History

aarch64: Consistently use lowercase for vector element specifiers

Martin Storsjö committed 3 years ago

Commit 184103b1 parent 393d1ee16 files · +2231 −2231Browse files

Unified · Split

16 files
  1. libavcodec/aarch64/aacpsdsp_neon.S +97 −97
  2. libavcodec/aarch64/h264cmc_neon.S +203 −203
  3. libavcodec/aarch64/h264dsp_neon.S +297 −297
  4. libavcodec/aarch64/h264idct_neon.S +195 −195
  5. libavcodec/aarch64/h264qpel_neon.S +278 −278
  6. libavcodec/aarch64/hpeldsp_neon.S +181 −181
  7. libavcodec/aarch64/me_cmp_neon.S +1 −1
  8. libavcodec/aarch64/neon.S +123 −123
  9. libavcodec/aarch64/sbrdsp_neon.S +147 −147
  10. libavcodec/aarch64/simple_idct_neon.S +193 −193
  11. libavfilter/aarch64/vf_nlmeans_neon.S +35 −35
  12. libavutil/aarch64/float_dsp_neon.S +100 −100
  13. libswresample/aarch64/resample.S +32 −32
  14. libswscale/aarch64/hscale.S +216 −216
  15. libswscale/aarch64/output.S +75 −75
  16. libswscale/aarch64/yuv2rgb_neon.S +58 −58
libavcodec/aarch64/aacpsdsp_neon.S+97−97
@@ -19,82 +19,82 @@ · 18 unmodified lines
1919 #include "libavutil/aarch64/asm.S"
2020
2121 function ff_ps_add_squares_neon, export=1
22−1: ld1 {v0.4S,v1.4S}, [x1], #32
23− fmul v0.4S, v0.4S, v0.4S
24− fmul v1.4S, v1.4S, v1.4S
25− faddp v2.4S, v0.4S, v1.4S
26− ld1 {v3.4S}, [x0]
27− fadd v3.4S, v3.4S, v2.4S
28− st1 {v3.4S}, [x0], #16
22+1: ld1 {v0.4s,v1.4s}, [x1], #32
23+ fmul v0.4s, v0.4s, v0.4s
24+ fmul v1.4s, v1.4s, v1.4s
25+ faddp v2.4s, v0.4s, v1.4s
26+ ld1 {v3.4s}, [x0]
27+ fadd v3.4s, v3.4s, v2.4s
28+ st1 {v3.4s}, [x0], #16
2929 subs w2, w2, #4
3030 b.gt 1b
3131 ret
3232 endfunc
3333
3434 function ff_ps_mul_pair_single_neon, export=1
35−1: ld1 {v0.4S,v1.4S}, [x1], #32
36− ld1 {v2.4S}, [x2], #16
37− zip1 v3.4S, v2.4S, v2.4S
38− zip2 v4.4S, v2.4S, v2.4S
39− fmul v0.4S, v0.4S, v3.4S
40− fmul v1.4S, v1.4S, v4.4S
41− st1 {v0.4S,v1.4S}, [x0], #32
35+1: ld1 {v0.4s,v1.4s}, [x1], #32
36+ ld1 {v2.4s}, [x2], #16
37+ zip1 v3.4s, v2.4s, v2.4s
38+ zip2 v4.4s, v2.4s, v2.4s
39+ fmul v0.4s, v0.4s, v3.4s
40+ fmul v1.4s, v1.4s, v4.4s
41+ st1 {v0.4s,v1.4s}, [x0], #32
4242 subs w3, w3, #4
4343 b.gt 1b
4444 ret
4545 endfunc
4646
4747 function ff_ps_stereo_interpolate_neon, export=1
48− ld1 {v0.4S}, [x2]
49− ld1 {v1.4S}, [x3]
50− zip1 v4.4S, v0.4S, v0.4S
51− zip2 v5.4S, v0.4S, v0.4S
52− zip1 v6.4S, v1.4S, v1.4S
53− zip2 v7.4S, v1.4S, v1.4S
54−1: ld1 {v2.2S}, [x0]
55− ld1 {v3.2S}, [x1]
56− fadd v4.4S, v4.4S, v6.4S
57− fadd v5.4S, v5.4S, v7.4S
58− mov v2.D[1], v2.D[0]
59− mov v3.D[1], v3.D[0]
60− fmul v2.4S, v2.4S, v4.4S
61− fmla v2.4S, v3.4S, v5.4S
62− st1 {v2.D}[0], [x0], #8
63− st1 {v2.D}[1], [x1], #8
48+ ld1 {v0.4s}, [x2]
49+ ld1 {v1.4s}, [x3]
50+ zip1 v4.4s, v0.4s, v0.4s
51+ zip2 v5.4s, v0.4s, v0.4s
52+ zip1 v6.4s, v1.4s, v1.4s
53+ zip2 v7.4s, v1.4s, v1.4s
54+1: ld1 {v2.2s}, [x0]
55+ ld1 {v3.2s}, [x1]
56+ fadd v4.4s, v4.4s, v6.4s
57+ fadd v5.4s, v5.4s, v7.4s
58+ mov v2.d[1], v2.d[0]
59+ mov v3.d[1], v3.d[0]
60+ fmul v2.4s, v2.4s, v4.4s
61+ fmla v2.4s, v3.4s, v5.4s
62+ st1 {v2.d}[0], [x0], #8
63+ st1 {v2.d}[1], [x1], #8
6464 subs w4, w4, #1
6565 b.gt 1b
6666 ret
6767 endfunc
6868
6969 function ff_ps_stereo_interpolate_ipdopd_neon, export=1
70− ld1 {v0.4S,v1.4S}, [x2]
71− ld1 {v6.4S,v7.4S}, [x3]
72− fneg v2.4S, v1.4S
73− fneg v3.4S, v7.4S
74− zip1 v16.4S, v0.4S, v0.4S
75− zip2 v17.4S, v0.4S, v0.4S
76− zip1 v18.4S, v2.4S, v1.4S
77− zip2 v19.4S, v2.4S, v1.4S
78− zip1 v20.4S, v6.4S, v6.4S
79− zip2 v21.4S, v6.4S, v6.4S
80− zip1 v22.4S, v3.4S, v7.4S
81− zip2 v23.4S, v3.4S, v7.4S
82−1: ld1 {v2.2S}, [x0]
83− ld1 {v3.2S}, [x1]
84− fadd v16.4S, v16.4S, v20.4S
85− fadd v17.4S, v17.4S, v21.4S
86− mov v2.D[1], v2.D[0]
87− mov v3.D[1], v3.D[0]
88− fmul v4.4S, v2.4S, v16.4S
89− fmla v4.4S, v3.4S, v17.4S
90− fadd v18.4S, v18.4S, v22.4S
91− fadd v19.4S, v19.4S, v23.4S
92− ext v2.16B, v2.16B, v2.16B, #4
93− ext v3.16B, v3.16B, v3.16B, #4
94− fmla v4.4S, v2.4S, v18.4S
95− fmla v4.4S, v3.4S, v19.4S
96− st1 {v4.D}[0], [x0], #8
97− st1 {v4.D}[1], [x1], #8
70+ ld1 {v0.4s,v1.4s}, [x2]
71+ ld1 {v6.4s,v7.4s}, [x3]
72+ fneg v2.4s, v1.4s
73+ fneg v3.4s, v7.4s
74+ zip1 v16.4s, v0.4s, v0.4s
75+ zip2 v17.4s, v0.4s, v0.4s
76+ zip1 v18.4s, v2.4s, v1.4s
77+ zip2 v19.4s, v2.4s, v1.4s
78+ zip1 v20.4s, v6.4s, v6.4s
79+ zip2 v21.4s, v6.4s, v6.4s
80+ zip1 v22.4s, v3.4s, v7.4s
81+ zip2 v23.4s, v3.4s, v7.4s
82+1: ld1 {v2.2s}, [x0]
83+ ld1 {v3.2s}, [x1]
84+ fadd v16.4s, v16.4s, v20.4s
85+ fadd v17.4s, v17.4s, v21.4s
86+ mov v2.d[1], v2.d[0]
87+ mov v3.d[1], v3.d[0]
88+ fmul v4.4s, v2.4s, v16.4s
89+ fmla v4.4s, v3.4s, v17.4s
90+ fadd v18.4s, v18.4s, v22.4s
91+ fadd v19.4s, v19.4s, v23.4s
92+ ext v2.16b, v2.16b, v2.16b, #4
93+ ext v3.16b, v3.16b, v3.16b, #4
94+ fmla v4.4s, v2.4s, v18.4s
95+ fmla v4.4s, v3.4s, v19.4s
96+ st1 {v4.d}[0], [x0], #8
97+ st1 {v4.d}[1], [x1], #8
9898 subs w4, w4, #1
9999 b.gt 1b
100100 ret
@@ -102,46 +102,46 @@ endfunc · 1 unmodified line
102102
103103 function ff_ps_hybrid_analysis_neon, export=1
104104 lsl x3, x3, #3
105− ld2 {v0.4S,v1.4S}, [x1], #32
106− ld2 {v2.2S,v3.2S}, [x1], #16
107− ld1 {v24.2S}, [x1], #8
108− ld2 {v4.2S,v5.2S}, [x1], #16
109− ld2 {v6.4S,v7.4S}, [x1]
110− rev64 v6.4S, v6.4S
111− rev64 v7.4S, v7.4S
112− ext v6.16B, v6.16B, v6.16B, #8
113− ext v7.16B, v7.16B, v7.16B, #8
114− rev64 v4.2S, v4.2S
115− rev64 v5.2S, v5.2S
116− mov v2.D[1], v3.D[0]
117− mov v4.D[1], v5.D[0]
118− mov v5.D[1], v2.D[0]
119− mov v3.D[1], v4.D[0]
120− fadd v16.4S, v0.4S, v6.4S
121− fadd v17.4S, v1.4S, v7.4S
122− fsub v18.4S, v1.4S, v7.4S
123− fsub v19.4S, v0.4S, v6.4S
124− fadd v22.4S, v2.4S, v4.4S
125− fsub v23.4S, v5.4S, v3.4S
126− trn1 v20.2D, v22.2D, v23.2D // {re4+re8, re5+re7, im8-im4, im7-im5}
127− trn2 v21.2D, v22.2D, v23.2D // {im4+im8, im5+im7, re4-re8, re5-re7}
128−1: ld2 {v2.4S,v3.4S}, [x2], #32
129− ld2 {v4.2S,v5.2S}, [x2], #16
130− ld1 {v6.2S}, [x2], #8
105+ ld2 {v0.4s,v1.4s}, [x1], #32
106+ ld2 {v2.2s,v3.2s}, [x1], #16
107+ ld1 {v24.2s}, [x1], #8
108+ ld2 {v4.2s,v5.2s}, [x1], #16
109+ ld2 {v6.4s,v7.4s}, [x1]
110+ rev64 v6.4s, v6.4s
111+ rev64 v7.4s, v7.4s
112+ ext v6.16b, v6.16b, v6.16b, #8
113+ ext v7.16b, v7.16b, v7.16b, #8
114+ rev64 v4.2s, v4.2s
115+ rev64 v5.2s, v5.2s
116+ mov v2.d[1], v3.d[0]
117+ mov v4.d[1], v5.d[0]
118+ mov v5.d[1], v2.d[0]
119+ mov v3.d[1], v4.d[0]
120+ fadd v16.4s, v0.4s, v6.4s
121+ fadd v17.4s, v1.4s, v7.4s
122+ fsub v18.4s, v1.4s, v7.4s
123+ fsub v19.4s, v0.4s, v6.4s
124+ fadd v22.4s, v2.4s, v4.4s
125+ fsub v23.4s, v5.4s, v3.4s
126+ trn1 v20.2d, v22.2d, v23.2d // {re4+re8, re5+re7, im8-im4, im7-im5}
127+ trn2 v21.2d, v22.2d, v23.2d // {im4+im8, im5+im7, re4-re8, re5-re7}
128+1: ld2 {v2.4s,v3.4s}, [x2], #32
129+ ld2 {v4.2s,v5.2s}, [x2], #16
130+ ld1 {v6.2s}, [x2], #8
131131 add x2, x2, #8
132− mov v4.D[1], v5.D[0]
133− mov v6.S[1], v6.S[0]
134− fmul v6.2S, v6.2S, v24.2S
135− fmul v0.4S, v2.4S, v16.4S
136− fmul v1.4S, v2.4S, v17.4S
137− fmls v0.4S, v3.4S, v18.4S
138− fmla v1.4S, v3.4S, v19.4S
139− fmla v0.4S, v4.4S, v20.4S
140− fmla v1.4S, v4.4S, v21.4S
141− faddp v0.4S, v0.4S, v1.4S
142− faddp v0.4S, v0.4S, v0.4S
143− fadd v0.2S, v0.2S, v6.2S
144− st1 {v0.2S}, [x0], x3
132+ mov v4.d[1], v5.d[0]
133+ mov v6.s[1], v6.s[0]
134+ fmul v6.2s, v6.2s, v24.2s
135+ fmul v0.4s, v2.4s, v16.4s
136+ fmul v1.4s, v2.4s, v17.4s
137+ fmls v0.4s, v3.4s, v18.4s
138+ fmla v1.4s, v3.4s, v19.4s
139+ fmla v0.4s, v4.4s, v20.4s
140+ fmla v1.4s, v4.4s, v21.4s
141+ faddp v0.4s, v0.4s, v1.4s
142+ faddp v0.4s, v0.4s, v0.4s
143+ fadd v0.2s, v0.2s, v6.2s
144+ st1 {v0.2s}, [x0], x3
145145 subs w4, w4, #1
146146 b.gt 1b
147147 ret
libavcodec/aarch64/h264cmc_neon.S+203−203
@@ -39,10 +39,10 @@ function ff_\type\()_\codec\()_chroma_mc8_neon, export=1 · 38 unmodified lines
3939 lsl w10, w10, #1
4040 add w9, w9, w10
4141 add x6, x6, w9, UXTW
42− ld1r {v22.8H}, [x6]
42+ ld1r {v22.8h}, [x6]
4343 .endif
4444 .ifc \codec,vc1
45− movi v22.8H, #28
45+ movi v22.8h, #28
4646 .endif
4747 mul w7, w4, w5
4848 lsl w14, w5, #3
@@ -55,139 +55,139 @@ function ff_\type\()_\codec\()_chroma_mc8_neon, export=1 · 6 unmodified lines
5555 add w4, w4, #64
5656 b.eq 2f
5757
58− dup v0.8B, w4
59− dup v1.8B, w12
60− ld1 {v4.8B, v5.8B}, [x1], x2
61− dup v2.8B, w6
62− dup v3.8B, w7
63− ext v5.8B, v4.8B, v5.8B, #1
64−1: ld1 {v6.8B, v7.8B}, [x1], x2
65− umull v16.8H, v4.8B, v0.8B
66− umlal v16.8H, v5.8B, v1.8B
67− ext v7.8B, v6.8B, v7.8B, #1
68− ld1 {v4.8B, v5.8B}, [x1], x2
69− umlal v16.8H, v6.8B, v2.8B
58+ dup v0.8b, w4
59+ dup v1.8b, w12
60+ ld1 {v4.8b, v5.8b}, [x1], x2
61+ dup v2.8b, w6
62+ dup v3.8b, w7
63+ ext v5.8b, v4.8b, v5.8b, #1
64+1: ld1 {v6.8b, v7.8b}, [x1], x2
65+ umull v16.8h, v4.8b, v0.8b
66+ umlal v16.8h, v5.8b, v1.8b
67+ ext v7.8b, v6.8b, v7.8b, #1
68+ ld1 {v4.8b, v5.8b}, [x1], x2
69+ umlal v16.8h, v6.8b, v2.8b
7070 prfm pldl1strm, [x1]
71− ext v5.8B, v4.8B, v5.8B, #1
72− umlal v16.8H, v7.8B, v3.8B
73− umull v17.8H, v6.8B, v0.8B
71+ ext v5.8b, v4.8b, v5.8b, #1
72+ umlal v16.8h, v7.8b, v3.8b
73+ umull v17.8h, v6.8b, v0.8b
7474 subs w3, w3, #2
75− umlal v17.8H, v7.8B, v1.8B
76− umlal v17.8H, v4.8B, v2.8B
77− umlal v17.8H, v5.8B, v3.8B
75+ umlal v17.8h, v7.8b, v1.8b
76+ umlal v17.8h, v4.8b, v2.8b
77+ umlal v17.8h, v5.8b, v3.8b
7878 prfm pldl1strm, [x1, x2]
7979 .ifc \codec,h264
80− rshrn v16.8B, v16.8H, #6
81− rshrn v17.8B, v17.8H, #6
80+ rshrn v16.8b, v16.8h, #6
81+ rshrn v17.8b, v17.8h, #6
8282 .else
83− add v16.8H, v16.8H, v22.8H
84− add v17.8H, v17.8H, v22.8H
85− shrn v16.8B, v16.8H, #6
86− shrn v17.8B, v17.8H, #6
83+ add v16.8h, v16.8h, v22.8h
84+ add v17.8h, v17.8h, v22.8h
85+ shrn v16.8b, v16.8h, #6
86+ shrn v17.8b, v17.8h, #6
8787 .endif
8888 .ifc \type,avg
89− ld1 {v20.8B}, [x8], x2
90− ld1 {v21.8B}, [x8], x2
91− urhadd v16.8B, v16.8B, v20.8B
92− urhadd v17.8B, v17.8B, v21.8B
89+ ld1 {v20.8b}, [x8], x2
90+ ld1 {v21.8b}, [x8], x2
91+ urhadd v16.8b, v16.8b, v20.8b
92+ urhadd v17.8b, v17.8b, v21.8b
9393 .endif
94− st1 {v16.8B}, [x0], x2
95− st1 {v17.8B}, [x0], x2
94+ st1 {v16.8b}, [x0], x2
95+ st1 {v17.8b}, [x0], x2
9696 b.gt 1b
9797 ret
9898
9999 2: adds w12, w12, w6
100− dup v0.8B, w4
100+ dup v0.8b, w4
101101 b.eq 5f
102102 tst w6, w6
103− dup v1.8B, w12
103+ dup v1.8b, w12
104104 b.eq 4f
105105
106− ld1 {v4.8B}, [x1], x2
107−3: ld1 {v6.8B}, [x1], x2
108− umull v16.8H, v4.8B, v0.8B
109− umlal v16.8H, v6.8B, v1.8B
110− ld1 {v4.8B}, [x1], x2
111− umull v17.8H, v6.8B, v0.8B
112− umlal v17.8H, v4.8B, v1.8B
106+ ld1 {v4.8b}, [x1], x2
107+3: ld1 {v6.8b}, [x1], x2
108+ umull v16.8h, v4.8b, v0.8b
109+ umlal v16.8h, v6.8b, v1.8b
110+ ld1 {v4.8b}, [x1], x2
111+ umull v17.8h, v6.8b, v0.8b
112+ umlal v17.8h, v4.8b, v1.8b
113113 prfm pldl1strm, [x1]
114114 .ifc \codec,h264
115− rshrn v16.8B, v16.8H, #6
116− rshrn v17.8B, v17.8H, #6
115+ rshrn v16.8b, v16.8h, #6
116+ rshrn v17.8b, v17.8h, #6
117117 .else
118− add v16.8H, v16.8H, v22.8H
119− add v17.8H, v17.8H, v22.8H
120− shrn v16.8B, v16.8H, #6
121− shrn v17.8B, v17.8H, #6
118+ add v16.8h, v16.8h, v22.8h
119+ add v17.8h, v17.8h, v22.8h
120+ shrn v16.8b, v16.8h, #6
121+ shrn v17.8b, v17.8h, #6
122122 .endif
123123 prfm pldl1strm, [x1, x2]
124124 .ifc \type,avg
125− ld1 {v20.8B}, [x8], x2
126− ld1 {v21.8B}, [x8], x2
127− urhadd v16.8B, v16.8B, v20.8B
128− urhadd v17.8B, v17.8B, v21.8B
125+ ld1 {v20.8b}, [x8], x2
126+ ld1 {v21.8b}, [x8], x2
127+ urhadd v16.8b, v16.8b, v20.8b
128+ urhadd v17.8b, v17.8b, v21.8b
129129 .endif
130130 subs w3, w3, #2
131− st1 {v16.8B}, [x0], x2
132− st1 {v17.8B}, [x0], x2
131+ st1 {v16.8b}, [x0], x2
132+ st1 {v17.8b}, [x0], x2
133133 b.gt 3b
134134 ret
135135
136−4: ld1 {v4.8B, v5.8B}, [x1], x2
137− ld1 {v6.8B, v7.8B}, [x1], x2
138− ext v5.8B, v4.8B, v5.8B, #1
139− ext v7.8B, v6.8B, v7.8B, #1
136+4: ld1 {v4.8b, v5.8b}, [x1], x2
137+ ld1 {v6.8b, v7.8b}, [x1], x2
138+ ext v5.8b, v4.8b, v5.8b, #1
139+ ext v7.8b, v6.8b, v7.8b, #1
140140 prfm pldl1strm, [x1]
141141 subs w3, w3, #2
142− umull v16.8H, v4.8B, v0.8B
143− umlal v16.8H, v5.8B, v1.8B
144− umull v17.8H, v6.8B, v0.8B
145− umlal v17.8H, v7.8B, v1.8B
142+ umull v16.8h, v4.8b, v0.8b
143+ umlal v16.8h, v5.8b, v1.8b
144+ umull v17.8h, v6.8b, v0.8b
145+ umlal v17.8h, v7.8b, v1.8b
146146 prfm pldl1strm, [x1, x2]
147147 .ifc \codec,h264
148− rshrn v16.8B, v16.8H, #6
149− rshrn v17.8B, v17.8H, #6
148+ rshrn v16.8b, v16.8h, #6
149+ rshrn v17.8b, v17.8h, #6
150150 .else
151− add v16.8H, v16.8H, v22.8H
152− add v17.8H, v17.8H, v22.8H
153− shrn v16.8B, v16.8H, #6
154− shrn v17.8B, v17.8H, #6
151+ add v16.8h, v16.8h, v22.8h
152+ add v17.8h, v17.8h, v22.8h
153+ shrn v16.8b, v16.8h, #6
154+ shrn v17.8b, v17.8h, #6
155155 .endif
156156 .ifc \type,avg
157− ld1 {v20.8B}, [x8], x2
158− ld1 {v21.8B}, [x8], x2
159− urhadd v16.8B, v16.8B, v20.8B
160− urhadd v17.8B, v17.8B, v21.8B
157+ ld1 {v20.8b}, [x8], x2
158+ ld1 {v21.8b}, [x8], x2
159+ urhadd v16.8b, v16.8b, v20.8b
160+ urhadd v17.8b, v17.8b, v21.8b
161161 .endif
162− st1 {v16.8B}, [x0], x2
163− st1 {v17.8B}, [x0], x2
162+ st1 {v16.8b}, [x0], x2
163+ st1 {v17.8b}, [x0], x2
164164 b.gt 4b
165165 ret
166166
167−5: ld1 {v4.8B}, [x1], x2
168− ld1 {v5.8B}, [x1], x2
167+5: ld1 {v4.8b}, [x1], x2
168+ ld1 {v5.8b}, [x1], x2
169169 prfm pldl1strm, [x1]
170170 subs w3, w3, #2
171− umull v16.8H, v4.8B, v0.8B
172− umull v17.8H, v5.8B, v0.8B
171+ umull v16.8h, v4.8b, v0.8b
172+ umull v17.8h, v5.8b, v0.8b
173173 prfm pldl1strm, [x1, x2]
174174 .ifc \codec,h264
175− rshrn v16.8B, v16.8H, #6
176− rshrn v17.8B, v17.8H, #6
175+ rshrn v16.8b, v16.8h, #6
176+ rshrn v17.8b, v17.8h, #6
177177 .else
178− add v16.8H, v16.8H, v22.8H
179− add v17.8H, v17.8H, v22.8H
180− shrn v16.8B, v16.8H, #6
181− shrn v17.8B, v17.8H, #6
178+ add v16.8h, v16.8h, v22.8h
179+ add v17.8h, v17.8h, v22.8h
180+ shrn v16.8b, v16.8h, #6
181+ shrn v17.8b, v17.8h, #6
182182 .endif
183183 .ifc \type,avg
184− ld1 {v20.8B}, [x8], x2
185− ld1 {v21.8B}, [x8], x2
186− urhadd v16.8B, v16.8B, v20.8B
187− urhadd v17.8B, v17.8B, v21.8B
184+ ld1 {v20.8b}, [x8], x2
185+ ld1 {v21.8b}, [x8], x2
186+ urhadd v16.8b, v16.8b, v20.8b
187+ urhadd v17.8b, v17.8b, v21.8b
188188 .endif
189− st1 {v16.8B}, [x0], x2
190− st1 {v17.8B}, [x0], x2
189+ st1 {v16.8b}, [x0], x2
190+ st1 {v17.8b}, [x0], x2
191191 b.gt 5b
192192 ret
193193 endfunc
@@ -209,10 +209,10 @@ function ff_\type\()_\codec\()_chroma_mc4_neon, export=1 · 15 unmodified lines
209209 lsl w10, w10, #1
210210 add w9, w9, w10
211211 add x6, x6, w9, UXTW
212− ld1r {v22.8H}, [x6]
212+ ld1r {v22.8h}, [x6]
213213 .endif
214214 .ifc \codec,vc1
215− movi v22.8H, #28
215+ movi v22.8h, #28
216216 .endif
217217 mul w7, w4, w5
218218 lsl w14, w5, #3
@@ -225,133 +225,133 @@ function ff_\type\()_\codec\()_chroma_mc4_neon, export=1 · 6 unmodified lines
225225 add w4, w4, #64
226226 b.eq 2f
227227
228− dup v24.8B, w4
229− dup v25.8B, w12
230− ld1 {v4.8B}, [x1], x2
231− dup v26.8B, w6
232− dup v27.8B, w7
233− ext v5.8B, v4.8B, v5.8B, #1
234− trn1 v0.2S, v24.2S, v25.2S
235− trn1 v2.2S, v26.2S, v27.2S
236− trn1 v4.2S, v4.2S, v5.2S
237−1: ld1 {v6.8B}, [x1], x2
238− ext v7.8B, v6.8B, v7.8B, #1
239− trn1 v6.2S, v6.2S, v7.2S
240− umull v18.8H, v4.8B, v0.8B
241− umlal v18.8H, v6.8B, v2.8B
242− ld1 {v4.8B}, [x1], x2
243− ext v5.8B, v4.8B, v5.8B, #1
244− trn1 v4.2S, v4.2S, v5.2S
228+ dup v24.8b, w4
229+ dup v25.8b, w12
230+ ld1 {v4.8b}, [x1], x2
231+ dup v26.8b, w6
232+ dup v27.8b, w7
233+ ext v5.8b, v4.8b, v5.8b, #1
234+ trn1 v0.2s, v24.2s, v25.2s
235+ trn1 v2.2s, v26.2s, v27.2s
236+ trn1 v4.2s, v4.2s, v5.2s
237+1: ld1 {v6.8b}, [x1], x2
238+ ext v7.8b, v6.8b, v7.8b, #1
239+ trn1 v6.2s, v6.2s, v7.2s
240+ umull v18.8h, v4.8b, v0.8b
241+ umlal v18.8h, v6.8b, v2.8b
242+ ld1 {v4.8b}, [x1], x2
243+ ext v5.8b, v4.8b, v5.8b, #1
244+ trn1 v4.2s, v4.2s, v5.2s
245245 prfm pldl1strm, [x1]
246− umull v19.8H, v6.8B, v0.8B
247− umlal v19.8H, v4.8B, v2.8B
248− trn1 v30.2D, v18.2D, v19.2D
249− trn2 v31.2D, v18.2D, v19.2D
250− add v18.8H, v30.8H, v31.8H
246+ umull v19.8h, v6.8b, v0.8b
247+ umlal v19.8h, v4.8b, v2.8b
248+ trn1 v30.2d, v18.2d, v19.2d
249+ trn2 v31.2d, v18.2d, v19.2d
250+ add v18.8h, v30.8h, v31.8h
251251 .ifc \codec,h264
252− rshrn v16.8B, v18.8H, #6
252+ rshrn v16.8b, v18.8h, #6
253253 .else
254− add v18.8H, v18.8H, v22.8H
255− shrn v16.8B, v18.8H, #6
254+ add v18.8h, v18.8h, v22.8h
255+ shrn v16.8b, v18.8h, #6
256256 .endif
257257 subs w3, w3, #2
258258 prfm pldl1strm, [x1, x2]
259259 .ifc \type,avg
260− ld1 {v20.S}[0], [x8], x2
261− ld1 {v20.S}[1], [x8], x2
262− urhadd v16.8B, v16.8B, v20.8B
260+ ld1 {v20.s}[0], [x8], x2
261+ ld1 {v20.s}[1], [x8], x2
262+ urhadd v16.8b, v16.8b, v20.8b
263263 .endif
264− st1 {v16.S}[0], [x0], x2
265− st1 {v16.S}[1], [x0], x2
264+ st1 {v16.s}[0], [x0], x2
265+ st1 {v16.s}[1], [x0], x2
266266 b.gt 1b
267267 ret
268268
269269 2: adds w12, w12, w6
270− dup v30.8B, w4
270+ dup v30.8b, w4
271271 b.eq 5f
272272 tst w6, w6
273− dup v31.8B, w12
274− trn1 v0.2S, v30.2S, v31.2S
275− trn2 v1.2S, v30.2S, v31.2S
273+ dup v31.8b, w12
274+ trn1 v0.2s, v30.2s, v31.2s
275+ trn2 v1.2s, v30.2s, v31.2s
276276 b.eq 4f
277277
278− ext v1.8B, v0.8B, v1.8B, #4
279− ld1 {v4.S}[0], [x1], x2
280−3: ld1 {v4.S}[1], [x1], x2
281− umull v18.8H, v4.8B, v0.8B
282− ld1 {v4.S}[0], [x1], x2
283− umull v19.8H, v4.8B, v1.8B
284− trn1 v30.2D, v18.2D, v19.2D
285− trn2 v31.2D, v18.2D, v19.2D
286− add v18.8H, v30.8H, v31.8H
278+ ext v1.8b, v0.8b, v1.8b, #4
279+ ld1 {v4.s}[0], [x1], x2
280+3: ld1 {v4.s}[1], [x1], x2
281+ umull v18.8h, v4.8b, v0.8b
282+ ld1 {v4.s}[0], [x1], x2
283+ umull v19.8h, v4.8b, v1.8b
284+ trn1 v30.2d, v18.2d, v19.2d
285+ trn2 v31.2d, v18.2d, v19.2d
286+ add v18.8h, v30.8h, v31.8h
287287 prfm pldl1strm, [x1]
288288 .ifc \codec,h264
289− rshrn v16.8B, v18.8H, #6
289+ rshrn v16.8b, v18.8h, #6
290290 .else
291− add v18.8H, v18.8H, v22.8H
292− shrn v16.8B, v18.8H, #6
291+ add v18.8h, v18.8h, v22.8h
292+ shrn v16.8b, v18.8h, #6
293293 .endif
294294 .ifc \type,avg
295− ld1 {v20.S}[0], [x8], x2
296− ld1 {v20.S}[1], [x8], x2
297− urhadd v16.8B, v16.8B, v20.8B
295+ ld1 {v20.s}[0], [x8], x2
296+ ld1 {v20.s}[1], [x8], x2
297+ urhadd v16.8b, v16.8b, v20.8b
298298 .endif
299299 subs w3, w3, #2
300300 prfm pldl1strm, [x1, x2]
301− st1 {v16.S}[0], [x0], x2
302− st1 {v16.S}[1], [x0], x2
301+ st1 {v16.s}[0], [x0], x2
302+ st1 {v16.s}[1], [x0], x2
303303 b.gt 3b
304304 ret
305305
306−4: ld1 {v4.8B}, [x1], x2
307− ld1 {v6.8B}, [x1], x2
308− ext v5.8B, v4.8B, v5.8B, #1
309− ext v7.8B, v6.8B, v7.8B, #1
310− trn1 v4.2S, v4.2S, v5.2S
311− trn1 v6.2S, v6.2S, v7.2S
312− umull v18.8H, v4.8B, v0.8B
313− umull v19.8H, v6.8B, v0.8B
306+4: ld1 {v4.8b}, [x1], x2
307+ ld1 {v6.8b}, [x1], x2
308+ ext v5.8b, v4.8b, v5.8b, #1
309+ ext v7.8b, v6.8b, v7.8b, #1
310+ trn1 v4.2s, v4.2s, v5.2s
311+ trn1 v6.2s, v6.2s, v7.2s
312+ umull v18.8h, v4.8b, v0.8b
313+ umull v19.8h, v6.8b, v0.8b
314314 subs w3, w3, #2
315− trn1 v30.2D, v18.2D, v19.2D
316− trn2 v31.2D, v18.2D, v19.2D
317− add v18.8H, v30.8H, v31.8H
315+ trn1 v30.2d, v18.2d, v19.2d
316+ trn2 v31.2d, v18.2d, v19.2d
317+ add v18.8h, v30.8h, v31.8h
318318 prfm pldl1strm, [x1]
319319 .ifc \codec,h264
320− rshrn v16.8B, v18.8H, #6
320+ rshrn v16.8b, v18.8h, #6
321321 .else
322− add v18.8H, v18.8H, v22.8H
323− shrn v16.8B, v18.8H, #6
322+ add v18.8h, v18.8h, v22.8h
323+ shrn v16.8b, v18.8h, #6
324324 .endif
325325 .ifc \type,avg
326− ld1 {v20.S}[0], [x8], x2
327− ld1 {v20.S}[1], [x8], x2
328− urhadd v16.8B, v16.8B, v20.8B
326+ ld1 {v20.s}[0], [x8], x2
327+ ld1 {v20.s}[1], [x8], x2
328+ urhadd v16.8b, v16.8b, v20.8b
329329 .endif
330330 prfm pldl1strm, [x1]
331− st1 {v16.S}[0], [x0], x2
332− st1 {v16.S}[1], [x0], x2
331+ st1 {v16.s}[0], [x0], x2
332+ st1 {v16.s}[1], [x0], x2
333333 b.gt 4b
334334 ret
335335
336−5: ld1 {v4.S}[0], [x1], x2
337− ld1 {v4.S}[1], [x1], x2
338− umull v18.8H, v4.8B, v30.8B
336+5: ld1 {v4.s}[0], [x1], x2
337+ ld1 {v4.s}[1], [x1], x2
338+ umull v18.8h, v4.8b, v30.8b
339339 subs w3, w3, #2
340340 prfm pldl1strm, [x1]
341341 .ifc \codec,h264
342− rshrn v16.8B, v18.8H, #6
342+ rshrn v16.8b, v18.8h, #6
343343 .else
344− add v18.8H, v18.8H, v22.8H
345− shrn v16.8B, v18.8H, #6
344+ add v18.8h, v18.8h, v22.8h
345+ shrn v16.8b, v18.8h, #6
346346 .endif
347347 .ifc \type,avg
348− ld1 {v20.S}[0], [x8], x2
349− ld1 {v20.S}[1], [x8], x2
350− urhadd v16.8B, v16.8B, v20.8B
348+ ld1 {v20.s}[0], [x8], x2
349+ ld1 {v20.s}[1], [x8], x2
350+ urhadd v16.8b, v16.8b, v20.8b
351351 .endif
352352 prfm pldl1strm, [x1]
353− st1 {v16.S}[0], [x0], x2
354− st1 {v16.S}[1], [x0], x2
353+ st1 {v16.s}[0], [x0], x2
354+ st1 {v16.s}[1], [x0], x2
355355 b.gt 5b
356356 ret
357357 endfunc
@@ -372,51 +372,51 @@ function ff_\type\()_h264_chroma_mc2_neon, export=1 · 14 unmodified lines
372372 sub w4, w7, w13
373373 sub w4, w4, w14
374374 add w4, w4, #64
375− dup v0.8B, w4
376− dup v2.8B, w12
377− dup v1.8B, w6
378− dup v3.8B, w7
379− trn1 v0.4H, v0.4H, v2.4H
380− trn1 v1.4H, v1.4H, v3.4H
375+ dup v0.8b, w4
376+ dup v2.8b, w12
377+ dup v1.8b, w6
378+ dup v3.8b, w7
379+ trn1 v0.4h, v0.4h, v2.4h
380+ trn1 v1.4h, v1.4h, v3.4h
381381 1:
382− ld1 {v4.S}[0], [x1], x2
383− ld1 {v4.S}[1], [x1], x2
384− rev64 v5.2S, v4.2S
385− ld1 {v5.S}[1], [x1]
386− ext v6.8B, v4.8B, v5.8B, #1
387− ext v7.8B, v5.8B, v4.8B, #1
388− trn1 v4.4H, v4.4H, v6.4H
389− trn1 v5.4H, v5.4H, v7.4H
390− umull v16.8H, v4.8B, v0.8B
391− umlal v16.8H, v5.8B, v1.8B
382+ ld1 {v4.s}[0], [x1], x2
383+ ld1 {v4.s}[1], [x1], x2
384+ rev64 v5.2s, v4.2s
385+ ld1 {v5.s}[1], [x1]
386+ ext v6.8b, v4.8b, v5.8b, #1
387+ ext v7.8b, v5.8b, v4.8b, #1
388+ trn1 v4.4h, v4.4h, v6.4h
389+ trn1 v5.4h, v5.4h, v7.4h
390+ umull v16.8h, v4.8b, v0.8b
391+ umlal v16.8h, v5.8b, v1.8b
392392 .ifc \type,avg
393− ld1 {v18.H}[0], [x0], x2
394− ld1 {v18.H}[2], [x0]
393+ ld1 {v18.h}[0], [x0], x2
394+ ld1 {v18.h}[2], [x0]
395395 sub x0, x0, x2
396396 .endif
397− rev64 v17.4S, v16.4S
398− add v16.8H, v16.8H, v17.8H
399− rshrn v16.8B, v16.8H, #6
397+ rev64 v17.4s, v16.4s
398+ add v16.8h, v16.8h, v17.8h
399+ rshrn v16.8b, v16.8h, #6
400400 .ifc \type,avg
401− urhadd v16.8B, v16.8B, v18.8B
401+ urhadd v16.8b, v16.8b, v18.8b
402402 .endif
403− st1 {v16.H}[0], [x0], x2
404− st1 {v16.H}[2], [x0], x2
403+ st1 {v16.h}[0], [x0], x2
404+ st1 {v16.h}[2], [x0], x2
405405 subs w3, w3, #2
406406 b.gt 1b
407407 ret
408408
409409 2:
410− ld1 {v16.H}[0], [x1], x2
411− ld1 {v16.H}[1], [x1], x2
410+ ld1 {v16.h}[0], [x1], x2
411+ ld1 {v16.h}[1], [x1], x2
412412 .ifc \type,avg
413− ld1 {v18.H}[0], [x0], x2
414− ld1 {v18.H}[1], [x0]
413+ ld1 {v18.h}[0], [x0], x2
414+ ld1 {v18.h}[1], [x0]
415415 sub x0, x0, x2
416− urhadd v16.8B, v16.8B, v18.8B
416+ urhadd v16.8b, v16.8b, v18.8b
417417 .endif
418− st1 {v16.H}[0], [x0], x2
419− st1 {v16.H}[1], [x0], x2
418+ st1 {v16.h}[0], [x0], x2
419+ st1 {v16.h}[1], [x0], x2
420420 subs w3, w3, #2
421421 b.gt 2b
422422 ret
libavcodec/aarch64/h264dsp_neon.S+297−297
@@ -27,7 +27,7 @@ · 26 unmodified lines
2727 cmp w2, #0
2828 ldr w6, [x4]
2929 ccmp w3, #0, #0, ne
30− mov v24.S[0], w6
30+ mov v24.s[0], w6
3131 and w8, w6, w6, lsl #16
3232 b.eq 1f
3333 ands w8, w8, w8, lsl #8
@@ -38,95 +38,95 @@ · 4 unmodified lines
3838 .endm
3939
4040 .macro h264_loop_filter_luma
41− dup v22.16B, w2 // alpha
42− uxtl v24.8H, v24.8B
43− uabd v21.16B, v16.16B, v0.16B // abs(p0 - q0)
44− uxtl v24.4S, v24.4H
45− uabd v28.16B, v18.16B, v16.16B // abs(p1 - p0)
46− sli v24.8H, v24.8H, #8
47− uabd v30.16B, v2.16B, v0.16B // abs(q1 - q0)
48− sli v24.4S, v24.4S, #16
49− cmhi v21.16B, v22.16B, v21.16B // < alpha
50− dup v22.16B, w3 // beta
51− cmlt v23.16B, v24.16B, #0
52− cmhi v28.16B, v22.16B, v28.16B // < beta
53− cmhi v30.16B, v22.16B, v30.16B // < beta
54− bic v21.16B, v21.16B, v23.16B
55− uabd v17.16B, v20.16B, v16.16B // abs(p2 - p0)
56− and v21.16B, v21.16B, v28.16B
57− uabd v19.16B, v4.16B, v0.16B // abs(q2 - q0)
58− and v21.16B, v21.16B, v30.16B // < beta
41+ dup v22.16b, w2 // alpha
42+ uxtl v24.8h, v24.8b
43+ uabd v21.16b, v16.16b, v0.16b // abs(p0 - q0)
44+ uxtl v24.4s, v24.4h
45+ uabd v28.16b, v18.16b, v16.16b // abs(p1 - p0)
46+ sli v24.8h, v24.8h, #8
47+ uabd v30.16b, v2.16b, v0.16b // abs(q1 - q0)
48+ sli v24.4s, v24.4s, #16
49+ cmhi v21.16b, v22.16b, v21.16b // < alpha
50+ dup v22.16b, w3 // beta
51+ cmlt v23.16b, v24.16b, #0
52+ cmhi v28.16b, v22.16b, v28.16b // < beta
53+ cmhi v30.16b, v22.16b, v30.16b // < beta
54+ bic v21.16b, v21.16b, v23.16b
55+ uabd v17.16b, v20.16b, v16.16b // abs(p2 - p0)
56+ and v21.16b, v21.16b, v28.16b
57+ uabd v19.16b, v4.16b, v0.16b // abs(q2 - q0)
58+ and v21.16b, v21.16b, v30.16b // < beta
5959 shrn v30.8b, v21.8h, #4
6060 mov x7, v30.d[0]
61− cmhi v17.16B, v22.16B, v17.16B // < beta
62− cmhi v19.16B, v22.16B, v19.16B // < beta
61+ cmhi v17.16b, v22.16b, v17.16b // < beta
62+ cmhi v19.16b, v22.16b, v19.16b // < beta
6363 cbz x7, 9f
64− and v17.16B, v17.16B, v21.16B
65− and v19.16B, v19.16B, v21.16B
66− and v24.16B, v24.16B, v21.16B
67− urhadd v28.16B, v16.16B, v0.16B
68− sub v21.16B, v24.16B, v17.16B
69− uqadd v23.16B, v18.16B, v24.16B
70− uhadd v20.16B, v20.16B, v28.16B
71− sub v21.16B, v21.16B, v19.16B
72− uhadd v28.16B, v4.16B, v28.16B
73− umin v23.16B, v23.16B, v20.16B
74− uqsub v22.16B, v18.16B, v24.16B
75− uqadd v4.16B, v2.16B, v24.16B
76− umax v23.16B, v23.16B, v22.16B
77− uqsub v22.16B, v2.16B, v24.16B
78− umin v28.16B, v4.16B, v28.16B
79− uxtl v4.8H, v0.8B
80− umax v28.16B, v28.16B, v22.16B
81− uxtl2 v20.8H, v0.16B
82− usubw v4.8H, v4.8H, v16.8B
83− usubw2 v20.8H, v20.8H, v16.16B
84− shl v4.8H, v4.8H, #2
85− shl v20.8H, v20.8H, #2
86− uaddw v4.8H, v4.8H, v18.8B
87− uaddw2 v20.8H, v20.8H, v18.16B
88− usubw v4.8H, v4.8H, v2.8B
89− usubw2 v20.8H, v20.8H, v2.16B
90− rshrn v4.8B, v4.8H, #3
91− rshrn2 v4.16B, v20.8H, #3
92− bsl v17.16B, v23.16B, v18.16B
93− bsl v19.16B, v28.16B, v2.16B
94− neg v23.16B, v21.16B
95− uxtl v28.8H, v16.8B
96− smin v4.16B, v4.16B, v21.16B
97− uxtl2 v21.8H, v16.16B
98− smax v4.16B, v4.16B, v23.16B
99− uxtl v22.8H, v0.8B
100− uxtl2 v24.8H, v0.16B
101− saddw v28.8H, v28.8H, v4.8B
102− saddw2 v21.8H, v21.8H, v4.16B
103− ssubw v22.8H, v22.8H, v4.8B
104− ssubw2 v24.8H, v24.8H, v4.16B
105− sqxtun v16.8B, v28.8H
106− sqxtun2 v16.16B, v21.8H
107− sqxtun v0.8B, v22.8H
108− sqxtun2 v0.16B, v24.8H
64+ and v17.16b, v17.16b, v21.16b
65+ and v19.16b, v19.16b, v21.16b
66+ and v24.16b, v24.16b, v21.16b
67+ urhadd v28.16b, v16.16b, v0.16b
68+ sub v21.16b, v24.16b, v17.16b
69+ uqadd v23.16b, v18.16b, v24.16b
70+ uhadd v20.16b, v20.16b, v28.16b
71+ sub v21.16b, v21.16b, v19.16b
72+ uhadd v28.16b, v4.16b, v28.16b
73+ umin v23.16b, v23.16b, v20.16b
74+ uqsub v22.16b, v18.16b, v24.16b
75+ uqadd v4.16b, v2.16b, v24.16b
76+ umax v23.16b, v23.16b, v22.16b
77+ uqsub v22.16b, v2.16b, v24.16b
78+ umin v28.16b, v4.16b, v28.16b
79+ uxtl v4.8h, v0.8b
80+ umax v28.16b, v28.16b, v22.16b
81+ uxtl2 v20.8h, v0.16b
82+ usubw v4.8h, v4.8h, v16.8b
83+ usubw2 v20.8h, v20.8h, v16.16b
84+ shl v4.8h, v4.8h, #2
85+ shl v20.8h, v20.8h, #2
86+ uaddw v4.8h, v4.8h, v18.8b
87+ uaddw2 v20.8h, v20.8h, v18.16b
88+ usubw v4.8h, v4.8h, v2.8b
89+ usubw2 v20.8h, v20.8h, v2.16b
90+ rshrn v4.8b, v4.8h, #3
91+ rshrn2 v4.16b, v20.8h, #3
92+ bsl v17.16b, v23.16b, v18.16b
93+ bsl v19.16b, v28.16b, v2.16b
94+ neg v23.16b, v21.16b
95+ uxtl v28.8h, v16.8b
96+ smin v4.16b, v4.16b, v21.16b
97+ uxtl2 v21.8h, v16.16b
98+ smax v4.16b, v4.16b, v23.16b
99+ uxtl v22.8h, v0.8b
100+ uxtl2 v24.8h, v0.16b
101+ saddw v28.8h, v28.8h, v4.8b
102+ saddw2 v21.8h, v21.8h, v4.16b
103+ ssubw v22.8h, v22.8h, v4.8b
104+ ssubw2 v24.8h, v24.8h, v4.16b
105+ sqxtun v16.8b, v28.8h
106+ sqxtun2 v16.16b, v21.8h
107+ sqxtun v0.8b, v22.8h
108+ sqxtun2 v0.16b, v24.8h
109109 .endm
110110
111111 function ff_h264_v_loop_filter_luma_neon, export=1
112112 h264_loop_filter_start
113113
114− ld1 {v0.16B}, [x0], x1
115− ld1 {v2.16B}, [x0], x1
116− ld1 {v4.16B}, [x0], x1
114+ ld1 {v0.16b}, [x0], x1
115+ ld1 {v2.16b}, [x0], x1
116+ ld1 {v4.16b}, [x0], x1
117117 sub x0, x0, x1, lsl #2
118118 sub x0, x0, x1, lsl #1
119− ld1 {v20.16B}, [x0], x1
120− ld1 {v18.16B}, [x0], x1
121− ld1 {v16.16B}, [x0], x1
119+ ld1 {v20.16b}, [x0], x1
120+ ld1 {v18.16b}, [x0], x1
121+ ld1 {v16.16b}, [x0], x1
122122
123123 h264_loop_filter_luma
124124
125125 sub x0, x0, x1, lsl #1
126− st1 {v17.16B}, [x0], x1
127− st1 {v16.16B}, [x0], x1
128− st1 {v0.16B}, [x0], x1
129− st1 {v19.16B}, [x0]
126+ st1 {v17.16b}, [x0], x1
127+ st1 {v16.16b}, [x0], x1
128+ st1 {v0.16b}, [x0], x1
129+ st1 {v19.16b}, [x0]
130130 9:
131131 ret
132132 endfunc
@@ -135,22 +135,22 @@ function ff_h264_h_loop_filter_luma_neon, export=1 · 2 unmodified lines
135135 h264_loop_filter_start
136136
137137 sub x0, x0, #4
138− ld1 {v6.8B}, [x0], x1
139− ld1 {v20.8B}, [x0], x1
140− ld1 {v18.8B}, [x0], x1
141− ld1 {v16.8B}, [x0], x1
142− ld1 {v0.8B}, [x0], x1
143− ld1 {v2.8B}, [x0], x1
144− ld1 {v4.8B}, [x0], x1
145− ld1 {v26.8B}, [x0], x1
146− ld1 {v6.D}[1], [x0], x1
147− ld1 {v20.D}[1], [x0], x1
148− ld1 {v18.D}[1], [x0], x1
149− ld1 {v16.D}[1], [x0], x1
150− ld1 {v0.D}[1], [x0], x1
151− ld1 {v2.D}[1], [x0], x1
152− ld1 {v4.D}[1], [x0], x1
153− ld1 {v26.D}[1], [x0], x1
138+ ld1 {v6.8b}, [x0], x1
139+ ld1 {v20.8b}, [x0], x1
140+ ld1 {v18.8b}, [x0], x1
141+ ld1 {v16.8b}, [x0], x1
142+ ld1 {v0.8b}, [x0], x1
143+ ld1 {v2.8b}, [x0], x1
144+ ld1 {v4.8b}, [x0], x1
145+ ld1 {v26.8b}, [x0], x1
146+ ld1 {v6.d}[1], [x0], x1
147+ ld1 {v20.d}[1], [x0], x1
148+ ld1 {v18.d}[1], [x0], x1
149+ ld1 {v16.d}[1], [x0], x1
150+ ld1 {v0.d}[1], [x0], x1
151+ ld1 {v2.d}[1], [x0], x1
152+ ld1 {v4.d}[1], [x0], x1
153+ ld1 {v26.d}[1], [x0], x1
154154
155155 transpose_8x16B v6, v20, v18, v16, v0, v2, v4, v26, v21, v23
156156
@@ -160,22 +160,22 @@ function ff_h264_h_loop_filter_luma_neon, export=1 · 3 unmodified lines
160160
161161 sub x0, x0, x1, lsl #4
162162 add x0, x0, #2
163− st1 {v17.S}[0], [x0], x1
164− st1 {v16.S}[0], [x0], x1
165− st1 {v0.S}[0], [x0], x1
166− st1 {v19.S}[0], [x0], x1
167− st1 {v17.S}[1], [x0], x1
168− st1 {v16.S}[1], [x0], x1
169− st1 {v0.S}[1], [x0], x1
170− st1 {v19.S}[1], [x0], x1
171− st1 {v17.S}[2], [x0], x1
172− st1 {v16.S}[2], [x0], x1
173− st1 {v0.S}[2], [x0], x1
174− st1 {v19.S}[2], [x0], x1
175− st1 {v17.S}[3], [x0], x1
176− st1 {v16.S}[3], [x0], x1
177− st1 {v0.S}[3], [x0], x1
178− st1 {v19.S}[3], [x0], x1
163+ st1 {v17.s}[0], [x0], x1
164+ st1 {v16.s}[0], [x0], x1
165+ st1 {v0.s}[0], [x0], x1
166+ st1 {v19.s}[0], [x0], x1
167+ st1 {v17.s}[1], [x0], x1
168+ st1 {v16.s}[1], [x0], x1
169+ st1 {v0.s}[1], [x0], x1
170+ st1 {v19.s}[1], [x0], x1
171+ st1 {v17.s}[2], [x0], x1
172+ st1 {v16.s}[2], [x0], x1
173+ st1 {v0.s}[2], [x0], x1
174+ st1 {v19.s}[2], [x0], x1
175+ st1 {v17.s}[3], [x0], x1
176+ st1 {v16.s}[3], [x0], x1
177+ st1 {v0.s}[3], [x0], x1
178+ st1 {v19.s}[3], [x0], x1
179179 9:
180180 ret
181181 endfunc
@@ -377,52 +377,52 @@ function ff_h264_h_loop_filter_luma_intra_neon, export=1 · 195 unmodified lines
377377 endfunc
378378
379379 .macro h264_loop_filter_chroma
380− dup v22.8B, w2 // alpha
381− dup v23.8B, w3 // beta
382− uxtl v24.8H, v24.8B
383− uabd v26.8B, v16.8B, v0.8B // abs(p0 - q0)
384− uabd v28.8B, v18.8B, v16.8B // abs(p1 - p0)
385− uabd v30.8B, v2.8B, v0.8B // abs(q1 - q0)
386− cmhi v26.8B, v22.8B, v26.8B // < alpha
387− cmhi v28.8B, v23.8B, v28.8B // < beta
388− cmhi v30.8B, v23.8B, v30.8B // < beta
389− uxtl v4.8H, v0.8B
390− and v26.8B, v26.8B, v28.8B
391− usubw v4.8H, v4.8H, v16.8B
392− and v26.8B, v26.8B, v30.8B
393− shl v4.8H, v4.8H, #2
380+ dup v22.8b, w2 // alpha
381+ dup v23.8b, w3 // beta
382+ uxtl v24.8h, v24.8b
383+ uabd v26.8b, v16.8b, v0.8b // abs(p0 - q0)
384+ uabd v28.8b, v18.8b, v16.8b // abs(p1 - p0)
385+ uabd v30.8b, v2.8b, v0.8b // abs(q1 - q0)
386+ cmhi v26.8b, v22.8b, v26.8b // < alpha
387+ cmhi v28.8b, v23.8b, v28.8b // < beta
388+ cmhi v30.8b, v23.8b, v30.8b // < beta
389+ uxtl v4.8h, v0.8b
390+ and v26.8b, v26.8b, v28.8b
391+ usubw v4.8h, v4.8h, v16.8b
392+ and v26.8b, v26.8b, v30.8b
393+ shl v4.8h, v4.8h, #2
394394 mov x8, v26.d[0]
395− sli v24.8H, v24.8H, #8
396− uaddw v4.8H, v4.8H, v18.8B
395+ sli v24.8h, v24.8h, #8
396+ uaddw v4.8h, v4.8h, v18.8b
397397 cbz x8, 9f
398− usubw v4.8H, v4.8H, v2.8B
399− rshrn v4.8B, v4.8H, #3
400− smin v4.8B, v4.8B, v24.8B
401− neg v25.8B, v24.8B
402− smax v4.8B, v4.8B, v25.8B
403− uxtl v22.8H, v0.8B
404− and v4.8B, v4.8B, v26.8B
405− uxtl v28.8H, v16.8B
406− saddw v28.8H, v28.8H, v4.8B
407− ssubw v22.8H, v22.8H, v4.8B
408− sqxtun v16.8B, v28.8H
409− sqxtun v0.8B, v22.8H
398+ usubw v4.8h, v4.8h, v2.8b
399+ rshrn v4.8b, v4.8h, #3
400+ smin v4.8b, v4.8b, v24.8b
401+ neg v25.8b, v24.8b
402+ smax v4.8b, v4.8b, v25.8b
403+ uxtl v22.8h, v0.8b
404+ and v4.8b, v4.8b, v26.8b
405+ uxtl v28.8h, v16.8b
406+ saddw v28.8h, v28.8h, v4.8b
407+ ssubw v22.8h, v22.8h, v4.8b
408+ sqxtun v16.8b, v28.8h
409+ sqxtun v0.8b, v22.8h
410410 .endm
411411
412412 function ff_h264_v_loop_filter_chroma_neon, export=1
413413 h264_loop_filter_start
414414
415415 sub x0, x0, x1, lsl #1
416− ld1 {v18.8B}, [x0], x1
417− ld1 {v16.8B}, [x0], x1
418− ld1 {v0.8B}, [x0], x1
419− ld1 {v2.8B}, [x0]
416+ ld1 {v18.8b}, [x0], x1
417+ ld1 {v16.8b}, [x0], x1
418+ ld1 {v0.8b}, [x0], x1
419+ ld1 {v2.8b}, [x0]
420420
421421 h264_loop_filter_chroma
422422
423423 sub x0, x0, x1, lsl #1
424− st1 {v16.8B}, [x0], x1
425− st1 {v0.8B}, [x0], x1
424+ st1 {v16.8b}, [x0], x1
425+ st1 {v0.8b}, [x0], x1
426426 9:
427427 ret
428428 endfunc
@@ -432,14 +432,14 @@ function ff_h264_h_loop_filter_chroma_neon, export=1 · 3 unmodified lines
432432
433433 sub x0, x0, #2
434434 h_loop_filter_chroma420:
435− ld1 {v18.S}[0], [x0], x1
436− ld1 {v16.S}[0], [x0], x1
437− ld1 {v0.S}[0], [x0], x1
438− ld1 {v2.S}[0], [x0], x1
439− ld1 {v18.S}[1], [x0], x1
440− ld1 {v16.S}[1], [x0], x1
441− ld1 {v0.S}[1], [x0], x1
442− ld1 {v2.S}[1], [x0], x1
435+ ld1 {v18.s}[0], [x0], x1
436+ ld1 {v16.s}[0], [x0], x1
437+ ld1 {v0.s}[0], [x0], x1
438+ ld1 {v2.s}[0], [x0], x1
439+ ld1 {v18.s}[1], [x0], x1
440+ ld1 {v16.s}[1], [x0], x1
441+ ld1 {v0.s}[1], [x0], x1
442+ ld1 {v2.s}[1], [x0], x1
443443
444444 transpose_4x8B v18, v16, v0, v2, v28, v29, v30, v31
445445
@@ -448,14 +448,14 @@ h_loop_filter_chroma420: · 2 unmodified lines
448448 transpose_4x8B v18, v16, v0, v2, v28, v29, v30, v31
449449
450450 sub x0, x0, x1, lsl #3
451− st1 {v18.S}[0], [x0], x1
452− st1 {v16.S}[0], [x0], x1
453− st1 {v0.S}[0], [x0], x1
454− st1 {v2.S}[0], [x0], x1
455− st1 {v18.S}[1], [x0], x1
456− st1 {v16.S}[1], [x0], x1
457− st1 {v0.S}[1], [x0], x1
458− st1 {v2.S}[1], [x0], x1
451+ st1 {v18.s}[0], [x0], x1
452+ st1 {v16.s}[0], [x0], x1
453+ st1 {v0.s}[0], [x0], x1
454+ st1 {v2.s}[0], [x0], x1
455+ st1 {v18.s}[1], [x0], x1
456+ st1 {v16.s}[1], [x0], x1
457+ st1 {v0.s}[1], [x0], x1
458+ st1 {v2.s}[1], [x0], x1
459459 9:
460460 ret
461461 endfunc
@@ -584,102 +584,102 @@ function ff_h264_h_loop_filter_chroma422_intra_neon, export=1 · 122 unmodified lines
584584 endfunc
585585
586586 .macro biweight_16 macs, macd
587− dup v0.16B, w5
588− dup v1.16B, w6
589− mov v4.16B, v16.16B
590− mov v6.16B, v16.16B
587+ dup v0.16b, w5
588+ dup v1.16b, w6
589+ mov v4.16b, v16.16b
590+ mov v6.16b, v16.16b
591591 1: subs w3, w3, #2
592− ld1 {v20.16B}, [x0], x2
593− \macd v4.8H, v0.8B, v20.8B
592+ ld1 {v20.16b}, [x0], x2
593+ \macd v4.8h, v0.8b, v20.8b
594594 \macd\()2 v6.8H, v0.16B, v20.16B
595− ld1 {v22.16B}, [x1], x2
596− \macs v4.8H, v1.8B, v22.8B
595+ ld1 {v22.16b}, [x1], x2
596+ \macs v4.8h, v1.8b, v22.8b
597597 \macs\()2 v6.8H, v1.16B, v22.16B
598− mov v24.16B, v16.16B
599− ld1 {v28.16B}, [x0], x2
600− mov v26.16B, v16.16B
601− \macd v24.8H, v0.8B, v28.8B
598+ mov v24.16b, v16.16b
599+ ld1 {v28.16b}, [x0], x2
600+ mov v26.16b, v16.16b
601+ \macd v24.8h, v0.8b, v28.8b
602602 \macd\()2 v26.8H, v0.16B, v28.16B
603− ld1 {v30.16B}, [x1], x2
604− \macs v24.8H, v1.8B, v30.8B
603+ ld1 {v30.16b}, [x1], x2
604+ \macs v24.8h, v1.8b, v30.8b
605605 \macs\()2 v26.8H, v1.16B, v30.16B
606− sshl v4.8H, v4.8H, v18.8H
607− sshl v6.8H, v6.8H, v18.8H
608− sqxtun v4.8B, v4.8H
609− sqxtun2 v4.16B, v6.8H
610− sshl v24.8H, v24.8H, v18.8H
611− sshl v26.8H, v26.8H, v18.8H
612− sqxtun v24.8B, v24.8H
613− sqxtun2 v24.16B, v26.8H
614− mov v6.16B, v16.16B
615− st1 {v4.16B}, [x7], x2
616− mov v4.16B, v16.16B
617− st1 {v24.16B}, [x7], x2
606+ sshl v4.8h, v4.8h, v18.8h
607+ sshl v6.8h, v6.8h, v18.8h
608+ sqxtun v4.8b, v4.8h
609+ sqxtun2 v4.16b, v6.8h
610+ sshl v24.8h, v24.8h, v18.8h
611+ sshl v26.8h, v26.8h, v18.8h
612+ sqxtun v24.8b, v24.8h
613+ sqxtun2 v24.16b, v26.8h
614+ mov v6.16b, v16.16b
615+ st1 {v4.16b}, [x7], x2
616+ mov v4.16b, v16.16b
617+ st1 {v24.16b}, [x7], x2
618618 b.ne 1b
619619 ret
620620 .endm
621621
622622 .macro biweight_8 macs, macd
623− dup v0.8B, w5
624− dup v1.8B, w6
625− mov v2.16B, v16.16B
626− mov v20.16B, v16.16B
623+ dup v0.8b, w5
624+ dup v1.8b, w6
625+ mov v2.16b, v16.16b
626+ mov v20.16b, v16.16b
627627 1: subs w3, w3, #2
628− ld1 {v4.8B}, [x0], x2
629− \macd v2.8H, v0.8B, v4.8B
630− ld1 {v5.8B}, [x1], x2
631− \macs v2.8H, v1.8B, v5.8B
632− ld1 {v6.8B}, [x0], x2
633− \macd v20.8H, v0.8B, v6.8B
634− ld1 {v7.8B}, [x1], x2
635− \macs v20.8H, v1.8B, v7.8B
636− sshl v2.8H, v2.8H, v18.8H
637− sqxtun v2.8B, v2.8H
638− sshl v20.8H, v20.8H, v18.8H
639− sqxtun v4.8B, v20.8H
640− mov v20.16B, v16.16B
641− st1 {v2.8B}, [x7], x2
642− mov v2.16B, v16.16B
643− st1 {v4.8B}, [x7], x2
628+ ld1 {v4.8b}, [x0], x2
629+ \macd v2.8h, v0.8b, v4.8b
630+ ld1 {v5.8b}, [x1], x2
631+ \macs v2.8h, v1.8b, v5.8b
632+ ld1 {v6.8b}, [x0], x2
633+ \macd v20.8h, v0.8b, v6.8b
634+ ld1 {v7.8b}, [x1], x2
635+ \macs v20.8h, v1.8b, v7.8b
636+ sshl v2.8h, v2.8h, v18.8h
637+ sqxtun v2.8b, v2.8h
638+ sshl v20.8h, v20.8h, v18.8h
639+ sqxtun v4.8b, v20.8h
640+ mov v20.16b, v16.16b
641+ st1 {v2.8b}, [x7], x2
642+ mov v2.16b, v16.16b
643+ st1 {v4.8b}, [x7], x2
644644 b.ne 1b
645645 ret
646646 .endm
647647
648648 .macro biweight_4 macs, macd
649− dup v0.8B, w5
650− dup v1.8B, w6
651− mov v2.16B, v16.16B
652− mov v20.16B,v16.16B
649+ dup v0.8b, w5
650+ dup v1.8b, w6
651+ mov v2.16b, v16.16b
652+ mov v20.16b,v16.16b
653653 1: subs w3, w3, #4
654− ld1 {v4.S}[0], [x0], x2
655− ld1 {v4.S}[1], [x0], x2
656− \macd v2.8H, v0.8B, v4.8B
657− ld1 {v5.S}[0], [x1], x2
658− ld1 {v5.S}[1], [x1], x2
659− \macs v2.8H, v1.8B, v5.8B
654+ ld1 {v4.s}[0], [x0], x2
655+ ld1 {v4.s}[1], [x0], x2
656+ \macd v2.8h, v0.8b, v4.8b
657+ ld1 {v5.s}[0], [x1], x2
658+ ld1 {v5.s}[1], [x1], x2
659+ \macs v2.8h, v1.8b, v5.8b
660660 b.lt 2f
661− ld1 {v6.S}[0], [x0], x2
662− ld1 {v6.S}[1], [x0], x2
663− \macd v20.8H, v0.8B, v6.8B
664− ld1 {v7.S}[0], [x1], x2
665− ld1 {v7.S}[1], [x1], x2
666− \macs v20.8H, v1.8B, v7.8B
667− sshl v2.8H, v2.8H, v18.8H
668− sqxtun v2.8B, v2.8H
669− sshl v20.8H, v20.8H, v18.8H
670− sqxtun v4.8B, v20.8H
671− mov v20.16B, v16.16B
672− st1 {v2.S}[0], [x7], x2
673− st1 {v2.S}[1], [x7], x2
674− mov v2.16B, v16.16B
675− st1 {v4.S}[0], [x7], x2
676− st1 {v4.S}[1], [x7], x2
661+ ld1 {v6.s}[0], [x0], x2
662+ ld1 {v6.s}[1], [x0], x2
663+ \macd v20.8h, v0.8b, v6.8b
664+ ld1 {v7.s}[0], [x1], x2
665+ ld1 {v7.s}[1], [x1], x2
666+ \macs v20.8h, v1.8b, v7.8b
667+ sshl v2.8h, v2.8h, v18.8h
668+ sqxtun v2.8b, v2.8h
669+ sshl v20.8h, v20.8h, v18.8h
670+ sqxtun v4.8b, v20.8h
671+ mov v20.16b, v16.16b
672+ st1 {v2.s}[0], [x7], x2
673+ st1 {v2.s}[1], [x7], x2
674+ mov v2.16b, v16.16b
675+ st1 {v4.s}[0], [x7], x2
676+ st1 {v4.s}[1], [x7], x2
677677 b.ne 1b
678678 ret
679−2: sshl v2.8H, v2.8H, v18.8H
680− sqxtun v2.8B, v2.8H
681− st1 {v2.S}[0], [x7], x2
682− st1 {v2.S}[1], [x7], x2
679+2: sshl v2.8h, v2.8h, v18.8h
680+ sqxtun v2.8b, v2.8h
681+ st1 {v2.s}[0], [x7], x2
682+ st1 {v2.s}[1], [x7], x2
683683 ret
684684 .endm
685685
@@ -689,10 +689,10 @@ function ff_biweight_h264_pixels_\w\()_neon, export=1 · 3 unmodified lines
689689 add w7, w7, #1
690690 eor w8, w8, w6, lsr #30
691691 orr w7, w7, #1
692− dup v18.8H, w4
692+ dup v18.8h, w4
693693 lsl w7, w7, w4
694− not v18.16B, v18.16B
695− dup v16.8H, w7
694+ not v18.16b, v18.16b
695+ dup v16.8h, w7
696696 mov x7, x0
697697 cbz w8, 10f
698698 subs w8, w8, #1
@@ -716,78 +716,78 @@ endfunc · 17 unmodified lines
716716 biweight_func 4
717717
718718 .macro weight_16 add
719− dup v0.16B, w4
719+ dup v0.16b, w4
720720 1: subs w2, w2, #2
721− ld1 {v20.16B}, [x0], x1
722− umull v4.8H, v0.8B, v20.8B
723− umull2 v6.8H, v0.16B, v20.16B
724− ld1 {v28.16B}, [x0], x1
725− umull v24.8H, v0.8B, v28.8B
726− umull2 v26.8H, v0.16B, v28.16B
727− \add v4.8H, v16.8H, v4.8H
728− srshl v4.8H, v4.8H, v18.8H
729− \add v6.8H, v16.8H, v6.8H
730− srshl v6.8H, v6.8H, v18.8H
731− sqxtun v4.8B, v4.8H
732− sqxtun2 v4.16B, v6.8H
733− \add v24.8H, v16.8H, v24.8H
734− srshl v24.8H, v24.8H, v18.8H
735− \add v26.8H, v16.8H, v26.8H
736− srshl v26.8H, v26.8H, v18.8H
737− sqxtun v24.8B, v24.8H
738− sqxtun2 v24.16B, v26.8H
739− st1 {v4.16B}, [x5], x1
740− st1 {v24.16B}, [x5], x1
721+ ld1 {v20.16b}, [x0], x1
722+ umull v4.8h, v0.8b, v20.8b
723+ umull2 v6.8h, v0.16b, v20.16b
724+ ld1 {v28.16b}, [x0], x1
725+ umull v24.8h, v0.8b, v28.8b
726+ umull2 v26.8h, v0.16b, v28.16b
727+ \add v4.8h, v16.8h, v4.8h
728+ srshl v4.8h, v4.8h, v18.8h
729+ \add v6.8h, v16.8h, v6.8h
730+ srshl v6.8h, v6.8h, v18.8h
731+ sqxtun v4.8b, v4.8h
732+ sqxtun2 v4.16b, v6.8h
733+ \add v24.8h, v16.8h, v24.8h
734+ srshl v24.8h, v24.8h, v18.8h
735+ \add v26.8h, v16.8h, v26.8h
736+ srshl v26.8h, v26.8h, v18.8h
737+ sqxtun v24.8b, v24.8h
738+ sqxtun2 v24.16b, v26.8h
739+ st1 {v4.16b}, [x5], x1
740+ st1 {v24.16b}, [x5], x1
741741 b.ne 1b
742742 ret
743743 .endm
744744
745745 .macro weight_8 add
746− dup v0.8B, w4
746+ dup v0.8b, w4
747747 1: subs w2, w2, #2
748− ld1 {v4.8B}, [x0], x1
749− umull v2.8H, v0.8B, v4.8B
750− ld1 {v6.8B}, [x0], x1
751− umull v20.8H, v0.8B, v6.8B
752− \add v2.8H, v16.8H, v2.8H
753− srshl v2.8H, v2.8H, v18.8H
754− sqxtun v2.8B, v2.8H
755− \add v20.8H, v16.8H, v20.8H
756− srshl v20.8H, v20.8H, v18.8H
757− sqxtun v4.8B, v20.8H
758− st1 {v2.8B}, [x5], x1
759− st1 {v4.8B}, [x5], x1
748+ ld1 {v4.8b}, [x0], x1
749+ umull v2.8h, v0.8b, v4.8b
750+ ld1 {v6.8b}, [x0], x1
751+ umull v20.8h, v0.8b, v6.8b
752+ \add v2.8h, v16.8h, v2.8h
753+ srshl v2.8h, v2.8h, v18.8h
754+ sqxtun v2.8b, v2.8h
755+ \add v20.8h, v16.8h, v20.8h
756+ srshl v20.8h, v20.8h, v18.8h
757+ sqxtun v4.8b, v20.8h
758+ st1 {v2.8b}, [x5], x1
759+ st1 {v4.8b}, [x5], x1
760760 b.ne 1b
761761 ret
762762 .endm
763763
764764 .macro weight_4 add
765− dup v0.8B, w4
765+ dup v0.8b, w4
766766 1: subs w2, w2, #4
767− ld1 {v4.S}[0], [x0], x1
768− ld1 {v4.S}[1], [x0], x1
769− umull v2.8H, v0.8B, v4.8B
767+ ld1 {v4.s}[0], [x0], x1
768+ ld1 {v4.s}[1], [x0], x1
769+ umull v2.8h, v0.8b, v4.8b
770770 b.lt 2f
771− ld1 {v6.S}[0], [x0], x1
772− ld1 {v6.S}[1], [x0], x1
773− umull v20.8H, v0.8B, v6.8B
774− \add v2.8H, v16.8H, v2.8H
775− srshl v2.8H, v2.8H, v18.8H
776− sqxtun v2.8B, v2.8H
777− \add v20.8H, v16.8H, v20.8H
778− srshl v20.8H, v20.8h, v18.8H
779− sqxtun v4.8B, v20.8H
780− st1 {v2.S}[0], [x5], x1
781− st1 {v2.S}[1], [x5], x1
782− st1 {v4.S}[0], [x5], x1
783− st1 {v4.S}[1], [x5], x1
771+ ld1 {v6.s}[0], [x0], x1
772+ ld1 {v6.s}[1], [x0], x1
773+ umull v20.8h, v0.8b, v6.8b
774+ \add v2.8h, v16.8h, v2.8h
775+ srshl v2.8h, v2.8h, v18.8h
776+ sqxtun v2.8b, v2.8h
777+ \add v20.8h, v16.8h, v20.8h
778+ srshl v20.8h, v20.8h, v18.8h
779+ sqxtun v4.8b, v20.8h
780+ st1 {v2.s}[0], [x5], x1
781+ st1 {v2.s}[1], [x5], x1
782+ st1 {v4.s}[0], [x5], x1
783+ st1 {v4.s}[1], [x5], x1
784784 b.ne 1b
785785 ret
786−2: \add v2.8H, v16.8H, v2.8H
787− srshl v2.8H, v2.8H, v18.8H
788− sqxtun v2.8B, v2.8H
789− st1 {v2.S}[0], [x5], x1
790− st1 {v2.S}[1], [x5], x1
786+2: \add v2.8h, v16.8h, v2.8h
787+ srshl v2.8h, v2.8h, v18.8h
788+ sqxtun v2.8b, v2.8h
789+ st1 {v2.s}[0], [x5], x1
790+ st1 {v2.s}[1], [x5], x1
791791 ret
792792 .endm
793793
@@ -796,18 +796,18 @@ function ff_weight_h264_pixels_\w\()_neon, export=1 · 2 unmodified lines
796796 cmp w3, #1
797797 mov w6, #1
798798 lsl w5, w5, w3
799− dup v16.8H, w5
799+ dup v16.8h, w5
800800 mov x5, x0
801801 b.le 20f
802802 sub w6, w6, w3
803− dup v18.8H, w6
803+ dup v18.8h, w6
804804 cmp w4, #0
805805 b.lt 10f
806806 weight_\w shadd
807807 10: neg w4, w4
808808 weight_\w shsub
809809 20: neg w6, w3
810− dup v18.8H, w6
810+ dup v18.8h, w6
811811 cmp w4, #0
812812 b.lt 10f
813813 weight_\w add
@@ -825,7 +825,7 @@ endfunc · 11 unmodified lines
825825 ldr w6, [x4]
826826 ccmp w3, #0, #0, ne
827827 lsl w2, w2, #2
828− mov v24.S[0], w6
828+ mov v24.s[0], w6
829829 lsl w3, w3, #2
830830 and w8, w6, w6, lsl #16
831831 b.eq 1f
libavcodec/aarch64/h264idct_neon.S+195−195
@@ -25,54 +25,54 @@ · 24 unmodified lines
2525 function ff_h264_idct_add_neon, export=1
2626 .L_ff_h264_idct_add_neon:
2727 AARCH64_VALID_CALL_TARGET
28− ld1 {v0.4H, v1.4H, v2.4H, v3.4H}, [x1]
28+ ld1 {v0.4h, v1.4h, v2.4h, v3.4h}, [x1]
2929 sxtw x2, w2
30− movi v30.8H, #0
30+ movi v30.8h, #0
3131
32− add v4.4H, v0.4H, v2.4H
33− sshr v16.4H, v1.4H, #1
34− st1 {v30.8H}, [x1], #16
35− sshr v17.4H, v3.4H, #1
36− st1 {v30.8H}, [x1], #16
37− sub v5.4H, v0.4H, v2.4H
38− sub v6.4H, v16.4H, v3.4H
39− add v7.4H, v1.4H, v17.4H
40− add v0.4H, v4.4H, v7.4H
41− add v1.4H, v5.4H, v6.4H
42− sub v2.4H, v5.4H, v6.4H
43− sub v3.4H, v4.4H, v7.4H
32+ add v4.4h, v0.4h, v2.4h
33+ sshr v16.4h, v1.4h, #1
34+ st1 {v30.8h}, [x1], #16
35+ sshr v17.4h, v3.4h, #1
36+ st1 {v30.8h}, [x1], #16
37+ sub v5.4h, v0.4h, v2.4h
38+ sub v6.4h, v16.4h, v3.4h
39+ add v7.4h, v1.4h, v17.4h
40+ add v0.4h, v4.4h, v7.4h
41+ add v1.4h, v5.4h, v6.4h
42+ sub v2.4h, v5.4h, v6.4h
43+ sub v3.4h, v4.4h, v7.4h
4444
4545 transpose_4x4H v0, v1, v2, v3, v4, v5, v6, v7
4646
47− add v4.4H, v0.4H, v2.4H
48− ld1 {v18.S}[0], [x0], x2
49− sshr v16.4H, v3.4H, #1
50− sshr v17.4H, v1.4H, #1
51− ld1 {v18.S}[1], [x0], x2
52− sub v5.4H, v0.4H, v2.4H
53− ld1 {v19.S}[1], [x0], x2
54− add v6.4H, v16.4H, v1.4H
55− ins v4.D[1], v5.D[0]
56− sub v7.4H, v17.4H, v3.4H
57− ld1 {v19.S}[0], [x0], x2
58− ins v6.D[1], v7.D[0]
47+ add v4.4h, v0.4h, v2.4h
48+ ld1 {v18.s}[0], [x0], x2
49+ sshr v16.4h, v3.4h, #1
50+ sshr v17.4h, v1.4h, #1
51+ ld1 {v18.s}[1], [x0], x2
52+ sub v5.4h, v0.4h, v2.4h
53+ ld1 {v19.s}[1], [x0], x2
54+ add v6.4h, v16.4h, v1.4h
55+ ins v4.d[1], v5.d[0]
56+ sub v7.4h, v17.4h, v3.4h
57+ ld1 {v19.s}[0], [x0], x2
58+ ins v6.d[1], v7.d[0]
5959 sub x0, x0, x2, lsl #2
60− add v0.8H, v4.8H, v6.8H
61− sub v1.8H, v4.8H, v6.8H
60+ add v0.8h, v4.8h, v6.8h
61+ sub v1.8h, v4.8h, v6.8h
6262
63− srshr v0.8H, v0.8H, #6
64− srshr v1.8H, v1.8H, #6
63+ srshr v0.8h, v0.8h, #6
64+ srshr v1.8h, v1.8h, #6
6565
66− uaddw v0.8H, v0.8H, v18.8B
67− uaddw v1.8H, v1.8H, v19.8B
66+ uaddw v0.8h, v0.8h, v18.8b
67+ uaddw v1.8h, v1.8h, v19.8b
6868
69− sqxtun v0.8B, v0.8H
70− sqxtun v1.8B, v1.8H
69+ sqxtun v0.8b, v0.8h
70+ sqxtun v1.8b, v1.8h
7171
72− st1 {v0.S}[0], [x0], x2
73− st1 {v0.S}[1], [x0], x2
74− st1 {v1.S}[1], [x0], x2
75− st1 {v1.S}[0], [x0], x2
72+ st1 {v0.s}[0], [x0], x2
73+ st1 {v0.s}[1], [x0], x2
74+ st1 {v1.s}[1], [x0], x2
75+ st1 {v1.s}[0], [x0], x2
7676
7777 sub x1, x1, #32
7878 ret
@@ -83,22 +83,22 @@ function ff_h264_idct_dc_add_neon, export=1 · 4 unmodified lines
8383 AARCH64_VALID_CALL_TARGET
8484 sxtw x2, w2
8585 mov w3, #0
86− ld1r {v2.8H}, [x1]
86+ ld1r {v2.8h}, [x1]
8787 strh w3, [x1]
88− srshr v2.8H, v2.8H, #6
89− ld1 {v0.S}[0], [x0], x2
90− ld1 {v0.S}[1], [x0], x2
91− uaddw v3.8H, v2.8H, v0.8B
92− ld1 {v1.S}[0], [x0], x2
93− ld1 {v1.S}[1], [x0], x2
94− uaddw v4.8H, v2.8H, v1.8B
95− sqxtun v0.8B, v3.8H
96− sqxtun v1.8B, v4.8H
88+ srshr v2.8h, v2.8h, #6
89+ ld1 {v0.s}[0], [x0], x2
90+ ld1 {v0.s}[1], [x0], x2
91+ uaddw v3.8h, v2.8h, v0.8b
92+ ld1 {v1.s}[0], [x0], x2
93+ ld1 {v1.s}[1], [x0], x2
94+ uaddw v4.8h, v2.8h, v1.8b
95+ sqxtun v0.8b, v3.8h
96+ sqxtun v1.8b, v4.8h
9797 sub x0, x0, x2, lsl #2
98− st1 {v0.S}[0], [x0], x2
99− st1 {v0.S}[1], [x0], x2
100− st1 {v1.S}[0], [x0], x2
101− st1 {v1.S}[1], [x0], x2
98+ st1 {v0.s}[0], [x0], x2
99+ st1 {v0.s}[1], [x0], x2
100+ st1 {v1.s}[0], [x0], x2
101+ st1 {v1.s}[1], [x0], x2
102102 ret
103103 endfunc
104104
@@ -194,71 +194,71 @@ endfunc · 89 unmodified lines
194194 .if \pass == 0
195195 va .req v18
196196 vb .req v30
197− sshr v18.8H, v26.8H, #1
198− add v16.8H, v24.8H, v28.8H
199− ld1 {v30.8H, v31.8H}, [x1]
200− st1 {v19.8H}, [x1], #16
201− st1 {v19.8H}, [x1], #16
202− sub v17.8H, v24.8H, v28.8H
203− sshr v19.8H, v30.8H, #1
204− sub v18.8H, v18.8H, v30.8H
205− add v19.8H, v19.8H, v26.8H
197+ sshr v18.8h, v26.8h, #1
198+ add v16.8h, v24.8h, v28.8h
199+ ld1 {v30.8h, v31.8h}, [x1]
200+ st1 {v19.8h}, [x1], #16
201+ st1 {v19.8h}, [x1], #16
202+ sub v17.8h, v24.8h, v28.8h
203+ sshr v19.8h, v30.8h, #1
204+ sub v18.8h, v18.8h, v30.8h
205+ add v19.8h, v19.8h, v26.8h
206206 .else
207207 va .req v30
208208 vb .req v18
209− sshr v30.8H, v26.8H, #1
210− sshr v19.8H, v18.8H, #1
211− add v16.8H, v24.8H, v28.8H
212− sub v17.8H, v24.8H, v28.8H
213− sub v30.8H, v30.8H, v18.8H
214− add v19.8H, v19.8H, v26.8H
209+ sshr v30.8h, v26.8h, #1
210+ sshr v19.8h, v18.8h, #1
211+ add v16.8h, v24.8h, v28.8h
212+ sub v17.8h, v24.8h, v28.8h
213+ sub v30.8h, v30.8h, v18.8h
214+ add v19.8h, v19.8h, v26.8h
215215 .endif
216− add v26.8H, v17.8H, va.8H
217− sub v28.8H, v17.8H, va.8H
218− add v24.8H, v16.8H, v19.8H
219− sub vb.8H, v16.8H, v19.8H
220− sub v16.8H, v29.8H, v27.8H
221− add v17.8H, v31.8H, v25.8H
222− sub va.8H, v31.8H, v25.8H
223− add v19.8H, v29.8H, v27.8H
224− sub v16.8H, v16.8H, v31.8H
225− sub v17.8H, v17.8H, v27.8H
226− add va.8H, va.8H, v29.8H
227− add v19.8H, v19.8H, v25.8H
228− sshr v25.8H, v25.8H, #1
229− sshr v27.8H, v27.8H, #1
230− sshr v29.8H, v29.8H, #1
231− sshr v31.8H, v31.8H, #1
232− sub v16.8H, v16.8H, v31.8H
233− sub v17.8H, v17.8H, v27.8H
234− add va.8H, va.8H, v29.8H
235− add v19.8H, v19.8H, v25.8H
236− sshr v25.8H, v16.8H, #2
237− sshr v27.8H, v17.8H, #2
238− sshr v29.8H, va.8H, #2
239− sshr v31.8H, v19.8H, #2
240− sub v19.8H, v19.8H, v25.8H
241− sub va.8H, v27.8H, va.8H
242− add v17.8H, v17.8H, v29.8H
243− add v16.8H, v16.8H, v31.8H
216+ add v26.8h, v17.8h, va.8h
217+ sub v28.8h, v17.8h, va.8h
218+ add v24.8h, v16.8h, v19.8h
219+ sub vb.8h, v16.8h, v19.8h
220+ sub v16.8h, v29.8h, v27.8h
221+ add v17.8h, v31.8h, v25.8h
222+ sub va.8h, v31.8h, v25.8h
223+ add v19.8h, v29.8h, v27.8h
224+ sub v16.8h, v16.8h, v31.8h
225+ sub v17.8h, v17.8h, v27.8h
226+ add va.8h, va.8h, v29.8h
227+ add v19.8h, v19.8h, v25.8h
228+ sshr v25.8h, v25.8h, #1
229+ sshr v27.8h, v27.8h, #1
230+ sshr v29.8h, v29.8h, #1
231+ sshr v31.8h, v31.8h, #1
232+ sub v16.8h, v16.8h, v31.8h
233+ sub v17.8h, v17.8h, v27.8h
234+ add va.8h, va.8h, v29.8h
235+ add v19.8h, v19.8h, v25.8h
236+ sshr v25.8h, v16.8h, #2
237+ sshr v27.8h, v17.8h, #2
238+ sshr v29.8h, va.8h, #2
239+ sshr v31.8h, v19.8h, #2
240+ sub v19.8h, v19.8h, v25.8h
241+ sub va.8h, v27.8h, va.8h
242+ add v17.8h, v17.8h, v29.8h
243+ add v16.8h, v16.8h, v31.8h
244244 .if \pass == 0
245− sub v31.8H, v24.8H, v19.8H
246− add v24.8H, v24.8H, v19.8H
247− add v25.8H, v26.8H, v18.8H
248− sub v18.8H, v26.8H, v18.8H
249− add v26.8H, v28.8H, v17.8H
250− add v27.8H, v30.8H, v16.8H
251− sub v29.8H, v28.8H, v17.8H
252− sub v28.8H, v30.8H, v16.8H
245+ sub v31.8h, v24.8h, v19.8h
246+ add v24.8h, v24.8h, v19.8h
247+ add v25.8h, v26.8h, v18.8h
248+ sub v18.8h, v26.8h, v18.8h
249+ add v26.8h, v28.8h, v17.8h
250+ add v27.8h, v30.8h, v16.8h
251+ sub v29.8h, v28.8h, v17.8h
252+ sub v28.8h, v30.8h, v16.8h
253253 .else
254− sub v31.8H, v24.8H, v19.8H
255− add v24.8H, v24.8H, v19.8H
256− add v25.8H, v26.8H, v30.8H
257− sub v30.8H, v26.8H, v30.8H
258− add v26.8H, v28.8H, v17.8H
259− sub v29.8H, v28.8H, v17.8H
260− add v27.8H, v18.8H, v16.8H
261− sub v28.8H, v18.8H, v16.8H
254+ sub v31.8h, v24.8h, v19.8h
255+ add v24.8h, v24.8h, v19.8h
256+ add v25.8h, v26.8h, v30.8h
257+ sub v30.8h, v26.8h, v30.8h
258+ add v26.8h, v28.8h, v17.8h
259+ sub v29.8h, v28.8h, v17.8h
260+ add v27.8h, v18.8h, v16.8h
261+ sub v28.8h, v18.8h, v16.8h
262262 .endif
263263 .unreq va
264264 .unreq vb
@@ -267,63 +267,63 @@ endfunc · 2 unmodified lines
267267 function ff_h264_idct8_add_neon, export=1
268268 .L_ff_h264_idct8_add_neon:
269269 AARCH64_VALID_CALL_TARGET
270− movi v19.8H, #0
270+ movi v19.8h, #0
271271 sxtw x2, w2
272− ld1 {v24.8H, v25.8H}, [x1]
273− st1 {v19.8H}, [x1], #16
274− st1 {v19.8H}, [x1], #16
275− ld1 {v26.8H, v27.8H}, [x1]
276− st1 {v19.8H}, [x1], #16
277− st1 {v19.8H}, [x1], #16
278− ld1 {v28.8H, v29.8H}, [x1]
279− st1 {v19.8H}, [x1], #16
280− st1 {v19.8H}, [x1], #16
272+ ld1 {v24.8h, v25.8h}, [x1]
273+ st1 {v19.8h}, [x1], #16
274+ st1 {v19.8h}, [x1], #16
275+ ld1 {v26.8h, v27.8h}, [x1]
276+ st1 {v19.8h}, [x1], #16
277+ st1 {v19.8h}, [x1], #16
278+ ld1 {v28.8h, v29.8h}, [x1]
279+ st1 {v19.8h}, [x1], #16
280+ st1 {v19.8h}, [x1], #16
281281
282282 idct8x8_cols 0
283283 transpose_8x8H v24, v25, v26, v27, v28, v29, v18, v31, v6, v7
284284 idct8x8_cols 1
285285
286286 mov x3, x0
287− srshr v24.8H, v24.8H, #6
288− ld1 {v0.8B}, [x0], x2
289− srshr v25.8H, v25.8H, #6
290− ld1 {v1.8B}, [x0], x2
291− srshr v26.8H, v26.8H, #6
292− ld1 {v2.8B}, [x0], x2
293− srshr v27.8H, v27.8H, #6
294− ld1 {v3.8B}, [x0], x2
295− srshr v28.8H, v28.8H, #6
296− ld1 {v4.8B}, [x0], x2
297− srshr v29.8H, v29.8H, #6
298− ld1 {v5.8B}, [x0], x2
299− srshr v30.8H, v30.8H, #6
300− ld1 {v6.8B}, [x0], x2
301− srshr v31.8H, v31.8H, #6
302− ld1 {v7.8B}, [x0], x2
303− uaddw v24.8H, v24.8H, v0.8B
304− uaddw v25.8H, v25.8H, v1.8B
305− uaddw v26.8H, v26.8H, v2.8B
306− sqxtun v0.8B, v24.8H
307− uaddw v27.8H, v27.8H, v3.8B
308− sqxtun v1.8B, v25.8H
309− uaddw v28.8H, v28.8H, v4.8B
310− sqxtun v2.8B, v26.8H
311− st1 {v0.8B}, [x3], x2
312− uaddw v29.8H, v29.8H, v5.8B
313− sqxtun v3.8B, v27.8H
314− st1 {v1.8B}, [x3], x2
315− uaddw v30.8H, v30.8H, v6.8B
316− sqxtun v4.8B, v28.8H
317− st1 {v2.8B}, [x3], x2
318− uaddw v31.8H, v31.8H, v7.8B
319− sqxtun v5.8B, v29.8H
320− st1 {v3.8B}, [x3], x2
321− sqxtun v6.8B, v30.8H
322− sqxtun v7.8B, v31.8H
323− st1 {v4.8B}, [x3], x2
324− st1 {v5.8B}, [x3], x2
325− st1 {v6.8B}, [x3], x2
326− st1 {v7.8B}, [x3], x2
287+ srshr v24.8h, v24.8h, #6
288+ ld1 {v0.8b}, [x0], x2
289+ srshr v25.8h, v25.8h, #6
290+ ld1 {v1.8b}, [x0], x2
291+ srshr v26.8h, v26.8h, #6
292+ ld1 {v2.8b}, [x0], x2
293+ srshr v27.8h, v27.8h, #6
294+ ld1 {v3.8b}, [x0], x2
295+ srshr v28.8h, v28.8h, #6
296+ ld1 {v4.8b}, [x0], x2
297+ srshr v29.8h, v29.8h, #6
298+ ld1 {v5.8b}, [x0], x2
299+ srshr v30.8h, v30.8h, #6
300+ ld1 {v6.8b}, [x0], x2
301+ srshr v31.8h, v31.8h, #6
302+ ld1 {v7.8b}, [x0], x2
303+ uaddw v24.8h, v24.8h, v0.8b
304+ uaddw v25.8h, v25.8h, v1.8b
305+ uaddw v26.8h, v26.8h, v2.8b
306+ sqxtun v0.8b, v24.8h
307+ uaddw v27.8h, v27.8h, v3.8b
308+ sqxtun v1.8b, v25.8h
309+ uaddw v28.8h, v28.8h, v4.8b
310+ sqxtun v2.8b, v26.8h
311+ st1 {v0.8b}, [x3], x2
312+ uaddw v29.8h, v29.8h, v5.8b
313+ sqxtun v3.8b, v27.8h
314+ st1 {v1.8b}, [x3], x2
315+ uaddw v30.8h, v30.8h, v6.8b
316+ sqxtun v4.8b, v28.8h
317+ st1 {v2.8b}, [x3], x2
318+ uaddw v31.8h, v31.8h, v7.8b
319+ sqxtun v5.8b, v29.8h
320+ st1 {v3.8b}, [x3], x2
321+ sqxtun v6.8b, v30.8h
322+ sqxtun v7.8b, v31.8h
323+ st1 {v4.8b}, [x3], x2
324+ st1 {v5.8b}, [x3], x2
325+ st1 {v6.8b}, [x3], x2
326+ st1 {v7.8b}, [x3], x2
327327
328328 sub x1, x1, #128
329329 ret
@@ -334,42 +334,42 @@ function ff_h264_idct8_dc_add_neon, export=1 · 4 unmodified lines
334334 AARCH64_VALID_CALL_TARGET
335335 mov w3, #0
336336 sxtw x2, w2
337− ld1r {v31.8H}, [x1]
337+ ld1r {v31.8h}, [x1]
338338 strh w3, [x1]
339− ld1 {v0.8B}, [x0], x2
340− srshr v31.8H, v31.8H, #6
341− ld1 {v1.8B}, [x0], x2
342− ld1 {v2.8B}, [x0], x2
343− uaddw v24.8H, v31.8H, v0.8B
344− ld1 {v3.8B}, [x0], x2
345− uaddw v25.8H, v31.8H, v1.8B
346− ld1 {v4.8B}, [x0], x2
347− uaddw v26.8H, v31.8H, v2.8B
348− ld1 {v5.8B}, [x0], x2
349− uaddw v27.8H, v31.8H, v3.8B
350− ld1 {v6.8B}, [x0], x2
351− uaddw v28.8H, v31.8H, v4.8B
352− ld1 {v7.8B}, [x0], x2
353− uaddw v29.8H, v31.8H, v5.8B
354− uaddw v30.8H, v31.8H, v6.8B
355− uaddw v31.8H, v31.8H, v7.8B
356− sqxtun v0.8B, v24.8H
357− sqxtun v1.8B, v25.8H
358− sqxtun v2.8B, v26.8H
359− sqxtun v3.8B, v27.8H
339+ ld1 {v0.8b}, [x0], x2
340+ srshr v31.8h, v31.8h, #6
341+ ld1 {v1.8b}, [x0], x2
342+ ld1 {v2.8b}, [x0], x2
343+ uaddw v24.8h, v31.8h, v0.8b
344+ ld1 {v3.8b}, [x0], x2
345+ uaddw v25.8h, v31.8h, v1.8b
346+ ld1 {v4.8b}, [x0], x2
347+ uaddw v26.8h, v31.8h, v2.8b
348+ ld1 {v5.8b}, [x0], x2
349+ uaddw v27.8h, v31.8h, v3.8b
350+ ld1 {v6.8b}, [x0], x2
351+ uaddw v28.8h, v31.8h, v4.8b
352+ ld1 {v7.8b}, [x0], x2
353+ uaddw v29.8h, v31.8h, v5.8b
354+ uaddw v30.8h, v31.8h, v6.8b
355+ uaddw v31.8h, v31.8h, v7.8b
356+ sqxtun v0.8b, v24.8h
357+ sqxtun v1.8b, v25.8h
358+ sqxtun v2.8b, v26.8h
359+ sqxtun v3.8b, v27.8h
360360 sub x0, x0, x2, lsl #3
361− st1 {v0.8B}, [x0], x2
362− sqxtun v4.8B, v28.8H
363− st1 {v1.8B}, [x0], x2
364− sqxtun v5.8B, v29.8H
365− st1 {v2.8B}, [x0], x2
366− sqxtun v6.8B, v30.8H
367− st1 {v3.8B}, [x0], x2
368− sqxtun v7.8B, v31.8H
369− st1 {v4.8B}, [x0], x2
370− st1 {v5.8B}, [x0], x2
371− st1 {v6.8B}, [x0], x2
372− st1 {v7.8B}, [x0], x2
361+ st1 {v0.8b}, [x0], x2
362+ sqxtun v4.8b, v28.8h
363+ st1 {v1.8b}, [x0], x2
364+ sqxtun v5.8b, v29.8h
365+ st1 {v2.8b}, [x0], x2
366+ sqxtun v6.8b, v30.8h
367+ st1 {v3.8b}, [x0], x2
368+ sqxtun v7.8b, v31.8h
369+ st1 {v4.8b}, [x0], x2
370+ st1 {v5.8b}, [x0], x2
371+ st1 {v6.8b}, [x0], x2
372+ st1 {v7.8b}, [x0], x2
373373 ret
374374 endfunc
375375
libavcodec/aarch64/h264qpel_neon.S+278−278
@@ -27,127 +27,127 @@ · 26 unmodified lines
2727 .macro lowpass_const r
2828 movz \r, #20, lsl #16
2929 movk \r, #5
30− mov v6.S[0], \r
30+ mov v6.s[0], \r
3131 .endm
3232
3333 //trashes v0-v5
3434 .macro lowpass_8 r0, r1, r2, r3, d0, d1, narrow=1
35− ext v2.8B, \r0\().8B, \r1\().8B, #2
36− ext v3.8B, \r0\().8B, \r1\().8B, #3
37− uaddl v2.8H, v2.8B, v3.8B
38− ext v4.8B, \r0\().8B, \r1\().8B, #1
39− ext v5.8B, \r0\().8B, \r1\().8B, #4
40− uaddl v4.8H, v4.8B, v5.8B
41− ext v1.8B, \r0\().8B, \r1\().8B, #5
42− uaddl \d0\().8H, \r0\().8B, v1.8B
43− ext v0.8B, \r2\().8B, \r3\().8B, #2
44− mla \d0\().8H, v2.8H, v6.H[1]
45− ext v1.8B, \r2\().8B, \r3\().8B, #3
46− uaddl v0.8H, v0.8B, v1.8B
47− ext v1.8B, \r2\().8B, \r3\().8B, #1
48− mls \d0\().8H, v4.8H, v6.H[0]
49− ext v3.8B, \r2\().8B, \r3\().8B, #4
50− uaddl v1.8H, v1.8B, v3.8B
51− ext v2.8B, \r2\().8B, \r3\().8B, #5
52− uaddl \d1\().8H, \r2\().8B, v2.8B
53− mla \d1\().8H, v0.8H, v6.H[1]
54− mls \d1\().8H, v1.8H, v6.H[0]
35+ ext v2.8b, \r0\().8b, \r1\().8b, #2
36+ ext v3.8b, \r0\().8b, \r1\().8b, #3
37+ uaddl v2.8h, v2.8b, v3.8b
38+ ext v4.8b, \r0\().8b, \r1\().8b, #1
39+ ext v5.8b, \r0\().8b, \r1\().8b, #4
40+ uaddl v4.8h, v4.8b, v5.8b
41+ ext v1.8b, \r0\().8b, \r1\().8b, #5
42+ uaddl \d0\().8h, \r0\().8b, v1.8b
43+ ext v0.8b, \r2\().8b, \r3\().8b, #2
44+ mla \d0\().8h, v2.8h, v6.h[1]
45+ ext v1.8b, \r2\().8b, \r3\().8b, #3
46+ uaddl v0.8h, v0.8b, v1.8b
47+ ext v1.8b, \r2\().8b, \r3\().8b, #1
48+ mls \d0\().8h, v4.8h, v6.h[0]
49+ ext v3.8b, \r2\().8b, \r3\().8b, #4
50+ uaddl v1.8h, v1.8b, v3.8b
51+ ext v2.8b, \r2\().8b, \r3\().8b, #5
52+ uaddl \d1\().8h, \r2\().8b, v2.8b
53+ mla \d1\().8h, v0.8h, v6.h[1]
54+ mls \d1\().8h, v1.8h, v6.h[0]
5555 .if \narrow
56− sqrshrun \d0\().8B, \d0\().8H, #5
57− sqrshrun \d1\().8B, \d1\().8H, #5
56+ sqrshrun \d0\().8b, \d0\().8h, #5
57+ sqrshrun \d1\().8b, \d1\().8h, #5
5858 .endif
5959 .endm
6060
6161 //trashes v0-v4
6262 .macro lowpass_8_v r0, r1, r2, r3, r4, r5, r6, d0, d1, narrow=1
63− uaddl v2.8H, \r2\().8B, \r3\().8B
64− uaddl v0.8H, \r3\().8B, \r4\().8B
65− uaddl v4.8H, \r1\().8B, \r4\().8B
66− uaddl v1.8H, \r2\().8B, \r5\().8B
67− uaddl \d0\().8H, \r0\().8B, \r5\().8B
68− uaddl \d1\().8H, \r1\().8B, \r6\().8B
69− mla \d0\().8H, v2.8H, v6.H[1]
70− mls \d0\().8H, v4.8H, v6.H[0]
71− mla \d1\().8H, v0.8H, v6.H[1]
72− mls \d1\().8H, v1.8H, v6.H[0]
63+ uaddl v2.8h, \r2\().8b, \r3\().8b
64+ uaddl v0.8h, \r3\().8b, \r4\().8b
65+ uaddl v4.8h, \r1\().8b, \r4\().8b
66+ uaddl v1.8h, \r2\().8b, \r5\().8b
67+ uaddl \d0\().8h, \r0\().8b, \r5\().8b
68+ uaddl \d1\().8h, \r1\().8b, \r6\().8b
69+ mla \d0\().8h, v2.8h, v6.h[1]
70+ mls \d0\().8h, v4.8h, v6.h[0]
71+ mla \d1\().8h, v0.8h, v6.h[1]
72+ mls \d1\().8h, v1.8h, v6.h[0]
7373 .if \narrow
74− sqrshrun \d0\().8B, \d0\().8H, #5
75− sqrshrun \d1\().8B, \d1\().8H, #5
74+ sqrshrun \d0\().8b, \d0\().8h, #5
75+ sqrshrun \d1\().8b, \d1\().8h, #5
7676 .endif
7777 .endm
7878
7979 //trashes v0-v5, v7, v30-v31
8080 .macro lowpass_8H r0, r1
81− ext v0.16B, \r0\().16B, \r0\().16B, #2
82− ext v1.16B, \r0\().16B, \r0\().16B, #3
83− uaddl v0.8H, v0.8B, v1.8B
84− ext v2.16B, \r0\().16B, \r0\().16B, #1
85− ext v3.16B, \r0\().16B, \r0\().16B, #4
86− uaddl v2.8H, v2.8B, v3.8B
87− ext v30.16B, \r0\().16B, \r0\().16B, #5
88− uaddl \r0\().8H, \r0\().8B, v30.8B
89− ext v4.16B, \r1\().16B, \r1\().16B, #2
90− mla \r0\().8H, v0.8H, v6.H[1]
91− ext v5.16B, \r1\().16B, \r1\().16B, #3
92− uaddl v4.8H, v4.8B, v5.8B
93− ext v7.16B, \r1\().16B, \r1\().16B, #1
94− mls \r0\().8H, v2.8H, v6.H[0]
95− ext v0.16B, \r1\().16B, \r1\().16B, #4
96− uaddl v7.8H, v7.8B, v0.8B
97− ext v31.16B, \r1\().16B, \r1\().16B, #5
98− uaddl \r1\().8H, \r1\().8B, v31.8B
99− mla \r1\().8H, v4.8H, v6.H[1]
100− mls \r1\().8H, v7.8H, v6.H[0]
81+ ext v0.16b, \r0\().16b, \r0\().16b, #2
82+ ext v1.16b, \r0\().16b, \r0\().16b, #3
83+ uaddl v0.8h, v0.8b, v1.8b
84+ ext v2.16b, \r0\().16b, \r0\().16b, #1
85+ ext v3.16b, \r0\().16b, \r0\().16b, #4
86+ uaddl v2.8h, v2.8b, v3.8b
87+ ext v30.16b, \r0\().16b, \r0\().16b, #5
88+ uaddl \r0\().8h, \r0\().8b, v30.8b
89+ ext v4.16b, \r1\().16b, \r1\().16b, #2
90+ mla \r0\().8h, v0.8h, v6.h[1]
91+ ext v5.16b, \r1\().16b, \r1\().16b, #3
92+ uaddl v4.8h, v4.8b, v5.8b
93+ ext v7.16b, \r1\().16b, \r1\().16b, #1
94+ mls \r0\().8h, v2.8h, v6.h[0]
95+ ext v0.16b, \r1\().16b, \r1\().16b, #4
96+ uaddl v7.8h, v7.8b, v0.8b
97+ ext v31.16b, \r1\().16b, \r1\().16b, #5
98+ uaddl \r1\().8h, \r1\().8b, v31.8b
99+ mla \r1\().8h, v4.8h, v6.h[1]
100+ mls \r1\().8h, v7.8h, v6.h[0]
101101 .endm
102102
103103 // trashes v2-v5, v30
104104 .macro lowpass_8_1 r0, r1, d0, narrow=1
105− ext v2.8B, \r0\().8B, \r1\().8B, #2
106− ext v3.8B, \r0\().8B, \r1\().8B, #3
107− uaddl v2.8H, v2.8B, v3.8B
108− ext v4.8B, \r0\().8B, \r1\().8B, #1
109− ext v5.8B, \r0\().8B, \r1\().8B, #4
110− uaddl v4.8H, v4.8B, v5.8B
111− ext v30.8B, \r0\().8B, \r1\().8B, #5
112− uaddl \d0\().8H, \r0\().8B, v30.8B
113− mla \d0\().8H, v2.8H, v6.H[1]
114− mls \d0\().8H, v4.8H, v6.H[0]
105+ ext v2.8b, \r0\().8b, \r1\().8b, #2
106+ ext v3.8b, \r0\().8b, \r1\().8b, #3
107+ uaddl v2.8h, v2.8b, v3.8b
108+ ext v4.8b, \r0\().8b, \r1\().8b, #1
109+ ext v5.8b, \r0\().8b, \r1\().8b, #4
110+ uaddl v4.8h, v4.8b, v5.8b
111+ ext v30.8b, \r0\().8b, \r1\().8b, #5
112+ uaddl \d0\().8h, \r0\().8b, v30.8b
113+ mla \d0\().8h, v2.8h, v6.h[1]
114+ mls \d0\().8h, v4.8h, v6.h[0]
115115 .if \narrow
116− sqrshrun \d0\().8B, \d0\().8H, #5
116+ sqrshrun \d0\().8b, \d0\().8h, #5
117117 .endif
118118 .endm
119119
120120 // trashed v0-v7
121121 .macro lowpass_8.16 r0, r1, r2, r3, r4, r5
122− saddl v5.4S, \r2\().4H, \r3\().4H
123− saddl2 v1.4S, \r2\().8H, \r3\().8H
124− saddl v6.4S, \r1\().4H, \r4\().4H
125− saddl2 v2.4S, \r1\().8H, \r4\().8H
126− saddl v0.4S, \r0\().4H, \r5\().4H
127− saddl2 v4.4S, \r0\().8H, \r5\().8H
128−
129− shl v3.4S, v5.4S, #4
130− shl v5.4S, v5.4S, #2
131− shl v7.4S, v6.4S, #2
132− add v5.4S, v5.4S, v3.4S
133− add v6.4S, v6.4S, v7.4S
134−
135− shl v3.4S, v1.4S, #4
136− shl v1.4S, v1.4S, #2
137− shl v7.4S, v2.4S, #2
138− add v1.4S, v1.4S, v3.4S
139− add v2.4S, v2.4S, v7.4S
140−
141− add v5.4S, v5.4S, v0.4S
142− sub v5.4S, v5.4S, v6.4S
143−
144− add v1.4S, v1.4S, v4.4S
145− sub v1.4S, v1.4S, v2.4S
146−
147− rshrn v5.4H, v5.4S, #10
148− rshrn2 v5.8H, v1.4S, #10
149−
150− sqxtun \r0\().8B, v5.8H
122+ saddl v5.4s, \r2\().4h, \r3\().4h
123+ saddl2 v1.4s, \r2\().8h, \r3\().8h
124+ saddl v6.4s, \r1\().4h, \r4\().4h
125+ saddl2 v2.4s, \r1\().8h, \r4\().8h
126+ saddl v0.4s, \r0\().4h, \r5\().4h
127+ saddl2 v4.4s, \r0\().8h, \r5\().8h
128+
129+ shl v3.4s, v5.4s, #4
130+ shl v5.4s, v5.4s, #2
131+ shl v7.4s, v6.4s, #2
132+ add v5.4s, v5.4s, v3.4s
133+ add v6.4s, v6.4s, v7.4s
134+
135+ shl v3.4s, v1.4s, #4
136+ shl v1.4s, v1.4s, #2
137+ shl v7.4s, v2.4s, #2
138+ add v1.4s, v1.4s, v3.4s
139+ add v2.4s, v2.4s, v7.4s
140+
141+ add v5.4s, v5.4s, v0.4s
142+ sub v5.4s, v5.4s, v6.4s
143+
144+ add v1.4s, v1.4s, v4.4s
145+ sub v1.4s, v1.4s, v2.4s
146+
147+ rshrn v5.4h, v5.4s, #10
148+ rshrn2 v5.8h, v1.4s, #10
149+
150+ sqxtun \r0\().8b, v5.8h
151151 .endm
152152
153153 function put_h264_qpel16_h_lowpass_neon_packed
@@ -176,19 +176,19 @@ function \type\()_h264_qpel16_h_lowpass_neon · 22 unmodified lines
176176 endfunc
177177
178178 function \type\()_h264_qpel8_h_lowpass_neon
179−1: ld1 {v28.8B, v29.8B}, [x1], x2
180− ld1 {v16.8B, v17.8B}, [x1], x2
179+1: ld1 {v28.8b, v29.8b}, [x1], x2
180+ ld1 {v16.8b, v17.8b}, [x1], x2
181181 subs x12, x12, #2
182182 lowpass_8 v28, v29, v16, v17, v28, v16
183183 .ifc \type,avg
184− ld1 {v2.8B}, [x0], x3
185− ld1 {v3.8B}, [x0]
186− urhadd v28.8B, v28.8B, v2.8B
187− urhadd v16.8B, v16.8B, v3.8B
184+ ld1 {v2.8b}, [x0], x3
185+ ld1 {v3.8b}, [x0]
186+ urhadd v28.8b, v28.8b, v2.8b
187+ urhadd v16.8b, v16.8b, v3.8b
188188 sub x0, x0, x3
189189 .endif
190− st1 {v28.8B}, [x0], x3
191− st1 {v16.8B}, [x0], x3
190+ st1 {v28.8b}, [x0], x3
191+ st1 {v16.8b}, [x0], x3
192192 b.ne 1b
193193 ret
194194 endfunc
@@ -213,23 +213,23 @@ function \type\()_h264_qpel16_h_lowpass_l2_neon · 18 unmodified lines
213213 endfunc
214214
215215 function \type\()_h264_qpel8_h_lowpass_l2_neon
216−1: ld1 {v26.8B, v27.8B}, [x1], x2
217− ld1 {v16.8B, v17.8B}, [x1], x2
218− ld1 {v28.8B}, [x3], x2
219− ld1 {v29.8B}, [x3], x2
216+1: ld1 {v26.8b, v27.8b}, [x1], x2
217+ ld1 {v16.8b, v17.8b}, [x1], x2
218+ ld1 {v28.8b}, [x3], x2
219+ ld1 {v29.8b}, [x3], x2
220220 subs x12, x12, #2
221221 lowpass_8 v26, v27, v16, v17, v26, v27
222− urhadd v26.8B, v26.8B, v28.8B
223− urhadd v27.8B, v27.8B, v29.8B
222+ urhadd v26.8b, v26.8b, v28.8b
223+ urhadd v27.8b, v27.8b, v29.8b
224224 .ifc \type,avg
225− ld1 {v2.8B}, [x0], x2
226− ld1 {v3.8B}, [x0]
227− urhadd v26.8B, v26.8B, v2.8B
228− urhadd v27.8B, v27.8B, v3.8B
225+ ld1 {v2.8b}, [x0], x2
226+ ld1 {v3.8b}, [x0]
227+ urhadd v26.8b, v26.8b, v2.8b
228+ urhadd v27.8b, v27.8b, v3.8b
229229 sub x0, x0, x2
230230 .endif
231− st1 {v26.8B}, [x0], x2
232− st1 {v27.8B}, [x0], x2
231+ st1 {v26.8b}, [x0], x2
232+ st1 {v27.8b}, [x0], x2
233233 b.ne 1b
234234 ret
235235 endfunc
@@ -270,52 +270,52 @@ function \type\()_h264_qpel16_v_lowpass_neon · 34 unmodified lines
270270 endfunc
271271
272272 function \type\()_h264_qpel8_v_lowpass_neon
273− ld1 {v16.8B}, [x1], x3
274− ld1 {v17.8B}, [x1], x3
275− ld1 {v18.8B}, [x1], x3
276− ld1 {v19.8B}, [x1], x3
277− ld1 {v20.8B}, [x1], x3
278− ld1 {v21.8B}, [x1], x3
279− ld1 {v22.8B}, [x1], x3
280− ld1 {v23.8B}, [x1], x3
281− ld1 {v24.8B}, [x1], x3
282− ld1 {v25.8B}, [x1], x3
283− ld1 {v26.8B}, [x1], x3
284− ld1 {v27.8B}, [x1], x3
285− ld1 {v28.8B}, [x1]
273+ ld1 {v16.8b}, [x1], x3
274+ ld1 {v17.8b}, [x1], x3
275+ ld1 {v18.8b}, [x1], x3
276+ ld1 {v19.8b}, [x1], x3
277+ ld1 {v20.8b}, [x1], x3
278+ ld1 {v21.8b}, [x1], x3
279+ ld1 {v22.8b}, [x1], x3
280+ ld1 {v23.8b}, [x1], x3
281+ ld1 {v24.8b}, [x1], x3
282+ ld1 {v25.8b}, [x1], x3
283+ ld1 {v26.8b}, [x1], x3
284+ ld1 {v27.8b}, [x1], x3
285+ ld1 {v28.8b}, [x1]
286286
287287 lowpass_8_v v16, v17, v18, v19, v20, v21, v22, v16, v17
288288 lowpass_8_v v18, v19, v20, v21, v22, v23, v24, v18, v19
289289 lowpass_8_v v20, v21, v22, v23, v24, v25, v26, v20, v21
290290 lowpass_8_v v22, v23, v24, v25, v26, v27, v28, v22, v23
291291 .ifc \type,avg
292− ld1 {v24.8B}, [x0], x2
293− ld1 {v25.8B}, [x0], x2
294− ld1 {v26.8B}, [x0], x2
295− urhadd v16.8B, v16.8B, v24.8B
296− ld1 {v27.8B}, [x0], x2
297− urhadd v17.8B, v17.8B, v25.8B
298− ld1 {v28.8B}, [x0], x2
299− urhadd v18.8B, v18.8B, v26.8B
300− ld1 {v29.8B}, [x0], x2
301− urhadd v19.8B, v19.8B, v27.8B
302− ld1 {v30.8B}, [x0], x2
303− urhadd v20.8B, v20.8B, v28.8B
304− ld1 {v31.8B}, [x0], x2
305− urhadd v21.8B, v21.8B, v29.8B
306− urhadd v22.8B, v22.8B, v30.8B
307− urhadd v23.8B, v23.8B, v31.8B
292+ ld1 {v24.8b}, [x0], x2
293+ ld1 {v25.8b}, [x0], x2
294+ ld1 {v26.8b}, [x0], x2
295+ urhadd v16.8b, v16.8b, v24.8b
296+ ld1 {v27.8b}, [x0], x2
297+ urhadd v17.8b, v17.8b, v25.8b
298+ ld1 {v28.8b}, [x0], x2
299+ urhadd v18.8b, v18.8b, v26.8b
300+ ld1 {v29.8b}, [x0], x2
301+ urhadd v19.8b, v19.8b, v27.8b
302+ ld1 {v30.8b}, [x0], x2
303+ urhadd v20.8b, v20.8b, v28.8b
304+ ld1 {v31.8b}, [x0], x2
305+ urhadd v21.8b, v21.8b, v29.8b
306+ urhadd v22.8b, v22.8b, v30.8b
307+ urhadd v23.8b, v23.8b, v31.8b
308308 sub x0, x0, x2, lsl #3
309309 .endif
310310
311− st1 {v16.8B}, [x0], x2
312− st1 {v17.8B}, [x0], x2
313− st1 {v18.8B}, [x0], x2
314− st1 {v19.8B}, [x0], x2
315− st1 {v20.8B}, [x0], x2
316− st1 {v21.8B}, [x0], x2
317− st1 {v22.8B}, [x0], x2
318− st1 {v23.8B}, [x0], x2
311+ st1 {v16.8b}, [x0], x2
312+ st1 {v17.8b}, [x0], x2
313+ st1 {v18.8b}, [x0], x2
314+ st1 {v19.8b}, [x0], x2
315+ st1 {v20.8b}, [x0], x2
316+ st1 {v21.8b}, [x0], x2
317+ st1 {v22.8b}, [x0], x2
318+ st1 {v23.8b}, [x0], x2
319319
320320 ret
321321 endfunc
@@ -343,70 +343,70 @@ function \type\()_h264_qpel16_v_lowpass_l2_neon · 21 unmodified lines
343343 endfunc
344344
345345 function \type\()_h264_qpel8_v_lowpass_l2_neon
346− ld1 {v16.8B}, [x1], x3
347− ld1 {v17.8B}, [x1], x3
348− ld1 {v18.8B}, [x1], x3
349− ld1 {v19.8B}, [x1], x3
350− ld1 {v20.8B}, [x1], x3
351− ld1 {v21.8B}, [x1], x3
352− ld1 {v22.8B}, [x1], x3
353− ld1 {v23.8B}, [x1], x3
354− ld1 {v24.8B}, [x1], x3
355− ld1 {v25.8B}, [x1], x3
356− ld1 {v26.8B}, [x1], x3
357− ld1 {v27.8B}, [x1], x3
358− ld1 {v28.8B}, [x1]
346+ ld1 {v16.8b}, [x1], x3
347+ ld1 {v17.8b}, [x1], x3
348+ ld1 {v18.8b}, [x1], x3
349+ ld1 {v19.8b}, [x1], x3
350+ ld1 {v20.8b}, [x1], x3
351+ ld1 {v21.8b}, [x1], x3
352+ ld1 {v22.8b}, [x1], x3
353+ ld1 {v23.8b}, [x1], x3
354+ ld1 {v24.8b}, [x1], x3
355+ ld1 {v25.8b}, [x1], x3
356+ ld1 {v26.8b}, [x1], x3
357+ ld1 {v27.8b}, [x1], x3
358+ ld1 {v28.8b}, [x1]
359359
360360 lowpass_8_v v16, v17, v18, v19, v20, v21, v22, v16, v17
361361 lowpass_8_v v18, v19, v20, v21, v22, v23, v24, v18, v19
362362 lowpass_8_v v20, v21, v22, v23, v24, v25, v26, v20, v21
363363 lowpass_8_v v22, v23, v24, v25, v26, v27, v28, v22, v23
364364
365− ld1 {v24.8B}, [x12], x2
366− ld1 {v25.8B}, [x12], x2
367− ld1 {v26.8B}, [x12], x2
368− ld1 {v27.8B}, [x12], x2
369− ld1 {v28.8B}, [x12], x2
370− urhadd v16.8B, v24.8B, v16.8B
371− urhadd v17.8B, v25.8B, v17.8B
372− ld1 {v29.8B}, [x12], x2
373− urhadd v18.8B, v26.8B, v18.8B
374− urhadd v19.8B, v27.8B, v19.8B
375− ld1 {v30.8B}, [x12], x2
376− urhadd v20.8B, v28.8B, v20.8B
377− urhadd v21.8B, v29.8B, v21.8B
378− ld1 {v31.8B}, [x12], x2
379− urhadd v22.8B, v30.8B, v22.8B
380− urhadd v23.8B, v31.8B, v23.8B
365+ ld1 {v24.8b}, [x12], x2
366+ ld1 {v25.8b}, [x12], x2
367+ ld1 {v26.8b}, [x12], x2
368+ ld1 {v27.8b}, [x12], x2
369+ ld1 {v28.8b}, [x12], x2
370+ urhadd v16.8b, v24.8b, v16.8b
371+ urhadd v17.8b, v25.8b, v17.8b
372+ ld1 {v29.8b}, [x12], x2
373+ urhadd v18.8b, v26.8b, v18.8b
374+ urhadd v19.8b, v27.8b, v19.8b
375+ ld1 {v30.8b}, [x12], x2
376+ urhadd v20.8b, v28.8b, v20.8b
377+ urhadd v21.8b, v29.8b, v21.8b
378+ ld1 {v31.8b}, [x12], x2
379+ urhadd v22.8b, v30.8b, v22.8b
380+ urhadd v23.8b, v31.8b, v23.8b
381381
382382 .ifc \type,avg
383− ld1 {v24.8B}, [x0], x3
384− ld1 {v25.8B}, [x0], x3
385− ld1 {v26.8B}, [x0], x3
386− urhadd v16.8B, v16.8B, v24.8B
387− ld1 {v27.8B}, [x0], x3
388− urhadd v17.8B, v17.8B, v25.8B
389− ld1 {v28.8B}, [x0], x3
390− urhadd v18.8B, v18.8B, v26.8B
391− ld1 {v29.8B}, [x0], x3
392− urhadd v19.8B, v19.8B, v27.8B
393− ld1 {v30.8B}, [x0], x3
394− urhadd v20.8B, v20.8B, v28.8B
395− ld1 {v31.8B}, [x0], x3
396− urhadd v21.8B, v21.8B, v29.8B
397− urhadd v22.8B, v22.8B, v30.8B
398− urhadd v23.8B, v23.8B, v31.8B
383+ ld1 {v24.8b}, [x0], x3
384+ ld1 {v25.8b}, [x0], x3
385+ ld1 {v26.8b}, [x0], x3
386+ urhadd v16.8b, v16.8b, v24.8b
387+ ld1 {v27.8b}, [x0], x3
388+ urhadd v17.8b, v17.8b, v25.8b
389+ ld1 {v28.8b}, [x0], x3
390+ urhadd v18.8b, v18.8b, v26.8b
391+ ld1 {v29.8b}, [x0], x3
392+ urhadd v19.8b, v19.8b, v27.8b
393+ ld1 {v30.8b}, [x0], x3
394+ urhadd v20.8b, v20.8b, v28.8b
395+ ld1 {v31.8b}, [x0], x3
396+ urhadd v21.8b, v21.8b, v29.8b
397+ urhadd v22.8b, v22.8b, v30.8b
398+ urhadd v23.8b, v23.8b, v31.8b
399399 sub x0, x0, x3, lsl #3
400400 .endif
401401
402− st1 {v16.8B}, [x0], x3
403− st1 {v17.8B}, [x0], x3
404− st1 {v18.8B}, [x0], x3
405− st1 {v19.8B}, [x0], x3
406− st1 {v20.8B}, [x0], x3
407− st1 {v21.8B}, [x0], x3
408− st1 {v22.8B}, [x0], x3
409− st1 {v23.8B}, [x0], x3
402+ st1 {v16.8b}, [x0], x3
403+ st1 {v17.8b}, [x0], x3
404+ st1 {v18.8b}, [x0], x3
405+ st1 {v19.8b}, [x0], x3
406+ st1 {v20.8b}, [x0], x3
407+ st1 {v21.8b}, [x0], x3
408+ st1 {v22.8b}, [x0], x3
409+ st1 {v23.8b}, [x0], x3
410410
411411 ret
412412 endfunc
@@ -417,19 +417,19 @@ endfunc · 4 unmodified lines
417417
418418 function put_h264_qpel8_hv_lowpass_neon_top
419419 lowpass_const w12
420− ld1 {v16.8H}, [x1], x3
421− ld1 {v17.8H}, [x1], x3
422− ld1 {v18.8H}, [x1], x3
423− ld1 {v19.8H}, [x1], x3
424− ld1 {v20.8H}, [x1], x3
425− ld1 {v21.8H}, [x1], x3
426− ld1 {v22.8H}, [x1], x3
427− ld1 {v23.8H}, [x1], x3
428− ld1 {v24.8H}, [x1], x3
429− ld1 {v25.8H}, [x1], x3
430− ld1 {v26.8H}, [x1], x3
431− ld1 {v27.8H}, [x1], x3
432− ld1 {v28.8H}, [x1]
420+ ld1 {v16.8h}, [x1], x3
421+ ld1 {v17.8h}, [x1], x3
422+ ld1 {v18.8h}, [x1], x3
423+ ld1 {v19.8h}, [x1], x3
424+ ld1 {v20.8h}, [x1], x3
425+ ld1 {v21.8h}, [x1], x3
426+ ld1 {v22.8h}, [x1], x3
427+ ld1 {v23.8h}, [x1], x3
428+ ld1 {v24.8h}, [x1], x3
429+ ld1 {v25.8h}, [x1], x3
430+ ld1 {v26.8h}, [x1], x3
431+ ld1 {v27.8h}, [x1], x3
432+ ld1 {v28.8h}, [x1]
433433 lowpass_8H v16, v17
434434 lowpass_8H v18, v19
435435 lowpass_8H v20, v21
@@ -458,33 +458,33 @@ function \type\()_h264_qpel8_hv_lowpass_neon · 22 unmodified lines
458458 mov x10, x30
459459 bl put_h264_qpel8_hv_lowpass_neon_top
460460 .ifc \type,avg
461− ld1 {v0.8B}, [x0], x2
462− ld1 {v1.8B}, [x0], x2
463− ld1 {v2.8B}, [x0], x2
464− urhadd v16.8B, v16.8B, v0.8B
465− ld1 {v3.8B}, [x0], x2
466− urhadd v17.8B, v17.8B, v1.8B
467− ld1 {v4.8B}, [x0], x2
468− urhadd v18.8B, v18.8B, v2.8B
469− ld1 {v5.8B}, [x0], x2
470− urhadd v19.8B, v19.8B, v3.8B
471− ld1 {v6.8B}, [x0], x2
472− urhadd v20.8B, v20.8B, v4.8B
473− ld1 {v7.8B}, [x0], x2
474− urhadd v21.8B, v21.8B, v5.8B
475− urhadd v22.8B, v22.8B, v6.8B
476− urhadd v23.8B, v23.8B, v7.8B
461+ ld1 {v0.8b}, [x0], x2
462+ ld1 {v1.8b}, [x0], x2
463+ ld1 {v2.8b}, [x0], x2
464+ urhadd v16.8b, v16.8b, v0.8b
465+ ld1 {v3.8b}, [x0], x2
466+ urhadd v17.8b, v17.8b, v1.8b
467+ ld1 {v4.8b}, [x0], x2
468+ urhadd v18.8b, v18.8b, v2.8b
469+ ld1 {v5.8b}, [x0], x2
470+ urhadd v19.8b, v19.8b, v3.8b
471+ ld1 {v6.8b}, [x0], x2
472+ urhadd v20.8b, v20.8b, v4.8b
473+ ld1 {v7.8b}, [x0], x2
474+ urhadd v21.8b, v21.8b, v5.8b
475+ urhadd v22.8b, v22.8b, v6.8b
476+ urhadd v23.8b, v23.8b, v7.8b
477477 sub x0, x0, x2, lsl #3
478478 .endif
479479
480− st1 {v16.8B}, [x0], x2
481− st1 {v17.8B}, [x0], x2
482− st1 {v18.8B}, [x0], x2
483− st1 {v19.8B}, [x0], x2
484− st1 {v20.8B}, [x0], x2
485− st1 {v21.8B}, [x0], x2
486− st1 {v22.8B}, [x0], x2
487− st1 {v23.8B}, [x0], x2
480+ st1 {v16.8b}, [x0], x2
481+ st1 {v17.8b}, [x0], x2
482+ st1 {v18.8b}, [x0], x2
483+ st1 {v19.8b}, [x0], x2
484+ st1 {v20.8b}, [x0], x2
485+ st1 {v21.8b}, [x0], x2
486+ st1 {v22.8b}, [x0], x2
487+ st1 {v23.8b}, [x0], x2
488488
489489 ret x10
490490 endfunc
@@ -498,45 +498,45 @@ function \type\()_h264_qpel8_hv_lowpass_l2_neon · 7 unmodified lines
498498 mov x10, x30
499499 bl put_h264_qpel8_hv_lowpass_neon_top
500500
501− ld1 {v0.8B, v1.8B}, [x2], #16
502− ld1 {v2.8B, v3.8B}, [x2], #16
503− urhadd v0.8B, v0.8B, v16.8B
504− urhadd v1.8B, v1.8B, v17.8B
505− ld1 {v4.8B, v5.8B}, [x2], #16
506− urhadd v2.8B, v2.8B, v18.8B
507− urhadd v3.8B, v3.8B, v19.8B
508− ld1 {v6.8B, v7.8B}, [x2], #16
509− urhadd v4.8B, v4.8B, v20.8B
510− urhadd v5.8B, v5.8B, v21.8B
511− urhadd v6.8B, v6.8B, v22.8B
512− urhadd v7.8B, v7.8B, v23.8B
501+ ld1 {v0.8b, v1.8b}, [x2], #16
502+ ld1 {v2.8b, v3.8b}, [x2], #16
503+ urhadd v0.8b, v0.8b, v16.8b
504+ urhadd v1.8b, v1.8b, v17.8b
505+ ld1 {v4.8b, v5.8b}, [x2], #16
506+ urhadd v2.8b, v2.8b, v18.8b
507+ urhadd v3.8b, v3.8b, v19.8b
508+ ld1 {v6.8b, v7.8b}, [x2], #16
509+ urhadd v4.8b, v4.8b, v20.8b
510+ urhadd v5.8b, v5.8b, v21.8b
511+ urhadd v6.8b, v6.8b, v22.8b
512+ urhadd v7.8b, v7.8b, v23.8b
513513 .ifc \type,avg
514− ld1 {v16.8B}, [x0], x3
515− ld1 {v17.8B}, [x0], x3
516− ld1 {v18.8B}, [x0], x3
517− urhadd v0.8B, v0.8B, v16.8B
518− ld1 {v19.8B}, [x0], x3
519− urhadd v1.8B, v1.8B, v17.8B
520− ld1 {v20.8B}, [x0], x3
521− urhadd v2.8B, v2.8B, v18.8B
522− ld1 {v21.8B}, [x0], x3
523− urhadd v3.8B, v3.8B, v19.8B
524− ld1 {v22.8B}, [x0], x3
525− urhadd v4.8B, v4.8B, v20.8B
526− ld1 {v23.8B}, [x0], x3
527− urhadd v5.8B, v5.8B, v21.8B
528− urhadd v6.8B, v6.8B, v22.8B
529− urhadd v7.8B, v7.8B, v23.8B
514+ ld1 {v16.8b}, [x0], x3
515+ ld1 {v17.8b}, [x0], x3
516+ ld1 {v18.8b}, [x0], x3
517+ urhadd v0.8b, v0.8b, v16.8b
518+ ld1 {v19.8b}, [x0], x3
519+ urhadd v1.8b, v1.8b, v17.8b
520+ ld1 {v20.8b}, [x0], x3
521+ urhadd v2.8b, v2.8b, v18.8b
522+ ld1 {v21.8b}, [x0], x3
523+ urhadd v3.8b, v3.8b, v19.8b
524+ ld1 {v22.8b}, [x0], x3
525+ urhadd v4.8b, v4.8b, v20.8b
526+ ld1 {v23.8b}, [x0], x3
527+ urhadd v5.8b, v5.8b, v21.8b
528+ urhadd v6.8b, v6.8b, v22.8b
529+ urhadd v7.8b, v7.8b, v23.8b
530530 sub x0, x0, x3, lsl #3
531531 .endif
532− st1 {v0.8B}, [x0], x3
533− st1 {v1.8B}, [x0], x3
534− st1 {v2.8B}, [x0], x3
535− st1 {v3.8B}, [x0], x3
536− st1 {v4.8B}, [x0], x3
537− st1 {v5.8B}, [x0], x3
538− st1 {v6.8B}, [x0], x3
539− st1 {v7.8B}, [x0], x3
532+ st1 {v0.8b}, [x0], x3
533+ st1 {v1.8b}, [x0], x3
534+ st1 {v2.8b}, [x0], x3
535+ st1 {v3.8b}, [x0], x3
536+ st1 {v4.8b}, [x0], x3
537+ st1 {v5.8b}, [x0], x3
538+ st1 {v6.8b}, [x0], x3
539+ st1 {v7.8b}, [x0], x3
540540
541541 ret x10
542542 endfunc
libavcodec/aarch64/hpeldsp_neon.S+181−181
@@ -26,295 +26,295 @@ · 25 unmodified lines
2626 .if \avg
2727 mov x12, x0
2828 .endif
29−1: ld1 {v0.16B}, [x1], x2
30− ld1 {v1.16B}, [x1], x2
31− ld1 {v2.16B}, [x1], x2
32− ld1 {v3.16B}, [x1], x2
29+1: ld1 {v0.16b}, [x1], x2
30+ ld1 {v1.16b}, [x1], x2
31+ ld1 {v2.16b}, [x1], x2
32+ ld1 {v3.16b}, [x1], x2
3333 .if \avg
34− ld1 {v4.16B}, [x12], x2
35− urhadd v0.16B, v0.16B, v4.16B
36− ld1 {v5.16B}, [x12], x2
37− urhadd v1.16B, v1.16B, v5.16B
38− ld1 {v6.16B}, [x12], x2
39− urhadd v2.16B, v2.16B, v6.16B
40− ld1 {v7.16B}, [x12], x2
41− urhadd v3.16B, v3.16B, v7.16B
34+ ld1 {v4.16b}, [x12], x2
35+ urhadd v0.16b, v0.16b, v4.16b
36+ ld1 {v5.16b}, [x12], x2
37+ urhadd v1.16b, v1.16b, v5.16b
38+ ld1 {v6.16b}, [x12], x2
39+ urhadd v2.16b, v2.16b, v6.16b
40+ ld1 {v7.16b}, [x12], x2
41+ urhadd v3.16b, v3.16b, v7.16b
4242 .endif
4343 subs w3, w3, #4
44− st1 {v0.16B}, [x0], x2
45− st1 {v1.16B}, [x0], x2
46− st1 {v2.16B}, [x0], x2
47− st1 {v3.16B}, [x0], x2
44+ st1 {v0.16b}, [x0], x2
45+ st1 {v1.16b}, [x0], x2
46+ st1 {v2.16b}, [x0], x2
47+ st1 {v3.16b}, [x0], x2
4848 b.ne 1b
4949 ret
5050 .endm
5151
5252 .macro pixels16_x2 rnd=1, avg=0
53−1: ld1 {v0.16B, v1.16B}, [x1], x2
54− ld1 {v2.16B, v3.16B}, [x1], x2
53+1: ld1 {v0.16b, v1.16b}, [x1], x2
54+ ld1 {v2.16b, v3.16b}, [x1], x2
5555 subs w3, w3, #2
56− ext v1.16B, v0.16B, v1.16B, #1
57− avg v0.16B, v0.16B, v1.16B
58− ext v3.16B, v2.16B, v3.16B, #1
59− avg v2.16B, v2.16B, v3.16B
56+ ext v1.16b, v0.16b, v1.16b, #1
57+ avg v0.16b, v0.16b, v1.16b
58+ ext v3.16b, v2.16b, v3.16b, #1
59+ avg v2.16b, v2.16b, v3.16b
6060 .if \avg
61− ld1 {v1.16B}, [x0], x2
62− ld1 {v3.16B}, [x0]
63− urhadd v0.16B, v0.16B, v1.16B
64− urhadd v2.16B, v2.16B, v3.16B
61+ ld1 {v1.16b}, [x0], x2
62+ ld1 {v3.16b}, [x0]
63+ urhadd v0.16b, v0.16b, v1.16b
64+ urhadd v2.16b, v2.16b, v3.16b
6565 sub x0, x0, x2
6666 .endif
67− st1 {v0.16B}, [x0], x2
68− st1 {v2.16B}, [x0], x2
67+ st1 {v0.16b}, [x0], x2
68+ st1 {v2.16b}, [x0], x2
6969 b.ne 1b
7070 ret
7171 .endm
7272
7373 .macro pixels16_y2 rnd=1, avg=0
7474 sub w3, w3, #2
75− ld1 {v0.16B}, [x1], x2
76− ld1 {v1.16B}, [x1], x2
75+ ld1 {v0.16b}, [x1], x2
76+ ld1 {v1.16b}, [x1], x2
7777 1: subs w3, w3, #2
78− avg v2.16B, v0.16B, v1.16B
79− ld1 {v0.16B}, [x1], x2
80− avg v3.16B, v0.16B, v1.16B
81− ld1 {v1.16B}, [x1], x2
78+ avg v2.16b, v0.16b, v1.16b
79+ ld1 {v0.16b}, [x1], x2
80+ avg v3.16b, v0.16b, v1.16b
81+ ld1 {v1.16b}, [x1], x2
8282 .if \avg
83− ld1 {v4.16B}, [x0], x2
84− ld1 {v5.16B}, [x0]
85− urhadd v2.16B, v2.16B, v4.16B
86− urhadd v3.16B, v3.16B, v5.16B
83+ ld1 {v4.16b}, [x0], x2
84+ ld1 {v5.16b}, [x0]
85+ urhadd v2.16b, v2.16b, v4.16b
86+ urhadd v3.16b, v3.16b, v5.16b
8787 sub x0, x0, x2
8888 .endif
89− st1 {v2.16B}, [x0], x2
90− st1 {v3.16B}, [x0], x2
89+ st1 {v2.16b}, [x0], x2
90+ st1 {v3.16b}, [x0], x2
9191 b.ne 1b
9292
93− avg v2.16B, v0.16B, v1.16B
94− ld1 {v0.16B}, [x1], x2
95− avg v3.16B, v0.16B, v1.16B
93+ avg v2.16b, v0.16b, v1.16b
94+ ld1 {v0.16b}, [x1], x2
95+ avg v3.16b, v0.16b, v1.16b
9696 .if \avg
97− ld1 {v4.16B}, [x0], x2
98− ld1 {v5.16B}, [x0]
99− urhadd v2.16B, v2.16B, v4.16B
100− urhadd v3.16B, v3.16B, v5.16B
97+ ld1 {v4.16b}, [x0], x2
98+ ld1 {v5.16b}, [x0]
99+ urhadd v2.16b, v2.16b, v4.16b
100+ urhadd v3.16b, v3.16b, v5.16b
101101 sub x0, x0, x2
102102 .endif
103− st1 {v2.16B}, [x0], x2
104− st1 {v3.16B}, [x0], x2
103+ st1 {v2.16b}, [x0], x2
104+ st1 {v3.16b}, [x0], x2
105105
106106 ret
107107 .endm
108108
109109 .macro pixels16_xy2 rnd=1, avg=0
110110 sub w3, w3, #2
111− ld1 {v0.16B, v1.16B}, [x1], x2
112− ld1 {v4.16B, v5.16B}, [x1], x2
111+ ld1 {v0.16b, v1.16b}, [x1], x2
112+ ld1 {v4.16b, v5.16b}, [x1], x2
113113 NRND movi v26.8H, #1
114− ext v1.16B, v0.16B, v1.16B, #1
115− ext v5.16B, v4.16B, v5.16B, #1
116− uaddl v16.8H, v0.8B, v1.8B
117− uaddl2 v20.8H, v0.16B, v1.16B
118− uaddl v18.8H, v4.8B, v5.8B
119− uaddl2 v22.8H, v4.16B, v5.16B
114+ ext v1.16b, v0.16b, v1.16b, #1
115+ ext v5.16b, v4.16b, v5.16b, #1
116+ uaddl v16.8h, v0.8b, v1.8b
117+ uaddl2 v20.8h, v0.16b, v1.16b
118+ uaddl v18.8h, v4.8b, v5.8b
119+ uaddl2 v22.8h, v4.16b, v5.16b
120120 1: subs w3, w3, #2
121− ld1 {v0.16B, v1.16B}, [x1], x2
122− add v24.8H, v16.8H, v18.8H
121+ ld1 {v0.16b, v1.16b}, [x1], x2
122+ add v24.8h, v16.8h, v18.8h
123123 NRND add v24.8H, v24.8H, v26.8H
124− ext v30.16B, v0.16B, v1.16B, #1
125− add v1.8H, v20.8H, v22.8H
126− mshrn v28.8B, v24.8H, #2
124+ ext v30.16b, v0.16b, v1.16b, #1
125+ add v1.8h, v20.8h, v22.8h
126+ mshrn v28.8b, v24.8h, #2
127127 NRND add v1.8H, v1.8H, v26.8H
128− mshrn2 v28.16B, v1.8H, #2
128+ mshrn2 v28.16b, v1.8h, #2
129129 .if \avg
130− ld1 {v16.16B}, [x0]
131− urhadd v28.16B, v28.16B, v16.16B
130+ ld1 {v16.16b}, [x0]
131+ urhadd v28.16b, v28.16b, v16.16b
132132 .endif
133− uaddl v16.8H, v0.8B, v30.8B
134− ld1 {v2.16B, v3.16B}, [x1], x2
135− uaddl2 v20.8H, v0.16B, v30.16B
136− st1 {v28.16B}, [x0], x2
137− add v24.8H, v16.8H, v18.8H
133+ uaddl v16.8h, v0.8b, v30.8b
134+ ld1 {v2.16b, v3.16b}, [x1], x2
135+ uaddl2 v20.8h, v0.16b, v30.16b
136+ st1 {v28.16b}, [x0], x2
137+ add v24.8h, v16.8h, v18.8h
138138 NRND add v24.8H, v24.8H, v26.8H
139− ext v3.16B, v2.16B, v3.16B, #1
140− add v0.8H, v20.8H, v22.8H
141− mshrn v30.8B, v24.8H, #2
139+ ext v3.16b, v2.16b, v3.16b, #1
140+ add v0.8h, v20.8h, v22.8h
141+ mshrn v30.8b, v24.8h, #2
142142 NRND add v0.8H, v0.8H, v26.8H
143− mshrn2 v30.16B, v0.8H, #2
143+ mshrn2 v30.16b, v0.8h, #2
144144 .if \avg
145− ld1 {v18.16B}, [x0]
146− urhadd v30.16B, v30.16B, v18.16B
145+ ld1 {v18.16b}, [x0]
146+ urhadd v30.16b, v30.16b, v18.16b
147147 .endif
148− uaddl v18.8H, v2.8B, v3.8B
149− uaddl2 v22.8H, v2.16B, v3.16B
150− st1 {v30.16B}, [x0], x2
148+ uaddl v18.8h, v2.8b, v3.8b
149+ uaddl2 v22.8h, v2.16b, v3.16b
150+ st1 {v30.16b}, [x0], x2
151151 b.gt 1b
152152
153− ld1 {v0.16B, v1.16B}, [x1], x2
154− add v24.8H, v16.8H, v18.8H
153+ ld1 {v0.16b, v1.16b}, [x1], x2
154+ add v24.8h, v16.8h, v18.8h
155155 NRND add v24.8H, v24.8H, v26.8H
156− ext v30.16B, v0.16B, v1.16B, #1
157− add v1.8H, v20.8H, v22.8H
158− mshrn v28.8B, v24.8H, #2
156+ ext v30.16b, v0.16b, v1.16b, #1
157+ add v1.8h, v20.8h, v22.8h
158+ mshrn v28.8b, v24.8h, #2
159159 NRND add v1.8H, v1.8H, v26.8H
160− mshrn2 v28.16B, v1.8H, #2
160+ mshrn2 v28.16b, v1.8h, #2
161161 .if \avg
162− ld1 {v16.16B}, [x0]
163− urhadd v28.16B, v28.16B, v16.16B
162+ ld1 {v16.16b}, [x0]
163+ urhadd v28.16b, v28.16b, v16.16b
164164 .endif
165− uaddl v16.8H, v0.8B, v30.8B
166− uaddl2 v20.8H, v0.16B, v30.16B
167− st1 {v28.16B}, [x0], x2
168− add v24.8H, v16.8H, v18.8H
165+ uaddl v16.8h, v0.8b, v30.8b
166+ uaddl2 v20.8h, v0.16b, v30.16b
167+ st1 {v28.16b}, [x0], x2
168+ add v24.8h, v16.8h, v18.8h
169169 NRND add v24.8H, v24.8H, v26.8H
170− add v0.8H, v20.8H, v22.8H
171− mshrn v30.8B, v24.8H, #2
170+ add v0.8h, v20.8h, v22.8h
171+ mshrn v30.8b, v24.8h, #2
172172 NRND add v0.8H, v0.8H, v26.8H
173− mshrn2 v30.16B, v0.8H, #2
173+ mshrn2 v30.16b, v0.8h, #2
174174 .if \avg
175− ld1 {v18.16B}, [x0]
176− urhadd v30.16B, v30.16B, v18.16B
175+ ld1 {v18.16b}, [x0]
176+ urhadd v30.16b, v30.16b, v18.16b
177177 .endif
178− st1 {v30.16B}, [x0], x2
178+ st1 {v30.16b}, [x0], x2
179179
180180 ret
181181 .endm
182182
183183 .macro pixels8 rnd=1, avg=0
184−1: ld1 {v0.8B}, [x1], x2
185− ld1 {v1.8B}, [x1], x2
186− ld1 {v2.8B}, [x1], x2
187− ld1 {v3.8B}, [x1], x2
184+1: ld1 {v0.8b}, [x1], x2
185+ ld1 {v1.8b}, [x1], x2
186+ ld1 {v2.8b}, [x1], x2
187+ ld1 {v3.8b}, [x1], x2
188188 .if \avg
189− ld1 {v4.8B}, [x0], x2
190− urhadd v0.8B, v0.8B, v4.8B
191− ld1 {v5.8B}, [x0], x2
192− urhadd v1.8B, v1.8B, v5.8B
193− ld1 {v6.8B}, [x0], x2
194− urhadd v2.8B, v2.8B, v6.8B
195− ld1 {v7.8B}, [x0], x2
196− urhadd v3.8B, v3.8B, v7.8B
189+ ld1 {v4.8b}, [x0], x2
190+ urhadd v0.8b, v0.8b, v4.8b
191+ ld1 {v5.8b}, [x0], x2
192+ urhadd v1.8b, v1.8b, v5.8b
193+ ld1 {v6.8b}, [x0], x2
194+ urhadd v2.8b, v2.8b, v6.8b
195+ ld1 {v7.8b}, [x0], x2
196+ urhadd v3.8b, v3.8b, v7.8b
197197 sub x0, x0, x2, lsl #2
198198 .endif
199199 subs w3, w3, #4
200− st1 {v0.8B}, [x0], x2
201− st1 {v1.8B}, [x0], x2
202− st1 {v2.8B}, [x0], x2
203− st1 {v3.8B}, [x0], x2
200+ st1 {v0.8b}, [x0], x2
201+ st1 {v1.8b}, [x0], x2
202+ st1 {v2.8b}, [x0], x2
203+ st1 {v3.8b}, [x0], x2
204204 b.ne 1b
205205 ret
206206 .endm
207207
208208 .macro pixels8_x2 rnd=1, avg=0
209−1: ld1 {v0.8B, v1.8B}, [x1], x2
210− ext v1.8B, v0.8B, v1.8B, #1
211− ld1 {v2.8B, v3.8B}, [x1], x2
212− ext v3.8B, v2.8B, v3.8B, #1
209+1: ld1 {v0.8b, v1.8b}, [x1], x2
210+ ext v1.8b, v0.8b, v1.8b, #1
211+ ld1 {v2.8b, v3.8b}, [x1], x2
212+ ext v3.8b, v2.8b, v3.8b, #1
213213 subs w3, w3, #2
214− avg v0.8B, v0.8B, v1.8B
215− avg v2.8B, v2.8B, v3.8B
214+ avg v0.8b, v0.8b, v1.8b
215+ avg v2.8b, v2.8b, v3.8b
216216 .if \avg
217− ld1 {v4.8B}, [x0], x2
218− ld1 {v5.8B}, [x0]
219− urhadd v0.8B, v0.8B, v4.8B
220− urhadd v2.8B, v2.8B, v5.8B
217+ ld1 {v4.8b}, [x0], x2
218+ ld1 {v5.8b}, [x0]
219+ urhadd v0.8b, v0.8b, v4.8b
220+ urhadd v2.8b, v2.8b, v5.8b
221221 sub x0, x0, x2
222222 .endif
223− st1 {v0.8B}, [x0], x2
224− st1 {v2.8B}, [x0], x2
223+ st1 {v0.8b}, [x0], x2
224+ st1 {v2.8b}, [x0], x2
225225 b.ne 1b
226226 ret
227227 .endm
228228
229229 .macro pixels8_y2 rnd=1, avg=0
230230 sub w3, w3, #2
231− ld1 {v0.8B}, [x1], x2
232− ld1 {v1.8B}, [x1], x2
231+ ld1 {v0.8b}, [x1], x2
232+ ld1 {v1.8b}, [x1], x2
233233 1: subs w3, w3, #2
234− avg v4.8B, v0.8B, v1.8B
235− ld1 {v0.8B}, [x1], x2
236− avg v5.8B, v0.8B, v1.8B
237− ld1 {v1.8B}, [x1], x2
234+ avg v4.8b, v0.8b, v1.8b
235+ ld1 {v0.8b}, [x1], x2
236+ avg v5.8b, v0.8b, v1.8b
237+ ld1 {v1.8b}, [x1], x2
238238 .if \avg
239− ld1 {v2.8B}, [x0], x2
240− ld1 {v3.8B}, [x0]
241− urhadd v4.8B, v4.8B, v2.8B
242− urhadd v5.8B, v5.8B, v3.8B
239+ ld1 {v2.8b}, [x0], x2
240+ ld1 {v3.8b}, [x0]
241+ urhadd v4.8b, v4.8b, v2.8b
242+ urhadd v5.8b, v5.8b, v3.8b
243243 sub x0, x0, x2
244244 .endif
245− st1 {v4.8B}, [x0], x2
246− st1 {v5.8B}, [x0], x2
245+ st1 {v4.8b}, [x0], x2
246+ st1 {v5.8b}, [x0], x2
247247 b.ne 1b
248248
249− avg v4.8B, v0.8B, v1.8B
250− ld1 {v0.8B}, [x1], x2
251− avg v5.8B, v0.8B, v1.8B
249+ avg v4.8b, v0.8b, v1.8b
250+ ld1 {v0.8b}, [x1], x2
251+ avg v5.8b, v0.8b, v1.8b
252252 .if \avg
253− ld1 {v2.8B}, [x0], x2
254− ld1 {v3.8B}, [x0]
255− urhadd v4.8B, v4.8B, v2.8B
256− urhadd v5.8B, v5.8B, v3.8B
253+ ld1 {v2.8b}, [x0], x2
254+ ld1 {v3.8b}, [x0]
255+ urhadd v4.8b, v4.8b, v2.8b
256+ urhadd v5.8b, v5.8b, v3.8b
257257 sub x0, x0, x2
258258 .endif
259− st1 {v4.8B}, [x0], x2
260− st1 {v5.8B}, [x0], x2
259+ st1 {v4.8b}, [x0], x2
260+ st1 {v5.8b}, [x0], x2
261261
262262 ret
263263 .endm
264264
265265 .macro pixels8_xy2 rnd=1, avg=0
266266 sub w3, w3, #2
267− ld1 {v0.16B}, [x1], x2
268− ld1 {v1.16B}, [x1], x2
267+ ld1 {v0.16b}, [x1], x2
268+ ld1 {v1.16b}, [x1], x2
269269 NRND movi v19.8H, #1
270− ext v4.16B, v0.16B, v4.16B, #1
271− ext v6.16B, v1.16B, v6.16B, #1
272− uaddl v16.8H, v0.8B, v4.8B
273− uaddl v17.8H, v1.8B, v6.8B
270+ ext v4.16b, v0.16b, v4.16b, #1
271+ ext v6.16b, v1.16b, v6.16b, #1
272+ uaddl v16.8h, v0.8b, v4.8b
273+ uaddl v17.8h, v1.8b, v6.8b
274274 1: subs w3, w3, #2
275− ld1 {v0.16B}, [x1], x2
276− add v18.8H, v16.8H, v17.8H
277− ext v4.16B, v0.16B, v4.16B, #1
275+ ld1 {v0.16b}, [x1], x2
276+ add v18.8h, v16.8h, v17.8h
277+ ext v4.16b, v0.16b, v4.16b, #1
278278 NRND add v18.8H, v18.8H, v19.8H
279− uaddl v16.8H, v0.8B, v4.8B
280− mshrn v5.8B, v18.8H, #2
281− ld1 {v1.16B}, [x1], x2
282− add v18.8H, v16.8H, v17.8H
279+ uaddl v16.8h, v0.8b, v4.8b
280+ mshrn v5.8b, v18.8h, #2
281+ ld1 {v1.16b}, [x1], x2
282+ add v18.8h, v16.8h, v17.8h
283283 .if \avg
284− ld1 {v7.8B}, [x0]
285− urhadd v5.8B, v5.8B, v7.8B
284+ ld1 {v7.8b}, [x0]
285+ urhadd v5.8b, v5.8b, v7.8b
286286 .endif
287287 NRND add v18.8H, v18.8H, v19.8H
288− st1 {v5.8B}, [x0], x2
289− mshrn v7.8B, v18.8H, #2
288+ st1 {v5.8b}, [x0], x2
289+ mshrn v7.8b, v18.8h, #2
290290 .if \avg
291− ld1 {v5.8B}, [x0]
292− urhadd v7.8B, v7.8B, v5.8B
291+ ld1 {v5.8b}, [x0]
292+ urhadd v7.8b, v7.8b, v5.8b
293293 .endif
294− ext v6.16B, v1.16B, v6.16B, #1
295− uaddl v17.8H, v1.8B, v6.8B
296− st1 {v7.8B}, [x0], x2
294+ ext v6.16b, v1.16b, v6.16b, #1
295+ uaddl v17.8h, v1.8b, v6.8b
296+ st1 {v7.8b}, [x0], x2
297297 b.gt 1b
298298
299− ld1 {v0.16B}, [x1], x2
300− add v18.8H, v16.8H, v17.8H
301− ext v4.16B, v0.16B, v4.16B, #1
299+ ld1 {v0.16b}, [x1], x2
300+ add v18.8h, v16.8h, v17.8h
301+ ext v4.16b, v0.16b, v4.16b, #1
302302 NRND add v18.8H, v18.8H, v19.8H
303− uaddl v16.8H, v0.8B, v4.8B
304− mshrn v5.8B, v18.8H, #2
305− add v18.8H, v16.8H, v17.8H
303+ uaddl v16.8h, v0.8b, v4.8b
304+ mshrn v5.8b, v18.8h, #2
305+ add v18.8h, v16.8h, v17.8h
306306 .if \avg
307− ld1 {v7.8B}, [x0]
308− urhadd v5.8B, v5.8B, v7.8B
307+ ld1 {v7.8b}, [x0]
308+ urhadd v5.8b, v5.8b, v7.8b
309309 .endif
310310 NRND add v18.8H, v18.8H, v19.8H
311− st1 {v5.8B}, [x0], x2
312− mshrn v7.8B, v18.8H, #2
311+ st1 {v5.8b}, [x0], x2
312+ mshrn v7.8b, v18.8h, #2
313313 .if \avg
314− ld1 {v5.8B}, [x0]
315− urhadd v7.8B, v7.8B, v5.8B
314+ ld1 {v5.8b}, [x0]
315+ urhadd v7.8b, v7.8b, v5.8b
316316 .endif
317− st1 {v7.8B}, [x0], x2
317+ st1 {v7.8b}, [x0], x2
318318
319319 ret
320320 .endm
libavcodec/aarch64/me_cmp_neon.S+1−1
@@ -1099,7 +1099,7 @@ function vsse_intra16_neon, export=1 · 1098 unmodified lines
10991099 cbnz w4, 2b
11001100
11011101 3:
1102− add v16.4s, v16.4s, v17.4S
1102+ add v16.4s, v16.4s, v17.4s
11031103 uaddlv d17, v16.4s
11041104 fmov w0, s17
11051105
libavcodec/aarch64/neon.S+123−123
@@ -28,146 +28,146 @@ · 27 unmodified lines
2828 .endm
2929
3030 .macro transpose_8x8B r0, r1, r2, r3, r4, r5, r6, r7, r8, r9
31− trn1 \r8\().8B, \r0\().8B, \r1\().8B
32− trn2 \r9\().8B, \r0\().8B, \r1\().8B
33− trn1 \r1\().8B, \r2\().8B, \r3\().8B
34− trn2 \r3\().8B, \r2\().8B, \r3\().8B
35− trn1 \r0\().8B, \r4\().8B, \r5\().8B
36− trn2 \r5\().8B, \r4\().8B, \r5\().8B
37− trn1 \r2\().8B, \r6\().8B, \r7\().8B
38− trn2 \r7\().8B, \r6\().8B, \r7\().8B
39−
40− trn1 \r4\().4H, \r0\().4H, \r2\().4H
41− trn2 \r2\().4H, \r0\().4H, \r2\().4H
42− trn1 \r6\().4H, \r5\().4H, \r7\().4H
43− trn2 \r7\().4H, \r5\().4H, \r7\().4H
44− trn1 \r5\().4H, \r9\().4H, \r3\().4H
45− trn2 \r9\().4H, \r9\().4H, \r3\().4H
46− trn1 \r3\().4H, \r8\().4H, \r1\().4H
47− trn2 \r8\().4H, \r8\().4H, \r1\().4H
48−
49− trn1 \r0\().2S, \r3\().2S, \r4\().2S
50− trn2 \r4\().2S, \r3\().2S, \r4\().2S
51−
52− trn1 \r1\().2S, \r5\().2S, \r6\().2S
53− trn2 \r5\().2S, \r5\().2S, \r6\().2S
54−
55− trn2 \r6\().2S, \r8\().2S, \r2\().2S
56− trn1 \r2\().2S, \r8\().2S, \r2\().2S
57−
58− trn1 \r3\().2S, \r9\().2S, \r7\().2S
59− trn2 \r7\().2S, \r9\().2S, \r7\().2S
31+ trn1 \r8\().8b, \r0\().8b, \r1\().8b
32+ trn2 \r9\().8b, \r0\().8b, \r1\().8b
33+ trn1 \r1\().8b, \r2\().8b, \r3\().8b
34+ trn2 \r3\().8b, \r2\().8b, \r3\().8b
35+ trn1 \r0\().8b, \r4\().8b, \r5\().8b
36+ trn2 \r5\().8b, \r4\().8b, \r5\().8b
37+ trn1 \r2\().8b, \r6\().8b, \r7\().8b
38+ trn2 \r7\().8b, \r6\().8b, \r7\().8b
39+
40+ trn1 \r4\().4h, \r0\().4h, \r2\().4h
41+ trn2 \r2\().4h, \r0\().4h, \r2\().4h
42+ trn1 \r6\().4h, \r5\().4h, \r7\().4h
43+ trn2 \r7\().4h, \r5\().4h, \r7\().4h
44+ trn1 \r5\().4h, \r9\().4h, \r3\().4h
45+ trn2 \r9\().4h, \r9\().4h, \r3\().4h
46+ trn1 \r3\().4h, \r8\().4h, \r1\().4h
47+ trn2 \r8\().4h, \r8\().4h, \r1\().4h
48+
49+ trn1 \r0\().2s, \r3\().2s, \r4\().2s
50+ trn2 \r4\().2s, \r3\().2s, \r4\().2s
51+
52+ trn1 \r1\().2s, \r5\().2s, \r6\().2s
53+ trn2 \r5\().2s, \r5\().2s, \r6\().2s
54+
55+ trn2 \r6\().2s, \r8\().2s, \r2\().2s
56+ trn1 \r2\().2s, \r8\().2s, \r2\().2s
57+
58+ trn1 \r3\().2s, \r9\().2s, \r7\().2s
59+ trn2 \r7\().2s, \r9\().2s, \r7\().2s
6060 .endm
6161
6262 .macro transpose_8x16B r0, r1, r2, r3, r4, r5, r6, r7, t0, t1
63− trn1 \t0\().16B, \r0\().16B, \r1\().16B
64− trn2 \t1\().16B, \r0\().16B, \r1\().16B
65− trn1 \r1\().16B, \r2\().16B, \r3\().16B
66− trn2 \r3\().16B, \r2\().16B, \r3\().16B
67− trn1 \r0\().16B, \r4\().16B, \r5\().16B
68− trn2 \r5\().16B, \r4\().16B, \r5\().16B
69− trn1 \r2\().16B, \r6\().16B, \r7\().16B
70− trn2 \r7\().16B, \r6\().16B, \r7\().16B
71−
72− trn1 \r4\().8H, \r0\().8H, \r2\().8H
73− trn2 \r2\().8H, \r0\().8H, \r2\().8H
74− trn1 \r6\().8H, \r5\().8H, \r7\().8H
75− trn2 \r7\().8H, \r5\().8H, \r7\().8H
76− trn1 \r5\().8H, \t1\().8H, \r3\().8H
77− trn2 \t1\().8H, \t1\().8H, \r3\().8H
78− trn1 \r3\().8H, \t0\().8H, \r1\().8H
79− trn2 \t0\().8H, \t0\().8H, \r1\().8H
80−
81− trn1 \r0\().4S, \r3\().4S, \r4\().4S
82− trn2 \r4\().4S, \r3\().4S, \r4\().4S
83−
84− trn1 \r1\().4S, \r5\().4S, \r6\().4S
85− trn2 \r5\().4S, \r5\().4S, \r6\().4S
86−
87− trn2 \r6\().4S, \t0\().4S, \r2\().4S
88− trn1 \r2\().4S, \t0\().4S, \r2\().4S
89−
90− trn1 \r3\().4S, \t1\().4S, \r7\().4S
91− trn2 \r7\().4S, \t1\().4S, \r7\().4S
63+ trn1 \t0\().16b, \r0\().16b, \r1\().16b
64+ trn2 \t1\().16b, \r0\().16b, \r1\().16b
65+ trn1 \r1\().16b, \r2\().16b, \r3\().16b
66+ trn2 \r3\().16b, \r2\().16b, \r3\().16b
67+ trn1 \r0\().16b, \r4\().16b, \r5\().16b
68+ trn2 \r5\().16b, \r4\().16b, \r5\().16b
69+ trn1 \r2\().16b, \r6\().16b, \r7\().16b
70+ trn2 \r7\().16b, \r6\().16b, \r7\().16b
71+
72+ trn1 \r4\().8h, \r0\().8h, \r2\().8h
73+ trn2 \r2\().8h, \r0\().8h, \r2\().8h
74+ trn1 \r6\().8h, \r5\().8h, \r7\().8h
75+ trn2 \r7\().8h, \r5\().8h, \r7\().8h
76+ trn1 \r5\().8h, \t1\().8h, \r3\().8h
77+ trn2 \t1\().8h, \t1\().8h, \r3\().8h
78+ trn1 \r3\().8h, \t0\().8h, \r1\().8h
79+ trn2 \t0\().8h, \t0\().8h, \r1\().8h
80+
81+ trn1 \r0\().4s, \r3\().4s, \r4\().4s
82+ trn2 \r4\().4s, \r3\().4s, \r4\().4s
83+
84+ trn1 \r1\().4s, \r5\().4s, \r6\().4s
85+ trn2 \r5\().4s, \r5\().4s, \r6\().4s
86+
87+ trn2 \r6\().4s, \t0\().4s, \r2\().4s
88+ trn1 \r2\().4s, \t0\().4s, \r2\().4s
89+
90+ trn1 \r3\().4s, \t1\().4s, \r7\().4s
91+ trn2 \r7\().4s, \t1\().4s, \r7\().4s
9292 .endm
9393
9494 .macro transpose_4x16B r0, r1, r2, r3, t4, t5, t6, t7
95− trn1 \t4\().16B, \r0\().16B, \r1\().16B
96− trn2 \t5\().16B, \r0\().16B, \r1\().16B
97− trn1 \t6\().16B, \r2\().16B, \r3\().16B
98− trn2 \t7\().16B, \r2\().16B, \r3\().16B
99−
100− trn1 \r0\().8H, \t4\().8H, \t6\().8H
101− trn2 \r2\().8H, \t4\().8H, \t6\().8H
102− trn1 \r1\().8H, \t5\().8H, \t7\().8H
103− trn2 \r3\().8H, \t5\().8H, \t7\().8H
95+ trn1 \t4\().16b, \r0\().16b, \r1\().16b
96+ trn2 \t5\().16b, \r0\().16b, \r1\().16b
97+ trn1 \t6\().16b, \r2\().16b, \r3\().16b
98+ trn2 \t7\().16b, \r2\().16b, \r3\().16b
99+
100+ trn1 \r0\().8h, \t4\().8h, \t6\().8h
101+ trn2 \r2\().8h, \t4\().8h, \t6\().8h
102+ trn1 \r1\().8h, \t5\().8h, \t7\().8h
103+ trn2 \r3\().8h, \t5\().8h, \t7\().8h
104104 .endm
105105
106106 .macro transpose_4x8B r0, r1, r2, r3, t4, t5, t6, t7
107− trn1 \t4\().8B, \r0\().8B, \r1\().8B
108− trn2 \t5\().8B, \r0\().8B, \r1\().8B
109− trn1 \t6\().8B, \r2\().8B, \r3\().8B
110− trn2 \t7\().8B, \r2\().8B, \r3\().8B
111−
112− trn1 \r0\().4H, \t4\().4H, \t6\().4H
113− trn2 \r2\().4H, \t4\().4H, \t6\().4H
114− trn1 \r1\().4H, \t5\().4H, \t7\().4H
115− trn2 \r3\().4H, \t5\().4H, \t7\().4H
107+ trn1 \t4\().8b, \r0\().8b, \r1\().8b
108+ trn2 \t5\().8b, \r0\().8b, \r1\().8b
109+ trn1 \t6\().8b, \r2\().8b, \r3\().8b
110+ trn2 \t7\().8b, \r2\().8b, \r3\().8b
111+
112+ trn1 \r0\().4h, \t4\().4h, \t6\().4h
113+ trn2 \r2\().4h, \t4\().4h, \t6\().4h
114+ trn1 \r1\().4h, \t5\().4h, \t7\().4h
115+ trn2 \r3\().4h, \t5\().4h, \t7\().4h
116116 .endm
117117
118118 .macro transpose_4x4H r0, r1, r2, r3, r4, r5, r6, r7
119− trn1 \r4\().4H, \r0\().4H, \r1\().4H
120− trn2 \r5\().4H, \r0\().4H, \r1\().4H
121− trn1 \r6\().4H, \r2\().4H, \r3\().4H
122− trn2 \r7\().4H, \r2\().4H, \r3\().4H
123−
124− trn1 \r0\().2S, \r4\().2S, \r6\().2S
125− trn2 \r2\().2S, \r4\().2S, \r6\().2S
126− trn1 \r1\().2S, \r5\().2S, \r7\().2S
127− trn2 \r3\().2S, \r5\().2S, \r7\().2S
119+ trn1 \r4\().4h, \r0\().4h, \r1\().4h
120+ trn2 \r5\().4h, \r0\().4h, \r1\().4h
121+ trn1 \r6\().4h, \r2\().4h, \r3\().4h
122+ trn2 \r7\().4h, \r2\().4h, \r3\().4h
123+
124+ trn1 \r0\().2s, \r4\().2s, \r6\().2s
125+ trn2 \r2\().2s, \r4\().2s, \r6\().2s
126+ trn1 \r1\().2s, \r5\().2s, \r7\().2s
127+ trn2 \r3\().2s, \r5\().2s, \r7\().2s
128128 .endm
129129
130130 .macro transpose_4x8H r0, r1, r2, r3, t4, t5, t6, t7
131− trn1 \t4\().8H, \r0\().8H, \r1\().8H
132− trn2 \t5\().8H, \r0\().8H, \r1\().8H
133− trn1 \t6\().8H, \r2\().8H, \r3\().8H
134− trn2 \t7\().8H, \r2\().8H, \r3\().8H
135−
136− trn1 \r0\().4S, \t4\().4S, \t6\().4S
137− trn2 \r2\().4S, \t4\().4S, \t6\().4S
138− trn1 \r1\().4S, \t5\().4S, \t7\().4S
139− trn2 \r3\().4S, \t5\().4S, \t7\().4S
131+ trn1 \t4\().8h, \r0\().8h, \r1\().8h
132+ trn2 \t5\().8h, \r0\().8h, \r1\().8h
133+ trn1 \t6\().8h, \r2\().8h, \r3\().8h
134+ trn2 \t7\().8h, \r2\().8h, \r3\().8h
135+
136+ trn1 \r0\().4s, \t4\().4s, \t6\().4s
137+ trn2 \r2\().4s, \t4\().4s, \t6\().4s
138+ trn1 \r1\().4s, \t5\().4s, \t7\().4s
139+ trn2 \r3\().4s, \t5\().4s, \t7\().4s
140140 .endm
141141
142142 .macro transpose_8x8H r0, r1, r2, r3, r4, r5, r6, r7, r8, r9
143− trn1 \r8\().8H, \r0\().8H, \r1\().8H
144− trn2 \r9\().8H, \r0\().8H, \r1\().8H
145− trn1 \r1\().8H, \r2\().8H, \r3\().8H
146− trn2 \r3\().8H, \r2\().8H, \r3\().8H
147− trn1 \r0\().8H, \r4\().8H, \r5\().8H
148− trn2 \r5\().8H, \r4\().8H, \r5\().8H
149− trn1 \r2\().8H, \r6\().8H, \r7\().8H
150− trn2 \r7\().8H, \r6\().8H, \r7\().8H
151−
152− trn1 \r4\().4S, \r0\().4S, \r2\().4S
153− trn2 \r2\().4S, \r0\().4S, \r2\().4S
154− trn1 \r6\().4S, \r5\().4S, \r7\().4S
155− trn2 \r7\().4S, \r5\().4S, \r7\().4S
156− trn1 \r5\().4S, \r9\().4S, \r3\().4S
157− trn2 \r9\().4S, \r9\().4S, \r3\().4S
158− trn1 \r3\().4S, \r8\().4S, \r1\().4S
159− trn2 \r8\().4S, \r8\().4S, \r1\().4S
160−
161− trn1 \r0\().2D, \r3\().2D, \r4\().2D
162− trn2 \r4\().2D, \r3\().2D, \r4\().2D
163−
164− trn1 \r1\().2D, \r5\().2D, \r6\().2D
165− trn2 \r5\().2D, \r5\().2D, \r6\().2D
166−
167− trn2 \r6\().2D, \r8\().2D, \r2\().2D
168− trn1 \r2\().2D, \r8\().2D, \r2\().2D
169−
170− trn1 \r3\().2D, \r9\().2D, \r7\().2D
171− trn2 \r7\().2D, \r9\().2D, \r7\().2D
143+ trn1 \r8\().8h, \r0\().8h, \r1\().8h
144+ trn2 \r9\().8h, \r0\().8h, \r1\().8h
145+ trn1 \r1\().8h, \r2\().8h, \r3\().8h
146+ trn2 \r3\().8h, \r2\().8h, \r3\().8h
147+ trn1 \r0\().8h, \r4\().8h, \r5\().8h
148+ trn2 \r5\().8h, \r4\().8h, \r5\().8h
149+ trn1 \r2\().8h, \r6\().8h, \r7\().8h
150+ trn2 \r7\().8h, \r6\().8h, \r7\().8h
151+
152+ trn1 \r4\().4s, \r0\().4s, \r2\().4s
153+ trn2 \r2\().4s, \r0\().4s, \r2\().4s
154+ trn1 \r6\().4s, \r5\().4s, \r7\().4s
155+ trn2 \r7\().4s, \r5\().4s, \r7\().4s
156+ trn1 \r5\().4s, \r9\().4s, \r3\().4s
157+ trn2 \r9\().4s, \r9\().4s, \r3\().4s
158+ trn1 \r3\().4s, \r8\().4s, \r1\().4s
159+ trn2 \r8\().4s, \r8\().4s, \r1\().4s
160+
161+ trn1 \r0\().2d, \r3\().2d, \r4\().2d
162+ trn2 \r4\().2d, \r3\().2d, \r4\().2d
163+
164+ trn1 \r1\().2d, \r5\().2d, \r6\().2d
165+ trn2 \r5\().2d, \r5\().2d, \r6\().2d
166+
167+ trn2 \r6\().2d, \r8\().2d, \r2\().2d
168+ trn1 \r2\().2d, \r8\().2d, \r2\().2d
169+
170+ trn1 \r3\().2d, \r9\().2d, \r7\().2d
171+ trn2 \r7\().2d, \r9\().2d, \r7\().2d
172172
173173 .endm
libavcodec/aarch64/sbrdsp_neon.S+147−147
@@ -46,49 +46,49 @@ function ff_sbr_sum64x5_neon, export=1 · 45 unmodified lines
4646 add x3, x0, #192*4
4747 add x4, x0, #256*4
4848 mov x5, #64
49−1: ld1 {v0.4S}, [x0]
50− ld1 {v1.4S}, [x1], #16
51− fadd v0.4S, v0.4S, v1.4S
52− ld1 {v2.4S}, [x2], #16
53− fadd v0.4S, v0.4S, v2.4S
54− ld1 {v3.4S}, [x3], #16
55− fadd v0.4S, v0.4S, v3.4S
56− ld1 {v4.4S}, [x4], #16
57− fadd v0.4S, v0.4S, v4.4S
58− st1 {v0.4S}, [x0], #16
49+1: ld1 {v0.4s}, [x0]
50+ ld1 {v1.4s}, [x1], #16
51+ fadd v0.4s, v0.4s, v1.4s
52+ ld1 {v2.4s}, [x2], #16
53+ fadd v0.4s, v0.4s, v2.4s
54+ ld1 {v3.4s}, [x3], #16
55+ fadd v0.4s, v0.4s, v3.4s
56+ ld1 {v4.4s}, [x4], #16
57+ fadd v0.4s, v0.4s, v4.4s
58+ st1 {v0.4s}, [x0], #16
5959 subs x5, x5, #4
6060 b.gt 1b
6161 ret
6262 endfunc
6363
6464 function ff_sbr_sum_square_neon, export=1
65− movi v0.4S, #0
66−1: ld1 {v1.4S}, [x0], #16
67− fmla v0.4S, v1.4S, v1.4S
65+ movi v0.4s, #0
66+1: ld1 {v1.4s}, [x0], #16
67+ fmla v0.4s, v1.4s, v1.4s
6868 subs w1, w1, #2
6969 b.gt 1b
70− faddp v0.4S, v0.4S, v0.4S
71− faddp v0.4S, v0.4S, v0.4S
70+ faddp v0.4s, v0.4s, v0.4s
71+ faddp v0.4s, v0.4s, v0.4s
7272 ret
7373 endfunc
7474
7575 function ff_sbr_neg_odd_64_neon, export=1
7676 mov x1, x0
77− movi v5.4S, #1<<7, lsl #24
78− ld2 {v0.4S, v1.4S}, [x0], #32
79− eor v1.16B, v1.16B, v5.16B
80− ld2 {v2.4S, v3.4S}, [x0], #32
77+ movi v5.4s, #1<<7, lsl #24
78+ ld2 {v0.4s, v1.4s}, [x0], #32
79+ eor v1.16b, v1.16b, v5.16b
80+ ld2 {v2.4s, v3.4s}, [x0], #32
8181 .rept 3
82− st2 {v0.4S, v1.4S}, [x1], #32
83− eor v3.16B, v3.16B, v5.16B
84− ld2 {v0.4S, v1.4S}, [x0], #32
85− st2 {v2.4S, v3.4S}, [x1], #32
86− eor v1.16B, v1.16B, v5.16B
87− ld2 {v2.4S, v3.4S}, [x0], #32
82+ st2 {v0.4s, v1.4s}, [x1], #32
83+ eor v3.16b, v3.16b, v5.16b
84+ ld2 {v0.4s, v1.4s}, [x0], #32
85+ st2 {v2.4s, v3.4s}, [x1], #32
86+ eor v1.16b, v1.16b, v5.16b
87+ ld2 {v2.4s, v3.4s}, [x0], #32
8888 .endr
89− eor v3.16B, v3.16B, v5.16B
90− st2 {v0.4S, v1.4S}, [x1], #32
91− st2 {v2.4S, v3.4S}, [x1], #32
89+ eor v3.16b, v3.16b, v5.16b
90+ st2 {v0.4s, v1.4s}, [x1], #32
91+ st2 {v2.4s, v3.4s}, [x1], #32
9292 ret
9393 endfunc
9494
@@ -97,26 +97,26 @@ function ff_sbr_qmf_pre_shuffle_neon, export=1 · 2 unmodified lines
9797 add x2, x0, #64*4
9898 mov x3, #-16
9999 mov x4, #-4
100− movi v6.4S, #1<<7, lsl #24
101− ld1 {v0.2S}, [x0], #8
102− st1 {v0.2S}, [x2], #8
100+ movi v6.4s, #1<<7, lsl #24
101+ ld1 {v0.2s}, [x0], #8
102+ st1 {v0.2s}, [x2], #8
103103 .rept 7
104− ld1 {v1.4S}, [x1], x3
105− ld1 {v2.4S}, [x0], #16
106− eor v1.16B, v1.16B, v6.16B
107− rev64 v1.4S, v1.4S
108− ext v1.16B, v1.16B, v1.16B, #8
109− st2 {v1.4S, v2.4S}, [x2], #32
104+ ld1 {v1.4s}, [x1], x3
105+ ld1 {v2.4s}, [x0], #16
106+ eor v1.16b, v1.16b, v6.16b
107+ rev64 v1.4s, v1.4s
108+ ext v1.16b, v1.16b, v1.16b, #8
109+ st2 {v1.4s, v2.4s}, [x2], #32
110110 .endr
111111 add x1, x1, #8
112− ld1 {v1.2S}, [x1], x4
113− ld1 {v2.2S}, [x0], #8
114− ld1 {v1.S}[3], [x1]
115− ld1 {v2.S}[2], [x0]
116− eor v1.16B, v1.16B, v6.16B
117− rev64 v1.4S, v1.4S
118− st2 {v1.2S, v2.2S}, [x2], #16
119− st2 {v1.S, v2.S}[2], [x2]
112+ ld1 {v1.2s}, [x1], x4
113+ ld1 {v2.2s}, [x0], #8
114+ ld1 {v1.s}[3], [x1]
115+ ld1 {v2.s}[2], [x0]
116+ eor v1.16b, v1.16b, v6.16b
117+ rev64 v1.4s, v1.4s
118+ st2 {v1.2s, v2.2s}, [x2], #16
119+ st2 {v1.s, v2.s}[2], [x2]
120120 ret
121121 endfunc
122122
@@ -124,13 +124,13 @@ function ff_sbr_qmf_post_shuffle_neon, export=1 · 1 unmodified line
124124 add x2, x1, #60*4
125125 mov x3, #-16
126126 mov x4, #32
127− movi v6.4S, #1<<7, lsl #24
128−1: ld1 {v0.4S}, [x2], x3
129− ld1 {v1.4S}, [x1], #16
130− eor v0.16B, v0.16B, v6.16B
131− rev64 v0.4S, v0.4S
132− ext v0.16B, v0.16B, v0.16B, #8
133− st2 {v0.4S, v1.4S}, [x0], #32
127+ movi v6.4s, #1<<7, lsl #24
128+1: ld1 {v0.4s}, [x2], x3
129+ ld1 {v1.4s}, [x1], #16
130+ eor v0.16b, v0.16b, v6.16b
131+ rev64 v0.4s, v0.4s
132+ ext v0.16b, v0.16b, v0.16b, #8
133+ st2 {v0.4s, v1.4s}, [x0], #32
134134 subs x4, x4, #4
135135 b.gt 1b
136136 ret
@@ -141,13 +141,13 @@ function ff_sbr_qmf_deint_neg_neon, export=1 · 4 unmodified lines
141141 add x2, x0, #60*4
142142 mov x3, #-32
143143 mov x4, #32
144− movi v2.4S, #1<<7, lsl #24
145−1: ld2 {v0.4S, v1.4S}, [x1], x3
146− eor v0.16B, v0.16B, v2.16B
147− rev64 v1.4S, v1.4S
148− ext v1.16B, v1.16B, v1.16B, #8
149− st1 {v0.4S}, [x2]
150− st1 {v1.4S}, [x0], #16
144+ movi v2.4s, #1<<7, lsl #24
145+1: ld2 {v0.4s, v1.4s}, [x1], x3
146+ eor v0.16b, v0.16b, v2.16b
147+ rev64 v1.4s, v1.4s
148+ ext v1.16b, v1.16b, v1.16b, #8
149+ st1 {v0.4s}, [x2]
150+ st1 {v1.4s}, [x0], #16
151151 sub x2, x2, #16
152152 subs x4, x4, #4
153153 b.gt 1b
@@ -159,16 +159,16 @@ function ff_sbr_qmf_deint_bfly_neon, export=1 · 5 unmodified lines
159159 add x3, x0, #124*4
160160 mov x4, #64
161161 mov x5, #-16
162−1: ld1 {v0.4S}, [x1], #16
163− ld1 {v1.4S}, [x2], x5
164− rev64 v2.4S, v0.4S
165− ext v2.16B, v2.16B, v2.16B, #8
166− rev64 v3.4S, v1.4S
167− ext v3.16B, v3.16B, v3.16B, #8
168− fadd v1.4S, v1.4S, v2.4S
169− fsub v0.4S, v0.4S, v3.4S
170− st1 {v0.4S}, [x0], #16
171− st1 {v1.4S}, [x3], x5
162+1: ld1 {v0.4s}, [x1], #16
163+ ld1 {v1.4s}, [x2], x5
164+ rev64 v2.4s, v0.4s
165+ ext v2.16b, v2.16b, v2.16b, #8
166+ rev64 v3.4s, v1.4s
167+ ext v3.16b, v3.16b, v3.16b, #8
168+ fadd v1.4s, v1.4s, v2.4s
169+ fsub v0.4s, v0.4s, v3.4s
170+ st1 {v0.4s}, [x0], #16
171+ st1 {v1.4s}, [x3], x5
172172 subs x4, x4, #4
173173 b.gt 1b
174174 ret
@@ -178,32 +178,32 @@ function ff_sbr_hf_gen_neon, export=1 · 3 unmodified lines
178178 sxtw x4, w4
179179 sxtw x5, w5
180180 movrel x6, factors
181− ld1 {v7.4S}, [x6]
182− dup v1.4S, v0.S[0]
183− mov v2.8B, v1.8B
184− mov v2.S[2], v7.S[0]
185− mov v2.S[3], v7.S[0]
186− fmul v1.4S, v1.4S, v2.4S
187− ld1 {v0.D}[0], [x3]
188− ld1 {v0.D}[1], [x2]
189− fmul v0.4S, v0.4S, v1.4S
190− fmul v1.4S, v0.4S, v7.4S
191− rev64 v0.4S, v0.4S
181+ ld1 {v7.4s}, [x6]
182+ dup v1.4s, v0.s[0]
183+ mov v2.8b, v1.8b
184+ mov v2.s[2], v7.s[0]
185+ mov v2.s[3], v7.s[0]
186+ fmul v1.4s, v1.4s, v2.4s
187+ ld1 {v0.d}[0], [x3]
188+ ld1 {v0.d}[1], [x2]
189+ fmul v0.4s, v0.4s, v1.4s
190+ fmul v1.4s, v0.4s, v7.4s
191+ rev64 v0.4s, v0.4s
192192 sub x7, x5, x4
193193 add x0, x0, x4, lsl #3
194194 add x1, x1, x4, lsl #3
195195 sub x1, x1, #16
196−1: ld1 {v2.4S}, [x1], #16
197− ld1 {v3.2S}, [x1]
198− fmul v4.4S, v2.4S, v1.4S
199− fmul v5.4S, v2.4S, v0.4S
200− faddp v4.4S, v4.4S, v4.4S
201− faddp v5.4S, v5.4S, v5.4S
202− faddp v4.4S, v4.4S, v4.4S
203− faddp v5.4S, v5.4S, v5.4S
204− mov v4.S[1], v5.S[0]
205− fadd v4.2S, v4.2S, v3.2S
206− st1 {v4.2S}, [x0], #8
196+1: ld1 {v2.4s}, [x1], #16
197+ ld1 {v3.2s}, [x1]
198+ fmul v4.4s, v2.4s, v1.4s
199+ fmul v5.4s, v2.4s, v0.4s
200+ faddp v4.4s, v4.4s, v4.4s
201+ faddp v5.4s, v5.4s, v5.4s
202+ faddp v4.4s, v4.4s, v4.4s
203+ faddp v5.4s, v5.4s, v5.4s
204+ mov v4.s[1], v5.s[0]
205+ fadd v4.2s, v4.2s, v3.2s
206+ st1 {v4.2s}, [x0], #8
207207 sub x1, x1, #8
208208 subs x7, x7, #1
209209 b.gt 1b
@@ -215,10 +215,10 @@ function ff_sbr_hf_g_filt_neon, export=1 · 5 unmodified lines
215215 sxtw x4, w4
216216 mov x5, #40*2*4
217217 add x1, x1, x4, lsl #3
218−1: ld1 {v0.2S}, [x1], x5
219− ld1 {v1.S}[0], [x2], #4
220− fmul v2.4S, v0.4S, v1.S[0]
221− st1 {v2.2S}, [x0], #8
218+1: ld1 {v0.2s}, [x1], x5
219+ ld1 {v1.s}[0], [x2], #4
220+ fmul v2.4s, v0.4s, v1.s[0]
221+ st1 {v2.2s}, [x0], #8
222222 subs x3, x3, #1
223223 b.gt 1b
224224 ret
@@ -227,46 +227,46 @@ endfunc · 2 unmodified lines
227227 function ff_sbr_autocorrelate_neon, export=1
228228 mov x2, #38
229229 movrel x3, factors
230− ld1 {v0.4S}, [x3]
231− movi v1.4S, #0
232− movi v2.4S, #0
233− movi v3.4S, #0
234− ld1 {v4.2S}, [x0], #8
235− ld1 {v5.2S}, [x0], #8
236− fmul v16.2S, v4.2S, v4.2S
237− fmul v17.2S, v5.2S, v4.S[0]
238− fmul v18.2S, v5.2S, v4.S[1]
239−1: ld1 {v5.D}[1], [x0], #8
240− fmla v1.2S, v4.2S, v4.2S
241− fmla v2.4S, v5.4S, v4.S[0]
242− fmla v3.4S, v5.4S, v4.S[1]
243− mov v4.D[0], v5.D[0]
244− mov v5.D[0], v5.D[1]
230+ ld1 {v0.4s}, [x3]
231+ movi v1.4s, #0
232+ movi v2.4s, #0
233+ movi v3.4s, #0
234+ ld1 {v4.2s}, [x0], #8
235+ ld1 {v5.2s}, [x0], #8
236+ fmul v16.2s, v4.2s, v4.2s
237+ fmul v17.2s, v5.2s, v4.s[0]
238+ fmul v18.2s, v5.2s, v4.s[1]
239+1: ld1 {v5.d}[1], [x0], #8
240+ fmla v1.2s, v4.2s, v4.2s
241+ fmla v2.4s, v5.4s, v4.s[0]
242+ fmla v3.4s, v5.4s, v4.s[1]
243+ mov v4.d[0], v5.d[0]
244+ mov v5.d[0], v5.d[1]
245245 subs x2, x2, #1
246246 b.gt 1b
247− fmul v19.2S, v4.2S, v4.2S
248− fmul v20.2S, v5.2S, v4.S[0]
249− fmul v21.2S, v5.2S, v4.S[1]
250− fadd v22.4S, v2.4S, v20.4S
251− fsub v22.4S, v22.4S, v17.4S
252− fadd v23.4S, v3.4S, v21.4S
253− fsub v23.4S, v23.4S, v18.4S
254− rev64 v23.4S, v23.4S
255− fmul v23.4S, v23.4S, v0.4S
256− fadd v22.4S, v22.4S, v23.4S
257− st1 {v22.4S}, [x1], #16
258− fadd v23.2S, v1.2S, v19.2S
259− fsub v23.2S, v23.2S, v16.2S
260− faddp v23.2S, v23.2S, v23.2S
261− st1 {v23.S}[0], [x1]
247+ fmul v19.2s, v4.2s, v4.2s
248+ fmul v20.2s, v5.2s, v4.s[0]
249+ fmul v21.2s, v5.2s, v4.s[1]
250+ fadd v22.4s, v2.4s, v20.4s
251+ fsub v22.4s, v22.4s, v17.4s
252+ fadd v23.4s, v3.4s, v21.4s
253+ fsub v23.4s, v23.4s, v18.4s
254+ rev64 v23.4s, v23.4s
255+ fmul v23.4s, v23.4s, v0.4s
256+ fadd v22.4s, v22.4s, v23.4s
257+ st1 {v22.4s}, [x1], #16
258+ fadd v23.2s, v1.2s, v19.2s
259+ fsub v23.2s, v23.2s, v16.2s
260+ faddp v23.2s, v23.2s, v23.2s
261+ st1 {v23.s}[0], [x1]
262262 add x1, x1, #8
263− rev64 v3.2S, v3.2S
264− fmul v3.2S, v3.2S, v0.2S
265− fadd v2.2S, v2.2S, v3.2S
266− st1 {v2.2S}, [x1]
263+ rev64 v3.2s, v3.2s
264+ fmul v3.2s, v3.2s, v0.2s
265+ fadd v2.2s, v2.2s, v3.2s
266+ st1 {v2.2s}, [x1]
267267 add x1, x1, #16
268− faddp v1.2S, v1.2S, v1.2S
269− st1 {v1.S}[0], [x1]
268+ faddp v1.2s, v1.2s, v1.2s
269+ st1 {v1.s}[0], [x1]
270270 ret
271271 endfunc
272272
@@ -278,25 +278,25 @@ endfunc · 5 unmodified lines
278278 1: and x3, x3, #0x1ff
279279 add x8, x7, x3, lsl #3
280280 add x3, x3, #2
281− ld1 {v2.4S}, [x0]
282− ld1 {v3.2S}, [x1], #8
283− ld1 {v4.2S}, [x2], #8
284− ld1 {v5.4S}, [x8]
285− mov v6.16B, v2.16B
286− zip1 v3.4S, v3.4S, v3.4S
287− zip1 v4.4S, v4.4S, v4.4S
288− fmla v6.4S, v1.4S, v3.4S
289− fmla v2.4S, v5.4S, v4.4S
290− fcmeq v7.4S, v3.4S, #0
291− bif v2.16B, v6.16B, v7.16B
292− st1 {v2.4S}, [x0], #16
281+ ld1 {v2.4s}, [x0]
282+ ld1 {v3.2s}, [x1], #8
283+ ld1 {v4.2s}, [x2], #8
284+ ld1 {v5.4s}, [x8]
285+ mov v6.16b, v2.16b
286+ zip1 v3.4s, v3.4s, v3.4s
287+ zip1 v4.4s, v4.4s, v4.4s
288+ fmla v6.4s, v1.4s, v3.4s
289+ fmla v2.4s, v5.4s, v4.4s
290+ fcmeq v7.4s, v3.4s, #0
291+ bif v2.16b, v6.16b, v7.16b
292+ st1 {v2.4s}, [x0], #16
293293 subs x5, x5, #2
294294 b.gt 1b
295295 .endm
296296
297297 function ff_sbr_hf_apply_noise_0_neon, export=1
298298 movrel x9, phi_noise_0
299− ld1 {v1.4S}, [x9]
299+ ld1 {v1.4s}, [x9]
300300 apply_noise_common
301301 ret
302302 endfunc
@@ -305,14 +305,14 @@ function ff_sbr_hf_apply_noise_1_neon, export=1 · 2 unmodified lines
305305 movrel x9, phi_noise_1
306306 and x4, x4, #1
307307 add x9, x9, x4, lsl #4
308− ld1 {v1.4S}, [x9]
308+ ld1 {v1.4s}, [x9]
309309 apply_noise_common
310310 ret
311311 endfunc
312312
313313 function ff_sbr_hf_apply_noise_2_neon, export=1
314314 movrel x9, phi_noise_2
315− ld1 {v1.4S}, [x9]
315+ ld1 {v1.4s}, [x9]
316316 apply_noise_common
317317 ret
318318 endfunc
@@ -321,7 +321,7 @@ function ff_sbr_hf_apply_noise_3_neon, export=1 · 2 unmodified lines
321321 movrel x9, phi_noise_3
322322 and x4, x4, #1
323323 add x9, x9, x4, lsl #4
324− ld1 {v1.4S}, [x9]
324+ ld1 {v1.4s}, [x9]
325325 apply_noise_common
326326 ret
327327 endfunc
libavcodec/aarch64/simple_idct_neon.S+193−193
@@ -54,7 +54,7 @@ endconst · 53 unmodified lines
5454 prfm pldl1keep, [\data]
5555 mov x10, x30
5656 movrel x3, idct_coeff_neon
57− ld1 {v0.2D}, [x3]
57+ ld1 {v0.2d}, [x3]
5858 .endm
5959
6060 .macro idct_end
@@ -74,146 +74,146 @@ endconst · 13 unmodified lines
7474 .endm
7575
7676 .macro idct_col4_top y1, y2, y3, y4, i, l
77− smull\i v7.4S, \y3\l, z2
78− smull\i v16.4S, \y3\l, z6
79− smull\i v17.4S, \y2\l, z1
80− add v19.4S, v23.4S, v7.4S
81− smull\i v18.4S, \y2\l, z3
82− add v20.4S, v23.4S, v16.4S
83− smull\i v5.4S, \y2\l, z5
84− sub v21.4S, v23.4S, v16.4S
85− smull\i v6.4S, \y2\l, z7
86− sub v22.4S, v23.4S, v7.4S
87−
88− smlal\i v17.4S, \y4\l, z3
89− smlsl\i v18.4S, \y4\l, z7
90− smlsl\i v5.4S, \y4\l, z1
91− smlsl\i v6.4S, \y4\l, z5
77+ smull\i v7.4s, \y3\l, z2
78+ smull\i v16.4s, \y3\l, z6
79+ smull\i v17.4s, \y2\l, z1
80+ add v19.4s, v23.4s, v7.4s
81+ smull\i v18.4s, \y2\l, z3
82+ add v20.4s, v23.4s, v16.4s
83+ smull\i v5.4s, \y2\l, z5
84+ sub v21.4s, v23.4s, v16.4s
85+ smull\i v6.4s, \y2\l, z7
86+ sub v22.4s, v23.4s, v7.4s
87+
88+ smlal\i v17.4s, \y4\l, z3
89+ smlsl\i v18.4s, \y4\l, z7
90+ smlsl\i v5.4s, \y4\l, z1
91+ smlsl\i v6.4s, \y4\l, z5
9292 .endm
9393
9494 .macro idct_row4_neon y1, y2, y3, y4, pass
95− ld1 {\y1\().2D,\y2\().2D}, [x2], #32
96− movi v23.4S, #1<<2, lsl #8
97− orr v5.16B, \y1\().16B, \y2\().16B
98− ld1 {\y3\().2D,\y4\().2D}, [x2], #32
99− orr v6.16B, \y3\().16B, \y4\().16B
100− orr v5.16B, v5.16B, v6.16B
101− mov x3, v5.D[1]
102− smlal v23.4S, \y1\().4H, z4
95+ ld1 {\y1\().2d,\y2\().2d}, [x2], #32
96+ movi v23.4s, #1<<2, lsl #8
97+ orr v5.16b, \y1\().16b, \y2\().16b
98+ ld1 {\y3\().2d,\y4\().2d}, [x2], #32
99+ orr v6.16b, \y3\().16b, \y4\().16b
100+ orr v5.16b, v5.16b, v6.16b
101+ mov x3, v5.d[1]
102+ smlal v23.4s, \y1\().4h, z4
103103
104− idct_col4_top \y1, \y2, \y3, \y4, 1, .4H
104+ idct_col4_top \y1, \y2, \y3, \y4, 1, .4h
105105
106106 cmp x3, #0
107107 b.eq \pass\()f
108108
109− smull2 v7.4S, \y1\().8H, z4
110− smlal2 v17.4S, \y2\().8H, z5
111− smlsl2 v18.4S, \y2\().8H, z1
112− smull2 v16.4S, \y3\().8H, z2
113− smlal2 v5.4S, \y2\().8H, z7
114− add v19.4S, v19.4S, v7.4S
115− sub v20.4S, v20.4S, v7.4S
116− sub v21.4S, v21.4S, v7.4S
117− add v22.4S, v22.4S, v7.4S
118− smlal2 v6.4S, \y2\().8H, z3
119− smull2 v7.4S, \y3\().8H, z6
120− smlal2 v17.4S, \y4\().8H, z7
121− smlsl2 v18.4S, \y4\().8H, z5
122− smlal2 v5.4S, \y4\().8H, z3
123− smlsl2 v6.4S, \y4\().8H, z1
124− add v19.4S, v19.4S, v7.4S
125− sub v20.4S, v20.4S, v16.4S
126− add v21.4S, v21.4S, v16.4S
127− sub v22.4S, v22.4S, v7.4S
109+ smull2 v7.4s, \y1\().8h, z4
110+ smlal2 v17.4s, \y2\().8h, z5
111+ smlsl2 v18.4s, \y2\().8h, z1
112+ smull2 v16.4s, \y3\().8h, z2
113+ smlal2 v5.4s, \y2\().8h, z7
114+ add v19.4s, v19.4s, v7.4s
115+ sub v20.4s, v20.4s, v7.4s
116+ sub v21.4s, v21.4s, v7.4s
117+ add v22.4s, v22.4s, v7.4s
118+ smlal2 v6.4s, \y2\().8h, z3
119+ smull2 v7.4s, \y3\().8h, z6
120+ smlal2 v17.4s, \y4\().8h, z7
121+ smlsl2 v18.4s, \y4\().8h, z5
122+ smlal2 v5.4s, \y4\().8h, z3
123+ smlsl2 v6.4s, \y4\().8h, z1
124+ add v19.4s, v19.4s, v7.4s
125+ sub v20.4s, v20.4s, v16.4s
126+ add v21.4s, v21.4s, v16.4s
127+ sub v22.4s, v22.4s, v7.4s
128128
129129 \pass: add \y3\().4S, v19.4S, v17.4S
130− add \y4\().4S, v20.4S, v18.4S
131− shrn \y1\().4H, \y3\().4S, #ROW_SHIFT
132− shrn \y2\().4H, \y4\().4S, #ROW_SHIFT
133− add v7.4S, v21.4S, v5.4S
134− add v16.4S, v22.4S, v6.4S
135− shrn \y3\().4H, v7.4S, #ROW_SHIFT
136− shrn \y4\().4H, v16.4S, #ROW_SHIFT
137− sub v22.4S, v22.4S, v6.4S
138− sub v19.4S, v19.4S, v17.4S
139− sub v21.4S, v21.4S, v5.4S
140− shrn2 \y1\().8H, v22.4S, #ROW_SHIFT
141− sub v20.4S, v20.4S, v18.4S
142− shrn2 \y2\().8H, v21.4S, #ROW_SHIFT
143− shrn2 \y3\().8H, v20.4S, #ROW_SHIFT
144− shrn2 \y4\().8H, v19.4S, #ROW_SHIFT
145−
146− trn1 v16.8H, \y1\().8H, \y2\().8H
147− trn2 v17.8H, \y1\().8H, \y2\().8H
148− trn1 v18.8H, \y3\().8H, \y4\().8H
149− trn2 v19.8H, \y3\().8H, \y4\().8H
150− trn1 \y1\().4S, v16.4S, v18.4S
151− trn1 \y2\().4S, v17.4S, v19.4S
152− trn2 \y3\().4S, v16.4S, v18.4S
153− trn2 \y4\().4S, v17.4S, v19.4S
130+ add \y4\().4s, v20.4s, v18.4s
131+ shrn \y1\().4h, \y3\().4s, #ROW_SHIFT
132+ shrn \y2\().4h, \y4\().4s, #ROW_SHIFT
133+ add v7.4s, v21.4s, v5.4s
134+ add v16.4s, v22.4s, v6.4s
135+ shrn \y3\().4h, v7.4s, #ROW_SHIFT
136+ shrn \y4\().4h, v16.4s, #ROW_SHIFT
137+ sub v22.4s, v22.4s, v6.4s
138+ sub v19.4s, v19.4s, v17.4s
139+ sub v21.4s, v21.4s, v5.4s
140+ shrn2 \y1\().8h, v22.4s, #ROW_SHIFT
141+ sub v20.4s, v20.4s, v18.4s
142+ shrn2 \y2\().8h, v21.4s, #ROW_SHIFT
143+ shrn2 \y3\().8h, v20.4s, #ROW_SHIFT
144+ shrn2 \y4\().8h, v19.4s, #ROW_SHIFT
145+
146+ trn1 v16.8h, \y1\().8h, \y2\().8h
147+ trn2 v17.8h, \y1\().8h, \y2\().8h
148+ trn1 v18.8h, \y3\().8h, \y4\().8h
149+ trn2 v19.8h, \y3\().8h, \y4\().8h
150+ trn1 \y1\().4s, v16.4s, v18.4s
151+ trn1 \y2\().4s, v17.4s, v19.4s
152+ trn2 \y3\().4s, v16.4s, v18.4s
153+ trn2 \y4\().4s, v17.4s, v19.4s
154154 .endm
155155
156156 .macro declare_idct_col4_neon i, l
157157 function idct_col4_neon\i
158− dup v23.4H, z4c
158+ dup v23.4h, z4c
159159 .if \i == 1
160− add v23.4H, v23.4H, v24.4H
160+ add v23.4h, v23.4h, v24.4h
161161 .else
162− mov v5.D[0], v24.D[1]
163− add v23.4H, v23.4H, v5.4H
162+ mov v5.d[0], v24.d[1]
163+ add v23.4h, v23.4h, v5.4h
164164 .endif
165− smull v23.4S, v23.4H, z4
165+ smull v23.4s, v23.4h, z4
166166
167167 idct_col4_top v24, v25, v26, v27, \i, \l
168168
169− mov x4, v28.D[\i - 1]
170− mov x5, v29.D[\i - 1]
169+ mov x4, v28.d[\i - 1]
170+ mov x5, v29.d[\i - 1]
171171 cmp x4, #0
172172 b.eq 1f
173173
174− smull\i v7.4S, v28\l, z4
175− add v19.4S, v19.4S, v7.4S
176− sub v20.4S, v20.4S, v7.4S
177− sub v21.4S, v21.4S, v7.4S
178− add v22.4S, v22.4S, v7.4S
174+ smull\i v7.4s, v28\l, z4
175+ add v19.4s, v19.4s, v7.4s
176+ sub v20.4s, v20.4s, v7.4s
177+ sub v21.4s, v21.4s, v7.4s
178+ add v22.4s, v22.4s, v7.4s
179179
180−1: mov x4, v30.D[\i - 1]
180+1: mov x4, v30.d[\i - 1]
181181 cmp x5, #0
182182 b.eq 2f
183183
184− smlal\i v17.4S, v29\l, z5
185− smlsl\i v18.4S, v29\l, z1
186− smlal\i v5.4S, v29\l, z7
187− smlal\i v6.4S, v29\l, z3
184+ smlal\i v17.4s, v29\l, z5
185+ smlsl\i v18.4s, v29\l, z1
186+ smlal\i v5.4s, v29\l, z7
187+ smlal\i v6.4s, v29\l, z3
188188
189−2: mov x5, v31.D[\i - 1]
189+2: mov x5, v31.d[\i - 1]
190190 cmp x4, #0
191191 b.eq 3f
192192
193− smull\i v7.4S, v30\l, z6
194− smull\i v16.4S, v30\l, z2
195− add v19.4S, v19.4S, v7.4S
196− sub v22.4S, v22.4S, v7.4S
197− sub v20.4S, v20.4S, v16.4S
198− add v21.4S, v21.4S, v16.4S
193+ smull\i v7.4s, v30\l, z6
194+ smull\i v16.4s, v30\l, z2
195+ add v19.4s, v19.4s, v7.4s
196+ sub v22.4s, v22.4s, v7.4s
197+ sub v20.4s, v20.4s, v16.4s
198+ add v21.4s, v21.4s, v16.4s
199199
200200 3: cmp x5, #0
201201 b.eq 4f
202202
203− smlal\i v17.4S, v31\l, z7
204− smlsl\i v18.4S, v31\l, z5
205− smlal\i v5.4S, v31\l, z3
206− smlsl\i v6.4S, v31\l, z1
203+ smlal\i v17.4s, v31\l, z7
204+ smlsl\i v18.4s, v31\l, z5
205+ smlal\i v5.4s, v31\l, z3
206+ smlsl\i v6.4s, v31\l, z1
207207
208−4: addhn v7.4H, v19.4S, v17.4S
209− addhn2 v7.8H, v20.4S, v18.4S
210− subhn v18.4H, v20.4S, v18.4S
211− subhn2 v18.8H, v19.4S, v17.4S
208+4: addhn v7.4h, v19.4s, v17.4s
209+ addhn2 v7.8h, v20.4s, v18.4s
210+ subhn v18.4h, v20.4s, v18.4s
211+ subhn2 v18.8h, v19.4s, v17.4s
212212
213− addhn v16.4H, v21.4S, v5.4S
214− addhn2 v16.8H, v22.4S, v6.4S
215− subhn v17.4H, v22.4S, v6.4S
216− subhn2 v17.8H, v21.4S, v5.4S
213+ addhn v16.4h, v21.4s, v5.4s
214+ addhn2 v16.8h, v22.4s, v6.4s
215+ subhn v17.4h, v22.4s, v6.4s
216+ subhn2 v17.8h, v21.4s, v5.4s
217217
218218 ret
219219 endfunc
@@ -229,33 +229,33 @@ function ff_simple_idct_put_neon, export=1 · 9 unmodified lines
229229 idct_row4_neon v28, v29, v30, v31, 2
230230 bl idct_col4_neon1
231231
232− sqshrun v1.8B, v7.8H, #COL_SHIFT-16
233− sqshrun2 v1.16B, v16.8H, #COL_SHIFT-16
234− sqshrun v3.8B, v17.8H, #COL_SHIFT-16
235− sqshrun2 v3.16B, v18.8H, #COL_SHIFT-16
232+ sqshrun v1.8b, v7.8h, #COL_SHIFT-16
233+ sqshrun2 v1.16b, v16.8h, #COL_SHIFT-16
234+ sqshrun v3.8b, v17.8h, #COL_SHIFT-16
235+ sqshrun2 v3.16b, v18.8h, #COL_SHIFT-16
236236
237237 bl idct_col4_neon2
238238
239− sqshrun v2.8B, v7.8H, #COL_SHIFT-16
240− sqshrun2 v2.16B, v16.8H, #COL_SHIFT-16
241− sqshrun v4.8B, v17.8H, #COL_SHIFT-16
242− sqshrun2 v4.16B, v18.8H, #COL_SHIFT-16
239+ sqshrun v2.8b, v7.8h, #COL_SHIFT-16
240+ sqshrun2 v2.16b, v16.8h, #COL_SHIFT-16
241+ sqshrun v4.8b, v17.8h, #COL_SHIFT-16
242+ sqshrun2 v4.16b, v18.8h, #COL_SHIFT-16
243243
244− zip1 v16.4S, v1.4S, v2.4S
245− zip2 v17.4S, v1.4S, v2.4S
244+ zip1 v16.4s, v1.4s, v2.4s
245+ zip2 v17.4s, v1.4s, v2.4s
246246
247− st1 {v16.D}[0], [x0], x1
248− st1 {v16.D}[1], [x0], x1
247+ st1 {v16.d}[0], [x0], x1
248+ st1 {v16.d}[1], [x0], x1
249249
250− zip1 v18.4S, v3.4S, v4.4S
251− zip2 v19.4S, v3.4S, v4.4S
250+ zip1 v18.4s, v3.4s, v4.4s
251+ zip2 v19.4s, v3.4s, v4.4s
252252
253− st1 {v17.D}[0], [x0], x1
254− st1 {v17.D}[1], [x0], x1
255− st1 {v18.D}[0], [x0], x1
256− st1 {v18.D}[1], [x0], x1
257− st1 {v19.D}[0], [x0], x1
258− st1 {v19.D}[1], [x0], x1
253+ st1 {v17.d}[0], [x0], x1
254+ st1 {v17.d}[1], [x0], x1
255+ st1 {v18.d}[0], [x0], x1
256+ st1 {v18.d}[1], [x0], x1
257+ st1 {v19.d}[0], [x0], x1
258+ st1 {v19.d}[1], [x0], x1
259259
260260 idct_end
261261 endfunc
@@ -267,59 +267,59 @@ function ff_simple_idct_add_neon, export=1 · 5 unmodified lines
267267 idct_row4_neon v28, v29, v30, v31, 2
268268 bl idct_col4_neon1
269269
270− sshr v1.8H, v7.8H, #COL_SHIFT-16
271− sshr v2.8H, v16.8H, #COL_SHIFT-16
272− sshr v3.8H, v17.8H, #COL_SHIFT-16
273− sshr v4.8H, v18.8H, #COL_SHIFT-16
270+ sshr v1.8h, v7.8h, #COL_SHIFT-16
271+ sshr v2.8h, v16.8h, #COL_SHIFT-16
272+ sshr v3.8h, v17.8h, #COL_SHIFT-16
273+ sshr v4.8h, v18.8h, #COL_SHIFT-16
274274
275275 bl idct_col4_neon2
276276
277− sshr v7.8H, v7.8H, #COL_SHIFT-16
278− sshr v16.8H, v16.8H, #COL_SHIFT-16
279− sshr v17.8H, v17.8H, #COL_SHIFT-16
280− sshr v18.8H, v18.8H, #COL_SHIFT-16
277+ sshr v7.8h, v7.8h, #COL_SHIFT-16
278+ sshr v16.8h, v16.8h, #COL_SHIFT-16
279+ sshr v17.8h, v17.8h, #COL_SHIFT-16
280+ sshr v18.8h, v18.8h, #COL_SHIFT-16
281281
282282 mov x9, x0
283− ld1 {v19.D}[0], [x0], x1
284− zip1 v23.2D, v1.2D, v7.2D
285− zip2 v24.2D, v1.2D, v7.2D
286− ld1 {v19.D}[1], [x0], x1
287− zip1 v25.2D, v2.2D, v16.2D
288− zip2 v26.2D, v2.2D, v16.2D
289− ld1 {v20.D}[0], [x0], x1
290− zip1 v27.2D, v3.2D, v17.2D
291− zip2 v28.2D, v3.2D, v17.2D
292− ld1 {v20.D}[1], [x0], x1
293− zip1 v29.2D, v4.2D, v18.2D
294− zip2 v30.2D, v4.2D, v18.2D
295− ld1 {v21.D}[0], [x0], x1
296− uaddw v23.8H, v23.8H, v19.8B
297− uaddw2 v24.8H, v24.8H, v19.16B
298− ld1 {v21.D}[1], [x0], x1
299− sqxtun v23.8B, v23.8H
300− sqxtun2 v23.16B, v24.8H
301− ld1 {v22.D}[0], [x0], x1
302− uaddw v24.8H, v25.8H, v20.8B
303− uaddw2 v25.8H, v26.8H, v20.16B
304− ld1 {v22.D}[1], [x0], x1
305− sqxtun v24.8B, v24.8H
306− sqxtun2 v24.16B, v25.8H
307− st1 {v23.D}[0], [x9], x1
308− uaddw v25.8H, v27.8H, v21.8B
309− uaddw2 v26.8H, v28.8H, v21.16B
310− st1 {v23.D}[1], [x9], x1
311− sqxtun v25.8B, v25.8H
312− sqxtun2 v25.16B, v26.8H
313− st1 {v24.D}[0], [x9], x1
314− uaddw v26.8H, v29.8H, v22.8B
315− uaddw2 v27.8H, v30.8H, v22.16B
316− st1 {v24.D}[1], [x9], x1
317− sqxtun v26.8B, v26.8H
318− sqxtun2 v26.16B, v27.8H
319− st1 {v25.D}[0], [x9], x1
320− st1 {v25.D}[1], [x9], x1
321− st1 {v26.D}[0], [x9], x1
322− st1 {v26.D}[1], [x9], x1
283+ ld1 {v19.d}[0], [x0], x1
284+ zip1 v23.2d, v1.2d, v7.2d
285+ zip2 v24.2d, v1.2d, v7.2d
286+ ld1 {v19.d}[1], [x0], x1
287+ zip1 v25.2d, v2.2d, v16.2d
288+ zip2 v26.2d, v2.2d, v16.2d
289+ ld1 {v20.d}[0], [x0], x1
290+ zip1 v27.2d, v3.2d, v17.2d
291+ zip2 v28.2d, v3.2d, v17.2d
292+ ld1 {v20.d}[1], [x0], x1
293+ zip1 v29.2d, v4.2d, v18.2d
294+ zip2 v30.2d, v4.2d, v18.2d
295+ ld1 {v21.d}[0], [x0], x1
296+ uaddw v23.8h, v23.8h, v19.8b
297+ uaddw2 v24.8h, v24.8h, v19.16b
298+ ld1 {v21.d}[1], [x0], x1
299+ sqxtun v23.8b, v23.8h
300+ sqxtun2 v23.16b, v24.8h
301+ ld1 {v22.d}[0], [x0], x1
302+ uaddw v24.8h, v25.8h, v20.8b
303+ uaddw2 v25.8h, v26.8h, v20.16b
304+ ld1 {v22.d}[1], [x0], x1
305+ sqxtun v24.8b, v24.8h
306+ sqxtun2 v24.16b, v25.8h
307+ st1 {v23.d}[0], [x9], x1
308+ uaddw v25.8h, v27.8h, v21.8b
309+ uaddw2 v26.8h, v28.8h, v21.16b
310+ st1 {v23.d}[1], [x9], x1
311+ sqxtun v25.8b, v25.8h
312+ sqxtun2 v25.16b, v26.8h
313+ st1 {v24.d}[0], [x9], x1
314+ uaddw v26.8h, v29.8h, v22.8b
315+ uaddw2 v27.8h, v30.8h, v22.16b
316+ st1 {v24.d}[1], [x9], x1
317+ sqxtun v26.8b, v26.8h
318+ sqxtun2 v26.16b, v27.8h
319+ st1 {v25.d}[0], [x9], x1
320+ st1 {v25.d}[1], [x9], x1
321+ st1 {v26.d}[0], [x9], x1
322+ st1 {v26.d}[1], [x9], x1
323323
324324 idct_end
325325 endfunc
@@ -333,30 +333,30 @@ function ff_simple_idct_neon, export=1 · 7 unmodified lines
333333 sub x2, x2, #128
334334 bl idct_col4_neon1
335335
336− sshr v1.8H, v7.8H, #COL_SHIFT-16
337− sshr v2.8H, v16.8H, #COL_SHIFT-16
338− sshr v3.8H, v17.8H, #COL_SHIFT-16
339− sshr v4.8H, v18.8H, #COL_SHIFT-16
336+ sshr v1.8h, v7.8h, #COL_SHIFT-16
337+ sshr v2.8h, v16.8h, #COL_SHIFT-16
338+ sshr v3.8h, v17.8h, #COL_SHIFT-16
339+ sshr v4.8h, v18.8h, #COL_SHIFT-16
340340
341341 bl idct_col4_neon2
342342
343− sshr v7.8H, v7.8H, #COL_SHIFT-16
344− sshr v16.8H, v16.8H, #COL_SHIFT-16
345− sshr v17.8H, v17.8H, #COL_SHIFT-16
346− sshr v18.8H, v18.8H, #COL_SHIFT-16
347−
348− zip1 v23.2D, v1.2D, v7.2D
349− zip2 v24.2D, v1.2D, v7.2D
350− st1 {v23.2D,v24.2D}, [x2], #32
351− zip1 v25.2D, v2.2D, v16.2D
352− zip2 v26.2D, v2.2D, v16.2D
353− st1 {v25.2D,v26.2D}, [x2], #32
354− zip1 v27.2D, v3.2D, v17.2D
355− zip2 v28.2D, v3.2D, v17.2D
356− st1 {v27.2D,v28.2D}, [x2], #32
357− zip1 v29.2D, v4.2D, v18.2D
358− zip2 v30.2D, v4.2D, v18.2D
359− st1 {v29.2D,v30.2D}, [x2], #32
343+ sshr v7.8h, v7.8h, #COL_SHIFT-16
344+ sshr v16.8h, v16.8h, #COL_SHIFT-16
345+ sshr v17.8h, v17.8h, #COL_SHIFT-16
346+ sshr v18.8h, v18.8h, #COL_SHIFT-16
347+
348+ zip1 v23.2d, v1.2d, v7.2d
349+ zip2 v24.2d, v1.2d, v7.2d
350+ st1 {v23.2d,v24.2d}, [x2], #32
351+ zip1 v25.2d, v2.2d, v16.2d
352+ zip2 v26.2d, v2.2d, v16.2d
353+ st1 {v25.2d,v26.2d}, [x2], #32
354+ zip1 v27.2d, v3.2d, v17.2d
355+ zip2 v28.2d, v3.2d, v17.2d
356+ st1 {v27.2d,v28.2d}, [x2], #32
357+ zip1 v29.2d, v4.2d, v18.2d
358+ zip2 v30.2d, v4.2d, v18.2d
359+ st1 {v29.2d,v30.2d}, [x2], #32
360360
361361 idct_end
362362 endfunc
libavfilter/aarch64/vf_nlmeans_neon.S+35−35
@@ -22,19 +22,19 @@ · 21 unmodified lines
2222
2323 // acc_sum_store(ABCD) = {X+A, X+A+B, X+A+B+C, X+A+B+C+D}
2424 .macro acc_sum_store x, xb
25− dup v24.4S, v24.S[3] // ...X -> XXXX
26− ext v25.16B, v26.16B, \xb, #12 // ext(0000,ABCD,12)=0ABC
27− add v24.4S, v24.4S, \x // XXXX+ABCD={X+A,X+B,X+C,X+D}
28− add v24.4S, v24.4S, v25.4S // {X+A,X+B+A,X+C+B,X+D+C} (+0ABC)
29− ext v25.16B, v26.16B, v25.16B, #12 // ext(0000,0ABC,12)=00AB
30− add v24.4S, v24.4S, v25.4S // {X+A,X+B+A,X+C+B+A,X+D+C+B} (+00AB)
31− ext v25.16B, v26.16B, v25.16B, #12 // ext(0000,00AB,12)=000A
32− add v24.4S, v24.4S, v25.4S // {X+A,X+B+A,X+C+B+A,X+D+C+B+A} (+000A)
33− st1 {v24.4S}, [x0], #16 // write 4x32-bit final values
25+ dup v24.4s, v24.s[3] // ...X -> XXXX
26+ ext v25.16b, v26.16b, \xb, #12 // ext(0000,ABCD,12)=0ABC
27+ add v24.4s, v24.4s, \x // XXXX+ABCD={X+A,X+B,X+C,X+D}
28+ add v24.4s, v24.4s, v25.4s // {X+A,X+B+A,X+C+B,X+D+C} (+0ABC)
29+ ext v25.16b, v26.16b, v25.16b, #12 // ext(0000,0ABC,12)=00AB
30+ add v24.4s, v24.4s, v25.4s // {X+A,X+B+A,X+C+B+A,X+D+C+B} (+00AB)
31+ ext v25.16b, v26.16b, v25.16b, #12 // ext(0000,00AB,12)=000A
32+ add v24.4s, v24.4s, v25.4s // {X+A,X+B+A,X+C+B+A,X+D+C+B+A} (+000A)
33+ st1 {v24.4s}, [x0], #16 // write 4x32-bit final values
3434 .endm
3535
3636 function ff_compute_safe_ssd_integral_image_neon, export=1
37− movi v26.4S, #0 // used as zero for the "rotations" in acc_sum_store
37+ movi v26.4s, #0 // used as zero for the "rotations" in acc_sum_store
3838 sub x3, x3, w6, UXTW // s1 padding (s1_linesize - w)
3939 sub x5, x5, w6, UXTW // s2 padding (s2_linesize - w)
4040 sub x9, x0, w1, UXTW #2 // dst_top
@@ -43,31 +43,31 @@ function ff_compute_safe_ssd_integral_image_neon, export=1 · 2 unmodified lines
4343 1: mov w10, w6 // width copy for each line
4444 sub x0, x0, #16 // beginning of the dst line minus 4 sums
4545 sub x8, x9, #4 // dst_top-1
46− ld1 {v24.4S}, [x0], #16 // load ...X (contextual last sums)
47−2: ld1 {v0.16B}, [x2], #16 // s1[x + 0..15]
48− ld1 {v1.16B}, [x4], #16 // s2[x + 0..15]
49− ld1 {v16.4S,v17.4S}, [x8], #32 // dst_top[x + 0..7 - 1]
50− usubl v2.8H, v0.8B, v1.8B // d[x + 0..7] = s1[x + 0..7] - s2[x + 0..7]
51− usubl2 v3.8H, v0.16B, v1.16B // d[x + 8..15] = s1[x + 8..15] - s2[x + 8..15]
52− ld1 {v18.4S,v19.4S}, [x8], #32 // dst_top[x + 8..15 - 1]
53− smull v4.4S, v2.4H, v2.4H // d[x + 0..3]^2
54− smull2 v5.4S, v2.8H, v2.8H // d[x + 4..7]^2
55− ld1 {v20.4S,v21.4S}, [x9], #32 // dst_top[x + 0..7]
56− smull v6.4S, v3.4H, v3.4H // d[x + 8..11]^2
57− smull2 v7.4S, v3.8H, v3.8H // d[x + 12..15]^2
58− ld1 {v22.4S,v23.4S}, [x9], #32 // dst_top[x + 8..15]
59− sub v0.4S, v20.4S, v16.4S // dst_top[x + 0..3] - dst_top[x + 0..3 - 1]
60− sub v1.4S, v21.4S, v17.4S // dst_top[x + 4..7] - dst_top[x + 4..7 - 1]
61− add v0.4S, v0.4S, v4.4S // + d[x + 0..3]^2
62− add v1.4S, v1.4S, v5.4S // + d[x + 4..7]^2
63− sub v2.4S, v22.4S, v18.4S // dst_top[x + 8..11] - dst_top[x + 8..11 - 1]
64− sub v3.4S, v23.4S, v19.4S // dst_top[x + 12..15] - dst_top[x + 12..15 - 1]
65− add v2.4S, v2.4S, v6.4S // + d[x + 8..11]^2
66− add v3.4S, v3.4S, v7.4S // + d[x + 12..15]^2
67− acc_sum_store v0.4S, v0.16B // accumulate and store dst[ 0..3]
68− acc_sum_store v1.4S, v1.16B // accumulate and store dst[ 4..7]
69− acc_sum_store v2.4S, v2.16B // accumulate and store dst[ 8..11]
70− acc_sum_store v3.4S, v3.16B // accumulate and store dst[12..15]
46+ ld1 {v24.4s}, [x0], #16 // load ...X (contextual last sums)
47+2: ld1 {v0.16b}, [x2], #16 // s1[x + 0..15]
48+ ld1 {v1.16b}, [x4], #16 // s2[x + 0..15]
49+ ld1 {v16.4s,v17.4s}, [x8], #32 // dst_top[x + 0..7 - 1]
50+ usubl v2.8h, v0.8b, v1.8b // d[x + 0..7] = s1[x + 0..7] - s2[x + 0..7]
51+ usubl2 v3.8h, v0.16b, v1.16b // d[x + 8..15] = s1[x + 8..15] - s2[x + 8..15]
52+ ld1 {v18.4s,v19.4s}, [x8], #32 // dst_top[x + 8..15 - 1]
53+ smull v4.4s, v2.4h, v2.4h // d[x + 0..3]^2
54+ smull2 v5.4s, v2.8h, v2.8h // d[x + 4..7]^2
55+ ld1 {v20.4s,v21.4s}, [x9], #32 // dst_top[x + 0..7]
56+ smull v6.4s, v3.4h, v3.4h // d[x + 8..11]^2
57+ smull2 v7.4s, v3.8h, v3.8h // d[x + 12..15]^2
58+ ld1 {v22.4s,v23.4s}, [x9], #32 // dst_top[x + 8..15]
59+ sub v0.4s, v20.4s, v16.4s // dst_top[x + 0..3] - dst_top[x + 0..3 - 1]
60+ sub v1.4s, v21.4s, v17.4s // dst_top[x + 4..7] - dst_top[x + 4..7 - 1]
61+ add v0.4s, v0.4s, v4.4s // + d[x + 0..3]^2
62+ add v1.4s, v1.4s, v5.4s // + d[x + 4..7]^2
63+ sub v2.4s, v22.4s, v18.4s // dst_top[x + 8..11] - dst_top[x + 8..11 - 1]
64+ sub v3.4s, v23.4s, v19.4s // dst_top[x + 12..15] - dst_top[x + 12..15 - 1]
65+ add v2.4s, v2.4s, v6.4s // + d[x + 8..11]^2
66+ add v3.4s, v3.4s, v7.4s // + d[x + 12..15]^2
67+ acc_sum_store v0.4s, v0.16b // accumulate and store dst[ 0..3]
68+ acc_sum_store v1.4s, v1.16b // accumulate and store dst[ 4..7]
69+ acc_sum_store v2.4s, v2.16b // accumulate and store dst[ 8..11]
70+ acc_sum_store v3.4s, v3.16b // accumulate and store dst[12..15]
7171 subs w10, w10, #16 // width dec
7272 b.ne 2b // loop til next line
7373 add x2, x2, x3 // skip to next line (s1)
libavutil/aarch64/float_dsp_neon.S+100−100
@@ -25,16 +25,16 @@ · 24 unmodified lines
2525
2626 function ff_vector_fmul_neon, export=1
2727 1: subs w3, w3, #16
28− ld1 {v0.4S, v1.4S}, [x1], #32
29− ld1 {v2.4S, v3.4S}, [x1], #32
30− ld1 {v4.4S, v5.4S}, [x2], #32
31− ld1 {v6.4S, v7.4S}, [x2], #32
32− fmul v16.4S, v0.4S, v4.4S
33− fmul v17.4S, v1.4S, v5.4S
34− fmul v18.4S, v2.4S, v6.4S
35− fmul v19.4S, v3.4S, v7.4S
36− st1 {v16.4S, v17.4S}, [x0], #32
37− st1 {v18.4S, v19.4S}, [x0], #32
28+ ld1 {v0.4s, v1.4s}, [x1], #32
29+ ld1 {v2.4s, v3.4s}, [x1], #32
30+ ld1 {v4.4s, v5.4s}, [x2], #32
31+ ld1 {v6.4s, v7.4s}, [x2], #32
32+ fmul v16.4s, v0.4s, v4.4s
33+ fmul v17.4s, v1.4s, v5.4s
34+ fmul v18.4s, v2.4s, v6.4s
35+ fmul v19.4s, v3.4s, v7.4s
36+ st1 {v16.4s, v17.4s}, [x0], #32
37+ st1 {v18.4s, v19.4s}, [x0], #32
3838 b.ne 1b
3939 ret
4040 endfunc
@@ -42,16 +42,16 @@ endfunc · 1 unmodified line
4242 function ff_vector_fmac_scalar_neon, export=1
4343 mov x3, #-32
4444 1: subs w2, w2, #16
45− ld1 {v16.4S, v17.4S}, [x0], #32
46− ld1 {v18.4S, v19.4S}, [x0], x3
47− ld1 {v4.4S, v5.4S}, [x1], #32
48− ld1 {v6.4S, v7.4S}, [x1], #32
49− fmla v16.4S, v4.4S, v0.S[0]
50− fmla v17.4S, v5.4S, v0.S[0]
51− fmla v18.4S, v6.4S, v0.S[0]
52− fmla v19.4S, v7.4S, v0.S[0]
53− st1 {v16.4S, v17.4S}, [x0], #32
54− st1 {v18.4S, v19.4S}, [x0], #32
45+ ld1 {v16.4s, v17.4s}, [x0], #32
46+ ld1 {v18.4s, v19.4s}, [x0], x3
47+ ld1 {v4.4s, v5.4s}, [x1], #32
48+ ld1 {v6.4s, v7.4s}, [x1], #32
49+ fmla v16.4s, v4.4s, v0.s[0]
50+ fmla v17.4s, v5.4s, v0.s[0]
51+ fmla v18.4s, v6.4s, v0.s[0]
52+ fmla v19.4s, v7.4s, v0.s[0]
53+ st1 {v16.4s, v17.4s}, [x0], #32
54+ st1 {v18.4s, v19.4s}, [x0], #32
5555 b.ne 1b
5656 ret
5757 endfunc
@@ -59,43 +59,43 @@ endfunc · 1 unmodified line
5959 function ff_vector_fmul_scalar_neon, export=1
6060 mov w4, #15
6161 bics w3, w2, w4
62− dup v16.4S, v0.S[0]
62+ dup v16.4s, v0.s[0]
6363 b.eq 3f
64− ld1 {v0.4S, v1.4S}, [x1], #32
64+ ld1 {v0.4s, v1.4s}, [x1], #32
6565 1: subs w3, w3, #16
66− fmul v0.4S, v0.4S, v16.4S
67− ld1 {v2.4S, v3.4S}, [x1], #32
68− fmul v1.4S, v1.4S, v16.4S
69− fmul v2.4S, v2.4S, v16.4S
70− st1 {v0.4S, v1.4S}, [x0], #32
71− fmul v3.4S, v3.4S, v16.4S
66+ fmul v0.4s, v0.4s, v16.4s
67+ ld1 {v2.4s, v3.4s}, [x1], #32
68+ fmul v1.4s, v1.4s, v16.4s
69+ fmul v2.4s, v2.4s, v16.4s
70+ st1 {v0.4s, v1.4s}, [x0], #32
71+ fmul v3.4s, v3.4s, v16.4s
7272 b.eq 2f
73− ld1 {v0.4S, v1.4S}, [x1], #32
74− st1 {v2.4S, v3.4S}, [x0], #32
73+ ld1 {v0.4s, v1.4s}, [x1], #32
74+ st1 {v2.4s, v3.4s}, [x0], #32
7575 b 1b
7676 2: ands w2, w2, #15
77− st1 {v2.4S, v3.4S}, [x0], #32
77+ st1 {v2.4s, v3.4s}, [x0], #32
7878 b.eq 4f
79−3: ld1 {v0.4S}, [x1], #16
80− fmul v0.4S, v0.4S, v16.4S
81− st1 {v0.4S}, [x0], #16
79+3: ld1 {v0.4s}, [x1], #16
80+ fmul v0.4s, v0.4s, v16.4s
81+ st1 {v0.4s}, [x0], #16
8282 subs w2, w2, #4
8383 b.gt 3b
8484 4: ret
8585 endfunc
8686
8787 function ff_vector_dmul_scalar_neon, export=1
88− dup v16.2D, v0.D[0]
89− ld1 {v0.2D, v1.2D}, [x1], #32
88+ dup v16.2d, v0.d[0]
89+ ld1 {v0.2d, v1.2d}, [x1], #32
9090 1: subs w2, w2, #8
91− fmul v0.2D, v0.2D, v16.2D
92− ld1 {v2.2D, v3.2D}, [x1], #32
93− fmul v1.2D, v1.2D, v16.2D
94− fmul v2.2D, v2.2D, v16.2D
95− st1 {v0.2D, v1.2D}, [x0], #32
96− fmul v3.2D, v3.2D, v16.2D
97− ld1 {v0.2D, v1.2D}, [x1], #32
98− st1 {v2.2D, v3.2D}, [x0], #32
91+ fmul v0.2d, v0.2d, v16.2d
92+ ld1 {v2.2d, v3.2d}, [x1], #32
93+ fmul v1.2d, v1.2d, v16.2d
94+ fmul v2.2d, v2.2d, v16.2d
95+ st1 {v0.2d, v1.2d}, [x0], #32
96+ fmul v3.2d, v3.2d, v16.2d
97+ ld1 {v0.2d, v1.2d}, [x1], #32
98+ st1 {v2.2d, v3.2d}, [x0], #32
9999 b.gt 1b
100100 ret
101101 endfunc
@@ -108,49 +108,49 @@ function ff_vector_fmul_window_neon, export=1 · 6 unmodified lines
108108 add x6, x3, x5, lsl #3 // win + 8 * (len - 2)
109109 add x5, x0, x5, lsl #3 // dst + 8 * (len - 2)
110110 mov x7, #-16
111− ld1 {v0.4S}, [x1], #16 // s0
112− ld1 {v2.4S}, [x3], #16 // wi
113− ld1 {v1.4S}, [x2], x7 // s1
114−1: ld1 {v3.4S}, [x6], x7 // wj
111+ ld1 {v0.4s}, [x1], #16 // s0
112+ ld1 {v2.4s}, [x3], #16 // wi
113+ ld1 {v1.4s}, [x2], x7 // s1
114+1: ld1 {v3.4s}, [x6], x7 // wj
115115 subs x4, x4, #4
116− fmul v17.4S, v0.4S, v2.4S // s0 * wi
117− rev64 v4.4S, v1.4S
118− rev64 v5.4S, v3.4S
119− rev64 v17.4S, v17.4S
120− ext v4.16B, v4.16B, v4.16B, #8 // s1_r
121− ext v5.16B, v5.16B, v5.16B, #8 // wj_r
122− ext v17.16B, v17.16B, v17.16B, #8 // (s0 * wi)_rev
123− fmul v16.4S, v0.4S, v5.4S // s0 * wj_r
124− fmla v17.4S, v1.4S, v3.4S // (s0 * wi)_rev + s1 * wj
116+ fmul v17.4s, v0.4s, v2.4s // s0 * wi
117+ rev64 v4.4s, v1.4s
118+ rev64 v5.4s, v3.4s
119+ rev64 v17.4s, v17.4s
120+ ext v4.16b, v4.16b, v4.16b, #8 // s1_r
121+ ext v5.16b, v5.16b, v5.16b, #8 // wj_r
122+ ext v17.16b, v17.16b, v17.16b, #8 // (s0 * wi)_rev
123+ fmul v16.4s, v0.4s, v5.4s // s0 * wj_r
124+ fmla v17.4s, v1.4s, v3.4s // (s0 * wi)_rev + s1 * wj
125125 b.eq 2f
126− ld1 {v0.4S}, [x1], #16
127− fmls v16.4S, v4.4S, v2.4S // s0 * wj_r - s1_r * wi
128− st1 {v17.4S}, [x5], x7
129− ld1 {v2.4S}, [x3], #16
130− ld1 {v1.4S}, [x2], x7
131− st1 {v16.4S}, [x0], #16
126+ ld1 {v0.4s}, [x1], #16
127+ fmls v16.4s, v4.4s, v2.4s // s0 * wj_r - s1_r * wi
128+ st1 {v17.4s}, [x5], x7
129+ ld1 {v2.4s}, [x3], #16
130+ ld1 {v1.4s}, [x2], x7
131+ st1 {v16.4s}, [x0], #16
132132 b 1b
133133 2:
134− fmls v16.4S, v4.4S, v2.4S // s0 * wj_r - s1_r * wi
135− st1 {v17.4S}, [x5], x7
136− st1 {v16.4S}, [x0], #16
134+ fmls v16.4s, v4.4s, v2.4s // s0 * wj_r - s1_r * wi
135+ st1 {v17.4s}, [x5], x7
136+ st1 {v16.4s}, [x0], #16
137137 ret
138138 endfunc
139139
140140 function ff_vector_fmul_add_neon, export=1
141− ld1 {v0.4S, v1.4S}, [x1], #32
142− ld1 {v2.4S, v3.4S}, [x2], #32
143− ld1 {v4.4S, v5.4S}, [x3], #32
141+ ld1 {v0.4s, v1.4s}, [x1], #32
142+ ld1 {v2.4s, v3.4s}, [x2], #32
143+ ld1 {v4.4s, v5.4s}, [x3], #32
144144 1: subs w4, w4, #8
145− fmla v4.4S, v0.4S, v2.4S
146− fmla v5.4S, v1.4S, v3.4S
145+ fmla v4.4s, v0.4s, v2.4s
146+ fmla v5.4s, v1.4s, v3.4s
147147 b.eq 2f
148− ld1 {v0.4S, v1.4S}, [x1], #32
149− ld1 {v2.4S, v3.4S}, [x2], #32
150− st1 {v4.4S, v5.4S}, [x0], #32
151− ld1 {v4.4S, v5.4S}, [x3], #32
148+ ld1 {v0.4s, v1.4s}, [x1], #32
149+ ld1 {v2.4s, v3.4s}, [x2], #32
150+ st1 {v4.4s, v5.4s}, [x0], #32
151+ ld1 {v4.4s, v5.4s}, [x3], #32
152152 b 1b
153−2: st1 {v4.4S, v5.4S}, [x0], #32
153+2: st1 {v4.4s, v5.4s}, [x0], #32
154154 ret
155155 endfunc
156156
@@ -159,44 +159,44 @@ function ff_vector_fmul_reverse_neon, export=1 · 2 unmodified lines
159159 add x2, x2, x3, lsl #2
160160 sub x2, x2, #32
161161 mov x4, #-32
162− ld1 {v2.4S, v3.4S}, [x2], x4
163− ld1 {v0.4S, v1.4S}, [x1], #32
162+ ld1 {v2.4s, v3.4s}, [x2], x4
163+ ld1 {v0.4s, v1.4s}, [x1], #32
164164 1: subs x3, x3, #8
165− rev64 v3.4S, v3.4S
166− rev64 v2.4S, v2.4S
167− ext v3.16B, v3.16B, v3.16B, #8
168− ext v2.16B, v2.16B, v2.16B, #8
169− fmul v16.4S, v0.4S, v3.4S
170− fmul v17.4S, v1.4S, v2.4S
165+ rev64 v3.4s, v3.4s
166+ rev64 v2.4s, v2.4s
167+ ext v3.16b, v3.16b, v3.16b, #8
168+ ext v2.16b, v2.16b, v2.16b, #8
169+ fmul v16.4s, v0.4s, v3.4s
170+ fmul v17.4s, v1.4s, v2.4s
171171 b.eq 2f
172− ld1 {v2.4S, v3.4S}, [x2], x4
173− ld1 {v0.4S, v1.4S}, [x1], #32
174− st1 {v16.4S, v17.4S}, [x0], #32
172+ ld1 {v2.4s, v3.4s}, [x2], x4
173+ ld1 {v0.4s, v1.4s}, [x1], #32
174+ st1 {v16.4s, v17.4s}, [x0], #32
175175 b 1b
176−2: st1 {v16.4S, v17.4S}, [x0], #32
176+2: st1 {v16.4s, v17.4s}, [x0], #32
177177 ret
178178 endfunc
179179
180180 function ff_butterflies_float_neon, export=1
181−1: ld1 {v0.4S}, [x0]
182− ld1 {v1.4S}, [x1]
181+1: ld1 {v0.4s}, [x0]
182+ ld1 {v1.4s}, [x1]
183183 subs w2, w2, #4
184− fsub v2.4S, v0.4S, v1.4S
185− fadd v3.4S, v0.4S, v1.4S
186− st1 {v2.4S}, [x1], #16
187− st1 {v3.4S}, [x0], #16
184+ fsub v2.4s, v0.4s, v1.4s
185+ fadd v3.4s, v0.4s, v1.4s
186+ st1 {v2.4s}, [x1], #16
187+ st1 {v3.4s}, [x0], #16
188188 b.gt 1b
189189 ret
190190 endfunc
191191
192192 function ff_scalarproduct_float_neon, export=1
193− movi v2.4S, #0
194−1: ld1 {v0.4S}, [x0], #16
195− ld1 {v1.4S}, [x1], #16
193+ movi v2.4s, #0
194+1: ld1 {v0.4s}, [x0], #16
195+ ld1 {v1.4s}, [x1], #16
196196 subs w2, w2, #4
197− fmla v2.4S, v0.4S, v1.4S
197+ fmla v2.4s, v0.4s, v1.4s
198198 b.gt 1b
199− faddp v0.4S, v2.4S, v2.4S
200− faddp s0, v0.2S
199+ faddp v0.4s, v2.4s, v2.4s
200+ faddp s0, v0.2s
201201 ret
202202 endfunc
libswresample/aarch64/resample.S+32−32
@@ -21,57 +21,57 @@ · 20 unmodified lines
2121 #include "libavutil/aarch64/asm.S"
2222
2323 function ff_resample_common_apply_filter_x4_float_neon, export=1
24− movi v0.4S, #0 // accumulator
25−1: ld1 {v1.4S}, [x1], #16 // src[0..3]
26− ld1 {v2.4S}, [x2], #16 // filter[0..3]
27− fmla v0.4S, v1.4S, v2.4S // accumulator += src[0..3] * filter[0..3]
24+ movi v0.4s, #0 // accumulator
25+1: ld1 {v1.4s}, [x1], #16 // src[0..3]
26+ ld1 {v2.4s}, [x2], #16 // filter[0..3]
27+ fmla v0.4s, v1.4s, v2.4s // accumulator += src[0..3] * filter[0..3]
2828 subs w3, w3, #4 // filter_length -= 4
2929 b.gt 1b // loop until filter_length
30− faddp v0.4S, v0.4S, v0.4S // pair adding of the 4x32-bit accumulated values
31− faddp v0.4S, v0.4S, v0.4S // pair adding of the 4x32-bit accumulated values
32− st1 {v0.S}[0], [x0], #4 // write accumulator
30+ faddp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values
31+ faddp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values
32+ st1 {v0.s}[0], [x0], #4 // write accumulator
3333 ret
3434 endfunc
3535
3636 function ff_resample_common_apply_filter_x8_float_neon, export=1
37− movi v0.4S, #0 // accumulator
38−1: ld1 {v1.4S}, [x1], #16 // src[0..3]
39− ld1 {v2.4S}, [x2], #16 // filter[0..3]
40− ld1 {v3.4S}, [x1], #16 // src[4..7]
41− ld1 {v4.4S}, [x2], #16 // filter[4..7]
42− fmla v0.4S, v1.4S, v2.4S // accumulator += src[0..3] * filter[0..3]
43− fmla v0.4S, v3.4S, v4.4S // accumulator += src[4..7] * filter[4..7]
37+ movi v0.4s, #0 // accumulator
38+1: ld1 {v1.4s}, [x1], #16 // src[0..3]
39+ ld1 {v2.4s}, [x2], #16 // filter[0..3]
40+ ld1 {v3.4s}, [x1], #16 // src[4..7]
41+ ld1 {v4.4s}, [x2], #16 // filter[4..7]
42+ fmla v0.4s, v1.4s, v2.4s // accumulator += src[0..3] * filter[0..3]
43+ fmla v0.4s, v3.4s, v4.4s // accumulator += src[4..7] * filter[4..7]
4444 subs w3, w3, #8 // filter_length -= 8
4545 b.gt 1b // loop until filter_length
46− faddp v0.4S, v0.4S, v0.4S // pair adding of the 4x32-bit accumulated values
47− faddp v0.4S, v0.4S, v0.4S // pair adding of the 4x32-bit accumulated values
48− st1 {v0.S}[0], [x0], #4 // write accumulator
46+ faddp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values
47+ faddp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values
48+ st1 {v0.s}[0], [x0], #4 // write accumulator
4949 ret
5050 endfunc
5151
5252 function ff_resample_common_apply_filter_x4_s16_neon, export=1
53− movi v0.4S, #0 // accumulator
54−1: ld1 {v1.4H}, [x1], #8 // src[0..3]
55− ld1 {v2.4H}, [x2], #8 // filter[0..3]
56− smlal v0.4S, v1.4H, v2.4H // accumulator += src[0..3] * filter[0..3]
53+ movi v0.4s, #0 // accumulator
54+1: ld1 {v1.4h}, [x1], #8 // src[0..3]
55+ ld1 {v2.4h}, [x2], #8 // filter[0..3]
56+ smlal v0.4s, v1.4h, v2.4h // accumulator += src[0..3] * filter[0..3]
5757 subs w3, w3, #4 // filter_length -= 4
5858 b.gt 1b // loop until filter_length
59− addp v0.4S, v0.4S, v0.4S // pair adding of the 4x32-bit accumulated values
60− addp v0.4S, v0.4S, v0.4S // pair adding of the 4x32-bit accumulated values
61− st1 {v0.S}[0], [x0], #4 // write accumulator
59+ addp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values
60+ addp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values
61+ st1 {v0.s}[0], [x0], #4 // write accumulator
6262 ret
6363 endfunc
6464
6565 function ff_resample_common_apply_filter_x8_s16_neon, export=1
66− movi v0.4S, #0 // accumulator
67−1: ld1 {v1.8H}, [x1], #16 // src[0..7]
68− ld1 {v2.8H}, [x2], #16 // filter[0..7]
69− smlal v0.4S, v1.4H, v2.4H // accumulator += src[0..3] * filter[0..3]
70− smlal2 v0.4S, v1.8H, v2.8H // accumulator += src[4..7] * filter[4..7]
66+ movi v0.4s, #0 // accumulator
67+1: ld1 {v1.8h}, [x1], #16 // src[0..7]
68+ ld1 {v2.8h}, [x2], #16 // filter[0..7]
69+ smlal v0.4s, v1.4h, v2.4h // accumulator += src[0..3] * filter[0..3]
70+ smlal2 v0.4s, v1.8h, v2.8h // accumulator += src[4..7] * filter[4..7]
7171 subs w3, w3, #8 // filter_length -= 8
7272 b.gt 1b // loop until filter_length
73− addp v0.4S, v0.4S, v0.4S // pair adding of the 4x32-bit accumulated values
74− addp v0.4S, v0.4S, v0.4S // pair adding of the 4x32-bit accumulated values
75− st1 {v0.S}[0], [x0], #4 // write accumulator
73+ addp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values
74+ addp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values
75+ st1 {v0.s}[0], [x0], #4 // write accumulator
7676 ret
7777 endfunc
libswscale/aarch64/hscale.S+216−216
@@ -50,43 +50,43 @@ function ff_hscale8to15_X8_neon, export=1 · 49 unmodified lines
5050 add x12, x16, x7 // filter1 = filter0 + filterSize*2
5151 add x13, x12, x7 // filter2 = filter1 + filterSize*2
5252 add x4, x13, x7 // filter3 = filter2 + filterSize*2
53− movi v0.2D, #0 // val sum part 1 (for dst[0])
54− movi v1.2D, #0 // val sum part 2 (for dst[1])
55− movi v2.2D, #0 // val sum part 3 (for dst[2])
56− movi v3.2D, #0 // val sum part 4 (for dst[3])
53+ movi v0.2d, #0 // val sum part 1 (for dst[0])
54+ movi v1.2d, #0 // val sum part 2 (for dst[1])
55+ movi v2.2d, #0 // val sum part 3 (for dst[2])
56+ movi v3.2d, #0 // val sum part 4 (for dst[3])
5757 add x17, x3, w8, UXTW // srcp + filterPos[0]
5858 add x8, x3, w0, UXTW // srcp + filterPos[1]
5959 add x0, x3, w11, UXTW // srcp + filterPos[2]
6060 add x11, x3, w9, UXTW // srcp + filterPos[3]
6161 mov w15, w6 // filterSize counter
62−2: ld1 {v4.8B}, [x17], #8 // srcp[filterPos[0] + {0..7}]
63− ld1 {v5.8H}, [x16], #16 // load 8x16-bit filter values, part 1
64− ld1 {v6.8B}, [x8], #8 // srcp[filterPos[1] + {0..7}]
65− ld1 {v7.8H}, [x12], #16 // load 8x16-bit at filter+filterSize
66− uxtl v4.8H, v4.8B // unpack part 1 to 16-bit
67− smlal v0.4S, v4.4H, v5.4H // v0 accumulates srcp[filterPos[0] + {0..3}] * filter[{0..3}]
68− smlal2 v0.4S, v4.8H, v5.8H // v0 accumulates srcp[filterPos[0] + {4..7}] * filter[{4..7}]
69− ld1 {v16.8B}, [x0], #8 // srcp[filterPos[2] + {0..7}]
70− ld1 {v17.8H}, [x13], #16 // load 8x16-bit at filter+2*filterSize
71− uxtl v6.8H, v6.8B // unpack part 2 to 16-bit
72− smlal v1.4S, v6.4H, v7.4H // v1 accumulates srcp[filterPos[1] + {0..3}] * filter[{0..3}]
73− uxtl v16.8H, v16.8B // unpack part 3 to 16-bit
74− smlal v2.4S, v16.4H, v17.4H // v2 accumulates srcp[filterPos[2] + {0..3}] * filter[{0..3}]
75− smlal2 v2.4S, v16.8H, v17.8H // v2 accumulates srcp[filterPos[2] + {4..7}] * filter[{4..7}]
76− ld1 {v18.8B}, [x11], #8 // srcp[filterPos[3] + {0..7}]
77− smlal2 v1.4S, v6.8H, v7.8H // v1 accumulates srcp[filterPos[1] + {4..7}] * filter[{4..7}]
78− ld1 {v19.8H}, [x4], #16 // load 8x16-bit at filter+3*filterSize
62+2: ld1 {v4.8b}, [x17], #8 // srcp[filterPos[0] + {0..7}]
63+ ld1 {v5.8h}, [x16], #16 // load 8x16-bit filter values, part 1
64+ ld1 {v6.8b}, [x8], #8 // srcp[filterPos[1] + {0..7}]
65+ ld1 {v7.8h}, [x12], #16 // load 8x16-bit at filter+filterSize
66+ uxtl v4.8h, v4.8b // unpack part 1 to 16-bit
67+ smlal v0.4s, v4.4h, v5.4h // v0 accumulates srcp[filterPos[0] + {0..3}] * filter[{0..3}]
68+ smlal2 v0.4s, v4.8h, v5.8h // v0 accumulates srcp[filterPos[0] + {4..7}] * filter[{4..7}]
69+ ld1 {v16.8b}, [x0], #8 // srcp[filterPos[2] + {0..7}]
70+ ld1 {v17.8h}, [x13], #16 // load 8x16-bit at filter+2*filterSize
71+ uxtl v6.8h, v6.8b // unpack part 2 to 16-bit
72+ smlal v1.4s, v6.4h, v7.4h // v1 accumulates srcp[filterPos[1] + {0..3}] * filter[{0..3}]
73+ uxtl v16.8h, v16.8b // unpack part 3 to 16-bit
74+ smlal v2.4s, v16.4h, v17.4h // v2 accumulates srcp[filterPos[2] + {0..3}] * filter[{0..3}]
75+ smlal2 v2.4s, v16.8h, v17.8h // v2 accumulates srcp[filterPos[2] + {4..7}] * filter[{4..7}]
76+ ld1 {v18.8b}, [x11], #8 // srcp[filterPos[3] + {0..7}]
77+ smlal2 v1.4s, v6.8h, v7.8h // v1 accumulates srcp[filterPos[1] + {4..7}] * filter[{4..7}]
78+ ld1 {v19.8h}, [x4], #16 // load 8x16-bit at filter+3*filterSize
7979 subs w15, w15, #8 // j -= 8: processed 8/filterSize
80− uxtl v18.8H, v18.8B // unpack part 4 to 16-bit
81− smlal v3.4S, v18.4H, v19.4H // v3 accumulates srcp[filterPos[3] + {0..3}] * filter[{0..3}]
82− smlal2 v3.4S, v18.8H, v19.8H // v3 accumulates srcp[filterPos[3] + {4..7}] * filter[{4..7}]
80+ uxtl v18.8h, v18.8b // unpack part 4 to 16-bit
81+ smlal v3.4s, v18.4h, v19.4h // v3 accumulates srcp[filterPos[3] + {0..3}] * filter[{0..3}]
82+ smlal2 v3.4s, v18.8h, v19.8h // v3 accumulates srcp[filterPos[3] + {4..7}] * filter[{4..7}]
8383 b.gt 2b // inner loop if filterSize not consumed completely
84− addp v0.4S, v0.4S, v1.4S // part01 horizontal pair adding
85− addp v2.4S, v2.4S, v3.4S // part23 horizontal pair adding
86− addp v0.4S, v0.4S, v2.4S // part0123 horizontal pair adding
84+ addp v0.4s, v0.4s, v1.4s // part01 horizontal pair adding
85+ addp v2.4s, v2.4s, v3.4s // part23 horizontal pair adding
86+ addp v0.4s, v0.4s, v2.4s // part0123 horizontal pair adding
8787 subs w2, w2, #4 // dstW -= 4
88− sqshrn v0.4H, v0.4S, #7 // shift and clip the 2x16-bit final values
89− st1 {v0.4H}, [x1], #8 // write to destination part0123
88+ sqshrn v0.4h, v0.4s, #7 // shift and clip the 2x16-bit final values
89+ st1 {v0.4h}, [x1], #8 // write to destination part0123
9090 b.gt 1b // loop until end of line
9191 ret
9292 endfunc
@@ -245,7 +245,7 @@ function ff_hscale8to15_4_neon, export=1 · 152 unmodified lines
245245 stp w14, w15, [sp, #24] // *scratch_mem = { src[filterPos[idx + 6]][0..3], src[filterPos[idx + 7]][0..3] }
246246
247247 1:
248− ld4 {v16.8B, v17.8B, v18.8B, v19.8B}, [sp] // transpose 8 bytes each from src into 4 registers
248+ ld4 {v16.8b, v17.8b, v18.8b, v19.8b}, [sp] // transpose 8 bytes each from src into 4 registers
249249
250250 // load 8 values from filterPos to be used as offsets into src
251251 ldp w8, w9, [x5] // filterPos[idx + 0][0..3], [idx + 1][0..3], next iteration
@@ -253,74 +253,74 @@ function ff_hscale8to15_4_neon, export=1 · 1 unmodified line
253253 ldp w12, w13, [x5, #16] // filterPos[idx + 4][0..3], [idx + 5][0..3], next iteration
254254 ldp w14, w15, [x5, #24] // filterPos[idx + 6][0..3], [idx + 7][0..3], next iteration
255255
256− movi v0.2D, #0 // Clear madd accumulator for idx 0..3
257− movi v5.2D, #0 // Clear madd accumulator for idx 4..7
256+ movi v0.2d, #0 // Clear madd accumulator for idx 0..3
257+ movi v5.2d, #0 // Clear madd accumulator for idx 4..7
258258
259− ld4 {v1.8H, v2.8H, v3.8H, v4.8H}, [x4], #64 // load filter idx + 0..7
259+ ld4 {v1.8h, v2.8h, v3.8h, v4.8h}, [x4], #64 // load filter idx + 0..7
260260
261261 add x5, x5, #32 // advance filterPos
262262
263263 // interleaved SIMD and prefetching intended to keep ld/st and vector pipelines busy
264− uxtl v16.8H, v16.8B // unsigned extend long, covert src data to 16-bit
265− uxtl v17.8H, v17.8B // unsigned extend long, covert src data to 16-bit
264+ uxtl v16.8h, v16.8b // unsigned extend long, covert src data to 16-bit
265+ uxtl v17.8h, v17.8b // unsigned extend long, covert src data to 16-bit
266266 ldr w8, [x3, w8, UXTW] // src[filterPos[idx + 0]], next iteration
267267 ldr w9, [x3, w9, UXTW] // src[filterPos[idx + 1]], next iteration
268− uxtl v18.8H, v18.8B // unsigned extend long, covert src data to 16-bit
269− uxtl v19.8H, v19.8B // unsigned extend long, covert src data to 16-bit
268+ uxtl v18.8h, v18.8b // unsigned extend long, covert src data to 16-bit
269+ uxtl v19.8h, v19.8b // unsigned extend long, covert src data to 16-bit
270270 ldr w10, [x3, w10, UXTW] // src[filterPos[idx + 2]], next iteration
271271 ldr w11, [x3, w11, UXTW] // src[filterPos[idx + 3]], next iteration
272272
273− smlal v0.4S, v1.4H, v16.4H // multiply accumulate inner loop j = 0, idx = 0..3
274− smlal v0.4S, v2.4H, v17.4H // multiply accumulate inner loop j = 1, idx = 0..3
273+ smlal v0.4s, v1.4h, v16.4h // multiply accumulate inner loop j = 0, idx = 0..3
274+ smlal v0.4s, v2.4h, v17.4h // multiply accumulate inner loop j = 1, idx = 0..3
275275 ldr w12, [x3, w12, UXTW] // src[filterPos[idx + 4]], next iteration
276276 ldr w13, [x3, w13, UXTW] // src[filterPos[idx + 5]], next iteration
277− smlal v0.4S, v3.4H, v18.4H // multiply accumulate inner loop j = 2, idx = 0..3
278− smlal v0.4S, v4.4H, v19.4H // multiply accumulate inner loop j = 3, idx = 0..3
277+ smlal v0.4s, v3.4h, v18.4h // multiply accumulate inner loop j = 2, idx = 0..3
278+ smlal v0.4s, v4.4h, v19.4h // multiply accumulate inner loop j = 3, idx = 0..3
279279 ldr w14, [x3, w14, UXTW] // src[filterPos[idx + 6]], next iteration
280280 ldr w15, [x3, w15, UXTW] // src[filterPos[idx + 7]], next iteration
281281
282− smlal2 v5.4S, v1.8H, v16.8H // multiply accumulate inner loop j = 0, idx = 4..7
283− smlal2 v5.4S, v2.8H, v17.8H // multiply accumulate inner loop j = 1, idx = 4..7
282+ smlal2 v5.4s, v1.8h, v16.8h // multiply accumulate inner loop j = 0, idx = 4..7
283+ smlal2 v5.4s, v2.8h, v17.8h // multiply accumulate inner loop j = 1, idx = 4..7
284284 stp w8, w9, [sp] // *scratch_mem = { src[filterPos[idx + 0]][0..3], src[filterPos[idx + 1]][0..3] }
285285 stp w10, w11, [sp, #8] // *scratch_mem = { src[filterPos[idx + 2]][0..3], src[filterPos[idx + 3]][0..3] }
286− smlal2 v5.4S, v3.8H, v18.8H // multiply accumulate inner loop j = 2, idx = 4..7
287− smlal2 v5.4S, v4.8H, v19.8H // multiply accumulate inner loop j = 3, idx = 4..7
286+ smlal2 v5.4s, v3.8h, v18.8h // multiply accumulate inner loop j = 2, idx = 4..7
287+ smlal2 v5.4s, v4.8h, v19.8h // multiply accumulate inner loop j = 3, idx = 4..7
288288 stp w12, w13, [sp, #16] // *scratch_mem = { src[filterPos[idx + 4]][0..3], src[filterPos[idx + 5]][0..3] }
289289 stp w14, w15, [sp, #24] // *scratch_mem = { src[filterPos[idx + 6]][0..3], src[filterPos[idx + 7]][0..3] }
290290
291291 sub w2, w2, #8 // dstW -= 8
292− sqshrn v0.4H, v0.4S, #7 // shift and clip the 2x16-bit final values
293− sqshrn v1.4H, v5.4S, #7 // shift and clip the 2x16-bit final values
294− st1 {v0.4H, v1.4H}, [x1], #16 // write to dst[idx + 0..7]
292+ sqshrn v0.4h, v0.4s, #7 // shift and clip the 2x16-bit final values
293+ sqshrn v1.4h, v5.4s, #7 // shift and clip the 2x16-bit final values
294+ st1 {v0.4h, v1.4h}, [x1], #16 // write to dst[idx + 0..7]
295295 cmp w2, #16 // continue on main loop if there are at least 16 iterations left
296296 b.ge 1b
297297
298298 // last full iteration
299− ld4 {v16.8B, v17.8B, v18.8B, v19.8B}, [sp]
300− ld4 {v1.8H, v2.8H, v3.8H, v4.8H}, [x4], #64 // load filter idx + 0..7
299+ ld4 {v16.8b, v17.8b, v18.8b, v19.8b}, [sp]
300+ ld4 {v1.8h, v2.8h, v3.8h, v4.8h}, [x4], #64 // load filter idx + 0..7
301301
302− movi v0.2D, #0 // Clear madd accumulator for idx 0..3
303− movi v5.2D, #0 // Clear madd accumulator for idx 4..7
302+ movi v0.2d, #0 // Clear madd accumulator for idx 0..3
303+ movi v5.2d, #0 // Clear madd accumulator for idx 4..7
304304
305− uxtl v16.8H, v16.8B // unsigned extend long, covert src data to 16-bit
306− uxtl v17.8H, v17.8B // unsigned extend long, covert src data to 16-bit
307− uxtl v18.8H, v18.8B // unsigned extend long, covert src data to 16-bit
308− uxtl v19.8H, v19.8B // unsigned extend long, covert src data to 16-bit
305+ uxtl v16.8h, v16.8b // unsigned extend long, covert src data to 16-bit
306+ uxtl v17.8h, v17.8b // unsigned extend long, covert src data to 16-bit
307+ uxtl v18.8h, v18.8b // unsigned extend long, covert src data to 16-bit
308+ uxtl v19.8h, v19.8b // unsigned extend long, covert src data to 16-bit
309309
310− smlal v0.4S, v1.4H, v16.4H // multiply accumulate inner loop j = 0, idx = 0..3
311− smlal v0.4S, v2.4H, v17.4H // multiply accumulate inner loop j = 1, idx = 0..3
312− smlal v0.4S, v3.4H, v18.4H // multiply accumulate inner loop j = 2, idx = 0..3
313− smlal v0.4S, v4.4H, v19.4H // multiply accumulate inner loop j = 3, idx = 0..3
310+ smlal v0.4s, v1.4h, v16.4h // multiply accumulate inner loop j = 0, idx = 0..3
311+ smlal v0.4s, v2.4h, v17.4h // multiply accumulate inner loop j = 1, idx = 0..3
312+ smlal v0.4s, v3.4h, v18.4h // multiply accumulate inner loop j = 2, idx = 0..3
313+ smlal v0.4s, v4.4h, v19.4h // multiply accumulate inner loop j = 3, idx = 0..3
314314
315− smlal2 v5.4S, v1.8H, v16.8H // multiply accumulate inner loop j = 0, idx = 4..7
316− smlal2 v5.4S, v2.8H, v17.8H // multiply accumulate inner loop j = 1, idx = 4..7
317− smlal2 v5.4S, v3.8H, v18.8H // multiply accumulate inner loop j = 2, idx = 4..7
318− smlal2 v5.4S, v4.8H, v19.8H // multiply accumulate inner loop j = 3, idx = 4..7
315+ smlal2 v5.4s, v1.8h, v16.8h // multiply accumulate inner loop j = 0, idx = 4..7
316+ smlal2 v5.4s, v2.8h, v17.8h // multiply accumulate inner loop j = 1, idx = 4..7
317+ smlal2 v5.4s, v3.8h, v18.8h // multiply accumulate inner loop j = 2, idx = 4..7
318+ smlal2 v5.4s, v4.8h, v19.8h // multiply accumulate inner loop j = 3, idx = 4..7
319319
320320 subs w2, w2, #8 // dstW -= 8
321− sqshrn v0.4H, v0.4S, #7 // shift and clip the 2x16-bit final values
322− sqshrn v1.4H, v5.4S, #7 // shift and clip the 2x16-bit final values
323− st1 {v0.4H, v1.4H}, [x1], #16 // write to dst[idx + 0..7]
321+ sqshrn v0.4h, v0.4s, #7 // shift and clip the 2x16-bit final values
322+ sqshrn v1.4h, v5.4s, #7 // shift and clip the 2x16-bit final values
323+ st1 {v0.4h, v1.4h}, [x1], #16 // write to dst[idx + 0..7]
324324
325325 cbnz w2, 2f // if >0 iterations remain, jump to the wrap up section
326326
@@ -332,15 +332,15 @@ function ff_hscale8to15_4_neon, export=1 · 5 unmodified lines
332332 // load src
333333 ldr w8, [x5], #4 // filterPos[i]
334334 add x9, x3, w8, UXTW // calculate the address for src load
335− ld1 {v5.S}[0], [x9] // src[filterPos[i] + 0..3]
335+ ld1 {v5.s}[0], [x9] // src[filterPos[i] + 0..3]
336336 // load filter
337− ld1 {v6.4H}, [x4], #8 // filter[filterSize * i + 0..3]
337+ ld1 {v6.4h}, [x4], #8 // filter[filterSize * i + 0..3]
338338
339− uxtl v5.8H, v5.8B // unsigned exten long, convert src data to 16-bit
340− smull v0.4S, v5.4H, v6.4H // 4 iterations of src[...] * filter[...]
341− addv s0, v0.4S // add up products of src and filter values
339+ uxtl v5.8h, v5.8b // unsigned exten long, convert src data to 16-bit
340+ smull v0.4s, v5.4h, v6.4h // 4 iterations of src[...] * filter[...]
341+ addv s0, v0.4s // add up products of src and filter values
342342 sqshrn h0, s0, #7 // shift and clip the 2x16-bit final value
343− st1 {v0.H}[0], [x1], #2 // dst[i] = ...
343+ st1 {v0.h}[0], [x1], #2 // dst[i] = ...
344344 sub w2, w2, #1 // dstW--
345345 cbnz w2, 2b
346346
@@ -445,12 +445,12 @@ function ff_hscale8to19_4_neon, export=1 · 98 unmodified lines
445445 smull v5.4s, v0.4h, v28.4h
446446 smull2 v6.4s, v0.8h, v28.8h
447447 uxtl v2.8h, v2.8b
448− smlal v5.4s, v1.4h, v29.4H
449− smlal2 v6.4s, v1.8h, v29.8H
448+ smlal v5.4s, v1.4h, v29.4h
449+ smlal2 v6.4s, v1.8h, v29.8h
450450 uxtl v3.8h, v3.8b
451− smlal v5.4s, v2.4h, v30.4H
452− smlal2 v6.4s, v2.8h, v30.8H
453− smlal v5.4s, v3.4h, v31.4H
451+ smlal v5.4s, v2.4h, v30.4h
452+ smlal2 v6.4s, v2.8h, v30.8h
453+ smlal v5.4s, v3.4h, v31.4h
454454 smlal2 v6.4s, v3.8h, v31.8h
455455
456456 sshr v5.4s, v5.4s, #3
@@ -472,8 +472,8 @@ function ff_hscale8to19_4_neon, export=1 · 15 unmodified lines
472472 ld1 {v0.s}[0], [x9] // load 4 * uint8_t* into one single
473473 ld1 {v31.4h}, [x4], #8
474474 uxtl v0.8h, v0.8b
475− smull v5.4s, v0.4h, v31.4H
476− saddlv d0, v5.4S
475+ smull v5.4s, v0.4h, v31.4h
476+ saddlv d0, v5.4s
477477 sqshrn s0, d0, #3
478478 smin v0.4s, v0.4s, v18.4s
479479 st1 {v0.s}[0], [x1], #4
@@ -499,42 +499,42 @@ function ff_hscale8to19_X8_neon, export=1 · 19 unmodified lines
499499 ldr w11, [x5], #4 // filterPos[idx + 2]
500500 add x4, x13, x7 // filter3 = filter2 + filterSize*2
501501 ldr w9, [x5], #4 // filterPos[idx + 3]
502− movi v0.2D, #0 // val sum part 1 (for dst[0])
503− movi v1.2D, #0 // val sum part 2 (for dst[1])
504− movi v2.2D, #0 // val sum part 3 (for dst[2])
505− movi v3.2D, #0 // val sum part 4 (for dst[3])
502+ movi v0.2d, #0 // val sum part 1 (for dst[0])
503+ movi v1.2d, #0 // val sum part 2 (for dst[1])
504+ movi v2.2d, #0 // val sum part 3 (for dst[2])
505+ movi v3.2d, #0 // val sum part 4 (for dst[3])
506506 add x17, x3, w8, UXTW // srcp + filterPos[0]
507507 add x8, x3, w0, UXTW // srcp + filterPos[1]
508508 add x0, x3, w11, UXTW // srcp + filterPos[2]
509509 add x11, x3, w9, UXTW // srcp + filterPos[3]
510510 mov w15, w6 // filterSize counter
511−2: ld1 {v4.8B}, [x17], #8 // srcp[filterPos[0] + {0..7}]
512− ld1 {v5.8H}, [x16], #16 // load 8x16-bit filter values, part 1
513− uxtl v4.8H, v4.8B // unpack part 1 to 16-bit
514− smlal v0.4S, v4.4H, v5.4H // v0 accumulates srcp[filterPos[0] + {0..3}] * filter[{0..3}]
515− ld1 {v6.8B}, [x8], #8 // srcp[filterPos[1] + {0..7}]
516− smlal2 v0.4S, v4.8H, v5.8H // v0 accumulates srcp[filterPos[0] + {4..7}] * filter[{4..7}]
517− ld1 {v7.8H}, [x12], #16 // load 8x16-bit at filter+filterSize
518− ld1 {v16.8B}, [x0], #8 // srcp[filterPos[2] + {0..7}]
519− uxtl v6.8H, v6.8B // unpack part 2 to 16-bit
520− ld1 {v17.8H}, [x13], #16 // load 8x16-bit at filter+2*filterSize
521− uxtl v16.8H, v16.8B // unpack part 3 to 16-bit
522− smlal v1.4S, v6.4H, v7.4H // v1 accumulates srcp[filterPos[1] + {0..3}] * filter[{0..3}]
523− ld1 {v18.8B}, [x11], #8 // srcp[filterPos[3] + {0..7}]
524− smlal v2.4S, v16.4H, v17.4H // v2 accumulates srcp[filterPos[2] + {0..3}] * filter[{0..3}]
525− ld1 {v19.8H}, [x4], #16 // load 8x16-bit at filter+3*filterSize
526− smlal2 v2.4S, v16.8H, v17.8H // v2 accumulates srcp[filterPos[2] + {4..7}] * filter[{4..7}]
527− uxtl v18.8H, v18.8B // unpack part 4 to 16-bit
528− smlal2 v1.4S, v6.8H, v7.8H // v1 accumulates srcp[filterPos[1] + {4..7}] * filter[{4..7}]
529− smlal v3.4S, v18.4H, v19.4H // v3 accumulates srcp[filterPos[3] + {0..3}] * filter[{0..3}]
511+2: ld1 {v4.8b}, [x17], #8 // srcp[filterPos[0] + {0..7}]
512+ ld1 {v5.8h}, [x16], #16 // load 8x16-bit filter values, part 1
513+ uxtl v4.8h, v4.8b // unpack part 1 to 16-bit
514+ smlal v0.4s, v4.4h, v5.4h // v0 accumulates srcp[filterPos[0] + {0..3}] * filter[{0..3}]
515+ ld1 {v6.8b}, [x8], #8 // srcp[filterPos[1] + {0..7}]
516+ smlal2 v0.4s, v4.8h, v5.8h // v0 accumulates srcp[filterPos[0] + {4..7}] * filter[{4..7}]
517+ ld1 {v7.8h}, [x12], #16 // load 8x16-bit at filter+filterSize
518+ ld1 {v16.8b}, [x0], #8 // srcp[filterPos[2] + {0..7}]
519+ uxtl v6.8h, v6.8b // unpack part 2 to 16-bit
520+ ld1 {v17.8h}, [x13], #16 // load 8x16-bit at filter+2*filterSize
521+ uxtl v16.8h, v16.8b // unpack part 3 to 16-bit
522+ smlal v1.4s, v6.4h, v7.4h // v1 accumulates srcp[filterPos[1] + {0..3}] * filter[{0..3}]
523+ ld1 {v18.8b}, [x11], #8 // srcp[filterPos[3] + {0..7}]
524+ smlal v2.4s, v16.4h, v17.4h // v2 accumulates srcp[filterPos[2] + {0..3}] * filter[{0..3}]
525+ ld1 {v19.8h}, [x4], #16 // load 8x16-bit at filter+3*filterSize
526+ smlal2 v2.4s, v16.8h, v17.8h // v2 accumulates srcp[filterPos[2] + {4..7}] * filter[{4..7}]
527+ uxtl v18.8h, v18.8b // unpack part 4 to 16-bit
528+ smlal2 v1.4s, v6.8h, v7.8h // v1 accumulates srcp[filterPos[1] + {4..7}] * filter[{4..7}]
529+ smlal v3.4s, v18.4h, v19.4h // v3 accumulates srcp[filterPos[3] + {0..3}] * filter[{0..3}]
530530 subs w15, w15, #8 // j -= 8: processed 8/filterSize
531− smlal2 v3.4S, v18.8H, v19.8H // v3 accumulates srcp[filterPos[3] + {4..7}] * filter[{4..7}]
531+ smlal2 v3.4s, v18.8h, v19.8h // v3 accumulates srcp[filterPos[3] + {4..7}] * filter[{4..7}]
532532 b.gt 2b // inner loop if filterSize not consumed completely
533− addp v0.4S, v0.4S, v1.4S // part01 horizontal pair adding
534− addp v2.4S, v2.4S, v3.4S // part23 horizontal pair adding
535− addp v0.4S, v0.4S, v2.4S // part0123 horizontal pair adding
533+ addp v0.4s, v0.4s, v1.4s // part01 horizontal pair adding
534+ addp v2.4s, v2.4s, v3.4s // part23 horizontal pair adding
535+ addp v0.4s, v0.4s, v2.4s // part0123 horizontal pair adding
536536 subs w2, w2, #4 // dstW -= 4
537− sshr v0.4s, v0.4S, #3 // shift and clip the 2x16-bit final values
537+ sshr v0.4s, v0.4s, #3 // shift and clip the 2x16-bit final values
538538 smin v0.4s, v0.4s, v20.4s
539539 st1 {v0.4s}, [x1], #16 // write to destination part0123
540540 b.gt 1b // loop until end of line
@@ -588,16 +588,16 @@ function ff_hscale8to19_X4_neon, export=1 · 47 unmodified lines
588588 smlal2 v16.4s, v4.8h, v31.8h // multiplication of upper half for idx 0
589589 ldr d6, [x10], #8 // load src values for idx 2
590590 ldr q29, [x14, x16] // load filter values for idx 2
591− smlal v17.4s, v5.4h, v30.4H // multiplication of lower half for idx 1
591+ smlal v17.4s, v5.4h, v30.4h // multiplication of lower half for idx 1
592592 ldr d7, [x11], #8 // load src values for idx 3
593− smlal2 v17.4s, v5.8h, v30.8H // multiplication of upper half for idx 1
594− uxtl v6.8h, v6.8B // extend tpye to matchi the filter's size
593+ smlal2 v17.4s, v5.8h, v30.8h // multiplication of upper half for idx 1
594+ uxtl v6.8h, v6.8b // extend tpye to matchi the filter's size
595595 ldr q28, [x15, x16] // load filter values for idx 3
596596 smlal v18.4s, v6.4h, v29.4h // multiplication of lower half for idx 2
597− uxtl v7.8h, v7.8B
598− smlal2 v18.4s, v6.8h, v29.8H // multiplication of upper half for idx 2
597+ uxtl v7.8h, v7.8b
598+ smlal2 v18.4s, v6.8h, v29.8h // multiplication of upper half for idx 2
599599 sub w0, w0, #8
600− smlal v19.4s, v7.4h, v28.4H // multiplication of lower half for idx 3
600+ smlal v19.4s, v7.4h, v28.4h // multiplication of lower half for idx 3
601601 cmp w0, #8
602602 smlal2 v19.4s, v7.8h, v28.8h // multiplication of upper half for idx 3
603603 add x16, x16, #16 // advance filter values indexing
@@ -618,11 +618,11 @@ function ff_hscale8to19_X4_neon, export=1 · 14 unmodified lines
618618 uxtl v5.8h, v5.8b // extend type to match the filter' size
619619 ldr s6, [x10] // load src values for idx 2
620620 smlal v17.4s, v5.4h, v30.4h
621− uxtl v6.8h, v6.8B // extend type to match the filter's size
621+ uxtl v6.8h, v6.8b // extend type to match the filter's size
622622 ldr d29, [x14, x17] // load filter values for idx 2
623623 ldr s7, [x11] // load src values for idx 3
624624 addp v16.4s, v16.4s, v17.4s
625− uxtl v7.8h, v7.8B
625+ uxtl v7.8h, v7.8b
626626 ldr d28, [x15, x17] // load filter values for idx 3
627627 smlal v18.4s, v6.4h, v29.4h
628628 smlal v19.4s, v7.4h, v28.4h
@@ -700,31 +700,31 @@ function ff_hscale16to15_4_neon_asm, export=1 · 71 unmodified lines
700700 // Extending to 32 bits is necessary, as unit16_t values can't
701701 // be represented as int16_t without type promotion.
702702 uxtl v26.4s, v0.4h
703− sxtl v27.4s, v28.4H
703+ sxtl v27.4s, v28.4h
704704 uxtl2 v0.4s, v0.8h
705705 mul v5.4s, v26.4s, v27.4s
706− sxtl2 v28.4s, v28.8H
706+ sxtl2 v28.4s, v28.8h
707707 uxtl v26.4s, v1.4h
708708 mul v6.4s, v0.4s, v28.4s
709709
710− sxtl v27.4s, v29.4H
710+ sxtl v27.4s, v29.4h
711711 uxtl2 v0.4s, v1.8h
712712 mla v5.4s, v27.4s, v26.4s
713− sxtl2 v28.4s, v29.8H
713+ sxtl2 v28.4s, v29.8h
714714 uxtl v26.4s, v2.4h
715715 mla v6.4s, v28.4s, v0.4s
716716
717− sxtl v27.4s, v30.4H
717+ sxtl v27.4s, v30.4h
718718 uxtl2 v0.4s, v2.8h
719719 mla v5.4s, v27.4s, v26.4s
720− sxtl2 v28.4s, v30.8H
720+ sxtl2 v28.4s, v30.8h
721721 uxtl v26.4s, v3.4h
722722 mla v6.4s, v28.4s, v0.4s
723723
724− sxtl v27.4s, v31.4H
724+ sxtl v27.4s, v31.4h
725725 uxtl2 v0.4s, v3.8h
726726 mla v5.4s, v27.4s, v26.4s
727− sxtl2 v28.4s, v31.8H
727+ sxtl2 v28.4s, v31.8h
728728 sub w2, w2, #8
729729 mla v6.4s, v28.4s, v0.4s
730730
@@ -775,31 +775,31 @@ function ff_hscale16to15_4_neon_asm, export=1 · 44 unmodified lines
775775 ld4 {v28.8h, v29.8h, v30.8h, v31.8h}, [x4], #64
776776
777777 uxtl v26.4s, v0.4h
778− sxtl v27.4s, v28.4H
778+ sxtl v27.4s, v28.4h
779779 uxtl2 v0.4s, v0.8h
780780 mul v5.4s, v26.4s, v27.4s
781− sxtl2 v28.4s, v28.8H
781+ sxtl2 v28.4s, v28.8h
782782 uxtl v26.4s, v1.4h
783783 mul v6.4s, v0.4s, v28.4s
784784
785− sxtl v27.4s, v29.4H
785+ sxtl v27.4s, v29.4h
786786 uxtl2 v0.4s, v1.8h
787787 mla v5.4s, v26.4s, v27.4s
788− sxtl2 v28.4s, v29.8H
788+ sxtl2 v28.4s, v29.8h
789789 uxtl v26.4s, v2.4h
790790 mla v6.4s, v0.4s, v28.4s
791791
792− sxtl v27.4s, v30.4H
792+ sxtl v27.4s, v30.4h
793793 uxtl2 v0.4s, v2.8h
794794 mla v5.4s, v26.4s, v27.4s
795− sxtl2 v28.4s, v30.8H
795+ sxtl2 v28.4s, v30.8h
796796 uxtl v26.4s, v3.4h
797797 mla v6.4s, v0.4s, v28.4s
798798
799− sxtl v27.4s, v31.4H
799+ sxtl v27.4s, v31.4h
800800 uxtl2 v0.4s, v3.8h
801801 mla v5.4s, v26.4s, v27.4s
802− sxtl2 v28.4s, v31.8H
802+ sxtl2 v28.4s, v31.8h
803803 subs w2, w2, #8
804804 mla v6.4s, v0.4s, v28.4s
805805
@@ -807,7 +807,7 @@ function ff_hscale16to15_4_neon_asm, export=1 · 1 unmodified line
807807 sshl v6.4s, v6.4s, v17.4s
808808 smin v5.4s, v5.4s, v18.4s
809809 smin v6.4s, v6.4s, v18.4s
810− xtn v5.4h, v5.4S
810+ xtn v5.4h, v5.4s
811811 xtn2 v5.8h, v6.4s
812812
813813 st1 {v5.8h}, [x1], #16
@@ -826,7 +826,7 @@ function ff_hscale16to15_4_neon_asm, export=1 · 12 unmodified lines
826826 uxtl v0.4s, v0.4h
827827 sxtl v31.4s, v31.4h
828828 mul v5.4s, v0.4s, v31.4s
829− addv s0, v5.4S
829+ addv s0, v5.4s
830830 sshl v0.4s, v0.4s, v17.4s
831831 smin v0.4s, v0.4s, v18.4s
832832 st1 {v0.h}[0], [x1], #2
@@ -865,58 +865,58 @@ function ff_hscale16to15_X8_neon_asm, export=1 · 32 unmodified lines
865865 add x12, x16, x7 // filter1 = filter0 + filterSize*2
866866 add x13, x12, x7 // filter2 = filter1 + filterSize*2
867867 add x4, x13, x7 // filter3 = filter2 + filterSize*2
868− movi v0.2D, #0 // val sum part 1 (for dst[0])
869− movi v1.2D, #0 // val sum part 2 (for dst[1])
870− movi v2.2D, #0 // val sum part 3 (for dst[2])
871− movi v3.2D, #0 // val sum part 4 (for dst[3])
868+ movi v0.2d, #0 // val sum part 1 (for dst[0])
869+ movi v1.2d, #0 // val sum part 2 (for dst[1])
870+ movi v2.2d, #0 // val sum part 3 (for dst[2])
871+ movi v3.2d, #0 // val sum part 4 (for dst[3])
872872 add x17, x3, w8, UXTW // srcp + filterPos[0]
873873 add x8, x3, w10, UXTW // srcp + filterPos[1]
874874 add x10, x3, w11, UXTW // srcp + filterPos[2]
875875 add x11, x3, w9, UXTW // srcp + filterPos[3]
876876 mov w15, w6 // filterSize counter
877−2: ld1 {v4.8H}, [x17], #16 // srcp[filterPos[0] + {0..7}]
878− ld1 {v5.8H}, [x16], #16 // load 8x16-bit filter values, part 1
879− ld1 {v6.8H}, [x8], #16 // srcp[filterPos[1] + {0..7}]
880− ld1 {v7.8H}, [x12], #16 // load 8x16-bit at filter+filterSize
881− uxtl v24.4s, v4.4H // extend srcp lower half to 32 bits to preserve sign
882− sxtl v25.4s, v5.4H // extend filter lower half to 32 bits to match srcp size
877+2: ld1 {v4.8h}, [x17], #16 // srcp[filterPos[0] + {0..7}]
878+ ld1 {v5.8h}, [x16], #16 // load 8x16-bit filter values, part 1
879+ ld1 {v6.8h}, [x8], #16 // srcp[filterPos[1] + {0..7}]
880+ ld1 {v7.8h}, [x12], #16 // load 8x16-bit at filter+filterSize
881+ uxtl v24.4s, v4.4h // extend srcp lower half to 32 bits to preserve sign
882+ sxtl v25.4s, v5.4h // extend filter lower half to 32 bits to match srcp size
883883 uxtl2 v4.4s, v4.8h // extend srcp upper half to 32 bits
884884 mla v0.4s, v24.4s, v25.4s // multiply accumulate lower half of v4 * v5
885885 sxtl2 v5.4s, v5.8h // extend filter upper half to 32 bits
886886 uxtl v26.4s, v6.4h // extend srcp lower half to 32 bits
887− mla v0.4S, v4.4s, v5.4s // multiply accumulate upper half of v4 * v5
888− sxtl v27.4s, v7.4H // exted filter lower half
889− uxtl2 v6.4s, v6.8H // extend srcp upper half
887+ mla v0.4s, v4.4s, v5.4s // multiply accumulate upper half of v4 * v5
888+ sxtl v27.4s, v7.4h // exted filter lower half
889+ uxtl2 v6.4s, v6.8h // extend srcp upper half
890890 sxtl2 v7.4s, v7.8h // extend filter upper half
891− ld1 {v16.8H}, [x10], #16 // srcp[filterPos[2] + {0..7}]
892− mla v1.4S, v26.4s, v27.4s // v1 accumulates srcp[filterPos[1] + {0..3}] * filter[{0..3}]
893− ld1 {v17.8H}, [x13], #16 // load 8x16-bit at filter+2*filterSize
894− uxtl v22.4s, v16.4H // extend srcp lower half
895− sxtl v23.4s, v17.4H // extend filter lower half
896− uxtl2 v16.4s, v16.8H // extend srcp upper half
891+ ld1 {v16.8h}, [x10], #16 // srcp[filterPos[2] + {0..7}]
892+ mla v1.4s, v26.4s, v27.4s // v1 accumulates srcp[filterPos[1] + {0..3}] * filter[{0..3}]
893+ ld1 {v17.8h}, [x13], #16 // load 8x16-bit at filter+2*filterSize
894+ uxtl v22.4s, v16.4h // extend srcp lower half
895+ sxtl v23.4s, v17.4h // extend filter lower half
896+ uxtl2 v16.4s, v16.8h // extend srcp upper half
897897 sxtl2 v17.4s, v17.8h // extend filter upper half
898− mla v2.4S, v22.4s, v23.4s // v2 accumulates srcp[filterPos[2] + {0..3}] * filter[{0..3}]
899− mla v2.4S, v16.4s, v17.4s // v2 accumulates srcp[filterPos[2] + {4..7}] * filter[{4..7}]
900− ld1 {v18.8H}, [x11], #16 // srcp[filterPos[3] + {0..7}]
901− mla v1.4S, v6.4s, v7.4s // v1 accumulates srcp[filterPos[1] + {4..7}] * filter[{4..7}]
902− ld1 {v19.8H}, [x4], #16 // load 8x16-bit at filter+3*filterSize
898+ mla v2.4s, v22.4s, v23.4s // v2 accumulates srcp[filterPos[2] + {0..3}] * filter[{0..3}]
899+ mla v2.4s, v16.4s, v17.4s // v2 accumulates srcp[filterPos[2] + {4..7}] * filter[{4..7}]
900+ ld1 {v18.8h}, [x11], #16 // srcp[filterPos[3] + {0..7}]
901+ mla v1.4s, v6.4s, v7.4s // v1 accumulates srcp[filterPos[1] + {4..7}] * filter[{4..7}]
902+ ld1 {v19.8h}, [x4], #16 // load 8x16-bit at filter+3*filterSize
903903 subs w15, w15, #8 // j -= 8: processed 8/filterSize
904− uxtl v28.4s, v18.4H // extend srcp lower half
905− sxtl v29.4s, v19.4H // extend filter lower half
906− uxtl2 v18.4s, v18.8H // extend srcp upper half
904+ uxtl v28.4s, v18.4h // extend srcp lower half
905+ sxtl v29.4s, v19.4h // extend filter lower half
906+ uxtl2 v18.4s, v18.8h // extend srcp upper half
907907 sxtl2 v19.4s, v19.8h // extend filter upper half
908− mla v3.4S, v28.4s, v29.4s // v3 accumulates srcp[filterPos[3] + {0..3}] * filter[{0..3}]
909− mla v3.4S, v18.4s, v19.4s // v3 accumulates srcp[filterPos[3] + {4..7}] * filter[{4..7}]
908+ mla v3.4s, v28.4s, v29.4s // v3 accumulates srcp[filterPos[3] + {0..3}] * filter[{0..3}]
909+ mla v3.4s, v18.4s, v19.4s // v3 accumulates srcp[filterPos[3] + {4..7}] * filter[{4..7}]
910910 b.gt 2b // inner loop if filterSize not consumed completely
911− addp v0.4S, v0.4S, v1.4S // part01 horizontal pair adding
912− addp v2.4S, v2.4S, v3.4S // part23 horizontal pair adding
913− addp v0.4S, v0.4S, v2.4S // part0123 horizontal pair adding
911+ addp v0.4s, v0.4s, v1.4s // part01 horizontal pair adding
912+ addp v2.4s, v2.4s, v3.4s // part23 horizontal pair adding
913+ addp v0.4s, v0.4s, v2.4s // part0123 horizontal pair adding
914914 subs w2, w2, #4 // dstW -= 4
915915 sshl v0.4s, v0.4s, v21.4s // shift right (effectively rigth, as shift is negative); overflow expected
916916 smin v0.4s, v0.4s, v20.4s // apply min (do not use sqshl)
917917 xtn v0.4h, v0.4s // narrow down to 16 bits
918918
919− st1 {v0.4H}, [x1], #8 // write to destination part0123
919+ st1 {v0.4h}, [x1], #8 // write to destination part0123
920920 b.gt 1b // loop until end of line
921921 ret
922922 endfunc
@@ -1108,31 +1108,31 @@ function ff_hscale16to19_4_neon_asm, export=1 · 185 unmodified lines
11081108 // Extending to 32 bits is necessary, as unit16_t values can't
11091109 // be represented as int16_t without type promotion.
11101110 uxtl v26.4s, v0.4h
1111− sxtl v27.4s, v28.4H
1111+ sxtl v27.4s, v28.4h
11121112 uxtl2 v0.4s, v0.8h
11131113 mul v5.4s, v26.4s, v27.4s
1114− sxtl2 v28.4s, v28.8H
1114+ sxtl2 v28.4s, v28.8h
11151115 uxtl v26.4s, v1.4h
11161116 mul v6.4s, v0.4s, v28.4s
11171117
1118− sxtl v27.4s, v29.4H
1118+ sxtl v27.4s, v29.4h
11191119 uxtl2 v0.4s, v1.8h
11201120 mla v5.4s, v27.4s, v26.4s
1121− sxtl2 v28.4s, v29.8H
1121+ sxtl2 v28.4s, v29.8h
11221122 uxtl v26.4s, v2.4h
11231123 mla v6.4s, v28.4s, v0.4s
11241124
1125− sxtl v27.4s, v30.4H
1125+ sxtl v27.4s, v30.4h
11261126 uxtl2 v0.4s, v2.8h
11271127 mla v5.4s, v27.4s, v26.4s
1128− sxtl2 v28.4s, v30.8H
1128+ sxtl2 v28.4s, v30.8h
11291129 uxtl v26.4s, v3.4h
11301130 mla v6.4s, v28.4s, v0.4s
11311131
1132− sxtl v27.4s, v31.4H
1132+ sxtl v27.4s, v31.4h
11331133 uxtl2 v0.4s, v3.8h
11341134 mla v5.4s, v27.4s, v26.4s
1135− sxtl2 v28.4s, v31.8H
1135+ sxtl2 v28.4s, v31.8h
11361136 sub w2, w2, #8
11371137 mla v6.4s, v28.4s, v0.4s
11381138
@@ -1181,31 +1181,31 @@ function ff_hscale16to19_4_neon_asm, export=1 · 42 unmodified lines
11811181 ld4 {v28.8h, v29.8h, v30.8h, v31.8h}, [x4], #64
11821182
11831183 uxtl v26.4s, v0.4h
1184− sxtl v27.4s, v28.4H
1184+ sxtl v27.4s, v28.4h
11851185 uxtl2 v0.4s, v0.8h
11861186 mul v5.4s, v26.4s, v27.4s
1187− sxtl2 v28.4s, v28.8H
1187+ sxtl2 v28.4s, v28.8h
11881188 uxtl v26.4s, v1.4h
11891189 mul v6.4s, v0.4s, v28.4s
11901190
1191− sxtl v27.4s, v29.4H
1191+ sxtl v27.4s, v29.4h
11921192 uxtl2 v0.4s, v1.8h
11931193 mla v5.4s, v26.4s, v27.4s
1194− sxtl2 v28.4s, v29.8H
1194+ sxtl2 v28.4s, v29.8h
11951195 uxtl v26.4s, v2.4h
11961196 mla v6.4s, v0.4s, v28.4s
11971197
1198− sxtl v27.4s, v30.4H
1198+ sxtl v27.4s, v30.4h
11991199 uxtl2 v0.4s, v2.8h
12001200 mla v5.4s, v26.4s, v27.4s
1201− sxtl2 v28.4s, v30.8H
1201+ sxtl2 v28.4s, v30.8h
12021202 uxtl v26.4s, v3.4h
12031203 mla v6.4s, v0.4s, v28.4s
12041204
1205− sxtl v27.4s, v31.4H
1205+ sxtl v27.4s, v31.4h
12061206 uxtl2 v0.4s, v3.8h
12071207 mla v5.4s, v26.4s, v27.4s
1208− sxtl2 v28.4s, v31.8H
1208+ sxtl2 v28.4s, v31.8h
12091209 subs w2, w2, #8
12101210 mla v6.4s, v0.4s, v28.4s
12111211
@@ -1232,7 +1232,7 @@ function ff_hscale16to19_4_neon_asm, export=1 · 20 unmodified lines
12321232 sxtl v31.4s, v31.4h
12331233 subs w2, w2, #1
12341234 mul v5.4s, v0.4s, v31.4s
1235− addv s0, v5.4S
1235+ addv s0, v5.4s
12361236 sshl v0.4s, v0.4s, v17.4s
12371237 smin v0.4s, v0.4s, v18.4s
12381238 st1 {v0.s}[0], [x1], #4
@@ -1270,52 +1270,52 @@ function ff_hscale16to19_X8_neon_asm, export=1 · 31 unmodified lines
12701270 add x13, x12, x7 // filter2 = filter1 + filterSize*2
12711271 lsl w10, w10, #1
12721272 add x4, x13, x7 // filter3 = filter2 + filterSize*2
1273− movi v0.2D, #0 // val sum part 1 (for dst[0])
1274− movi v1.2D, #0 // val sum part 2 (for dst[1])
1275− movi v2.2D, #0 // val sum part 3 (for dst[2])
1276− movi v3.2D, #0 // val sum part 4 (for dst[3])
1273+ movi v0.2d, #0 // val sum part 1 (for dst[0])
1274+ movi v1.2d, #0 // val sum part 2 (for dst[1])
1275+ movi v2.2d, #0 // val sum part 3 (for dst[2])
1276+ movi v3.2d, #0 // val sum part 4 (for dst[3])
12771277 add x17, x3, w8, UXTW // srcp + filterPos[0]
12781278 add x8, x3, w10, UXTW // srcp + filterPos[1]
12791279 add x10, x3, w11, UXTW // srcp + filterPos[2]
12801280 add x11, x3, w9, UXTW // srcp + filterPos[3]
12811281 mov w15, w6 // filterSize counter
1282−2: ld1 {v4.8H}, [x17], #16 // srcp[filterPos[0] + {0..7}]
1283− ld1 {v5.8H}, [x16], #16 // load 8x16-bit filter values, part 1
1284− ld1 {v6.8H}, [x8], #16 // srcp[filterPos[1] + {0..7}]
1285− ld1 {v7.8H}, [x12], #16 // load 8x16-bit at filter+filterSize
1286− uxtl v24.4s, v4.4H // extend srcp lower half to 32 bits to preserve sign
1287− sxtl v25.4s, v5.4H // extend filter lower half to 32 bits to match srcp size
1282+2: ld1 {v4.8h}, [x17], #16 // srcp[filterPos[0] + {0..7}]
1283+ ld1 {v5.8h}, [x16], #16 // load 8x16-bit filter values, part 1
1284+ ld1 {v6.8h}, [x8], #16 // srcp[filterPos[1] + {0..7}]
1285+ ld1 {v7.8h}, [x12], #16 // load 8x16-bit at filter+filterSize
1286+ uxtl v24.4s, v4.4h // extend srcp lower half to 32 bits to preserve sign
1287+ sxtl v25.4s, v5.4h // extend filter lower half to 32 bits to match srcp size
12881288 uxtl2 v4.4s, v4.8h // extend srcp upper half to 32 bits
12891289 mla v0.4s, v24.4s, v25.4s // multiply accumulate lower half of v4 * v5
12901290 sxtl2 v5.4s, v5.8h // extend filter upper half to 32 bits
12911291 uxtl v26.4s, v6.4h // extend srcp lower half to 32 bits
1292− mla v0.4S, v4.4s, v5.4s // multiply accumulate upper half of v4 * v5
1293− sxtl v27.4s, v7.4H // exted filter lower half
1294− uxtl2 v6.4s, v6.8H // extend srcp upper half
1292+ mla v0.4s, v4.4s, v5.4s // multiply accumulate upper half of v4 * v5
1293+ sxtl v27.4s, v7.4h // exted filter lower half
1294+ uxtl2 v6.4s, v6.8h // extend srcp upper half
12951295 sxtl2 v7.4s, v7.8h // extend filter upper half
1296− ld1 {v16.8H}, [x10], #16 // srcp[filterPos[2] + {0..7}]
1297− mla v1.4S, v26.4s, v27.4s // v1 accumulates srcp[filterPos[1] + {0..3}] * filter[{0..3}]
1298− ld1 {v17.8H}, [x13], #16 // load 8x16-bit at filter+2*filterSize
1299− uxtl v22.4s, v16.4H // extend srcp lower half
1300− sxtl v23.4s, v17.4H // extend filter lower half
1301− uxtl2 v16.4s, v16.8H // extend srcp upper half
1296+ ld1 {v16.8h}, [x10], #16 // srcp[filterPos[2] + {0..7}]
1297+ mla v1.4s, v26.4s, v27.4s // v1 accumulates srcp[filterPos[1] + {0..3}] * filter[{0..3}]
1298+ ld1 {v17.8h}, [x13], #16 // load 8x16-bit at filter+2*filterSize
1299+ uxtl v22.4s, v16.4h // extend srcp lower half
1300+ sxtl v23.4s, v17.4h // extend filter lower half
1301+ uxtl2 v16.4s, v16.8h // extend srcp upper half
13021302 sxtl2 v17.4s, v17.8h // extend filter upper half
1303− mla v2.4S, v22.4s, v23.4s // v2 accumulates srcp[filterPos[2] + {0..3}] * filter[{0..3}]
1304− mla v2.4S, v16.4s, v17.4s // v2 accumulates srcp[filterPos[2] + {4..7}] * filter[{4..7}]
1305− ld1 {v18.8H}, [x11], #16 // srcp[filterPos[3] + {0..7}]
1306− mla v1.4S, v6.4s, v7.4s // v1 accumulates srcp[filterPos[1] + {4..7}] * filter[{4..7}]
1307− ld1 {v19.8H}, [x4], #16 // load 8x16-bit at filter+3*filterSize
1303+ mla v2.4s, v22.4s, v23.4s // v2 accumulates srcp[filterPos[2] + {0..3}] * filter[{0..3}]
1304+ mla v2.4s, v16.4s, v17.4s // v2 accumulates srcp[filterPos[2] + {4..7}] * filter[{4..7}]
1305+ ld1 {v18.8h}, [x11], #16 // srcp[filterPos[3] + {0..7}]
1306+ mla v1.4s, v6.4s, v7.4s // v1 accumulates srcp[filterPos[1] + {4..7}] * filter[{4..7}]
1307+ ld1 {v19.8h}, [x4], #16 // load 8x16-bit at filter+3*filterSize
13081308 subs w15, w15, #8 // j -= 8: processed 8/filterSize
1309− uxtl v28.4s, v18.4H // extend srcp lower half
1310− sxtl v29.4s, v19.4H // extend filter lower half
1311− uxtl2 v18.4s, v18.8H // extend srcp upper half
1309+ uxtl v28.4s, v18.4h // extend srcp lower half
1310+ sxtl v29.4s, v19.4h // extend filter lower half
1311+ uxtl2 v18.4s, v18.8h // extend srcp upper half
13121312 sxtl2 v19.4s, v19.8h // extend filter upper half
1313− mla v3.4S, v28.4s, v29.4s // v3 accumulates srcp[filterPos[3] + {0..3}] * filter[{0..3}]
1314− mla v3.4S, v18.4s, v19.4s // v3 accumulates srcp[filterPos[3] + {4..7}] * filter[{4..7}]
1313+ mla v3.4s, v28.4s, v29.4s // v3 accumulates srcp[filterPos[3] + {0..3}] * filter[{0..3}]
1314+ mla v3.4s, v18.4s, v19.4s // v3 accumulates srcp[filterPos[3] + {4..7}] * filter[{4..7}]
13151315 b.gt 2b // inner loop if filterSize not consumed completely
1316− addp v0.4S, v0.4S, v1.4S // part01 horizontal pair adding
1317− addp v2.4S, v2.4S, v3.4S // part23 horizontal pair adding
1318− addp v0.4S, v0.4S, v2.4S // part0123 horizontal pair adding
1316+ addp v0.4s, v0.4s, v1.4s // part01 horizontal pair adding
1317+ addp v2.4s, v2.4s, v3.4s // part23 horizontal pair adding
1318+ addp v0.4s, v0.4s, v2.4s // part0123 horizontal pair adding
13191319 subs w2, w2, #4 // dstW -= 4
13201320 sshl v0.4s, v0.4s, v21.4s // shift right (effectively rigth, as shift is negative); overflow expected
13211321 smin v0.4s, v0.4s, v20.4s // apply min (do not use sqshl)
libswscale/aarch64/output.S+75−75
@@ -29,13 +29,13 @@ function ff_yuv2planeX_8_neon, export=1 · 28 unmodified lines
2929 // x5 - const uint8_t *dither,
3030 // w6 - int offset
3131
32− ld1 {v0.8B}, [x5] // load 8x8-bit dither
32+ ld1 {v0.8b}, [x5] // load 8x8-bit dither
3333 and w6, w6, #7
3434 cbz w6, 1f // check if offsetting present
35− ext v0.8B, v0.8B, v0.8B, #3 // honor offsetting which can be 0 or 3 only
36−1: uxtl v0.8H, v0.8B // extend dither to 16-bit
37− ushll v1.4S, v0.4H, #12 // extend dither to 32-bit with left shift by 12 (part 1)
38− ushll2 v2.4S, v0.8H, #12 // extend dither to 32-bit with left shift by 12 (part 2)
35+ ext v0.8b, v0.8b, v0.8b, #3 // honor offsetting which can be 0 or 3 only
36+1: uxtl v0.8h, v0.8b // extend dither to 16-bit
37+ ushll v1.4s, v0.4h, #12 // extend dither to 32-bit with left shift by 12 (part 1)
38+ ushll2 v2.4s, v0.8h, #12 // extend dither to 32-bit with left shift by 12 (part 2)
3939 cmp w1, #8 // if filterSize == 8, branch to specialized version
4040 b.eq 6f
4141 cmp w1, #4 // if filterSize == 4, branch to specialized version
@@ -48,8 +48,8 @@ function ff_yuv2planeX_8_neon, export=1 · 6 unmodified lines
4848 mov x7, #0 // i = 0
4949 tbnz w1, #0, 4f // if filterSize % 2 != 0 branch to specialized version
5050 // fs % 2 == 0
51−2: mov v3.16B, v1.16B // initialize accumulator part 1 with dithering value
52− mov v4.16B, v2.16B // initialize accumulator part 2 with dithering value
51+2: mov v3.16b, v1.16b // initialize accumulator part 1 with dithering value
52+ mov v4.16b, v2.16b // initialize accumulator part 2 with dithering value
5353 mov w8, w1 // tmpfilterSize = filterSize
5454 mov x9, x2 // srcp = src
5555 mov x10, x0 // filterp = filter
@@ -57,12 +57,12 @@ function ff_yuv2planeX_8_neon, export=1 · 1 unmodified line
5757 ldr s7, [x10], #4 // read 2x16-bit coeff X and Y at filter[j] and filter[j+1]
5858 add x11, x11, x7, lsl #1 // &src[j ][i]
5959 add x12, x12, x7, lsl #1 // &src[j+1][i]
60− ld1 {v5.8H}, [x11] // read 8x16-bit @ src[j ][i + {0..7}]: A,B,C,D,E,F,G,H
61− ld1 {v6.8H}, [x12] // read 8x16-bit @ src[j+1][i + {0..7}]: I,J,K,L,M,N,O,P
62− smlal v3.4S, v5.4H, v7.H[0] // val0 += {A,B,C,D} * X
63− smlal2 v4.4S, v5.8H, v7.H[0] // val1 += {E,F,G,H} * X
64− smlal v3.4S, v6.4H, v7.H[1] // val0 += {I,J,K,L} * Y
65− smlal2 v4.4S, v6.8H, v7.H[1] // val1 += {M,N,O,P} * Y
60+ ld1 {v5.8h}, [x11] // read 8x16-bit @ src[j ][i + {0..7}]: A,B,C,D,E,F,G,H
61+ ld1 {v6.8h}, [x12] // read 8x16-bit @ src[j+1][i + {0..7}]: I,J,K,L,M,N,O,P
62+ smlal v3.4s, v5.4h, v7.h[0] // val0 += {A,B,C,D} * X
63+ smlal2 v4.4s, v5.8h, v7.h[0] // val1 += {E,F,G,H} * X
64+ smlal v3.4s, v6.4h, v7.h[1] // val0 += {I,J,K,L} * Y
65+ smlal2 v4.4s, v6.8h, v7.h[1] // val1 += {M,N,O,P} * Y
6666 subs w8, w8, #2 // tmpfilterSize -= 2
6767 b.gt 3b // loop until filterSize consumed
6868
@@ -77,17 +77,17 @@ function ff_yuv2planeX_8_neon, export=1 · 8 unmodified lines
7777
7878 // If filter size is odd (most likely == 1), then use this section.
7979 // fs % 2 != 0
80−4: mov v3.16B, v1.16B // initialize accumulator part 1 with dithering value
81− mov v4.16B, v2.16B // initialize accumulator part 2 with dithering value
80+4: mov v3.16b, v1.16b // initialize accumulator part 1 with dithering value
81+ mov v4.16b, v2.16b // initialize accumulator part 2 with dithering value
8282 mov w8, w1 // tmpfilterSize = filterSize
8383 mov x9, x2 // srcp = src
8484 mov x10, x0 // filterp = filter
8585 5: ldr x11, [x9], #8 // get 1 pointer: src[j]
8686 ldr h6, [x10], #2 // read 1 16 bit coeff X at filter[j]
8787 add x11, x11, x7, lsl #1 // &src[j ][i]
88− ld1 {v5.8H}, [x11] // read 8x16-bit @ src[j ][i + {0..7}]: A,B,C,D,E,F,G,H
89− smlal v3.4S, v5.4H, v6.H[0] // val0 += {A,B,C,D} * X
90− smlal2 v4.4S, v5.8H, v6.H[0] // val1 += {E,F,G,H} * X
88+ ld1 {v5.8h}, [x11] // read 8x16-bit @ src[j ][i + {0..7}]: A,B,C,D,E,F,G,H
89+ smlal v3.4s, v5.4h, v6.h[0] // val0 += {A,B,C,D} * X
90+ smlal2 v4.4s, v5.8h, v6.h[0] // val1 += {E,F,G,H} * X
9191 subs w8, w8, #1 // tmpfilterSize -= 2
9292 b.gt 5b // loop until filterSize consumed
9393
@@ -107,36 +107,36 @@ function ff_yuv2planeX_8_neon, export=1 · 13 unmodified lines
107107 ldp x12, x13, [x2, #48] // load 2 pointers: src[j+6] and src[j+7]
108108
109109 // load 8x16-bit values for filter[j], where j=0..7
110− ld1 {v6.8H}, [x0]
110+ ld1 {v6.8h}, [x0]
111111 7:
112− mov v3.16B, v1.16B // initialize accumulator part 1 with dithering value
113− mov v4.16B, v2.16B // initialize accumulator part 2 with dithering value
114−
115− ld1 {v24.8H}, [x5], #16 // load 8x16-bit values for src[j + 0][i + {0..7}]
116− ld1 {v25.8H}, [x6], #16 // load 8x16-bit values for src[j + 1][i + {0..7}]
117− ld1 {v26.8H}, [x7], #16 // load 8x16-bit values for src[j + 2][i + {0..7}]
118− ld1 {v27.8H}, [x9], #16 // load 8x16-bit values for src[j + 3][i + {0..7}]
119− ld1 {v28.8H}, [x10], #16 // load 8x16-bit values for src[j + 4][i + {0..7}]
120− ld1 {v29.8H}, [x11], #16 // load 8x16-bit values for src[j + 5][i + {0..7}]
121− ld1 {v30.8H}, [x12], #16 // load 8x16-bit values for src[j + 6][i + {0..7}]
122− ld1 {v31.8H}, [x13], #16 // load 8x16-bit values for src[j + 7][i + {0..7}]
123−
124− smlal v3.4S, v24.4H, v6.H[0] // val0 += src[0][i + {0..3}] * filter[0]
125− smlal2 v4.4S, v24.8H, v6.H[0] // val1 += src[0][i + {4..7}] * filter[0]
126− smlal v3.4S, v25.4H, v6.H[1] // val0 += src[1][i + {0..3}] * filter[1]
127− smlal2 v4.4S, v25.8H, v6.H[1] // val1 += src[1][i + {4..7}] * filter[1]
128− smlal v3.4S, v26.4H, v6.H[2] // val0 += src[2][i + {0..3}] * filter[2]
129− smlal2 v4.4S, v26.8H, v6.H[2] // val1 += src[2][i + {4..7}] * filter[2]
130− smlal v3.4S, v27.4H, v6.H[3] // val0 += src[3][i + {0..3}] * filter[3]
131− smlal2 v4.4S, v27.8H, v6.H[3] // val1 += src[3][i + {4..7}] * filter[3]
132− smlal v3.4S, v28.4H, v6.H[4] // val0 += src[4][i + {0..3}] * filter[4]
133− smlal2 v4.4S, v28.8H, v6.H[4] // val1 += src[4][i + {4..7}] * filter[4]
134− smlal v3.4S, v29.4H, v6.H[5] // val0 += src[5][i + {0..3}] * filter[5]
135− smlal2 v4.4S, v29.8H, v6.H[5] // val1 += src[5][i + {4..7}] * filter[5]
136− smlal v3.4S, v30.4H, v6.H[6] // val0 += src[6][i + {0..3}] * filter[6]
137− smlal2 v4.4S, v30.8H, v6.H[6] // val1 += src[6][i + {4..7}] * filter[6]
138− smlal v3.4S, v31.4H, v6.H[7] // val0 += src[7][i + {0..3}] * filter[7]
139− smlal2 v4.4S, v31.8H, v6.H[7] // val1 += src[7][i + {4..7}] * filter[7]
112+ mov v3.16b, v1.16b // initialize accumulator part 1 with dithering value
113+ mov v4.16b, v2.16b // initialize accumulator part 2 with dithering value
114+
115+ ld1 {v24.8h}, [x5], #16 // load 8x16-bit values for src[j + 0][i + {0..7}]
116+ ld1 {v25.8h}, [x6], #16 // load 8x16-bit values for src[j + 1][i + {0..7}]
117+ ld1 {v26.8h}, [x7], #16 // load 8x16-bit values for src[j + 2][i + {0..7}]
118+ ld1 {v27.8h}, [x9], #16 // load 8x16-bit values for src[j + 3][i + {0..7}]
119+ ld1 {v28.8h}, [x10], #16 // load 8x16-bit values for src[j + 4][i + {0..7}]
120+ ld1 {v29.8h}, [x11], #16 // load 8x16-bit values for src[j + 5][i + {0..7}]
121+ ld1 {v30.8h}, [x12], #16 // load 8x16-bit values for src[j + 6][i + {0..7}]
122+ ld1 {v31.8h}, [x13], #16 // load 8x16-bit values for src[j + 7][i + {0..7}]
123+
124+ smlal v3.4s, v24.4h, v6.h[0] // val0 += src[0][i + {0..3}] * filter[0]
125+ smlal2 v4.4s, v24.8h, v6.h[0] // val1 += src[0][i + {4..7}] * filter[0]
126+ smlal v3.4s, v25.4h, v6.h[1] // val0 += src[1][i + {0..3}] * filter[1]
127+ smlal2 v4.4s, v25.8h, v6.h[1] // val1 += src[1][i + {4..7}] * filter[1]
128+ smlal v3.4s, v26.4h, v6.h[2] // val0 += src[2][i + {0..3}] * filter[2]
129+ smlal2 v4.4s, v26.8h, v6.h[2] // val1 += src[2][i + {4..7}] * filter[2]
130+ smlal v3.4s, v27.4h, v6.h[3] // val0 += src[3][i + {0..3}] * filter[3]
131+ smlal2 v4.4s, v27.8h, v6.h[3] // val1 += src[3][i + {4..7}] * filter[3]
132+ smlal v3.4s, v28.4h, v6.h[4] // val0 += src[4][i + {0..3}] * filter[4]
133+ smlal2 v4.4s, v28.8h, v6.h[4] // val1 += src[4][i + {4..7}] * filter[4]
134+ smlal v3.4s, v29.4h, v6.h[5] // val0 += src[5][i + {0..3}] * filter[5]
135+ smlal2 v4.4s, v29.8h, v6.h[5] // val1 += src[5][i + {4..7}] * filter[5]
136+ smlal v3.4s, v30.4h, v6.h[6] // val0 += src[6][i + {0..3}] * filter[6]
137+ smlal2 v4.4s, v30.8h, v6.h[6] // val1 += src[6][i + {4..7}] * filter[6]
138+ smlal v3.4s, v31.4h, v6.h[7] // val0 += src[7][i + {0..3}] * filter[7]
139+ smlal2 v4.4s, v31.8h, v6.h[7] // val1 += src[7][i + {4..7}] * filter[7]
140140
141141 sqshrun v3.4h, v3.4s, #16 // clip16(val0>>16)
142142 sqshrun2 v3.8h, v4.4s, #16 // clip16(val1>>16)
@@ -151,24 +151,24 @@ function ff_yuv2planeX_8_neon, export=1 · 8 unmodified lines
151151 ldp x7, x9, [x2, #16] // load 2 pointers: src[j+2] and src[j+3]
152152
153153 // load 4x16-bit values for filter[j], where j=0..3 and replicated across lanes
154− ld1 {v6.4H}, [x0]
154+ ld1 {v6.4h}, [x0]
155155 9:
156− mov v3.16B, v1.16B // initialize accumulator part 1 with dithering value
157− mov v4.16B, v2.16B // initialize accumulator part 2 with dithering value
158−
159− ld1 {v24.8H}, [x5], #16 // load 8x16-bit values for src[j + 0][i + {0..7}]
160− ld1 {v25.8H}, [x6], #16 // load 8x16-bit values for src[j + 1][i + {0..7}]
161− ld1 {v26.8H}, [x7], #16 // load 8x16-bit values for src[j + 2][i + {0..7}]
162− ld1 {v27.8H}, [x9], #16 // load 8x16-bit values for src[j + 3][i + {0..7}]
163−
164− smlal v3.4S, v24.4H, v6.H[0] // val0 += src[0][i + {0..3}] * filter[0]
165− smlal2 v4.4S, v24.8H, v6.H[0] // val1 += src[0][i + {4..7}] * filter[0]
166− smlal v3.4S, v25.4H, v6.H[1] // val0 += src[1][i + {0..3}] * filter[1]
167− smlal2 v4.4S, v25.8H, v6.H[1] // val1 += src[1][i + {4..7}] * filter[1]
168− smlal v3.4S, v26.4H, v6.H[2] // val0 += src[2][i + {0..3}] * filter[2]
169− smlal2 v4.4S, v26.8H, v6.H[2] // val1 += src[2][i + {4..7}] * filter[2]
170− smlal v3.4S, v27.4H, v6.H[3] // val0 += src[3][i + {0..3}] * filter[3]
171− smlal2 v4.4S, v27.8H, v6.H[3] // val1 += src[3][i + {4..7}] * filter[3]
156+ mov v3.16b, v1.16b // initialize accumulator part 1 with dithering value
157+ mov v4.16b, v2.16b // initialize accumulator part 2 with dithering value
158+
159+ ld1 {v24.8h}, [x5], #16 // load 8x16-bit values for src[j + 0][i + {0..7}]
160+ ld1 {v25.8h}, [x6], #16 // load 8x16-bit values for src[j + 1][i + {0..7}]
161+ ld1 {v26.8h}, [x7], #16 // load 8x16-bit values for src[j + 2][i + {0..7}]
162+ ld1 {v27.8h}, [x9], #16 // load 8x16-bit values for src[j + 3][i + {0..7}]
163+
164+ smlal v3.4s, v24.4h, v6.h[0] // val0 += src[0][i + {0..3}] * filter[0]
165+ smlal2 v4.4s, v24.8h, v6.h[0] // val1 += src[0][i + {4..7}] * filter[0]
166+ smlal v3.4s, v25.4h, v6.h[1] // val0 += src[1][i + {0..3}] * filter[1]
167+ smlal2 v4.4s, v25.8h, v6.h[1] // val1 += src[1][i + {4..7}] * filter[1]
168+ smlal v3.4s, v26.4h, v6.h[2] // val0 += src[2][i + {0..3}] * filter[2]
169+ smlal2 v4.4s, v26.8h, v6.h[2] // val1 += src[2][i + {4..7}] * filter[2]
170+ smlal v3.4s, v27.4h, v6.h[3] // val0 += src[3][i + {0..3}] * filter[3]
171+ smlal2 v4.4s, v27.8h, v6.h[3] // val1 += src[3][i + {4..7}] * filter[3]
172172
173173 sqshrun v3.4h, v3.4s, #16 // clip16(val0>>16)
174174 sqshrun2 v3.8h, v4.4s, #16 // clip16(val1>>16)
@@ -184,16 +184,16 @@ function ff_yuv2planeX_8_neon, export=1 · 9 unmodified lines
184184 // load 2x16-bit values for filter[j], where j=0..1 and replicated across lanes
185185 ldr s6, [x0]
186186 11:
187− mov v3.16B, v1.16B // initialize accumulator part 1 with dithering value
188− mov v4.16B, v2.16B // initialize accumulator part 2 with dithering value
187+ mov v3.16b, v1.16b // initialize accumulator part 1 with dithering value
188+ mov v4.16b, v2.16b // initialize accumulator part 2 with dithering value
189189
190− ld1 {v24.8H}, [x5], #16 // load 8x16-bit values for src[j + 0][i + {0..7}]
191− ld1 {v25.8H}, [x6], #16 // load 8x16-bit values for src[j + 1][i + {0..7}]
190+ ld1 {v24.8h}, [x5], #16 // load 8x16-bit values for src[j + 0][i + {0..7}]
191+ ld1 {v25.8h}, [x6], #16 // load 8x16-bit values for src[j + 1][i + {0..7}]
192192
193− smlal v3.4S, v24.4H, v6.H[0] // val0 += src[0][i + {0..3}] * filter[0]
194− smlal2 v4.4S, v24.8H, v6.H[0] // val1 += src[0][i + {4..7}] * filter[0]
195− smlal v3.4S, v25.4H, v6.H[1] // val0 += src[1][i + {0..3}] * filter[1]
196− smlal2 v4.4S, v25.8H, v6.H[1] // val1 += src[1][i + {4..7}] * filter[1]
193+ smlal v3.4s, v24.4h, v6.h[0] // val0 += src[0][i + {0..3}] * filter[0]
194+ smlal2 v4.4s, v24.8h, v6.h[0] // val1 += src[0][i + {4..7}] * filter[0]
195+ smlal v3.4s, v25.4h, v6.h[1] // val0 += src[1][i + {0..3}] * filter[1]
196+ smlal2 v4.4s, v25.8h, v6.h[1] // val1 += src[1][i + {4..7}] * filter[1]
197197
198198 sqshrun v3.4h, v3.4s, #16 // clip16(val0>>16)
199199 sqshrun2 v3.8h, v4.4s, #16 // clip16(val1>>16)
@@ -210,11 +210,11 @@ function ff_yuv2plane1_8_neon, export=1 · 10 unmodified lines
210210 // w2 - int dstW,
211211 // x3 - const uint8_t *dither,
212212 // w4 - int offset
213− ld1 {v0.8B}, [x3] // load 8x8-bit dither
213+ ld1 {v0.8b}, [x3] // load 8x8-bit dither
214214 and w4, w4, #7
215215 cbz w4, 1f // check if offsetting present
216− ext v0.8B, v0.8B, v0.8B, #3 // honor offsetting which can be 0 or 3 only
217−1: uxtl v0.8H, v0.8B // extend dither to 32-bit
216+ ext v0.8b, v0.8b, v0.8b, #3 // honor offsetting which can be 0 or 3 only
217+1: uxtl v0.8h, v0.8b // extend dither to 32-bit
218218 uxtl v1.4s, v0.4h
219219 uxtl2 v2.4s, v0.8h
220220 2:
libswscale/aarch64/yuv2rgb_neon.S+58−58
@@ -33,9 +33,9 @@ · 32 unmodified lines
3333 .macro load_args_nv12
3434 ldr x8, [sp] // table
3535 load_yoff_ycoeff 8, 16 // y_offset, y_coeff
36− ld1 {v1.1D}, [x8]
37− dup v0.8H, w10
38− dup v3.8H, w9
36+ ld1 {v1.1d}, [x8]
37+ dup v0.8h, w10
38+ dup v3.8h, w9
3939 sub w3, w3, w0, lsl #2 // w3 = linesize - width * 4 (padding)
4040 sub w5, w5, w0 // w5 = linesizeY - width (paddingY)
4141 sub w7, w7, w0 // w7 = linesizeC - width (paddingC)
@@ -51,9 +51,9 @@ · 9 unmodified lines
5151 ldr w14, [sp, #8] // linesizeV
5252 ldr x8, [sp, #16] // table
5353 load_yoff_ycoeff 24, 32 // y_offset, y_coeff
54− ld1 {v1.1D}, [x8]
55− dup v0.8H, w10
56− dup v3.8H, w9
54+ ld1 {v1.1d}, [x8]
55+ dup v0.8h, w10
56+ dup v3.8h, w9
5757 sub w3, w3, w0, lsl #2 // w3 = linesize - width * 4 (padding)
5858 sub w5, w5, w0 // w5 = linesizeY - width (paddingY)
5959 sub w7, w7, w0, lsr #1 // w7 = linesizeU - width / 2 (paddingU)
@@ -67,9 +67,9 @@ · 7 unmodified lines
6767 ldr w14, [sp, #8] // linesizeV
6868 ldr x8, [sp, #16] // table
6969 load_yoff_ycoeff 24, 32 // y_offset, y_coeff
70− ld1 {v1.1D}, [x8]
71− dup v0.8H, w10
72− dup v3.8H, w9
70+ ld1 {v1.1d}, [x8]
71+ dup v0.8h, w10
72+ dup v3.8h, w9
7373 sub w3, w3, w0, lsl #2 // w3 = linesize - width * 4 (padding)
7474 sub w5, w5, w0 // w5 = linesizeY - width (paddingY)
7575 sub w7, w7, w0, lsr #1 // w7 = linesizeU - width / 2 (paddingU)
@@ -77,22 +77,22 @@ · 1 unmodified line
7777 .endm
7878
7979 .macro load_chroma_nv12
80− ld2 {v16.8B, v17.8B}, [x6], #16
81− ushll v18.8H, v16.8B, #3
82− ushll v19.8H, v17.8B, #3
80+ ld2 {v16.8b, v17.8b}, [x6], #16
81+ ushll v18.8h, v16.8b, #3
82+ ushll v19.8h, v17.8b, #3
8383 .endm
8484
8585 .macro load_chroma_nv21
86− ld2 {v16.8B, v17.8B}, [x6], #16
87− ushll v19.8H, v16.8B, #3
88− ushll v18.8H, v17.8B, #3
86+ ld2 {v16.8b, v17.8b}, [x6], #16
87+ ushll v19.8h, v16.8b, #3
88+ ushll v18.8h, v17.8b, #3
8989 .endm
9090
9191 .macro load_chroma_yuv420p
92− ld1 {v16.8B}, [ x6], #8
93− ld1 {v17.8B}, [x13], #8
94− ushll v18.8H, v16.8B, #3
95− ushll v19.8H, v17.8B, #3
92+ ld1 {v16.8b}, [ x6], #8
93+ ld1 {v17.8b}, [x13], #8
94+ ushll v18.8h, v16.8b, #3
95+ ushll v19.8h, v17.8b, #3
9696 .endm
9797
9898 .macro load_chroma_yuv422p
@@ -123,18 +123,18 @@ · 24 unmodified lines
123123 .endm
124124
125125 .macro compute_rgba r1 g1 b1 a1 r2 g2 b2 a2
126− add v20.8H, v26.8H, v20.8H // Y1 + R1
127− add v21.8H, v27.8H, v21.8H // Y2 + R2
128− add v22.8H, v26.8H, v22.8H // Y1 + G1
129− add v23.8H, v27.8H, v23.8H // Y2 + G2
130− add v24.8H, v26.8H, v24.8H // Y1 + B1
131− add v25.8H, v27.8H, v25.8H // Y2 + B2
132− sqrshrun \r1, v20.8H, #1 // clip_u8((Y1 + R1) >> 1)
133− sqrshrun \r2, v21.8H, #1 // clip_u8((Y2 + R1) >> 1)
134− sqrshrun \g1, v22.8H, #1 // clip_u8((Y1 + G1) >> 1)
135− sqrshrun \g2, v23.8H, #1 // clip_u8((Y2 + G1) >> 1)
136− sqrshrun \b1, v24.8H, #1 // clip_u8((Y1 + B1) >> 1)
137− sqrshrun \b2, v25.8H, #1 // clip_u8((Y2 + B1) >> 1)
126+ add v20.8h, v26.8h, v20.8h // Y1 + R1
127+ add v21.8h, v27.8h, v21.8h // Y2 + R2
128+ add v22.8h, v26.8h, v22.8h // Y1 + G1
129+ add v23.8h, v27.8h, v23.8h // Y2 + G2
130+ add v24.8h, v26.8h, v24.8h // Y1 + B1
131+ add v25.8h, v27.8h, v25.8h // Y2 + B2
132+ sqrshrun \r1, v20.8h, #1 // clip_u8((Y1 + R1) >> 1)
133+ sqrshrun \r2, v21.8h, #1 // clip_u8((Y2 + R1) >> 1)
134+ sqrshrun \g1, v22.8h, #1 // clip_u8((Y1 + G1) >> 1)
135+ sqrshrun \g2, v23.8h, #1 // clip_u8((Y2 + G1) >> 1)
136+ sqrshrun \b1, v24.8h, #1 // clip_u8((Y1 + B1) >> 1)
137+ sqrshrun \b2, v25.8h, #1 // clip_u8((Y2 + B1) >> 1)
138138 movi \a1, #255
139139 movi \a2, #255
140140 .endm
@@ -146,47 +146,47 @@ function ff_\ifmt\()_to_\ofmt\()_neon, export=1 · 5 unmodified lines
146146 1:
147147 mov w8, w0 // w8 = width
148148 2:
149− movi v5.8H, #4, lsl #8 // 128 * (1<<3)
149+ movi v5.8h, #4, lsl #8 // 128 * (1<<3)
150150 load_chroma_\ifmt
151− sub v18.8H, v18.8H, v5.8H // U*(1<<3) - 128*(1<<3)
152− sub v19.8H, v19.8H, v5.8H // V*(1<<3) - 128*(1<<3)
153− sqdmulh v20.8H, v19.8H, v1.H[0] // V * v2r (R)
154− sqdmulh v22.8H, v18.8H, v1.H[1] // U * u2g
155− sqdmulh v19.8H, v19.8H, v1.H[2] // V * v2g
156− add v22.8H, v22.8H, v19.8H // U * u2g + V * v2g (G)
157− sqdmulh v24.8H, v18.8H, v1.H[3] // U * u2b (B)
158− zip2 v21.8H, v20.8H, v20.8H // R2
159− zip1 v20.8H, v20.8H, v20.8H // R1
160− zip2 v23.8H, v22.8H, v22.8H // G2
161− zip1 v22.8H, v22.8H, v22.8H // G1
162− zip2 v25.8H, v24.8H, v24.8H // B2
163− zip1 v24.8H, v24.8H, v24.8H // B1
164− ld1 {v2.16B}, [x4], #16 // load luma
165− ushll v26.8H, v2.8B, #3 // Y1*(1<<3)
166− ushll2 v27.8H, v2.16B, #3 // Y2*(1<<3)
167− sub v26.8H, v26.8H, v3.8H // Y1*(1<<3) - y_offset
168− sub v27.8H, v27.8H, v3.8H // Y2*(1<<3) - y_offset
169− sqdmulh v26.8H, v26.8H, v0.8H // ((Y1*(1<<3) - y_offset) * y_coeff) >> 15
170− sqdmulh v27.8H, v27.8H, v0.8H // ((Y2*(1<<3) - y_offset) * y_coeff) >> 15
151+ sub v18.8h, v18.8h, v5.8h // U*(1<<3) - 128*(1<<3)
152+ sub v19.8h, v19.8h, v5.8h // V*(1<<3) - 128*(1<<3)
153+ sqdmulh v20.8h, v19.8h, v1.h[0] // V * v2r (R)
154+ sqdmulh v22.8h, v18.8h, v1.h[1] // U * u2g
155+ sqdmulh v19.8h, v19.8h, v1.h[2] // V * v2g
156+ add v22.8h, v22.8h, v19.8h // U * u2g + V * v2g (G)
157+ sqdmulh v24.8h, v18.8h, v1.h[3] // U * u2b (B)
158+ zip2 v21.8h, v20.8h, v20.8h // R2
159+ zip1 v20.8h, v20.8h, v20.8h // R1
160+ zip2 v23.8h, v22.8h, v22.8h // G2
161+ zip1 v22.8h, v22.8h, v22.8h // G1
162+ zip2 v25.8h, v24.8h, v24.8h // B2
163+ zip1 v24.8h, v24.8h, v24.8h // B1
164+ ld1 {v2.16b}, [x4], #16 // load luma
165+ ushll v26.8h, v2.8b, #3 // Y1*(1<<3)
166+ ushll2 v27.8h, v2.16b, #3 // Y2*(1<<3)
167+ sub v26.8h, v26.8h, v3.8h // Y1*(1<<3) - y_offset
168+ sub v27.8h, v27.8h, v3.8h // Y2*(1<<3) - y_offset
169+ sqdmulh v26.8h, v26.8h, v0.8h // ((Y1*(1<<3) - y_offset) * y_coeff) >> 15
170+ sqdmulh v27.8h, v27.8h, v0.8h // ((Y2*(1<<3) - y_offset) * y_coeff) >> 15
171171
172172 .ifc \ofmt,argb // 1 2 3 0
173− compute_rgba v5.8B,v6.8B,v7.8B,v4.8B, v17.8B,v18.8B,v19.8B,v16.8B
173+ compute_rgba v5.8b,v6.8b,v7.8b,v4.8b, v17.8b,v18.8b,v19.8b,v16.8b
174174 .endif
175175
176176 .ifc \ofmt,rgba // 0 1 2 3
177− compute_rgba v4.8B,v5.8B,v6.8B,v7.8B, v16.8B,v17.8B,v18.8B,v19.8B
177+ compute_rgba v4.8b,v5.8b,v6.8b,v7.8b, v16.8b,v17.8b,v18.8b,v19.8b
178178 .endif
179179
180180 .ifc \ofmt,abgr // 3 2 1 0
181− compute_rgba v7.8B,v6.8B,v5.8B,v4.8B, v19.8B,v18.8B,v17.8B,v16.8B
181+ compute_rgba v7.8b,v6.8b,v5.8b,v4.8b, v19.8b,v18.8b,v17.8b,v16.8b
182182 .endif
183183
184184 .ifc \ofmt,bgra // 2 1 0 3
185− compute_rgba v6.8B,v5.8B,v4.8B,v7.8B, v18.8B,v17.8B,v16.8B,v19.8B
185+ compute_rgba v6.8b,v5.8b,v4.8b,v7.8b, v18.8b,v17.8b,v16.8b,v19.8b
186186 .endif
187187
188− st4 { v4.8B, v5.8B, v6.8B, v7.8B}, [x2], #32
189− st4 {v16.8B,v17.8B,v18.8B,v19.8B}, [x2], #32
188+ st4 { v4.8b, v5.8b, v6.8b, v7.8b}, [x2], #32
189+ st4 {v16.8b,v17.8b,v18.8b,v19.8b}, [x2], #32
190190 subs w8, w8, #16 // width -= 16
191191 b.gt 2b
192192 add x2, x2, w3, SXTW // dst += padding

Keyboard shortcuts

General

Show keyboard shortcuts
?
Focus the current search
/

Issues and pull requests

Open filters
f

Issues, pull requests and Inbox

Select next item
j↓
Select previous item
k↑
Open selected item
Enter
Dismiss filters or clear selection
Esc

Repository code

Go to file
t