc_farrow *src_obj, int16_t **output_buffers,
* Data is Q1.15, coefficients are Q1.30. Prod will be Qx.45.
*/
for (n = 0; n < src_obj->filter_length; n++)
prod += (int64_t)(*buffer_p--) * (*filter_p++);
prod += (int64_t)(*buffer_p++) * (*filter_p++);

/* Shift left after accumulation, because interim
* results might saturate during filtering prod = prod
Expand Down Expand Up @@ -96,7 +96,7 @@ void asrc_fir_filter32(struct asrc_farrow *src_obj, int32_t **output_buffers,
* quality. The product is Qx.54.
*/
for (n = 0; n < src_obj->filter_length; n++)
prod += (int64_t)(*buffer_p--) * (*filter_p++ >> 8);
prod += (int64_t)(*buffer_p++) * (*filter_p++ >> 8);

/* Shift left after accumulation, because interim
* results might saturate during filtering prod = prod
Expand Down
8 changes: 4 additions & 4 deletions src/audio/asrc/asrc_farrow_hifi3.c
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,10 +1,10 @@
// SPDX-License-Identifier: BSD-3-Clause
//
// Copyright(c) 2012-2019 Intel Corporation. All rights reserved.
// Copyright(c) 2012-2025 Intel Corporation.

#include <sof/common.h>

#if SOF_USE_HIFI(3, ASRC) || SOF_USE_HIFI(4, ASRC) || SOF_USE_HIFI(5, ASRC)
#if SOF_USE_HIFI(3, ASRC) || SOF_USE_HIFI(4, ASRC)

#include "asrc_farrow.h"
#include <xtensa/tie/xt_hifi3.h>
Expand Down Expand Up @@ -56,7 +56,7 @@ void asrc_fir_filter16(struct asrc_farrow *src_obj, int16_t **output_buffers,
/* Iterate over the filter bins */
for (n = 0; n < n_limit; n++) {
/* Read four buffered samples at once */
AE_LA16X4_RIP(buffer0123, align_buffer, buffer_p);
AE_LA16X4_IP(buffer0123, align_buffer, buffer_p);

/* Store four bins of the impulse response */
AE_LA32X2_IP(filter01, align_filter, filter_p);
Expand Down Expand Up @@ -141,7 +141,7 @@ void asrc_fir_filter32(struct asrc_farrow *src_obj, int32_t **output_buffers,
/* Iterate over the filter bins */
for (n = 0; n < n_limit; n++) {
/* Read two buffered samples at once */
AE_LA32X2_RIP(buffer01, align_buffer, buffer_p);
AE_LA32X2_IP(buffer01, align_buffer, buffer_p);

/* Store two bins of the impulse response */
AE_LA32X2_IP(filter01, align_filter, filter_p);
Expand Down
Loading
Loading