ffmpeg/FFmpeg
1
0
Fork 0
mirror of https://git.ffmpeg.org/ffmpeg.git synced 2024-09-19 12:56:40 +00:00
Code Issues Packages Projects Releases Wiki Activity

avcodec/loongarch/h264chroma, vc1dsp_lasx: Add wrapper for __lasx_xvldx

Browse Source 
__lasx_xvldx does not accept a pointer to const (in fact,
no function in lasxintrin.h does so), although it is not allowed
to modify the pointed-to buffer. Therefore this commit adds a wrapper
for it in order to constify the H264Chroma API in a later commit.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
This commit is contained in:
Andreas Rheinhardt 2022-08-02 01:18:08 +02:00
parent b3bbbb14d0
commit 2c8dc7e953
3 changed files with 58 additions and 53 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

90
libavcodec/loongarch/h264chroma_lasx.c
View File

@ -51,7 +51,7 @@ static av_always_inline void avc_chroma_hv_8x4_lasx(uint8_t *src, uint8_t *dst,
__m256i coeff_vt_vec1 = __lasx_xvreplgr2vr_h(coef_ver1);
DUP2_ARG2(__lasx_xvld, chroma_mask_arr, 0, src, 0, mask, src0);
DUP4_ARG2(__lasx_xvldx, src, stride, src, stride_2x, src, stride_3x, src, stride_4x,
DUP4_ARG2(LASX_XVLDX, src, stride, src, stride_2x, src, stride_3x, src, stride_4x,
src1, src2, src3, src4);
DUP2_ARG3(__lasx_xvpermi_q, src2, src1, 0x20, src4, src3, 0x20, src1, src3);
src0 = __lasx_xvshuf_b(src0, src0, mask);
@ -91,10 +91,10 @@ static av_always_inline void avc_chroma_hv_8x8_lasx(uint8_t *src, uint8_t *dst,
__m256i coeff_vt_vec1 = __lasx_xvreplgr2vr_h(coef_ver1);
DUP2_ARG2(__lasx_xvld, chroma_mask_arr, 0, src, 0, mask, src0);
DUP4_ARG2(__lasx_xvldx, src, stride, src, stride_2x, src, stride_3x, src, stride_4x,
DUP4_ARG2(LASX_XVLDX, src, stride, src, stride_2x, src, stride_3x, src, stride_4x,
src1, src2, src3, src4);
src += stride_4x;
DUP4_ARG2(__lasx_xvldx, src, stride, src, stride_2x, src, stride_3x, src, stride_4x,
DUP4_ARG2(LASX_XVLDX, src, stride, src, stride_2x, src, stride_3x, src, stride_4x,
src5, src6, src7, src8);
DUP4_ARG3(__lasx_xvpermi_q, src2, src1, 0x20, src4, src3, 0x20, src6, src5, 0x20,
src8, src7, 0x20, src1, src3, src5, src7);
@ -141,8 +141,8 @@ static av_always_inline void avc_chroma_hz_8x4_lasx(uint8_t *src, uint8_t *dst,
coeff_vec = __lasx_xvslli_b(coeff_vec, 3);
DUP2_ARG2(__lasx_xvld, chroma_mask_arr, 0, src, 0, mask, src0);
DUP2_ARG2(__lasx_xvldx, src, stride, src, stride_2x, src1, src2);
src3 = __lasx_xvldx(src, stride_3x);
DUP2_ARG2(LASX_XVLDX, src, stride, src, stride_2x, src1, src2);
src3 = LASX_XVLDX(src, stride_3x);
DUP2_ARG3(__lasx_xvpermi_q, src1, src0, 0x20, src3, src2, 0x20, src0, src2);
DUP2_ARG3(__lasx_xvshuf_b, src0, src0, mask, src2, src2, mask, src0, src2);
DUP2_ARG2(__lasx_xvdp2_h_bu, src0, coeff_vec, src2, coeff_vec, res0, res1);
@ -170,11 +170,11 @@ static av_always_inline void avc_chroma_hz_8x8_lasx(uint8_t *src, uint8_t *dst,
coeff_vec = __lasx_xvslli_b(coeff_vec, 3);
DUP2_ARG2(__lasx_xvld, chroma_mask_arr, 0, src, 0, mask, src0);
DUP4_ARG2(__lasx_xvldx, src, stride, src, stride_2x, src, stride_3x, src, stride_4x,
DUP4_ARG2(LASX_XVLDX, src, stride, src, stride_2x, src, stride_3x, src, stride_4x,
src1, src2, src3, src4);
src += stride_4x;
DUP2_ARG2(__lasx_xvldx, src, stride, src, stride_2x, src5, src6);
src7 = __lasx_xvldx(src, stride_3x);
DUP2_ARG2(LASX_XVLDX, src, stride, src, stride_2x, src5, src6);
src7 = LASX_XVLDX(src, stride_3x);
DUP4_ARG3(__lasx_xvpermi_q, src1, src0, 0x20, src3, src2, 0x20, src5, src4, 0x20,
src7, src6, 0x20, src0, src2, src4, src6);
DUP4_ARG3(__lasx_xvshuf_b, src0, src0, mask, src2, src2, mask, src4, src4, mask,
@ -212,7 +212,7 @@ static av_always_inline void avc_chroma_hz_nonmult_lasx(uint8_t *src,
coeff_vec = __lasx_xvslli_b(coeff_vec, 3);
for (row = height >> 2; row--;) {
DUP4_ARG2(__lasx_xvldx, src, 0, src, stride, src, stride_2x, src, stride_3x,
DUP4_ARG2(LASX_XVLDX, src, 0, src, stride, src, stride_2x, src, stride_3x,
src0, src1, src2, src3);
src += stride_4x;
DUP2_ARG3(__lasx_xvpermi_q, src1, src0, 0x20, src3, src2, 0x20, src0, src2);
@ -228,7 +228,7 @@ static av_always_inline void avc_chroma_hz_nonmult_lasx(uint8_t *src,
if ((height & 3)) {
src0 = __lasx_xvld(src, 0);
src1 = __lasx_xvldx(src, stride);
src1 = LASX_XVLDX(src, stride);
src1 = __lasx_xvpermi_q(src1, src0, 0x20);
src0 = __lasx_xvshuf_b(src1, src1, mask);
res0 = __lasx_xvdp2_h_bu(src0, coeff_vec);
@ -253,7 +253,7 @@ static av_always_inline void avc_chroma_vt_8x4_lasx(uint8_t *src, uint8_t *dst,
coeff_vec = __lasx_xvslli_b(coeff_vec, 3);
src0 = __lasx_xvld(src, 0);
src += stride;
DUP4_ARG2(__lasx_xvldx, src, 0, src, stride, src, stride_2x, src, stride_3x,
DUP4_ARG2(LASX_XVLDX, src, 0, src, stride, src, stride_2x, src, stride_3x,
src1, src2, src3, src4);
DUP4_ARG3(__lasx_xvpermi_q, src1, src0, 0x20, src2, src1, 0x20, src3, src2, 0x20,
src4, src3, 0x20, src0, src1, src2, src3);
@ -282,10 +282,10 @@ static av_always_inline void avc_chroma_vt_8x8_lasx(uint8_t *src, uint8_t *dst,
coeff_vec = __lasx_xvslli_b(coeff_vec, 3);
src0 = __lasx_xvld(src, 0);
src += stride;
DUP4_ARG2(__lasx_xvldx, src, 0, src, stride, src, stride_2x, src, stride_3x,
DUP4_ARG2(LASX_XVLDX, src, 0, src, stride, src, stride_2x, src, stride_3x,
src1, src2, src3, src4);
src += stride_4x;
DUP4_ARG2(__lasx_xvldx, src, 0, src, stride, src, stride_2x, src, stride_3x,
DUP4_ARG2(LASX_XVLDX, src, 0, src, stride, src, stride_2x, src, stride_3x,
src5, src6, src7, src8);
DUP4_ARG3(__lasx_xvpermi_q, src1, src0, 0x20, src2, src1, 0x20, src3, src2, 0x20,
src4, src3, 0x20, src0, src1, src2, src3);
@ -402,7 +402,7 @@ static void avc_chroma_hv_4x2_lasx(uint8_t *src, uint8_t *dst, ptrdiff_t stride,
__m256i coeff_vt_vec = __lasx_xvpermi_q(coeff_vt_vec1, coeff_vt_vec0, 0x02);
DUP2_ARG2(__lasx_xvld, chroma_mask_arr, 32, src, 0, mask, src0);
DUP2_ARG2(__lasx_xvldx, src, stride, src, stride_2, src1, src2);
DUP2_ARG2(LASX_XVLDX, src, stride, src, stride_2, src1, src2);
DUP2_ARG3(__lasx_xvshuf_b, src1, src0, mask, src2, src1, mask, src0, src1);
src0 = __lasx_xvpermi_q(src0, src1, 0x02);
res_hz = __lasx_xvdp2_h_bu(src0, coeff_hz_vec);
@ -431,7 +431,7 @@ static void avc_chroma_hv_4x4_lasx(uint8_t *src, uint8_t *dst, ptrdiff_t stride,
__m256i coeff_vt_vec1 = __lasx_xvreplgr2vr_h(coef_ver1);
DUP2_ARG2(__lasx_xvld, chroma_mask_arr, 32, src, 0, mask, src0);
DUP4_ARG2(__lasx_xvldx, src, stride, src, stride_2, src, stride_3,
DUP4_ARG2(LASX_XVLDX, src, stride, src, stride_2, src, stride_3,
src, stride_4, src1, src2, src3, src4);
DUP4_ARG3(__lasx_xvshuf_b, src1, src0, mask, src2, src1, mask, src3, src2, mask,
src4, src3, mask, src0, src1, src2, src3);
@ -464,10 +464,10 @@ static void avc_chroma_hv_4x8_lasx(uint8_t *src, uint8_t * dst, ptrdiff_t stride
__m256i coeff_vt_vec1 = __lasx_xvreplgr2vr_h(coef_ver1);
DUP2_ARG2(__lasx_xvld, chroma_mask_arr, 32, src, 0, mask, src0);
DUP4_ARG2(__lasx_xvldx, src, stride, src, stride_2, src, stride_3,
DUP4_ARG2(LASX_XVLDX, src, stride, src, stride_2, src, stride_3,
src, stride_4, src1, src2, src3, src4);
src += stride_4;
DUP4_ARG2(__lasx_xvldx, src, stride, src, stride_2, src, stride_3,
DUP4_ARG2(LASX_XVLDX, src, stride, src, stride_2, src, stride_3,
src, stride_4, src5, src6, src7, src8);
DUP4_ARG3(__lasx_xvshuf_b, src1, src0, mask, src2, src1, mask, src3, src2, mask,
src4, src3, mask, src0, src1, src2, src3);
@ -519,7 +519,7 @@ static void avc_chroma_hz_4x2_lasx(uint8_t *src, uint8_t *dst, ptrdiff_t stride,
__m256i coeff_vec = __lasx_xvilvl_b(coeff_vec0, coeff_vec1);
DUP2_ARG2(__lasx_xvld, chroma_mask_arr, 32, src, 0, mask, src0);
src1 = __lasx_xvldx(src, stride);
src1 = LASX_XVLDX(src, stride);
src0 = __lasx_xvshuf_b(src1, src0, mask);
res = __lasx_xvdp2_h_bu(src0, coeff_vec);
res = __lasx_xvslli_h(res, 3);
@ -540,8 +540,8 @@ static void avc_chroma_hz_4x4_lasx(uint8_t *src, uint8_t *dst, ptrdiff_t stride,
__m256i coeff_vec = __lasx_xvilvl_b(coeff_vec0, coeff_vec1);
DUP2_ARG2(__lasx_xvld, chroma_mask_arr, 32, src, 0, mask, src0);
DUP2_ARG2(__lasx_xvldx, src, stride, src, stride_2, src1, src2);
src3 = __lasx_xvldx(src, stride_3);
DUP2_ARG2(LASX_XVLDX, src, stride, src, stride_2, src1, src2);
src3 = LASX_XVLDX(src, stride_3);
DUP2_ARG3(__lasx_xvshuf_b, src1, src0, mask, src3, src2, mask, src0, src2);
src0 = __lasx_xvpermi_q(src0, src2, 0x02);
res = __lasx_xvdp2_h_bu(src0, coeff_vec);
@ -567,11 +567,11 @@ static void avc_chroma_hz_4x8_lasx(uint8_t *src, uint8_t *dst, ptrdiff_t stride,
coeff_vec = __lasx_xvslli_b(coeff_vec, 3);
DUP2_ARG2(__lasx_xvld, chroma_mask_arr, 32, src, 0, mask, src0);
DUP4_ARG2(__lasx_xvldx, src, stride, src, stride_2, src, stride_3,
DUP4_ARG2(LASX_XVLDX, src, stride, src, stride_2, src, stride_3,
src, stride_4, src1, src2, src3, src4);
src += stride_4;
DUP2_ARG2(__lasx_xvldx, src, stride, src, stride_2, src5, src6);
src7 = __lasx_xvldx(src, stride_3);
DUP2_ARG2(LASX_XVLDX, src, stride, src, stride_2, src5, src6);
src7 = LASX_XVLDX(src, stride_3);
DUP4_ARG3(__lasx_xvshuf_b, src1, src0, mask, src3, src2, mask, src5, src4, mask,
src7, src6, mask, src0, src2, src4, src6);
DUP2_ARG3(__lasx_xvpermi_q, src0, src2, 0x02, src4, src6, 0x02, src0, src4);
@ -625,7 +625,7 @@ static void avc_chroma_vt_4x2_lasx(uint8_t *src, uint8_t *dst, ptrdiff_t stride,
__m256i coeff_vec = __lasx_xvilvl_b(coeff_vec0, coeff_vec1);
src0 = __lasx_xvld(src, 0);
DUP2_ARG2(__lasx_xvldx, src, stride, src, stride << 1, src1, src2);
DUP2_ARG2(LASX_XVLDX, src, stride, src, stride << 1, src1, src2);
DUP2_ARG2(__lasx_xvilvl_b, src1, src0, src2, src1, tmp0, tmp1);
tmp0 = __lasx_xvilvl_d(tmp1, tmp0);
res = __lasx_xvdp2_h_bu(tmp0, coeff_vec);
@ -649,7 +649,7 @@ static void avc_chroma_vt_4x4_lasx(uint8_t *src, uint8_t *dst, ptrdiff_t stride,
__m256i coeff_vec = __lasx_xvilvl_b(coeff_vec0, coeff_vec1);
src0 = __lasx_xvld(src, 0);
DUP4_ARG2(__lasx_xvldx, src, stride, src, stride_2, src, stride_3,
DUP4_ARG2(LASX_XVLDX, src, stride, src, stride_2, src, stride_3,
src, stride_4, src1, src2, src3, src4);
DUP4_ARG2(__lasx_xvilvl_b, src1, src0, src2, src1, src3, src2, src4, src3,
tmp0, tmp1, tmp2, tmp3);
@ -679,10 +679,10 @@ static void avc_chroma_vt_4x8_lasx(uint8_t *src, uint8_t *dst, ptrdiff_t stride,
coeff_vec = __lasx_xvslli_b(coeff_vec, 3);
src0 = __lasx_xvld(src, 0);
DUP4_ARG2(__lasx_xvldx, src, stride, src, stride_2, src, stride_3,
DUP4_ARG2(LASX_XVLDX, src, stride, src, stride_2, src, stride_3,
src, stride_4, src1, src2, src3, src4);
src += stride_4;
DUP4_ARG2(__lasx_xvldx, src, stride, src, stride_2, src, stride_3,
DUP4_ARG2(LASX_XVLDX, src, stride, src, stride_2, src, stride_3,
src, stride_4, src5, src6, src7, src8);
DUP4_ARG2(__lasx_xvilvl_b, src1, src0, src2, src1, src3, src2, src4, src3,
tmp0, tmp1, tmp2, tmp3);
@ -860,7 +860,7 @@ static av_always_inline void avc_chroma_hv_and_aver_dst_8x4_lasx(uint8_t *src,
__m256i coeff_vt_vec1 = __lasx_xvreplgr2vr_h(coef_ver1);
DUP2_ARG2(__lasx_xvld, chroma_mask_arr, 0, src, 0, mask, src0);
DUP4_ARG2(__lasx_xvldx, src, stride, src, stride_2x, src, stride_3x, src, stride_4x,
DUP4_ARG2(LASX_XVLDX, src, stride, src, stride_2x, src, stride_3x, src, stride_4x,
src1, src2, src3, src4);
DUP2_ARG3(__lasx_xvpermi_q, src2, src1, 0x20, src4, src3, 0x20, src1, src3);
src0 = __lasx_xvshuf_b(src0, src0, mask);
@ -874,7 +874,7 @@ static av_always_inline void avc_chroma_hv_and_aver_dst_8x4_lasx(uint8_t *src,
res_vt0 = __lasx_xvmadd_h(res_vt0, res_hz0, coeff_vt_vec1);
res_vt1 = __lasx_xvmadd_h(res_vt1, res_hz1, coeff_vt_vec1);
out = __lasx_xvssrarni_bu_h(res_vt1, res_vt0, 6);
DUP4_ARG2(__lasx_xvldx, dst, 0, dst, stride, dst, stride_2x, dst, stride_3x,
DUP4_ARG2(LASX_XVLDX, dst, 0, dst, stride, dst, stride_2x, dst, stride_3x,
tp0, tp1, tp2, tp3);
DUP2_ARG2(__lasx_xvilvl_d, tp2, tp0, tp3, tp1, tp0, tp2);
tp0 = __lasx_xvpermi_q(tp2, tp0, 0x20);
@ -907,10 +907,10 @@ static av_always_inline void avc_chroma_hv_and_aver_dst_8x8_lasx(uint8_t *src,
DUP2_ARG2(__lasx_xvld, chroma_mask_arr, 0, src, 0, mask, src0);
src += stride;
DUP4_ARG2(__lasx_xvldx, src, 0, src, stride, src, stride_2x, src, stride_3x,
DUP4_ARG2(LASX_XVLDX, src, 0, src, stride, src, stride_2x, src, stride_3x,
src1, src2, src3, src4);
src += stride_4x;
DUP4_ARG2(__lasx_xvldx, src, 0, src, stride, src, stride_2x, src, stride_3x,
DUP4_ARG2(LASX_XVLDX, src, 0, src, stride, src, stride_2x, src, stride_3x,
src5, src6, src7, src8);
DUP4_ARG3(__lasx_xvpermi_q, src2, src1, 0x20, src4, src3, 0x20, src6, src5, 0x20,
src8, src7, 0x20, src1, src3, src5, src7);
@ -934,12 +934,12 @@ static av_always_inline void avc_chroma_hv_and_aver_dst_8x8_lasx(uint8_t *src,
res_vt3 = __lasx_xvmadd_h(res_vt3, res_hz3, coeff_vt_vec1);
DUP2_ARG3(__lasx_xvssrarni_bu_h, res_vt1, res_vt0, 6, res_vt3, res_vt2, 6,
out0, out1);
DUP4_ARG2(__lasx_xvldx, dst, 0, dst, stride, dst, stride_2x, dst, stride_3x,
DUP4_ARG2(LASX_XVLDX, dst, 0, dst, stride, dst, stride_2x, dst, stride_3x,
tp0, tp1, tp2, tp3);
DUP2_ARG2(__lasx_xvilvl_d, tp2, tp0, tp3, tp1, tp0, tp2);
dst0 = __lasx_xvpermi_q(tp2, tp0, 0x20);
dst += stride_4x;
DUP4_ARG2(__lasx_xvldx, dst, 0, dst, stride, dst, stride_2x, dst, stride_3x,
DUP4_ARG2(LASX_XVLDX, dst, 0, dst, stride, dst, stride_2x, dst, stride_3x,
tp0, tp1, tp2, tp3);
dst -= stride_4x;
DUP2_ARG2(__lasx_xvilvl_d, tp2, tp0, tp3, tp1, tp0, tp2);
@ -973,13 +973,13 @@ static av_always_inline void avc_chroma_hz_and_aver_dst_8x4_lasx(uint8_t *src,
coeff_vec = __lasx_xvslli_b(coeff_vec, 3);
mask = __lasx_xvld(chroma_mask_arr, 0);
DUP4_ARG2(__lasx_xvldx, src, 0, src, stride, src, stride_2x, src, stride_3x,
DUP4_ARG2(LASX_XVLDX, src, 0, src, stride, src, stride_2x, src, stride_3x,
src0, src1, src2, src3);
DUP2_ARG3(__lasx_xvpermi_q, src1, src0, 0x20, src3, src2, 0x20, src0, src2);
DUP2_ARG3(__lasx_xvshuf_b, src0, src0, mask, src2, src2, mask, src0, src2);
DUP2_ARG2(__lasx_xvdp2_h_bu, src0, coeff_vec, src2, coeff_vec, res0, res1);
out = __lasx_xvssrarni_bu_h(res1, res0, 6);
DUP4_ARG2(__lasx_xvldx, dst, 0, dst, stride, dst, stride_2x, dst, stride_3x,
DUP4_ARG2(LASX_XVLDX, dst, 0, dst, stride, dst, stride_2x, dst, stride_3x,
tp0, tp1, tp2, tp3);
DUP2_ARG2(__lasx_xvilvl_d, tp2, tp0, tp3, tp1, tp0, tp2);
tp0 = __lasx_xvpermi_q(tp2, tp0, 0x20);
@ -1008,10 +1008,10 @@ static av_always_inline void avc_chroma_hz_and_aver_dst_8x8_lasx(uint8_t *src,
coeff_vec = __lasx_xvslli_b(coeff_vec, 3);
mask = __lasx_xvld(chroma_mask_arr, 0);
DUP4_ARG2(__lasx_xvldx, src, 0, src, stride, src, stride_2x, src, stride_3x,
DUP4_ARG2(LASX_XVLDX, src, 0, src, stride, src, stride_2x, src, stride_3x,
src0, src1, src2, src3);
src += stride_4x;
DUP4_ARG2(__lasx_xvldx, src, 0, src, stride, src, stride_2x, src, stride_3x,
DUP4_ARG2(LASX_XVLDX, src, 0, src, stride, src, stride_2x, src, stride_3x,
src4, src5, src6, src7);
DUP4_ARG3(__lasx_xvpermi_q, src1, src0, 0x20, src3, src2, 0x20, src5, src4, 0x20,
src7, src6, 0x20, src0, src2, src4, src6);
@ -1020,12 +1020,12 @@ static av_always_inline void avc_chroma_hz_and_aver_dst_8x8_lasx(uint8_t *src,
DUP4_ARG2(__lasx_xvdp2_h_bu, src0, coeff_vec, src2, coeff_vec, src4, coeff_vec, src6,
coeff_vec, res0, res1, res2, res3);
DUP2_ARG3(__lasx_xvssrarni_bu_h, res1, res0, 6, res3, res2, 6, out0, out1);
DUP4_ARG2(__lasx_xvldx, dst, 0, dst, stride, dst, stride_2x, dst, stride_3x,
DUP4_ARG2(LASX_XVLDX, dst, 0, dst, stride, dst, stride_2x, dst, stride_3x,
tp0, tp1, tp2, tp3);
DUP2_ARG2(__lasx_xvilvl_d, tp2, tp0, tp3, tp1, tp0, tp2);
dst0 = __lasx_xvpermi_q(tp2, tp0, 0x20);
dst += stride_4x;
DUP4_ARG2(__lasx_xvldx, dst, 0, dst, stride, dst, stride_2x, dst, stride_3x,
DUP4_ARG2(LASX_XVLDX, dst, 0, dst, stride, dst, stride_2x, dst, stride_3x,
tp0, tp1, tp2, tp3);
dst -= stride_4x;
DUP2_ARG2(__lasx_xvilvl_d, tp2, tp0, tp3, tp1, tp0, tp2);
@ -1059,14 +1059,14 @@ static av_always_inline void avc_chroma_vt_and_aver_dst_8x4_lasx(uint8_t *src,
coeff_vec = __lasx_xvslli_b(coeff_vec, 3);
src0 = __lasx_xvld(src, 0);
DUP4_ARG2(__lasx_xvldx, src, stride, src, stride_2x, src, stride_3x, src, stride_4x,
DUP4_ARG2(LASX_XVLDX, src, stride, src, stride_2x, src, stride_3x, src, stride_4x,
src1, src2, src3, src4);
DUP4_ARG3(__lasx_xvpermi_q, src1, src0, 0x20, src2, src1, 0x20, src3, src2, 0x20,
src4, src3, 0x20, src0, src1, src2, src3);
DUP2_ARG2(__lasx_xvilvl_b, src1, src0, src3, src2, src0, src2);
DUP2_ARG2(__lasx_xvdp2_h_bu, src0, coeff_vec, src2, coeff_vec, res0, res1);
out = __lasx_xvssrarni_bu_h(res1, res0, 6);
DUP4_ARG2(__lasx_xvldx, dst, 0, dst, stride, dst, stride_2x, dst, stride_3x,
DUP4_ARG2(LASX_XVLDX, dst, 0, dst, stride, dst, stride_2x, dst, stride_3x,
tp0, tp1, tp2, tp3);
DUP2_ARG2(__lasx_xvilvl_d, tp2, tp0, tp3, tp1, tp0, tp2);
tp0 = __lasx_xvpermi_q(tp2, tp0, 0x20);
@ -1095,10 +1095,10 @@ static av_always_inline void avc_chroma_vt_and_aver_dst_8x8_lasx(uint8_t *src,
coeff_vec = __lasx_xvslli_b(coeff_vec, 3);
src0 = __lasx_xvld(src, 0);
src += stride;
DUP4_ARG2(__lasx_xvldx, src, 0, src, stride, src, stride_2x, src, stride_3x,
DUP4_ARG2(LASX_XVLDX, src, 0, src, stride, src, stride_2x, src, stride_3x,
src1, src2, src3, src4);
src += stride_4x;
DUP4_ARG2(__lasx_xvldx, src, 0, src, stride, src, stride_2x, src, stride_3x,
DUP4_ARG2(LASX_XVLDX, src, 0, src, stride, src, stride_2x, src, stride_3x,
src5, src6, src7, src8);
DUP4_ARG3(__lasx_xvpermi_q, src1, src0, 0x20, src2, src1, 0x20, src3, src2, 0x20,
src4, src3, 0x20, src0, src1, src2, src3);
@ -1109,12 +1109,12 @@ static av_always_inline void avc_chroma_vt_and_aver_dst_8x8_lasx(uint8_t *src,
DUP4_ARG2(__lasx_xvdp2_h_bu, src0, coeff_vec, src2, coeff_vec, src4, coeff_vec, src6,
coeff_vec, res0, res1, res2, res3);
DUP2_ARG3(__lasx_xvssrarni_bu_h, res1, res0, 6, res3, res2, 6, out0, out1);
DUP4_ARG2(__lasx_xvldx, dst, 0, dst, stride, dst, stride_2x, dst, stride_3x,
DUP4_ARG2(LASX_XVLDX, dst, 0, dst, stride, dst, stride_2x, dst, stride_3x,
tp0, tp1, tp2, tp3);
DUP2_ARG2(__lasx_xvilvl_d, tp2, tp0, tp3, tp1, tp0, tp2);
dst0 = __lasx_xvpermi_q(tp2, tp0, 0x20);
dst += stride_4x;
DUP4_ARG2(__lasx_xvldx, dst, 0, dst, stride, dst, stride_2x, dst, stride_3x,
DUP4_ARG2(LASX_XVLDX, dst, 0, dst, stride, dst, stride_2x, dst, stride_3x,
tp0, tp1, tp2, tp3);
dst -= stride_4x;
DUP2_ARG2(__lasx_xvilvl_d, tp2, tp0, tp3, tp1, tp0, tp2);

16
libavcodec/loongarch/vc1dsp_lasx.c
View File

@ -831,20 +831,20 @@ static void put_vc1_mspel_mc_h_lasx(uint8_t *dst, const uint8_t *src,
const_para1_2 = __lasx_xvreplgr2vr_h(*(para_v + 1));
in0 = __lasx_xvld(_src, 0);
DUP2_ARG2(__lasx_xvldx, _src, stride, _src, stride2, in1, in2);
in3 = __lasx_xvldx(_src, stride3);
DUP2_ARG2(LASX_XVLDX, _src, stride, _src, stride2, in1, in2);
in3 = LASX_XVLDX(_src, stride3);
_src += stride4;
in4 = __lasx_xvld(_src, 0);
DUP2_ARG2(__lasx_xvldx, _src, stride, _src, stride2, in5, in6);
in7 = __lasx_xvldx(_src, stride3);
DUP2_ARG2(LASX_XVLDX, _src, stride, _src, stride2, in5, in6);
in7 = LASX_XVLDX(_src, stride3);
_src += stride4;
in8 = __lasx_xvld(_src, 0);
DUP2_ARG2(__lasx_xvldx, _src, stride, _src, stride2, in9, in10);
in11 = __lasx_xvldx(_src, stride3);
DUP2_ARG2(LASX_XVLDX, _src, stride, _src, stride2, in9, in10);
in11 = LASX_XVLDX(_src, stride3);
_src += stride4;
in12 = __lasx_xvld(_src, 0);
DUP2_ARG2(__lasx_xvldx, _src, stride, _src, stride2, in13, in14);
in15 = __lasx_xvldx(_src, stride3);
DUP2_ARG2(LASX_XVLDX, _src, stride, _src, stride2, in13, in14);
in15 = LASX_XVLDX(_src, stride3);
DUP4_ARG2(__lasx_xvilvl_b, in2, in0, in3, in1, in6, in4, in7, in5,
tmp0_m, tmp1_m, tmp2_m, tmp3_m);
DUP4_ARG2(__lasx_xvilvl_b, in10, in8, in11, in9, in14, in12, in15, in13,

5
libavutil/loongarch/loongson_intrinsics.h
View File

@ -721,6 +721,11 @@ static inline __m128i __lsx_vclip255_w(__m128i _in) {
#ifdef __loongarch_asx
#include <lasxintrin.h>
/* __lasx_xvldx() in lasxintrin.h does not accept a const void*;
* remove the following once it does. */
#define LASX_XVLDX(ptr, stride) __lasx_xvldx((void*)ptr, stride)
/*
* =============================================================================
* Description : Dot product of byte vector elements