Commit ea1167c3 authored by Yi Luo's avatar Yi Luo
Browse files

Remove unused copies of transform related source code

- Library size reduces: 165 kB, 292 kB (HBD).

Change-Id: I50cb630dde326bd2a28c0db4b7e2d53c2fd94a2a
parent 57a7baf6
......@@ -30,8 +30,6 @@ AV1_COMMON_SRCS-yes += common/filter.h
AV1_COMMON_SRCS-yes += common/filter.c
AV1_COMMON_SRCS-yes += common/idct.h
AV1_COMMON_SRCS-yes += common/idct.c
AV1_COMMON_SRCS-yes += common/av1_inv_txfm.h
AV1_COMMON_SRCS-yes += common/av1_inv_txfm.c
AV1_COMMON_SRCS-yes += common/loopfilter.h
AV1_COMMON_SRCS-yes += common/thread_common.h
AV1_COMMON_SRCS-yes += common/mv.h
......@@ -61,8 +59,6 @@ AV1_COMMON_SRCS-yes += common/common_data.h
AV1_COMMON_SRCS-yes += common/scan.c
AV1_COMMON_SRCS-yes += common/scan.h
# TODO(angiebird) the forward transform belongs under encoder/
AV1_COMMON_SRCS-$(CONFIG_AV1_ENCODER) += common/av1_fwd_txfm.h
AV1_COMMON_SRCS-$(CONFIG_AV1_ENCODER) += common/av1_fwd_txfm.c
AV1_COMMON_SRCS-yes += common/av1_txfm.h
AV1_COMMON_SRCS-yes += common/av1_fwd_txfm1d.h
AV1_COMMON_SRCS-yes += common/av1_fwd_txfm1d.c
......@@ -123,9 +119,6 @@ AV1_COMMON_SRCS-$(HAVE_MSA) += common/mips/msa/idct16x16_msa.c
AV1_COMMON_SRCS-$(HAVE_SSE2) += common/x86/idct_intrin_sse2.c
ifeq ($(CONFIG_AV1_ENCODER),yes)
AV1_COMMON_SRCS-$(HAVE_SSE2) += common/x86/av1_fwd_txfm_sse2.c
AV1_COMMON_SRCS-$(HAVE_SSE2) += common/x86/av1_fwd_dct32x32_impl_sse2.h
AV1_COMMON_SRCS-$(HAVE_SSE2) += common/x86/av1_fwd_txfm_impl_sse2.h
AV1_COMMON_SRCS-$(HAVE_SSE4_1) += common/x86/av1_txfm1d_sse4.h
AV1_COMMON_SRCS-$(HAVE_SSE4_1) += common/x86/av1_fwd_txfm1d_sse4.c
AV1_COMMON_SRCS-$(HAVE_SSE4_1) += common/x86/av1_fwd_txfm2d_sse4.c
......@@ -143,7 +136,4 @@ ifeq ($(CONFIG_FILTER_INTRA),yes)
AV1_COMMON_SRCS-$(HAVE_SSE4_1) += common/x86/filterintra_sse4.c
endif
AV1_COMMON_SRCS-$(HAVE_SSE2) += common/x86/av1_inv_txfm_sse2.c
AV1_COMMON_SRCS-$(HAVE_SSE2) += common/x86/av1_inv_txfm_sse2.h
$(eval $(call rtcd_h_template,av1_rtcd,av1/common/av1_rtcd_defs.pl))
This diff is collapsed.
/*
* Copyright (c) 2016, Alliance for Open Media. All rights reserved
*
* This source code is subject to the terms of the BSD 2 Clause License and
* the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
* was not distributed with this source code in the LICENSE file, you can
* obtain it at www.aomedia.org/license/software. If the Alliance for Open
* Media Patent License 1.0 was not distributed with this source code in the
* PATENTS file, you can obtain it at www.aomedia.org/license/patent.
*/
#ifndef AV1_COMMON_AV1_FWD_TXFM_H_
#define AV1_COMMON_AV1_FWD_TXFM_H_
#include "aom_dsp/txfm_common.h"
#include "aom_dsp/fwd_txfm.h"
void av1_fdct32(const tran_high_t *input, tran_high_t *output, int round);
#endif // AV1_COMMON_AV1_FWD_TXFM_H_
This diff is collapsed.
/*
* Copyright (c) 2016, Alliance for Open Media. All rights reserved
*
* This source code is subject to the terms of the BSD 2 Clause License and
* the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
* was not distributed with this source code in the LICENSE file, you can
* obtain it at www.aomedia.org/license/software. If the Alliance for Open
* Media Patent License 1.0 was not distributed with this source code in the
* PATENTS file, you can obtain it at www.aomedia.org/license/patent.
*/
#ifndef AOM_DSP_INV_TXFM_H_
#define AOM_DSP_INV_TXFM_H_
#include <assert.h>
#include "./aom_config.h"
#include "aom_dsp/txfm_common.h"
#include "aom_dsp/inv_txfm.h"
#include "aom_ports/mem.h"
#ifdef __cplusplus
extern "C" {
#endif
static INLINE tran_high_t check_range(tran_high_t input) {
#if CONFIG_COEFFICIENT_RANGE_CHECKING
// For valid input streams, intermediate stage coefficients should always
// stay within the range of a signed 16 bit integer. Coefficients can go out
// of this range for invalid/corrupt streams. However, strictly checking
// this range for every intermediate coefficient can burdensome for a decoder,
// therefore the following assertion is only enabled when configured with
// --enable-coefficient-range-checking.
assert(INT16_MIN <= input);
assert(input <= INT16_MAX);
#endif // CONFIG_COEFFICIENT_RANGE_CHECKING
return input;
}
static INLINE tran_high_t dct_const_round_shift(tran_high_t input) {
tran_high_t rv = ROUND_POWER_OF_TWO(input, DCT_CONST_BITS);
return rv;
}
#if CONFIG_AOM_HIGHBITDEPTH
static INLINE tran_high_t highbd_check_range(tran_high_t input, int bd) {
#if CONFIG_COEFFICIENT_RANGE_CHECKING
// For valid highbitdepth streams, intermediate stage coefficients will
// stay within the ranges:
// - 8 bit: signed 16 bit integer
// - 10 bit: signed 18 bit integer
// - 12 bit: signed 20 bit integer
const int32_t int_max = (1 << (7 + bd)) - 1;
const int32_t int_min = -int_max - 1;
assert(int_min <= input);
assert(input <= int_max);
(void)int_min;
#endif // CONFIG_COEFFICIENT_RANGE_CHECKING
(void)bd;
return input;
}
static INLINE tran_high_t highbd_dct_const_round_shift(tran_high_t input) {
tran_high_t rv = ROUND_POWER_OF_TWO(input, DCT_CONST_BITS);
return rv;
}
#endif // CONFIG_AOM_HIGHBITDEPTH
#if CONFIG_EMULATE_HARDWARE
// When CONFIG_EMULATE_HARDWARE is 1 the transform performs a
// non-normative method to handle overflows. A stream that causes
// overflows in the inverse transform is considered invalid,
// and a hardware implementer is free to choose any reasonable
// method to handle overflows. However to aid in hardware
// verification they can use a specific implementation of the
// WRAPLOW() macro below that is identical to their intended
// hardware implementation (and also use configure options to trigger
// the C-implementation of the transform).
//
// The particular WRAPLOW implementation below performs strict
// overflow wrapping to match common hardware implementations.
// bd of 8 uses trans_low with 16bits, need to remove 16bits
// bd of 10 uses trans_low with 18bits, need to remove 14bits
// bd of 12 uses trans_low with 20bits, need to remove 12bits
// bd of x uses trans_low with 8+x bits, need to remove 24-x bits
#define WRAPLOW(x) ((((int32_t)check_range(x)) << 16) >> 16)
#if CONFIG_AOM_HIGHBITDEPTH
#define HIGHBD_WRAPLOW(x, bd) \
((((int32_t)highbd_check_range((x), bd)) << (24 - bd)) >> (24 - bd))
#endif // CONFIG_AOM_HIGHBITDEPTH
#else // CONFIG_EMULATE_HARDWARE
#define WRAPLOW(x) ((int32_t)check_range(x))
#if CONFIG_AOM_HIGHBITDEPTH
#define HIGHBD_WRAPLOW(x, bd) ((int32_t)highbd_check_range((x), bd))
#endif // CONFIG_AOM_HIGHBITDEPTH
#endif // CONFIG_EMULATE_HARDWARE
void av1_idct4_c(const tran_low_t *input, tran_low_t *output);
void av1_idct8_c(const tran_low_t *input, tran_low_t *output);
void av1_idct16_c(const tran_low_t *input, tran_low_t *output);
void av1_idct32_c(const tran_low_t *input, tran_low_t *output);
void av1_iadst4_c(const tran_low_t *input, tran_low_t *output);
void av1_iadst8_c(const tran_low_t *input, tran_low_t *output);
void av1_iadst16_c(const tran_low_t *input, tran_low_t *output);
#if CONFIG_AOM_HIGHBITDEPTH
void av1_highbd_idct4_c(const tran_low_t *input, tran_low_t *output, int bd);
void av1_highbd_idct8_c(const tran_low_t *input, tran_low_t *output, int bd);
void av1_highbd_idct16_c(const tran_low_t *input, tran_low_t *output, int bd);
void av1_highbd_iadst4_c(const tran_low_t *input, tran_low_t *output, int bd);
void av1_highbd_iadst8_c(const tran_low_t *input, tran_low_t *output, int bd);
void av1_highbd_iadst16_c(const tran_low_t *input, tran_low_t *output, int bd);
static INLINE uint16_t highbd_clip_pixel_add(uint16_t dest, tran_high_t trans,
int bd) {
trans = HIGHBD_WRAPLOW(trans, bd);
return clip_pixel_highbd(dest + (int)trans, bd);
}
#endif
static INLINE uint8_t clip_pixel_add(uint8_t dest, tran_high_t trans) {
trans = WRAPLOW(trans);
return clip_pixel(dest + (int)trans);
}
#ifdef __cplusplus
} // extern "C"
#endif
#endif // AOM_DSP_INV_TXFM_H_
......@@ -414,62 +414,6 @@ if (aom_config("CONFIG_EXT_TX") eq "yes") {
specialize qw/av1_fht32x16 sse2/;
}
if (aom_config("CONFIG_EMULATE_HARDWARE") eq "yes") {
add_proto qw/void av1_fdct4x4/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_fdct4x4/;
add_proto qw/void av1_fdct4x4_1/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_fdct4x4_1/;
add_proto qw/void av1_fdct8x8/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_fdct8x8/;
add_proto qw/void av1_fdct8x8_1/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_fdct8x8_1/;
add_proto qw/void av1_fdct16x16/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_fdct16x16/;
add_proto qw/void av1_fdct16x16_1/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_fdct16x16_1/;
add_proto qw/void av1_fdct32x32/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_fdct32x32/;
add_proto qw/void av1_fdct32x32_rd/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_fdct32x32_rd/;
add_proto qw/void av1_fdct32x32_1/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_fdct32x32_1/;
} else {
add_proto qw/void av1_fdct4x4/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_fdct4x4 sse2/;
add_proto qw/void av1_fdct4x4_1/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_fdct4x4_1 sse2/;
add_proto qw/void av1_fdct8x8/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_fdct8x8 sse2/;
add_proto qw/void av1_fdct8x8_1/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_fdct8x8_1 sse2/;
add_proto qw/void av1_fdct16x16/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_fdct16x16 sse2/;
add_proto qw/void av1_fdct16x16_1/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_fdct16x16_1 sse2/;
add_proto qw/void av1_fdct32x32/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_fdct32x32 sse2/;
add_proto qw/void av1_fdct32x32_rd/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_fdct32x32_rd sse2/;
add_proto qw/void av1_fdct32x32_1/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_fdct32x32_1 sse2/;
}
if (aom_config("CONFIG_AOM_HIGHBITDEPTH") ne "yes") {
if (aom_config("CONFIG_EXT_TX") ne "yes") {
specialize qw/av1_fht4x4 msa/;
......@@ -478,243 +422,9 @@ if (aom_config("CONFIG_AOM_HIGHBITDEPTH") ne "yes") {
}
}
if (aom_config("CONFIG_AOM_HIGHBITDEPTH") eq "yes") {
if (aom_config("CONFIG_EMULATE_HARDWARE") eq "yes") {
add_proto qw/void av1_highbd_fdct4x4/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_highbd_fdct4x4/;
add_proto qw/void av1_highbd_fdct8x8/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_highbd_fdct8x8/;
add_proto qw/void av1_highbd_fdct8x8_1/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_highbd_fdct8x8_1/;
add_proto qw/void av1_highbd_fdct16x16/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_highbd_fdct16x16/;
add_proto qw/void av1_highbd_fdct16x16_1/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_highbd_fdct16x16_1/;
add_proto qw/void av1_highbd_fdct32x32/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_highbd_fdct32x32/;
add_proto qw/void av1_highbd_fdct32x32_rd/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_highbd_fdct32x32_rd/;
add_proto qw/void av1_highbd_fdct32x32_1/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_highbd_fdct32x32_1/;
} else {
add_proto qw/void av1_highbd_fdct4x4/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_highbd_fdct4x4 sse2/;
add_proto qw/void av1_highbd_fdct8x8/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_highbd_fdct8x8 sse2/;
add_proto qw/void av1_highbd_fdct8x8_1/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_highbd_fdct8x8_1/;
add_proto qw/void av1_highbd_fdct16x16/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_highbd_fdct16x16 sse2/;
add_proto qw/void av1_highbd_fdct16x16_1/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_highbd_fdct16x16_1/;
add_proto qw/void av1_highbd_fdct32x32/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_highbd_fdct32x32 sse2/;
add_proto qw/void av1_highbd_fdct32x32_rd/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_highbd_fdct32x32_rd sse2/;
add_proto qw/void av1_highbd_fdct32x32_1/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_highbd_fdct32x32_1/;
}
}
add_proto qw/void av1_fwd_idtx/, "const int16_t *src_diff, tran_low_t *coeff, int stride, int bs, int tx_type";
specialize qw/av1_fwd_idtx/;
# Inverse transform
if (aom_config("CONFIG_AOM_HIGHBITDEPTH") eq "yes") {
# Note as optimized versions of these functions are added we need to add a check to ensure
# that when CONFIG_EMULATE_HARDWARE is on, it defaults to the C versions only.
add_proto qw/void av1_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct4x4_1_add/;
add_proto qw/void av1_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct4x4_16_add/;
add_proto qw/void av1_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct8x8_1_add/;
add_proto qw/void av1_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct8x8_64_add/;
add_proto qw/void av1_idct8x8_12_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct8x8_12_add/;
add_proto qw/void av1_idct16x16_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct16x16_1_add/;
add_proto qw/void av1_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct16x16_256_add/;
add_proto qw/void av1_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct16x16_10_add/;
add_proto qw/void av1_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct32x32_1024_add/;
add_proto qw/void av1_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct32x32_34_add/;
add_proto qw/void av1_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct32x32_1_add/;
add_proto qw/void av1_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_iwht4x4_1_add/;
add_proto qw/void av1_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_iwht4x4_16_add/;
add_proto qw/void av1_highbd_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/av1_highbd_idct4x4_1_add/;
add_proto qw/void av1_highbd_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/av1_highbd_idct8x8_1_add/;
add_proto qw/void av1_highbd_idct16x16_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/av1_highbd_idct16x16_1_add/;
add_proto qw/void av1_highbd_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/av1_highbd_idct32x32_1024_add/;
add_proto qw/void av1_highbd_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/av1_highbd_idct32x32_34_add/;
add_proto qw/void av1_highbd_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/av1_highbd_idct32x32_1_add/;
add_proto qw/void av1_highbd_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/av1_highbd_iwht4x4_1_add/;
add_proto qw/void av1_highbd_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/av1_highbd_iwht4x4_16_add/;
# Force C versions if CONFIG_EMULATE_HARDWARE is 1
if (aom_config("CONFIG_EMULATE_HARDWARE") eq "yes") {
add_proto qw/void av1_highbd_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/av1_highbd_idct4x4_16_add/;
add_proto qw/void av1_highbd_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/av1_highbd_idct8x8_64_add/;
add_proto qw/void av1_highbd_idct8x8_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/av1_highbd_idct8x8_10_add/;
add_proto qw/void av1_highbd_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/av1_highbd_idct16x16_256_add/;
add_proto qw/void av1_highbd_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/av1_highbd_idct16x16_10_add/;
} else {
add_proto qw/void av1_highbd_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/av1_highbd_idct4x4_16_add sse2/;
add_proto qw/void av1_highbd_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/av1_highbd_idct8x8_64_add sse2/;
add_proto qw/void av1_highbd_idct8x8_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/av1_highbd_idct8x8_10_add sse2/;
add_proto qw/void av1_highbd_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/av1_highbd_idct16x16_256_add sse2/;
add_proto qw/void av1_highbd_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/av1_highbd_idct16x16_10_add sse2/;
} # CONFIG_EMULATE_HARDWARE
} else {
# Force C versions if CONFIG_EMULATE_HARDWARE is 1
if (aom_config("CONFIG_EMULATE_HARDWARE") eq "yes") {
add_proto qw/void av1_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct4x4_1_add/;
add_proto qw/void av1_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct4x4_16_add/;
add_proto qw/void av1_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct8x8_1_add/;
add_proto qw/void av1_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct8x8_64_add/;
add_proto qw/void av1_idct8x8_12_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct8x8_12_add/;
add_proto qw/void av1_idct16x16_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct16x16_1_add/;
add_proto qw/void av1_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct16x16_256_add/;
add_proto qw/void av1_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct16x16_10_add/;
add_proto qw/void av1_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct32x32_1024_add/;
add_proto qw/void av1_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct32x32_34_add/;
add_proto qw/void av1_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct32x32_1_add/;
add_proto qw/void av1_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_iwht4x4_1_add/;
add_proto qw/void av1_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_iwht4x4_16_add/;
} else {
add_proto qw/void av1_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct4x4_1_add sse2/;
add_proto qw/void av1_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct4x4_16_add sse2/;
add_proto qw/void av1_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct8x8_1_add sse2/;
add_proto qw/void av1_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct8x8_64_add sse2/;
add_proto qw/void av1_idct8x8_12_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct8x8_12_add sse2/;
add_proto qw/void av1_idct16x16_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct16x16_1_add sse2/;
add_proto qw/void av1_idct16x16_256_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct16x16_256_add sse2/;
add_proto qw/void av1_idct16x16_10_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct16x16_10_add sse2/;
add_proto qw/void av1_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct32x32_1024_add sse2/;
add_proto qw/void av1_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct32x32_34_add sse2/;
add_proto qw/void av1_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_idct32x32_1_add sse2/;
add_proto qw/void av1_iwht4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_iwht4x4_1_add/;
add_proto qw/void av1_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/av1_iwht4x4_16_add/;
} # CONFIG_EMULATE_HARDWARE
} # CONFIG_AOM_HIGHBITDEPTH
if (aom_config("CONFIG_AOM_HIGHBITDEPTH") eq "yes") {
#fwd txfm
add_proto qw/void av1_fwd_txfm2d_4x4/, "const int16_t *input, int32_t *output, int stride, int tx_type, int bd";
......
This diff is collapsed.
This diff is collapsed.
......@@ -28,13 +28,7 @@ typedef void (*TxfmFuncSSE2)(const __m128i *input, __m128i *output,
static INLINE TxfmFuncSSE2 fwd_txfm_type_to_func(TXFM_TYPE txfm_type) {
switch (txfm_type) {
case TXFM_TYPE_DCT4: return av1_fdct4_new_sse4_1; break;
case TXFM_TYPE_DCT8: return av1_fdct8_new_sse4_1; break;
case TXFM_TYPE_DCT16: return av1_fdct16_new_sse4_1; break;
case TXFM_TYPE_DCT32: return av1_fdct32_new_sse4_1; break;
case TXFM_TYPE_ADST4: return av1_fadst4_new_sse4_1; break;
case TXFM_TYPE_ADST8: return av1_fadst8_new_sse4_1; break;
case TXFM_TYPE_ADST16: return av1_fadst16_new_sse4_1; break;
case TXFM_TYPE_ADST32: return av1_fadst32_new_sse4_1; break;
default: assert(0);
}
......
This diff is collapsed.
/*
* Copyright (c) 2016, Alliance for Open Media. All rights reserved
*
* This source code is subject to the terms of the BSD 2 Clause License and
* the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
* was not distributed with this source code in the LICENSE file, you can
* obtain it at www.aomedia.org/license/software. If the Alliance for Open
* Media Patent License 1.0 was not distributed with this source code in the
* PATENTS file, you can obtain it at www.aomedia.org/license/patent.
*/
#include <emmintrin.h> // SSE2
#include "./aom_config.h"
#include "aom_dsp/aom_dsp_common.h"
#include "aom_dsp/x86/fwd_txfm_sse2.h"
void av1_fdct4x4_1_sse2(const int16_t *input, tran_low_t *output, int stride) {
__m128i in0, in1;
__m128i tmp;
const __m128i zero = _mm_setzero_si128();
in0 = _mm_loadl_epi64((const __m128i *)(input + 0 * stride));
in1 = _mm_loadl_epi64((const __m128i *)(input + 1 * stride));
in1 = _mm_unpacklo_epi64(
in1, _mm_loadl_epi64((const __m128i *)(input + 2 * stride)));
in0 = _mm_unpacklo_epi64(
in0, _mm_loadl_epi64((const __m128i *)(input + 3 * stride)));
tmp = _mm_add_epi16(in0, in1);
in0 = _mm_unpacklo_epi16(zero, tmp);
in1 = _mm_unpackhi_epi16(zero, tmp);
in0 = _mm_srai_epi32(in0, 16);
in1 = _mm_srai_epi32(in1, 16);
tmp = _mm_add_epi32(in0, in1);
in0 = _mm_unpacklo_epi32(tmp, zero);
in1 = _mm_unpackhi_epi32(tmp, zero);
tmp = _mm_add_epi32(in0, in1);
in0 = _mm_srli_si128(tmp, 8);
in1 = _mm_add_epi32(tmp, in0);
in0 = _mm_slli_epi32(in1, 1);
store_output(&in0, output);
}
void av1_fdct8x8_1_sse2(const int16_t *input, tran_low_t *output, int stride) {
__m128i in0 = _mm_load_si128((const __m128i *)(input + 0 * stride));
__m128i in1 = _mm_load_si128((const __m128i *)(input + 1 * stride));
__m128i in2 = _mm_load_si128((const __m128i *)(input + 2 * stride));
__m128i in3 = _mm_load_si128((const __m128i *)(input + 3 * stride));
__m128i u0, u1, sum;
u0 = _mm_add_epi16(in0, in1);
u1 = _mm_add_epi16(in2, in3);
in0 = _mm_load_si128((const __m128i *)(input + 4 * stride));
in1 = _mm_load_si128((const __m128i *)(input + 5 * stride));
in2 = _mm_load_si128((const __m128i *)(input + 6 * stride));
in3 = _mm_load_si128((const __m128i *)(input + 7 * stride));
sum = _mm_add_epi16(u0, u1);
in0 = _mm_add_epi16(in0, in1);
in2 = _mm_add_epi16(in2, in3);
sum = _mm_add_epi16(sum, in0);
u0 = _mm_setzero_si128();
sum = _mm_add_epi16(sum, in2);
in0 = _mm_unpacklo_epi16(u0, sum);
in1 = _mm_unpackhi_epi16(u0, sum);
in0 = _mm_srai_epi32(in0, 16);
in1 = _mm_srai_epi32(in1, 16);
sum = _mm_add_epi32(in0, in1);
in0 = _mm_unpacklo_epi32(sum, u0);
in1 = _mm_unpackhi_epi32(sum, u0);
sum = _mm_add_epi32(in0, in1);
in0 = _mm_srli_si128(sum, 8);
in1 = _mm_add_epi32(sum, in0);
store_output(&in1, output);
}
void av1_fdct16x16_1_sse2(const int16_t *input, tran_low_t *output,
int stride) {
__m128i in0, in1, in2, in3;
__m128i u0, u1;