avutil/mips: refactor msa SLDI_Bn_0 and SLDI_Bn macros.

Changing details as following: 1. The previous order of parameters are irregular and difficult to understand. Adjust the order of the parameters according to the rule: (RTYPE, input registers, input mask/input index/..., output registers). Most of the existing msa macros follow the rule. 2. Remove the redundant macro SLDI_Bn_0 and use SLDI_Bn instead.

gxw committed 7 years ago

Commit 92fc0bf1 parent 77937a412 files · +222 −215Browse files

Unified · Split

12 files

libavcodec/mips/h264dsp_msa.c +5 −4
libavcodec/mips/h264qpel_msa.c +32 −32
libavcodec/mips/hevc_lpf_sao_msa.c +36 −34
libavcodec/mips/hevcpred_msa.c +16 −14
libavcodec/mips/hpeldsp_msa.c +42 −24
libavcodec/mips/me_cmp_msa.c +4 −4
libavcodec/mips/qpeldsp_msa.c +42 −42
libavcodec/mips/vp8_mc_msa.c +2 −2
libavcodec/mips/vp9_idct_msa.c +2 −1
libavcodec/mips/vp9_lpf_msa.c +2 −1
libavcodec/mips/vp9_mc_msa.c +8 −8
libavutil/mips/generic_macros_msa.h +31 −49

libavcodec/mips/h264dsp_msa.c+5−4

@@ -618,7 +618,7 @@ static void avc_biwgt_8x16_msa(uint8_t src, uint8_t dst, int32_t stride, · 617 unmodified lines
618	618	\
619	619	out0 = (v16u8) __msa_ilvr_b((v16i8) in1, (v16i8) in0); \
620	620	out1 = (v16u8) __msa_sldi_b(zero_m, (v16i8) out0, 2); \
621		− SLDI_B2_0_UB(out1, out2, out2, out3, 2); \
	621	+ SLDI_B2_UB(zero_m, out1, zero_m, out2, 2, out2, out3); \
622	622	}
623	623
624	624	#define AVC_LPF_H_2BYTE_CHROMA_422(src, stride, tc_val, alpha, beta, res) \
@@ -1023,7 +1023,8 @@ static void avc_h_loop_filter_luma_mbaff_intra_msa(uint8_t *src, int32_t stride, · 398 unmodified lines
1023	1023
1024	1024	ILVR_W2_SB(tmp2, tmp0, tmp3, tmp1, src6, src3);
1025	1025	ILVL_W2_SB(tmp2, tmp0, tmp3, tmp1, src1, src5);
1026		− SLDI_B4_0_SB(src6, src1, src3, src5, src0, src2, src4, src7, 8);
	1026	+ SLDI_B4_SB(zeros, src6, zeros, src1, zeros, src3, zeros, src5,
	1027	+ 8, src0, src2, src4, src7);
1027	1028
1028	1029	p0_asub_q0 = __msa_asub_u_b((v16u8) src2, (v16u8) src3);
1029	1030	p1_asub_p0 = __msa_asub_u_b((v16u8) src1, (v16u8) src2);
@@ -1114,10 +1115,10 @@ static void avc_h_loop_filter_luma_mbaff_intra_msa(uint8_t *src, int32_t stride, · 84 unmodified lines
1114	1115	ILVRL_H2_SH(zeros, dst2_x, tmp2, tmp3);
1115	1116
1116	1117	ILVR_W2_UB(tmp2, tmp0, tmp3, tmp1, dst0, dst4);
1117		− SLDI_B2_0_UB(dst0, dst4, dst1, dst5, 8);
	1118	+ SLDI_B2_UB(zeros, dst0, zeros, dst4, 8, dst1, dst5);
1118	1119	dst2_x = (v16u8) __msa_ilvl_w((v4i32) tmp2, (v4i32) tmp0);
1119	1120	dst2_y = (v16u8) __msa_ilvl_w((v4i32) tmp3, (v4i32) tmp1);
1120		− SLDI_B2_0_UB(dst2_x, dst2_y, dst3_x, dst3_y, 8);
	1121	+ SLDI_B2_UB(zeros, dst2_x, zeros, dst2_y, 8, dst3_x, dst3_y);
1121	1122
1122	1123	out0 = __msa_copy_u_w((v4i32) dst0, 0);
1123	1124	out1 = __msa_copy_u_h((v8i16) dst0, 2);

libavcodec/mips/h264qpel_msa.c+32−32

@@ -790,8 +790,8 @@ void ff_put_h264_qpel16_mc10_msa(uint8_t dst, const uint8_t src, · 789 unmodified lines
790	790	minus5b, res4, res5, res6, res7);
791	791	DPADD_SB4_SH(vec2, vec5, vec8, vec11, plus20b, plus20b, plus20b,
792	792	plus20b, res4, res5, res6, res7);
793		− SLDI_B2_SB(src1, src3, src0, src2, src0, src2, 2);
794		− SLDI_B2_SB(src5, src7, src4, src6, src4, src6, 2);
	793	+ SLDI_B4_SB(src1, src0, src3, src2, src5, src4, src7, src6, 2,
	794	+ src0, src2, src4, src6);
795	795	SRARI_H4_SH(res0, res1, res2, res3, 5);
796	796	SRARI_H4_SH(res4, res5, res6, res7, 5);
797	797	SAT_SH4_SH(res0, res1, res2, res3, 7);
@@ -858,8 +858,8 @@ void ff_put_h264_qpel16_mc30_msa(uint8_t dst, const uint8_t src, · 60 unmodified lines
858	858	minus5b, res4, res5, res6, res7);
859	859	DPADD_SB4_SH(vec2, vec5, vec8, vec11, plus20b, plus20b, plus20b,
860	860	plus20b, res4, res5, res6, res7);
861		− SLDI_B2_SB(src1, src3, src0, src2, src0, src2, 3);
862		− SLDI_B2_SB(src5, src7, src4, src6, src4, src6, 3);
	861	+ SLDI_B4_SB(src1, src0, src3, src2, src5, src4, src7, src6, 3,
	862	+ src0, src2, src4, src6);
863	863	SRARI_H4_SH(res0, res1, res2, res3, 5);
864	864	SRARI_H4_SH(res4, res5, res6, res7, 5);
865	865	SAT_SH4_SH(res0, res1, res2, res3, 7);
@@ -911,10 +911,10 @@ void ff_put_h264_qpel8_mc10_msa(uint8_t dst, const uint8_t src, · 45 unmodified lines
911	911	VSHF_B2_SB(src6, src6, src7, src7, mask2, mask2, vec10, vec11);
912	912	DPADD_SB4_SH(vec8, vec9, vec10, vec11, plus20b, plus20b, plus20b, plus20b,
913	913	res4, res5, res6, res7);
914		− SLDI_B2_SB(src0, src1, src0, src1, src0, src1, 2);
915		− SLDI_B2_SB(src2, src3, src2, src3, src2, src3, 2);
916		− SLDI_B2_SB(src4, src5, src4, src5, src4, src5, 2);
917		− SLDI_B2_SB(src6, src7, src6, src7, src6, src7, 2);
	914	+ SLDI_B4_SB(src0, src0, src1, src1, src2, src2, src3, src3, 2,
	915	+ src0, src1, src2, src3);
	916	+ SLDI_B4_SB(src4, src4, src5, src5, src6, src6, src7, src7, 2,
	917	+ src4, src5, src6, src7);
918	918	PCKEV_D2_SB(src1, src0, src3, src2, src0, src1);
919	919	PCKEV_D2_SB(src5, src4, src7, src6, src4, src5);
920	920	SRARI_H4_SH(res0, res1, res2, res3, 5);
@@ -966,10 +966,10 @@ void ff_put_h264_qpel8_mc30_msa(uint8_t dst, const uint8_t src, · 45 unmodified lines
966	966	VSHF_B2_SB(src6, src6, src7, src7, mask2, mask2, vec10, vec11);
967	967	DPADD_SB4_SH(vec8, vec9, vec10, vec11, plus20b, plus20b, plus20b, plus20b,
968	968	res4, res5, res6, res7);
969		− SLDI_B2_SB(src0, src1, src0, src1, src0, src1, 3);
970		− SLDI_B2_SB(src2, src3, src2, src3, src2, src3, 3);
971		− SLDI_B2_SB(src4, src5, src4, src5, src4, src5, 3);
972		− SLDI_B2_SB(src6, src7, src6, src7, src6, src7, 3);
	969	+ SLDI_B4_SB(src0, src0, src1, src1, src2, src2, src3, src3, 3,
	970	+ src0, src1, src2, src3);
	971	+ SLDI_B4_SB(src4, src4, src5, src5, src6, src6, src7, src7, 3,
	972	+ src4, src5, src6, src7);
973	973	PCKEV_D2_SB(src1, src0, src3, src2, src0, src1);
974	974	PCKEV_D2_SB(src5, src4, src7, src6, src4, src5);
975	975	SRARI_H4_SH(res0, res1, res2, res3, 5);
@@ -1007,8 +1007,8 @@ void ff_put_h264_qpel4_mc10_msa(uint8_t dst, const uint8_t src, · 31 unmodified lines
1007	1007	SRARI_H2_SH(res0, res1, 5);
1008	1008	SAT_SH2_SH(res0, res1, 7);
1009	1009	res = __msa_pckev_b((v16i8) res1, (v16i8) res0);
1010		− SLDI_B2_SB(src0, src1, src0, src1, src0, src1, 2);
1011		− SLDI_B2_SB(src2, src3, src2, src3, src2, src3, 2);
	1010	+ SLDI_B4_SB(src0, src0, src1, src1, src2, src2, src3, src3, 2,
	1011	+ src0, src1, src2, src3);
1012	1012	src0 = (v16i8) __msa_insve_w((v4i32) src0, 1, (v4i32) src1);
1013	1013	src1 = (v16i8) __msa_insve_w((v4i32) src2, 1, (v4i32) src3);
1014	1014	src0 = (v16i8) __msa_insve_d((v2i64) src0, 1, (v2i64) src1);
@@ -1038,8 +1038,8 @@ void ff_put_h264_qpel4_mc30_msa(uint8_t dst, const uint8_t src, · 23 unmodified lines
1038	1038	SRARI_H2_SH(res0, res1, 5);
1039	1039	SAT_SH2_SH(res0, res1, 7);
1040	1040	res = __msa_pckev_b((v16i8) res1, (v16i8) res0);
1041		− SLDI_B2_SB(src0, src1, src0, src1, src0, src1, 3);
1042		− SLDI_B2_SB(src2, src3, src2, src3, src2, src3, 3);
	1041	+ SLDI_B4_SB(src0, src0, src1, src1, src2, src2, src3, src3, 3,
	1042	+ src0, src1, src2, src3);
1043	1043	src0 = (v16i8) __msa_insve_w((v4i32) src0, 1, (v4i32) src1);
1044	1044	src1 = (v16i8) __msa_insve_w((v4i32) src2, 1, (v4i32) src3);
1045	1045	src0 = (v16i8) __msa_insve_d((v2i64) src0, 1, (v2i64) src1);
@@ -3194,8 +3194,8 @@ void ff_avg_h264_qpel16_mc10_msa(uint8_t dst, const uint8_t src, · 2148 unmodified lines
3194	3194	minus5b, res4, res5, res6, res7);
3195	3195	DPADD_SB4_SH(vec2, vec5, vec8, vec11, plus20b, plus20b, plus20b,
3196	3196	plus20b, res4, res5, res6, res7);
3197		− SLDI_B2_SB(src1, src3, src0, src2, src0, src2, 2);
3198		− SLDI_B2_SB(src5, src7, src4, src6, src4, src6, 2);
	3197	+ SLDI_B4_SB(src1, src0, src3, src2, src5, src4, src7, src6, 2,
	3198	+ src0, src2, src4, src6);
3199	3199	SRARI_H4_SH(res0, res1, res2, res3, 5);
3200	3200	SRARI_H4_SH(res4, res5, res6, res7, 5);
3201	3201	SAT_SH4_SH(res0, res1, res2, res3, 7);
@@ -3266,8 +3266,8 @@ void ff_avg_h264_qpel16_mc30_msa(uint8_t dst, const uint8_t src, · 64 unmodified lines
3266	3266	minus5b, res4, res5, res6, res7);
3267	3267	DPADD_SB4_SH(vec2, vec5, vec8, vec11, plus20b, plus20b, plus20b,
3268	3268	plus20b, res4, res5, res6, res7);
3269		− SLDI_B2_SB(src1, src3, src0, src2, src0, src2, 3);
3270		− SLDI_B2_SB(src5, src7, src4, src6, src4, src6, 3);
	3269	+ SLDI_B4_SB(src1, src0, src3, src2, src5, src4, src7, src6, 3,
	3270	+ src0, src2, src4, src6);
3271	3271	SRARI_H4_SH(res0, res1, res2, res3, 5);
3272	3272	SRARI_H4_SH(res4, res5, res6, res7, 5);
3273	3273	SAT_SH4_SH(res0, res1, res2, res3, 7);
@@ -3323,10 +3323,10 @@ void ff_avg_h264_qpel8_mc10_msa(uint8_t dst, const uint8_t src, · 49 unmodified lines
3323	3323	VSHF_B2_SB(src6, src6, src7, src7, mask2, mask2, vec10, vec11);
3324	3324	DPADD_SB4_SH(vec8, vec9, vec10, vec11, plus20b, plus20b, plus20b, plus20b,
3325	3325	res4, res5, res6, res7);
3326		− SLDI_B2_SB(src0, src1, src0, src1, src0, src1, 2);
3327		− SLDI_B2_SB(src2, src3, src2, src3, src2, src3, 2);
3328		− SLDI_B2_SB(src4, src5, src4, src5, src4, src5, 2);
3329		− SLDI_B2_SB(src6, src7, src6, src7, src6, src7, 2);
	3326	+ SLDI_B4_SB(src0, src0, src1, src1, src2, src2, src3, src3, 2,
	3327	+ src0, src1, src2, src3);
	3328	+ SLDI_B4_SB(src4, src4, src5, src5, src6, src6, src7, src7, 2,
	3329	+ src4, src5, src6, src7);
3330	3330	PCKEV_D2_SB(src1, src0, src3, src2, src0, src1);
3331	3331	PCKEV_D2_SB(src5, src4, src7, src6, src4, src5);
3332	3332	SRARI_H4_SH(res0, res1, res2, res3, 5);
@@ -3388,10 +3388,10 @@ void ff_avg_h264_qpel8_mc30_msa(uint8_t dst, const uint8_t src, · 55 unmodified lines
3388	3388	VSHF_B2_SB(src6, src6, src7, src7, mask2, mask2, vec10, vec11);
3389	3389	DPADD_SB4_SH(vec8, vec9, vec10, vec11, plus20b, plus20b, plus20b, plus20b,
3390	3390	res4, res5, res6, res7);
3391		− SLDI_B2_SB(src0, src1, src0, src1, src0, src1, 3);
3392		− SLDI_B2_SB(src2, src3, src2, src3, src2, src3, 3);
3393		− SLDI_B2_SB(src4, src5, src4, src5, src4, src5, 3);
3394		− SLDI_B2_SB(src6, src7, src6, src7, src6, src7, 3);
	3391	+ SLDI_B4_SB(src0, src0, src1, src1, src2, src2, src3, src3, 3,
	3392	+ src0, src1, src2, src3);
	3393	+ SLDI_B4_SB(src4, src4, src5, src5, src6, src6, src7, src7, 3,
	3394	+ src4, src5, src6, src7);
3395	3395	PCKEV_D2_SB(src1, src0, src3, src2, src0, src1);
3396	3396	PCKEV_D2_SB(src5, src4, src7, src6, src4, src5);
3397	3397	SRARI_H4_SH(res0, res1, res2, res3, 5);
@@ -3439,8 +3439,8 @@ void ff_avg_h264_qpel4_mc10_msa(uint8_t dst, const uint8_t src, · 41 unmodified lines
3439	3439	SRARI_H2_SH(out0, out1, 5);
3440	3440	SAT_SH2_SH(out0, out1, 7);
3441	3441	res = __msa_pckev_b((v16i8) out1, (v16i8) out0);
3442		− SLDI_B2_SB(src0, src1, src0, src1, src0, src1, 2);
3443		− SLDI_B2_SB(src2, src3, src2, src3, src2, src3, 2);
	3442	+ SLDI_B4_SB(src0, src0, src1, src1, src2, src2, src3, src3, 2,
	3443	+ src0, src1, src2, src3);
3444	3444	src0 = (v16i8) __msa_insve_w((v4i32) src0, 1, (v4i32) src1);
3445	3445	src1 = (v16i8) __msa_insve_w((v4i32) src2, 1, (v4i32) src3);
3446	3446	src0 = (v16i8) __msa_insve_d((v2i64) src0, 1, (v2i64) src1);
@@ -3475,8 +3475,8 @@ void ff_avg_h264_qpel4_mc30_msa(uint8_t dst, const uint8_t src, · 28 unmodified lines
3475	3475	SRARI_H2_SH(out0, out1, 5);
3476	3476	SAT_SH2_SH(out0, out1, 7);
3477	3477	res = __msa_pckev_b((v16i8) out1, (v16i8) out0);
3478		− SLDI_B2_SB(src0, src1, src0, src1, src0, src1, 3);
3479		− SLDI_B2_SB(src2, src3, src2, src3, src2, src3, 3);
	3478	+ SLDI_B4_SB(src0, src0, src1, src1, src2, src2, src3, src3, 3,
	3479	+ src0, src1, src2, src3);
3480	3480	src0 = (v16i8) __msa_insve_w((v4i32) src0, 1, (v4i32) src1);
3481	3481	src1 = (v16i8) __msa_insve_w((v4i32) src2, 1, (v4i32) src3);
3482	3482	src0 = (v16i8) __msa_insve_d((v2i64) src0, 1, (v2i64) src1);

libavcodec/mips/hevc_lpf_sao_msa.c+36−34

@@ -1357,6 +1357,7 @@ static void hevc_sao_edge_filter_0degree_8width_msa(uint8_t *dst, · 1356 unmodified lines
1357	1357	v16u8 cmp_minus10, diff_minus10, diff_minus11;
1358	1358	v16u8 src0, src1, dst0, src_minus10, src_minus11, src_plus10, src_plus11;
1359	1359	v16i8 offset, sao_offset = LD_SB(sao_offset_val);
	1360	+ v16i8 zeros = { 0 };
1360	1361
1361	1362	sao_offset = __msa_pckev_b(sao_offset, sao_offset);
1362	1363	src -= 1;
@@ -1367,8 +1368,8 @@ static void hevc_sao_edge_filter_0degree_8width_msa(uint8_t *dst, · 4 unmodified lines
1367	1368	for (height -= 2; height; height -= 2) {
1368	1369	src += (src_stride << 1);
1369	1370
1370		− SLDI_B2_0_UB(src_minus10, src_minus11, src0, src1, 1);
1371		− SLDI_B2_0_UB(src_minus10, src_minus11, src_plus10, src_plus11, 2);
	1371	+ SLDI_B2_UB(zeros, src_minus10, zeros, src_minus11, 1, src0, src1);
	1372	+ SLDI_B2_UB(zeros, src_minus10, zeros, src_minus11, 2, src_plus10, src_plus11);
1372	1373
1373	1374	PCKEV_D2_UB(src_minus11, src_minus10, src_plus11, src_plus10,
1374	1375	src_minus10, src_plus10);
@@ -1404,8 +1405,8 @@ static void hevc_sao_edge_filter_0degree_8width_msa(uint8_t *dst, · 29 unmodified lines
1404	1405	dst += dst_stride;
1405	1406	}
1406	1407
1407		− SLDI_B2_0_UB(src_minus10, src_minus11, src0, src1, 1);
1408		− SLDI_B2_0_UB(src_minus10, src_minus11, src_plus10, src_plus11, 2);
	1408	+ SLDI_B2_UB(zeros, src_minus10, zeros, src_minus11, 1, src0, src1);
	1409	+ SLDI_B2_UB(zeros, src_minus10, zeros, src_minus11, 2, src_plus10, src_plus11);
1409	1410
1410	1411	PCKEV_D2_UB(src_minus11, src_minus10, src_plus11, src_plus10, src_minus10,
1411	1412	src_plus10);
@@ -1473,14 +1474,12 @@ static void hevc_sao_edge_filter_0degree_16multiple_msa(uint8_t *dst, · 61 unmodified lines
1473	1474	dst_ptr = dst + v_cnt;
1474	1475	LD_UB4(src_minus1, src_stride, src10, src11, src12, src13);
1475	1476
1476		− SLDI_B2_SB(src10, src11, src_minus10, src_minus11, src_zero0,
1477		− src_zero1, 1);
1478		− SLDI_B2_SB(src12, src13, src_minus12, src_minus13, src_zero2,
1479		− src_zero3, 1);
1480		− SLDI_B2_SB(src10, src11, src_minus10, src_minus11, src_plus10,
1481		− src_plus11, 2);
1482		− SLDI_B2_SB(src12, src13, src_minus12, src_minus13, src_plus12,
1483		− src_plus13, 2);
	1477	+ SLDI_B4_SB(src10, src_minus10, src11, src_minus11,
	1478	+ src12, src_minus12, src13, src_minus13, 1,
	1479	+ src_zero0, src_zero1, src_zero2, src_zero3);
	1480	+ SLDI_B4_SB(src10, src_minus10, src11, src_minus11,
	1481	+ src12, src_minus12, src13, src_minus13, 2,
	1482	+ src_plus10, src_plus11, src_plus12, src_plus13);
1484	1483
1485	1484	cmp_minus10 = ((v16u8) src_zero0 == src_minus10);
1486	1485	cmp_plus10 = ((v16u8) src_zero0 == (v16u8) src_plus10);
@@ -1880,6 +1879,7 @@ static void hevc_sao_edge_filter_45degree_4width_msa(uint8_t *dst, · 393 unmodified lines
1880	1879	v16u8 src_minus11, src10, src11;
1881	1880	v16i8 src_plus0, src_zero0, src_plus1, src_zero1, dst0;
1882	1881	v8i16 offset_mask0, offset_mask1;
	1882	+ v16i8 zeros = { 0 };
1883	1883
1884	1884	sao_offset = __msa_pckev_b(sao_offset, sao_offset);
1885	1885
@@ -1892,8 +1892,8 @@ static void hevc_sao_edge_filter_45degree_4width_msa(uint8_t *dst, · 6 unmodified lines
1892	1892	for (height -= 2; height; height -= 2) {
1893	1893	src_orig += (src_stride << 1);
1894	1894
1895		− SLDI_B2_0_SB(src_minus11, src10, src_zero0, src_zero1, 1);
1896		− SLDI_B2_0_SB(src10, src11, src_plus0, src_plus1, 2);
	1895	+ SLDI_B2_SB(zeros, src_minus11, zeros, src10, 1, src_zero0, src_zero1);
	1896	+ SLDI_B2_SB(zeros, src10, zeros, src11, 2, src_plus0, src_plus1);
1897	1897
1898	1898	ILVR_B2_UB(src_plus0, src_minus10, src_plus1, src_minus11, src_minus10,
1899	1899	src_minus11);
@@ -1938,8 +1938,8 @@ static void hevc_sao_edge_filter_45degree_4width_msa(uint8_t *dst, · 38 unmodified lines
1938	1938	dst += dst_stride;
1939	1939	}
1940	1940
1941		− SLDI_B2_0_SB(src_minus11, src10, src_zero0, src_zero1, 1);
1942		− SLDI_B2_0_SB(src10, src11, src_plus0, src_plus1, 2);
	1941	+ SLDI_B2_SB(zeros, src_minus11, zeros, src10, 1, src_zero0, src_zero1);
	1942	+ SLDI_B2_SB(zeros, src10, zeros, src11, 2, src_plus0, src_plus1);
1943	1943
1944	1944	ILVR_B2_UB(src_plus0, src_minus10, src_plus1, src_minus11, src_minus10,
1945	1945	src_minus11);
@@ -1992,6 +1992,7 @@ static void hevc_sao_edge_filter_45degree_8width_msa(uint8_t *dst, · 46 unmodified lines
1992	1992	v16u8 src_minus10, src10, src_minus11, src11;
1993	1993	v16i8 src_zero0, src_plus10, src_zero1, src_plus11, dst0;
1994	1994	v8i16 offset_mask0, offset_mask1;
	1995	+ v16i8 zeros = { 0 };
1995	1996
1996	1997	sao_offset = __msa_pckev_b(sao_offset, sao_offset);
1997	1998	src_orig = src - 1;
@@ -2003,8 +2004,8 @@ static void hevc_sao_edge_filter_45degree_8width_msa(uint8_t *dst, · 5 unmodified lines
2003	2004	for (height -= 2; height; height -= 2) {
2004	2005	src_orig += (src_stride << 1);
2005	2006
2006		− SLDI_B2_0_SB(src_minus11, src10, src_zero0, src_zero1, 1);
2007		− SLDI_B2_0_SB(src10, src11, src_plus10, src_plus11, 2);
	2007	+ SLDI_B2_SB(zeros, src_minus11, zeros, src10, 1, src_zero0, src_zero1);
	2008	+ SLDI_B2_SB(zeros, src10, zeros, src11, 2, src_plus10, src_plus11);
2008	2009
2009	2010	ILVR_B2_UB(src_plus10, src_minus10, src_plus11, src_minus11,
2010	2011	src_minus10, src_minus11);
@@ -2048,8 +2049,8 @@ static void hevc_sao_edge_filter_45degree_8width_msa(uint8_t *dst, · 37 unmodified lines
2048	2049	dst += dst_stride;
2049	2050	}
2050	2051
2051		− SLDI_B2_0_SB(src_minus11, src10, src_zero0, src_zero1, 1);
2052		− SLDI_B2_0_SB(src10, src11, src_plus10, src_plus11, 2);
	2052	+ SLDI_B2_SB(zeros, src_minus11, zeros, src10, 1, src_zero0, src_zero1);
	2053	+ SLDI_B2_SB(zeros, src10, zeros, src11, 2, src_plus10, src_plus11);
2053	2054	ILVR_B2_UB(src_plus10, src_minus10, src_plus11, src_minus11, src_minus10,
2054	2055	src_minus11);
2055	2056	ILVR_B2_SB(src_zero0, src_zero0, src_zero1, src_zero1, src_zero0,
@@ -2130,12 +2131,11 @@ static void hevc_sao_edge_filter_45degree_16multiple_msa(uint8_t *dst, · 74 unmodified lines
2130	2131	src_plus13 = LD_UB(src + 1 + v_cnt + (src_stride << 2));
2131	2132	src_orig += 16;
2132	2133
2133		− SLDI_B2_SB(src10, src11, src_minus11, src_minus12, src_zero0,
2134		− src_zero1, 1);
2135		− SLDI_B2_SB(src12, src13, src_minus13, src_minus14, src_zero2,
2136		− src_zero3, 1);
2137		− SLDI_B2_SB(src11, src12, src_minus12, src_minus13, src_plus10,
2138		− src_plus11, 2);
	2134	+ SLDI_B4_SB(src10, src_minus11, src11, src_minus12,
	2135	+ src12, src_minus13, src13, src_minus14, 1,
	2136	+ src_zero0, src_zero1, src_zero2, src_zero3);
	2137	+ SLDI_B2_SB(src11, src_minus12, src12, src_minus13, 2, src_plus10,
	2138	+ src_plus11);
2139	2139
2140	2140	src_plus12 = __msa_sldi_b((v16i8) src13, (v16i8) src_minus14, 2);
2141	2141
@@ -2228,6 +2228,7 @@ static void hevc_sao_edge_filter_135degree_4width_msa(uint8_t *dst, · 86 unmodified lines
2228	2228	v16u8 cmp_minus10, diff_minus10, cmp_minus11, diff_minus11;
2229	2229	v16u8 src_minus10, src10, src_minus11, src11;
2230	2230	v8i16 offset_mask0, offset_mask1;
	2231	+ v16i8 zeros = { 0 };
2231	2232
2232	2233	sao_offset = __msa_pckev_b(sao_offset, sao_offset);
2233	2234	src_orig = src - 1;
@@ -2239,8 +2240,8 @@ static void hevc_sao_edge_filter_135degree_4width_msa(uint8_t *dst, · 5 unmodified lines
2239	2240	for (height -= 2; height; height -= 2) {
2240	2241	src_orig += (src_stride << 1);
2241	2242
2242		− SLDI_B2_0_SB(src_minus11, src10, src_zero0, src_zero1, 1);
2243		− SLDI_B2_0_UB(src_minus10, src_minus11, src_minus10, src_minus11, 2);
	2243	+ SLDI_B2_SB(zeros, src_minus11, zeros, src10, 1, src_zero0, src_zero1);
	2244	+ SLDI_B2_UB(zeros, src_minus10, zeros, src_minus11, 2, src_minus10, src_minus11);
2244	2245
2245	2246	ILVR_B2_UB(src10, src_minus10, src11, src_minus11, src_minus10,
2246	2247	src_minus11);
@@ -2286,8 +2287,8 @@ static void hevc_sao_edge_filter_135degree_4width_msa(uint8_t *dst, · 39 unmodified lines
2286	2287	dst += dst_stride;
2287	2288	}
2288	2289
2289		− SLDI_B2_0_SB(src_minus11, src10, src_zero0, src_zero1, 1);
2290		− SLDI_B2_0_UB(src_minus10, src_minus11, src_minus10, src_minus11, 2);
	2290	+ SLDI_B2_SB(zeros, src_minus11, zeros, src10, 1, src_zero0, src_zero1);
	2291	+ SLDI_B2_UB(zeros, src_minus10, zeros, src_minus11, 2, src_minus10, src_minus11);
2291	2292
2292	2293	ILVR_B2_UB(src10, src_minus10, src11, src_minus11, src_minus10,
2293	2294	src_minus11);
@@ -2342,6 +2343,7 @@ static void hevc_sao_edge_filter_135degree_8width_msa(uint8_t *dst, · 48 unmodified lines
2342	2343	v16u8 src_minus10, src10, src_minus11, src11;
2343	2344	v16i8 src_zero0, src_zero1, dst0;
2344	2345	v8i16 offset_mask0, offset_mask1;
	2346	+ v16i8 zeros = { 0 };
2345	2347
2346	2348	sao_offset = __msa_pckev_b(sao_offset, sao_offset);
2347	2349	src_orig = src - 1;
@@ -2353,8 +2355,8 @@ static void hevc_sao_edge_filter_135degree_8width_msa(uint8_t *dst, · 5 unmodified lines
2353	2355	for (height -= 2; height; height -= 2) {
2354	2356	src_orig += (src_stride << 1);
2355	2357
2356		− SLDI_B2_0_SB(src_minus11, src10, src_zero0, src_zero1, 1);
2357		− SLDI_B2_0_UB(src_minus10, src_minus11, src_minus10, src_minus11, 2);
	2358	+ SLDI_B2_SB(zeros, src_minus11, zeros, src10, 1, src_zero0, src_zero1);
	2359	+ SLDI_B2_UB(zeros, src_minus10, zeros, src_minus11, 2, src_minus10, src_minus11);
2358	2360	ILVR_B2_UB(src10, src_minus10, src11, src_minus11, src_minus10,
2359	2361	src_minus11);
2360	2362	ILVR_B2_SB(src_zero0, src_zero0, src_zero1, src_zero1, src_zero0,
@@ -2398,8 +2400,8 @@ static void hevc_sao_edge_filter_135degree_8width_msa(uint8_t *dst, · 37 unmodified lines
2398	2400	dst += dst_stride;
2399	2401	}
2400	2402
2401		− SLDI_B2_0_SB(src_minus11, src10, src_zero0, src_zero1, 1);
2402		− SLDI_B2_0_UB(src_minus10, src_minus11, src_minus10, src_minus11, 2);
	2403	+ SLDI_B2_SB(zeros, src_minus11, zeros, src10, 1, src_zero0, src_zero1);
	2404	+ SLDI_B2_UB(zeros, src_minus10, zeros, src_minus11, 2, src_minus10, src_minus11);
2403	2405	ILVR_B2_UB(src10, src_minus10, src11, src_minus11, src_minus10,
2404	2406	src_minus11);
2405	2407	ILVR_B2_SB(src_zero0, src_zero0, src_zero1, src_zero1, src_zero0,

libavcodec/mips/hevcpred_msa.c+16−14

@@ -998,7 +998,8 @@ static void hevc_intra_pred_angular_upper_4width_msa(const uint8_t *src_top, · 997 unmodified lines
998	998	ILVR_D2_SH(fact3, fact1, fact7, fact5, fact1, fact3);
999	999	ILVR_B4_SH(zero, top0, zero, top1, zero, top2, zero, top3,
1000	1000	diff0, diff2, diff4, diff6);
1001		− SLDI_B4_0_SH(diff0, diff2, diff4, diff6, diff1, diff3, diff5, diff7, 2);
	1001	+ SLDI_B4_SH(zero, diff0, zero, diff2, zero, diff4, zero, diff6, 2,
	1002	+ diff1, diff3, diff5, diff7);
1002	1003	ILVR_D2_SH(diff2, diff0, diff6, diff4, diff0, diff2);
1003	1004	ILVR_D2_SH(diff3, diff1, diff7, diff5, diff1, diff3);
1004	1005	MUL2(diff1, fact0, diff3, fact2, diff1, diff3);
@@ -1093,8 +1094,8 @@ static void hevc_intra_pred_angular_upper_8width_msa(const uint8_t *src_top, · 88 unmodified lines
1093	1094	UNPCK_UB_SH(top2, diff4, diff5);
1094	1095	UNPCK_UB_SH(top3, diff6, diff7);
1095	1096
1096		− SLDI_B2_SH(diff1, diff3, diff0, diff2, diff1, diff3, 2);
1097		− SLDI_B2_SH(diff5, diff7, diff4, diff6, diff5, diff7, 2);
	1097	+ SLDI_B4_SH(diff1, diff0, diff3, diff2, diff5, diff4, diff7, diff6, 2,
	1098	+ diff1, diff3, diff5, diff7);
1098	1099	MUL4(diff1, fact0, diff3, fact2, diff5, fact4, diff7, fact6,
1099	1100	diff1, diff3, diff5, diff7);
1100	1101
@@ -1186,8 +1187,8 @@ static void hevc_intra_pred_angular_upper_16width_msa(const uint8_t *src_top, · 85 unmodified lines
1186	1187	fact6 = __msa_fill_h(fact_val3);
1187	1188	fact7 = __msa_fill_h(32 - fact_val3);
1188	1189
1189		− SLDI_B2_UB(top1, top3, top0, top2, top1, top3, 1);
1190		− SLDI_B2_UB(top5, top7, top4, top6, top5, top7, 1);
	1190	+ SLDI_B4_UB(top1, top0, top3, top2, top5, top4, top7, top6, 1,
	1191	+ top1, top3, top5, top7);
1191	1192	UNPCK_UB_SH(top0, diff0, diff1);
1192	1193	UNPCK_UB_SH(top1, diff2, diff3);
1193	1194	UNPCK_UB_SH(top2, diff4, diff5);
@@ -1297,8 +1298,8 @@ static void hevc_intra_pred_angular_upper_32width_msa(const uint8_t *src_top, · 103 unmodified lines
1297	1298	top2 = top1;
1298	1299	top6 = top5;
1299	1300
1300		− SLDI_B2_UB(top1, top3, top0, top2, top1, top3, 1);
1301		− SLDI_B2_UB(top5, top7, top4, top6, top5, top7, 1);
	1301	+ SLDI_B4_UB(top1, top0, top3, top2, top5, top4, top7, top6, 1,
	1302	+ top1, top3, top5, top7);
1302	1303	UNPCK_UB_SH(top0, diff0, diff1);
1303	1304	UNPCK_UB_SH(top1, diff2, diff3);
1304	1305	UNPCK_UB_SH(top2, diff4, diff5);
@@ -1407,7 +1408,8 @@ static void hevc_intra_pred_angular_lower_4width_msa(const uint8_t *src_top, · 102 unmodified lines
1407	1408	ILVR_D2_SH(fact3, fact1, fact7, fact5, fact1, fact3);
1408	1409	ILVR_B4_SH(zero, top0, zero, top1, zero, top2, zero, top3,
1409	1410	diff0, diff2, diff4, diff6);
1410		− SLDI_B4_0_SH(diff0, diff2, diff4, diff6, diff1, diff3, diff5, diff7, 2);
	1411	+ SLDI_B4_SH(zero, diff0, zero, diff2, zero, diff4, zero, diff6, 2,
	1412	+ diff1, diff3, diff5, diff7);
1411	1413	ILVR_D2_SH(diff2, diff0, diff6, diff4, diff0, diff2);
1412	1414	ILVR_D2_SH(diff3, diff1, diff7, diff5, diff1, diff3);
1413	1415	MUL2(diff1, fact0, diff3, fact2, diff1, diff3);
@@ -1511,8 +1513,8 @@ static void hevc_intra_pred_angular_lower_8width_msa(const uint8_t *src_top, · 97 unmodified lines
1511	1513	UNPCK_UB_SH(top1, diff2, diff3);
1512	1514	UNPCK_UB_SH(top2, diff4, diff5);
1513	1515	UNPCK_UB_SH(top3, diff6, diff7);
1514		− SLDI_B2_SH(diff1, diff3, diff0, diff2, diff1, diff3, 2);
1515		− SLDI_B2_SH(diff5, diff7, diff4, diff6, diff5, diff7, 2);
	1516	+ SLDI_B4_SH(diff1, diff0, diff3, diff2, diff5, diff4, diff7, diff6, 2,
	1517	+ diff1, diff3, diff5, diff7);
1516	1518	MUL4(diff1, fact0, diff3, fact2, diff5, fact4, diff7, fact6,
1517	1519	diff1, diff3, diff5, diff7);
1518	1520
@@ -1606,8 +1608,8 @@ static void hevc_intra_pred_angular_lower_16width_msa(const uint8_t *src_top, · 87 unmodified lines
1606	1608	fact6 = __msa_fill_h(fact_val3);
1607	1609	fact7 = __msa_fill_h(32 - fact_val3);
1608	1610
1609		− SLDI_B2_SB(top1, top3, top0, top2, top1, top3, 1);
1610		− SLDI_B2_SB(top5, top7, top4, top6, top5, top7, 1);
	1611	+ SLDI_B4_SB(top1, top0, top3, top2, top5, top4, top7, top6, 1,
	1612	+ top1, top3, top5, top7);
1611	1613
1612	1614	UNPCK_UB_SH(top0, diff0, diff1);
1613	1615	UNPCK_UB_SH(top1, diff2, diff3);
@@ -1713,8 +1715,8 @@ static void hevc_intra_pred_angular_lower_32width_msa(const uint8_t *src_top, · 99 unmodified lines
1713	1715	top2 = top1;
1714	1716	top6 = top5;
1715	1717
1716		− SLDI_B2_SB(top1, top3, top0, top2, top1, top3, 1);
1717		− SLDI_B2_SB(top5, top7, top4, top6, top5, top7, 1);
	1718	+ SLDI_B4_SB(top1, top0, top3, top2, top5, top4, top7, top6, 1,
	1719	+ top1, top3, top5, top7);
1718	1720
1719	1721	UNPCK_UB_SH(top0, diff0, diff1);
1720	1722	UNPCK_UB_SH(top1, diff2, diff3);

libavcodec/mips/hpeldsp_msa.c+42−24

@@ -59,12 +59,13 @@ static void common_hz_bil_4w_msa(const uint8_t *src, int32_t src_stride, · 58 unmodified lines
59	59	uint8_t loop_cnt;
60	60	uint32_t out0, out1;
61	61	v16u8 src0, src1, src0_sld1, src1_sld1, res0, res1;
	62	+ v16i8 zeros = { 0 };
62	63
63	64	for (loop_cnt = (height >> 1); loop_cnt--;) {
64	65	LD_UB2(src, src_stride, src0, src1);
65	66	src += (2 * src_stride);
66	67
67		− SLDI_B2_0_UB(src0, src1, src0_sld1, src1_sld1, 1);
	68	+ SLDI_B2_UB(zeros, src0, zeros, src1, 1, src0_sld1, src1_sld1);
68	69	AVER_UB2_UB(src0_sld1, src0, src1_sld1, src1, res0, res1);
69	70
70	71	out0 = __msa_copy_u_w((v4i32) res0, 0);
@@ -82,13 +83,14 @@ static void common_hz_bil_8w_msa(const uint8_t *src, int32_t src_stride, · 11 unmodified lines
82	83	{
83	84	uint8_t loop_cnt;
84	85	v16i8 src0, src1, src2, src3, src0_sld1, src1_sld1, src2_sld1, src3_sld1;
	86	+ v16i8 zeros = { 0 };
85	87
86	88	for (loop_cnt = (height >> 2); loop_cnt--;) {
87	89	LD_SB4(src, src_stride, src0, src1, src2, src3);
88	90	src += (4 * src_stride);
89	91
90		− SLDI_B4_0_SB(src0, src1, src2, src3,
91		− src0_sld1, src1_sld1, src2_sld1, src3_sld1, 1);
	92	+ SLDI_B4_SB(zeros, src0, zeros, src1, zeros, src2, zeros, src3, 1,
	93	+ src0_sld1, src1_sld1, src2_sld1, src3_sld1);
92	94	AVER_ST8x4_UB(src0, src0_sld1, src1, src1_sld1,
93	95	src2, src2_sld1, src3, src3_sld1, dst, dst_stride);
94	96	dst += (4 * dst_stride);
@@ -125,14 +127,15 @@ static void common_hz_bil_no_rnd_8x8_msa(const uint8_t *src, int32_t src_stride, · 30 unmodified lines
125	127	v16i8 src0, src1, src2, src3, src4, src5, src6, src7;
126	128	v16i8 src0_sld1, src1_sld1, src2_sld1, src3_sld1;
127	129	v16i8 src4_sld1, src5_sld1, src6_sld1, src7_sld1;
	130	+ v16i8 zeros = { 0 };
128	131
129	132	LD_SB8(src, src_stride, src0, src1, src2, src3, src4, src5, src6, src7);
130	133	src += (8 * src_stride);
131	134
132		− SLDI_B4_0_SB(src0, src1, src2, src3,
133		− src0_sld1, src1_sld1, src2_sld1, src3_sld1, 1);
134		− SLDI_B4_0_SB(src4, src5, src6, src7,
135		− src4_sld1, src5_sld1, src6_sld1, src7_sld1, 1);
	135	+ SLDI_B4_SB(zeros, src0, zeros, src1, zeros, src2, zeros, src3, 1,
	136	+ src0_sld1, src1_sld1, src2_sld1, src3_sld1);
	137	+ SLDI_B4_SB(zeros, src4, zeros, src5, zeros, src6, zeros, src7, 1,
	138	+ src4_sld1, src5_sld1, src6_sld1, src7_sld1);
136	139
137	140	AVE_ST8x4_UB(src0, src0_sld1, src1, src1_sld1,
138	141	src2, src2_sld1, src3, src3_sld1, dst, dst_stride);
@@ -145,10 +148,11 @@ static void common_hz_bil_no_rnd_4x8_msa(const uint8_t *src, int32_t src_stride, · 6 unmodified lines
145	148	uint8_t *dst, int32_t dst_stride)
146	149	{
147	150	v16i8 src0, src1, src2, src3, src0_sld1, src1_sld1, src2_sld1, src3_sld1;
	151	+ v16i8 zeros = { 0 };
148	152
149	153	LD_SB4(src, src_stride, src0, src1, src2, src3);
150		− SLDI_B4_0_SB(src0, src1, src2, src3,
151		− src0_sld1, src1_sld1, src2_sld1, src3_sld1, 1);
	154	+ SLDI_B4_SB(zeros, src0, zeros, src1, zeros, src2, zeros, src3, 1,
	155	+ src0_sld1, src1_sld1, src2_sld1, src3_sld1);
152	156	AVE_ST8x4_UB(src0, src0_sld1, src1, src1_sld1,
153	157	src2, src2_sld1, src3, src3_sld1, dst, dst_stride);
154	158	}
@@ -216,12 +220,13 @@ static void common_hz_bil_and_aver_dst_4w_msa(const uint8_t *src, · 61 unmodified lines
216	220	v16u8 src0, src1, src0_sld1, src1_sld1, res0, res1;
217	221	v16u8 tmp0 = { 0 };
218	222	v16u8 tmp1 = { 0 };
	223	+ v16i8 zeros = { 0 };
219	224
220	225	for (loop_cnt = (height >> 1); loop_cnt--;) {
221	226	LD_UB2(src, src_stride, src0, src1);
222	227	src += (2 * src_stride);
223	228
224		− SLDI_B2_0_UB(src0, src1, src0_sld1, src1_sld1, 1);
	229	+ SLDI_B2_UB(zeros, src0, zeros, src1, 1, src0_sld1, src1_sld1);
225	230
226	231	dst0 = LW(dst);
227	232	dst1 = LW(dst + dst_stride);
@@ -247,13 +252,14 @@ static void common_hz_bil_and_aver_dst_8w_msa(const uint8_t *src, · 19 unmodified lines
247	252	{
248	253	uint8_t loop_cnt;
249	254	v16i8 src0, src1, src2, src3, src0_sld1, src1_sld1, src2_sld1, src3_sld1;
	255	+ v16i8 zeros = { 0 };
250	256
251	257	for (loop_cnt = (height >> 2); loop_cnt--;) {
252	258	LD_SB4(src, src_stride, src0, src1, src2, src3);
253	259	src += (4 * src_stride);
254	260
255		− SLDI_B4_0_SB(src0, src1, src2, src3,
256		− src0_sld1, src1_sld1, src2_sld1, src3_sld1, 1);
	261	+ SLDI_B4_SB(zeros, src0, zeros, src1, zeros, src2, zeros, src3, 1,
	262	+ src0_sld1, src1_sld1, src2_sld1, src3_sld1);
257	263
258	264	AVER_DST_ST8x4_UB(src0, src0_sld1, src1, src1_sld1, src2, src2_sld1,
259	265	src3, src3_sld1, dst, dst_stride);
@@ -529,6 +535,7 @@ static void common_hv_bil_4w_msa(const uint8_t *src, int32_t src_stride, · 269 unmodified lines
529	535	v16i8 src0, src1, src2, src0_sld1, src1_sld1, src2_sld1;
530	536	v16u8 src0_r, src1_r, src2_r, res;
531	537	v8u16 add0, add1, add2, sum0, sum1;
	538	+ v16i8 zeros = { 0 };
532	539
533	540	src0 = LD_SB(src);
534	541	src += src_stride;
@@ -537,7 +544,8 @@ static void common_hv_bil_4w_msa(const uint8_t *src, int32_t src_stride, · 2 unmodified lines
537	544	LD_SB2(src, src_stride, src1, src2);
538	545	src += (2 * src_stride);
539	546
540		− SLDI_B3_0_SB(src0, src1, src2, src0_sld1, src1_sld1, src2_sld1, 1);
	547	+ SLDI_B3_SB(zeros, src0, zeros, src1, zeros, src2, 1, src0_sld1,
	548	+ src1_sld1, src2_sld1);
541	549	ILVR_B3_UB(src0_sld1, src0, src1_sld1, src1, src2_sld1, src2,
542	550	src0_r, src1_r, src2_r);
543	551	HADD_UB3_UH(src0_r, src1_r, src2_r, add0, add1, add2);
@@ -565,6 +573,7 @@ static void common_hv_bil_8w_msa(const uint8_t *src, int32_t src_stride, · 21 unmodified lines
565	573	v16u8 src0_r, src1_r, src2_r, src3_r, src4_r;
566	574	v8u16 add0, add1, add2, add3, add4;
567	575	v8u16 sum0, sum1, sum2, sum3;
	576	+ v16i8 zeros = { 0 };
568	577
569	578	src0 = LD_SB(src);
570	579	src += src_stride;
@@ -573,8 +582,9 @@ static void common_hv_bil_8w_msa(const uint8_t *src, int32_t src_stride, · 2 unmodified lines
573	582	LD_SB4(src, src_stride, src1, src2, src3, src4);
574	583	src += (4 * src_stride);
575	584
576		− SLDI_B3_0_SB(src0, src1, src2, src0_sld1, src1_sld1, src2_sld1, 1);
577		− SLDI_B2_0_SB(src3, src4, src3_sld1, src4_sld1, 1);
	585	+ SLDI_B3_SB(zeros, src0, zeros, src1, zeros, src2, 1, src0_sld1,
	586	+ src1_sld1, src2_sld1);
	587	+ SLDI_B2_SB(zeros, src3, zeros, src4, 1, src3_sld1, src4_sld1);
578	588	ILVR_B3_UB(src0_sld1, src0, src1_sld1, src1, src2_sld1, src2, src0_r,
579	589	src1_r, src2_r);
580	590	ILVR_B2_UB(src3_sld1, src3, src4_sld1, src4, src3_r, src4_r);
@@ -659,15 +669,17 @@ static void common_hv_bil_no_rnd_8x8_msa(const uint8_t *src, int32_t src_stride, · 78 unmodified lines
659	669	v8u16 add0, add1, add2, add3, add4, add5, add6, add7, add8;
660	670	v8u16 sum0, sum1, sum2, sum3, sum4, sum5, sum6, sum7;
661	671	v16i8 out0, out1;
	672	+ v16i8 zeros = { 0 };
662	673
663	674	LD_UB8(src, src_stride, src0, src1, src2, src3, src4, src5, src6, src7);
664	675	src += (8 * src_stride);
665	676	src8 = LD_UB(src);
666	677
667		− SLDI_B4_0_UB(src0, src1, src2, src3, src0_sld1, src1_sld1, src2_sld1,
668		− src3_sld1, 1);
669		− SLDI_B3_0_UB(src4, src5, src6, src4_sld1, src5_sld1, src6_sld1, 1);
670		− SLDI_B2_0_UB(src7, src8, src7_sld1, src8_sld1, 1);
	678	+ SLDI_B4_UB(zeros, src0, zeros, src1, zeros, src2, zeros, src3, 1,
	679	+ src0_sld1, src1_sld1, src2_sld1, src3_sld1);
	680	+ SLDI_B3_UB(zeros, src4, zeros, src5, zeros, src6, 1, src4_sld1,
	681	+ src5_sld1, src6_sld1);
	682	+ SLDI_B2_UB(zeros, src7, zeros, src8, 1, src7_sld1, src8_sld1);
671	683	ILVR_B4_UH(src0_sld1, src0, src1_sld1, src1, src2_sld1, src2, src3_sld1,
672	684	src3, src0_r, src1_r, src2_r, src3_r);
673	685	ILVR_B3_UH(src4_sld1, src4, src5_sld1, src5, src6_sld1, src6, src4_r,
@@ -703,13 +715,15 @@ static void common_hv_bil_no_rnd_4x8_msa(const uint8_t *src, int32_t src_stride, · 29 unmodified lines
703	715	v8u16 add0, add1, add2, add3, add4;
704	716	v8u16 sum0, sum1, sum2, sum3;
705	717	v16i8 out0, out1;
	718	+ v16i8 zeros = { 0 };
706	719
707	720	LD_SB4(src, src_stride, src0, src1, src2, src3);
708	721	src += (4 * src_stride);
709	722	src4 = LD_SB(src);
710	723
711		− SLDI_B3_0_SB(src0, src1, src2, src0_sld1, src1_sld1, src2_sld1, 1);
712		− SLDI_B2_0_SB(src3, src4, src3_sld1, src4_sld1, 1);
	724	+ SLDI_B3_SB(zeros, src0, zeros, src1, zeros, src2, 1, src0_sld1,
	725	+ src1_sld1, src2_sld1);
	726	+ SLDI_B2_SB(zeros, src3, zeros, src4, 1, src3_sld1, src4_sld1);
713	727	ILVR_B3_UH(src0_sld1, src0, src1_sld1, src1, src2_sld1, src2, src0_r,
714	728	src1_r, src2_r);
715	729	ILVR_B2_UH(src3_sld1, src3, src4_sld1, src4, src3_r, src4_r);
@@ -918,6 +932,7 @@ static void common_hv_bil_and_aver_dst_4w_msa(const uint8_t *src, · 202 unmodified lines
918	932	v16u8 src0_r, src1_r, src2_r;
919	933	v8u16 add0, add1, add2, sum0, sum1;
920	934	v16u8 dst0, dst1, res0, res1;
	935	+ v16i8 zeros = { 0 };
921	936
922	937	src0 = LD_SB(src);
923	938	src += src_stride;
@@ -927,7 +942,8 @@ static void common_hv_bil_and_aver_dst_4w_msa(const uint8_t *src, · 3 unmodified lines
927	942	src += (2 * src_stride);
928	943
929	944	LD_UB2(dst, dst_stride, dst0, dst1);
930		− SLDI_B3_0_SB(src0, src1, src2, src0_sld1, src1_sld1, src2_sld1, 1);
	945	+ SLDI_B3_SB(zeros, src0, zeros, src1, zeros, src2, 1, src0_sld1,
	946	+ src1_sld1, src2_sld1);
931	947	ILVR_B3_UB(src0_sld1, src0, src1_sld1, src1, src2_sld1, src2, src0_r,
932	948	src1_r, src2_r);
933	949	HADD_UB3_UH(src0_r, src1_r, src2_r, add0, add1, add2);
@@ -959,6 +975,7 @@ static void common_hv_bil_and_aver_dst_8w_msa(const uint8_t *src, · 25 unmodified lines
959	975	v16u8 src0_r, src1_r, src2_r, src3_r, src4_r;
960	976	v8u16 add0, add1, add2, add3, add4;
961	977	v8u16 sum0, sum1, sum2, sum3;
	978	+ v16i8 zeros = { 0 };
962	979
963	980	src0 = LD_SB(src);
964	981	src += src_stride;
@@ -968,8 +985,9 @@ static void common_hv_bil_and_aver_dst_8w_msa(const uint8_t *src, · 3 unmodified lines
968	985	src += (4 * src_stride);
969	986
970	987	LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
971		− SLDI_B3_0_SB(src0, src1, src2, src0_sld1, src1_sld1, src2_sld1, 1);
972		− SLDI_B2_0_SB(src3, src4, src3_sld1, src4_sld1, 1);
	988	+ SLDI_B3_SB(zeros, src0, zeros, src1, zeros, src2, 1, src0_sld1,
	989	+ src1_sld1, src2_sld1);
	990	+ SLDI_B2_SB(zeros, src3, zeros, src4, 1, src3_sld1, src4_sld1);
973	991	ILVR_B3_UB(src0_sld1, src0, src1_sld1, src1, src2_sld1, src2, src0_r,
974	992	src1_r, src2_r);
975	993	ILVR_B2_UB(src3_sld1, src3, src4_sld1, src4, src3_r, src4_r);

libavcodec/mips/me_cmp_msa.c+4−4

@@ -87,8 +87,8 @@ static uint32_t sad_horiz_bilinear_filter_8width_msa(uint8_t *src, · 86 unmodified lines
87	87
88	88	PCKEV_D2_UB(src1, src0, src3, src2, src0, src1);
89	89	PCKEV_D2_UB(ref1, ref0, ref3, ref2, ref4, ref5);
90		− SLDI_B2_UB(ref0, ref1, ref0, ref1, ref0, ref1, 1);
91		− SLDI_B2_UB(ref2, ref3, ref2, ref3, ref2, ref3, 1);
	90	+ SLDI_B4_UB(ref0, ref0, ref1, ref1, ref2, ref2, ref3, ref3, 1,
	91	+ ref0, ref1, ref2, ref3);
92	92	PCKEV_D2_UB(ref1, ref0, ref3, ref2, ref0, ref1);
93	93	AVER_UB2_UB(ref4, ref0, ref5, ref1, comp0, comp1);
94	94	sad += SAD_UB2_UH(src0, src1, comp0, comp1);
@@ -100,8 +100,8 @@ static uint32_t sad_horiz_bilinear_filter_8width_msa(uint8_t *src, · 5 unmodified lines
100	100
101	101	PCKEV_D2_UB(src1, src0, src3, src2, src0, src1);
102	102	PCKEV_D2_UB(ref1, ref0, ref3, ref2, ref4, ref5);
103		− SLDI_B2_UB(ref0, ref1, ref0, ref1, ref0, ref1, 1);
104		− SLDI_B2_UB(ref2, ref3, ref2, ref3, ref2, ref3, 1);
	103	+ SLDI_B4_UB(ref0, ref0, ref1, ref1, ref2, ref2, ref3, ref3, 1,
	104	+ ref0, ref1, ref2, ref3);
105	105	PCKEV_D2_UB(ref1, ref0, ref3, ref2, ref0, ref1);
106	106	AVER_UB2_UB(ref4, ref0, ref5, ref1, comp0, comp1);
107	107	sad += SAD_UB2_UH(src0, src1, comp0, comp1);

libavcodec/mips/qpeldsp_msa.c+42−42

@@ -480,8 +480,8 @@ static void horiz_mc_qpel_aver_src1_8width_msa(const uint8_t *src, · 479 unmodified lines
480	480	res1 = APPLY_HORIZ_QPEL_FILTER_8BYTE(inp2, inp3,
481	481	mask0, mask1, mask2, mask3,
482	482	const20, const6, const3);
483		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
484		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	483	+ SLDI_B4_UB(inp0, inp0, inp1, inp1, inp2, inp2, inp3, inp3, 1,
	484	+ inp0, inp1, inp2, inp3);
485	485	inp0 = (v16u8) __msa_insve_d((v2i64) inp0, 1, (v2i64) inp1);
486	486	inp2 = (v16u8) __msa_insve_d((v2i64) inp2, 1, (v2i64) inp3);
487	487	AVER_UB2_UB(inp0, res0, inp2, res1, res0, res1);
@@ -710,8 +710,8 @@ static void horiz_mc_qpel_no_rnd_aver_src1_8width_msa(const uint8_t *src, · 222 unmodified lines
710	710	res1 = APPLY_HORIZ_QPEL_NO_ROUND_FILTER_8BYTE(inp2, inp3, mask0, mask1,
711	711	mask2, mask3, const20,
712	712	const6, const3);
713		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
714		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	713	+ SLDI_B4_UB(inp0, inp0, inp1, inp1, inp2, inp2, inp3, inp3, 1,
	714	+ inp0, inp1, inp2, inp3);
715	715	inp0 = (v16u8) __msa_insve_d((v2i64) inp0, 1, (v2i64) inp1);
716	716	inp2 = (v16u8) __msa_insve_d((v2i64) inp2, 1, (v2i64) inp3);
717	717	res0 = __msa_ave_u_b(inp0, res0);
@@ -948,8 +948,8 @@ static void horiz_mc_qpel_avg_dst_aver_src1_8width_msa(const uint8_t *src, · 230 unmodified lines
948	948	mask0, mask1, mask2, mask3,
949	949	const20, const6, const3);
950	950	LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
951		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
952		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	951	+ SLDI_B4_UB(inp0, inp0, inp1, inp1, inp2, inp2, inp3, inp3, 1,
	952	+ inp0, inp1, inp2, inp3);
953	953	inp0 = (v16u8) __msa_insve_d((v2i64) inp0, 1, (v2i64) inp1);
954	954	inp2 = (v16u8) __msa_insve_d((v2i64) inp2, 1, (v2i64) inp3);
955	955	dst0 = (v16u8) __msa_insve_d((v2i64) dst0, 1, (v2i64) dst1);
@@ -3094,7 +3094,7 @@ static void hv_mc_qpel_no_rnd_aver_hv_src10_8x8_msa(const uint8_t *src, · 2138 unmodified lines
3094	3094	res0 = APPLY_HORIZ_QPEL_NO_ROUND_FILTER_8BYTE(inp0, inp1, mask0, mask1,
3095	3095	mask2, mask3, const20,
3096	3096	const6, const3);
3097		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
	3097	+ SLDI_B2_UB(inp0, inp0, inp1, inp1, 1, inp0, inp1);
3098	3098
3099	3099	inp0 = (v16u8) __msa_insve_d((v2i64) inp0, 1, (v2i64) inp1);
3100	3100	horiz0 = __msa_ave_u_b(inp0, res0);
@@ -3104,7 +3104,7 @@ static void hv_mc_qpel_no_rnd_aver_hv_src10_8x8_msa(const uint8_t *src, · 3 unmodified lines
3104	3104	res1 = APPLY_HORIZ_QPEL_NO_ROUND_FILTER_8BYTE(inp2, inp3, mask0, mask1,
3105	3105	mask2, mask3, const20,
3106	3106	const6, const3);
3107		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	3107	+ SLDI_B2_UB(inp2, inp2, inp3, inp3, 1, inp2, inp3);
3108	3108
3109	3109	inp2 = (v16u8) __msa_insve_d((v2i64) inp2, 1, (v2i64) inp3);
3110	3110	horiz2 = __msa_ave_u_b(inp2, res1);
@@ -3114,7 +3114,7 @@ static void hv_mc_qpel_no_rnd_aver_hv_src10_8x8_msa(const uint8_t *src, · 3 unmodified lines
3114	3114	res0 = APPLY_HORIZ_QPEL_NO_ROUND_FILTER_8BYTE(inp0, inp1, mask0, mask1,
3115	3115	mask2, mask3, const20,
3116	3116	const6, const3);
3117		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
	3117	+ SLDI_B2_UB(inp0, inp0, inp1, inp1, 1, inp0, inp1);
3118	3118
3119	3119	inp0 = (v16u8) __msa_insve_d((v2i64) inp0, 1, (v2i64) inp1);
3120	3120	horiz4 = __msa_ave_u_b(inp0, res0);
@@ -3134,7 +3134,7 @@ static void hv_mc_qpel_no_rnd_aver_hv_src10_8x8_msa(const uint8_t *src, · 13 unmodified lines
3134	3134	res1 = APPLY_HORIZ_QPEL_NO_ROUND_FILTER_8BYTE(inp2, inp3, mask0, mask1,
3135	3135	mask2, mask3, const20,
3136	3136	const6, const3);
3137		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	3137	+ SLDI_B2_UB(inp2, inp2, inp3, inp3, 1, inp2, inp3);
3138	3138
3139	3139	inp2 = (v16u8) __msa_insve_d((v2i64) inp2, 1, (v2i64) inp3);
3140	3140	horiz6 = __msa_ave_u_b(inp2, res1);
@@ -3389,7 +3389,7 @@ static void hv_mc_qpel_no_rnd_aver_h_src1_8x8_msa(const uint8_t *src, · 248 unmodified lines
3389	3389	res0 = APPLY_HORIZ_QPEL_NO_ROUND_FILTER_8BYTE(inp0, inp1, mask0, mask1,
3390	3390	mask2, mask3, const20,
3391	3391	const6, const3);
3392		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
	3392	+ SLDI_B2_UB(inp0, inp0, inp1, inp1, 1, inp0, inp1);
3393	3393
3394	3394	inp0 = (v16u8) __msa_insve_d((v2i64) inp0, 1, (v2i64) inp1);
3395	3395	horiz0 = __msa_ave_u_b(inp0, res0);
@@ -3399,7 +3399,7 @@ static void hv_mc_qpel_no_rnd_aver_h_src1_8x8_msa(const uint8_t *src, · 3 unmodified lines
3399	3399	res1 = APPLY_HORIZ_QPEL_NO_ROUND_FILTER_8BYTE(inp2, inp3, mask0, mask1,
3400	3400	mask2, mask3, const20,
3401	3401	const6, const3);
3402		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	3402	+ SLDI_B2_UB(inp2, inp2, inp3, inp3, 1, inp2, inp3);
3403	3403
3404	3404	inp2 = (v16u8) __msa_insve_d((v2i64) inp2, 1, (v2i64) inp3);
3405	3405	horiz2 = __msa_ave_u_b(inp2, res1);
@@ -3409,7 +3409,7 @@ static void hv_mc_qpel_no_rnd_aver_h_src1_8x8_msa(const uint8_t *src, · 3 unmodified lines
3409	3409	res0 = APPLY_HORIZ_QPEL_NO_ROUND_FILTER_8BYTE(inp0, inp1, mask0, mask1,
3410	3410	mask2, mask3, const20,
3411	3411	const6, const3);
3412		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
	3412	+ SLDI_B2_UB(inp0, inp0, inp1, inp1, 1, inp0, inp1);
3413	3413
3414	3414	inp0 = (v16u8) __msa_insve_d((v2i64) inp0, 1, (v2i64) inp1);
3415	3415	horiz4 = __msa_ave_u_b(inp0, res0);
@@ -3427,7 +3427,7 @@ static void hv_mc_qpel_no_rnd_aver_h_src1_8x8_msa(const uint8_t *src, · 11 unmodified lines
3427	3427	res1 = APPLY_HORIZ_QPEL_NO_ROUND_FILTER_8BYTE(inp2, inp3, mask0, mask1,
3428	3428	mask2, mask3, const20,
3429	3429	const6, const3);
3430		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	3430	+ SLDI_B2_UB(inp2, inp2, inp3, inp3, 1, inp2, inp3);
3431	3431
3432	3432	inp2 = (v16u8) __msa_insve_d((v2i64) inp2, 1, (v2i64) inp3);
3433	3433	horiz6 = __msa_ave_u_b(inp2, res1);
@@ -3691,7 +3691,7 @@ static void hv_mc_qpel_no_rnd_aver_hv_src11_8x8_msa(const uint8_t *src, · 257 unmodified lines
3691	3691	res0 = APPLY_HORIZ_QPEL_NO_ROUND_FILTER_8BYTE(inp0, inp1, mask0, mask1,
3692	3692	mask2, mask3, const20,
3693	3693	const6, const3);
3694		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
	3694	+ SLDI_B2_UB(inp0, inp0, inp1, inp1, 1, inp0, inp1);
3695	3695
3696	3696	inp0 = (v16u8) __msa_insve_d((v2i64) inp0, 1, (v2i64) inp1);
3697	3697	horiz0 = __msa_ave_u_b(inp0, res0);
@@ -3701,7 +3701,7 @@ static void hv_mc_qpel_no_rnd_aver_hv_src11_8x8_msa(const uint8_t *src, · 3 unmodified lines
3701	3701	res1 = APPLY_HORIZ_QPEL_NO_ROUND_FILTER_8BYTE(inp2, inp3, mask0, mask1,
3702	3702	mask2, mask3, const20,
3703	3703	const6, const3);
3704		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	3704	+ SLDI_B2_UB(inp2, inp2, inp3, inp3, 1, inp2, inp3);
3705	3705
3706	3706	inp2 = (v16u8) __msa_insve_d((v2i64) inp2, 1, (v2i64) inp3);
3707	3707	horiz2 = __msa_ave_u_b(inp2, res1);
@@ -3712,7 +3712,7 @@ static void hv_mc_qpel_no_rnd_aver_hv_src11_8x8_msa(const uint8_t *src, · 4 unmodified lines
3712	3712	mask2, mask3, const20,
3713	3713	const6, const3);
3714	3714
3715		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
	3715	+ SLDI_B2_UB(inp0, inp0, inp1, inp1, 1, inp0, inp1);
3716	3716	inp0 = (v16u8) __msa_insve_d((v2i64) inp0, 1, (v2i64) inp1);
3717	3717	horiz4 = __msa_ave_u_b(inp0, res0);
3718	3718	horiz5 = (v16u8) __msa_splati_d((v2i64) horiz4, 1);
@@ -3731,7 +3731,7 @@ static void hv_mc_qpel_no_rnd_aver_hv_src11_8x8_msa(const uint8_t *src, · 12 unmodified lines
3731	3731	res1 = APPLY_HORIZ_QPEL_NO_ROUND_FILTER_8BYTE(inp2, inp3, mask0, mask1,
3732	3732	mask2, mask3, const20,
3733	3733	const6, const3);
3734		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	3734	+ SLDI_B2_UB(inp2, inp2, inp3, inp3, 1, inp2, inp3);
3735	3735
3736	3736	inp2 = (v16u8) __msa_insve_d((v2i64) inp2, 1, (v2i64) inp3);
3737	3737	horiz6 = __msa_ave_u_b(inp2, res1);
@@ -4134,12 +4134,12 @@ static void hv_mc_qpel_aver_hv_src10_8x8_msa(const uint8_t *src, · 396 unmodified lines
4134	4134	const20, const6, const3);
4135	4135	res1 = APPLY_HORIZ_QPEL_FILTER_8BYTE(inp2, inp3, mask0, mask1, mask2, mask3,
4136	4136	const20, const6, const3);
4137		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
	4137	+ SLDI_B2_UB(inp0, inp0, inp1, inp1, 1, inp0, inp1);
4138	4138
4139	4139	inp0 = (v16u8) __msa_insve_d((v2i64) inp0, 1, (v2i64) inp1);
4140	4140	horiz0 = __msa_aver_u_b(inp0, res0);
4141	4141	horiz1 = (v16u8) __msa_splati_d((v2i64) horiz0, 1);
4142		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	4142	+ SLDI_B2_UB(inp2, inp2, inp3, inp3, 1, inp2, inp3);
4143	4143
4144	4144	inp2 = (v16u8) __msa_insve_d((v2i64) inp2, 1, (v2i64) inp3);
4145	4145	horiz2 = __msa_aver_u_b(inp2, res1);
@@ -4150,12 +4150,12 @@ static void hv_mc_qpel_aver_hv_src10_8x8_msa(const uint8_t *src, · 4 unmodified lines
4150	4150	const20, const6, const3);
4151	4151	res1 = APPLY_HORIZ_QPEL_FILTER_8BYTE(inp2, inp3, mask0, mask1, mask2, mask3,
4152	4152	const20, const6, const3);
4153		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
	4153	+ SLDI_B2_UB(inp0, inp0, inp1, inp1, 1, inp0, inp1);
4154	4154
4155	4155	inp0 = (v16u8) __msa_insve_d((v2i64) inp0, 1, (v2i64) inp1);
4156	4156	horiz4 = __msa_aver_u_b(inp0, res0);
4157	4157	horiz5 = (v16u8) __msa_splati_d((v2i64) horiz4, 1);
4158		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	4158	+ SLDI_B2_UB(inp2, inp2, inp3, inp3, 1, inp2, inp3);
4159	4159
4160	4160	inp2 = (v16u8) __msa_insve_d((v2i64) inp2, 1, (v2i64) inp3);
4161	4161	horiz6 = __msa_aver_u_b(inp2, res1);
@@ -4410,12 +4410,12 @@ static void hv_mc_qpel_aver_h_src1_8x8_msa(const uint8_t *src, · 248 unmodified lines
4410	4410	const20, const6, const3);
4411	4411	res1 = APPLY_HORIZ_QPEL_FILTER_8BYTE(inp2, inp3, mask0, mask1, mask2, mask3,
4412	4412	const20, const6, const3);
4413		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
	4413	+ SLDI_B2_UB(inp0, inp0, inp1, inp1, 1, inp0, inp1);
4414	4414
4415	4415	inp0 = (v16u8) __msa_insve_d((v2i64) inp0, 1, (v2i64) inp1);
4416	4416	horiz0 = __msa_aver_u_b(inp0, res0);
4417	4417	horiz1 = (v16u8) __msa_splati_d((v2i64) horiz0, 1);
4418		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	4418	+ SLDI_B2_UB(inp2, inp2, inp3, inp3, 1, inp2, inp3);
4419	4419
4420	4420	inp2 = (v16u8) __msa_insve_d((v2i64) inp2, 1, (v2i64) inp3);
4421	4421	horiz2 = __msa_aver_u_b(inp2, res1);
@@ -4426,12 +4426,12 @@ static void hv_mc_qpel_aver_h_src1_8x8_msa(const uint8_t *src, · 4 unmodified lines
4426	4426	const20, const6, const3);
4427	4427	res1 = APPLY_HORIZ_QPEL_FILTER_8BYTE(inp2, inp3, mask0, mask1, mask2, mask3,
4428	4428	const20, const6, const3);
4429		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
	4429	+ SLDI_B2_UB(inp0, inp0, inp1, inp1, 1, inp0, inp1);
4430	4430
4431	4431	inp0 = (v16u8) __msa_insve_d((v2i64) inp0, 1, (v2i64) inp1);
4432	4432	horiz4 = __msa_aver_u_b(inp0, res0);
4433	4433	horiz5 = (v16u8) __msa_splati_d((v2i64) horiz4, 1);
4434		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	4434	+ SLDI_B2_UB(inp2, inp2, inp3, inp3, 1, inp2, inp3);
4435	4435
4436	4436	inp2 = (v16u8) __msa_insve_d((v2i64) inp2, 1, (v2i64) inp3);
4437	4437	horiz6 = __msa_aver_u_b(inp2, res1);
@@ -4690,14 +4690,14 @@ static void hv_mc_qpel_aver_hv_src11_8x8_msa(const uint8_t *src, · 252 unmodified lines
4690	4690	res0 = APPLY_HORIZ_QPEL_FILTER_8BYTE(inp0, inp1,
4691	4691	mask0, mask1, mask2, mask3,
4692	4692	const20, const6, const3);
4693		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
	4693	+ SLDI_B2_UB(inp0, inp0, inp1, inp1, 1, inp0, inp1);
4694	4694
4695	4695	inp0 = (v16u8) __msa_insve_d((v2i64) inp0, 1, (v2i64) inp1);
4696	4696	horiz0 = __msa_aver_u_b(inp0, res0);
4697	4697	horiz1 = (v16u8) __msa_splati_d((v2i64) horiz0, 1);
4698	4698	res1 = APPLY_HORIZ_QPEL_FILTER_8BYTE(inp2, inp3, mask0, mask1, mask2, mask3,
4699	4699	const20, const6, const3);
4700		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	4700	+ SLDI_B2_UB(inp2, inp2, inp3, inp3, 1, inp2, inp3);
4701	4701
4702	4702	inp2 = (v16u8) __msa_insve_d((v2i64) inp2, 1, (v2i64) inp3);
4703	4703	horiz2 = __msa_aver_u_b(inp2, res1);
@@ -4706,7 +4706,7 @@ static void hv_mc_qpel_aver_hv_src11_8x8_msa(const uint8_t *src, · 2 unmodified lines
4706	4706	src += (2 * src_stride);
4707	4707	res0 = APPLY_HORIZ_QPEL_FILTER_8BYTE(inp0, inp1, mask0, mask1, mask2, mask3,
4708	4708	const20, const6, const3);
4709		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
	4709	+ SLDI_B2_UB(inp0, inp0, inp1, inp1, 1, inp0, inp1);
4710	4710
4711	4711	inp0 = (v16u8) __msa_insve_d((v2i64) inp0, 1, (v2i64) inp1);
4712	4712	horiz4 = __msa_aver_u_b(inp0, res0);
@@ -4725,7 +4725,7 @@ static void hv_mc_qpel_aver_hv_src11_8x8_msa(const uint8_t *src, · 12 unmodified lines
4725	4725
4726	4726	res1 = APPLY_HORIZ_QPEL_FILTER_8BYTE(inp2, inp3, mask0, mask1, mask2, mask3,
4727	4727	const20, const6, const3);
4728		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	4728	+ SLDI_B2_UB(inp2, inp2, inp3, inp3, 1, inp2, inp3);
4729	4729
4730	4730	inp2 = (v16u8) __msa_insve_d((v2i64) inp2, 1, (v2i64) inp3);
4731	4731	horiz6 = __msa_aver_u_b(inp2, res1);
@@ -5020,7 +5020,7 @@ static void hv_mc_qpel_avg_dst_aver_hv_src10_8x8_msa(const uint8_t *src, · 288 unmodified lines
5020	5020
5021	5021	LD_UB2(src, src_stride, inp2, inp3);
5022	5022	src += (2 * src_stride);
5023		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
	5023	+ SLDI_B2_UB(inp0, inp0, inp1, inp1, 1, inp0, inp1);
5024	5024
5025	5025	inp0 = (v16u8) __msa_ilvr_d((v2i64) inp1, (v2i64) inp0);
5026	5026	horiz0 = __msa_aver_u_b(inp0, res0);
@@ -5029,7 +5029,7 @@ static void hv_mc_qpel_avg_dst_aver_hv_src10_8x8_msa(const uint8_t *src, · 2 unmodified lines
5029	5029	const20, const6, const3);
5030	5030	LD_UB2(src, src_stride, inp0, inp1);
5031	5031	src += (2 * src_stride);
5032		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	5032	+ SLDI_B2_UB(inp2, inp2, inp3, inp3, 1, inp2, inp3);
5033	5033
5034	5034	inp2 = (v16u8) __msa_ilvr_d((v2i64) inp3, (v2i64) inp2);
5035	5035	horiz2 = __msa_aver_u_b(inp2, res1);
@@ -5037,7 +5037,7 @@ static void hv_mc_qpel_avg_dst_aver_hv_src10_8x8_msa(const uint8_t *src, · 1 unmodified line
5037	5037	res0 = APPLY_HORIZ_QPEL_FILTER_8BYTE(inp0, inp1, mask0, mask1, mask2, mask3,
5038	5038	const20, const6, const3);
5039	5039
5040		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
	5040	+ SLDI_B2_UB(inp0, inp0, inp1, inp1, 1, inp0, inp1);
5041	5041
5042	5042	inp0 = (v16u8) __msa_ilvr_d((v2i64) inp1, (v2i64) inp0);
5043	5043	horiz4 = __msa_aver_u_b(inp0, res0);
@@ -5060,7 +5060,7 @@ static void hv_mc_qpel_avg_dst_aver_hv_src10_8x8_msa(const uint8_t *src, · 16 unmodified lines
5060	5060	res1 = APPLY_HORIZ_QPEL_FILTER_8BYTE(inp2, inp3, mask0, mask1, mask2, mask3,
5061	5061	const20, const6, const3);
5062	5062
5063		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	5063	+ SLDI_B2_UB(inp2, inp2, inp3, inp3, 1, inp2, inp3);
5064	5064
5065	5065	inp2 = (v16u8) __msa_ilvr_d((v2i64) inp3, (v2i64) inp2);
5066	5066	horiz6 = __msa_aver_u_b(inp2, res1);
@@ -5347,7 +5347,7 @@ static void hv_mc_qpel_avg_dst_aver_h_src1_8x8_msa(const uint8_t *src, · 280 unmodified lines
5347	5347	const20, const6, const3);
5348	5348	LD_UB2(src, src_stride, inp2, inp3);
5349	5349	src += (2 * src_stride);
5350		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
	5350	+ SLDI_B2_UB(inp0, inp0, inp1, inp1, 1, inp0, inp1);
5351	5351
5352	5352	inp0 = (v16u8) __msa_ilvr_d((v2i64) inp1, (v2i64) inp0);
5353	5353	horiz0 = __msa_aver_u_b(inp0, res0);
@@ -5356,7 +5356,7 @@ static void hv_mc_qpel_avg_dst_aver_h_src1_8x8_msa(const uint8_t *src, · 2 unmodified lines
5356	5356	const20, const6, const3);
5357	5357	LD_UB2(src, src_stride, inp0, inp1);
5358	5358	src += (2 * src_stride);
5359		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	5359	+ SLDI_B2_UB(inp2, inp2, inp3, inp3, 1, inp2, inp3);
5360	5360
5361	5361	inp2 = (v16u8) __msa_ilvr_d((v2i64) inp3, (v2i64) inp2);
5362	5362	horiz2 = __msa_aver_u_b(inp2, res1);
@@ -5364,7 +5364,7 @@ static void hv_mc_qpel_avg_dst_aver_h_src1_8x8_msa(const uint8_t *src, · 1 unmodified line
5364	5364	res0 = APPLY_HORIZ_QPEL_FILTER_8BYTE(inp0, inp1, mask0, mask1, mask2, mask3,
5365	5365	const20, const6, const3);
5366	5366
5367		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
	5367	+ SLDI_B2_UB(inp0, inp0, inp1, inp1, 1, inp0, inp1);
5368	5368
5369	5369	inp0 = (v16u8) __msa_ilvr_d((v2i64) inp1, (v2i64) inp0);
5370	5370	horiz4 = __msa_aver_u_b(inp0, res0);
@@ -5385,7 +5385,7 @@ static void hv_mc_qpel_avg_dst_aver_h_src1_8x8_msa(const uint8_t *src, · 14 unmodified lines
5385	5385	res1 = APPLY_HORIZ_QPEL_FILTER_8BYTE(inp2, inp3, mask0, mask1, mask2, mask3,
5386	5386	const20, const6, const3);
5387	5387
5388		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	5388	+ SLDI_B2_UB(inp2, inp2, inp3, inp3, 1, inp2, inp3);
5389	5389
5390	5390	inp2 = (v16u8) __msa_ilvr_d((v2i64) inp3, (v2i64) inp2);
5391	5391	horiz6 = __msa_aver_u_b(inp2, res1);
@@ -5684,7 +5684,7 @@ static void hv_mc_qpel_avg_dst_aver_hv_src11_8x8_msa(const uint8_t *src, · 292 unmodified lines
5684	5684	const20, const6, const3);
5685	5685	LD_UB2(src, src_stride, inp2, inp3);
5686	5686	src += (2 * src_stride);
5687		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
	5687	+ SLDI_B2_UB(inp0, inp0, inp1, inp1, 1, inp0, inp1);
5688	5688
5689	5689	inp0 = (v16u8) __msa_ilvr_d((v2i64) inp1, (v2i64) inp0);
5690	5690	horiz0 = __msa_aver_u_b(inp0, res0);
@@ -5693,14 +5693,14 @@ static void hv_mc_qpel_avg_dst_aver_hv_src11_8x8_msa(const uint8_t *src, · 2 unmodified lines
5693	5693	const20, const6, const3);
5694	5694	LD_UB2(src, src_stride, inp0, inp1);
5695	5695	src += (2 * src_stride);
5696		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	5696	+ SLDI_B2_UB(inp2, inp2, inp3, inp3, 1, inp2, inp3);
5697	5697
5698	5698	inp2 = (v16u8) __msa_ilvr_d((v2i64) inp3, (v2i64) inp2);
5699	5699	horiz2 = __msa_aver_u_b(inp2, res1);
5700	5700	horiz3 = (v16u8) __msa_splati_d((v2i64) horiz2, 1);
5701	5701	res0 = APPLY_HORIZ_QPEL_FILTER_8BYTE(inp0, inp1, mask0, mask1, mask2, mask3,
5702	5702	const20, const6, const3);
5703		− SLDI_B2_UB(inp0, inp1, inp0, inp1, inp0, inp1, 1);
	5703	+ SLDI_B2_UB(inp0, inp0, inp1, inp1, 1, inp0, inp1);
5704	5704
5705	5705	inp0 = (v16u8) __msa_ilvr_d((v2i64) inp1, (v2i64) inp0);
5706	5706	horiz4 = __msa_aver_u_b(inp0, res0);
@@ -5721,7 +5721,7 @@ static void hv_mc_qpel_avg_dst_aver_hv_src11_8x8_msa(const uint8_t *src, · 14 unmodified lines
5721	5721	src += (2 * src_stride);
5722	5722	res1 = APPLY_HORIZ_QPEL_FILTER_8BYTE(inp2, inp3, mask0, mask1, mask2, mask3,
5723	5723	const20, const6, const3);
5724		− SLDI_B2_UB(inp2, inp3, inp2, inp3, inp2, inp3, 1);
	5724	+ SLDI_B2_UB(inp2, inp2, inp3, inp3, 1, inp2, inp3);
5725	5725
5726	5726	inp2 = (v16u8) __msa_ilvr_d((v2i64) inp3, (v2i64) inp2);
5727	5727	horiz6 = __msa_aver_u_b(inp2, res1);

libavcodec/mips/vp8_mc_msa.c+2−2

@@ -1995,8 +1995,8 @@ static void common_hv_2ht_2vt_4x8_msa(uint8_t *src, int32_t src_stride, · 1994 unmodified lines
1995	1995	hz_out4 = HORIZ_2TAP_FILT_UH(src4, src5, mask, filt_hz, 7);
1996	1996	hz_out6 = HORIZ_2TAP_FILT_UH(src6, src7, mask, filt_hz, 7);
1997	1997	hz_out8 = HORIZ_2TAP_FILT_UH(src8, src8, mask, filt_hz, 7);
1998		− SLDI_B3_UH(hz_out2, hz_out4, hz_out6, hz_out0, hz_out2, hz_out4, hz_out1,
1999		− hz_out3, hz_out5, 8);
	1998	+ SLDI_B3_UH(hz_out2, hz_out0, hz_out4, hz_out2, hz_out6, hz_out4, 8, hz_out1,
	1999	+ hz_out3, hz_out5);
2000	2000	hz_out7 = (v8u16) __msa_pckod_d((v2i64) hz_out8, (v2i64) hz_out6);
2001	2001
2002	2002	ILVEV_B2_UB(hz_out0, hz_out1, hz_out2, hz_out3, vec0, vec1);

libavcodec/mips/vp9_idct_msa.c+2−1

@@ -249,6 +249,7 @@ static const int32_t sinpi_4_9 = 15212; · 248 unmodified lines
249	249	v8i16 c0_m, c1_m, c2_m, c3_m; \
250	250	v8i16 step0_m, step1_m; \
251	251	v4i32 tmp0_m, tmp1_m, tmp2_m, tmp3_m; \
	252	+ v16i8 zeros = { 0 }; \
252	253	\
253	254	c0_m = VP9_SET_COSPI_PAIR(cospi_16_64, cospi_16_64); \
254	255	c1_m = VP9_SET_COSPI_PAIR(cospi_16_64, -cospi_16_64); \
@@ -262,7 +263,7 @@ static const int32_t sinpi_4_9 = 15212; · 7 unmodified lines
262	263	SRARI_W4_SW(tmp0_m, tmp1_m, tmp2_m, tmp3_m, VP9_DCT_CONST_BITS); \
263	264	\
264	265	PCKEV_H2_SW(tmp1_m, tmp0_m, tmp3_m, tmp2_m, tmp0_m, tmp2_m); \
265		− SLDI_B2_0_SW(tmp0_m, tmp2_m, tmp1_m, tmp3_m, 8); \
	266	+ SLDI_B2_SW(zeros, tmp0_m, zeros, tmp2_m, 8, tmp1_m, tmp3_m); \
266	267	BUTTERFLY_4((v8i16) tmp0_m, (v8i16) tmp1_m, \
267	268	(v8i16) tmp2_m, (v8i16) tmp3_m, \
268	269	out0, out1, out2, out3); \

libavcodec/mips/vp9_lpf_msa.c+2−1

@@ -1673,6 +1673,7 @@ static void vp9_transpose_16x8_to_8x16(uint8_t *input, int32_t in_pitch, · 1672 unmodified lines
1673	1673	v16u8 p7_org, p6_org, p5_org, p4_org, p3_org, p2_org, p1_org, p0_org;
1674	1674	v16i8 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
1675	1675	v16u8 p7, p6, p5, p4, p3, p2, p1, p0, q0, q1, q2, q3, q4, q5, q6, q7;
	1676	+ v16i8 zeros = { 0 };
1676	1677
1677	1678	LD_UB8(input, in_pitch,
1678	1679	p7_org, p6_org, p5_org, p4_org, p3_org, p2_org, p1_org, p0_org);
@@ -1686,7 +1687,7 @@ static void vp9_transpose_16x8_to_8x16(uint8_t *input, int32_t in_pitch, · 7 unmodified lines
1686	1687	ILVL_B2_SB(tmp1, tmp0, tmp3, tmp2, tmp5, tmp7);
1687	1688	ILVR_W2_UB(tmp6, tmp4, tmp7, tmp5, q0, q4);
1688	1689	ILVL_W2_UB(tmp6, tmp4, tmp7, tmp5, q2, q6);
1689		− SLDI_B4_0_UB(q0, q2, q4, q6, q1, q3, q5, q7, 8);
	1690	+ SLDI_B4_UB(zeros, q0, zeros, q2, zeros, q4, zeros, q6, 8, q1, q3, q5, q7);
1690	1691
1691	1692	ST_UB8(p7, p6, p5, p4, p3, p2, p1, p0, output, out_pitch);
1692	1693	output += (8 * out_pitch);

libavcodec/mips/vp9_mc_msa.c+8−8

@@ -795,7 +795,7 @@ static void common_hv_8ht_8vt_4w_msa(const uint8_t *src, int32_t src_stride, · 794 unmodified lines
795	795	filt_hz1, filt_hz2, filt_hz3);
796	796	hz_out5 = HORIZ_8TAP_FILT(src5, src6, mask0, mask1, mask2, mask3, filt_hz0,
797	797	filt_hz1, filt_hz2, filt_hz3);
798		− SLDI_B2_SH(hz_out2, hz_out4, hz_out0, hz_out2, hz_out1, hz_out3, 8);
	798	+ SLDI_B2_SH(hz_out2, hz_out0, hz_out4, hz_out2, 8, hz_out1, hz_out3);
799	799
800	800	filt = LD_SH(filter_vert);
801	801	SPLATI_H4_SH(filt, 0, 1, 2, 3, filt_vt0, filt_vt1, filt_vt2, filt_vt3);
@@ -1585,7 +1585,7 @@ static void common_hv_8ht_8vt_and_aver_dst_4w_msa(const uint8_t *src, · 783 unmodified lines
1585	1585	filt_hz1, filt_hz2, filt_hz3);
1586	1586	hz_out5 = HORIZ_8TAP_FILT(src5, src6, mask0, mask1, mask2, mask3, filt_hz0,
1587	1587	filt_hz1, filt_hz2, filt_hz3);
1588		− SLDI_B2_SH(hz_out2, hz_out4, hz_out0, hz_out2, hz_out1, hz_out3, 8);
	1588	+ SLDI_B2_SH(hz_out2, hz_out0, hz_out4, hz_out2, 8, hz_out1, hz_out3);
1589	1589
1590	1590	filt = LD_SH(filter_vert);
1591	1591	SPLATI_H4_SH(filt, 0, 1, 2, 3, filt_vt0, filt_vt1, filt_vt2, filt_vt3);
@@ -2093,7 +2093,7 @@ void ff_put_bilin_64h_msa(uint8_t *dst, ptrdiff_t dst_stride, · 501 unmodified lines
2093	2093	src4 = LD_SB(src + 32);
2094	2094	src6 = LD_SB(src + 48);
2095	2095	src7 = LD_SB(src + 56);
2096		− SLDI_B3_SB(src2, src4, src6, src0, src2, src4, src1, src3, src5, 8);
	2096	+ SLDI_B3_SB(src2, src0, src4, src2, src6, src4, 8, src1, src3, src5);
2097	2097	src += src_stride;
2098	2098
2099	2099	VSHF_B2_UB(src0, src0, src1, src1, mask, mask, vec0, vec1);
@@ -2544,8 +2544,8 @@ static void common_hv_2ht_2vt_4x8_msa(const uint8_t *src, int32_t src_stride, · 444 unmodified lines
2544	2544	hz_out4 = HORIZ_2TAP_FILT_UH(src4, src5, mask, filt_hz, 7);
2545	2545	hz_out6 = HORIZ_2TAP_FILT_UH(src6, src7, mask, filt_hz, 7);
2546	2546	hz_out8 = HORIZ_2TAP_FILT_UH(src8, src8, mask, filt_hz, 7);
2547		− SLDI_B3_UH(hz_out2, hz_out4, hz_out6, hz_out0, hz_out2, hz_out4, hz_out1,
2548		− hz_out3, hz_out5, 8);
	2547	+ SLDI_B3_UH(hz_out2, hz_out0, hz_out4, hz_out2, hz_out6, hz_out4, 8, hz_out1,
	2548	+ hz_out3, hz_out5);
2549	2549	hz_out7 = (v8u16) __msa_pckod_d((v2i64) hz_out8, (v2i64) hz_out6);
2550	2550
2551	2551	ILVEV_B2_UB(hz_out0, hz_out1, hz_out2, hz_out3, vec0, vec1);
@@ -3146,7 +3146,7 @@ void ff_avg_bilin_64h_msa(uint8_t *dst, ptrdiff_t dst_stride, · 594 unmodified lines
3146	3146	for (loop_cnt = height; loop_cnt--;) {
3147	3147	LD_SB4(src, 16, src0, src2, src4, src6);
3148	3148	src7 = LD_SB(src + 56);
3149		− SLDI_B3_SB(src2, src4, src6, src0, src2, src4, src1, src3, src5, 8);
	3149	+ SLDI_B3_SB(src2, src0, src4, src2, src6, src4, 8, src1, src3, src5);
3150	3150	src += src_stride;
3151	3151
3152	3152	VSHF_B2_UB(src0, src0, src1, src1, mask, mask, vec0, vec1);
@@ -3655,8 +3655,8 @@ static void common_hv_2ht_2vt_and_aver_dst_4x8_msa(const uint8_t *src, · 502 unmodified lines
3655	3655	hz_out4 = HORIZ_2TAP_FILT_UH(src4, src5, mask, filt_hz, 7);
3656	3656	hz_out6 = HORIZ_2TAP_FILT_UH(src6, src7, mask, filt_hz, 7);
3657	3657	hz_out8 = HORIZ_2TAP_FILT_UH(src8, src8, mask, filt_hz, 7);
3658		− SLDI_B3_UH(hz_out2, hz_out4, hz_out6, hz_out0, hz_out2, hz_out4, hz_out1,
3659		− hz_out3, hz_out5, 8);
	3658	+ SLDI_B3_UH(hz_out2, hz_out0, hz_out4, hz_out2, hz_out6, hz_out4, 8, hz_out1,
	3659	+ hz_out3, hz_out5);
3660	3660	hz_out7 = (v8u16) __msa_pckod_d((v2i64) hz_out8, (v2i64) hz_out6);
3661	3661
3662	3662	LW4(dst, dst_stride, tp0, tp1, tp2, tp3);

libavutil/mips/generic_macros_msa.h+31−49

@@ -602,67 +602,48 @@ · 601 unmodified lines
602	602	}
603	603	#define AVER_UB4_UB(...) AVER_UB4(v16u8, __VA_ARGS__)
604	604
605		−/* Description : Immediate number of columns to slide with zero
606		− Arguments : Inputs - in0, in1, slide_val
607		− Outputs - out0, out1
	605	+/* Description : Immediate number of columns to slide
	606	+ Arguments : Inputs - s, d, slide_val
	607	+ Outputs - out
608	608	Return Type - as per RTYPE
609		− Details : Byte elements from 'zero_m' vector are slide into 'in0' by
	609	+ Details : Byte elements from 'd' vector are slide into 's' by
610	610	number of elements specified by 'slide_val'
611	611	*/
612		−#define SLDI_B2_0(RTYPE, in0, in1, out0, out1, slide_val) \
613		−{ \
614		− v16i8 zero_m = { 0 }; \
615		− out0 = (RTYPE) __msa_sldi_b((v16i8) zero_m, (v16i8) in0, slide_val); \
616		− out1 = (RTYPE) __msa_sldi_b((v16i8) zero_m, (v16i8) in1, slide_val); \
617		−}
618		−#define SLDI_B2_0_UB(...) SLDI_B2_0(v16u8, __VA_ARGS__)
619		−#define SLDI_B2_0_SB(...) SLDI_B2_0(v16i8, __VA_ARGS__)
620		−#define SLDI_B2_0_SW(...) SLDI_B2_0(v4i32, __VA_ARGS__)
621		−
622		−#define SLDI_B3_0(RTYPE, in0, in1, in2, out0, out1, out2, slide_val) \
623		−{ \
624		− v16i8 zero_m = { 0 }; \
625		− SLDI_B2_0(RTYPE, in0, in1, out0, out1, slide_val); \
626		− out2 = (RTYPE) __msa_sldi_b((v16i8) zero_m, (v16i8) in2, slide_val); \
627		−}
628		−#define SLDI_B3_0_UB(...) SLDI_B3_0(v16u8, __VA_ARGS__)
629		−#define SLDI_B3_0_SB(...) SLDI_B3_0(v16i8, __VA_ARGS__)
630		−
631		−#define SLDI_B4_0(RTYPE, in0, in1, in2, in3, \
632		− out0, out1, out2, out3, slide_val) \
633		−{ \
634		− SLDI_B2_0(RTYPE, in0, in1, out0, out1, slide_val); \
635		− SLDI_B2_0(RTYPE, in2, in3, out2, out3, slide_val); \
	612	+#define SLDI_B(RTYPE, d, s, slide_val, out) \
	613	+{ \
	614	+ out = (RTYPE) __msa_sldi_b((v16i8) d, (v16i8) s, slide_val); \
636	615	}
637		−#define SLDI_B4_0_UB(...) SLDI_B4_0(v16u8, __VA_ARGS__)
638		−#define SLDI_B4_0_SB(...) SLDI_B4_0(v16i8, __VA_ARGS__)
639		−#define SLDI_B4_0_SH(...) SLDI_B4_0(v8i16, __VA_ARGS__)
640	616
641		−/* Description : Immediate number of columns to slide
642		− Arguments : Inputs - in0_0, in0_1, in1_0, in1_1, slide_val
643		− Outputs - out0, out1
644		− Return Type - as per RTYPE
645		− Details : Byte elements from 'in0_0' vector are slide into 'in1_0' by
646		− number of elements specified by 'slide_val'
647		−*/
648		−#define SLDI_B2(RTYPE, in0_0, in0_1, in1_0, in1_1, out0, out1, slide_val) \
649		−{ \
650		− out0 = (RTYPE) __msa_sldi_b((v16i8) in0_0, (v16i8) in1_0, slide_val); \
651		− out1 = (RTYPE) __msa_sldi_b((v16i8) in0_1, (v16i8) in1_1, slide_val); \
	617	+#define SLDI_B2(RTYPE, d0, s0, d1, s1, slide_val, out0, out1) \
	618	+{ \
	619	+ SLDI_B(RTYPE, d0, s0, slide_val, out0) \
	620	+ SLDI_B(RTYPE, d1, s1, slide_val, out1) \
652	621	}
653	622	#define SLDI_B2_UB(...) SLDI_B2(v16u8, __VA_ARGS__)
654	623	#define SLDI_B2_SB(...) SLDI_B2(v16i8, __VA_ARGS__)
655	624	#define SLDI_B2_SH(...) SLDI_B2(v8i16, __VA_ARGS__)
	625	+#define SLDI_B2_SW(...) SLDI_B2(v4i32, __VA_ARGS__)
656	626
657		−#define SLDI_B3(RTYPE, in0_0, in0_1, in0_2, in1_0, in1_1, in1_2, \
658		− out0, out1, out2, slide_val) \
659		−{ \
660		− SLDI_B2(RTYPE, in0_0, in0_1, in1_0, in1_1, out0, out1, slide_val) \
661		− out2 = (RTYPE) __msa_sldi_b((v16i8) in0_2, (v16i8) in1_2, slide_val); \
	627	+#define SLDI_B3(RTYPE, d0, s0, d1, s1, d2, s2, slide_val, \
	628	+ out0, out1, out2) \
	629	+{ \
	630	+ SLDI_B2(RTYPE, d0, s0, d1, s1, slide_val, out0, out1) \
	631	+ SLDI_B(RTYPE, d2, s2, slide_val, out2) \
662	632	}
	633	+#define SLDI_B3_UB(...) SLDI_B3(v16u8, __VA_ARGS__)
663	634	#define SLDI_B3_SB(...) SLDI_B3(v16i8, __VA_ARGS__)
664	635	#define SLDI_B3_UH(...) SLDI_B3(v8u16, __VA_ARGS__)
665	636
	637	+#define SLDI_B4(RTYPE, d0, s0, d1, s1, d2, s2, d3, s3, \
	638	+ slide_val, out0, out1, out2, out3) \
	639	+{ \
	640	+ SLDI_B2(RTYPE, d0, s0, d1, s1, slide_val, out0, out1) \
	641	+ SLDI_B2(RTYPE, d2, s2, d3, s3, slide_val, out2, out3) \
	642	+}
	643	+#define SLDI_B4_UB(...) SLDI_B4(v16u8, __VA_ARGS__)
	644	+#define SLDI_B4_SB(...) SLDI_B4(v16i8, __VA_ARGS__)
	645	+#define SLDI_B4_SH(...) SLDI_B4(v8i16, __VA_ARGS__)
	646	+
666	647	/* Description : Shuffle byte vector elements as per mask vector
667	648	Arguments : Inputs - in0, in1, in2, in3, mask0, mask1
668	649	Outputs - out0, out1
@@ -2412,6 +2393,7 @@ · 1743 unmodified lines
2412	2393	{ \
2413	2394	v16i8 tmp0_m, tmp1_m, tmp2_m, tmp3_m; \
2414	2395	v16i8 tmp4_m, tmp5_m, tmp6_m, tmp7_m; \
	2396	+ v16i8 zeros = { 0 }; \
2415	2397	\
2416	2398	ILVR_B4_SB(in2, in0, in3, in1, in6, in4, in7, in5, \
2417	2399	tmp0_m, tmp1_m, tmp2_m, tmp3_m); \
@@ -2419,8 +2401,8 @@ · 1 unmodified line
2419	2401	ILVRL_B2_SB(tmp3_m, tmp2_m, tmp6_m, tmp7_m); \
2420	2402	ILVRL_W2(RTYPE, tmp6_m, tmp4_m, out0, out2); \
2421	2403	ILVRL_W2(RTYPE, tmp7_m, tmp5_m, out4, out6); \
2422		− SLDI_B2_0(RTYPE, out0, out2, out1, out3, 8); \
2423		− SLDI_B2_0(RTYPE, out4, out6, out5, out7, 8); \
	2404	+ SLDI_B4(RTYPE, zeros, out0, zeros, out2, zeros, out4, zeros, out6, \
	2405	+ 8, out1, out3, out5, out7); \
2424	2406	}
2425	2407	#define TRANSPOSE8x8_UB_UB(...) TRANSPOSE8x8_UB(v16u8, __VA_ARGS__)
2426	2408	#define TRANSPOSE8x8_UB_UH(...) TRANSPOSE8x8_UB(v8u16, __VA_ARGS__)

avutil/mips: refactor msa SLDI_Bn_0 and SLDI_Bn macros.

Keyboard shortcuts

General

Issues and pull requests

Issues, pull requests and Inbox

Repository code