#include "libavcodec/vp9dsp.h"
#include "libavutil/loongarch/loongson_intrinsics.h"
#include "vp9dsp_loongarch.h"
#include "libavutil/attributes.h"

Macros
#define	VP9_DCT_CONST_BITS 14

#define	ALLOC_ALIGNED(align) __attribute__ ((aligned(align)))

#define	ROUND_POWER_OF_TWO(value, n) (((value) + (1 << ((n) - 1))) >> (n))

#define	VP9_DOTP_CONST_PAIR(reg0, reg1, cnst0, cnst1, out0, out1)

#define	VP9_SET_COSPI_PAIR(c0_h, c1_h)

#define	VP9_ADDBLK_ST8x4_UB(dst, dst_stride, in0, in1, in2, in3)

#define	VP9_UNPCK_UB_SH(in, out_h, out_l)

#define	VP9_ILVLTRANS4x8_H(in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, out2, out3, out4, out5, out6, out7)

#define	VP9_MADD(inp0, inp1, inp2, inp3, cst0, cst1, cst2, cst3, out0, out1, out2, out3)

#define	VP9_SET_CONST_PAIR(mask_h, idx1_h, idx2_h)

#define	VP9_IDCT8x8_1D(in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, out2, out3, out4, out5, out6, out7)

#define	SUB(a, b) __lsx_vsub_h(a, b)

Functions
static av_always_inline void	vp9_idct8x8_1_add_lsx (int16_t input, uint8_t dst, int32_t dst_stride)

static void	vp9_idct8x8_12_colcol_addblk_lsx (int16_t input, uint8_t dst, int32_t dst_stride)

static void	vp9_idct8x8_colcol_addblk_lsx (int16_t input, uint8_t dst, int32_t dst_stride)

static void	vp9_idct16_1d_columns_addblk_lsx (int16_t input, uint8_t dst, int32_t dst_stride)

static void	vp9_idct16_1d_columns_lsx (int16_t input, int16_t output)

static void	vp9_idct16x16_1_add_lsx (int16_t input, uint8_t dst, int32_t dst_stride)

static void	vp9_idct16x16_10_colcol_addblk_lsx (int16_t input, uint8_t dst, int32_t dst_stride)

static void	vp9_idct16x16_colcol_addblk_lsx (int16_t input, uint8_t dst, int32_t dst_stride)

static void	vp9_idct_butterfly_transpose_store (int16_t tmp_buf, int16_t tmp_eve_buf, int16_t tmp_odd_buf, int16_t dst)

static void	vp9_idct8x32_column_even_process_store (int16_t tmp_buf, int16_t tmp_eve_buf)

static void	vp9_idct8x32_column_odd_process_store (int16_t tmp_buf, int16_t tmp_odd_buf)

static void	vp9_idct8x32_column_butterfly_addblk (int16_t tmp_eve_buf, int16_t tmp_odd_buf, uint8_t *dst, int32_t dst_stride)

static void	vp9_idct8x32_1d_columns_addblk_lsx (int16_t input, uint8_t dst, int32_t dst_stride)

static void	vp9_idct8x32_1d_columns_lsx (int16_t input, int16_t output, int16_t *tmp_buf)

static void	vp9_idct32x32_1_add_lsx (int16_t input, uint8_t dst, int32_t dst_stride)

static void	vp9_idct32x32_34_colcol_addblk_lsx (int16_t input, uint8_t dst, int32_t dst_stride)

static void	vp9_idct32x32_colcol_addblk_lsx (int16_t input, uint8_t dst, int32_t dst_stride)

void	ff_idct_idct_8x8_add_lsx (uint8_t dst, ptrdiff_t stride, int16_t block, int eob)

void	ff_idct_idct_16x16_add_lsx (uint8_t dst, ptrdiff_t stride, int16_t block, int eob)

void	ff_idct_idct_32x32_add_lsx (uint8_t dst, ptrdiff_t stride, int16_t block, int eob)

Variables
const int32_t	cospi_1_64 = 16364

const int32_t	cospi_2_64 = 16305

const int32_t	cospi_3_64 = 16207

const int32_t	cospi_4_64 = 16069

const int32_t	cospi_5_64 = 15893

const int32_t	cospi_6_64 = 15679

const int32_t	cospi_7_64 = 15426

const int32_t	cospi_8_64 = 15137

const int32_t	cospi_9_64 = 14811

const int32_t	cospi_10_64 = 14449

const int32_t	cospi_11_64 = 14053

const int32_t	cospi_12_64 = 13623

const int32_t	cospi_13_64 = 13160

const int32_t	cospi_14_64 = 12665

const int32_t	cospi_15_64 = 12140

const int32_t	cospi_16_64 = 11585

const int32_t	cospi_17_64 = 11003

const int32_t	cospi_18_64 = 10394

const int32_t	cospi_19_64 = 9760

const int32_t	cospi_20_64 = 9102

const int32_t	cospi_21_64 = 8423

const int32_t	cospi_22_64 = 7723

const int32_t	cospi_23_64 = 7005

const int32_t	cospi_24_64 = 6270

const int32_t	cospi_25_64 = 5520

const int32_t	cospi_26_64 = 4756

const int32_t	cospi_27_64 = 3981

const int32_t	cospi_28_64 = 3196

const int32_t	cospi_29_64 = 2404

const int32_t	cospi_30_64 = 1606

const int32_t	cospi_31_64 = 804

const int32_t	sinpi_1_9 = 5283

const int32_t	sinpi_2_9 = 9929

const int32_t	sinpi_3_9 = 13377

const int32_t	sinpi_4_9 = 15212

Macro Definition Documentation

◆ VP9_DCT_CONST_BITS

#define VP9_DCT_CONST_BITS 14

Definition at line 27 of file vp9_idct_lsx.c.

◆ ALLOC_ALIGNED

#define ALLOC_ALIGNED ( align ) __attribute__ ((aligned(align)))

Definition at line 28 of file vp9_idct_lsx.c.

◆ ROUND_POWER_OF_TWO

#define ROUND_POWER_OF_TWO	(	value,
		n
	)	(((value) + (1 << ((n) - 1))) >> (n))

Definition at line 29 of file vp9_idct_lsx.c.

◆ VP9_DOTP_CONST_PAIR

#define VP9_DOTP_CONST_PAIR	(	reg0,
		reg1,
		cnst0,
		cnst1,
		out0,
		out1
	)

Value:

{                                                                  \
    __m128i k0_m = __lsx_vreplgr2vr_h(cnst0);                      \
    __m128i s0_m, s1_m, s2_m, s3_m;                                \
                                                                   \
    s0_m = __lsx_vreplgr2vr_h(cnst1);                              \
    k0_m = __lsx_vpackev_h(s0_m, k0_m);                            \
                                                                   \
    s1_m = __lsx_vilvl_h(__lsx_vneg_h(reg1), reg0);                \
    s0_m = __lsx_vilvh_h(__lsx_vneg_h(reg1), reg0);                \
    s3_m = __lsx_vilvl_h(reg0, reg1);                              \
    s2_m = __lsx_vilvh_h(reg0, reg1);                              \
    DUP2_ARG2(__lsx_vdp2_w_h, s1_m, k0_m, s0_m, k0_m, s1_m, s0_m); \
    DUP2_ARG2(__lsx_vsrari_w, s1_m, VP9_DCT_CONST_BITS,            \
              s0_m, VP9_DCT_CONST_BITS, s1_m, s0_m);               \
    out0 = __lsx_vpickev_h(s0_m, s1_m);                            \
    DUP2_ARG2(__lsx_vdp2_w_h, s3_m, k0_m, s2_m, k0_m, s1_m, s0_m); \
    DUP2_ARG2(__lsx_vsrari_w, s1_m, VP9_DCT_CONST_BITS,            \
              s0_m, VP9_DCT_CONST_BITS, s1_m, s0_m);               \
    out1 = __lsx_vpickev_h(s0_m, s1_m);                            \
}

Definition at line 68 of file vp9_idct_lsx.c.

◆ VP9_SET_COSPI_PAIR

#define VP9_SET_COSPI_PAIR	(	c0_h,
		c1_h
	)

Value:

( {                                       \
    __m128i out0_m, r0_m, r1_m;           \
                                          \
    r0_m = __lsx_vreplgr2vr_h(c0_h);      \
    r1_m = __lsx_vreplgr2vr_h(c1_h);      \
    out0_m = __lsx_vpackev_h(r1_m, r0_m); \
                                          \
    out0_m;                               \
} )

Definition at line 90 of file vp9_idct_lsx.c.

◆ VP9_ADDBLK_ST8x4_UB

#define VP9_ADDBLK_ST8x4_UB	(	dst,
		dst_stride,
		in0,
		in1,
		in2,
		in3
	)

Value:

{                                                                     \
    uint8_t *dst_m = (uint8_t *) (dst);                               \
    __m128i dst0_m, dst1_m, dst2_m, dst3_m;                           \
    __m128i tmp0_m, tmp1_m;                                           \
    __m128i res0_m, res1_m, res2_m, res3_m;                           \
    __m128i zero_m = __lsx_vldi(0);                                   \
    DUP4_ARG2(__lsx_vld, dst_m, 0, dst_m + dst_stride, 0,             \
              dst_m + 2 * dst_stride, 0, dst_m + 3 * dst_stride, 0,   \
              dst0_m, dst1_m, dst2_m, dst3_m);                        \
    DUP4_ARG2(__lsx_vilvl_b, zero_m, dst0_m, zero_m, dst1_m, zero_m,  \
              dst2_m, zero_m, dst3_m, res0_m, res1_m, res2_m, res3_m);\
    DUP4_ARG2(__lsx_vadd_h, res0_m, in0, res1_m, in1, res2_m, in2,    \
              res3_m, in3, res0_m, res1_m, res2_m, res3_m);           \
    DUP4_ARG1(__lsx_vclip255_h, res0_m, res1_m, res2_m, res3_m,       \
              res0_m, res1_m, res2_m, res3_m);                        \
    DUP2_ARG2(__lsx_vpickev_b, res1_m, res0_m, res3_m, res2_m,        \
              tmp0_m, tmp1_m);                                        \
    __lsx_vstelm_d(tmp0_m, dst_m, 0, 0);                              \
    __lsx_vstelm_d(tmp0_m, dst_m + dst_stride, 0, 1);                 \
    __lsx_vstelm_d(tmp1_m, dst_m + 2 * dst_stride, 0, 0);             \
    __lsx_vstelm_d(tmp1_m, dst_m + 3 * dst_stride, 0, 1);             \
}

Definition at line 101 of file vp9_idct_lsx.c.

◆ VP9_UNPCK_UB_SH

#define VP9_UNPCK_UB_SH	(	in,
		out_h,
		out_l
	)

Value:

{                                         \
    __m128i zero = __lsx_vldi(0);         \
    out_l = __lsx_vilvl_b(zero, in);      \
    out_h = __lsx_vilvh_b(zero, in);      \
}

Definition at line 125 of file vp9_idct_lsx.c.

◆ VP9_ILVLTRANS4x8_H

#define VP9_ILVLTRANS4x8_H	(	in0,
		in1,
		in2,
		in3,
		in4,
		in5,
		in6,
		in7,
		out0,
		out1,
		out2,
		out3,
		out4,
		out5,
		out6,
		out7
	)

Value:

{                                                                           \
    __m128i tmp0_m, tmp1_m, tmp2_m, tmp3_m;                                 \
    __m128i tmp0_n, tmp1_n, tmp2_n, tmp3_n;                                 \
    __m128i zero_m = __lsx_vldi(0);                                         \
                                                                            \
    DUP4_ARG2(__lsx_vilvl_h, in1, in0, in3, in2, in5, in4, in7, in6,        \
              tmp0_n, tmp1_n, tmp2_n, tmp3_n);                              \
    tmp0_m = __lsx_vilvl_w(tmp1_n, tmp0_n);                                 \
    tmp2_m = __lsx_vilvh_w(tmp1_n, tmp0_n);                                 \
    tmp1_m = __lsx_vilvl_w(tmp3_n, tmp2_n);                                 \
    tmp3_m = __lsx_vilvh_w(tmp3_n, tmp2_n);                                 \
                                                                            \
    out0 = __lsx_vilvl_d(tmp1_m, tmp0_m);                                   \
    out1 = __lsx_vilvh_d(tmp1_m, tmp0_m);                                   \
    out2 = __lsx_vilvl_d(tmp3_m, tmp2_m);                                   \
    out3 = __lsx_vilvh_d(tmp3_m, tmp2_m);                                   \
                                                                            \
    out4 = zero_m;                                                          \
    out5 = zero_m;                                                          \
    out6 = zero_m;                                                          \
    out7 = zero_m;                                                          \
}

Definition at line 132 of file vp9_idct_lsx.c.

◆ VP9_MADD

#define VP9_MADD	(	inp0,
		inp1,
		inp2,
		inp3,
		cst0,
		cst1,
		cst2,
		cst3,
		out0,
		out1,
		out2,
		out3
	)

Value:

{                                                                           \
    __m128i madd_s0_m, madd_s1_m, madd_s2_m, madd_s3_m;                     \
    __m128i tmp0_m, tmp1_m, tmp2_m, tmp3_m;                                 \
                                                                            \
    madd_s1_m = __lsx_vilvl_h(inp1, inp0);                                  \
    madd_s0_m = __lsx_vilvh_h(inp1, inp0);                                  \
    madd_s3_m = __lsx_vilvl_h(inp3, inp2);                                  \
    madd_s2_m = __lsx_vilvh_h(inp3, inp2);                                  \
    DUP4_ARG2(__lsx_vdp2_w_h, madd_s1_m, cst0, madd_s0_m, cst0,             \
              madd_s1_m, cst1, madd_s0_m, cst1, tmp0_m, tmp1_m,             \
              tmp2_m, tmp3_m);                                              \
    DUP4_ARG2(__lsx_vsrari_w, tmp0_m, VP9_DCT_CONST_BITS, tmp1_m,           \
              VP9_DCT_CONST_BITS, tmp2_m, VP9_DCT_CONST_BITS, tmp3_m,       \
              VP9_DCT_CONST_BITS, tmp0_m, tmp1_m, tmp2_m, tmp3_m);          \
    DUP2_ARG2(__lsx_vpickev_h, tmp1_m, tmp0_m, tmp3_m, tmp2_m, out0, out1); \
    DUP4_ARG2(__lsx_vdp2_w_h, madd_s3_m, cst2, madd_s2_m, cst2, madd_s3_m,  \
              cst3, madd_s2_m, cst3, tmp0_m, tmp1_m, tmp2_m, tmp3_m);       \
    DUP4_ARG2(__lsx_vsrari_w, tmp0_m, VP9_DCT_CONST_BITS,                   \
              tmp1_m, VP9_DCT_CONST_BITS, tmp2_m, VP9_DCT_CONST_BITS,       \
              tmp3_m, VP9_DCT_CONST_BITS, tmp0_m, tmp1_m, tmp2_m, tmp3_m);  \
    DUP2_ARG2(__lsx_vpickev_h, tmp1_m, tmp0_m, tmp3_m, tmp2_m, out2, out3); \
}

Definition at line 158 of file vp9_idct_lsx.c.

◆ VP9_SET_CONST_PAIR

#define VP9_SET_CONST_PAIR	(	mask_h,
		idx1_h,
		idx2_h
	)

Value:

( {                                                                          \
    __m128i c0_m, c1_m;                                                      \
                                                                             \
    DUP2_ARG2(__lsx_vreplvei_h, mask_h, idx1_h, mask_h, idx2_h, c0_m, c1_m); \
    c0_m = __lsx_vpackev_h(c1_m, c0_m);                                      \
                                                                             \
    c0_m;                                                                    \
} )

Definition at line 183 of file vp9_idct_lsx.c.