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
19#include "libavutil/aarch64/asm.S"19#include "libavutil/aarch64/asm.S"
20 20
21function ff_ps_add_squares_neon, export=121function ff_ps_add_squares_neon, export=1
221: ld1 {v0.4S,v1.4S}, [x1], #32221: ld1 {v0.4s,v1.4s}, [x1], #32
23 fmul v0.4S, v0.4S, v0.4S23 fmul v0.4s, v0.4s, v0.4s
24 fmul v1.4S, v1.4S, v1.4S24 fmul v1.4s, v1.4s, v1.4s
25 faddp v2.4S, v0.4S, v1.4S25 faddp v2.4s, v0.4s, v1.4s
26 ld1 {v3.4S}, [x0]26 ld1 {v3.4s}, [x0]
27 fadd v3.4S, v3.4S, v2.4S27 fadd v3.4s, v3.4s, v2.4s
28 st1 {v3.4S}, [x0], #1628 st1 {v3.4s}, [x0], #16
29 subs w2, w2, #429 subs w2, w2, #4
30 b.gt 1b30 b.gt 1b
31 ret31 ret
32endfunc32endfunc
33 33
34function ff_ps_mul_pair_single_neon, export=134function ff_ps_mul_pair_single_neon, export=1
351: ld1 {v0.4S,v1.4S}, [x1], #32351: ld1 {v0.4s,v1.4s}, [x1], #32
36 ld1 {v2.4S}, [x2], #1636 ld1 {v2.4s}, [x2], #16
37 zip1 v3.4S, v2.4S, v2.4S37 zip1 v3.4s, v2.4s, v2.4s
38 zip2 v4.4S, v2.4S, v2.4S38 zip2 v4.4s, v2.4s, v2.4s
39 fmul v0.4S, v0.4S, v3.4S39 fmul v0.4s, v0.4s, v3.4s
40 fmul v1.4S, v1.4S, v4.4S40 fmul v1.4s, v1.4s, v4.4s
41 st1 {v0.4S,v1.4S}, [x0], #3241 st1 {v0.4s,v1.4s}, [x0], #32
42 subs w3, w3, #442 subs w3, w3, #4
43 b.gt 1b43 b.gt 1b
44 ret44 ret
45endfunc45endfunc
46 46
47function ff_ps_stereo_interpolate_neon, export=147function ff_ps_stereo_interpolate_neon, export=1
48 ld1 {v0.4S}, [x2]48 ld1 {v0.4s}, [x2]
49 ld1 {v1.4S}, [x3]49 ld1 {v1.4s}, [x3]
50 zip1 v4.4S, v0.4S, v0.4S50 zip1 v4.4s, v0.4s, v0.4s
51 zip2 v5.4S, v0.4S, v0.4S51 zip2 v5.4s, v0.4s, v0.4s
52 zip1 v6.4S, v1.4S, v1.4S52 zip1 v6.4s, v1.4s, v1.4s
53 zip2 v7.4S, v1.4S, v1.4S53 zip2 v7.4s, v1.4s, v1.4s
541: ld1 {v2.2S}, [x0]541: ld1 {v2.2s}, [x0]
55 ld1 {v3.2S}, [x1]55 ld1 {v3.2s}, [x1]
56 fadd v4.4S, v4.4S, v6.4S56 fadd v4.4s, v4.4s, v6.4s
57 fadd v5.4S, v5.4S, v7.4S57 fadd v5.4s, v5.4s, v7.4s
58 mov v2.D[1], v2.D[0]58 mov v2.d[1], v2.d[0]
59 mov v3.D[1], v3.D[0]59 mov v3.d[1], v3.d[0]
60 fmul v2.4S, v2.4S, v4.4S60 fmul v2.4s, v2.4s, v4.4s
61 fmla v2.4S, v3.4S, v5.4S61 fmla v2.4s, v3.4s, v5.4s
62 st1 {v2.D}[0], [x0], #862 st1 {v2.d}[0], [x0], #8
63 st1 {v2.D}[1], [x1], #863 st1 {v2.d}[1], [x1], #8
64 subs w4, w4, #164 subs w4, w4, #1
65 b.gt 1b65 b.gt 1b
66 ret66 ret
67endfunc67endfunc
68 68
69function ff_ps_stereo_interpolate_ipdopd_neon, export=169function ff_ps_stereo_interpolate_ipdopd_neon, export=1
70 ld1 {v0.4S,v1.4S}, [x2]70 ld1 {v0.4s,v1.4s}, [x2]
71 ld1 {v6.4S,v7.4S}, [x3]71 ld1 {v6.4s,v7.4s}, [x3]
72 fneg v2.4S, v1.4S72 fneg v2.4s, v1.4s
73 fneg v3.4S, v7.4S73 fneg v3.4s, v7.4s
74 zip1 v16.4S, v0.4S, v0.4S74 zip1 v16.4s, v0.4s, v0.4s
75 zip2 v17.4S, v0.4S, v0.4S75 zip2 v17.4s, v0.4s, v0.4s
76 zip1 v18.4S, v2.4S, v1.4S76 zip1 v18.4s, v2.4s, v1.4s
77 zip2 v19.4S, v2.4S, v1.4S77 zip2 v19.4s, v2.4s, v1.4s
78 zip1 v20.4S, v6.4S, v6.4S78 zip1 v20.4s, v6.4s, v6.4s
79 zip2 v21.4S, v6.4S, v6.4S79 zip2 v21.4s, v6.4s, v6.4s
80 zip1 v22.4S, v3.4S, v7.4S80 zip1 v22.4s, v3.4s, v7.4s
81 zip2 v23.4S, v3.4S, v7.4S81 zip2 v23.4s, v3.4s, v7.4s
821: ld1 {v2.2S}, [x0]821: ld1 {v2.2s}, [x0]
83 ld1 {v3.2S}, [x1]83 ld1 {v3.2s}, [x1]
84 fadd v16.4S, v16.4S, v20.4S84 fadd v16.4s, v16.4s, v20.4s
85 fadd v17.4S, v17.4S, v21.4S85 fadd v17.4s, v17.4s, v21.4s
86 mov v2.D[1], v2.D[0]86 mov v2.d[1], v2.d[0]
87 mov v3.D[1], v3.D[0]87 mov v3.d[1], v3.d[0]
88 fmul v4.4S, v2.4S, v16.4S88 fmul v4.4s, v2.4s, v16.4s
89 fmla v4.4S, v3.4S, v17.4S89 fmla v4.4s, v3.4s, v17.4s
90 fadd v18.4S, v18.4S, v22.4S90 fadd v18.4s, v18.4s, v22.4s
91 fadd v19.4S, v19.4S, v23.4S91 fadd v19.4s, v19.4s, v23.4s
92 ext v2.16B, v2.16B, v2.16B, #492 ext v2.16b, v2.16b, v2.16b, #4
93 ext v3.16B, v3.16B, v3.16B, #493 ext v3.16b, v3.16b, v3.16b, #4
94 fmla v4.4S, v2.4S, v18.4S94 fmla v4.4s, v2.4s, v18.4s
95 fmla v4.4S, v3.4S, v19.4S95 fmla v4.4s, v3.4s, v19.4s
96 st1 {v4.D}[0], [x0], #896 st1 {v4.d}[0], [x0], #8
97 st1 {v4.D}[1], [x1], #897 st1 {v4.d}[1], [x1], #8
98 subs w4, w4, #198 subs w4, w4, #1
99 b.gt 1b99 b.gt 1b
100 ret100 ret
@@ -102,46 +102,46 @@ endfunc · 1 unmodified line
102 102
103function ff_ps_hybrid_analysis_neon, export=1103function ff_ps_hybrid_analysis_neon, export=1
104 lsl x3, x3, #3104 lsl x3, x3, #3
105 ld2 {v0.4S,v1.4S}, [x1], #32105 ld2 {v0.4s,v1.4s}, [x1], #32
106 ld2 {v2.2S,v3.2S}, [x1], #16106 ld2 {v2.2s,v3.2s}, [x1], #16
107 ld1 {v24.2S}, [x1], #8107 ld1 {v24.2s}, [x1], #8
108 ld2 {v4.2S,v5.2S}, [x1], #16108 ld2 {v4.2s,v5.2s}, [x1], #16
109 ld2 {v6.4S,v7.4S}, [x1]109 ld2 {v6.4s,v7.4s}, [x1]
110 rev64 v6.4S, v6.4S110 rev64 v6.4s, v6.4s
111 rev64 v7.4S, v7.4S111 rev64 v7.4s, v7.4s
112 ext v6.16B, v6.16B, v6.16B, #8112 ext v6.16b, v6.16b, v6.16b, #8
113 ext v7.16B, v7.16B, v7.16B, #8113 ext v7.16b, v7.16b, v7.16b, #8
114 rev64 v4.2S, v4.2S114 rev64 v4.2s, v4.2s
115 rev64 v5.2S, v5.2S115 rev64 v5.2s, v5.2s
116 mov v2.D[1], v3.D[0]116 mov v2.d[1], v3.d[0]
117 mov v4.D[1], v5.D[0]117 mov v4.d[1], v5.d[0]
118 mov v5.D[1], v2.D[0]118 mov v5.d[1], v2.d[0]
119 mov v3.D[1], v4.D[0]119 mov v3.d[1], v4.d[0]
120 fadd v16.4S, v0.4S, v6.4S120 fadd v16.4s, v0.4s, v6.4s
121 fadd v17.4S, v1.4S, v7.4S121 fadd v17.4s, v1.4s, v7.4s
122 fsub v18.4S, v1.4S, v7.4S122 fsub v18.4s, v1.4s, v7.4s
123 fsub v19.4S, v0.4S, v6.4S123 fsub v19.4s, v0.4s, v6.4s
124 fadd v22.4S, v2.4S, v4.4S124 fadd v22.4s, v2.4s, v4.4s
125 fsub v23.4S, v5.4S, v3.4S125 fsub v23.4s, v5.4s, v3.4s
126 trn1 v20.2D, v22.2D, v23.2D // {re4+re8, re5+re7, im8-im4, im7-im5}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}127 trn2 v21.2d, v22.2d, v23.2d // {im4+im8, im5+im7, re4-re8, re5-re7}
1281: ld2 {v2.4S,v3.4S}, [x2], #321281: ld2 {v2.4s,v3.4s}, [x2], #32
129 ld2 {v4.2S,v5.2S}, [x2], #16129 ld2 {v4.2s,v5.2s}, [x2], #16
130 ld1 {v6.2S}, [x2], #8130 ld1 {v6.2s}, [x2], #8
131 add x2, x2, #8131 add x2, x2, #8
132 mov v4.D[1], v5.D[0]132 mov v4.d[1], v5.d[0]
133 mov v6.S[1], v6.S[0]133 mov v6.s[1], v6.s[0]
134 fmul v6.2S, v6.2S, v24.2S134 fmul v6.2s, v6.2s, v24.2s
135 fmul v0.4S, v2.4S, v16.4S135 fmul v0.4s, v2.4s, v16.4s
136 fmul v1.4S, v2.4S, v17.4S136 fmul v1.4s, v2.4s, v17.4s
137 fmls v0.4S, v3.4S, v18.4S137 fmls v0.4s, v3.4s, v18.4s
138 fmla v1.4S, v3.4S, v19.4S138 fmla v1.4s, v3.4s, v19.4s
139 fmla v0.4S, v4.4S, v20.4S139 fmla v0.4s, v4.4s, v20.4s
140 fmla v1.4S, v4.4S, v21.4S140 fmla v1.4s, v4.4s, v21.4s
141 faddp v0.4S, v0.4S, v1.4S141 faddp v0.4s, v0.4s, v1.4s
142 faddp v0.4S, v0.4S, v0.4S142 faddp v0.4s, v0.4s, v0.4s
143 fadd v0.2S, v0.2S, v6.2S143 fadd v0.2s, v0.2s, v6.2s
144 st1 {v0.2S}, [x0], x3144 st1 {v0.2s}, [x0], x3
145 subs w4, w4, #1145 subs w4, w4, #1
146 b.gt 1b146 b.gt 1b
147 ret147 ret
libavcodec/aarch64/h264cmc_neon.S+203−203
@@ -39,10 +39,10 @@ function ff_\type\()_\codec\()_chroma_mc8_neon, export=1 · 38 unmodified lines
39 lsl w10, w10, #139 lsl w10, w10, #1
40 add w9, w9, w1040 add w9, w9, w10
41 add x6, x6, w9, UXTW41 add x6, x6, w9, UXTW
42 ld1r {v22.8H}, [x6]42 ld1r {v22.8h}, [x6]
43 .endif43 .endif
44 .ifc \codec,vc144 .ifc \codec,vc1
45 movi v22.8H, #2845 movi v22.8h, #28
46 .endif46 .endif
47 mul w7, w4, w547 mul w7, w4, w5
48 lsl w14, w5, #348 lsl w14, w5, #3
@@ -55,139 +55,139 @@ function ff_\type\()_\codec\()_chroma_mc8_neon, export=1 · 6 unmodified lines
55 add w4, w4, #6455 add w4, w4, #64
56 b.eq 2f56 b.eq 2f
57 57
58 dup v0.8B, w458 dup v0.8b, w4
59 dup v1.8B, w1259 dup v1.8b, w12
60 ld1 {v4.8B, v5.8B}, [x1], x260 ld1 {v4.8b, v5.8b}, [x1], x2
61 dup v2.8B, w661 dup v2.8b, w6
62 dup v3.8B, w762 dup v3.8b, w7
63 ext v5.8B, v4.8B, v5.8B, #163 ext v5.8b, v4.8b, v5.8b, #1
641: ld1 {v6.8B, v7.8B}, [x1], x2641: ld1 {v6.8b, v7.8b}, [x1], x2
65 umull v16.8H, v4.8B, v0.8B65 umull v16.8h, v4.8b, v0.8b
66 umlal v16.8H, v5.8B, v1.8B66 umlal v16.8h, v5.8b, v1.8b
67 ext v7.8B, v6.8B, v7.8B, #167 ext v7.8b, v6.8b, v7.8b, #1
68 ld1 {v4.8B, v5.8B}, [x1], x268 ld1 {v4.8b, v5.8b}, [x1], x2
69 umlal v16.8H, v6.8B, v2.8B69 umlal v16.8h, v6.8b, v2.8b
70 prfm pldl1strm, [x1]70 prfm pldl1strm, [x1]
71 ext v5.8B, v4.8B, v5.8B, #171 ext v5.8b, v4.8b, v5.8b, #1
72 umlal v16.8H, v7.8B, v3.8B72 umlal v16.8h, v7.8b, v3.8b
73 umull v17.8H, v6.8B, v0.8B73 umull v17.8h, v6.8b, v0.8b
74 subs w3, w3, #274 subs w3, w3, #2
75 umlal v17.8H, v7.8B, v1.8B75 umlal v17.8h, v7.8b, v1.8b
76 umlal v17.8H, v4.8B, v2.8B76 umlal v17.8h, v4.8b, v2.8b
77 umlal v17.8H, v5.8B, v3.8B77 umlal v17.8h, v5.8b, v3.8b
78 prfm pldl1strm, [x1, x2]78 prfm pldl1strm, [x1, x2]
79 .ifc \codec,h26479 .ifc \codec,h264
80 rshrn v16.8B, v16.8H, #680 rshrn v16.8b, v16.8h, #6
81 rshrn v17.8B, v17.8H, #681 rshrn v17.8b, v17.8h, #6
82 .else82 .else
83 add v16.8H, v16.8H, v22.8H83 add v16.8h, v16.8h, v22.8h
84 add v17.8H, v17.8H, v22.8H84 add v17.8h, v17.8h, v22.8h
85 shrn v16.8B, v16.8H, #685 shrn v16.8b, v16.8h, #6
86 shrn v17.8B, v17.8H, #686 shrn v17.8b, v17.8h, #6
87 .endif87 .endif
88 .ifc \type,avg88 .ifc \type,avg
89 ld1 {v20.8B}, [x8], x289 ld1 {v20.8b}, [x8], x2
90 ld1 {v21.8B}, [x8], x290 ld1 {v21.8b}, [x8], x2
91 urhadd v16.8B, v16.8B, v20.8B91 urhadd v16.8b, v16.8b, v20.8b
92 urhadd v17.8B, v17.8B, v21.8B92 urhadd v17.8b, v17.8b, v21.8b
93 .endif93 .endif
94 st1 {v16.8B}, [x0], x294 st1 {v16.8b}, [x0], x2
95 st1 {v17.8B}, [x0], x295 st1 {v17.8b}, [x0], x2
96 b.gt 1b96 b.gt 1b
97 ret97 ret
98 98
992: adds w12, w12, w6992: adds w12, w12, w6
100 dup v0.8B, w4100 dup v0.8b, w4
101 b.eq 5f101 b.eq 5f
102 tst w6, w6102 tst w6, w6
103 dup v1.8B, w12103 dup v1.8b, w12
104 b.eq 4f104 b.eq 4f
105 105
106 ld1 {v4.8B}, [x1], x2106 ld1 {v4.8b}, [x1], x2
1073: ld1 {v6.8B}, [x1], x21073: ld1 {v6.8b}, [x1], x2
108 umull v16.8H, v4.8B, v0.8B108 umull v16.8h, v4.8b, v0.8b
109 umlal v16.8H, v6.8B, v1.8B109 umlal v16.8h, v6.8b, v1.8b
110 ld1 {v4.8B}, [x1], x2110 ld1 {v4.8b}, [x1], x2
111 umull v17.8H, v6.8B, v0.8B111 umull v17.8h, v6.8b, v0.8b
112 umlal v17.8H, v4.8B, v1.8B112 umlal v17.8h, v4.8b, v1.8b
113 prfm pldl1strm, [x1]113 prfm pldl1strm, [x1]
114 .ifc \codec,h264114 .ifc \codec,h264
115 rshrn v16.8B, v16.8H, #6115 rshrn v16.8b, v16.8h, #6
116 rshrn v17.8B, v17.8H, #6116 rshrn v17.8b, v17.8h, #6
117 .else117 .else
118 add v16.8H, v16.8H, v22.8H118 add v16.8h, v16.8h, v22.8h
119 add v17.8H, v17.8H, v22.8H119 add v17.8h, v17.8h, v22.8h
120 shrn v16.8B, v16.8H, #6120 shrn v16.8b, v16.8h, #6
121 shrn v17.8B, v17.8H, #6121 shrn v17.8b, v17.8h, #6
122 .endif122 .endif
123 prfm pldl1strm, [x1, x2]123 prfm pldl1strm, [x1, x2]
124 .ifc \type,avg124 .ifc \type,avg
125 ld1 {v20.8B}, [x8], x2125 ld1 {v20.8b}, [x8], x2
126 ld1 {v21.8B}, [x8], x2126 ld1 {v21.8b}, [x8], x2
127 urhadd v16.8B, v16.8B, v20.8B127 urhadd v16.8b, v16.8b, v20.8b
128 urhadd v17.8B, v17.8B, v21.8B128 urhadd v17.8b, v17.8b, v21.8b
129 .endif129 .endif
130 subs w3, w3, #2130 subs w3, w3, #2
131 st1 {v16.8B}, [x0], x2131 st1 {v16.8b}, [x0], x2
132 st1 {v17.8B}, [x0], x2132 st1 {v17.8b}, [x0], x2
133 b.gt 3b133 b.gt 3b
134 ret134 ret
135 135
1364: ld1 {v4.8B, v5.8B}, [x1], x21364: ld1 {v4.8b, v5.8b}, [x1], x2
137 ld1 {v6.8B, v7.8B}, [x1], x2137 ld1 {v6.8b, v7.8b}, [x1], x2
138 ext v5.8B, v4.8B, v5.8B, #1138 ext v5.8b, v4.8b, v5.8b, #1
139 ext v7.8B, v6.8B, v7.8B, #1139 ext v7.8b, v6.8b, v7.8b, #1
140 prfm pldl1strm, [x1]140 prfm pldl1strm, [x1]
141 subs w3, w3, #2141 subs w3, w3, #2
142 umull v16.8H, v4.8B, v0.8B142 umull v16.8h, v4.8b, v0.8b
143 umlal v16.8H, v5.8B, v1.8B143 umlal v16.8h, v5.8b, v1.8b
144 umull v17.8H, v6.8B, v0.8B144 umull v17.8h, v6.8b, v0.8b
145 umlal v17.8H, v7.8B, v1.8B145 umlal v17.8h, v7.8b, v1.8b
146 prfm pldl1strm, [x1, x2]146 prfm pldl1strm, [x1, x2]
147 .ifc \codec,h264147 .ifc \codec,h264
148 rshrn v16.8B, v16.8H, #6148 rshrn v16.8b, v16.8h, #6
149 rshrn v17.8B, v17.8H, #6149 rshrn v17.8b, v17.8h, #6
150 .else150 .else
151 add v16.8H, v16.8H, v22.8H151 add v16.8h, v16.8h, v22.8h
152 add v17.8H, v17.8H, v22.8H152 add v17.8h, v17.8h, v22.8h
153 shrn v16.8B, v16.8H, #6153 shrn v16.8b, v16.8h, #6
154 shrn v17.8B, v17.8H, #6154 shrn v17.8b, v17.8h, #6
155 .endif155 .endif
156 .ifc \type,avg156 .ifc \type,avg
157 ld1 {v20.8B}, [x8], x2157 ld1 {v20.8b}, [x8], x2
158 ld1 {v21.8B}, [x8], x2158 ld1 {v21.8b}, [x8], x2
159 urhadd v16.8B, v16.8B, v20.8B159 urhadd v16.8b, v16.8b, v20.8b
160 urhadd v17.8B, v17.8B, v21.8B160 urhadd v17.8b, v17.8b, v21.8b
161 .endif161 .endif
162 st1 {v16.8B}, [x0], x2162 st1 {v16.8b}, [x0], x2
163 st1 {v17.8B}, [x0], x2163 st1 {v17.8b}, [x0], x2
164 b.gt 4b164 b.gt 4b
165 ret165 ret
166 166
1675: ld1 {v4.8B}, [x1], x21675: ld1 {v4.8b}, [x1], x2
168 ld1 {v5.8B}, [x1], x2168 ld1 {v5.8b}, [x1], x2
169 prfm pldl1strm, [x1]169 prfm pldl1strm, [x1]
170 subs w3, w3, #2170 subs w3, w3, #2
171 umull v16.8H, v4.8B, v0.8B171 umull v16.8h, v4.8b, v0.8b
172 umull v17.8H, v5.8B, v0.8B172 umull v17.8h, v5.8b, v0.8b
173 prfm pldl1strm, [x1, x2]173 prfm pldl1strm, [x1, x2]
174 .ifc \codec,h264174 .ifc \codec,h264
175 rshrn v16.8B, v16.8H, #6175 rshrn v16.8b, v16.8h, #6
176 rshrn v17.8B, v17.8H, #6176 rshrn v17.8b, v17.8h, #6
177 .else177 .else
178 add v16.8H, v16.8H, v22.8H178 add v16.8h, v16.8h, v22.8h
179 add v17.8H, v17.8H, v22.8H179 add v17.8h, v17.8h, v22.8h
180 shrn v16.8B, v16.8H, #6180 shrn v16.8b, v16.8h, #6
181 shrn v17.8B, v17.8H, #6181 shrn v17.8b, v17.8h, #6
182 .endif182 .endif
183 .ifc \type,avg183 .ifc \type,avg
184 ld1 {v20.8B}, [x8], x2184 ld1 {v20.8b}, [x8], x2
185 ld1 {v21.8B}, [x8], x2185 ld1 {v21.8b}, [x8], x2
186 urhadd v16.8B, v16.8B, v20.8B186 urhadd v16.8b, v16.8b, v20.8b
187 urhadd v17.8B, v17.8B, v21.8B187 urhadd v17.8b, v17.8b, v21.8b
188 .endif188 .endif
189 st1 {v16.8B}, [x0], x2189 st1 {v16.8b}, [x0], x2
190 st1 {v17.8B}, [x0], x2190 st1 {v17.8b}, [x0], x2
191 b.gt 5b191 b.gt 5b
192 ret192 ret
193endfunc193endfunc
@@ -209,10 +209,10 @@ function ff_\type\()_\codec\()_chroma_mc4_neon, export=1 · 15 unmodified lines
209 lsl w10, w10, #1209 lsl w10, w10, #1
210 add w9, w9, w10210 add w9, w9, w10
211 add x6, x6, w9, UXTW211 add x6, x6, w9, UXTW
212 ld1r {v22.8H}, [x6]212 ld1r {v22.8h}, [x6]
213 .endif213 .endif
214 .ifc \codec,vc1214 .ifc \codec,vc1
215 movi v22.8H, #28215 movi v22.8h, #28
216 .endif216 .endif
217 mul w7, w4, w5217 mul w7, w4, w5
218 lsl w14, w5, #3218 lsl w14, w5, #3
@@ -225,133 +225,133 @@ function ff_\type\()_\codec\()_chroma_mc4_neon, export=1 · 6 unmodified lines
225 add w4, w4, #64225 add w4, w4, #64
226 b.eq 2f226 b.eq 2f
227 227
228 dup v24.8B, w4228 dup v24.8b, w4
229 dup v25.8B, w12229 dup v25.8b, w12
230 ld1 {v4.8B}, [x1], x2230 ld1 {v4.8b}, [x1], x2
231 dup v26.8B, w6231 dup v26.8b, w6
232 dup v27.8B, w7232 dup v27.8b, w7
233 ext v5.8B, v4.8B, v5.8B, #1233 ext v5.8b, v4.8b, v5.8b, #1
234 trn1 v0.2S, v24.2S, v25.2S234 trn1 v0.2s, v24.2s, v25.2s
235 trn1 v2.2S, v26.2S, v27.2S235 trn1 v2.2s, v26.2s, v27.2s
236 trn1 v4.2S, v4.2S, v5.2S236 trn1 v4.2s, v4.2s, v5.2s
2371: ld1 {v6.8B}, [x1], x22371: ld1 {v6.8b}, [x1], x2
238 ext v7.8B, v6.8B, v7.8B, #1238 ext v7.8b, v6.8b, v7.8b, #1
239 trn1 v6.2S, v6.2S, v7.2S239 trn1 v6.2s, v6.2s, v7.2s
240 umull v18.8H, v4.8B, v0.8B240 umull v18.8h, v4.8b, v0.8b
241 umlal v18.8H, v6.8B, v2.8B241 umlal v18.8h, v6.8b, v2.8b
242 ld1 {v4.8B}, [x1], x2242 ld1 {v4.8b}, [x1], x2
243 ext v5.8B, v4.8B, v5.8B, #1243 ext v5.8b, v4.8b, v5.8b, #1
244 trn1 v4.2S, v4.2S, v5.2S244 trn1 v4.2s, v4.2s, v5.2s
245 prfm pldl1strm, [x1]245 prfm pldl1strm, [x1]
246 umull v19.8H, v6.8B, v0.8B246 umull v19.8h, v6.8b, v0.8b
247 umlal v19.8H, v4.8B, v2.8B247 umlal v19.8h, v4.8b, v2.8b
248 trn1 v30.2D, v18.2D, v19.2D248 trn1 v30.2d, v18.2d, v19.2d
249 trn2 v31.2D, v18.2D, v19.2D249 trn2 v31.2d, v18.2d, v19.2d
250 add v18.8H, v30.8H, v31.8H250 add v18.8h, v30.8h, v31.8h
251 .ifc \codec,h264251 .ifc \codec,h264
252 rshrn v16.8B, v18.8H, #6252 rshrn v16.8b, v18.8h, #6
253 .else253 .else
254 add v18.8H, v18.8H, v22.8H254 add v18.8h, v18.8h, v22.8h
255 shrn v16.8B, v18.8H, #6255 shrn v16.8b, v18.8h, #6
256 .endif256 .endif
257 subs w3, w3, #2257 subs w3, w3, #2
258 prfm pldl1strm, [x1, x2]258 prfm pldl1strm, [x1, x2]
259 .ifc \type,avg259 .ifc \type,avg
260 ld1 {v20.S}[0], [x8], x2260 ld1 {v20.s}[0], [x8], x2
261 ld1 {v20.S}[1], [x8], x2261 ld1 {v20.s}[1], [x8], x2
262 urhadd v16.8B, v16.8B, v20.8B262 urhadd v16.8b, v16.8b, v20.8b
263 .endif263 .endif
264 st1 {v16.S}[0], [x0], x2264 st1 {v16.s}[0], [x0], x2
265 st1 {v16.S}[1], [x0], x2265 st1 {v16.s}[1], [x0], x2
266 b.gt 1b266 b.gt 1b
267 ret267 ret
268 268
2692: adds w12, w12, w62692: adds w12, w12, w6
270 dup v30.8B, w4270 dup v30.8b, w4
271 b.eq 5f271 b.eq 5f
272 tst w6, w6272 tst w6, w6
273 dup v31.8B, w12273 dup v31.8b, w12
274 trn1 v0.2S, v30.2S, v31.2S274 trn1 v0.2s, v30.2s, v31.2s
275 trn2 v1.2S, v30.2S, v31.2S275 trn2 v1.2s, v30.2s, v31.2s
276 b.eq 4f276 b.eq 4f
277 277
278 ext v1.8B, v0.8B, v1.8B, #4278 ext v1.8b, v0.8b, v1.8b, #4
279 ld1 {v4.S}[0], [x1], x2279 ld1 {v4.s}[0], [x1], x2
2803: ld1 {v4.S}[1], [x1], x22803: ld1 {v4.s}[1], [x1], x2
281 umull v18.8H, v4.8B, v0.8B281 umull v18.8h, v4.8b, v0.8b
282 ld1 {v4.S}[0], [x1], x2282 ld1 {v4.s}[0], [x1], x2
283 umull v19.8H, v4.8B, v1.8B283 umull v19.8h, v4.8b, v1.8b
284 trn1 v30.2D, v18.2D, v19.2D284 trn1 v30.2d, v18.2d, v19.2d
285 trn2 v31.2D, v18.2D, v19.2D285 trn2 v31.2d, v18.2d, v19.2d
286 add v18.8H, v30.8H, v31.8H286 add v18.8h, v30.8h, v31.8h
287 prfm pldl1strm, [x1]287 prfm pldl1strm, [x1]
288 .ifc \codec,h264288 .ifc \codec,h264
289 rshrn v16.8B, v18.8H, #6289 rshrn v16.8b, v18.8h, #6
290 .else290 .else
291 add v18.8H, v18.8H, v22.8H291 add v18.8h, v18.8h, v22.8h
292 shrn v16.8B, v18.8H, #6292 shrn v16.8b, v18.8h, #6
293 .endif293 .endif
294 .ifc \type,avg294 .ifc \type,avg
295 ld1 {v20.S}[0], [x8], x2295 ld1 {v20.s}[0], [x8], x2
296 ld1 {v20.S}[1], [x8], x2296 ld1 {v20.s}[1], [x8], x2
297 urhadd v16.8B, v16.8B, v20.8B297 urhadd v16.8b, v16.8b, v20.8b
298 .endif298 .endif
299 subs w3, w3, #2299 subs w3, w3, #2
300 prfm pldl1strm, [x1, x2]300 prfm pldl1strm, [x1, x2]
301 st1 {v16.S}[0], [x0], x2301 st1 {v16.s}[0], [x0], x2
302 st1 {v16.S}[1], [x0], x2302 st1 {v16.s}[1], [x0], x2
303 b.gt 3b303 b.gt 3b
304 ret304 ret
305 305
3064: ld1 {v4.8B}, [x1], x23064: ld1 {v4.8b}, [x1], x2
307 ld1 {v6.8B}, [x1], x2307 ld1 {v6.8b}, [x1], x2
308 ext v5.8B, v4.8B, v5.8B, #1308 ext v5.8b, v4.8b, v5.8b, #1
309 ext v7.8B, v6.8B, v7.8B, #1309 ext v7.8b, v6.8b, v7.8b, #1
310 trn1 v4.2S, v4.2S, v5.2S310 trn1 v4.2s, v4.2s, v5.2s
311 trn1 v6.2S, v6.2S, v7.2S311 trn1 v6.2s, v6.2s, v7.2s
312 umull v18.8H, v4.8B, v0.8B312 umull v18.8h, v4.8b, v0.8b
313 umull v19.8H, v6.8B, v0.8B313 umull v19.8h, v6.8b, v0.8b
314 subs w3, w3, #2314 subs w3, w3, #2
315 trn1 v30.2D, v18.2D, v19.2D315 trn1 v30.2d, v18.2d, v19.2d
316 trn2 v31.2D, v18.2D, v19.2D316 trn2 v31.2d, v18.2d, v19.2d
317 add v18.8H, v30.8H, v31.8H317 add v18.8h, v30.8h, v31.8h
318 prfm pldl1strm, [x1]318 prfm pldl1strm, [x1]
319 .ifc \codec,h264319 .ifc \codec,h264
320 rshrn v16.8B, v18.8H, #6320 rshrn v16.8b, v18.8h, #6
321 .else321 .else
322 add v18.8H, v18.8H, v22.8H322 add v18.8h, v18.8h, v22.8h
323 shrn v16.8B, v18.8H, #6323 shrn v16.8b, v18.8h, #6
324 .endif324 .endif
325 .ifc \type,avg325 .ifc \type,avg
326 ld1 {v20.S}[0], [x8], x2326 ld1 {v20.s}[0], [x8], x2
327 ld1 {v20.S}[1], [x8], x2327 ld1 {v20.s}[1], [x8], x2
328 urhadd v16.8B, v16.8B, v20.8B328 urhadd v16.8b, v16.8b, v20.8b
329 .endif329 .endif
330 prfm pldl1strm, [x1]330 prfm pldl1strm, [x1]
331 st1 {v16.S}[0], [x0], x2331 st1 {v16.s}[0], [x0], x2
332 st1 {v16.S}[1], [x0], x2332 st1 {v16.s}[1], [x0], x2
333 b.gt 4b333 b.gt 4b
334 ret334 ret
335 335
3365: ld1 {v4.S}[0], [x1], x23365: ld1 {v4.s}[0], [x1], x2
337 ld1 {v4.S}[1], [x1], x2337 ld1 {v4.s}[1], [x1], x2
338 umull v18.8H, v4.8B, v30.8B338 umull v18.8h, v4.8b, v30.8b
339 subs w3, w3, #2339 subs w3, w3, #2
340 prfm pldl1strm, [x1]340 prfm pldl1strm, [x1]
341 .ifc \codec,h264341 .ifc \codec,h264
342 rshrn v16.8B, v18.8H, #6342 rshrn v16.8b, v18.8h, #6
343 .else343 .else
344 add v18.8H, v18.8H, v22.8H344 add v18.8h, v18.8h, v22.8h
345 shrn v16.8B, v18.8H, #6345 shrn v16.8b, v18.8h, #6
346 .endif346 .endif
347 .ifc \type,avg347 .ifc \type,avg
348 ld1 {v20.S}[0], [x8], x2348 ld1 {v20.s}[0], [x8], x2
349 ld1 {v20.S}[1], [x8], x2349 ld1 {v20.s}[1], [x8], x2
350 urhadd v16.8B, v16.8B, v20.8B350 urhadd v16.8b, v16.8b, v20.8b
351 .endif351 .endif
352 prfm pldl1strm, [x1]352 prfm pldl1strm, [x1]
353 st1 {v16.S}[0], [x0], x2353 st1 {v16.s}[0], [x0], x2
354 st1 {v16.S}[1], [x0], x2354 st1 {v16.s}[1], [x0], x2
355 b.gt 5b355 b.gt 5b
356 ret356 ret
357endfunc357endfunc
@@ -372,51 +372,51 @@ function ff_\type\()_h264_chroma_mc2_neon, export=1 · 14 unmodified lines
372 sub w4, w7, w13372 sub w4, w7, w13
373 sub w4, w4, w14373 sub w4, w4, w14
374 add w4, w4, #64374 add w4, w4, #64
375 dup v0.8B, w4375 dup v0.8b, w4
376 dup v2.8B, w12376 dup v2.8b, w12
377 dup v1.8B, w6377 dup v1.8b, w6
378 dup v3.8B, w7378 dup v3.8b, w7
379 trn1 v0.4H, v0.4H, v2.4H379 trn1 v0.4h, v0.4h, v2.4h
380 trn1 v1.4H, v1.4H, v3.4H380 trn1 v1.4h, v1.4h, v3.4h
3811:3811:
382 ld1 {v4.S}[0], [x1], x2382 ld1 {v4.s}[0], [x1], x2
383 ld1 {v4.S}[1], [x1], x2383 ld1 {v4.s}[1], [x1], x2
384 rev64 v5.2S, v4.2S384 rev64 v5.2s, v4.2s
385 ld1 {v5.S}[1], [x1]385 ld1 {v5.s}[1], [x1]
386 ext v6.8B, v4.8B, v5.8B, #1386 ext v6.8b, v4.8b, v5.8b, #1
387 ext v7.8B, v5.8B, v4.8B, #1387 ext v7.8b, v5.8b, v4.8b, #1
388 trn1 v4.4H, v4.4H, v6.4H388 trn1 v4.4h, v4.4h, v6.4h
389 trn1 v5.4H, v5.4H, v7.4H389 trn1 v5.4h, v5.4h, v7.4h
390 umull v16.8H, v4.8B, v0.8B390 umull v16.8h, v4.8b, v0.8b
391 umlal v16.8H, v5.8B, v1.8B391 umlal v16.8h, v5.8b, v1.8b
392 .ifc \type,avg392 .ifc \type,avg
393 ld1 {v18.H}[0], [x0], x2393 ld1 {v18.h}[0], [x0], x2
394 ld1 {v18.H}[2], [x0]394 ld1 {v18.h}[2], [x0]
395 sub x0, x0, x2395 sub x0, x0, x2
396 .endif396 .endif
397 rev64 v17.4S, v16.4S397 rev64 v17.4s, v16.4s
398 add v16.8H, v16.8H, v17.8H398 add v16.8h, v16.8h, v17.8h
399 rshrn v16.8B, v16.8H, #6399 rshrn v16.8b, v16.8h, #6
400 .ifc \type,avg400 .ifc \type,avg
401 urhadd v16.8B, v16.8B, v18.8B401 urhadd v16.8b, v16.8b, v18.8b
402 .endif402 .endif
403 st1 {v16.H}[0], [x0], x2403 st1 {v16.h}[0], [x0], x2
404 st1 {v16.H}[2], [x0], x2404 st1 {v16.h}[2], [x0], x2
405 subs w3, w3, #2405 subs w3, w3, #2
406 b.gt 1b406 b.gt 1b
407 ret407 ret
408 408
4092:4092:
410 ld1 {v16.H}[0], [x1], x2410 ld1 {v16.h}[0], [x1], x2
411 ld1 {v16.H}[1], [x1], x2411 ld1 {v16.h}[1], [x1], x2
412 .ifc \type,avg412 .ifc \type,avg
413 ld1 {v18.H}[0], [x0], x2413 ld1 {v18.h}[0], [x0], x2
414 ld1 {v18.H}[1], [x0]414 ld1 {v18.h}[1], [x0]
415 sub x0, x0, x2415 sub x0, x0, x2
416 urhadd v16.8B, v16.8B, v18.8B416 urhadd v16.8b, v16.8b, v18.8b
417 .endif417 .endif
418 st1 {v16.H}[0], [x0], x2418 st1 {v16.h}[0], [x0], x2
419 st1 {v16.H}[1], [x0], x2419 st1 {v16.h}[1], [x0], x2
420 subs w3, w3, #2420 subs w3, w3, #2
421 b.gt 2b421 b.gt 2b
422 ret422 ret
libavcodec/aarch64/h264dsp_neon.S+297−297
@@ -27,7 +27,7 @@ · 26 unmodified lines
27 cmp w2, #027 cmp w2, #0
28 ldr w6, [x4]28 ldr w6, [x4]
29 ccmp w3, #0, #0, ne29 ccmp w3, #0, #0, ne
30 mov v24.S[0], w630 mov v24.s[0], w6
31 and w8, w6, w6, lsl #1631 and w8, w6, w6, lsl #16
32 b.eq 1f32 b.eq 1f
33 ands w8, w8, w8, lsl #833 ands w8, w8, w8, lsl #8
@@ -38,95 +38,95 @@ · 4 unmodified lines
38.endm38.endm
39 39
40.macro h264_loop_filter_luma40.macro h264_loop_filter_luma
41 dup v22.16B, w2 // alpha41 dup v22.16b, w2 // alpha
42 uxtl v24.8H, v24.8B42 uxtl v24.8h, v24.8b
43 uabd v21.16B, v16.16B, v0.16B // abs(p0 - q0)43 uabd v21.16b, v16.16b, v0.16b // abs(p0 - q0)
44 uxtl v24.4S, v24.4H44 uxtl v24.4s, v24.4h
45 uabd v28.16B, v18.16B, v16.16B // abs(p1 - p0)45 uabd v28.16b, v18.16b, v16.16b // abs(p1 - p0)
46 sli v24.8H, v24.8H, #846 sli v24.8h, v24.8h, #8
47 uabd v30.16B, v2.16B, v0.16B // abs(q1 - q0)47 uabd v30.16b, v2.16b, v0.16b // abs(q1 - q0)
48 sli v24.4S, v24.4S, #1648 sli v24.4s, v24.4s, #16
49 cmhi v21.16B, v22.16B, v21.16B // < alpha49 cmhi v21.16b, v22.16b, v21.16b // < alpha
50 dup v22.16B, w3 // beta50 dup v22.16b, w3 // beta
51 cmlt v23.16B, v24.16B, #051 cmlt v23.16b, v24.16b, #0
52 cmhi v28.16B, v22.16B, v28.16B // < beta52 cmhi v28.16b, v22.16b, v28.16b // < beta
53 cmhi v30.16B, v22.16B, v30.16B // < beta53 cmhi v30.16b, v22.16b, v30.16b // < beta
54 bic v21.16B, v21.16B, v23.16B54 bic v21.16b, v21.16b, v23.16b
55 uabd v17.16B, v20.16B, v16.16B // abs(p2 - p0)55 uabd v17.16b, v20.16b, v16.16b // abs(p2 - p0)
56 and v21.16B, v21.16B, v28.16B56 and v21.16b, v21.16b, v28.16b
57 uabd v19.16B, v4.16B, v0.16B // abs(q2 - q0)57 uabd v19.16b, v4.16b, v0.16b // abs(q2 - q0)
58 and v21.16B, v21.16B, v30.16B // < beta58 and v21.16b, v21.16b, v30.16b // < beta
59 shrn v30.8b, v21.8h, #459 shrn v30.8b, v21.8h, #4
60 mov x7, v30.d[0]60 mov x7, v30.d[0]
61 cmhi v17.16B, v22.16B, v17.16B // < beta61 cmhi v17.16b, v22.16b, v17.16b // < beta
62 cmhi v19.16B, v22.16B, v19.16B // < beta62 cmhi v19.16b, v22.16b, v19.16b // < beta
63 cbz x7, 9f63 cbz x7, 9f
64 and v17.16B, v17.16B, v21.16B64 and v17.16b, v17.16b, v21.16b
65 and v19.16B, v19.16B, v21.16B65 and v19.16b, v19.16b, v21.16b
66 and v24.16B, v24.16B, v21.16B66 and v24.16b, v24.16b, v21.16b
67 urhadd v28.16B, v16.16B, v0.16B67 urhadd v28.16b, v16.16b, v0.16b
68 sub v21.16B, v24.16B, v17.16B68 sub v21.16b, v24.16b, v17.16b
69 uqadd v23.16B, v18.16B, v24.16B69 uqadd v23.16b, v18.16b, v24.16b
70 uhadd v20.16B, v20.16B, v28.16B70 uhadd v20.16b, v20.16b, v28.16b
71 sub v21.16B, v21.16B, v19.16B71 sub v21.16b, v21.16b, v19.16b
72 uhadd v28.16B, v4.16B, v28.16B72 uhadd v28.16b, v4.16b, v28.16b
73 umin v23.16B, v23.16B, v20.16B73 umin v23.16b, v23.16b, v20.16b
74 uqsub v22.16B, v18.16B, v24.16B74 uqsub v22.16b, v18.16b, v24.16b
75 uqadd v4.16B, v2.16B, v24.16B75 uqadd v4.16b, v2.16b, v24.16b
76 umax v23.16B, v23.16B, v22.16B76 umax v23.16b, v23.16b, v22.16b
77 uqsub v22.16B, v2.16B, v24.16B77 uqsub v22.16b, v2.16b, v24.16b
78 umin v28.16B, v4.16B, v28.16B78 umin v28.16b, v4.16b, v28.16b
79 uxtl v4.8H, v0.8B79 uxtl v4.8h, v0.8b
80 umax v28.16B, v28.16B, v22.16B80 umax v28.16b, v28.16b, v22.16b
81 uxtl2 v20.8H, v0.16B81 uxtl2 v20.8h, v0.16b
82 usubw v4.8H, v4.8H, v16.8B82 usubw v4.8h, v4.8h, v16.8b
83 usubw2 v20.8H, v20.8H, v16.16B83 usubw2 v20.8h, v20.8h, v16.16b
84 shl v4.8H, v4.8H, #284 shl v4.8h, v4.8h, #2
85 shl v20.8H, v20.8H, #285 shl v20.8h, v20.8h, #2
86 uaddw v4.8H, v4.8H, v18.8B86 uaddw v4.8h, v4.8h, v18.8b
87 uaddw2 v20.8H, v20.8H, v18.16B87 uaddw2 v20.8h, v20.8h, v18.16b
88 usubw v4.8H, v4.8H, v2.8B88 usubw v4.8h, v4.8h, v2.8b
89 usubw2 v20.8H, v20.8H, v2.16B89 usubw2 v20.8h, v20.8h, v2.16b
90 rshrn v4.8B, v4.8H, #390 rshrn v4.8b, v4.8h, #3
91 rshrn2 v4.16B, v20.8H, #391 rshrn2 v4.16b, v20.8h, #3
92 bsl v17.16B, v23.16B, v18.16B92 bsl v17.16b, v23.16b, v18.16b
93 bsl v19.16B, v28.16B, v2.16B93 bsl v19.16b, v28.16b, v2.16b
94 neg v23.16B, v21.16B94 neg v23.16b, v21.16b
95 uxtl v28.8H, v16.8B95 uxtl v28.8h, v16.8b
96 smin v4.16B, v4.16B, v21.16B96 smin v4.16b, v4.16b, v21.16b
97 uxtl2 v21.8H, v16.16B97 uxtl2 v21.8h, v16.16b
98 smax v4.16B, v4.16B, v23.16B98 smax v4.16b, v4.16b, v23.16b
99 uxtl v22.8H, v0.8B99 uxtl v22.8h, v0.8b
100 uxtl2 v24.8H, v0.16B100 uxtl2 v24.8h, v0.16b
101 saddw v28.8H, v28.8H, v4.8B101 saddw v28.8h, v28.8h, v4.8b
102 saddw2 v21.8H, v21.8H, v4.16B102 saddw2 v21.8h, v21.8h, v4.16b
103 ssubw v22.8H, v22.8H, v4.8B103 ssubw v22.8h, v22.8h, v4.8b
104 ssubw2 v24.8H, v24.8H, v4.16B104 ssubw2 v24.8h, v24.8h, v4.16b
105 sqxtun v16.8B, v28.8H105 sqxtun v16.8b, v28.8h
106 sqxtun2 v16.16B, v21.8H106 sqxtun2 v16.16b, v21.8h
107 sqxtun v0.8B, v22.8H107 sqxtun v0.8b, v22.8h
108 sqxtun2 v0.16B, v24.8H108 sqxtun2 v0.16b, v24.8h
109.endm109.endm
110 110
111function ff_h264_v_loop_filter_luma_neon, export=1111function ff_h264_v_loop_filter_luma_neon, export=1
112 h264_loop_filter_start112 h264_loop_filter_start
113 113
114 ld1 {v0.16B}, [x0], x1114 ld1 {v0.16b}, [x0], x1
115 ld1 {v2.16B}, [x0], x1115 ld1 {v2.16b}, [x0], x1
116 ld1 {v4.16B}, [x0], x1116 ld1 {v4.16b}, [x0], x1
117 sub x0, x0, x1, lsl #2117 sub x0, x0, x1, lsl #2
118 sub x0, x0, x1, lsl #1118 sub x0, x0, x1, lsl #1
119 ld1 {v20.16B}, [x0], x1119 ld1 {v20.16b}, [x0], x1
120 ld1 {v18.16B}, [x0], x1120 ld1 {v18.16b}, [x0], x1
121 ld1 {v16.16B}, [x0], x1121 ld1 {v16.16b}, [x0], x1
122 122
123 h264_loop_filter_luma123 h264_loop_filter_luma
124 124
125 sub x0, x0, x1, lsl #1125 sub x0, x0, x1, lsl #1
126 st1 {v17.16B}, [x0], x1126 st1 {v17.16b}, [x0], x1
127 st1 {v16.16B}, [x0], x1127 st1 {v16.16b}, [x0], x1
128 st1 {v0.16B}, [x0], x1128 st1 {v0.16b}, [x0], x1
129 st1 {v19.16B}, [x0]129 st1 {v19.16b}, [x0]
1309:1309:
131 ret131 ret
132endfunc132endfunc
@@ -135,22 +135,22 @@ function ff_h264_h_loop_filter_luma_neon, export=1 · 2 unmodified lines
135 h264_loop_filter_start135 h264_loop_filter_start
136 136
137 sub x0, x0, #4137 sub x0, x0, #4
138 ld1 {v6.8B}, [x0], x1138 ld1 {v6.8b}, [x0], x1
139 ld1 {v20.8B}, [x0], x1139 ld1 {v20.8b}, [x0], x1
140 ld1 {v18.8B}, [x0], x1140 ld1 {v18.8b}, [x0], x1
141 ld1 {v16.8B}, [x0], x1141 ld1 {v16.8b}, [x0], x1
142 ld1 {v0.8B}, [x0], x1142 ld1 {v0.8b}, [x0], x1
143 ld1 {v2.8B}, [x0], x1143 ld1 {v2.8b}, [x0], x1
144 ld1 {v4.8B}, [x0], x1144 ld1 {v4.8b}, [x0], x1
145 ld1 {v26.8B}, [x0], x1145 ld1 {v26.8b}, [x0], x1
146 ld1 {v6.D}[1], [x0], x1146 ld1 {v6.d}[1], [x0], x1
147 ld1 {v20.D}[1], [x0], x1147 ld1 {v20.d}[1], [x0], x1
148 ld1 {v18.D}[1], [x0], x1148 ld1 {v18.d}[1], [x0], x1
149 ld1 {v16.D}[1], [x0], x1149 ld1 {v16.d}[1], [x0], x1
150 ld1 {v0.D}[1], [x0], x1150 ld1 {v0.d}[1], [x0], x1
151 ld1 {v2.D}[1], [x0], x1151 ld1 {v2.d}[1], [x0], x1
152 ld1 {v4.D}[1], [x0], x1152 ld1 {v4.d}[1], [x0], x1
153 ld1 {v26.D}[1], [x0], x1153 ld1 {v26.d}[1], [x0], x1
154 154
155 transpose_8x16B v6, v20, v18, v16, v0, v2, v4, v26, v21, v23155 transpose_8x16B v6, v20, v18, v16, v0, v2, v4, v26, v21, v23
156 156
@@ -160,22 +160,22 @@ function ff_h264_h_loop_filter_luma_neon, export=1 · 3 unmodified lines
160 160
161 sub x0, x0, x1, lsl #4161 sub x0, x0, x1, lsl #4
162 add x0, x0, #2162 add x0, x0, #2
163 st1 {v17.S}[0], [x0], x1163 st1 {v17.s}[0], [x0], x1
164 st1 {v16.S}[0], [x0], x1164 st1 {v16.s}[0], [x0], x1
165 st1 {v0.S}[0], [x0], x1165 st1 {v0.s}[0], [x0], x1
166 st1 {v19.S}[0], [x0], x1166 st1 {v19.s}[0], [x0], x1
167 st1 {v17.S}[1], [x0], x1167 st1 {v17.s}[1], [x0], x1
168 st1 {v16.S}[1], [x0], x1168 st1 {v16.s}[1], [x0], x1
169 st1 {v0.S}[1], [x0], x1169 st1 {v0.s}[1], [x0], x1
170 st1 {v19.S}[1], [x0], x1170 st1 {v19.s}[1], [x0], x1
171 st1 {v17.S}[2], [x0], x1171 st1 {v17.s}[2], [x0], x1
172 st1 {v16.S}[2], [x0], x1172 st1 {v16.s}[2], [x0], x1
173 st1 {v0.S}[2], [x0], x1173 st1 {v0.s}[2], [x0], x1
174 st1 {v19.S}[2], [x0], x1174 st1 {v19.s}[2], [x0], x1
175 st1 {v17.S}[3], [x0], x1175 st1 {v17.s}[3], [x0], x1
176 st1 {v16.S}[3], [x0], x1176 st1 {v16.s}[3], [x0], x1
177 st1 {v0.S}[3], [x0], x1177 st1 {v0.s}[3], [x0], x1
178 st1 {v19.S}[3], [x0], x1178 st1 {v19.s}[3], [x0], x1
1799:1799:
180 ret180 ret
181endfunc181endfunc
@@ -377,52 +377,52 @@ function ff_h264_h_loop_filter_luma_intra_neon, export=1 · 195 unmodified lines
377endfunc377endfunc
378 378
379.macro h264_loop_filter_chroma379.macro h264_loop_filter_chroma
380 dup v22.8B, w2 // alpha380 dup v22.8b, w2 // alpha
381 dup v23.8B, w3 // beta381 dup v23.8b, w3 // beta
382 uxtl v24.8H, v24.8B382 uxtl v24.8h, v24.8b
383 uabd v26.8B, v16.8B, v0.8B // abs(p0 - q0)383 uabd v26.8b, v16.8b, v0.8b // abs(p0 - q0)
384 uabd v28.8B, v18.8B, v16.8B // abs(p1 - p0)384 uabd v28.8b, v18.8b, v16.8b // abs(p1 - p0)
385 uabd v30.8B, v2.8B, v0.8B // abs(q1 - q0)385 uabd v30.8b, v2.8b, v0.8b // abs(q1 - q0)
386 cmhi v26.8B, v22.8B, v26.8B // < alpha386 cmhi v26.8b, v22.8b, v26.8b // < alpha
387 cmhi v28.8B, v23.8B, v28.8B // < beta387 cmhi v28.8b, v23.8b, v28.8b // < beta
388 cmhi v30.8B, v23.8B, v30.8B // < beta388 cmhi v30.8b, v23.8b, v30.8b // < beta
389 uxtl v4.8H, v0.8B389 uxtl v4.8h, v0.8b
390 and v26.8B, v26.8B, v28.8B390 and v26.8b, v26.8b, v28.8b
391 usubw v4.8H, v4.8H, v16.8B391 usubw v4.8h, v4.8h, v16.8b
392 and v26.8B, v26.8B, v30.8B392 and v26.8b, v26.8b, v30.8b
393 shl v4.8H, v4.8H, #2393 shl v4.8h, v4.8h, #2
394 mov x8, v26.d[0]394 mov x8, v26.d[0]
395 sli v24.8H, v24.8H, #8395 sli v24.8h, v24.8h, #8
396 uaddw v4.8H, v4.8H, v18.8B396 uaddw v4.8h, v4.8h, v18.8b
397 cbz x8, 9f397 cbz x8, 9f
398 usubw v4.8H, v4.8H, v2.8B398 usubw v4.8h, v4.8h, v2.8b
399 rshrn v4.8B, v4.8H, #3399 rshrn v4.8b, v4.8h, #3
400 smin v4.8B, v4.8B, v24.8B400 smin v4.8b, v4.8b, v24.8b
401 neg v25.8B, v24.8B401 neg v25.8b, v24.8b
402 smax v4.8B, v4.8B, v25.8B402 smax v4.8b, v4.8b, v25.8b
403 uxtl v22.8H, v0.8B403 uxtl v22.8h, v0.8b
404 and v4.8B, v4.8B, v26.8B404 and v4.8b, v4.8b, v26.8b
405 uxtl v28.8H, v16.8B405 uxtl v28.8h, v16.8b
406 saddw v28.8H, v28.8H, v4.8B406 saddw v28.8h, v28.8h, v4.8b
407 ssubw v22.8H, v22.8H, v4.8B407 ssubw v22.8h, v22.8h, v4.8b
408 sqxtun v16.8B, v28.8H408 sqxtun v16.8b, v28.8h
409 sqxtun v0.8B, v22.8H409 sqxtun v0.8b, v22.8h
410.endm410.endm
411 411
412function ff_h264_v_loop_filter_chroma_neon, export=1412function ff_h264_v_loop_filter_chroma_neon, export=1
413 h264_loop_filter_start413 h264_loop_filter_start
414 414
415 sub x0, x0, x1, lsl #1415 sub x0, x0, x1, lsl #1
416 ld1 {v18.8B}, [x0], x1416 ld1 {v18.8b}, [x0], x1
417 ld1 {v16.8B}, [x0], x1417 ld1 {v16.8b}, [x0], x1
418 ld1 {v0.8B}, [x0], x1418 ld1 {v0.8b}, [x0], x1
419 ld1 {v2.8B}, [x0]419 ld1 {v2.8b}, [x0]
420 420
421 h264_loop_filter_chroma421 h264_loop_filter_chroma
422 422
423 sub x0, x0, x1, lsl #1423 sub x0, x0, x1, lsl #1
424 st1 {v16.8B}, [x0], x1424 st1 {v16.8b}, [x0], x1
425 st1 {v0.8B}, [x0], x1425 st1 {v0.8b}, [x0], x1
4269:4269:
427 ret427 ret
428endfunc428endfunc
@@ -432,14 +432,14 @@ function ff_h264_h_loop_filter_chroma_neon, export=1 · 3 unmodified lines
432 432
433 sub x0, x0, #2433 sub x0, x0, #2
434h_loop_filter_chroma420:434h_loop_filter_chroma420:
435 ld1 {v18.S}[0], [x0], x1435 ld1 {v18.s}[0], [x0], x1
436 ld1 {v16.S}[0], [x0], x1436 ld1 {v16.s}[0], [x0], x1
437 ld1 {v0.S}[0], [x0], x1437 ld1 {v0.s}[0], [x0], x1
438 ld1 {v2.S}[0], [x0], x1438 ld1 {v2.s}[0], [x0], x1
439 ld1 {v18.S}[1], [x0], x1439 ld1 {v18.s}[1], [x0], x1
440 ld1 {v16.S}[1], [x0], x1440 ld1 {v16.s}[1], [x0], x1
441 ld1 {v0.S}[1], [x0], x1441 ld1 {v0.s}[1], [x0], x1
442 ld1 {v2.S}[1], [x0], x1442 ld1 {v2.s}[1], [x0], x1
443 443
444 transpose_4x8B v18, v16, v0, v2, v28, v29, v30, v31444 transpose_4x8B v18, v16, v0, v2, v28, v29, v30, v31
445 445
@@ -448,14 +448,14 @@ h_loop_filter_chroma420: · 2 unmodified lines
448 transpose_4x8B v18, v16, v0, v2, v28, v29, v30, v31448 transpose_4x8B v18, v16, v0, v2, v28, v29, v30, v31
449 449
450 sub x0, x0, x1, lsl #3450 sub x0, x0, x1, lsl #3
451 st1 {v18.S}[0], [x0], x1451 st1 {v18.s}[0], [x0], x1
452 st1 {v16.S}[0], [x0], x1452 st1 {v16.s}[0], [x0], x1
453 st1 {v0.S}[0], [x0], x1453 st1 {v0.s}[0], [x0], x1
454 st1 {v2.S}[0], [x0], x1454 st1 {v2.s}[0], [x0], x1
455 st1 {v18.S}[1], [x0], x1455 st1 {v18.s}[1], [x0], x1
456 st1 {v16.S}[1], [x0], x1456 st1 {v16.s}[1], [x0], x1
457 st1 {v0.S}[1], [x0], x1457 st1 {v0.s}[1], [x0], x1
458 st1 {v2.S}[1], [x0], x1458 st1 {v2.s}[1], [x0], x1
4599:4599:
460 ret460 ret
461endfunc461endfunc
@@ -584,102 +584,102 @@ function ff_h264_h_loop_filter_chroma422_intra_neon, export=1 · 122 unmodified lines
584endfunc584endfunc
585 585
586.macro biweight_16 macs, macd586.macro biweight_16 macs, macd
587 dup v0.16B, w5587 dup v0.16b, w5
588 dup v1.16B, w6588 dup v1.16b, w6
589 mov v4.16B, v16.16B589 mov v4.16b, v16.16b
590 mov v6.16B, v16.16B590 mov v6.16b, v16.16b
5911: subs w3, w3, #25911: subs w3, w3, #2
592 ld1 {v20.16B}, [x0], x2592 ld1 {v20.16b}, [x0], x2
593 \macd v4.8H, v0.8B, v20.8B593 \macd v4.8h, v0.8b, v20.8b
594 \macd\()2 v6.8H, v0.16B, v20.16B594 \macd\()2 v6.8H, v0.16B, v20.16B
595 ld1 {v22.16B}, [x1], x2595 ld1 {v22.16b}, [x1], x2
596 \macs v4.8H, v1.8B, v22.8B596 \macs v4.8h, v1.8b, v22.8b
597 \macs\()2 v6.8H, v1.16B, v22.16B597 \macs\()2 v6.8H, v1.16B, v22.16B
598 mov v24.16B, v16.16B598 mov v24.16b, v16.16b
599 ld1 {v28.16B}, [x0], x2599 ld1 {v28.16b}, [x0], x2
600 mov v26.16B, v16.16B600 mov v26.16b, v16.16b
601 \macd v24.8H, v0.8B, v28.8B601 \macd v24.8h, v0.8b, v28.8b
602 \macd\()2 v26.8H, v0.16B, v28.16B602 \macd\()2 v26.8H, v0.16B, v28.16B
603 ld1 {v30.16B}, [x1], x2603 ld1 {v30.16b}, [x1], x2
604 \macs v24.8H, v1.8B, v30.8B604 \macs v24.8h, v1.8b, v30.8b
605 \macs\()2 v26.8H, v1.16B, v30.16B605 \macs\()2 v26.8H, v1.16B, v30.16B
606 sshl v4.8H, v4.8H, v18.8H606 sshl v4.8h, v4.8h, v18.8h
607 sshl v6.8H, v6.8H, v18.8H607 sshl v6.8h, v6.8h, v18.8h
608 sqxtun v4.8B, v4.8H608 sqxtun v4.8b, v4.8h
609 sqxtun2 v4.16B, v6.8H609 sqxtun2 v4.16b, v6.8h
610 sshl v24.8H, v24.8H, v18.8H610 sshl v24.8h, v24.8h, v18.8h
611 sshl v26.8H, v26.8H, v18.8H611 sshl v26.8h, v26.8h, v18.8h
612 sqxtun v24.8B, v24.8H612 sqxtun v24.8b, v24.8h
613 sqxtun2 v24.16B, v26.8H613 sqxtun2 v24.16b, v26.8h
614 mov v6.16B, v16.16B614 mov v6.16b, v16.16b
615 st1 {v4.16B}, [x7], x2615 st1 {v4.16b}, [x7], x2
616 mov v4.16B, v16.16B616 mov v4.16b, v16.16b
617 st1 {v24.16B}, [x7], x2617 st1 {v24.16b}, [x7], x2
618 b.ne 1b618 b.ne 1b
619 ret619 ret
620.endm620.endm
621 621
622.macro biweight_8 macs, macd622.macro biweight_8 macs, macd
623 dup v0.8B, w5623 dup v0.8b, w5
624 dup v1.8B, w6624 dup v1.8b, w6
625 mov v2.16B, v16.16B625 mov v2.16b, v16.16b
626 mov v20.16B, v16.16B626 mov v20.16b, v16.16b
6271: subs w3, w3, #26271: subs w3, w3, #2
628 ld1 {v4.8B}, [x0], x2628 ld1 {v4.8b}, [x0], x2
629 \macd v2.8H, v0.8B, v4.8B629 \macd v2.8h, v0.8b, v4.8b
630 ld1 {v5.8B}, [x1], x2630 ld1 {v5.8b}, [x1], x2
631 \macs v2.8H, v1.8B, v5.8B631 \macs v2.8h, v1.8b, v5.8b
632 ld1 {v6.8B}, [x0], x2632 ld1 {v6.8b}, [x0], x2
633 \macd v20.8H, v0.8B, v6.8B633 \macd v20.8h, v0.8b, v6.8b
634 ld1 {v7.8B}, [x1], x2634 ld1 {v7.8b}, [x1], x2
635 \macs v20.8H, v1.8B, v7.8B635 \macs v20.8h, v1.8b, v7.8b
636 sshl v2.8H, v2.8H, v18.8H636 sshl v2.8h, v2.8h, v18.8h
637 sqxtun v2.8B, v2.8H637 sqxtun v2.8b, v2.8h
638 sshl v20.8H, v20.8H, v18.8H638 sshl v20.8h, v20.8h, v18.8h
639 sqxtun v4.8B, v20.8H639 sqxtun v4.8b, v20.8h
640 mov v20.16B, v16.16B640 mov v20.16b, v16.16b
641 st1 {v2.8B}, [x7], x2641 st1 {v2.8b}, [x7], x2
642 mov v2.16B, v16.16B642 mov v2.16b, v16.16b
643 st1 {v4.8B}, [x7], x2643 st1 {v4.8b}, [x7], x2
644 b.ne 1b644 b.ne 1b
645 ret645 ret
646.endm646.endm
647 647
648.macro biweight_4 macs, macd648.macro biweight_4 macs, macd
649 dup v0.8B, w5649 dup v0.8b, w5
650 dup v1.8B, w6650 dup v1.8b, w6
651 mov v2.16B, v16.16B651 mov v2.16b, v16.16b
652 mov v20.16B,v16.16B652 mov v20.16b,v16.16b
6531: subs w3, w3, #46531: subs w3, w3, #4
654 ld1 {v4.S}[0], [x0], x2654 ld1 {v4.s}[0], [x0], x2
655 ld1 {v4.S}[1], [x0], x2655 ld1 {v4.s}[1], [x0], x2
656 \macd v2.8H, v0.8B, v4.8B656 \macd v2.8h, v0.8b, v4.8b
657 ld1 {v5.S}[0], [x1], x2657 ld1 {v5.s}[0], [x1], x2
658 ld1 {v5.S}[1], [x1], x2658 ld1 {v5.s}[1], [x1], x2
659 \macs v2.8H, v1.8B, v5.8B659 \macs v2.8h, v1.8b, v5.8b
660 b.lt 2f660 b.lt 2f
661 ld1 {v6.S}[0], [x0], x2661 ld1 {v6.s}[0], [x0], x2
662 ld1 {v6.S}[1], [x0], x2662 ld1 {v6.s}[1], [x0], x2
663 \macd v20.8H, v0.8B, v6.8B663 \macd v20.8h, v0.8b, v6.8b
664 ld1 {v7.S}[0], [x1], x2664 ld1 {v7.s}[0], [x1], x2
665 ld1 {v7.S}[1], [x1], x2665 ld1 {v7.s}[1], [x1], x2
666 \macs v20.8H, v1.8B, v7.8B666 \macs v20.8h, v1.8b, v7.8b
667 sshl v2.8H, v2.8H, v18.8H667 sshl v2.8h, v2.8h, v18.8h
668 sqxtun v2.8B, v2.8H668 sqxtun v2.8b, v2.8h
669 sshl v20.8H, v20.8H, v18.8H669 sshl v20.8h, v20.8h, v18.8h
670 sqxtun v4.8B, v20.8H670 sqxtun v4.8b, v20.8h
671 mov v20.16B, v16.16B671 mov v20.16b, v16.16b
672 st1 {v2.S}[0], [x7], x2672 st1 {v2.s}[0], [x7], x2
673 st1 {v2.S}[1], [x7], x2673 st1 {v2.s}[1], [x7], x2
674 mov v2.16B, v16.16B674 mov v2.16b, v16.16b
675 st1 {v4.S}[0], [x7], x2675 st1 {v4.s}[0], [x7], x2
676 st1 {v4.S}[1], [x7], x2676 st1 {v4.s}[1], [x7], x2
677 b.ne 1b677 b.ne 1b
678 ret678 ret
6792: sshl v2.8H, v2.8H, v18.8H6792: sshl v2.8h, v2.8h, v18.8h
680 sqxtun v2.8B, v2.8H680 sqxtun v2.8b, v2.8h
681 st1 {v2.S}[0], [x7], x2681 st1 {v2.s}[0], [x7], x2
682 st1 {v2.S}[1], [x7], x2682 st1 {v2.s}[1], [x7], x2
683 ret683 ret
684.endm684.endm
685 685
@@ -689,10 +689,10 @@ function ff_biweight_h264_pixels_\w\()_neon, export=1 · 3 unmodified lines
689 add w7, w7, #1689 add w7, w7, #1
690 eor w8, w8, w6, lsr #30690 eor w8, w8, w6, lsr #30
691 orr w7, w7, #1691 orr w7, w7, #1
692 dup v18.8H, w4692 dup v18.8h, w4
693 lsl w7, w7, w4693 lsl w7, w7, w4
694 not v18.16B, v18.16B694 not v18.16b, v18.16b
695 dup v16.8H, w7695 dup v16.8h, w7
696 mov x7, x0696 mov x7, x0
697 cbz w8, 10f697 cbz w8, 10f
698 subs w8, w8, #1698 subs w8, w8, #1
@@ -716,78 +716,78 @@ endfunc · 17 unmodified lines
716 biweight_func 4716 biweight_func 4
717 717
718.macro weight_16 add718.macro weight_16 add
719 dup v0.16B, w4719 dup v0.16b, w4
7201: subs w2, w2, #27201: subs w2, w2, #2
721 ld1 {v20.16B}, [x0], x1721 ld1 {v20.16b}, [x0], x1
722 umull v4.8H, v0.8B, v20.8B722 umull v4.8h, v0.8b, v20.8b
723 umull2 v6.8H, v0.16B, v20.16B723 umull2 v6.8h, v0.16b, v20.16b
724 ld1 {v28.16B}, [x0], x1724 ld1 {v28.16b}, [x0], x1
725 umull v24.8H, v0.8B, v28.8B725 umull v24.8h, v0.8b, v28.8b
726 umull2 v26.8H, v0.16B, v28.16B726 umull2 v26.8h, v0.16b, v28.16b
727 \add v4.8H, v16.8H, v4.8H727 \add v4.8h, v16.8h, v4.8h
728 srshl v4.8H, v4.8H, v18.8H728 srshl v4.8h, v4.8h, v18.8h
729 \add v6.8H, v16.8H, v6.8H729 \add v6.8h, v16.8h, v6.8h
730 srshl v6.8H, v6.8H, v18.8H730 srshl v6.8h, v6.8h, v18.8h
731 sqxtun v4.8B, v4.8H731 sqxtun v4.8b, v4.8h
732 sqxtun2 v4.16B, v6.8H732 sqxtun2 v4.16b, v6.8h
733 \add v24.8H, v16.8H, v24.8H733 \add v24.8h, v16.8h, v24.8h
734 srshl v24.8H, v24.8H, v18.8H734 srshl v24.8h, v24.8h, v18.8h
735 \add v26.8H, v16.8H, v26.8H735 \add v26.8h, v16.8h, v26.8h
736 srshl v26.8H, v26.8H, v18.8H736 srshl v26.8h, v26.8h, v18.8h
737 sqxtun v24.8B, v24.8H737 sqxtun v24.8b, v24.8h
738 sqxtun2 v24.16B, v26.8H738 sqxtun2 v24.16b, v26.8h
739 st1 {v4.16B}, [x5], x1739 st1 {v4.16b}, [x5], x1
740 st1 {v24.16B}, [x5], x1740 st1 {v24.16b}, [x5], x1
741 b.ne 1b741 b.ne 1b
742 ret742 ret
743.endm743.endm
744 744
745.macro weight_8 add745.macro weight_8 add
746 dup v0.8B, w4746 dup v0.8b, w4
7471: subs w2, w2, #27471: subs w2, w2, #2
748 ld1 {v4.8B}, [x0], x1748 ld1 {v4.8b}, [x0], x1
749 umull v2.8H, v0.8B, v4.8B749 umull v2.8h, v0.8b, v4.8b
750 ld1 {v6.8B}, [x0], x1750 ld1 {v6.8b}, [x0], x1
751 umull v20.8H, v0.8B, v6.8B751 umull v20.8h, v0.8b, v6.8b
752 \add v2.8H, v16.8H, v2.8H752 \add v2.8h, v16.8h, v2.8h
753 srshl v2.8H, v2.8H, v18.8H753 srshl v2.8h, v2.8h, v18.8h
754 sqxtun v2.8B, v2.8H754 sqxtun v2.8b, v2.8h
755 \add v20.8H, v16.8H, v20.8H755 \add v20.8h, v16.8h, v20.8h
756 srshl v20.8H, v20.8H, v18.8H756 srshl v20.8h, v20.8h, v18.8h
757 sqxtun v4.8B, v20.8H757 sqxtun v4.8b, v20.8h
758 st1 {v2.8B}, [x5], x1758 st1 {v2.8b}, [x5], x1
759 st1 {v4.8B}, [x5], x1759 st1 {v4.8b}, [x5], x1
760 b.ne 1b760 b.ne 1b
761 ret761 ret
762.endm762.endm
763 763
764.macro weight_4 add764.macro weight_4 add
765 dup v0.8B, w4765 dup v0.8b, w4
7661: subs w2, w2, #47661: subs w2, w2, #4
767 ld1 {v4.S}[0], [x0], x1767 ld1 {v4.s}[0], [x0], x1
768 ld1 {v4.S}[1], [x0], x1768 ld1 {v4.s}[1], [x0], x1
769 umull v2.8H, v0.8B, v4.8B769 umull v2.8h, v0.8b, v4.8b
770 b.lt 2f770 b.lt 2f
771 ld1 {v6.S}[0], [x0], x1771 ld1 {v6.s}[0], [x0], x1
772 ld1 {v6.S}[1], [x0], x1772 ld1 {v6.s}[1], [x0], x1
773 umull v20.8H, v0.8B, v6.8B773 umull v20.8h, v0.8b, v6.8b
774 \add v2.8H, v16.8H, v2.8H774 \add v2.8h, v16.8h, v2.8h
775 srshl v2.8H, v2.8H, v18.8H775 srshl v2.8h, v2.8h, v18.8h
776 sqxtun v2.8B, v2.8H776 sqxtun v2.8b, v2.8h
777 \add v20.8H, v16.8H, v20.8H777 \add v20.8h, v16.8h, v20.8h
778 srshl v20.8H, v20.8h, v18.8H778 srshl v20.8h, v20.8h, v18.8h
779 sqxtun v4.8B, v20.8H779 sqxtun v4.8b, v20.8h
780 st1 {v2.S}[0], [x5], x1780 st1 {v2.s}[0], [x5], x1
781 st1 {v2.S}[1], [x5], x1781 st1 {v2.s}[1], [x5], x1
782 st1 {v4.S}[0], [x5], x1782 st1 {v4.s}[0], [x5], x1
783 st1 {v4.S}[1], [x5], x1783 st1 {v4.s}[1], [x5], x1
784 b.ne 1b784 b.ne 1b
785 ret785 ret
7862: \add v2.8H, v16.8H, v2.8H7862: \add v2.8h, v16.8h, v2.8h
787 srshl v2.8H, v2.8H, v18.8H787 srshl v2.8h, v2.8h, v18.8h
788 sqxtun v2.8B, v2.8H788 sqxtun v2.8b, v2.8h
789 st1 {v2.S}[0], [x5], x1789 st1 {v2.s}[0], [x5], x1
790 st1 {v2.S}[1], [x5], x1790 st1 {v2.s}[1], [x5], x1
791 ret791 ret
792.endm792.endm
793 793
@@ -796,18 +796,18 @@ function ff_weight_h264_pixels_\w\()_neon, export=1 · 2 unmodified lines
796 cmp w3, #1796 cmp w3, #1
797 mov w6, #1797 mov w6, #1
798 lsl w5, w5, w3798 lsl w5, w5, w3
799 dup v16.8H, w5799 dup v16.8h, w5
800 mov x5, x0800 mov x5, x0
801 b.le 20f801 b.le 20f
802 sub w6, w6, w3802 sub w6, w6, w3
803 dup v18.8H, w6803 dup v18.8h, w6
804 cmp w4, #0804 cmp w4, #0
805 b.lt 10f805 b.lt 10f
806 weight_\w shadd806 weight_\w shadd
80710: neg w4, w480710: neg w4, w4
808 weight_\w shsub808 weight_\w shsub
80920: neg w6, w380920: neg w6, w3
810 dup v18.8H, w6810 dup v18.8h, w6
811 cmp w4, #0811 cmp w4, #0
812 b.lt 10f812 b.lt 10f
813 weight_\w add813 weight_\w add
@@ -825,7 +825,7 @@ endfunc · 11 unmodified lines
825 ldr w6, [x4]825 ldr w6, [x4]
826 ccmp w3, #0, #0, ne826 ccmp w3, #0, #0, ne
827 lsl w2, w2, #2827 lsl w2, w2, #2
828 mov v24.S[0], w6828 mov v24.s[0], w6
829 lsl w3, w3, #2829 lsl w3, w3, #2
830 and w8, w6, w6, lsl #16830 and w8, w6, w6, lsl #16
831 b.eq 1f831 b.eq 1f
libavcodec/aarch64/h264idct_neon.S+195−195
@@ -25,54 +25,54 @@ · 24 unmodified lines
25function ff_h264_idct_add_neon, export=125function ff_h264_idct_add_neon, export=1
26.L_ff_h264_idct_add_neon:26.L_ff_h264_idct_add_neon:
27 AARCH64_VALID_CALL_TARGET27 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]
29 sxtw x2, w229 sxtw x2, w2
30 movi v30.8H, #030 movi v30.8h, #0
31 31
32 add v4.4H, v0.4H, v2.4H32 add v4.4h, v0.4h, v2.4h
33 sshr v16.4H, v1.4H, #133 sshr v16.4h, v1.4h, #1
34 st1 {v30.8H}, [x1], #1634 st1 {v30.8h}, [x1], #16
35 sshr v17.4H, v3.4H, #135 sshr v17.4h, v3.4h, #1
36 st1 {v30.8H}, [x1], #1636 st1 {v30.8h}, [x1], #16
37 sub v5.4H, v0.4H, v2.4H37 sub v5.4h, v0.4h, v2.4h
38 sub v6.4H, v16.4H, v3.4H38 sub v6.4h, v16.4h, v3.4h
39 add v7.4H, v1.4H, v17.4H39 add v7.4h, v1.4h, v17.4h
40 add v0.4H, v4.4H, v7.4H40 add v0.4h, v4.4h, v7.4h
41 add v1.4H, v5.4H, v6.4H41 add v1.4h, v5.4h, v6.4h
42 sub v2.4H, v5.4H, v6.4H42 sub v2.4h, v5.4h, v6.4h
43 sub v3.4H, v4.4H, v7.4H43 sub v3.4h, v4.4h, v7.4h
44 44
45 transpose_4x4H v0, v1, v2, v3, v4, v5, v6, v745 transpose_4x4H v0, v1, v2, v3, v4, v5, v6, v7
46 46
47 add v4.4H, v0.4H, v2.4H47 add v4.4h, v0.4h, v2.4h
48 ld1 {v18.S}[0], [x0], x248 ld1 {v18.s}[0], [x0], x2
49 sshr v16.4H, v3.4H, #149 sshr v16.4h, v3.4h, #1
50 sshr v17.4H, v1.4H, #150 sshr v17.4h, v1.4h, #1
51 ld1 {v18.S}[1], [x0], x251 ld1 {v18.s}[1], [x0], x2
52 sub v5.4H, v0.4H, v2.4H52 sub v5.4h, v0.4h, v2.4h
53 ld1 {v19.S}[1], [x0], x253 ld1 {v19.s}[1], [x0], x2
54 add v6.4H, v16.4H, v1.4H54 add v6.4h, v16.4h, v1.4h
55 ins v4.D[1], v5.D[0]55 ins v4.d[1], v5.d[0]
56 sub v7.4H, v17.4H, v3.4H56 sub v7.4h, v17.4h, v3.4h
57 ld1 {v19.S}[0], [x0], x257 ld1 {v19.s}[0], [x0], x2
58 ins v6.D[1], v7.D[0]58 ins v6.d[1], v7.d[0]
59 sub x0, x0, x2, lsl #259 sub x0, x0, x2, lsl #2
60 add v0.8H, v4.8H, v6.8H60 add v0.8h, v4.8h, v6.8h
61 sub v1.8H, v4.8H, v6.8H61 sub v1.8h, v4.8h, v6.8h
62 62
63 srshr v0.8H, v0.8H, #663 srshr v0.8h, v0.8h, #6
64 srshr v1.8H, v1.8H, #664 srshr v1.8h, v1.8h, #6
65 65
66 uaddw v0.8H, v0.8H, v18.8B66 uaddw v0.8h, v0.8h, v18.8b
67 uaddw v1.8H, v1.8H, v19.8B67 uaddw v1.8h, v1.8h, v19.8b
68 68
69 sqxtun v0.8B, v0.8H69 sqxtun v0.8b, v0.8h
70 sqxtun v1.8B, v1.8H70 sqxtun v1.8b, v1.8h
71 71
72 st1 {v0.S}[0], [x0], x272 st1 {v0.s}[0], [x0], x2
73 st1 {v0.S}[1], [x0], x273 st1 {v0.s}[1], [x0], x2
74 st1 {v1.S}[1], [x0], x274 st1 {v1.s}[1], [x0], x2
75 st1 {v1.S}[0], [x0], x275 st1 {v1.s}[0], [x0], x2
76 76
77 sub x1, x1, #3277 sub x1, x1, #32
78 ret78 ret
@@ -83,22 +83,22 @@ function ff_h264_idct_dc_add_neon, export=1 · 4 unmodified lines
83 AARCH64_VALID_CALL_TARGET83 AARCH64_VALID_CALL_TARGET
84 sxtw x2, w284 sxtw x2, w2
85 mov w3, #085 mov w3, #0
86 ld1r {v2.8H}, [x1]86 ld1r {v2.8h}, [x1]
87 strh w3, [x1]87 strh w3, [x1]
88 srshr v2.8H, v2.8H, #688 srshr v2.8h, v2.8h, #6
89 ld1 {v0.S}[0], [x0], x289 ld1 {v0.s}[0], [x0], x2
90 ld1 {v0.S}[1], [x0], x290 ld1 {v0.s}[1], [x0], x2
91 uaddw v3.8H, v2.8H, v0.8B91 uaddw v3.8h, v2.8h, v0.8b
92 ld1 {v1.S}[0], [x0], x292 ld1 {v1.s}[0], [x0], x2
93 ld1 {v1.S}[1], [x0], x293 ld1 {v1.s}[1], [x0], x2
94 uaddw v4.8H, v2.8H, v1.8B94 uaddw v4.8h, v2.8h, v1.8b
95 sqxtun v0.8B, v3.8H95 sqxtun v0.8b, v3.8h
96 sqxtun v1.8B, v4.8H96 sqxtun v1.8b, v4.8h
97 sub x0, x0, x2, lsl #297 sub x0, x0, x2, lsl #2
98 st1 {v0.S}[0], [x0], x298 st1 {v0.s}[0], [x0], x2
99 st1 {v0.S}[1], [x0], x299 st1 {v0.s}[1], [x0], x2
100 st1 {v1.S}[0], [x0], x2100 st1 {v1.s}[0], [x0], x2
101 st1 {v1.S}[1], [x0], x2101 st1 {v1.s}[1], [x0], x2
102 ret102 ret
103endfunc103endfunc
104 104
@@ -194,71 +194,71 @@ endfunc · 89 unmodified lines
194 .if \pass == 0194 .if \pass == 0
195 va .req v18195 va .req v18
196 vb .req v30196 vb .req v30
197 sshr v18.8H, v26.8H, #1197 sshr v18.8h, v26.8h, #1
198 add v16.8H, v24.8H, v28.8H198 add v16.8h, v24.8h, v28.8h
199 ld1 {v30.8H, v31.8H}, [x1]199 ld1 {v30.8h, v31.8h}, [x1]
200 st1 {v19.8H}, [x1], #16200 st1 {v19.8h}, [x1], #16
201 st1 {v19.8H}, [x1], #16201 st1 {v19.8h}, [x1], #16
202 sub v17.8H, v24.8H, v28.8H202 sub v17.8h, v24.8h, v28.8h
203 sshr v19.8H, v30.8H, #1203 sshr v19.8h, v30.8h, #1
204 sub v18.8H, v18.8H, v30.8H204 sub v18.8h, v18.8h, v30.8h
205 add v19.8H, v19.8H, v26.8H205 add v19.8h, v19.8h, v26.8h
206 .else206 .else
207 va .req v30207 va .req v30
208 vb .req v18208 vb .req v18
209 sshr v30.8H, v26.8H, #1209 sshr v30.8h, v26.8h, #1
210 sshr v19.8H, v18.8H, #1210 sshr v19.8h, v18.8h, #1
211 add v16.8H, v24.8H, v28.8H211 add v16.8h, v24.8h, v28.8h
212 sub v17.8H, v24.8H, v28.8H212 sub v17.8h, v24.8h, v28.8h
213 sub v30.8H, v30.8H, v18.8H213 sub v30.8h, v30.8h, v18.8h
214 add v19.8H, v19.8H, v26.8H214 add v19.8h, v19.8h, v26.8h
215 .endif215 .endif
216 add v26.8H, v17.8H, va.8H216 add v26.8h, v17.8h, va.8h
217 sub v28.8H, v17.8H, va.8H217 sub v28.8h, v17.8h, va.8h
218 add v24.8H, v16.8H, v19.8H218 add v24.8h, v16.8h, v19.8h
219 sub vb.8H, v16.8H, v19.8H219 sub vb.8h, v16.8h, v19.8h
220 sub v16.8H, v29.8H, v27.8H220 sub v16.8h, v29.8h, v27.8h
221 add v17.8H, v31.8H, v25.8H221 add v17.8h, v31.8h, v25.8h
222 sub va.8H, v31.8H, v25.8H222 sub va.8h, v31.8h, v25.8h
223 add v19.8H, v29.8H, v27.8H223 add v19.8h, v29.8h, v27.8h
224 sub v16.8H, v16.8H, v31.8H224 sub v16.8h, v16.8h, v31.8h
225 sub v17.8H, v17.8H, v27.8H225 sub v17.8h, v17.8h, v27.8h
226 add va.8H, va.8H, v29.8H226 add va.8h, va.8h, v29.8h
227 add v19.8H, v19.8H, v25.8H227 add v19.8h, v19.8h, v25.8h
228 sshr v25.8H, v25.8H, #1228 sshr v25.8h, v25.8h, #1
229 sshr v27.8H, v27.8H, #1229 sshr v27.8h, v27.8h, #1
230 sshr v29.8H, v29.8H, #1230 sshr v29.8h, v29.8h, #1
231 sshr v31.8H, v31.8H, #1231 sshr v31.8h, v31.8h, #1
232 sub v16.8H, v16.8H, v31.8H232 sub v16.8h, v16.8h, v31.8h
233 sub v17.8H, v17.8H, v27.8H233 sub v17.8h, v17.8h, v27.8h
234 add va.8H, va.8H, v29.8H234 add va.8h, va.8h, v29.8h
235 add v19.8H, v19.8H, v25.8H235 add v19.8h, v19.8h, v25.8h
236 sshr v25.8H, v16.8H, #2236 sshr v25.8h, v16.8h, #2
237 sshr v27.8H, v17.8H, #2237 sshr v27.8h, v17.8h, #2
238 sshr v29.8H, va.8H, #2238 sshr v29.8h, va.8h, #2
239 sshr v31.8H, v19.8H, #2239 sshr v31.8h, v19.8h, #2
240 sub v19.8H, v19.8H, v25.8H240 sub v19.8h, v19.8h, v25.8h
241 sub va.8H, v27.8H, va.8H241 sub va.8h, v27.8h, va.8h
242 add v17.8H, v17.8H, v29.8H242 add v17.8h, v17.8h, v29.8h
243 add v16.8H, v16.8H, v31.8H243 add v16.8h, v16.8h, v31.8h
244 .if \pass == 0244 .if \pass == 0
245 sub v31.8H, v24.8H, v19.8H245 sub v31.8h, v24.8h, v19.8h
246 add v24.8H, v24.8H, v19.8H246 add v24.8h, v24.8h, v19.8h
247 add v25.8H, v26.8H, v18.8H247 add v25.8h, v26.8h, v18.8h
248 sub v18.8H, v26.8H, v18.8H248 sub v18.8h, v26.8h, v18.8h
249 add v26.8H, v28.8H, v17.8H249 add v26.8h, v28.8h, v17.8h
250 add v27.8H, v30.8H, v16.8H250 add v27.8h, v30.8h, v16.8h
251 sub v29.8H, v28.8H, v17.8H251 sub v29.8h, v28.8h, v17.8h
252 sub v28.8H, v30.8H, v16.8H252 sub v28.8h, v30.8h, v16.8h
253 .else253 .else
254 sub v31.8H, v24.8H, v19.8H254 sub v31.8h, v24.8h, v19.8h
255 add v24.8H, v24.8H, v19.8H255 add v24.8h, v24.8h, v19.8h
256 add v25.8H, v26.8H, v30.8H256 add v25.8h, v26.8h, v30.8h
257 sub v30.8H, v26.8H, v30.8H257 sub v30.8h, v26.8h, v30.8h
258 add v26.8H, v28.8H, v17.8H258 add v26.8h, v28.8h, v17.8h
259 sub v29.8H, v28.8H, v17.8H259 sub v29.8h, v28.8h, v17.8h
260 add v27.8H, v18.8H, v16.8H260 add v27.8h, v18.8h, v16.8h
261 sub v28.8H, v18.8H, v16.8H261 sub v28.8h, v18.8h, v16.8h
262 .endif262 .endif
263 .unreq va263 .unreq va
264 .unreq vb264 .unreq vb
@@ -267,63 +267,63 @@ endfunc · 2 unmodified lines
267function ff_h264_idct8_add_neon, export=1267function ff_h264_idct8_add_neon, export=1
268.L_ff_h264_idct8_add_neon:268.L_ff_h264_idct8_add_neon:
269 AARCH64_VALID_CALL_TARGET269 AARCH64_VALID_CALL_TARGET
270 movi v19.8H, #0270 movi v19.8h, #0
271 sxtw x2, w2271 sxtw x2, w2
272 ld1 {v24.8H, v25.8H}, [x1]272 ld1 {v24.8h, v25.8h}, [x1]
273 st1 {v19.8H}, [x1], #16273 st1 {v19.8h}, [x1], #16
274 st1 {v19.8H}, [x1], #16274 st1 {v19.8h}, [x1], #16
275 ld1 {v26.8H, v27.8H}, [x1]275 ld1 {v26.8h, v27.8h}, [x1]
276 st1 {v19.8H}, [x1], #16276 st1 {v19.8h}, [x1], #16
277 st1 {v19.8H}, [x1], #16277 st1 {v19.8h}, [x1], #16
278 ld1 {v28.8H, v29.8H}, [x1]278 ld1 {v28.8h, v29.8h}, [x1]
279 st1 {v19.8H}, [x1], #16279 st1 {v19.8h}, [x1], #16
280 st1 {v19.8H}, [x1], #16280 st1 {v19.8h}, [x1], #16
281 281
282 idct8x8_cols 0282 idct8x8_cols 0
283 transpose_8x8H v24, v25, v26, v27, v28, v29, v18, v31, v6, v7283 transpose_8x8H v24, v25, v26, v27, v28, v29, v18, v31, v6, v7
284 idct8x8_cols 1284 idct8x8_cols 1
285 285
286 mov x3, x0286 mov x3, x0
287 srshr v24.8H, v24.8H, #6287 srshr v24.8h, v24.8h, #6
288 ld1 {v0.8B}, [x0], x2288 ld1 {v0.8b}, [x0], x2
289 srshr v25.8H, v25.8H, #6289 srshr v25.8h, v25.8h, #6
290 ld1 {v1.8B}, [x0], x2290 ld1 {v1.8b}, [x0], x2
291 srshr v26.8H, v26.8H, #6291 srshr v26.8h, v26.8h, #6
292 ld1 {v2.8B}, [x0], x2292 ld1 {v2.8b}, [x0], x2
293 srshr v27.8H, v27.8H, #6293 srshr v27.8h, v27.8h, #6
294 ld1 {v3.8B}, [x0], x2294 ld1 {v3.8b}, [x0], x2
295 srshr v28.8H, v28.8H, #6295 srshr v28.8h, v28.8h, #6
296 ld1 {v4.8B}, [x0], x2296 ld1 {v4.8b}, [x0], x2
297 srshr v29.8H, v29.8H, #6297 srshr v29.8h, v29.8h, #6
298 ld1 {v5.8B}, [x0], x2298 ld1 {v5.8b}, [x0], x2
299 srshr v30.8H, v30.8H, #6299 srshr v30.8h, v30.8h, #6
300 ld1 {v6.8B}, [x0], x2300 ld1 {v6.8b}, [x0], x2
301 srshr v31.8H, v31.8H, #6301 srshr v31.8h, v31.8h, #6
302 ld1 {v7.8B}, [x0], x2302 ld1 {v7.8b}, [x0], x2
303 uaddw v24.8H, v24.8H, v0.8B303 uaddw v24.8h, v24.8h, v0.8b
304 uaddw v25.8H, v25.8H, v1.8B304 uaddw v25.8h, v25.8h, v1.8b
305 uaddw v26.8H, v26.8H, v2.8B305 uaddw v26.8h, v26.8h, v2.8b
306 sqxtun v0.8B, v24.8H306 sqxtun v0.8b, v24.8h
307 uaddw v27.8H, v27.8H, v3.8B307 uaddw v27.8h, v27.8h, v3.8b
308 sqxtun v1.8B, v25.8H308 sqxtun v1.8b, v25.8h
309 uaddw v28.8H, v28.8H, v4.8B309 uaddw v28.8h, v28.8h, v4.8b
310 sqxtun v2.8B, v26.8H310 sqxtun v2.8b, v26.8h
311 st1 {v0.8B}, [x3], x2311 st1 {v0.8b}, [x3], x2
312 uaddw v29.8H, v29.8H, v5.8B312 uaddw v29.8h, v29.8h, v5.8b
313 sqxtun v3.8B, v27.8H313 sqxtun v3.8b, v27.8h
314 st1 {v1.8B}, [x3], x2314 st1 {v1.8b}, [x3], x2
315 uaddw v30.8H, v30.8H, v6.8B315 uaddw v30.8h, v30.8h, v6.8b
316 sqxtun v4.8B, v28.8H316 sqxtun v4.8b, v28.8h
317 st1 {v2.8B}, [x3], x2317 st1 {v2.8b}, [x3], x2
318 uaddw v31.8H, v31.8H, v7.8B318 uaddw v31.8h, v31.8h, v7.8b
319 sqxtun v5.8B, v29.8H319 sqxtun v5.8b, v29.8h
320 st1 {v3.8B}, [x3], x2320 st1 {v3.8b}, [x3], x2
321 sqxtun v6.8B, v30.8H321 sqxtun v6.8b, v30.8h
322 sqxtun v7.8B, v31.8H322 sqxtun v7.8b, v31.8h
323 st1 {v4.8B}, [x3], x2323 st1 {v4.8b}, [x3], x2
324 st1 {v5.8B}, [x3], x2324 st1 {v5.8b}, [x3], x2
325 st1 {v6.8B}, [x3], x2325 st1 {v6.8b}, [x3], x2
326 st1 {v7.8B}, [x3], x2326 st1 {v7.8b}, [x3], x2
327 327
328 sub x1, x1, #128328 sub x1, x1, #128
329 ret329 ret
@@ -334,42 +334,42 @@ function ff_h264_idct8_dc_add_neon, export=1 · 4 unmodified lines
334 AARCH64_VALID_CALL_TARGET334 AARCH64_VALID_CALL_TARGET
335 mov w3, #0335 mov w3, #0
336 sxtw x2, w2336 sxtw x2, w2
337 ld1r {v31.8H}, [x1]337 ld1r {v31.8h}, [x1]
338 strh w3, [x1]338 strh w3, [x1]
339 ld1 {v0.8B}, [x0], x2339 ld1 {v0.8b}, [x0], x2
340 srshr v31.8H, v31.8H, #6340 srshr v31.8h, v31.8h, #6
341 ld1 {v1.8B}, [x0], x2341 ld1 {v1.8b}, [x0], x2
342 ld1 {v2.8B}, [x0], x2342 ld1 {v2.8b}, [x0], x2
343 uaddw v24.8H, v31.8H, v0.8B343 uaddw v24.8h, v31.8h, v0.8b
344 ld1 {v3.8B}, [x0], x2344 ld1 {v3.8b}, [x0], x2
345 uaddw v25.8H, v31.8H, v1.8B345 uaddw v25.8h, v31.8h, v1.8b
346 ld1 {v4.8B}, [x0], x2346 ld1 {v4.8b}, [x0], x2
347 uaddw v26.8H, v31.8H, v2.8B347 uaddw v26.8h, v31.8h, v2.8b
348 ld1 {v5.8B}, [x0], x2348 ld1 {v5.8b}, [x0], x2
349 uaddw v27.8H, v31.8H, v3.8B349 uaddw v27.8h, v31.8h, v3.8b
350 ld1 {v6.8B}, [x0], x2350 ld1 {v6.8b}, [x0], x2
351 uaddw v28.8H, v31.8H, v4.8B351 uaddw v28.8h, v31.8h, v4.8b
352 ld1 {v7.8B}, [x0], x2352 ld1 {v7.8b}, [x0], x2
353 uaddw v29.8H, v31.8H, v5.8B353 uaddw v29.8h, v31.8h, v5.8b
354 uaddw v30.8H, v31.8H, v6.8B354 uaddw v30.8h, v31.8h, v6.8b
355 uaddw v31.8H, v31.8H, v7.8B355 uaddw v31.8h, v31.8h, v7.8b
356 sqxtun v0.8B, v24.8H356 sqxtun v0.8b, v24.8h
357 sqxtun v1.8B, v25.8H357 sqxtun v1.8b, v25.8h
358 sqxtun v2.8B, v26.8H358 sqxtun v2.8b, v26.8h
359 sqxtun v3.8B, v27.8H359 sqxtun v3.8b, v27.8h
360 sub x0, x0, x2, lsl #3360 sub x0, x0, x2, lsl #3
361 st1 {v0.8B}, [x0], x2361 st1 {v0.8b}, [x0], x2
362 sqxtun v4.8B, v28.8H362 sqxtun v4.8b, v28.8h
363 st1 {v1.8B}, [x0], x2363 st1 {v1.8b}, [x0], x2
364 sqxtun v5.8B, v29.8H364 sqxtun v5.8b, v29.8h
365 st1 {v2.8B}, [x0], x2365 st1 {v2.8b}, [x0], x2
366 sqxtun v6.8B, v30.8H366 sqxtun v6.8b, v30.8h
367 st1 {v3.8B}, [x0], x2367 st1 {v3.8b}, [x0], x2
368 sqxtun v7.8B, v31.8H368 sqxtun v7.8b, v31.8h
369 st1 {v4.8B}, [x0], x2369 st1 {v4.8b}, [x0], x2
370 st1 {v5.8B}, [x0], x2370 st1 {v5.8b}, [x0], x2
371 st1 {v6.8B}, [x0], x2371 st1 {v6.8b}, [x0], x2
372 st1 {v7.8B}, [x0], x2372 st1 {v7.8b}, [x0], x2
373 ret373 ret
374endfunc374endfunc
375 375
libavcodec/aarch64/h264qpel_neon.S+278−278
@@ -27,127 +27,127 @@ · 26 unmodified lines
27.macro lowpass_const r27.macro lowpass_const r
28 movz \r, #20, lsl #1628 movz \r, #20, lsl #16
29 movk \r, #529 movk \r, #5
30 mov v6.S[0], \r30 mov v6.s[0], \r
31.endm31.endm
32 32
33//trashes v0-v533//trashes v0-v5
34.macro lowpass_8 r0, r1, r2, r3, d0, d1, narrow=134.macro lowpass_8 r0, r1, r2, r3, d0, d1, narrow=1
35 ext v2.8B, \r0\().8B, \r1\().8B, #235 ext v2.8b, \r0\().8b, \r1\().8b, #2
36 ext v3.8B, \r0\().8B, \r1\().8B, #336 ext v3.8b, \r0\().8b, \r1\().8b, #3
37 uaddl v2.8H, v2.8B, v3.8B37 uaddl v2.8h, v2.8b, v3.8b
38 ext v4.8B, \r0\().8B, \r1\().8B, #138 ext v4.8b, \r0\().8b, \r1\().8b, #1
39 ext v5.8B, \r0\().8B, \r1\().8B, #439 ext v5.8b, \r0\().8b, \r1\().8b, #4
40 uaddl v4.8H, v4.8B, v5.8B40 uaddl v4.8h, v4.8b, v5.8b
41 ext v1.8B, \r0\().8B, \r1\().8B, #541 ext v1.8b, \r0\().8b, \r1\().8b, #5
42 uaddl \d0\().8H, \r0\().8B, v1.8B42 uaddl \d0\().8h, \r0\().8b, v1.8b
43 ext v0.8B, \r2\().8B, \r3\().8B, #243 ext v0.8b, \r2\().8b, \r3\().8b, #2
44 mla \d0\().8H, v2.8H, v6.H[1]44 mla \d0\().8h, v2.8h, v6.h[1]
45 ext v1.8B, \r2\().8B, \r3\().8B, #345 ext v1.8b, \r2\().8b, \r3\().8b, #3
46 uaddl v0.8H, v0.8B, v1.8B46 uaddl v0.8h, v0.8b, v1.8b
47 ext v1.8B, \r2\().8B, \r3\().8B, #147 ext v1.8b, \r2\().8b, \r3\().8b, #1
48 mls \d0\().8H, v4.8H, v6.H[0]48 mls \d0\().8h, v4.8h, v6.h[0]
49 ext v3.8B, \r2\().8B, \r3\().8B, #449 ext v3.8b, \r2\().8b, \r3\().8b, #4
50 uaddl v1.8H, v1.8B, v3.8B50 uaddl v1.8h, v1.8b, v3.8b
51 ext v2.8B, \r2\().8B, \r3\().8B, #551 ext v2.8b, \r2\().8b, \r3\().8b, #5
52 uaddl \d1\().8H, \r2\().8B, v2.8B52 uaddl \d1\().8h, \r2\().8b, v2.8b
53 mla \d1\().8H, v0.8H, v6.H[1]53 mla \d1\().8h, v0.8h, v6.h[1]
54 mls \d1\().8H, v1.8H, v6.H[0]54 mls \d1\().8h, v1.8h, v6.h[0]
55 .if \narrow55 .if \narrow
56 sqrshrun \d0\().8B, \d0\().8H, #556 sqrshrun \d0\().8b, \d0\().8h, #5
57 sqrshrun \d1\().8B, \d1\().8H, #557 sqrshrun \d1\().8b, \d1\().8h, #5
58 .endif58 .endif
59.endm59.endm
60 60
61//trashes v0-v461//trashes v0-v4
62.macro lowpass_8_v r0, r1, r2, r3, r4, r5, r6, d0, d1, narrow=162.macro lowpass_8_v r0, r1, r2, r3, r4, r5, r6, d0, d1, narrow=1
63 uaddl v2.8H, \r2\().8B, \r3\().8B63 uaddl v2.8h, \r2\().8b, \r3\().8b
64 uaddl v0.8H, \r3\().8B, \r4\().8B64 uaddl v0.8h, \r3\().8b, \r4\().8b
65 uaddl v4.8H, \r1\().8B, \r4\().8B65 uaddl v4.8h, \r1\().8b, \r4\().8b
66 uaddl v1.8H, \r2\().8B, \r5\().8B66 uaddl v1.8h, \r2\().8b, \r5\().8b
67 uaddl \d0\().8H, \r0\().8B, \r5\().8B67 uaddl \d0\().8h, \r0\().8b, \r5\().8b
68 uaddl \d1\().8H, \r1\().8B, \r6\().8B68 uaddl \d1\().8h, \r1\().8b, \r6\().8b
69 mla \d0\().8H, v2.8H, v6.H[1]69 mla \d0\().8h, v2.8h, v6.h[1]
70 mls \d0\().8H, v4.8H, v6.H[0]70 mls \d0\().8h, v4.8h, v6.h[0]
71 mla \d1\().8H, v0.8H, v6.H[1]71 mla \d1\().8h, v0.8h, v6.h[1]
72 mls \d1\().8H, v1.8H, v6.H[0]72 mls \d1\().8h, v1.8h, v6.h[0]
73 .if \narrow73 .if \narrow
74 sqrshrun \d0\().8B, \d0\().8H, #574 sqrshrun \d0\().8b, \d0\().8h, #5
75 sqrshrun \d1\().8B, \d1\().8H, #575 sqrshrun \d1\().8b, \d1\().8h, #5
76 .endif76 .endif
77.endm77.endm
78 78
79//trashes v0-v5, v7, v30-v3179//trashes v0-v5, v7, v30-v31
80.macro lowpass_8H r0, r180.macro lowpass_8H r0, r1
81 ext v0.16B, \r0\().16B, \r0\().16B, #281 ext v0.16b, \r0\().16b, \r0\().16b, #2
82 ext v1.16B, \r0\().16B, \r0\().16B, #382 ext v1.16b, \r0\().16b, \r0\().16b, #3
83 uaddl v0.8H, v0.8B, v1.8B83 uaddl v0.8h, v0.8b, v1.8b
84 ext v2.16B, \r0\().16B, \r0\().16B, #184 ext v2.16b, \r0\().16b, \r0\().16b, #1
85 ext v3.16B, \r0\().16B, \r0\().16B, #485 ext v3.16b, \r0\().16b, \r0\().16b, #4
86 uaddl v2.8H, v2.8B, v3.8B86 uaddl v2.8h, v2.8b, v3.8b
87 ext v30.16B, \r0\().16B, \r0\().16B, #587 ext v30.16b, \r0\().16b, \r0\().16b, #5
88 uaddl \r0\().8H, \r0\().8B, v30.8B88 uaddl \r0\().8h, \r0\().8b, v30.8b
89 ext v4.16B, \r1\().16B, \r1\().16B, #289 ext v4.16b, \r1\().16b, \r1\().16b, #2
90 mla \r0\().8H, v0.8H, v6.H[1]90 mla \r0\().8h, v0.8h, v6.h[1]
91 ext v5.16B, \r1\().16B, \r1\().16B, #391 ext v5.16b, \r1\().16b, \r1\().16b, #3
92 uaddl v4.8H, v4.8B, v5.8B92 uaddl v4.8h, v4.8b, v5.8b
93 ext v7.16B, \r1\().16B, \r1\().16B, #193 ext v7.16b, \r1\().16b, \r1\().16b, #1
94 mls \r0\().8H, v2.8H, v6.H[0]94 mls \r0\().8h, v2.8h, v6.h[0]
95 ext v0.16B, \r1\().16B, \r1\().16B, #495 ext v0.16b, \r1\().16b, \r1\().16b, #4
96 uaddl v7.8H, v7.8B, v0.8B96 uaddl v7.8h, v7.8b, v0.8b
97 ext v31.16B, \r1\().16B, \r1\().16B, #597 ext v31.16b, \r1\().16b, \r1\().16b, #5
98 uaddl \r1\().8H, \r1\().8B, v31.8B98 uaddl \r1\().8h, \r1\().8b, v31.8b
99 mla \r1\().8H, v4.8H, v6.H[1]99 mla \r1\().8h, v4.8h, v6.h[1]
100 mls \r1\().8H, v7.8H, v6.H[0]100 mls \r1\().8h, v7.8h, v6.h[0]
101.endm101.endm
102 102
103// trashes v2-v5, v30103// trashes v2-v5, v30
104.macro lowpass_8_1 r0, r1, d0, narrow=1104.macro lowpass_8_1 r0, r1, d0, narrow=1
105 ext v2.8B, \r0\().8B, \r1\().8B, #2105 ext v2.8b, \r0\().8b, \r1\().8b, #2
106 ext v3.8B, \r0\().8B, \r1\().8B, #3106 ext v3.8b, \r0\().8b, \r1\().8b, #3
107 uaddl v2.8H, v2.8B, v3.8B107 uaddl v2.8h, v2.8b, v3.8b
108 ext v4.8B, \r0\().8B, \r1\().8B, #1108 ext v4.8b, \r0\().8b, \r1\().8b, #1
109 ext v5.8B, \r0\().8B, \r1\().8B, #4109 ext v5.8b, \r0\().8b, \r1\().8b, #4
110 uaddl v4.8H, v4.8B, v5.8B110 uaddl v4.8h, v4.8b, v5.8b
111 ext v30.8B, \r0\().8B, \r1\().8B, #5111 ext v30.8b, \r0\().8b, \r1\().8b, #5
112 uaddl \d0\().8H, \r0\().8B, v30.8B112 uaddl \d0\().8h, \r0\().8b, v30.8b
113 mla \d0\().8H, v2.8H, v6.H[1]113 mla \d0\().8h, v2.8h, v6.h[1]
114 mls \d0\().8H, v4.8H, v6.H[0]114 mls \d0\().8h, v4.8h, v6.h[0]
115 .if \narrow115 .if \narrow
116 sqrshrun \d0\().8B, \d0\().8H, #5116 sqrshrun \d0\().8b, \d0\().8h, #5
117 .endif117 .endif
118.endm118.endm
119 119
120// trashed v0-v7120// trashed v0-v7
121.macro lowpass_8.16 r0, r1, r2, r3, r4, r5121.macro lowpass_8.16 r0, r1, r2, r3, r4, r5
122 saddl v5.4S, \r2\().4H, \r3\().4H122 saddl v5.4s, \r2\().4h, \r3\().4h
123 saddl2 v1.4S, \r2\().8H, \r3\().8H123 saddl2 v1.4s, \r2\().8h, \r3\().8h
124 saddl v6.4S, \r1\().4H, \r4\().4H124 saddl v6.4s, \r1\().4h, \r4\().4h
125 saddl2 v2.4S, \r1\().8H, \r4\().8H125 saddl2 v2.4s, \r1\().8h, \r4\().8h
126 saddl v0.4S, \r0\().4H, \r5\().4H126 saddl v0.4s, \r0\().4h, \r5\().4h
127 saddl2 v4.4S, \r0\().8H, \r5\().8H127 saddl2 v4.4s, \r0\().8h, \r5\().8h
128 128
129 shl v3.4S, v5.4S, #4129 shl v3.4s, v5.4s, #4
130 shl v5.4S, v5.4S, #2130 shl v5.4s, v5.4s, #2
131 shl v7.4S, v6.4S, #2131 shl v7.4s, v6.4s, #2
132 add v5.4S, v5.4S, v3.4S132 add v5.4s, v5.4s, v3.4s
133 add v6.4S, v6.4S, v7.4S133 add v6.4s, v6.4s, v7.4s
134 134
135 shl v3.4S, v1.4S, #4135 shl v3.4s, v1.4s, #4
136 shl v1.4S, v1.4S, #2136 shl v1.4s, v1.4s, #2
137 shl v7.4S, v2.4S, #2137 shl v7.4s, v2.4s, #2
138 add v1.4S, v1.4S, v3.4S138 add v1.4s, v1.4s, v3.4s
139 add v2.4S, v2.4S, v7.4S139 add v2.4s, v2.4s, v7.4s
140 140
141 add v5.4S, v5.4S, v0.4S141 add v5.4s, v5.4s, v0.4s
142 sub v5.4S, v5.4S, v6.4S142 sub v5.4s, v5.4s, v6.4s
143 143
144 add v1.4S, v1.4S, v4.4S144 add v1.4s, v1.4s, v4.4s
145 sub v1.4S, v1.4S, v2.4S145 sub v1.4s, v1.4s, v2.4s
146 146
147 rshrn v5.4H, v5.4S, #10147 rshrn v5.4h, v5.4s, #10
148 rshrn2 v5.8H, v1.4S, #10148 rshrn2 v5.8h, v1.4s, #10
149 149
150 sqxtun \r0\().8B, v5.8H150 sqxtun \r0\().8b, v5.8h
151.endm151.endm
152 152
153function put_h264_qpel16_h_lowpass_neon_packed153function put_h264_qpel16_h_lowpass_neon_packed
@@ -176,19 +176,19 @@ function \type\()_h264_qpel16_h_lowpass_neon · 22 unmodified lines
176endfunc176endfunc
177 177
178function \type\()_h264_qpel8_h_lowpass_neon178function \type\()_h264_qpel8_h_lowpass_neon
1791: ld1 {v28.8B, v29.8B}, [x1], x21791: ld1 {v28.8b, v29.8b}, [x1], x2
180 ld1 {v16.8B, v17.8B}, [x1], x2180 ld1 {v16.8b, v17.8b}, [x1], x2
181 subs x12, x12, #2181 subs x12, x12, #2
182 lowpass_8 v28, v29, v16, v17, v28, v16182 lowpass_8 v28, v29, v16, v17, v28, v16
183 .ifc \type,avg183 .ifc \type,avg
184 ld1 {v2.8B}, [x0], x3184 ld1 {v2.8b}, [x0], x3
185 ld1 {v3.8B}, [x0]185 ld1 {v3.8b}, [x0]
186 urhadd v28.8B, v28.8B, v2.8B186 urhadd v28.8b, v28.8b, v2.8b
187 urhadd v16.8B, v16.8B, v3.8B187 urhadd v16.8b, v16.8b, v3.8b
188 sub x0, x0, x3188 sub x0, x0, x3
189 .endif189 .endif
190 st1 {v28.8B}, [x0], x3190 st1 {v28.8b}, [x0], x3
191 st1 {v16.8B}, [x0], x3191 st1 {v16.8b}, [x0], x3
192 b.ne 1b192 b.ne 1b
193 ret193 ret
194endfunc194endfunc
@@ -213,23 +213,23 @@ function \type\()_h264_qpel16_h_lowpass_l2_neon · 18 unmodified lines
213endfunc213endfunc
214 214
215function \type\()_h264_qpel8_h_lowpass_l2_neon215function \type\()_h264_qpel8_h_lowpass_l2_neon
2161: ld1 {v26.8B, v27.8B}, [x1], x22161: ld1 {v26.8b, v27.8b}, [x1], x2
217 ld1 {v16.8B, v17.8B}, [x1], x2217 ld1 {v16.8b, v17.8b}, [x1], x2
218 ld1 {v28.8B}, [x3], x2218 ld1 {v28.8b}, [x3], x2
219 ld1 {v29.8B}, [x3], x2219 ld1 {v29.8b}, [x3], x2
220 subs x12, x12, #2220 subs x12, x12, #2
221 lowpass_8 v26, v27, v16, v17, v26, v27221 lowpass_8 v26, v27, v16, v17, v26, v27
222 urhadd v26.8B, v26.8B, v28.8B222 urhadd v26.8b, v26.8b, v28.8b
223 urhadd v27.8B, v27.8B, v29.8B223 urhadd v27.8b, v27.8b, v29.8b
224 .ifc \type,avg224 .ifc \type,avg
225 ld1 {v2.8B}, [x0], x2225 ld1 {v2.8b}, [x0], x2
226 ld1 {v3.8B}, [x0]226 ld1 {v3.8b}, [x0]
227 urhadd v26.8B, v26.8B, v2.8B227 urhadd v26.8b, v26.8b, v2.8b
228 urhadd v27.8B, v27.8B, v3.8B228 urhadd v27.8b, v27.8b, v3.8b
229 sub x0, x0, x2229 sub x0, x0, x2
230 .endif230 .endif
231 st1 {v26.8B}, [x0], x2231 st1 {v26.8b}, [x0], x2
232 st1 {v27.8B}, [x0], x2232 st1 {v27.8b}, [x0], x2
233 b.ne 1b233 b.ne 1b
234 ret234 ret
235endfunc235endfunc
@@ -270,52 +270,52 @@ function \type\()_h264_qpel16_v_lowpass_neon · 34 unmodified lines
270endfunc270endfunc
271 271
272function \type\()_h264_qpel8_v_lowpass_neon272function \type\()_h264_qpel8_v_lowpass_neon
273 ld1 {v16.8B}, [x1], x3273 ld1 {v16.8b}, [x1], x3
274 ld1 {v17.8B}, [x1], x3274 ld1 {v17.8b}, [x1], x3
275 ld1 {v18.8B}, [x1], x3275 ld1 {v18.8b}, [x1], x3
276 ld1 {v19.8B}, [x1], x3276 ld1 {v19.8b}, [x1], x3
277 ld1 {v20.8B}, [x1], x3277 ld1 {v20.8b}, [x1], x3
278 ld1 {v21.8B}, [x1], x3278 ld1 {v21.8b}, [x1], x3
279 ld1 {v22.8B}, [x1], x3279 ld1 {v22.8b}, [x1], x3
280 ld1 {v23.8B}, [x1], x3280 ld1 {v23.8b}, [x1], x3
281 ld1 {v24.8B}, [x1], x3281 ld1 {v24.8b}, [x1], x3
282 ld1 {v25.8B}, [x1], x3282 ld1 {v25.8b}, [x1], x3
283 ld1 {v26.8B}, [x1], x3283 ld1 {v26.8b}, [x1], x3
284 ld1 {v27.8B}, [x1], x3284 ld1 {v27.8b}, [x1], x3
285 ld1 {v28.8B}, [x1]285 ld1 {v28.8b}, [x1]
286 286
287 lowpass_8_v v16, v17, v18, v19, v20, v21, v22, v16, v17287 lowpass_8_v v16, v17, v18, v19, v20, v21, v22, v16, v17
288 lowpass_8_v v18, v19, v20, v21, v22, v23, v24, v18, v19288 lowpass_8_v v18, v19, v20, v21, v22, v23, v24, v18, v19
289 lowpass_8_v v20, v21, v22, v23, v24, v25, v26, v20, v21289 lowpass_8_v v20, v21, v22, v23, v24, v25, v26, v20, v21
290 lowpass_8_v v22, v23, v24, v25, v26, v27, v28, v22, v23290 lowpass_8_v v22, v23, v24, v25, v26, v27, v28, v22, v23
291 .ifc \type,avg291 .ifc \type,avg
292 ld1 {v24.8B}, [x0], x2292 ld1 {v24.8b}, [x0], x2
293 ld1 {v25.8B}, [x0], x2293 ld1 {v25.8b}, [x0], x2
294 ld1 {v26.8B}, [x0], x2294 ld1 {v26.8b}, [x0], x2
295 urhadd v16.8B, v16.8B, v24.8B295 urhadd v16.8b, v16.8b, v24.8b
296 ld1 {v27.8B}, [x0], x2296 ld1 {v27.8b}, [x0], x2
297 urhadd v17.8B, v17.8B, v25.8B297 urhadd v17.8b, v17.8b, v25.8b
298 ld1 {v28.8B}, [x0], x2298 ld1 {v28.8b}, [x0], x2
299 urhadd v18.8B, v18.8B, v26.8B299 urhadd v18.8b, v18.8b, v26.8b
300 ld1 {v29.8B}, [x0], x2300 ld1 {v29.8b}, [x0], x2
301 urhadd v19.8B, v19.8B, v27.8B301 urhadd v19.8b, v19.8b, v27.8b
302 ld1 {v30.8B}, [x0], x2302 ld1 {v30.8b}, [x0], x2
303 urhadd v20.8B, v20.8B, v28.8B303 urhadd v20.8b, v20.8b, v28.8b
304 ld1 {v31.8B}, [x0], x2304 ld1 {v31.8b}, [x0], x2
305 urhadd v21.8B, v21.8B, v29.8B305 urhadd v21.8b, v21.8b, v29.8b
306 urhadd v22.8B, v22.8B, v30.8B306 urhadd v22.8b, v22.8b, v30.8b
307 urhadd v23.8B, v23.8B, v31.8B307 urhadd v23.8b, v23.8b, v31.8b
308 sub x0, x0, x2, lsl #3308 sub x0, x0, x2, lsl #3
309 .endif309 .endif
310 310
311 st1 {v16.8B}, [x0], x2311 st1 {v16.8b}, [x0], x2
312 st1 {v17.8B}, [x0], x2312 st1 {v17.8b}, [x0], x2
313 st1 {v18.8B}, [x0], x2313 st1 {v18.8b}, [x0], x2
314 st1 {v19.8B}, [x0], x2314 st1 {v19.8b}, [x0], x2
315 st1 {v20.8B}, [x0], x2315 st1 {v20.8b}, [x0], x2
316 st1 {v21.8B}, [x0], x2316 st1 {v21.8b}, [x0], x2
317 st1 {v22.8B}, [x0], x2317 st1 {v22.8b}, [x0], x2
318 st1 {v23.8B}, [x0], x2318 st1 {v23.8b}, [x0], x2
319 319
320 ret320 ret
321endfunc321endfunc
@@ -343,70 +343,70 @@ function \type\()_h264_qpel16_v_lowpass_l2_neon · 21 unmodified lines
343endfunc343endfunc
344 344
345function \type\()_h264_qpel8_v_lowpass_l2_neon345function \type\()_h264_qpel8_v_lowpass_l2_neon
346 ld1 {v16.8B}, [x1], x3346 ld1 {v16.8b}, [x1], x3
347 ld1 {v17.8B}, [x1], x3347 ld1 {v17.8b}, [x1], x3
348 ld1 {v18.8B}, [x1], x3348 ld1 {v18.8b}, [x1], x3
349 ld1 {v19.8B}, [x1], x3349 ld1 {v19.8b}, [x1], x3
350 ld1 {v20.8B}, [x1], x3350 ld1 {v20.8b}, [x1], x3
351 ld1 {v21.8B}, [x1], x3351 ld1 {v21.8b}, [x1], x3
352 ld1 {v22.8B}, [x1], x3352 ld1 {v22.8b}, [x1], x3
353 ld1 {v23.8B}, [x1], x3353 ld1 {v23.8b}, [x1], x3
354 ld1 {v24.8B}, [x1], x3354 ld1 {v24.8b}, [x1], x3
355 ld1 {v25.8B}, [x1], x3355 ld1 {v25.8b}, [x1], x3
356 ld1 {v26.8B}, [x1], x3356 ld1 {v26.8b}, [x1], x3
357 ld1 {v27.8B}, [x1], x3357 ld1 {v27.8b}, [x1], x3
358 ld1 {v28.8B}, [x1]358 ld1 {v28.8b}, [x1]
359 359
360 lowpass_8_v v16, v17, v18, v19, v20, v21, v22, v16, v17360 lowpass_8_v v16, v17, v18, v19, v20, v21, v22, v16, v17
361 lowpass_8_v v18, v19, v20, v21, v22, v23, v24, v18, v19361 lowpass_8_v v18, v19, v20, v21, v22, v23, v24, v18, v19
362 lowpass_8_v v20, v21, v22, v23, v24, v25, v26, v20, v21362 lowpass_8_v v20, v21, v22, v23, v24, v25, v26, v20, v21
363 lowpass_8_v v22, v23, v24, v25, v26, v27, v28, v22, v23363 lowpass_8_v v22, v23, v24, v25, v26, v27, v28, v22, v23
364 364
365 ld1 {v24.8B}, [x12], x2365 ld1 {v24.8b}, [x12], x2
366 ld1 {v25.8B}, [x12], x2366 ld1 {v25.8b}, [x12], x2
367 ld1 {v26.8B}, [x12], x2367 ld1 {v26.8b}, [x12], x2
368 ld1 {v27.8B}, [x12], x2368 ld1 {v27.8b}, [x12], x2
369 ld1 {v28.8B}, [x12], x2369 ld1 {v28.8b}, [x12], x2
370 urhadd v16.8B, v24.8B, v16.8B370 urhadd v16.8b, v24.8b, v16.8b
371 urhadd v17.8B, v25.8B, v17.8B371 urhadd v17.8b, v25.8b, v17.8b
372 ld1 {v29.8B}, [x12], x2372 ld1 {v29.8b}, [x12], x2
373 urhadd v18.8B, v26.8B, v18.8B373 urhadd v18.8b, v26.8b, v18.8b
374 urhadd v19.8B, v27.8B, v19.8B374 urhadd v19.8b, v27.8b, v19.8b
375 ld1 {v30.8B}, [x12], x2375 ld1 {v30.8b}, [x12], x2
376 urhadd v20.8B, v28.8B, v20.8B376 urhadd v20.8b, v28.8b, v20.8b
377 urhadd v21.8B, v29.8B, v21.8B377 urhadd v21.8b, v29.8b, v21.8b
378 ld1 {v31.8B}, [x12], x2378 ld1 {v31.8b}, [x12], x2
379 urhadd v22.8B, v30.8B, v22.8B379 urhadd v22.8b, v30.8b, v22.8b
380 urhadd v23.8B, v31.8B, v23.8B380 urhadd v23.8b, v31.8b, v23.8b
381 381
382 .ifc \type,avg382 .ifc \type,avg
383 ld1 {v24.8B}, [x0], x3383 ld1 {v24.8b}, [x0], x3
384 ld1 {v25.8B}, [x0], x3384 ld1 {v25.8b}, [x0], x3
385 ld1 {v26.8B}, [x0], x3385 ld1 {v26.8b}, [x0], x3
386 urhadd v16.8B, v16.8B, v24.8B386 urhadd v16.8b, v16.8b, v24.8b
387 ld1 {v27.8B}, [x0], x3387 ld1 {v27.8b}, [x0], x3
388 urhadd v17.8B, v17.8B, v25.8B388 urhadd v17.8b, v17.8b, v25.8b
389 ld1 {v28.8B}, [x0], x3389 ld1 {v28.8b}, [x0], x3
390 urhadd v18.8B, v18.8B, v26.8B390 urhadd v18.8b, v18.8b, v26.8b
391 ld1 {v29.8B}, [x0], x3391 ld1 {v29.8b}, [x0], x3
392 urhadd v19.8B, v19.8B, v27.8B392 urhadd v19.8b, v19.8b, v27.8b
393 ld1 {v30.8B}, [x0], x3393 ld1 {v30.8b}, [x0], x3
394 urhadd v20.8B, v20.8B, v28.8B394 urhadd v20.8b, v20.8b, v28.8b
395 ld1 {v31.8B}, [x0], x3395 ld1 {v31.8b}, [x0], x3
396 urhadd v21.8B, v21.8B, v29.8B396 urhadd v21.8b, v21.8b, v29.8b
397 urhadd v22.8B, v22.8B, v30.8B397 urhadd v22.8b, v22.8b, v30.8b
398 urhadd v23.8B, v23.8B, v31.8B398 urhadd v23.8b, v23.8b, v31.8b
399 sub x0, x0, x3, lsl #3399 sub x0, x0, x3, lsl #3
400 .endif400 .endif
401 401
402 st1 {v16.8B}, [x0], x3402 st1 {v16.8b}, [x0], x3
403 st1 {v17.8B}, [x0], x3403 st1 {v17.8b}, [x0], x3
404 st1 {v18.8B}, [x0], x3404 st1 {v18.8b}, [x0], x3
405 st1 {v19.8B}, [x0], x3405 st1 {v19.8b}, [x0], x3
406 st1 {v20.8B}, [x0], x3406 st1 {v20.8b}, [x0], x3
407 st1 {v21.8B}, [x0], x3407 st1 {v21.8b}, [x0], x3
408 st1 {v22.8B}, [x0], x3408 st1 {v22.8b}, [x0], x3
409 st1 {v23.8B}, [x0], x3409 st1 {v23.8b}, [x0], x3
410 410
411 ret411 ret
412endfunc412endfunc
@@ -417,19 +417,19 @@ endfunc · 4 unmodified lines
417 417
418function put_h264_qpel8_hv_lowpass_neon_top418function put_h264_qpel8_hv_lowpass_neon_top
419 lowpass_const w12419 lowpass_const w12
420 ld1 {v16.8H}, [x1], x3420 ld1 {v16.8h}, [x1], x3
421 ld1 {v17.8H}, [x1], x3421 ld1 {v17.8h}, [x1], x3
422 ld1 {v18.8H}, [x1], x3422 ld1 {v18.8h}, [x1], x3
423 ld1 {v19.8H}, [x1], x3423 ld1 {v19.8h}, [x1], x3
424 ld1 {v20.8H}, [x1], x3424 ld1 {v20.8h}, [x1], x3
425 ld1 {v21.8H}, [x1], x3425 ld1 {v21.8h}, [x1], x3
426 ld1 {v22.8H}, [x1], x3426 ld1 {v22.8h}, [x1], x3
427 ld1 {v23.8H}, [x1], x3427 ld1 {v23.8h}, [x1], x3
428 ld1 {v24.8H}, [x1], x3428 ld1 {v24.8h}, [x1], x3
429 ld1 {v25.8H}, [x1], x3429 ld1 {v25.8h}, [x1], x3
430 ld1 {v26.8H}, [x1], x3430 ld1 {v26.8h}, [x1], x3
431 ld1 {v27.8H}, [x1], x3431 ld1 {v27.8h}, [x1], x3
432 ld1 {v28.8H}, [x1]432 ld1 {v28.8h}, [x1]
433 lowpass_8H v16, v17433 lowpass_8H v16, v17
434 lowpass_8H v18, v19434 lowpass_8H v18, v19
435 lowpass_8H v20, v21435 lowpass_8H v20, v21
@@ -458,33 +458,33 @@ function \type\()_h264_qpel8_hv_lowpass_neon · 22 unmodified lines
458 mov x10, x30458 mov x10, x30
459 bl put_h264_qpel8_hv_lowpass_neon_top459 bl put_h264_qpel8_hv_lowpass_neon_top
460 .ifc \type,avg460 .ifc \type,avg
461 ld1 {v0.8B}, [x0], x2461 ld1 {v0.8b}, [x0], x2
462 ld1 {v1.8B}, [x0], x2462 ld1 {v1.8b}, [x0], x2
463 ld1 {v2.8B}, [x0], x2463 ld1 {v2.8b}, [x0], x2
464 urhadd v16.8B, v16.8B, v0.8B464 urhadd v16.8b, v16.8b, v0.8b
465 ld1 {v3.8B}, [x0], x2465 ld1 {v3.8b}, [x0], x2
466 urhadd v17.8B, v17.8B, v1.8B466 urhadd v17.8b, v17.8b, v1.8b
467 ld1 {v4.8B}, [x0], x2467 ld1 {v4.8b}, [x0], x2
468 urhadd v18.8B, v18.8B, v2.8B468 urhadd v18.8b, v18.8b, v2.8b
469 ld1 {v5.8B}, [x0], x2469 ld1 {v5.8b}, [x0], x2
470 urhadd v19.8B, v19.8B, v3.8B470 urhadd v19.8b, v19.8b, v3.8b
471 ld1 {v6.8B}, [x0], x2471 ld1 {v6.8b}, [x0], x2
472 urhadd v20.8B, v20.8B, v4.8B472 urhadd v20.8b, v20.8b, v4.8b
473 ld1 {v7.8B}, [x0], x2473 ld1 {v7.8b}, [x0], x2
474 urhadd v21.8B, v21.8B, v5.8B474 urhadd v21.8b, v21.8b, v5.8b
475 urhadd v22.8B, v22.8B, v6.8B475 urhadd v22.8b, v22.8b, v6.8b
476 urhadd v23.8B, v23.8B, v7.8B476 urhadd v23.8b, v23.8b, v7.8b
477 sub x0, x0, x2, lsl #3477 sub x0, x0, x2, lsl #3
478 .endif478 .endif
479 479
480 st1 {v16.8B}, [x0], x2480 st1 {v16.8b}, [x0], x2
481 st1 {v17.8B}, [x0], x2481 st1 {v17.8b}, [x0], x2
482 st1 {v18.8B}, [x0], x2482 st1 {v18.8b}, [x0], x2
483 st1 {v19.8B}, [x0], x2483 st1 {v19.8b}, [x0], x2
484 st1 {v20.8B}, [x0], x2484 st1 {v20.8b}, [x0], x2
485 st1 {v21.8B}, [x0], x2485 st1 {v21.8b}, [x0], x2
486 st1 {v22.8B}, [x0], x2486 st1 {v22.8b}, [x0], x2
487 st1 {v23.8B}, [x0], x2487 st1 {v23.8b}, [x0], x2
488 488
489 ret x10489 ret x10
490endfunc490endfunc
@@ -498,45 +498,45 @@ function \type\()_h264_qpel8_hv_lowpass_l2_neon · 7 unmodified lines
498 mov x10, x30498 mov x10, x30
499 bl put_h264_qpel8_hv_lowpass_neon_top499 bl put_h264_qpel8_hv_lowpass_neon_top
500 500
501 ld1 {v0.8B, v1.8B}, [x2], #16501 ld1 {v0.8b, v1.8b}, [x2], #16
502 ld1 {v2.8B, v3.8B}, [x2], #16502 ld1 {v2.8b, v3.8b}, [x2], #16
503 urhadd v0.8B, v0.8B, v16.8B503 urhadd v0.8b, v0.8b, v16.8b
504 urhadd v1.8B, v1.8B, v17.8B504 urhadd v1.8b, v1.8b, v17.8b
505 ld1 {v4.8B, v5.8B}, [x2], #16505 ld1 {v4.8b, v5.8b}, [x2], #16
506 urhadd v2.8B, v2.8B, v18.8B506 urhadd v2.8b, v2.8b, v18.8b
507 urhadd v3.8B, v3.8B, v19.8B507 urhadd v3.8b, v3.8b, v19.8b
508 ld1 {v6.8B, v7.8B}, [x2], #16508 ld1 {v6.8b, v7.8b}, [x2], #16
509 urhadd v4.8B, v4.8B, v20.8B509 urhadd v4.8b, v4.8b, v20.8b
510 urhadd v5.8B, v5.8B, v21.8B510 urhadd v5.8b, v5.8b, v21.8b
511 urhadd v6.8B, v6.8B, v22.8B511 urhadd v6.8b, v6.8b, v22.8b
512 urhadd v7.8B, v7.8B, v23.8B512 urhadd v7.8b, v7.8b, v23.8b
513 .ifc \type,avg513 .ifc \type,avg
514 ld1 {v16.8B}, [x0], x3514 ld1 {v16.8b}, [x0], x3
515 ld1 {v17.8B}, [x0], x3515 ld1 {v17.8b}, [x0], x3
516 ld1 {v18.8B}, [x0], x3516 ld1 {v18.8b}, [x0], x3
517 urhadd v0.8B, v0.8B, v16.8B517 urhadd v0.8b, v0.8b, v16.8b
518 ld1 {v19.8B}, [x0], x3518 ld1 {v19.8b}, [x0], x3
519 urhadd v1.8B, v1.8B, v17.8B519 urhadd v1.8b, v1.8b, v17.8b
520 ld1 {v20.8B}, [x0], x3520 ld1 {v20.8b}, [x0], x3
521 urhadd v2.8B, v2.8B, v18.8B521 urhadd v2.8b, v2.8b, v18.8b
522 ld1 {v21.8B}, [x0], x3522 ld1 {v21.8b}, [x0], x3
523 urhadd v3.8B, v3.8B, v19.8B523 urhadd v3.8b, v3.8b, v19.8b
524 ld1 {v22.8B}, [x0], x3524 ld1 {v22.8b}, [x0], x3
525 urhadd v4.8B, v4.8B, v20.8B525 urhadd v4.8b, v4.8b, v20.8b
526 ld1 {v23.8B}, [x0], x3526 ld1 {v23.8b}, [x0], x3
527 urhadd v5.8B, v5.8B, v21.8B527 urhadd v5.8b, v5.8b, v21.8b
528 urhadd v6.8B, v6.8B, v22.8B528 urhadd v6.8b, v6.8b, v22.8b
529 urhadd v7.8B, v7.8B, v23.8B529 urhadd v7.8b, v7.8b, v23.8b
530 sub x0, x0, x3, lsl #3530 sub x0, x0, x3, lsl #3
531 .endif531 .endif
532 st1 {v0.8B}, [x0], x3532 st1 {v0.8b}, [x0], x3
533 st1 {v1.8B}, [x0], x3533 st1 {v1.8b}, [x0], x3
534 st1 {v2.8B}, [x0], x3534 st1 {v2.8b}, [x0], x3
535 st1 {v3.8B}, [x0], x3535 st1 {v3.8b}, [x0], x3
536 st1 {v4.8B}, [x0], x3536 st1 {v4.8b}, [x0], x3
537 st1 {v5.8B}, [x0], x3537 st1 {v5.8b}, [x0], x3
538 st1 {v6.8B}, [x0], x3538 st1 {v6.8b}, [x0], x3
539 st1 {v7.8B}, [x0], x3539 st1 {v7.8b}, [x0], x3
540 540
541 ret x10541 ret x10
542endfunc542endfunc
libavcodec/aarch64/hpeldsp_neon.S+181−181
@@ -26,295 +26,295 @@ · 25 unmodified lines
26 .if \avg26 .if \avg
27 mov x12, x027 mov x12, x0
28 .endif28 .endif
291: ld1 {v0.16B}, [x1], x2291: ld1 {v0.16b}, [x1], x2
30 ld1 {v1.16B}, [x1], x230 ld1 {v1.16b}, [x1], x2
31 ld1 {v2.16B}, [x1], x231 ld1 {v2.16b}, [x1], x2
32 ld1 {v3.16B}, [x1], x232 ld1 {v3.16b}, [x1], x2
33 .if \avg33 .if \avg
34 ld1 {v4.16B}, [x12], x234 ld1 {v4.16b}, [x12], x2
35 urhadd v0.16B, v0.16B, v4.16B35 urhadd v0.16b, v0.16b, v4.16b
36 ld1 {v5.16B}, [x12], x236 ld1 {v5.16b}, [x12], x2
37 urhadd v1.16B, v1.16B, v5.16B37 urhadd v1.16b, v1.16b, v5.16b
38 ld1 {v6.16B}, [x12], x238 ld1 {v6.16b}, [x12], x2
39 urhadd v2.16B, v2.16B, v6.16B39 urhadd v2.16b, v2.16b, v6.16b
40 ld1 {v7.16B}, [x12], x240 ld1 {v7.16b}, [x12], x2
41 urhadd v3.16B, v3.16B, v7.16B41 urhadd v3.16b, v3.16b, v7.16b
42 .endif42 .endif
43 subs w3, w3, #443 subs w3, w3, #4
44 st1 {v0.16B}, [x0], x244 st1 {v0.16b}, [x0], x2
45 st1 {v1.16B}, [x0], x245 st1 {v1.16b}, [x0], x2
46 st1 {v2.16B}, [x0], x246 st1 {v2.16b}, [x0], x2
47 st1 {v3.16B}, [x0], x247 st1 {v3.16b}, [x0], x2
48 b.ne 1b48 b.ne 1b
49 ret49 ret
50.endm50.endm
51 51
52.macro pixels16_x2 rnd=1, avg=052.macro pixels16_x2 rnd=1, avg=0
531: ld1 {v0.16B, v1.16B}, [x1], x2531: ld1 {v0.16b, v1.16b}, [x1], x2
54 ld1 {v2.16B, v3.16B}, [x1], x254 ld1 {v2.16b, v3.16b}, [x1], x2
55 subs w3, w3, #255 subs w3, w3, #2
56 ext v1.16B, v0.16B, v1.16B, #156 ext v1.16b, v0.16b, v1.16b, #1
57 avg v0.16B, v0.16B, v1.16B57 avg v0.16b, v0.16b, v1.16b
58 ext v3.16B, v2.16B, v3.16B, #158 ext v3.16b, v2.16b, v3.16b, #1
59 avg v2.16B, v2.16B, v3.16B59 avg v2.16b, v2.16b, v3.16b
60 .if \avg60 .if \avg
61 ld1 {v1.16B}, [x0], x261 ld1 {v1.16b}, [x0], x2
62 ld1 {v3.16B}, [x0]62 ld1 {v3.16b}, [x0]
63 urhadd v0.16B, v0.16B, v1.16B63 urhadd v0.16b, v0.16b, v1.16b
64 urhadd v2.16B, v2.16B, v3.16B64 urhadd v2.16b, v2.16b, v3.16b
65 sub x0, x0, x265 sub x0, x0, x2
66 .endif66 .endif
67 st1 {v0.16B}, [x0], x267 st1 {v0.16b}, [x0], x2
68 st1 {v2.16B}, [x0], x268 st1 {v2.16b}, [x0], x2
69 b.ne 1b69 b.ne 1b
70 ret70 ret
71.endm71.endm
72 72
73.macro pixels16_y2 rnd=1, avg=073.macro pixels16_y2 rnd=1, avg=0
74 sub w3, w3, #274 sub w3, w3, #2
75 ld1 {v0.16B}, [x1], x275 ld1 {v0.16b}, [x1], x2
76 ld1 {v1.16B}, [x1], x276 ld1 {v1.16b}, [x1], x2
771: subs w3, w3, #2771: subs w3, w3, #2
78 avg v2.16B, v0.16B, v1.16B78 avg v2.16b, v0.16b, v1.16b
79 ld1 {v0.16B}, [x1], x279 ld1 {v0.16b}, [x1], x2
80 avg v3.16B, v0.16B, v1.16B80 avg v3.16b, v0.16b, v1.16b
81 ld1 {v1.16B}, [x1], x281 ld1 {v1.16b}, [x1], x2
82 .if \avg82 .if \avg
83 ld1 {v4.16B}, [x0], x283 ld1 {v4.16b}, [x0], x2
84 ld1 {v5.16B}, [x0]84 ld1 {v5.16b}, [x0]
85 urhadd v2.16B, v2.16B, v4.16B85 urhadd v2.16b, v2.16b, v4.16b
86 urhadd v3.16B, v3.16B, v5.16B86 urhadd v3.16b, v3.16b, v5.16b
87 sub x0, x0, x287 sub x0, x0, x2
88 .endif88 .endif
89 st1 {v2.16B}, [x0], x289 st1 {v2.16b}, [x0], x2
90 st1 {v3.16B}, [x0], x290 st1 {v3.16b}, [x0], x2
91 b.ne 1b91 b.ne 1b
92 92
93 avg v2.16B, v0.16B, v1.16B93 avg v2.16b, v0.16b, v1.16b
94 ld1 {v0.16B}, [x1], x294 ld1 {v0.16b}, [x1], x2
95 avg v3.16B, v0.16B, v1.16B95 avg v3.16b, v0.16b, v1.16b
96 .if \avg96 .if \avg
97 ld1 {v4.16B}, [x0], x297 ld1 {v4.16b}, [x0], x2
98 ld1 {v5.16B}, [x0]98 ld1 {v5.16b}, [x0]
99 urhadd v2.16B, v2.16B, v4.16B99 urhadd v2.16b, v2.16b, v4.16b
100 urhadd v3.16B, v3.16B, v5.16B100 urhadd v3.16b, v3.16b, v5.16b
101 sub x0, x0, x2101 sub x0, x0, x2
102 .endif102 .endif
103 st1 {v2.16B}, [x0], x2103 st1 {v2.16b}, [x0], x2
104 st1 {v3.16B}, [x0], x2104 st1 {v3.16b}, [x0], x2
105 105
106 ret106 ret
107.endm107.endm
108 108
109.macro pixels16_xy2 rnd=1, avg=0109.macro pixels16_xy2 rnd=1, avg=0
110 sub w3, w3, #2110 sub w3, w3, #2
111 ld1 {v0.16B, v1.16B}, [x1], x2111 ld1 {v0.16b, v1.16b}, [x1], x2
112 ld1 {v4.16B, v5.16B}, [x1], x2112 ld1 {v4.16b, v5.16b}, [x1], x2
113NRND movi v26.8H, #1113NRND movi v26.8H, #1
114 ext v1.16B, v0.16B, v1.16B, #1114 ext v1.16b, v0.16b, v1.16b, #1
115 ext v5.16B, v4.16B, v5.16B, #1115 ext v5.16b, v4.16b, v5.16b, #1
116 uaddl v16.8H, v0.8B, v1.8B116 uaddl v16.8h, v0.8b, v1.8b
117 uaddl2 v20.8H, v0.16B, v1.16B117 uaddl2 v20.8h, v0.16b, v1.16b
118 uaddl v18.8H, v4.8B, v5.8B118 uaddl v18.8h, v4.8b, v5.8b
119 uaddl2 v22.8H, v4.16B, v5.16B119 uaddl2 v22.8h, v4.16b, v5.16b
1201: subs w3, w3, #21201: subs w3, w3, #2
121 ld1 {v0.16B, v1.16B}, [x1], x2121 ld1 {v0.16b, v1.16b}, [x1], x2
122 add v24.8H, v16.8H, v18.8H122 add v24.8h, v16.8h, v18.8h
123NRND add v24.8H, v24.8H, v26.8H123NRND add v24.8H, v24.8H, v26.8H
124 ext v30.16B, v0.16B, v1.16B, #1124 ext v30.16b, v0.16b, v1.16b, #1
125 add v1.8H, v20.8H, v22.8H125 add v1.8h, v20.8h, v22.8h
126 mshrn v28.8B, v24.8H, #2126 mshrn v28.8b, v24.8h, #2
127NRND add v1.8H, v1.8H, v26.8H127NRND add v1.8H, v1.8H, v26.8H
128 mshrn2 v28.16B, v1.8H, #2128 mshrn2 v28.16b, v1.8h, #2
129 .if \avg129 .if \avg
130 ld1 {v16.16B}, [x0]130 ld1 {v16.16b}, [x0]
131 urhadd v28.16B, v28.16B, v16.16B131 urhadd v28.16b, v28.16b, v16.16b
132 .endif132 .endif
133 uaddl v16.8H, v0.8B, v30.8B133 uaddl v16.8h, v0.8b, v30.8b
134 ld1 {v2.16B, v3.16B}, [x1], x2134 ld1 {v2.16b, v3.16b}, [x1], x2
135 uaddl2 v20.8H, v0.16B, v30.16B135 uaddl2 v20.8h, v0.16b, v30.16b
136 st1 {v28.16B}, [x0], x2136 st1 {v28.16b}, [x0], x2
137 add v24.8H, v16.8H, v18.8H137 add v24.8h, v16.8h, v18.8h
138NRND add v24.8H, v24.8H, v26.8H138NRND add v24.8H, v24.8H, v26.8H
139 ext v3.16B, v2.16B, v3.16B, #1139 ext v3.16b, v2.16b, v3.16b, #1
140 add v0.8H, v20.8H, v22.8H140 add v0.8h, v20.8h, v22.8h
141 mshrn v30.8B, v24.8H, #2141 mshrn v30.8b, v24.8h, #2
142NRND add v0.8H, v0.8H, v26.8H142NRND add v0.8H, v0.8H, v26.8H
143 mshrn2 v30.16B, v0.8H, #2143 mshrn2 v30.16b, v0.8h, #2
144 .if \avg144 .if \avg
145 ld1 {v18.16B}, [x0]145 ld1 {v18.16b}, [x0]
146 urhadd v30.16B, v30.16B, v18.16B146 urhadd v30.16b, v30.16b, v18.16b
147 .endif147 .endif
148 uaddl v18.8H, v2.8B, v3.8B148 uaddl v18.8h, v2.8b, v3.8b
149 uaddl2 v22.8H, v2.16B, v3.16B149 uaddl2 v22.8h, v2.16b, v3.16b
150 st1 {v30.16B}, [x0], x2150 st1 {v30.16b}, [x0], x2
151 b.gt 1b151 b.gt 1b
152 152
153 ld1 {v0.16B, v1.16B}, [x1], x2153 ld1 {v0.16b, v1.16b}, [x1], x2
154 add v24.8H, v16.8H, v18.8H154 add v24.8h, v16.8h, v18.8h
155NRND add v24.8H, v24.8H, v26.8H155NRND add v24.8H, v24.8H, v26.8H
156 ext v30.16B, v0.16B, v1.16B, #1156 ext v30.16b, v0.16b, v1.16b, #1
157 add v1.8H, v20.8H, v22.8H157 add v1.8h, v20.8h, v22.8h
158 mshrn v28.8B, v24.8H, #2158 mshrn v28.8b, v24.8h, #2
159NRND add v1.8H, v1.8H, v26.8H159NRND add v1.8H, v1.8H, v26.8H
160 mshrn2 v28.16B, v1.8H, #2160 mshrn2 v28.16b, v1.8h, #2
161 .if \avg161 .if \avg
162 ld1 {v16.16B}, [x0]162 ld1 {v16.16b}, [x0]
163 urhadd v28.16B, v28.16B, v16.16B163 urhadd v28.16b, v28.16b, v16.16b
164 .endif164 .endif
165 uaddl v16.8H, v0.8B, v30.8B165 uaddl v16.8h, v0.8b, v30.8b
166 uaddl2 v20.8H, v0.16B, v30.16B166 uaddl2 v20.8h, v0.16b, v30.16b
167 st1 {v28.16B}, [x0], x2167 st1 {v28.16b}, [x0], x2
168 add v24.8H, v16.8H, v18.8H168 add v24.8h, v16.8h, v18.8h
169NRND add v24.8H, v24.8H, v26.8H169NRND add v24.8H, v24.8H, v26.8H
170 add v0.8H, v20.8H, v22.8H170 add v0.8h, v20.8h, v22.8h
171 mshrn v30.8B, v24.8H, #2171 mshrn v30.8b, v24.8h, #2
172NRND add v0.8H, v0.8H, v26.8H172NRND add v0.8H, v0.8H, v26.8H
173 mshrn2 v30.16B, v0.8H, #2173 mshrn2 v30.16b, v0.8h, #2
174 .if \avg174 .if \avg
175 ld1 {v18.16B}, [x0]175 ld1 {v18.16b}, [x0]
176 urhadd v30.16B, v30.16B, v18.16B176 urhadd v30.16b, v30.16b, v18.16b
177 .endif177 .endif
178 st1 {v30.16B}, [x0], x2178 st1 {v30.16b}, [x0], x2
179 179
180 ret180 ret
181.endm181.endm
182 182
183.macro pixels8 rnd=1, avg=0183.macro pixels8 rnd=1, avg=0
1841: ld1 {v0.8B}, [x1], x21841: ld1 {v0.8b}, [x1], x2
185 ld1 {v1.8B}, [x1], x2185 ld1 {v1.8b}, [x1], x2
186 ld1 {v2.8B}, [x1], x2186 ld1 {v2.8b}, [x1], x2
187 ld1 {v3.8B}, [x1], x2187 ld1 {v3.8b}, [x1], x2
188 .if \avg188 .if \avg
189 ld1 {v4.8B}, [x0], x2189 ld1 {v4.8b}, [x0], x2
190 urhadd v0.8B, v0.8B, v4.8B190 urhadd v0.8b, v0.8b, v4.8b
191 ld1 {v5.8B}, [x0], x2191 ld1 {v5.8b}, [x0], x2
192 urhadd v1.8B, v1.8B, v5.8B192 urhadd v1.8b, v1.8b, v5.8b
193 ld1 {v6.8B}, [x0], x2193 ld1 {v6.8b}, [x0], x2
194 urhadd v2.8B, v2.8B, v6.8B194 urhadd v2.8b, v2.8b, v6.8b
195 ld1 {v7.8B}, [x0], x2195 ld1 {v7.8b}, [x0], x2
196 urhadd v3.8B, v3.8B, v7.8B196 urhadd v3.8b, v3.8b, v7.8b
197 sub x0, x0, x2, lsl #2197 sub x0, x0, x2, lsl #2
198 .endif198 .endif
199 subs w3, w3, #4199 subs w3, w3, #4
200 st1 {v0.8B}, [x0], x2200 st1 {v0.8b}, [x0], x2
201 st1 {v1.8B}, [x0], x2201 st1 {v1.8b}, [x0], x2
202 st1 {v2.8B}, [x0], x2202 st1 {v2.8b}, [x0], x2
203 st1 {v3.8B}, [x0], x2203 st1 {v3.8b}, [x0], x2
204 b.ne 1b204 b.ne 1b
205 ret205 ret
206.endm206.endm
207 207
208.macro pixels8_x2 rnd=1, avg=0208.macro pixels8_x2 rnd=1, avg=0
2091: ld1 {v0.8B, v1.8B}, [x1], x22091: ld1 {v0.8b, v1.8b}, [x1], x2
210 ext v1.8B, v0.8B, v1.8B, #1210 ext v1.8b, v0.8b, v1.8b, #1
211 ld1 {v2.8B, v3.8B}, [x1], x2211 ld1 {v2.8b, v3.8b}, [x1], x2
212 ext v3.8B, v2.8B, v3.8B, #1212 ext v3.8b, v2.8b, v3.8b, #1
213 subs w3, w3, #2213 subs w3, w3, #2
214 avg v0.8B, v0.8B, v1.8B214 avg v0.8b, v0.8b, v1.8b
215 avg v2.8B, v2.8B, v3.8B215 avg v2.8b, v2.8b, v3.8b
216 .if \avg216 .if \avg
217 ld1 {v4.8B}, [x0], x2217 ld1 {v4.8b}, [x0], x2
218 ld1 {v5.8B}, [x0]218 ld1 {v5.8b}, [x0]
219 urhadd v0.8B, v0.8B, v4.8B219 urhadd v0.8b, v0.8b, v4.8b
220 urhadd v2.8B, v2.8B, v5.8B220 urhadd v2.8b, v2.8b, v5.8b
221 sub x0, x0, x2221 sub x0, x0, x2
222 .endif222 .endif
223 st1 {v0.8B}, [x0], x2223 st1 {v0.8b}, [x0], x2
224 st1 {v2.8B}, [x0], x2224 st1 {v2.8b}, [x0], x2
225 b.ne 1b225 b.ne 1b
226 ret226 ret
227.endm227.endm
228 228
229.macro pixels8_y2 rnd=1, avg=0229.macro pixels8_y2 rnd=1, avg=0
230 sub w3, w3, #2230 sub w3, w3, #2
231 ld1 {v0.8B}, [x1], x2231 ld1 {v0.8b}, [x1], x2
232 ld1 {v1.8B}, [x1], x2232 ld1 {v1.8b}, [x1], x2
2331: subs w3, w3, #22331: subs w3, w3, #2
234 avg v4.8B, v0.8B, v1.8B234 avg v4.8b, v0.8b, v1.8b
235 ld1 {v0.8B}, [x1], x2235 ld1 {v0.8b}, [x1], x2
236 avg v5.8B, v0.8B, v1.8B236 avg v5.8b, v0.8b, v1.8b
237 ld1 {v1.8B}, [x1], x2237 ld1 {v1.8b}, [x1], x2
238 .if \avg238 .if \avg
239 ld1 {v2.8B}, [x0], x2239 ld1 {v2.8b}, [x0], x2
240 ld1 {v3.8B}, [x0]240 ld1 {v3.8b}, [x0]
241 urhadd v4.8B, v4.8B, v2.8B241 urhadd v4.8b, v4.8b, v2.8b
242 urhadd v5.8B, v5.8B, v3.8B242 urhadd v5.8b, v5.8b, v3.8b
243 sub x0, x0, x2243 sub x0, x0, x2
244 .endif244 .endif
245 st1 {v4.8B}, [x0], x2245 st1 {v4.8b}, [x0], x2
246 st1 {v5.8B}, [x0], x2246 st1 {v5.8b}, [x0], x2
247 b.ne 1b247 b.ne 1b
248 248
249 avg v4.8B, v0.8B, v1.8B249 avg v4.8b, v0.8b, v1.8b
250 ld1 {v0.8B}, [x1], x2250 ld1 {v0.8b}, [x1], x2
251 avg v5.8B, v0.8B, v1.8B251 avg v5.8b, v0.8b, v1.8b
252 .if \avg252 .if \avg
253 ld1 {v2.8B}, [x0], x2253 ld1 {v2.8b}, [x0], x2
254 ld1 {v3.8B}, [x0]254 ld1 {v3.8b}, [x0]
255 urhadd v4.8B, v4.8B, v2.8B255 urhadd v4.8b, v4.8b, v2.8b
256 urhadd v5.8B, v5.8B, v3.8B256 urhadd v5.8b, v5.8b, v3.8b
257 sub x0, x0, x2257 sub x0, x0, x2
258 .endif258 .endif
259 st1 {v4.8B}, [x0], x2259 st1 {v4.8b}, [x0], x2
260 st1 {v5.8B}, [x0], x2260 st1 {v5.8b}, [x0], x2
261 261
262 ret262 ret
263.endm263.endm
264 264
265.macro pixels8_xy2 rnd=1, avg=0265.macro pixels8_xy2 rnd=1, avg=0
266 sub w3, w3, #2266 sub w3, w3, #2
267 ld1 {v0.16B}, [x1], x2267 ld1 {v0.16b}, [x1], x2
268 ld1 {v1.16B}, [x1], x2268 ld1 {v1.16b}, [x1], x2
269NRND movi v19.8H, #1269NRND movi v19.8H, #1
270 ext v4.16B, v0.16B, v4.16B, #1270 ext v4.16b, v0.16b, v4.16b, #1
271 ext v6.16B, v1.16B, v6.16B, #1271 ext v6.16b, v1.16b, v6.16b, #1
272 uaddl v16.8H, v0.8B, v4.8B272 uaddl v16.8h, v0.8b, v4.8b
273 uaddl v17.8H, v1.8B, v6.8B273 uaddl v17.8h, v1.8b, v6.8b
2741: subs w3, w3, #22741: subs w3, w3, #2
275 ld1 {v0.16B}, [x1], x2275 ld1 {v0.16b}, [x1], x2
276 add v18.8H, v16.8H, v17.8H276 add v18.8h, v16.8h, v17.8h
277 ext v4.16B, v0.16B, v4.16B, #1277 ext v4.16b, v0.16b, v4.16b, #1
278NRND add v18.8H, v18.8H, v19.8H278NRND add v18.8H, v18.8H, v19.8H
279 uaddl v16.8H, v0.8B, v4.8B279 uaddl v16.8h, v0.8b, v4.8b
280 mshrn v5.8B, v18.8H, #2280 mshrn v5.8b, v18.8h, #2
281 ld1 {v1.16B}, [x1], x2281 ld1 {v1.16b}, [x1], x2
282 add v18.8H, v16.8H, v17.8H282 add v18.8h, v16.8h, v17.8h
283 .if \avg283 .if \avg
284 ld1 {v7.8B}, [x0]284 ld1 {v7.8b}, [x0]
285 urhadd v5.8B, v5.8B, v7.8B285 urhadd v5.8b, v5.8b, v7.8b
286 .endif286 .endif
287NRND add v18.8H, v18.8H, v19.8H287NRND add v18.8H, v18.8H, v19.8H
288 st1 {v5.8B}, [x0], x2288 st1 {v5.8b}, [x0], x2
289 mshrn v7.8B, v18.8H, #2289 mshrn v7.8b, v18.8h, #2
290 .if \avg290 .if \avg
291 ld1 {v5.8B}, [x0]291 ld1 {v5.8b}, [x0]
292 urhadd v7.8B, v7.8B, v5.8B292 urhadd v7.8b, v7.8b, v5.8b
293 .endif293 .endif
294 ext v6.16B, v1.16B, v6.16B, #1294 ext v6.16b, v1.16b, v6.16b, #1
295 uaddl v17.8H, v1.8B, v6.8B295 uaddl v17.8h, v1.8b, v6.8b
296 st1 {v7.8B}, [x0], x2296 st1 {v7.8b}, [x0], x2
297 b.gt 1b297 b.gt 1b
298 298
299 ld1 {v0.16B}, [x1], x2299 ld1 {v0.16b}, [x1], x2
300 add v18.8H, v16.8H, v17.8H300 add v18.8h, v16.8h, v17.8h
301 ext v4.16B, v0.16B, v4.16B, #1301 ext v4.16b, v0.16b, v4.16b, #1
302NRND add v18.8H, v18.8H, v19.8H302NRND add v18.8H, v18.8H, v19.8H
303 uaddl v16.8H, v0.8B, v4.8B303 uaddl v16.8h, v0.8b, v4.8b
304 mshrn v5.8B, v18.8H, #2304 mshrn v5.8b, v18.8h, #2
305 add v18.8H, v16.8H, v17.8H305 add v18.8h, v16.8h, v17.8h
306 .if \avg306 .if \avg
307 ld1 {v7.8B}, [x0]307 ld1 {v7.8b}, [x0]
308 urhadd v5.8B, v5.8B, v7.8B308 urhadd v5.8b, v5.8b, v7.8b
309 .endif309 .endif
310NRND add v18.8H, v18.8H, v19.8H310NRND add v18.8H, v18.8H, v19.8H
311 st1 {v5.8B}, [x0], x2311 st1 {v5.8b}, [x0], x2
312 mshrn v7.8B, v18.8H, #2312 mshrn v7.8b, v18.8h, #2
313 .if \avg313 .if \avg
314 ld1 {v5.8B}, [x0]314 ld1 {v5.8b}, [x0]
315 urhadd v7.8B, v7.8B, v5.8B315 urhadd v7.8b, v7.8b, v5.8b
316 .endif316 .endif
317 st1 {v7.8B}, [x0], x2317 st1 {v7.8b}, [x0], x2
318 318
319 ret319 ret
320.endm320.endm
libavcodec/aarch64/me_cmp_neon.S+1−1
@@ -1099,7 +1099,7 @@ function vsse_intra16_neon, export=1 · 1098 unmodified lines
1099 cbnz w4, 2b1099 cbnz w4, 2b
1100 1100
11013:11013:
1102 add v16.4s, v16.4s, v17.4S1102 add v16.4s, v16.4s, v17.4s
1103 uaddlv d17, v16.4s1103 uaddlv d17, v16.4s
1104 fmov w0, s171104 fmov w0, s17
1105 1105
libavcodec/aarch64/neon.S+123−123
@@ -28,146 +28,146 @@ · 27 unmodified lines
28.endm28.endm
29 29
30.macro transpose_8x8B r0, r1, r2, r3, r4, r5, r6, r7, r8, r930.macro transpose_8x8B r0, r1, r2, r3, r4, r5, r6, r7, r8, r9
31 trn1 \r8\().8B, \r0\().8B, \r1\().8B31 trn1 \r8\().8b, \r0\().8b, \r1\().8b
32 trn2 \r9\().8B, \r0\().8B, \r1\().8B32 trn2 \r9\().8b, \r0\().8b, \r1\().8b
33 trn1 \r1\().8B, \r2\().8B, \r3\().8B33 trn1 \r1\().8b, \r2\().8b, \r3\().8b
34 trn2 \r3\().8B, \r2\().8B, \r3\().8B34 trn2 \r3\().8b, \r2\().8b, \r3\().8b
35 trn1 \r0\().8B, \r4\().8B, \r5\().8B35 trn1 \r0\().8b, \r4\().8b, \r5\().8b
36 trn2 \r5\().8B, \r4\().8B, \r5\().8B36 trn2 \r5\().8b, \r4\().8b, \r5\().8b
37 trn1 \r2\().8B, \r6\().8B, \r7\().8B37 trn1 \r2\().8b, \r6\().8b, \r7\().8b
38 trn2 \r7\().8B, \r6\().8B, \r7\().8B38 trn2 \r7\().8b, \r6\().8b, \r7\().8b
39 39
40 trn1 \r4\().4H, \r0\().4H, \r2\().4H40 trn1 \r4\().4h, \r0\().4h, \r2\().4h
41 trn2 \r2\().4H, \r0\().4H, \r2\().4H41 trn2 \r2\().4h, \r0\().4h, \r2\().4h
42 trn1 \r6\().4H, \r5\().4H, \r7\().4H42 trn1 \r6\().4h, \r5\().4h, \r7\().4h
43 trn2 \r7\().4H, \r5\().4H, \r7\().4H43 trn2 \r7\().4h, \r5\().4h, \r7\().4h
44 trn1 \r5\().4H, \r9\().4H, \r3\().4H44 trn1 \r5\().4h, \r9\().4h, \r3\().4h
45 trn2 \r9\().4H, \r9\().4H, \r3\().4H45 trn2 \r9\().4h, \r9\().4h, \r3\().4h
46 trn1 \r3\().4H, \r8\().4H, \r1\().4H46 trn1 \r3\().4h, \r8\().4h, \r1\().4h
47 trn2 \r8\().4H, \r8\().4H, \r1\().4H47 trn2 \r8\().4h, \r8\().4h, \r1\().4h
48 48
49 trn1 \r0\().2S, \r3\().2S, \r4\().2S49 trn1 \r0\().2s, \r3\().2s, \r4\().2s
50 trn2 \r4\().2S, \r3\().2S, \r4\().2S50 trn2 \r4\().2s, \r3\().2s, \r4\().2s
51 51
52 trn1 \r1\().2S, \r5\().2S, \r6\().2S52 trn1 \r1\().2s, \r5\().2s, \r6\().2s
53 trn2 \r5\().2S, \r5\().2S, \r6\().2S53 trn2 \r5\().2s, \r5\().2s, \r6\().2s
54 54
55 trn2 \r6\().2S, \r8\().2S, \r2\().2S55 trn2 \r6\().2s, \r8\().2s, \r2\().2s
56 trn1 \r2\().2S, \r8\().2S, \r2\().2S56 trn1 \r2\().2s, \r8\().2s, \r2\().2s
57 57
58 trn1 \r3\().2S, \r9\().2S, \r7\().2S58 trn1 \r3\().2s, \r9\().2s, \r7\().2s
59 trn2 \r7\().2S, \r9\().2S, \r7\().2S59 trn2 \r7\().2s, \r9\().2s, \r7\().2s
60.endm60.endm
61 61
62.macro transpose_8x16B r0, r1, r2, r3, r4, r5, r6, r7, t0, t162.macro transpose_8x16B r0, r1, r2, r3, r4, r5, r6, r7, t0, t1
63 trn1 \t0\().16B, \r0\().16B, \r1\().16B63 trn1 \t0\().16b, \r0\().16b, \r1\().16b
64 trn2 \t1\().16B, \r0\().16B, \r1\().16B64 trn2 \t1\().16b, \r0\().16b, \r1\().16b
65 trn1 \r1\().16B, \r2\().16B, \r3\().16B65 trn1 \r1\().16b, \r2\().16b, \r3\().16b
66 trn2 \r3\().16B, \r2\().16B, \r3\().16B66 trn2 \r3\().16b, \r2\().16b, \r3\().16b
67 trn1 \r0\().16B, \r4\().16B, \r5\().16B67 trn1 \r0\().16b, \r4\().16b, \r5\().16b
68 trn2 \r5\().16B, \r4\().16B, \r5\().16B68 trn2 \r5\().16b, \r4\().16b, \r5\().16b
69 trn1 \r2\().16B, \r6\().16B, \r7\().16B69 trn1 \r2\().16b, \r6\().16b, \r7\().16b
70 trn2 \r7\().16B, \r6\().16B, \r7\().16B70 trn2 \r7\().16b, \r6\().16b, \r7\().16b
71 71
72 trn1 \r4\().8H, \r0\().8H, \r2\().8H72 trn1 \r4\().8h, \r0\().8h, \r2\().8h
73 trn2 \r2\().8H, \r0\().8H, \r2\().8H73 trn2 \r2\().8h, \r0\().8h, \r2\().8h
74 trn1 \r6\().8H, \r5\().8H, \r7\().8H74 trn1 \r6\().8h, \r5\().8h, \r7\().8h
75 trn2 \r7\().8H, \r5\().8H, \r7\().8H75 trn2 \r7\().8h, \r5\().8h, \r7\().8h
76 trn1 \r5\().8H, \t1\().8H, \r3\().8H76 trn1 \r5\().8h, \t1\().8h, \r3\().8h
77 trn2 \t1\().8H, \t1\().8H, \r3\().8H77 trn2 \t1\().8h, \t1\().8h, \r3\().8h
78 trn1 \r3\().8H, \t0\().8H, \r1\().8H78 trn1 \r3\().8h, \t0\().8h, \r1\().8h
79 trn2 \t0\().8H, \t0\().8H, \r1\().8H79 trn2 \t0\().8h, \t0\().8h, \r1\().8h
80 80
81 trn1 \r0\().4S, \r3\().4S, \r4\().4S81 trn1 \r0\().4s, \r3\().4s, \r4\().4s
82 trn2 \r4\().4S, \r3\().4S, \r4\().4S82 trn2 \r4\().4s, \r3\().4s, \r4\().4s
83 83
84 trn1 \r1\().4S, \r5\().4S, \r6\().4S84 trn1 \r1\().4s, \r5\().4s, \r6\().4s
85 trn2 \r5\().4S, \r5\().4S, \r6\().4S85 trn2 \r5\().4s, \r5\().4s, \r6\().4s
86 86
87 trn2 \r6\().4S, \t0\().4S, \r2\().4S87 trn2 \r6\().4s, \t0\().4s, \r2\().4s
88 trn1 \r2\().4S, \t0\().4S, \r2\().4S88 trn1 \r2\().4s, \t0\().4s, \r2\().4s
89 89
90 trn1 \r3\().4S, \t1\().4S, \r7\().4S90 trn1 \r3\().4s, \t1\().4s, \r7\().4s
91 trn2 \r7\().4S, \t1\().4S, \r7\().4S91 trn2 \r7\().4s, \t1\().4s, \r7\().4s
92.endm92.endm
93 93
94.macro transpose_4x16B r0, r1, r2, r3, t4, t5, t6, t794.macro transpose_4x16B r0, r1, r2, r3, t4, t5, t6, t7
95 trn1 \t4\().16B, \r0\().16B, \r1\().16B95 trn1 \t4\().16b, \r0\().16b, \r1\().16b
96 trn2 \t5\().16B, \r0\().16B, \r1\().16B96 trn2 \t5\().16b, \r0\().16b, \r1\().16b
97 trn1 \t6\().16B, \r2\().16B, \r3\().16B97 trn1 \t6\().16b, \r2\().16b, \r3\().16b
98 trn2 \t7\().16B, \r2\().16B, \r3\().16B98 trn2 \t7\().16b, \r2\().16b, \r3\().16b
99 99
100 trn1 \r0\().8H, \t4\().8H, \t6\().8H100 trn1 \r0\().8h, \t4\().8h, \t6\().8h
101 trn2 \r2\().8H, \t4\().8H, \t6\().8H101 trn2 \r2\().8h, \t4\().8h, \t6\().8h
102 trn1 \r1\().8H, \t5\().8H, \t7\().8H102 trn1 \r1\().8h, \t5\().8h, \t7\().8h
103 trn2 \r3\().8H, \t5\().8H, \t7\().8H103 trn2 \r3\().8h, \t5\().8h, \t7\().8h
104.endm104.endm
105 105
106.macro transpose_4x8B r0, r1, r2, r3, t4, t5, t6, t7106.macro transpose_4x8B r0, r1, r2, r3, t4, t5, t6, t7
107 trn1 \t4\().8B, \r0\().8B, \r1\().8B107 trn1 \t4\().8b, \r0\().8b, \r1\().8b
108 trn2 \t5\().8B, \r0\().8B, \r1\().8B108 trn2 \t5\().8b, \r0\().8b, \r1\().8b
109 trn1 \t6\().8B, \r2\().8B, \r3\().8B109 trn1 \t6\().8b, \r2\().8b, \r3\().8b
110 trn2 \t7\().8B, \r2\().8B, \r3\().8B110 trn2 \t7\().8b, \r2\().8b, \r3\().8b
111 111
112 trn1 \r0\().4H, \t4\().4H, \t6\().4H112 trn1 \r0\().4h, \t4\().4h, \t6\().4h
113 trn2 \r2\().4H, \t4\().4H, \t6\().4H113 trn2 \r2\().4h, \t4\().4h, \t6\().4h
114 trn1 \r1\().4H, \t5\().4H, \t7\().4H114 trn1 \r1\().4h, \t5\().4h, \t7\().4h
115 trn2 \r3\().4H, \t5\().4H, \t7\().4H115 trn2 \r3\().4h, \t5\().4h, \t7\().4h
116.endm116.endm
117 117
118.macro transpose_4x4H r0, r1, r2, r3, r4, r5, r6, r7118.macro transpose_4x4H r0, r1, r2, r3, r4, r5, r6, r7
119 trn1 \r4\().4H, \r0\().4H, \r1\().4H119 trn1 \r4\().4h, \r0\().4h, \r1\().4h
120 trn2 \r5\().4H, \r0\().4H, \r1\().4H120 trn2 \r5\().4h, \r0\().4h, \r1\().4h
121 trn1 \r6\().4H, \r2\().4H, \r3\().4H121 trn1 \r6\().4h, \r2\().4h, \r3\().4h
122 trn2 \r7\().4H, \r2\().4H, \r3\().4H122 trn2 \r7\().4h, \r2\().4h, \r3\().4h
123 123
124 trn1 \r0\().2S, \r4\().2S, \r6\().2S124 trn1 \r0\().2s, \r4\().2s, \r6\().2s
125 trn2 \r2\().2S, \r4\().2S, \r6\().2S125 trn2 \r2\().2s, \r4\().2s, \r6\().2s
126 trn1 \r1\().2S, \r5\().2S, \r7\().2S126 trn1 \r1\().2s, \r5\().2s, \r7\().2s
127 trn2 \r3\().2S, \r5\().2S, \r7\().2S127 trn2 \r3\().2s, \r5\().2s, \r7\().2s
128.endm128.endm
129 129
130.macro transpose_4x8H r0, r1, r2, r3, t4, t5, t6, t7130.macro transpose_4x8H r0, r1, r2, r3, t4, t5, t6, t7
131 trn1 \t4\().8H, \r0\().8H, \r1\().8H131 trn1 \t4\().8h, \r0\().8h, \r1\().8h
132 trn2 \t5\().8H, \r0\().8H, \r1\().8H132 trn2 \t5\().8h, \r0\().8h, \r1\().8h
133 trn1 \t6\().8H, \r2\().8H, \r3\().8H133 trn1 \t6\().8h, \r2\().8h, \r3\().8h
134 trn2 \t7\().8H, \r2\().8H, \r3\().8H134 trn2 \t7\().8h, \r2\().8h, \r3\().8h
135 135
136 trn1 \r0\().4S, \t4\().4S, \t6\().4S136 trn1 \r0\().4s, \t4\().4s, \t6\().4s
137 trn2 \r2\().4S, \t4\().4S, \t6\().4S137 trn2 \r2\().4s, \t4\().4s, \t6\().4s
138 trn1 \r1\().4S, \t5\().4S, \t7\().4S138 trn1 \r1\().4s, \t5\().4s, \t7\().4s
139 trn2 \r3\().4S, \t5\().4S, \t7\().4S139 trn2 \r3\().4s, \t5\().4s, \t7\().4s
140.endm140.endm
141 141
142.macro transpose_8x8H r0, r1, r2, r3, r4, r5, r6, r7, r8, r9142.macro transpose_8x8H r0, r1, r2, r3, r4, r5, r6, r7, r8, r9
143 trn1 \r8\().8H, \r0\().8H, \r1\().8H143 trn1 \r8\().8h, \r0\().8h, \r1\().8h
144 trn2 \r9\().8H, \r0\().8H, \r1\().8H144 trn2 \r9\().8h, \r0\().8h, \r1\().8h
145 trn1 \r1\().8H, \r2\().8H, \r3\().8H145 trn1 \r1\().8h, \r2\().8h, \r3\().8h
146 trn2 \r3\().8H, \r2\().8H, \r3\().8H146 trn2 \r3\().8h, \r2\().8h, \r3\().8h
147 trn1 \r0\().8H, \r4\().8H, \r5\().8H147 trn1 \r0\().8h, \r4\().8h, \r5\().8h
148 trn2 \r5\().8H, \r4\().8H, \r5\().8H148 trn2 \r5\().8h, \r4\().8h, \r5\().8h
149 trn1 \r2\().8H, \r6\().8H, \r7\().8H149 trn1 \r2\().8h, \r6\().8h, \r7\().8h
150 trn2 \r7\().8H, \r6\().8H, \r7\().8H150 trn2 \r7\().8h, \r6\().8h, \r7\().8h
151 151
152 trn1 \r4\().4S, \r0\().4S, \r2\().4S152 trn1 \r4\().4s, \r0\().4s, \r2\().4s
153 trn2 \r2\().4S, \r0\().4S, \r2\().4S153 trn2 \r2\().4s, \r0\().4s, \r2\().4s
154 trn1 \r6\().4S, \r5\().4S, \r7\().4S154 trn1 \r6\().4s, \r5\().4s, \r7\().4s
155 trn2 \r7\().4S, \r5\().4S, \r7\().4S155 trn2 \r7\().4s, \r5\().4s, \r7\().4s
156 trn1 \r5\().4S, \r9\().4S, \r3\().4S156 trn1 \r5\().4s, \r9\().4s, \r3\().4s
157 trn2 \r9\().4S, \r9\().4S, \r3\().4S157 trn2 \r9\().4s, \r9\().4s, \r3\().4s
158 trn1 \r3\().4S, \r8\().4S, \r1\().4S158 trn1 \r3\().4s, \r8\().4s, \r1\().4s
159 trn2 \r8\().4S, \r8\().4S, \r1\().4S159 trn2 \r8\().4s, \r8\().4s, \r1\().4s
160 160
161 trn1 \r0\().2D, \r3\().2D, \r4\().2D161 trn1 \r0\().2d, \r3\().2d, \r4\().2d
162 trn2 \r4\().2D, \r3\().2D, \r4\().2D162 trn2 \r4\().2d, \r3\().2d, \r4\().2d
163 163
164 trn1 \r1\().2D, \r5\().2D, \r6\().2D164 trn1 \r1\().2d, \r5\().2d, \r6\().2d
165 trn2 \r5\().2D, \r5\().2D, \r6\().2D165 trn2 \r5\().2d, \r5\().2d, \r6\().2d
166 166
167 trn2 \r6\().2D, \r8\().2D, \r2\().2D167 trn2 \r6\().2d, \r8\().2d, \r2\().2d
168 trn1 \r2\().2D, \r8\().2D, \r2\().2D168 trn1 \r2\().2d, \r8\().2d, \r2\().2d
169 169
170 trn1 \r3\().2D, \r9\().2D, \r7\().2D170 trn1 \r3\().2d, \r9\().2d, \r7\().2d
171 trn2 \r7\().2D, \r9\().2D, \r7\().2D171 trn2 \r7\().2d, \r9\().2d, \r7\().2d
172 172
173.endm173.endm
libavcodec/aarch64/sbrdsp_neon.S+147−147
@@ -46,49 +46,49 @@ function ff_sbr_sum64x5_neon, export=1 · 45 unmodified lines
46 add x3, x0, #192*446 add x3, x0, #192*4
47 add x4, x0, #256*447 add x4, x0, #256*4
48 mov x5, #6448 mov x5, #64
491: ld1 {v0.4S}, [x0]491: ld1 {v0.4s}, [x0]
50 ld1 {v1.4S}, [x1], #1650 ld1 {v1.4s}, [x1], #16
51 fadd v0.4S, v0.4S, v1.4S51 fadd v0.4s, v0.4s, v1.4s
52 ld1 {v2.4S}, [x2], #1652 ld1 {v2.4s}, [x2], #16
53 fadd v0.4S, v0.4S, v2.4S53 fadd v0.4s, v0.4s, v2.4s
54 ld1 {v3.4S}, [x3], #1654 ld1 {v3.4s}, [x3], #16
55 fadd v0.4S, v0.4S, v3.4S55 fadd v0.4s, v0.4s, v3.4s
56 ld1 {v4.4S}, [x4], #1656 ld1 {v4.4s}, [x4], #16
57 fadd v0.4S, v0.4S, v4.4S57 fadd v0.4s, v0.4s, v4.4s
58 st1 {v0.4S}, [x0], #1658 st1 {v0.4s}, [x0], #16
59 subs x5, x5, #459 subs x5, x5, #4
60 b.gt 1b60 b.gt 1b
61 ret61 ret
62endfunc62endfunc
63 63
64function ff_sbr_sum_square_neon, export=164function ff_sbr_sum_square_neon, export=1
65 movi v0.4S, #065 movi v0.4s, #0
661: ld1 {v1.4S}, [x0], #16661: ld1 {v1.4s}, [x0], #16
67 fmla v0.4S, v1.4S, v1.4S67 fmla v0.4s, v1.4s, v1.4s
68 subs w1, w1, #268 subs w1, w1, #2
69 b.gt 1b69 b.gt 1b
70 faddp v0.4S, v0.4S, v0.4S70 faddp v0.4s, v0.4s, v0.4s
71 faddp v0.4S, v0.4S, v0.4S71 faddp v0.4s, v0.4s, v0.4s
72 ret72 ret
73endfunc73endfunc
74 74
75function ff_sbr_neg_odd_64_neon, export=175function ff_sbr_neg_odd_64_neon, export=1
76 mov x1, x076 mov x1, x0
77 movi v5.4S, #1<<7, lsl #2477 movi v5.4s, #1<<7, lsl #24
78 ld2 {v0.4S, v1.4S}, [x0], #3278 ld2 {v0.4s, v1.4s}, [x0], #32
79 eor v1.16B, v1.16B, v5.16B79 eor v1.16b, v1.16b, v5.16b
80 ld2 {v2.4S, v3.4S}, [x0], #3280 ld2 {v2.4s, v3.4s}, [x0], #32
81.rept 381.rept 3
82 st2 {v0.4S, v1.4S}, [x1], #3282 st2 {v0.4s, v1.4s}, [x1], #32
83 eor v3.16B, v3.16B, v5.16B83 eor v3.16b, v3.16b, v5.16b
84 ld2 {v0.4S, v1.4S}, [x0], #3284 ld2 {v0.4s, v1.4s}, [x0], #32
85 st2 {v2.4S, v3.4S}, [x1], #3285 st2 {v2.4s, v3.4s}, [x1], #32
86 eor v1.16B, v1.16B, v5.16B86 eor v1.16b, v1.16b, v5.16b
87 ld2 {v2.4S, v3.4S}, [x0], #3287 ld2 {v2.4s, v3.4s}, [x0], #32
88.endr88.endr
89 eor v3.16B, v3.16B, v5.16B89 eor v3.16b, v3.16b, v5.16b
90 st2 {v0.4S, v1.4S}, [x1], #3290 st2 {v0.4s, v1.4s}, [x1], #32
91 st2 {v2.4S, v3.4S}, [x1], #3291 st2 {v2.4s, v3.4s}, [x1], #32
92 ret92 ret
93endfunc93endfunc
94 94
@@ -97,26 +97,26 @@ function ff_sbr_qmf_pre_shuffle_neon, export=1 · 2 unmodified lines
97 add x2, x0, #64*497 add x2, x0, #64*4
98 mov x3, #-1698 mov x3, #-16
99 mov x4, #-499 mov x4, #-4
100 movi v6.4S, #1<<7, lsl #24100 movi v6.4s, #1<<7, lsl #24
101 ld1 {v0.2S}, [x0], #8101 ld1 {v0.2s}, [x0], #8
102 st1 {v0.2S}, [x2], #8102 st1 {v0.2s}, [x2], #8
103.rept 7103.rept 7
104 ld1 {v1.4S}, [x1], x3104 ld1 {v1.4s}, [x1], x3
105 ld1 {v2.4S}, [x0], #16105 ld1 {v2.4s}, [x0], #16
106 eor v1.16B, v1.16B, v6.16B106 eor v1.16b, v1.16b, v6.16b
107 rev64 v1.4S, v1.4S107 rev64 v1.4s, v1.4s
108 ext v1.16B, v1.16B, v1.16B, #8108 ext v1.16b, v1.16b, v1.16b, #8
109 st2 {v1.4S, v2.4S}, [x2], #32109 st2 {v1.4s, v2.4s}, [x2], #32
110.endr110.endr
111 add x1, x1, #8111 add x1, x1, #8
112 ld1 {v1.2S}, [x1], x4112 ld1 {v1.2s}, [x1], x4
113 ld1 {v2.2S}, [x0], #8113 ld1 {v2.2s}, [x0], #8
114 ld1 {v1.S}[3], [x1]114 ld1 {v1.s}[3], [x1]
115 ld1 {v2.S}[2], [x0]115 ld1 {v2.s}[2], [x0]
116 eor v1.16B, v1.16B, v6.16B116 eor v1.16b, v1.16b, v6.16b
117 rev64 v1.4S, v1.4S117 rev64 v1.4s, v1.4s
118 st2 {v1.2S, v2.2S}, [x2], #16118 st2 {v1.2s, v2.2s}, [x2], #16
119 st2 {v1.S, v2.S}[2], [x2]119 st2 {v1.s, v2.s}[2], [x2]
120 ret120 ret
121endfunc121endfunc
122 122
@@ -124,13 +124,13 @@ function ff_sbr_qmf_post_shuffle_neon, export=1 · 1 unmodified line
124 add x2, x1, #60*4124 add x2, x1, #60*4
125 mov x3, #-16125 mov x3, #-16
126 mov x4, #32126 mov x4, #32
127 movi v6.4S, #1<<7, lsl #24127 movi v6.4s, #1<<7, lsl #24
1281: ld1 {v0.4S}, [x2], x31281: ld1 {v0.4s}, [x2], x3
129 ld1 {v1.4S}, [x1], #16129 ld1 {v1.4s}, [x1], #16
130 eor v0.16B, v0.16B, v6.16B130 eor v0.16b, v0.16b, v6.16b
131 rev64 v0.4S, v0.4S131 rev64 v0.4s, v0.4s
132 ext v0.16B, v0.16B, v0.16B, #8132 ext v0.16b, v0.16b, v0.16b, #8
133 st2 {v0.4S, v1.4S}, [x0], #32133 st2 {v0.4s, v1.4s}, [x0], #32
134 subs x4, x4, #4134 subs x4, x4, #4
135 b.gt 1b135 b.gt 1b
136 ret136 ret
@@ -141,13 +141,13 @@ function ff_sbr_qmf_deint_neg_neon, export=1 · 4 unmodified lines
141 add x2, x0, #60*4141 add x2, x0, #60*4
142 mov x3, #-32142 mov x3, #-32
143 mov x4, #32143 mov x4, #32
144 movi v2.4S, #1<<7, lsl #24144 movi v2.4s, #1<<7, lsl #24
1451: ld2 {v0.4S, v1.4S}, [x1], x31451: ld2 {v0.4s, v1.4s}, [x1], x3
146 eor v0.16B, v0.16B, v2.16B146 eor v0.16b, v0.16b, v2.16b
147 rev64 v1.4S, v1.4S147 rev64 v1.4s, v1.4s
148 ext v1.16B, v1.16B, v1.16B, #8148 ext v1.16b, v1.16b, v1.16b, #8
149 st1 {v0.4S}, [x2]149 st1 {v0.4s}, [x2]
150 st1 {v1.4S}, [x0], #16150 st1 {v1.4s}, [x0], #16
151 sub x2, x2, #16151 sub x2, x2, #16
152 subs x4, x4, #4152 subs x4, x4, #4
153 b.gt 1b153 b.gt 1b
@@ -159,16 +159,16 @@ function ff_sbr_qmf_deint_bfly_neon, export=1 · 5 unmodified lines
159 add x3, x0, #124*4159 add x3, x0, #124*4
160 mov x4, #64160 mov x4, #64
161 mov x5, #-16161 mov x5, #-16
1621: ld1 {v0.4S}, [x1], #161621: ld1 {v0.4s}, [x1], #16
163 ld1 {v1.4S}, [x2], x5163 ld1 {v1.4s}, [x2], x5
164 rev64 v2.4S, v0.4S164 rev64 v2.4s, v0.4s
165 ext v2.16B, v2.16B, v2.16B, #8165 ext v2.16b, v2.16b, v2.16b, #8
166 rev64 v3.4S, v1.4S166 rev64 v3.4s, v1.4s
167 ext v3.16B, v3.16B, v3.16B, #8167 ext v3.16b, v3.16b, v3.16b, #8
168 fadd v1.4S, v1.4S, v2.4S168 fadd v1.4s, v1.4s, v2.4s
169 fsub v0.4S, v0.4S, v3.4S169 fsub v0.4s, v0.4s, v3.4s
170 st1 {v0.4S}, [x0], #16170 st1 {v0.4s}, [x0], #16
171 st1 {v1.4S}, [x3], x5171 st1 {v1.4s}, [x3], x5
172 subs x4, x4, #4172 subs x4, x4, #4
173 b.gt 1b173 b.gt 1b
174 ret174 ret
@@ -178,32 +178,32 @@ function ff_sbr_hf_gen_neon, export=1 · 3 unmodified lines
178 sxtw x4, w4178 sxtw x4, w4
179 sxtw x5, w5179 sxtw x5, w5
180 movrel x6, factors180 movrel x6, factors
181 ld1 {v7.4S}, [x6]181 ld1 {v7.4s}, [x6]
182 dup v1.4S, v0.S[0]182 dup v1.4s, v0.s[0]
183 mov v2.8B, v1.8B183 mov v2.8b, v1.8b
184 mov v2.S[2], v7.S[0]184 mov v2.s[2], v7.s[0]
185 mov v2.S[3], v7.S[0]185 mov v2.s[3], v7.s[0]
186 fmul v1.4S, v1.4S, v2.4S186 fmul v1.4s, v1.4s, v2.4s
187 ld1 {v0.D}[0], [x3]187 ld1 {v0.d}[0], [x3]
188 ld1 {v0.D}[1], [x2]188 ld1 {v0.d}[1], [x2]
189 fmul v0.4S, v0.4S, v1.4S189 fmul v0.4s, v0.4s, v1.4s
190 fmul v1.4S, v0.4S, v7.4S190 fmul v1.4s, v0.4s, v7.4s
191 rev64 v0.4S, v0.4S191 rev64 v0.4s, v0.4s
192 sub x7, x5, x4192 sub x7, x5, x4
193 add x0, x0, x4, lsl #3193 add x0, x0, x4, lsl #3
194 add x1, x1, x4, lsl #3194 add x1, x1, x4, lsl #3
195 sub x1, x1, #16195 sub x1, x1, #16
1961: ld1 {v2.4S}, [x1], #161961: ld1 {v2.4s}, [x1], #16
197 ld1 {v3.2S}, [x1]197 ld1 {v3.2s}, [x1]
198 fmul v4.4S, v2.4S, v1.4S198 fmul v4.4s, v2.4s, v1.4s
199 fmul v5.4S, v2.4S, v0.4S199 fmul v5.4s, v2.4s, v0.4s
200 faddp v4.4S, v4.4S, v4.4S200 faddp v4.4s, v4.4s, v4.4s
201 faddp v5.4S, v5.4S, v5.4S201 faddp v5.4s, v5.4s, v5.4s
202 faddp v4.4S, v4.4S, v4.4S202 faddp v4.4s, v4.4s, v4.4s
203 faddp v5.4S, v5.4S, v5.4S203 faddp v5.4s, v5.4s, v5.4s
204 mov v4.S[1], v5.S[0]204 mov v4.s[1], v5.s[0]
205 fadd v4.2S, v4.2S, v3.2S205 fadd v4.2s, v4.2s, v3.2s
206 st1 {v4.2S}, [x0], #8206 st1 {v4.2s}, [x0], #8
207 sub x1, x1, #8207 sub x1, x1, #8
208 subs x7, x7, #1208 subs x7, x7, #1
209 b.gt 1b209 b.gt 1b
@@ -215,10 +215,10 @@ function ff_sbr_hf_g_filt_neon, export=1 · 5 unmodified lines
215 sxtw x4, w4215 sxtw x4, w4
216 mov x5, #40*2*4216 mov x5, #40*2*4
217 add x1, x1, x4, lsl #3217 add x1, x1, x4, lsl #3
2181: ld1 {v0.2S}, [x1], x52181: ld1 {v0.2s}, [x1], x5
219 ld1 {v1.S}[0], [x2], #4219 ld1 {v1.s}[0], [x2], #4
220 fmul v2.4S, v0.4S, v1.S[0]220 fmul v2.4s, v0.4s, v1.s[0]
221 st1 {v2.2S}, [x0], #8221 st1 {v2.2s}, [x0], #8
222 subs x3, x3, #1222 subs x3, x3, #1
223 b.gt 1b223 b.gt 1b
224 ret224 ret
@@ -227,46 +227,46 @@ endfunc · 2 unmodified lines
227function ff_sbr_autocorrelate_neon, export=1227function ff_sbr_autocorrelate_neon, export=1
228 mov x2, #38228 mov x2, #38
229 movrel x3, factors229 movrel x3, factors
230 ld1 {v0.4S}, [x3]230 ld1 {v0.4s}, [x3]
231 movi v1.4S, #0231 movi v1.4s, #0
232 movi v2.4S, #0232 movi v2.4s, #0
233 movi v3.4S, #0233 movi v3.4s, #0
234 ld1 {v4.2S}, [x0], #8234 ld1 {v4.2s}, [x0], #8
235 ld1 {v5.2S}, [x0], #8235 ld1 {v5.2s}, [x0], #8
236 fmul v16.2S, v4.2S, v4.2S236 fmul v16.2s, v4.2s, v4.2s
237 fmul v17.2S, v5.2S, v4.S[0]237 fmul v17.2s, v5.2s, v4.s[0]
238 fmul v18.2S, v5.2S, v4.S[1]238 fmul v18.2s, v5.2s, v4.s[1]
2391: ld1 {v5.D}[1], [x0], #82391: ld1 {v5.d}[1], [x0], #8
240 fmla v1.2S, v4.2S, v4.2S240 fmla v1.2s, v4.2s, v4.2s
241 fmla v2.4S, v5.4S, v4.S[0]241 fmla v2.4s, v5.4s, v4.s[0]
242 fmla v3.4S, v5.4S, v4.S[1]242 fmla v3.4s, v5.4s, v4.s[1]
243 mov v4.D[0], v5.D[0]243 mov v4.d[0], v5.d[0]
244 mov v5.D[0], v5.D[1]244 mov v5.d[0], v5.d[1]
245 subs x2, x2, #1245 subs x2, x2, #1
246 b.gt 1b246 b.gt 1b
247 fmul v19.2S, v4.2S, v4.2S247 fmul v19.2s, v4.2s, v4.2s
248 fmul v20.2S, v5.2S, v4.S[0]248 fmul v20.2s, v5.2s, v4.s[0]
249 fmul v21.2S, v5.2S, v4.S[1]249 fmul v21.2s, v5.2s, v4.s[1]
250 fadd v22.4S, v2.4S, v20.4S250 fadd v22.4s, v2.4s, v20.4s
251 fsub v22.4S, v22.4S, v17.4S251 fsub v22.4s, v22.4s, v17.4s
252 fadd v23.4S, v3.4S, v21.4S252 fadd v23.4s, v3.4s, v21.4s
253 fsub v23.4S, v23.4S, v18.4S253 fsub v23.4s, v23.4s, v18.4s
254 rev64 v23.4S, v23.4S254 rev64 v23.4s, v23.4s
255 fmul v23.4S, v23.4S, v0.4S255 fmul v23.4s, v23.4s, v0.4s
256 fadd v22.4S, v22.4S, v23.4S256 fadd v22.4s, v22.4s, v23.4s
257 st1 {v22.4S}, [x1], #16257 st1 {v22.4s}, [x1], #16
258 fadd v23.2S, v1.2S, v19.2S258 fadd v23.2s, v1.2s, v19.2s
259 fsub v23.2S, v23.2S, v16.2S259 fsub v23.2s, v23.2s, v16.2s
260 faddp v23.2S, v23.2S, v23.2S260 faddp v23.2s, v23.2s, v23.2s
261 st1 {v23.S}[0], [x1]261 st1 {v23.s}[0], [x1]
262 add x1, x1, #8262 add x1, x1, #8
263 rev64 v3.2S, v3.2S263 rev64 v3.2s, v3.2s
264 fmul v3.2S, v3.2S, v0.2S264 fmul v3.2s, v3.2s, v0.2s
265 fadd v2.2S, v2.2S, v3.2S265 fadd v2.2s, v2.2s, v3.2s
266 st1 {v2.2S}, [x1]266 st1 {v2.2s}, [x1]
267 add x1, x1, #16267 add x1, x1, #16
268 faddp v1.2S, v1.2S, v1.2S268 faddp v1.2s, v1.2s, v1.2s
269 st1 {v1.S}[0], [x1]269 st1 {v1.s}[0], [x1]
270 ret270 ret
271endfunc271endfunc
272 272
@@ -278,25 +278,25 @@ endfunc · 5 unmodified lines
2781: and x3, x3, #0x1ff2781: and x3, x3, #0x1ff
279 add x8, x7, x3, lsl #3279 add x8, x7, x3, lsl #3
280 add x3, x3, #2280 add x3, x3, #2
281 ld1 {v2.4S}, [x0]281 ld1 {v2.4s}, [x0]
282 ld1 {v3.2S}, [x1], #8282 ld1 {v3.2s}, [x1], #8
283 ld1 {v4.2S}, [x2], #8283 ld1 {v4.2s}, [x2], #8
284 ld1 {v5.4S}, [x8]284 ld1 {v5.4s}, [x8]
285 mov v6.16B, v2.16B285 mov v6.16b, v2.16b
286 zip1 v3.4S, v3.4S, v3.4S286 zip1 v3.4s, v3.4s, v3.4s
287 zip1 v4.4S, v4.4S, v4.4S287 zip1 v4.4s, v4.4s, v4.4s
288 fmla v6.4S, v1.4S, v3.4S288 fmla v6.4s, v1.4s, v3.4s
289 fmla v2.4S, v5.4S, v4.4S289 fmla v2.4s, v5.4s, v4.4s
290 fcmeq v7.4S, v3.4S, #0290 fcmeq v7.4s, v3.4s, #0
291 bif v2.16B, v6.16B, v7.16B291 bif v2.16b, v6.16b, v7.16b
292 st1 {v2.4S}, [x0], #16292 st1 {v2.4s}, [x0], #16
293 subs x5, x5, #2293 subs x5, x5, #2
294 b.gt 1b294 b.gt 1b
295.endm295.endm
296 296
297function ff_sbr_hf_apply_noise_0_neon, export=1297function ff_sbr_hf_apply_noise_0_neon, export=1
298 movrel x9, phi_noise_0298 movrel x9, phi_noise_0
299 ld1 {v1.4S}, [x9]299 ld1 {v1.4s}, [x9]
300 apply_noise_common300 apply_noise_common
301 ret301 ret
302endfunc302endfunc
@@ -305,14 +305,14 @@ function ff_sbr_hf_apply_noise_1_neon, export=1 · 2 unmodified lines
305 movrel x9, phi_noise_1305 movrel x9, phi_noise_1
306 and x4, x4, #1306 and x4, x4, #1
307 add x9, x9, x4, lsl #4307 add x9, x9, x4, lsl #4
308 ld1 {v1.4S}, [x9]308 ld1 {v1.4s}, [x9]
309 apply_noise_common309 apply_noise_common
310 ret310 ret
311endfunc311endfunc
312 312
313function ff_sbr_hf_apply_noise_2_neon, export=1313function ff_sbr_hf_apply_noise_2_neon, export=1
314 movrel x9, phi_noise_2314 movrel x9, phi_noise_2
315 ld1 {v1.4S}, [x9]315 ld1 {v1.4s}, [x9]
316 apply_noise_common316 apply_noise_common
317 ret317 ret
318endfunc318endfunc
@@ -321,7 +321,7 @@ function ff_sbr_hf_apply_noise_3_neon, export=1 · 2 unmodified lines
321 movrel x9, phi_noise_3321 movrel x9, phi_noise_3
322 and x4, x4, #1322 and x4, x4, #1
323 add x9, x9, x4, lsl #4323 add x9, x9, x4, lsl #4
324 ld1 {v1.4S}, [x9]324 ld1 {v1.4s}, [x9]
325 apply_noise_common325 apply_noise_common
326 ret326 ret
327endfunc327endfunc
libavcodec/aarch64/simple_idct_neon.S+193−193
@@ -54,7 +54,7 @@ endconst · 53 unmodified lines
54 prfm pldl1keep, [\data]54 prfm pldl1keep, [\data]
55 mov x10, x3055 mov x10, x30
56 movrel x3, idct_coeff_neon56 movrel x3, idct_coeff_neon
57 ld1 {v0.2D}, [x3]57 ld1 {v0.2d}, [x3]
58.endm58.endm
59 59
60.macro idct_end60.macro idct_end
@@ -74,146 +74,146 @@ endconst · 13 unmodified lines
74.endm74.endm
75 75
76.macro idct_col4_top y1, y2, y3, y4, i, l76.macro idct_col4_top y1, y2, y3, y4, i, l
77 smull\i v7.4S, \y3\l, z277 smull\i v7.4s, \y3\l, z2
78 smull\i v16.4S, \y3\l, z678 smull\i v16.4s, \y3\l, z6
79 smull\i v17.4S, \y2\l, z179 smull\i v17.4s, \y2\l, z1
80 add v19.4S, v23.4S, v7.4S80 add v19.4s, v23.4s, v7.4s
81 smull\i v18.4S, \y2\l, z381 smull\i v18.4s, \y2\l, z3
82 add v20.4S, v23.4S, v16.4S82 add v20.4s, v23.4s, v16.4s
83 smull\i v5.4S, \y2\l, z583 smull\i v5.4s, \y2\l, z5
84 sub v21.4S, v23.4S, v16.4S84 sub v21.4s, v23.4s, v16.4s
85 smull\i v6.4S, \y2\l, z785 smull\i v6.4s, \y2\l, z7
86 sub v22.4S, v23.4S, v7.4S86 sub v22.4s, v23.4s, v7.4s
87 87
88 smlal\i v17.4S, \y4\l, z388 smlal\i v17.4s, \y4\l, z3
89 smlsl\i v18.4S, \y4\l, z789 smlsl\i v18.4s, \y4\l, z7
90 smlsl\i v5.4S, \y4\l, z190 smlsl\i v5.4s, \y4\l, z1
91 smlsl\i v6.4S, \y4\l, z591 smlsl\i v6.4s, \y4\l, z5
92.endm92.endm
93 93
94.macro idct_row4_neon y1, y2, y3, y4, pass94.macro idct_row4_neon y1, y2, y3, y4, pass
95 ld1 {\y1\().2D,\y2\().2D}, [x2], #3295 ld1 {\y1\().2d,\y2\().2d}, [x2], #32
96 movi v23.4S, #1<<2, lsl #896 movi v23.4s, #1<<2, lsl #8
97 orr v5.16B, \y1\().16B, \y2\().16B97 orr v5.16b, \y1\().16b, \y2\().16b
98 ld1 {\y3\().2D,\y4\().2D}, [x2], #3298 ld1 {\y3\().2d,\y4\().2d}, [x2], #32
99 orr v6.16B, \y3\().16B, \y4\().16B99 orr v6.16b, \y3\().16b, \y4\().16b
100 orr v5.16B, v5.16B, v6.16B100 orr v5.16b, v5.16b, v6.16b
101 mov x3, v5.D[1]101 mov x3, v5.d[1]
102 smlal v23.4S, \y1\().4H, z4102 smlal v23.4s, \y1\().4h, z4
103 103
104 idct_col4_top \y1, \y2, \y3, \y4, 1, .4H104 idct_col4_top \y1, \y2, \y3, \y4, 1, .4h
105 105
106 cmp x3, #0106 cmp x3, #0
107 b.eq \pass\()f107 b.eq \pass\()f
108 108
109 smull2 v7.4S, \y1\().8H, z4109 smull2 v7.4s, \y1\().8h, z4
110 smlal2 v17.4S, \y2\().8H, z5110 smlal2 v17.4s, \y2\().8h, z5
111 smlsl2 v18.4S, \y2\().8H, z1111 smlsl2 v18.4s, \y2\().8h, z1
112 smull2 v16.4S, \y3\().8H, z2112 smull2 v16.4s, \y3\().8h, z2
113 smlal2 v5.4S, \y2\().8H, z7113 smlal2 v5.4s, \y2\().8h, z7
114 add v19.4S, v19.4S, v7.4S114 add v19.4s, v19.4s, v7.4s
115 sub v20.4S, v20.4S, v7.4S115 sub v20.4s, v20.4s, v7.4s
116 sub v21.4S, v21.4S, v7.4S116 sub v21.4s, v21.4s, v7.4s
117 add v22.4S, v22.4S, v7.4S117 add v22.4s, v22.4s, v7.4s
118 smlal2 v6.4S, \y2\().8H, z3118 smlal2 v6.4s, \y2\().8h, z3
119 smull2 v7.4S, \y3\().8H, z6119 smull2 v7.4s, \y3\().8h, z6
120 smlal2 v17.4S, \y4\().8H, z7120 smlal2 v17.4s, \y4\().8h, z7
121 smlsl2 v18.4S, \y4\().8H, z5121 smlsl2 v18.4s, \y4\().8h, z5
122 smlal2 v5.4S, \y4\().8H, z3122 smlal2 v5.4s, \y4\().8h, z3
123 smlsl2 v6.4S, \y4\().8H, z1123 smlsl2 v6.4s, \y4\().8h, z1
124 add v19.4S, v19.4S, v7.4S124 add v19.4s, v19.4s, v7.4s
125 sub v20.4S, v20.4S, v16.4S125 sub v20.4s, v20.4s, v16.4s
126 add v21.4S, v21.4S, v16.4S126 add v21.4s, v21.4s, v16.4s
127 sub v22.4S, v22.4S, v7.4S127 sub v22.4s, v22.4s, v7.4s
128 128
129\pass: add \y3\().4S, v19.4S, v17.4S129\pass: add \y3\().4S, v19.4S, v17.4S
130 add \y4\().4S, v20.4S, v18.4S130 add \y4\().4s, v20.4s, v18.4s
131 shrn \y1\().4H, \y3\().4S, #ROW_SHIFT131 shrn \y1\().4h, \y3\().4s, #ROW_SHIFT
132 shrn \y2\().4H, \y4\().4S, #ROW_SHIFT132 shrn \y2\().4h, \y4\().4s, #ROW_SHIFT
133 add v7.4S, v21.4S, v5.4S133 add v7.4s, v21.4s, v5.4s
134 add v16.4S, v22.4S, v6.4S134 add v16.4s, v22.4s, v6.4s
135 shrn \y3\().4H, v7.4S, #ROW_SHIFT135 shrn \y3\().4h, v7.4s, #ROW_SHIFT
136 shrn \y4\().4H, v16.4S, #ROW_SHIFT136 shrn \y4\().4h, v16.4s, #ROW_SHIFT
137 sub v22.4S, v22.4S, v6.4S137 sub v22.4s, v22.4s, v6.4s
138 sub v19.4S, v19.4S, v17.4S138 sub v19.4s, v19.4s, v17.4s
139 sub v21.4S, v21.4S, v5.4S139 sub v21.4s, v21.4s, v5.4s
140 shrn2 \y1\().8H, v22.4S, #ROW_SHIFT140 shrn2 \y1\().8h, v22.4s, #ROW_SHIFT
141 sub v20.4S, v20.4S, v18.4S141 sub v20.4s, v20.4s, v18.4s
142 shrn2 \y2\().8H, v21.4S, #ROW_SHIFT142 shrn2 \y2\().8h, v21.4s, #ROW_SHIFT
143 shrn2 \y3\().8H, v20.4S, #ROW_SHIFT143 shrn2 \y3\().8h, v20.4s, #ROW_SHIFT
144 shrn2 \y4\().8H, v19.4S, #ROW_SHIFT144 shrn2 \y4\().8h, v19.4s, #ROW_SHIFT
145 145
146 trn1 v16.8H, \y1\().8H, \y2\().8H146 trn1 v16.8h, \y1\().8h, \y2\().8h
147 trn2 v17.8H, \y1\().8H, \y2\().8H147 trn2 v17.8h, \y1\().8h, \y2\().8h
148 trn1 v18.8H, \y3\().8H, \y4\().8H148 trn1 v18.8h, \y3\().8h, \y4\().8h
149 trn2 v19.8H, \y3\().8H, \y4\().8H149 trn2 v19.8h, \y3\().8h, \y4\().8h
150 trn1 \y1\().4S, v16.4S, v18.4S150 trn1 \y1\().4s, v16.4s, v18.4s
151 trn1 \y2\().4S, v17.4S, v19.4S151 trn1 \y2\().4s, v17.4s, v19.4s
152 trn2 \y3\().4S, v16.4S, v18.4S152 trn2 \y3\().4s, v16.4s, v18.4s
153 trn2 \y4\().4S, v17.4S, v19.4S153 trn2 \y4\().4s, v17.4s, v19.4s
154.endm154.endm
155 155
156.macro declare_idct_col4_neon i, l156.macro declare_idct_col4_neon i, l
157function idct_col4_neon\i157function idct_col4_neon\i
158 dup v23.4H, z4c158 dup v23.4h, z4c
159.if \i == 1159.if \i == 1
160 add v23.4H, v23.4H, v24.4H160 add v23.4h, v23.4h, v24.4h
161.else161.else
162 mov v5.D[0], v24.D[1]162 mov v5.d[0], v24.d[1]
163 add v23.4H, v23.4H, v5.4H163 add v23.4h, v23.4h, v5.4h
164.endif164.endif
165 smull v23.4S, v23.4H, z4165 smull v23.4s, v23.4h, z4
166 166
167 idct_col4_top v24, v25, v26, v27, \i, \l167 idct_col4_top v24, v25, v26, v27, \i, \l
168 168
169 mov x4, v28.D[\i - 1]169 mov x4, v28.d[\i - 1]
170 mov x5, v29.D[\i - 1]170 mov x5, v29.d[\i - 1]
171 cmp x4, #0171 cmp x4, #0
172 b.eq 1f172 b.eq 1f
173 173
174 smull\i v7.4S, v28\l, z4174 smull\i v7.4s, v28\l, z4
175 add v19.4S, v19.4S, v7.4S175 add v19.4s, v19.4s, v7.4s
176 sub v20.4S, v20.4S, v7.4S176 sub v20.4s, v20.4s, v7.4s
177 sub v21.4S, v21.4S, v7.4S177 sub v21.4s, v21.4s, v7.4s
178 add v22.4S, v22.4S, v7.4S178 add v22.4s, v22.4s, v7.4s
179 179
1801: mov x4, v30.D[\i - 1]1801: mov x4, v30.d[\i - 1]
181 cmp x5, #0181 cmp x5, #0
182 b.eq 2f182 b.eq 2f
183 183
184 smlal\i v17.4S, v29\l, z5184 smlal\i v17.4s, v29\l, z5
185 smlsl\i v18.4S, v29\l, z1185 smlsl\i v18.4s, v29\l, z1
186 smlal\i v5.4S, v29\l, z7186 smlal\i v5.4s, v29\l, z7
187 smlal\i v6.4S, v29\l, z3187 smlal\i v6.4s, v29\l, z3
188 188
1892: mov x5, v31.D[\i - 1]1892: mov x5, v31.d[\i - 1]
190 cmp x4, #0190 cmp x4, #0
191 b.eq 3f191 b.eq 3f
192 192
193 smull\i v7.4S, v30\l, z6193 smull\i v7.4s, v30\l, z6
194 smull\i v16.4S, v30\l, z2194 smull\i v16.4s, v30\l, z2
195 add v19.4S, v19.4S, v7.4S195 add v19.4s, v19.4s, v7.4s
196 sub v22.4S, v22.4S, v7.4S196 sub v22.4s, v22.4s, v7.4s
197 sub v20.4S, v20.4S, v16.4S197 sub v20.4s, v20.4s, v16.4s
198 add v21.4S, v21.4S, v16.4S198 add v21.4s, v21.4s, v16.4s
199 199
2003: cmp x5, #02003: cmp x5, #0
201 b.eq 4f201 b.eq 4f
202 202
203 smlal\i v17.4S, v31\l, z7203 smlal\i v17.4s, v31\l, z7
204 smlsl\i v18.4S, v31\l, z5204 smlsl\i v18.4s, v31\l, z5
205 smlal\i v5.4S, v31\l, z3205 smlal\i v5.4s, v31\l, z3
206 smlsl\i v6.4S, v31\l, z1206 smlsl\i v6.4s, v31\l, z1
207 207
2084: addhn v7.4H, v19.4S, v17.4S2084: addhn v7.4h, v19.4s, v17.4s
209 addhn2 v7.8H, v20.4S, v18.4S209 addhn2 v7.8h, v20.4s, v18.4s
210 subhn v18.4H, v20.4S, v18.4S210 subhn v18.4h, v20.4s, v18.4s
211 subhn2 v18.8H, v19.4S, v17.4S211 subhn2 v18.8h, v19.4s, v17.4s
212 212
213 addhn v16.4H, v21.4S, v5.4S213 addhn v16.4h, v21.4s, v5.4s
214 addhn2 v16.8H, v22.4S, v6.4S214 addhn2 v16.8h, v22.4s, v6.4s
215 subhn v17.4H, v22.4S, v6.4S215 subhn v17.4h, v22.4s, v6.4s
216 subhn2 v17.8H, v21.4S, v5.4S216 subhn2 v17.8h, v21.4s, v5.4s
217 217
218 ret218 ret
219endfunc219endfunc
@@ -229,33 +229,33 @@ function ff_simple_idct_put_neon, export=1 · 9 unmodified lines
229 idct_row4_neon v28, v29, v30, v31, 2229 idct_row4_neon v28, v29, v30, v31, 2
230 bl idct_col4_neon1230 bl idct_col4_neon1
231 231
232 sqshrun v1.8B, v7.8H, #COL_SHIFT-16232 sqshrun v1.8b, v7.8h, #COL_SHIFT-16
233 sqshrun2 v1.16B, v16.8H, #COL_SHIFT-16233 sqshrun2 v1.16b, v16.8h, #COL_SHIFT-16
234 sqshrun v3.8B, v17.8H, #COL_SHIFT-16234 sqshrun v3.8b, v17.8h, #COL_SHIFT-16
235 sqshrun2 v3.16B, v18.8H, #COL_SHIFT-16235 sqshrun2 v3.16b, v18.8h, #COL_SHIFT-16
236 236
237 bl idct_col4_neon2237 bl idct_col4_neon2
238 238
239 sqshrun v2.8B, v7.8H, #COL_SHIFT-16239 sqshrun v2.8b, v7.8h, #COL_SHIFT-16
240 sqshrun2 v2.16B, v16.8H, #COL_SHIFT-16240 sqshrun2 v2.16b, v16.8h, #COL_SHIFT-16
241 sqshrun v4.8B, v17.8H, #COL_SHIFT-16241 sqshrun v4.8b, v17.8h, #COL_SHIFT-16
242 sqshrun2 v4.16B, v18.8H, #COL_SHIFT-16242 sqshrun2 v4.16b, v18.8h, #COL_SHIFT-16
243 243
244 zip1 v16.4S, v1.4S, v2.4S244 zip1 v16.4s, v1.4s, v2.4s
245 zip2 v17.4S, v1.4S, v2.4S245 zip2 v17.4s, v1.4s, v2.4s
246 246
247 st1 {v16.D}[0], [x0], x1247 st1 {v16.d}[0], [x0], x1
248 st1 {v16.D}[1], [x0], x1248 st1 {v16.d}[1], [x0], x1
249 249
250 zip1 v18.4S, v3.4S, v4.4S250 zip1 v18.4s, v3.4s, v4.4s
251 zip2 v19.4S, v3.4S, v4.4S251 zip2 v19.4s, v3.4s, v4.4s
252 252
253 st1 {v17.D}[0], [x0], x1253 st1 {v17.d}[0], [x0], x1
254 st1 {v17.D}[1], [x0], x1254 st1 {v17.d}[1], [x0], x1
255 st1 {v18.D}[0], [x0], x1255 st1 {v18.d}[0], [x0], x1
256 st1 {v18.D}[1], [x0], x1256 st1 {v18.d}[1], [x0], x1
257 st1 {v19.D}[0], [x0], x1257 st1 {v19.d}[0], [x0], x1
258 st1 {v19.D}[1], [x0], x1258 st1 {v19.d}[1], [x0], x1
259 259
260 idct_end260 idct_end
261endfunc261endfunc
@@ -267,59 +267,59 @@ function ff_simple_idct_add_neon, export=1 · 5 unmodified lines
267 idct_row4_neon v28, v29, v30, v31, 2267 idct_row4_neon v28, v29, v30, v31, 2
268 bl idct_col4_neon1268 bl idct_col4_neon1
269 269
270 sshr v1.8H, v7.8H, #COL_SHIFT-16270 sshr v1.8h, v7.8h, #COL_SHIFT-16
271 sshr v2.8H, v16.8H, #COL_SHIFT-16271 sshr v2.8h, v16.8h, #COL_SHIFT-16
272 sshr v3.8H, v17.8H, #COL_SHIFT-16272 sshr v3.8h, v17.8h, #COL_SHIFT-16
273 sshr v4.8H, v18.8H, #COL_SHIFT-16273 sshr v4.8h, v18.8h, #COL_SHIFT-16
274 274
275 bl idct_col4_neon2275 bl idct_col4_neon2
276 276
277 sshr v7.8H, v7.8H, #COL_SHIFT-16277 sshr v7.8h, v7.8h, #COL_SHIFT-16
278 sshr v16.8H, v16.8H, #COL_SHIFT-16278 sshr v16.8h, v16.8h, #COL_SHIFT-16
279 sshr v17.8H, v17.8H, #COL_SHIFT-16279 sshr v17.8h, v17.8h, #COL_SHIFT-16
280 sshr v18.8H, v18.8H, #COL_SHIFT-16280 sshr v18.8h, v18.8h, #COL_SHIFT-16
281 281
282 mov x9, x0282 mov x9, x0
283 ld1 {v19.D}[0], [x0], x1283 ld1 {v19.d}[0], [x0], x1
284 zip1 v23.2D, v1.2D, v7.2D284 zip1 v23.2d, v1.2d, v7.2d
285 zip2 v24.2D, v1.2D, v7.2D285 zip2 v24.2d, v1.2d, v7.2d
286 ld1 {v19.D}[1], [x0], x1286 ld1 {v19.d}[1], [x0], x1
287 zip1 v25.2D, v2.2D, v16.2D287 zip1 v25.2d, v2.2d, v16.2d
288 zip2 v26.2D, v2.2D, v16.2D288 zip2 v26.2d, v2.2d, v16.2d
289 ld1 {v20.D}[0], [x0], x1289 ld1 {v20.d}[0], [x0], x1
290 zip1 v27.2D, v3.2D, v17.2D290 zip1 v27.2d, v3.2d, v17.2d
291 zip2 v28.2D, v3.2D, v17.2D291 zip2 v28.2d, v3.2d, v17.2d
292 ld1 {v20.D}[1], [x0], x1292 ld1 {v20.d}[1], [x0], x1
293 zip1 v29.2D, v4.2D, v18.2D293 zip1 v29.2d, v4.2d, v18.2d
294 zip2 v30.2D, v4.2D, v18.2D294 zip2 v30.2d, v4.2d, v18.2d
295 ld1 {v21.D}[0], [x0], x1295 ld1 {v21.d}[0], [x0], x1
296 uaddw v23.8H, v23.8H, v19.8B296 uaddw v23.8h, v23.8h, v19.8b
297 uaddw2 v24.8H, v24.8H, v19.16B297 uaddw2 v24.8h, v24.8h, v19.16b
298 ld1 {v21.D}[1], [x0], x1298 ld1 {v21.d}[1], [x0], x1
299 sqxtun v23.8B, v23.8H299 sqxtun v23.8b, v23.8h
300 sqxtun2 v23.16B, v24.8H300 sqxtun2 v23.16b, v24.8h
301 ld1 {v22.D}[0], [x0], x1301 ld1 {v22.d}[0], [x0], x1
302 uaddw v24.8H, v25.8H, v20.8B302 uaddw v24.8h, v25.8h, v20.8b
303 uaddw2 v25.8H, v26.8H, v20.16B303 uaddw2 v25.8h, v26.8h, v20.16b
304 ld1 {v22.D}[1], [x0], x1304 ld1 {v22.d}[1], [x0], x1
305 sqxtun v24.8B, v24.8H305 sqxtun v24.8b, v24.8h
306 sqxtun2 v24.16B, v25.8H306 sqxtun2 v24.16b, v25.8h
307 st1 {v23.D}[0], [x9], x1307 st1 {v23.d}[0], [x9], x1
308 uaddw v25.8H, v27.8H, v21.8B308 uaddw v25.8h, v27.8h, v21.8b
309 uaddw2 v26.8H, v28.8H, v21.16B309 uaddw2 v26.8h, v28.8h, v21.16b
310 st1 {v23.D}[1], [x9], x1310 st1 {v23.d}[1], [x9], x1
311 sqxtun v25.8B, v25.8H311 sqxtun v25.8b, v25.8h
312 sqxtun2 v25.16B, v26.8H312 sqxtun2 v25.16b, v26.8h
313 st1 {v24.D}[0], [x9], x1313 st1 {v24.d}[0], [x9], x1
314 uaddw v26.8H, v29.8H, v22.8B314 uaddw v26.8h, v29.8h, v22.8b
315 uaddw2 v27.8H, v30.8H, v22.16B315 uaddw2 v27.8h, v30.8h, v22.16b
316 st1 {v24.D}[1], [x9], x1316 st1 {v24.d}[1], [x9], x1
317 sqxtun v26.8B, v26.8H317 sqxtun v26.8b, v26.8h
318 sqxtun2 v26.16B, v27.8H318 sqxtun2 v26.16b, v27.8h
319 st1 {v25.D}[0], [x9], x1319 st1 {v25.d}[0], [x9], x1
320 st1 {v25.D}[1], [x9], x1320 st1 {v25.d}[1], [x9], x1
321 st1 {v26.D}[0], [x9], x1321 st1 {v26.d}[0], [x9], x1
322 st1 {v26.D}[1], [x9], x1322 st1 {v26.d}[1], [x9], x1
323 323
324 idct_end324 idct_end
325endfunc325endfunc
@@ -333,30 +333,30 @@ function ff_simple_idct_neon, export=1 · 7 unmodified lines
333 sub x2, x2, #128333 sub x2, x2, #128
334 bl idct_col4_neon1334 bl idct_col4_neon1
335 335
336 sshr v1.8H, v7.8H, #COL_SHIFT-16336 sshr v1.8h, v7.8h, #COL_SHIFT-16
337 sshr v2.8H, v16.8H, #COL_SHIFT-16337 sshr v2.8h, v16.8h, #COL_SHIFT-16
338 sshr v3.8H, v17.8H, #COL_SHIFT-16338 sshr v3.8h, v17.8h, #COL_SHIFT-16
339 sshr v4.8H, v18.8H, #COL_SHIFT-16339 sshr v4.8h, v18.8h, #COL_SHIFT-16
340 340
341 bl idct_col4_neon2341 bl idct_col4_neon2
342 342
343 sshr v7.8H, v7.8H, #COL_SHIFT-16343 sshr v7.8h, v7.8h, #COL_SHIFT-16
344 sshr v16.8H, v16.8H, #COL_SHIFT-16344 sshr v16.8h, v16.8h, #COL_SHIFT-16
345 sshr v17.8H, v17.8H, #COL_SHIFT-16345 sshr v17.8h, v17.8h, #COL_SHIFT-16
346 sshr v18.8H, v18.8H, #COL_SHIFT-16346 sshr v18.8h, v18.8h, #COL_SHIFT-16
347 347
348 zip1 v23.2D, v1.2D, v7.2D348 zip1 v23.2d, v1.2d, v7.2d
349 zip2 v24.2D, v1.2D, v7.2D349 zip2 v24.2d, v1.2d, v7.2d
350 st1 {v23.2D,v24.2D}, [x2], #32350 st1 {v23.2d,v24.2d}, [x2], #32
351 zip1 v25.2D, v2.2D, v16.2D351 zip1 v25.2d, v2.2d, v16.2d
352 zip2 v26.2D, v2.2D, v16.2D352 zip2 v26.2d, v2.2d, v16.2d
353 st1 {v25.2D,v26.2D}, [x2], #32353 st1 {v25.2d,v26.2d}, [x2], #32
354 zip1 v27.2D, v3.2D, v17.2D354 zip1 v27.2d, v3.2d, v17.2d
355 zip2 v28.2D, v3.2D, v17.2D355 zip2 v28.2d, v3.2d, v17.2d
356 st1 {v27.2D,v28.2D}, [x2], #32356 st1 {v27.2d,v28.2d}, [x2], #32
357 zip1 v29.2D, v4.2D, v18.2D357 zip1 v29.2d, v4.2d, v18.2d
358 zip2 v30.2D, v4.2D, v18.2D358 zip2 v30.2d, v4.2d, v18.2d
359 st1 {v29.2D,v30.2D}, [x2], #32359 st1 {v29.2d,v30.2d}, [x2], #32
360 360
361 idct_end361 idct_end
362endfunc362endfunc
libavfilter/aarch64/vf_nlmeans_neon.S+35−35
@@ -22,19 +22,19 @@ · 21 unmodified lines
22 22
23// acc_sum_store(ABCD) = {X+A, X+A+B, X+A+B+C, X+A+B+C+D}23// acc_sum_store(ABCD) = {X+A, X+A+B, X+A+B+C, X+A+B+C+D}
24.macro acc_sum_store x, xb24.macro acc_sum_store x, xb
25 dup v24.4S, v24.S[3] // ...X -> XXXX25 dup v24.4s, v24.s[3] // ...X -> XXXX
26 ext v25.16B, v26.16B, \xb, #12 // ext(0000,ABCD,12)=0ABC26 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}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)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)=00AB29 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)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)=000A31 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)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 values33 st1 {v24.4s}, [x0], #16 // write 4x32-bit final values
34.endm34.endm
35 35
36function ff_compute_safe_ssd_integral_image_neon, export=136function ff_compute_safe_ssd_integral_image_neon, export=1
37 movi v26.4S, #0 // used as zero for the "rotations" in acc_sum_store37 movi v26.4s, #0 // used as zero for the "rotations" in acc_sum_store
38 sub x3, x3, w6, UXTW // s1 padding (s1_linesize - w)38 sub x3, x3, w6, UXTW // s1 padding (s1_linesize - w)
39 sub x5, x5, w6, UXTW // s2 padding (s2_linesize - w)39 sub x5, x5, w6, UXTW // s2 padding (s2_linesize - w)
40 sub x9, x0, w1, UXTW #2 // dst_top40 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
431: mov w10, w6 // width copy for each line431: mov w10, w6 // width copy for each line
44 sub x0, x0, #16 // beginning of the dst line minus 4 sums44 sub x0, x0, #16 // beginning of the dst line minus 4 sums
45 sub x8, x9, #4 // dst_top-145 sub x8, x9, #4 // dst_top-1
46 ld1 {v24.4S}, [x0], #16 // load ...X (contextual last sums)46 ld1 {v24.4s}, [x0], #16 // load ...X (contextual last sums)
472: ld1 {v0.16B}, [x2], #16 // s1[x + 0..15]472: ld1 {v0.16b}, [x2], #16 // s1[x + 0..15]
48 ld1 {v1.16B}, [x4], #16 // s2[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]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]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]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]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]^253 smull v4.4s, v2.4h, v2.4h // d[x + 0..3]^2
54 smull2 v5.4S, v2.8H, v2.8H // d[x + 4..7]^254 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]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]^256 smull v6.4s, v3.4h, v3.4h // d[x + 8..11]^2
57 smull2 v7.4S, v3.8H, v3.8H // d[x + 12..15]^257 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]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]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]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]^261 add v0.4s, v0.4s, v4.4s // + d[x + 0..3]^2
62 add v1.4S, v1.4S, v5.4S // + d[x + 4..7]^262 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]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]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]^265 add v2.4s, v2.4s, v6.4s // + d[x + 8..11]^2
66 add v3.4S, v3.4S, v7.4S // + d[x + 12..15]^266 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]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]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]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]70 acc_sum_store v3.4s, v3.16b // accumulate and store dst[12..15]
71 subs w10, w10, #16 // width dec71 subs w10, w10, #16 // width dec
72 b.ne 2b // loop til next line72 b.ne 2b // loop til next line
73 add x2, x2, x3 // skip to next line (s1)73 add x2, x2, x3 // skip to next line (s1)
libavutil/aarch64/float_dsp_neon.S+100−100
@@ -25,16 +25,16 @@ · 24 unmodified lines
25 25
26function ff_vector_fmul_neon, export=126function ff_vector_fmul_neon, export=1
271: subs w3, w3, #16271: subs w3, w3, #16
28 ld1 {v0.4S, v1.4S}, [x1], #3228 ld1 {v0.4s, v1.4s}, [x1], #32
29 ld1 {v2.4S, v3.4S}, [x1], #3229 ld1 {v2.4s, v3.4s}, [x1], #32
30 ld1 {v4.4S, v5.4S}, [x2], #3230 ld1 {v4.4s, v5.4s}, [x2], #32
31 ld1 {v6.4S, v7.4S}, [x2], #3231 ld1 {v6.4s, v7.4s}, [x2], #32
32 fmul v16.4S, v0.4S, v4.4S32 fmul v16.4s, v0.4s, v4.4s
33 fmul v17.4S, v1.4S, v5.4S33 fmul v17.4s, v1.4s, v5.4s
34 fmul v18.4S, v2.4S, v6.4S34 fmul v18.4s, v2.4s, v6.4s
35 fmul v19.4S, v3.4S, v7.4S35 fmul v19.4s, v3.4s, v7.4s
36 st1 {v16.4S, v17.4S}, [x0], #3236 st1 {v16.4s, v17.4s}, [x0], #32
37 st1 {v18.4S, v19.4S}, [x0], #3237 st1 {v18.4s, v19.4s}, [x0], #32
38 b.ne 1b38 b.ne 1b
39 ret39 ret
40endfunc40endfunc
@@ -42,16 +42,16 @@ endfunc · 1 unmodified line
42function ff_vector_fmac_scalar_neon, export=142function ff_vector_fmac_scalar_neon, export=1
43 mov x3, #-3243 mov x3, #-32
441: subs w2, w2, #16441: subs w2, w2, #16
45 ld1 {v16.4S, v17.4S}, [x0], #3245 ld1 {v16.4s, v17.4s}, [x0], #32
46 ld1 {v18.4S, v19.4S}, [x0], x346 ld1 {v18.4s, v19.4s}, [x0], x3
47 ld1 {v4.4S, v5.4S}, [x1], #3247 ld1 {v4.4s, v5.4s}, [x1], #32
48 ld1 {v6.4S, v7.4S}, [x1], #3248 ld1 {v6.4s, v7.4s}, [x1], #32
49 fmla v16.4S, v4.4S, v0.S[0]49 fmla v16.4s, v4.4s, v0.s[0]
50 fmla v17.4S, v5.4S, v0.S[0]50 fmla v17.4s, v5.4s, v0.s[0]
51 fmla v18.4S, v6.4S, v0.S[0]51 fmla v18.4s, v6.4s, v0.s[0]
52 fmla v19.4S, v7.4S, v0.S[0]52 fmla v19.4s, v7.4s, v0.s[0]
53 st1 {v16.4S, v17.4S}, [x0], #3253 st1 {v16.4s, v17.4s}, [x0], #32
54 st1 {v18.4S, v19.4S}, [x0], #3254 st1 {v18.4s, v19.4s}, [x0], #32
55 b.ne 1b55 b.ne 1b
56 ret56 ret
57endfunc57endfunc
@@ -59,43 +59,43 @@ endfunc · 1 unmodified line
59function ff_vector_fmul_scalar_neon, export=159function ff_vector_fmul_scalar_neon, export=1
60 mov w4, #1560 mov w4, #15
61 bics w3, w2, w461 bics w3, w2, w4
62 dup v16.4S, v0.S[0]62 dup v16.4s, v0.s[0]
63 b.eq 3f63 b.eq 3f
64 ld1 {v0.4S, v1.4S}, [x1], #3264 ld1 {v0.4s, v1.4s}, [x1], #32
651: subs w3, w3, #16651: subs w3, w3, #16
66 fmul v0.4S, v0.4S, v16.4S66 fmul v0.4s, v0.4s, v16.4s
67 ld1 {v2.4S, v3.4S}, [x1], #3267 ld1 {v2.4s, v3.4s}, [x1], #32
68 fmul v1.4S, v1.4S, v16.4S68 fmul v1.4s, v1.4s, v16.4s
69 fmul v2.4S, v2.4S, v16.4S69 fmul v2.4s, v2.4s, v16.4s
70 st1 {v0.4S, v1.4S}, [x0], #3270 st1 {v0.4s, v1.4s}, [x0], #32
71 fmul v3.4S, v3.4S, v16.4S71 fmul v3.4s, v3.4s, v16.4s
72 b.eq 2f72 b.eq 2f
73 ld1 {v0.4S, v1.4S}, [x1], #3273 ld1 {v0.4s, v1.4s}, [x1], #32
74 st1 {v2.4S, v3.4S}, [x0], #3274 st1 {v2.4s, v3.4s}, [x0], #32
75 b 1b75 b 1b
762: ands w2, w2, #15762: ands w2, w2, #15
77 st1 {v2.4S, v3.4S}, [x0], #3277 st1 {v2.4s, v3.4s}, [x0], #32
78 b.eq 4f78 b.eq 4f
793: ld1 {v0.4S}, [x1], #16793: ld1 {v0.4s}, [x1], #16
80 fmul v0.4S, v0.4S, v16.4S80 fmul v0.4s, v0.4s, v16.4s
81 st1 {v0.4S}, [x0], #1681 st1 {v0.4s}, [x0], #16
82 subs w2, w2, #482 subs w2, w2, #4
83 b.gt 3b83 b.gt 3b
844: ret844: ret
85endfunc85endfunc
86 86
87function ff_vector_dmul_scalar_neon, export=187function ff_vector_dmul_scalar_neon, export=1
88 dup v16.2D, v0.D[0]88 dup v16.2d, v0.d[0]
89 ld1 {v0.2D, v1.2D}, [x1], #3289 ld1 {v0.2d, v1.2d}, [x1], #32
901: subs w2, w2, #8901: subs w2, w2, #8
91 fmul v0.2D, v0.2D, v16.2D91 fmul v0.2d, v0.2d, v16.2d
92 ld1 {v2.2D, v3.2D}, [x1], #3292 ld1 {v2.2d, v3.2d}, [x1], #32
93 fmul v1.2D, v1.2D, v16.2D93 fmul v1.2d, v1.2d, v16.2d
94 fmul v2.2D, v2.2D, v16.2D94 fmul v2.2d, v2.2d, v16.2d
95 st1 {v0.2D, v1.2D}, [x0], #3295 st1 {v0.2d, v1.2d}, [x0], #32
96 fmul v3.2D, v3.2D, v16.2D96 fmul v3.2d, v3.2d, v16.2d
97 ld1 {v0.2D, v1.2D}, [x1], #3297 ld1 {v0.2d, v1.2d}, [x1], #32
98 st1 {v2.2D, v3.2D}, [x0], #3298 st1 {v2.2d, v3.2d}, [x0], #32
99 b.gt 1b99 b.gt 1b
100 ret100 ret
101endfunc101endfunc
@@ -108,49 +108,49 @@ function ff_vector_fmul_window_neon, export=1 · 6 unmodified lines
108 add x6, x3, x5, lsl #3 // win + 8 * (len - 2)108 add x6, x3, x5, lsl #3 // win + 8 * (len - 2)
109 add x5, x0, x5, lsl #3 // dst + 8 * (len - 2)109 add x5, x0, x5, lsl #3 // dst + 8 * (len - 2)
110 mov x7, #-16110 mov x7, #-16
111 ld1 {v0.4S}, [x1], #16 // s0111 ld1 {v0.4s}, [x1], #16 // s0
112 ld1 {v2.4S}, [x3], #16 // wi112 ld1 {v2.4s}, [x3], #16 // wi
113 ld1 {v1.4S}, [x2], x7 // s1113 ld1 {v1.4s}, [x2], x7 // s1
1141: ld1 {v3.4S}, [x6], x7 // wj1141: ld1 {v3.4s}, [x6], x7 // wj
115 subs x4, x4, #4115 subs x4, x4, #4
116 fmul v17.4S, v0.4S, v2.4S // s0 * wi116 fmul v17.4s, v0.4s, v2.4s // s0 * wi
117 rev64 v4.4S, v1.4S117 rev64 v4.4s, v1.4s
118 rev64 v5.4S, v3.4S118 rev64 v5.4s, v3.4s
119 rev64 v17.4S, v17.4S119 rev64 v17.4s, v17.4s
120 ext v4.16B, v4.16B, v4.16B, #8 // s1_r120 ext v4.16b, v4.16b, v4.16b, #8 // s1_r
121 ext v5.16B, v5.16B, v5.16B, #8 // wj_r121 ext v5.16b, v5.16b, v5.16b, #8 // wj_r
122 ext v17.16B, v17.16B, v17.16B, #8 // (s0 * wi)_rev122 ext v17.16b, v17.16b, v17.16b, #8 // (s0 * wi)_rev
123 fmul v16.4S, v0.4S, v5.4S // s0 * wj_r123 fmul v16.4s, v0.4s, v5.4s // s0 * wj_r
124 fmla v17.4S, v1.4S, v3.4S // (s0 * wi)_rev + s1 * wj124 fmla v17.4s, v1.4s, v3.4s // (s0 * wi)_rev + s1 * wj
125 b.eq 2f125 b.eq 2f
126 ld1 {v0.4S}, [x1], #16126 ld1 {v0.4s}, [x1], #16
127 fmls v16.4S, v4.4S, v2.4S // s0 * wj_r - s1_r * wi127 fmls v16.4s, v4.4s, v2.4s // s0 * wj_r - s1_r * wi
128 st1 {v17.4S}, [x5], x7128 st1 {v17.4s}, [x5], x7
129 ld1 {v2.4S}, [x3], #16129 ld1 {v2.4s}, [x3], #16
130 ld1 {v1.4S}, [x2], x7130 ld1 {v1.4s}, [x2], x7
131 st1 {v16.4S}, [x0], #16131 st1 {v16.4s}, [x0], #16
132 b 1b132 b 1b
1332:1332:
134 fmls v16.4S, v4.4S, v2.4S // s0 * wj_r - s1_r * wi134 fmls v16.4s, v4.4s, v2.4s // s0 * wj_r - s1_r * wi
135 st1 {v17.4S}, [x5], x7135 st1 {v17.4s}, [x5], x7
136 st1 {v16.4S}, [x0], #16136 st1 {v16.4s}, [x0], #16
137 ret137 ret
138endfunc138endfunc
139 139
140function ff_vector_fmul_add_neon, export=1140function ff_vector_fmul_add_neon, export=1
141 ld1 {v0.4S, v1.4S}, [x1], #32141 ld1 {v0.4s, v1.4s}, [x1], #32
142 ld1 {v2.4S, v3.4S}, [x2], #32142 ld1 {v2.4s, v3.4s}, [x2], #32
143 ld1 {v4.4S, v5.4S}, [x3], #32143 ld1 {v4.4s, v5.4s}, [x3], #32
1441: subs w4, w4, #81441: subs w4, w4, #8
145 fmla v4.4S, v0.4S, v2.4S145 fmla v4.4s, v0.4s, v2.4s
146 fmla v5.4S, v1.4S, v3.4S146 fmla v5.4s, v1.4s, v3.4s
147 b.eq 2f147 b.eq 2f
148 ld1 {v0.4S, v1.4S}, [x1], #32148 ld1 {v0.4s, v1.4s}, [x1], #32
149 ld1 {v2.4S, v3.4S}, [x2], #32149 ld1 {v2.4s, v3.4s}, [x2], #32
150 st1 {v4.4S, v5.4S}, [x0], #32150 st1 {v4.4s, v5.4s}, [x0], #32
151 ld1 {v4.4S, v5.4S}, [x3], #32151 ld1 {v4.4s, v5.4s}, [x3], #32
152 b 1b152 b 1b
1532: st1 {v4.4S, v5.4S}, [x0], #321532: st1 {v4.4s, v5.4s}, [x0], #32
154 ret154 ret
155endfunc155endfunc
156 156
@@ -159,44 +159,44 @@ function ff_vector_fmul_reverse_neon, export=1 · 2 unmodified lines
159 add x2, x2, x3, lsl #2159 add x2, x2, x3, lsl #2
160 sub x2, x2, #32160 sub x2, x2, #32
161 mov x4, #-32161 mov x4, #-32
162 ld1 {v2.4S, v3.4S}, [x2], x4162 ld1 {v2.4s, v3.4s}, [x2], x4
163 ld1 {v0.4S, v1.4S}, [x1], #32163 ld1 {v0.4s, v1.4s}, [x1], #32
1641: subs x3, x3, #81641: subs x3, x3, #8
165 rev64 v3.4S, v3.4S165 rev64 v3.4s, v3.4s
166 rev64 v2.4S, v2.4S166 rev64 v2.4s, v2.4s
167 ext v3.16B, v3.16B, v3.16B, #8167 ext v3.16b, v3.16b, v3.16b, #8
168 ext v2.16B, v2.16B, v2.16B, #8168 ext v2.16b, v2.16b, v2.16b, #8
169 fmul v16.4S, v0.4S, v3.4S169 fmul v16.4s, v0.4s, v3.4s
170 fmul v17.4S, v1.4S, v2.4S170 fmul v17.4s, v1.4s, v2.4s
171 b.eq 2f171 b.eq 2f
172 ld1 {v2.4S, v3.4S}, [x2], x4172 ld1 {v2.4s, v3.4s}, [x2], x4
173 ld1 {v0.4S, v1.4S}, [x1], #32173 ld1 {v0.4s, v1.4s}, [x1], #32
174 st1 {v16.4S, v17.4S}, [x0], #32174 st1 {v16.4s, v17.4s}, [x0], #32
175 b 1b175 b 1b
1762: st1 {v16.4S, v17.4S}, [x0], #321762: st1 {v16.4s, v17.4s}, [x0], #32
177 ret177 ret
178endfunc178endfunc
179 179
180function ff_butterflies_float_neon, export=1180function ff_butterflies_float_neon, export=1
1811: ld1 {v0.4S}, [x0]1811: ld1 {v0.4s}, [x0]
182 ld1 {v1.4S}, [x1]182 ld1 {v1.4s}, [x1]
183 subs w2, w2, #4183 subs w2, w2, #4
184 fsub v2.4S, v0.4S, v1.4S184 fsub v2.4s, v0.4s, v1.4s
185 fadd v3.4S, v0.4S, v1.4S185 fadd v3.4s, v0.4s, v1.4s
186 st1 {v2.4S}, [x1], #16186 st1 {v2.4s}, [x1], #16
187 st1 {v3.4S}, [x0], #16187 st1 {v3.4s}, [x0], #16
188 b.gt 1b188 b.gt 1b
189 ret189 ret
190endfunc190endfunc
191 191
192function ff_scalarproduct_float_neon, export=1192function ff_scalarproduct_float_neon, export=1
193 movi v2.4S, #0193 movi v2.4s, #0
1941: ld1 {v0.4S}, [x0], #161941: ld1 {v0.4s}, [x0], #16
195 ld1 {v1.4S}, [x1], #16195 ld1 {v1.4s}, [x1], #16
196 subs w2, w2, #4196 subs w2, w2, #4
197 fmla v2.4S, v0.4S, v1.4S197 fmla v2.4s, v0.4s, v1.4s
198 b.gt 1b198 b.gt 1b
199 faddp v0.4S, v2.4S, v2.4S199 faddp v0.4s, v2.4s, v2.4s
200 faddp s0, v0.2S200 faddp s0, v0.2s
201 ret201 ret
202endfunc202endfunc
libswresample/aarch64/resample.S+32−32
@@ -21,57 +21,57 @@ · 20 unmodified lines
21#include "libavutil/aarch64/asm.S"21#include "libavutil/aarch64/asm.S"
22 22
23function ff_resample_common_apply_filter_x4_float_neon, export=123function ff_resample_common_apply_filter_x4_float_neon, export=1
24 movi v0.4S, #0 // accumulator24 movi v0.4s, #0 // accumulator
251: ld1 {v1.4S}, [x1], #16 // src[0..3]251: ld1 {v1.4s}, [x1], #16 // src[0..3]
26 ld1 {v2.4S}, [x2], #16 // filter[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]27 fmla v0.4s, v1.4s, v2.4s // accumulator += src[0..3] * filter[0..3]
28 subs w3, w3, #4 // filter_length -= 428 subs w3, w3, #4 // filter_length -= 4
29 b.gt 1b // loop until filter_length29 b.gt 1b // loop until filter_length
30 faddp v0.4S, v0.4S, v0.4S // pair adding of the 4x32-bit accumulated values30 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 values31 faddp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values
32 st1 {v0.S}[0], [x0], #4 // write accumulator32 st1 {v0.s}[0], [x0], #4 // write accumulator
33 ret33 ret
34endfunc34endfunc
35 35
36function ff_resample_common_apply_filter_x8_float_neon, export=136function ff_resample_common_apply_filter_x8_float_neon, export=1
37 movi v0.4S, #0 // accumulator37 movi v0.4s, #0 // accumulator
381: ld1 {v1.4S}, [x1], #16 // src[0..3]381: ld1 {v1.4s}, [x1], #16 // src[0..3]
39 ld1 {v2.4S}, [x2], #16 // filter[0..3]39 ld1 {v2.4s}, [x2], #16 // filter[0..3]
40 ld1 {v3.4S}, [x1], #16 // src[4..7]40 ld1 {v3.4s}, [x1], #16 // src[4..7]
41 ld1 {v4.4S}, [x2], #16 // filter[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]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]43 fmla v0.4s, v3.4s, v4.4s // accumulator += src[4..7] * filter[4..7]
44 subs w3, w3, #8 // filter_length -= 844 subs w3, w3, #8 // filter_length -= 8
45 b.gt 1b // loop until filter_length45 b.gt 1b // loop until filter_length
46 faddp v0.4S, v0.4S, v0.4S // pair adding of the 4x32-bit accumulated values46 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 values47 faddp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values
48 st1 {v0.S}[0], [x0], #4 // write accumulator48 st1 {v0.s}[0], [x0], #4 // write accumulator
49 ret49 ret
50endfunc50endfunc
51 51
52function ff_resample_common_apply_filter_x4_s16_neon, export=152function ff_resample_common_apply_filter_x4_s16_neon, export=1
53 movi v0.4S, #0 // accumulator53 movi v0.4s, #0 // accumulator
541: ld1 {v1.4H}, [x1], #8 // src[0..3]541: ld1 {v1.4h}, [x1], #8 // src[0..3]
55 ld1 {v2.4H}, [x2], #8 // filter[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]56 smlal v0.4s, v1.4h, v2.4h // accumulator += src[0..3] * filter[0..3]
57 subs w3, w3, #4 // filter_length -= 457 subs w3, w3, #4 // filter_length -= 4
58 b.gt 1b // loop until filter_length58 b.gt 1b // loop until filter_length
59 addp v0.4S, v0.4S, v0.4S // pair adding of the 4x32-bit accumulated values59 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 values60 addp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values
61 st1 {v0.S}[0], [x0], #4 // write accumulator61 st1 {v0.s}[0], [x0], #4 // write accumulator
62 ret62 ret
63endfunc63endfunc
64 64
65function ff_resample_common_apply_filter_x8_s16_neon, export=165function ff_resample_common_apply_filter_x8_s16_neon, export=1
66 movi v0.4S, #0 // accumulator66 movi v0.4s, #0 // accumulator
671: ld1 {v1.8H}, [x1], #16 // src[0..7]671: ld1 {v1.8h}, [x1], #16 // src[0..7]
68 ld1 {v2.8H}, [x2], #16 // filter[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]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]70 smlal2 v0.4s, v1.8h, v2.8h // accumulator += src[4..7] * filter[4..7]
71 subs w3, w3, #8 // filter_length -= 871 subs w3, w3, #8 // filter_length -= 8
72 b.gt 1b // loop until filter_length72 b.gt 1b // loop until filter_length
73 addp v0.4S, v0.4S, v0.4S // pair adding of the 4x32-bit accumulated values73 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 values74 addp v0.4s, v0.4s, v0.4s // pair adding of the 4x32-bit accumulated values
75 st1 {v0.S}[0], [x0], #4 // write accumulator75 st1 {v0.s}[0], [x0], #4 // write accumulator
76 ret76 ret
77endfunc77endfunc
libswscale/aarch64/hscale.S+216−216
@@ -50,43 +50,43 @@ function ff_hscale8to15_X8_neon, export=1 · 49 unmodified lines
50 add x12, x16, x7 // filter1 = filter0 + filterSize*250 add x12, x16, x7 // filter1 = filter0 + filterSize*2
51 add x13, x12, x7 // filter2 = filter1 + filterSize*251 add x13, x12, x7 // filter2 = filter1 + filterSize*2
52 add x4, x13, x7 // filter3 = filter2 + filterSize*252 add x4, x13, x7 // filter3 = filter2 + filterSize*2
53 movi v0.2D, #0 // val sum part 1 (for dst[0])53 movi v0.2d, #0 // val sum part 1 (for dst[0])
54 movi v1.2D, #0 // val sum part 2 (for dst[1])54 movi v1.2d, #0 // val sum part 2 (for dst[1])
55 movi v2.2D, #0 // val sum part 3 (for dst[2])55 movi v2.2d, #0 // val sum part 3 (for dst[2])
56 movi v3.2D, #0 // val sum part 4 (for dst[3])56 movi v3.2d, #0 // val sum part 4 (for dst[3])
57 add x17, x3, w8, UXTW // srcp + filterPos[0]57 add x17, x3, w8, UXTW // srcp + filterPos[0]
58 add x8, x3, w0, UXTW // srcp + filterPos[1]58 add x8, x3, w0, UXTW // srcp + filterPos[1]
59 add x0, x3, w11, UXTW // srcp + filterPos[2]59 add x0, x3, w11, UXTW // srcp + filterPos[2]
60 add x11, x3, w9, UXTW // srcp + filterPos[3]60 add x11, x3, w9, UXTW // srcp + filterPos[3]
61 mov w15, w6 // filterSize counter61 mov w15, w6 // filterSize counter
622: ld1 {v4.8B}, [x17], #8 // srcp[filterPos[0] + {0..7}]622: ld1 {v4.8b}, [x17], #8 // srcp[filterPos[0] + {0..7}]
63 ld1 {v5.8H}, [x16], #16 // load 8x16-bit filter values, part 163 ld1 {v5.8h}, [x16], #16 // load 8x16-bit filter values, part 1
64 ld1 {v6.8B}, [x8], #8 // srcp[filterPos[1] + {0..7}]64 ld1 {v6.8b}, [x8], #8 // srcp[filterPos[1] + {0..7}]
65 ld1 {v7.8H}, [x12], #16 // load 8x16-bit at filter+filterSize65 ld1 {v7.8h}, [x12], #16 // load 8x16-bit at filter+filterSize
66 uxtl v4.8H, v4.8B // unpack part 1 to 16-bit66 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}]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}]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}]69 ld1 {v16.8b}, [x0], #8 // srcp[filterPos[2] + {0..7}]
70 ld1 {v17.8H}, [x13], #16 // load 8x16-bit at filter+2*filterSize70 ld1 {v17.8h}, [x13], #16 // load 8x16-bit at filter+2*filterSize
71 uxtl v6.8H, v6.8B // unpack part 2 to 16-bit71 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}]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-bit73 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}]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}]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}]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}]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*filterSize78 ld1 {v19.8h}, [x4], #16 // load 8x16-bit at filter+3*filterSize
79 subs w15, w15, #8 // j -= 8: processed 8/filterSize79 subs w15, w15, #8 // j -= 8: processed 8/filterSize
80 uxtl v18.8H, v18.8B // unpack part 4 to 16-bit80 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}]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}]82 smlal2 v3.4s, v18.8h, v19.8h // v3 accumulates srcp[filterPos[3] + {4..7}] * filter[{4..7}]
83 b.gt 2b // inner loop if filterSize not consumed completely83 b.gt 2b // inner loop if filterSize not consumed completely
84 addp v0.4S, v0.4S, v1.4S // part01 horizontal pair adding84 addp v0.4s, v0.4s, v1.4s // part01 horizontal pair adding
85 addp v2.4S, v2.4S, v3.4S // part23 horizontal pair adding85 addp v2.4s, v2.4s, v3.4s // part23 horizontal pair adding
86 addp v0.4S, v0.4S, v2.4S // part0123 horizontal pair adding86 addp v0.4s, v0.4s, v2.4s // part0123 horizontal pair adding
87 subs w2, w2, #4 // dstW -= 487 subs w2, w2, #4 // dstW -= 4
88 sqshrn v0.4H, v0.4S, #7 // shift and clip the 2x16-bit final values88 sqshrn v0.4h, v0.4s, #7 // shift and clip the 2x16-bit final values
89 st1 {v0.4H}, [x1], #8 // write to destination part012389 st1 {v0.4h}, [x1], #8 // write to destination part0123
90 b.gt 1b // loop until end of line90 b.gt 1b // loop until end of line
91 ret91 ret
92endfunc92endfunc
@@ -245,7 +245,7 @@ function ff_hscale8to15_4_neon, export=1 · 152 unmodified lines
245 stp w14, w15, [sp, #24] // *scratch_mem = { src[filterPos[idx + 6]][0..3], src[filterPos[idx + 7]][0..3] }245 stp w14, w15, [sp, #24] // *scratch_mem = { src[filterPos[idx + 6]][0..3], src[filterPos[idx + 7]][0..3] }
246 246
2471:2471:
248 ld4 {v16.8B, v17.8B, v18.8B, v19.8B}, [sp] // transpose 8 bytes each from src into 4 registers248 ld4 {v16.8b, v17.8b, v18.8b, v19.8b}, [sp] // transpose 8 bytes each from src into 4 registers
249 249
250 // load 8 values from filterPos to be used as offsets into src250 // load 8 values from filterPos to be used as offsets into src
251 ldp w8, w9, [x5] // filterPos[idx + 0][0..3], [idx + 1][0..3], next iteration251 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
253 ldp w12, w13, [x5, #16] // filterPos[idx + 4][0..3], [idx + 5][0..3], next iteration253 ldp w12, w13, [x5, #16] // filterPos[idx + 4][0..3], [idx + 5][0..3], next iteration
254 ldp w14, w15, [x5, #24] // filterPos[idx + 6][0..3], [idx + 7][0..3], next iteration254 ldp w14, w15, [x5, #24] // filterPos[idx + 6][0..3], [idx + 7][0..3], next iteration
255 255
256 movi v0.2D, #0 // Clear madd accumulator for idx 0..3256 movi v0.2d, #0 // Clear madd accumulator for idx 0..3
257 movi v5.2D, #0 // Clear madd accumulator for idx 4..7257 movi v5.2d, #0 // Clear madd accumulator for idx 4..7
258 258
259 ld4 {v1.8H, v2.8H, v3.8H, v4.8H}, [x4], #64 // load filter idx + 0..7259 ld4 {v1.8h, v2.8h, v3.8h, v4.8h}, [x4], #64 // load filter idx + 0..7
260 260
261 add x5, x5, #32 // advance filterPos261 add x5, x5, #32 // advance filterPos
262 262
263 // interleaved SIMD and prefetching intended to keep ld/st and vector pipelines busy263 // 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-bit264 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-bit265 uxtl v17.8h, v17.8b // unsigned extend long, covert src data to 16-bit
266 ldr w8, [x3, w8, UXTW] // src[filterPos[idx + 0]], next iteration266 ldr w8, [x3, w8, UXTW] // src[filterPos[idx + 0]], next iteration
267 ldr w9, [x3, w9, UXTW] // src[filterPos[idx + 1]], next iteration267 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-bit268 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-bit269 uxtl v19.8h, v19.8b // unsigned extend long, covert src data to 16-bit
270 ldr w10, [x3, w10, UXTW] // src[filterPos[idx + 2]], next iteration270 ldr w10, [x3, w10, UXTW] // src[filterPos[idx + 2]], next iteration
271 ldr w11, [x3, w11, UXTW] // src[filterPos[idx + 3]], next iteration271 ldr w11, [x3, w11, UXTW] // src[filterPos[idx + 3]], next iteration
272 272
273 smlal v0.4S, v1.4H, v16.4H // multiply accumulate inner loop j = 0, idx = 0..3273 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..3274 smlal v0.4s, v2.4h, v17.4h // multiply accumulate inner loop j = 1, idx = 0..3
275 ldr w12, [x3, w12, UXTW] // src[filterPos[idx + 4]], next iteration275 ldr w12, [x3, w12, UXTW] // src[filterPos[idx + 4]], next iteration
276 ldr w13, [x3, w13, UXTW] // src[filterPos[idx + 5]], next iteration276 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..3277 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..3278 smlal v0.4s, v4.4h, v19.4h // multiply accumulate inner loop j = 3, idx = 0..3
279 ldr w14, [x3, w14, UXTW] // src[filterPos[idx + 6]], next iteration279 ldr w14, [x3, w14, UXTW] // src[filterPos[idx + 6]], next iteration
280 ldr w15, [x3, w15, UXTW] // src[filterPos[idx + 7]], next iteration280 ldr w15, [x3, w15, UXTW] // src[filterPos[idx + 7]], next iteration
281 281
282 smlal2 v5.4S, v1.8H, v16.8H // multiply accumulate inner loop j = 0, idx = 4..7282 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..7283 smlal2 v5.4s, v2.8h, v17.8h // multiply accumulate inner loop j = 1, idx = 4..7
284 stp w8, w9, [sp] // *scratch_mem = { src[filterPos[idx + 0]][0..3], src[filterPos[idx + 1]][0..3] }284 stp w8, w9, [sp] // *scratch_mem = { src[filterPos[idx + 0]][0..3], src[filterPos[idx + 1]][0..3] }
285 stp w10, w11, [sp, #8] // *scratch_mem = { src[filterPos[idx + 2]][0..3], src[filterPos[idx + 3]][0..3] }285 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..7286 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..7287 smlal2 v5.4s, v4.8h, v19.8h // multiply accumulate inner loop j = 3, idx = 4..7
288 stp w12, w13, [sp, #16] // *scratch_mem = { src[filterPos[idx + 4]][0..3], src[filterPos[idx + 5]][0..3] }288 stp w12, w13, [sp, #16] // *scratch_mem = { src[filterPos[idx + 4]][0..3], src[filterPos[idx + 5]][0..3] }
289 stp w14, w15, [sp, #24] // *scratch_mem = { src[filterPos[idx + 6]][0..3], src[filterPos[idx + 7]][0..3] }289 stp w14, w15, [sp, #24] // *scratch_mem = { src[filterPos[idx + 6]][0..3], src[filterPos[idx + 7]][0..3] }
290 290
291 sub w2, w2, #8 // dstW -= 8291 sub w2, w2, #8 // dstW -= 8
292 sqshrn v0.4H, v0.4S, #7 // shift and clip the 2x16-bit final values292 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 values293 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]294 st1 {v0.4h, v1.4h}, [x1], #16 // write to dst[idx + 0..7]
295 cmp w2, #16 // continue on main loop if there are at least 16 iterations left295 cmp w2, #16 // continue on main loop if there are at least 16 iterations left
296 b.ge 1b296 b.ge 1b
297 297
298 // last full iteration298 // last full iteration
299 ld4 {v16.8B, v17.8B, v18.8B, v19.8B}, [sp]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..7300 ld4 {v1.8h, v2.8h, v3.8h, v4.8h}, [x4], #64 // load filter idx + 0..7
301 301
302 movi v0.2D, #0 // Clear madd accumulator for idx 0..3302 movi v0.2d, #0 // Clear madd accumulator for idx 0..3
303 movi v5.2D, #0 // Clear madd accumulator for idx 4..7303 movi v5.2d, #0 // Clear madd accumulator for idx 4..7
304 304
305 uxtl v16.8H, v16.8B // unsigned extend long, covert src data to 16-bit305 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-bit306 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-bit307 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-bit308 uxtl v19.8h, v19.8b // unsigned extend long, covert src data to 16-bit
309 309
310 smlal v0.4S, v1.4H, v16.4H // multiply accumulate inner loop j = 0, idx = 0..3310 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..3311 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..3312 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..3313 smlal v0.4s, v4.4h, v19.4h // multiply accumulate inner loop j = 3, idx = 0..3
314 314
315 smlal2 v5.4S, v1.8H, v16.8H // multiply accumulate inner loop j = 0, idx = 4..7315 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..7316 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..7317 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..7318 smlal2 v5.4s, v4.8h, v19.8h // multiply accumulate inner loop j = 3, idx = 4..7
319 319
320 subs w2, w2, #8 // dstW -= 8320 subs w2, w2, #8 // dstW -= 8
321 sqshrn v0.4H, v0.4S, #7 // shift and clip the 2x16-bit final values321 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 values322 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]323 st1 {v0.4h, v1.4h}, [x1], #16 // write to dst[idx + 0..7]
324 324
325 cbnz w2, 2f // if >0 iterations remain, jump to the wrap up section325 cbnz w2, 2f // if >0 iterations remain, jump to the wrap up section
326 326
@@ -332,15 +332,15 @@ function ff_hscale8to15_4_neon, export=1 · 5 unmodified lines
332 // load src332 // load src
333 ldr w8, [x5], #4 // filterPos[i]333 ldr w8, [x5], #4 // filterPos[i]
334 add x9, x3, w8, UXTW // calculate the address for src load334 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]
336 // load filter336 // load filter
337 ld1 {v6.4H}, [x4], #8 // filter[filterSize * i + 0..3]337 ld1 {v6.4h}, [x4], #8 // filter[filterSize * i + 0..3]
338 338
339 uxtl v5.8H, v5.8B // unsigned exten long, convert src data to 16-bit339 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[...]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 values341 addv s0, v0.4s // add up products of src and filter values
342 sqshrn h0, s0, #7 // shift and clip the 2x16-bit final value342 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] = ...
344 sub w2, w2, #1 // dstW--344 sub w2, w2, #1 // dstW--
345 cbnz w2, 2b345 cbnz w2, 2b
346 346
@@ -445,12 +445,12 @@ function ff_hscale8to19_4_neon, export=1 · 98 unmodified lines
445 smull v5.4s, v0.4h, v28.4h445 smull v5.4s, v0.4h, v28.4h
446 smull2 v6.4s, v0.8h, v28.8h446 smull2 v6.4s, v0.8h, v28.8h
447 uxtl v2.8h, v2.8b447 uxtl v2.8h, v2.8b
448 smlal v5.4s, v1.4h, v29.4H448 smlal v5.4s, v1.4h, v29.4h
449 smlal2 v6.4s, v1.8h, v29.8H449 smlal2 v6.4s, v1.8h, v29.8h
450 uxtl v3.8h, v3.8b450 uxtl v3.8h, v3.8b
451 smlal v5.4s, v2.4h, v30.4H451 smlal v5.4s, v2.4h, v30.4h
452 smlal2 v6.4s, v2.8h, v30.8H452 smlal2 v6.4s, v2.8h, v30.8h
453 smlal v5.4s, v3.4h, v31.4H453 smlal v5.4s, v3.4h, v31.4h
454 smlal2 v6.4s, v3.8h, v31.8h454 smlal2 v6.4s, v3.8h, v31.8h
455 455
456 sshr v5.4s, v5.4s, #3456 sshr v5.4s, v5.4s, #3
@@ -472,8 +472,8 @@ function ff_hscale8to19_4_neon, export=1 · 15 unmodified lines
472 ld1 {v0.s}[0], [x9] // load 4 * uint8_t* into one single472 ld1 {v0.s}[0], [x9] // load 4 * uint8_t* into one single
473 ld1 {v31.4h}, [x4], #8473 ld1 {v31.4h}, [x4], #8
474 uxtl v0.8h, v0.8b474 uxtl v0.8h, v0.8b
475 smull v5.4s, v0.4h, v31.4H475 smull v5.4s, v0.4h, v31.4h
476 saddlv d0, v5.4S476 saddlv d0, v5.4s
477 sqshrn s0, d0, #3477 sqshrn s0, d0, #3
478 smin v0.4s, v0.4s, v18.4s478 smin v0.4s, v0.4s, v18.4s
479 st1 {v0.s}[0], [x1], #4479 st1 {v0.s}[0], [x1], #4
@@ -499,42 +499,42 @@ function ff_hscale8to19_X8_neon, export=1 · 19 unmodified lines
499 ldr w11, [x5], #4 // filterPos[idx + 2]499 ldr w11, [x5], #4 // filterPos[idx + 2]
500 add x4, x13, x7 // filter3 = filter2 + filterSize*2500 add x4, x13, x7 // filter3 = filter2 + filterSize*2
501 ldr w9, [x5], #4 // filterPos[idx + 3]501 ldr w9, [x5], #4 // filterPos[idx + 3]
502 movi v0.2D, #0 // val sum part 1 (for dst[0])502 movi v0.2d, #0 // val sum part 1 (for dst[0])
503 movi v1.2D, #0 // val sum part 2 (for dst[1])503 movi v1.2d, #0 // val sum part 2 (for dst[1])
504 movi v2.2D, #0 // val sum part 3 (for dst[2])504 movi v2.2d, #0 // val sum part 3 (for dst[2])
505 movi v3.2D, #0 // val sum part 4 (for dst[3])505 movi v3.2d, #0 // val sum part 4 (for dst[3])
506 add x17, x3, w8, UXTW // srcp + filterPos[0]506 add x17, x3, w8, UXTW // srcp + filterPos[0]
507 add x8, x3, w0, UXTW // srcp + filterPos[1]507 add x8, x3, w0, UXTW // srcp + filterPos[1]
508 add x0, x3, w11, UXTW // srcp + filterPos[2]508 add x0, x3, w11, UXTW // srcp + filterPos[2]
509 add x11, x3, w9, UXTW // srcp + filterPos[3]509 add x11, x3, w9, UXTW // srcp + filterPos[3]
510 mov w15, w6 // filterSize counter510 mov w15, w6 // filterSize counter
5112: ld1 {v4.8B}, [x17], #8 // srcp[filterPos[0] + {0..7}]5112: ld1 {v4.8b}, [x17], #8 // srcp[filterPos[0] + {0..7}]
512 ld1 {v5.8H}, [x16], #16 // load 8x16-bit filter values, part 1512 ld1 {v5.8h}, [x16], #16 // load 8x16-bit filter values, part 1
513 uxtl v4.8H, v4.8B // unpack part 1 to 16-bit513 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}]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}]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}]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+filterSize517 ld1 {v7.8h}, [x12], #16 // load 8x16-bit at filter+filterSize
518 ld1 {v16.8B}, [x0], #8 // srcp[filterPos[2] + {0..7}]518 ld1 {v16.8b}, [x0], #8 // srcp[filterPos[2] + {0..7}]
519 uxtl v6.8H, v6.8B // unpack part 2 to 16-bit519 uxtl v6.8h, v6.8b // unpack part 2 to 16-bit
520 ld1 {v17.8H}, [x13], #16 // load 8x16-bit at filter+2*filterSize520 ld1 {v17.8h}, [x13], #16 // load 8x16-bit at filter+2*filterSize
521 uxtl v16.8H, v16.8B // unpack part 3 to 16-bit521 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}]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}]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}]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*filterSize525 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}]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-bit527 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}]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}]529 smlal v3.4s, v18.4h, v19.4h // v3 accumulates srcp[filterPos[3] + {0..3}] * filter[{0..3}]
530 subs w15, w15, #8 // j -= 8: processed 8/filterSize530 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}]
532 b.gt 2b // inner loop if filterSize not consumed completely532 b.gt 2b // inner loop if filterSize not consumed completely
533 addp v0.4S, v0.4S, v1.4S // part01 horizontal pair adding533 addp v0.4s, v0.4s, v1.4s // part01 horizontal pair adding
534 addp v2.4S, v2.4S, v3.4S // part23 horizontal pair adding534 addp v2.4s, v2.4s, v3.4s // part23 horizontal pair adding
535 addp v0.4S, v0.4S, v2.4S // part0123 horizontal pair adding535 addp v0.4s, v0.4s, v2.4s // part0123 horizontal pair adding
536 subs w2, w2, #4 // dstW -= 4536 subs w2, w2, #4 // dstW -= 4
537 sshr v0.4s, v0.4S, #3 // shift and clip the 2x16-bit final values537 sshr v0.4s, v0.4s, #3 // shift and clip the 2x16-bit final values
538 smin v0.4s, v0.4s, v20.4s538 smin v0.4s, v0.4s, v20.4s
539 st1 {v0.4s}, [x1], #16 // write to destination part0123539 st1 {v0.4s}, [x1], #16 // write to destination part0123
540 b.gt 1b // loop until end of line540 b.gt 1b // loop until end of line
@@ -588,16 +588,16 @@ function ff_hscale8to19_X4_neon, export=1 · 47 unmodified lines
588 smlal2 v16.4s, v4.8h, v31.8h // multiplication of upper half for idx 0588 smlal2 v16.4s, v4.8h, v31.8h // multiplication of upper half for idx 0
589 ldr d6, [x10], #8 // load src values for idx 2589 ldr d6, [x10], #8 // load src values for idx 2
590 ldr q29, [x14, x16] // load filter values for idx 2590 ldr q29, [x14, x16] // load filter values for idx 2
591 smlal v17.4s, v5.4h, v30.4H // multiplication of lower half for idx 1591 smlal v17.4s, v5.4h, v30.4h // multiplication of lower half for idx 1
592 ldr d7, [x11], #8 // load src values for idx 3592 ldr d7, [x11], #8 // load src values for idx 3
593 smlal2 v17.4s, v5.8h, v30.8H // multiplication of upper half for idx 1593 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 size594 uxtl v6.8h, v6.8b // extend tpye to matchi the filter's size
595 ldr q28, [x15, x16] // load filter values for idx 3595 ldr q28, [x15, x16] // load filter values for idx 3
596 smlal v18.4s, v6.4h, v29.4h // multiplication of lower half for idx 2596 smlal v18.4s, v6.4h, v29.4h // multiplication of lower half for idx 2
597 uxtl v7.8h, v7.8B597 uxtl v7.8h, v7.8b
598 smlal2 v18.4s, v6.8h, v29.8H // multiplication of upper half for idx 2598 smlal2 v18.4s, v6.8h, v29.8h // multiplication of upper half for idx 2
599 sub w0, w0, #8599 sub w0, w0, #8
600 smlal v19.4s, v7.4h, v28.4H // multiplication of lower half for idx 3600 smlal v19.4s, v7.4h, v28.4h // multiplication of lower half for idx 3
601 cmp w0, #8601 cmp w0, #8
602 smlal2 v19.4s, v7.8h, v28.8h // multiplication of upper half for idx 3602 smlal2 v19.4s, v7.8h, v28.8h // multiplication of upper half for idx 3
603 add x16, x16, #16 // advance filter values indexing603 add x16, x16, #16 // advance filter values indexing
@@ -618,11 +618,11 @@ function ff_hscale8to19_X4_neon, export=1 · 14 unmodified lines
618 uxtl v5.8h, v5.8b // extend type to match the filter' size618 uxtl v5.8h, v5.8b // extend type to match the filter' size
619 ldr s6, [x10] // load src values for idx 2619 ldr s6, [x10] // load src values for idx 2
620 smlal v17.4s, v5.4h, v30.4h620 smlal v17.4s, v5.4h, v30.4h
621 uxtl v6.8h, v6.8B // extend type to match the filter's size621 uxtl v6.8h, v6.8b // extend type to match the filter's size
622 ldr d29, [x14, x17] // load filter values for idx 2622 ldr d29, [x14, x17] // load filter values for idx 2
623 ldr s7, [x11] // load src values for idx 3623 ldr s7, [x11] // load src values for idx 3
624 addp v16.4s, v16.4s, v17.4s624 addp v16.4s, v16.4s, v17.4s
625 uxtl v7.8h, v7.8B625 uxtl v7.8h, v7.8b
626 ldr d28, [x15, x17] // load filter values for idx 3626 ldr d28, [x15, x17] // load filter values for idx 3
627 smlal v18.4s, v6.4h, v29.4h627 smlal v18.4s, v6.4h, v29.4h
628 smlal v19.4s, v7.4h, v28.4h628 smlal v19.4s, v7.4h, v28.4h
@@ -700,31 +700,31 @@ function ff_hscale16to15_4_neon_asm, export=1 · 71 unmodified lines
700 // Extending to 32 bits is necessary, as unit16_t values can't700 // Extending to 32 bits is necessary, as unit16_t values can't
701 // be represented as int16_t without type promotion.701 // be represented as int16_t without type promotion.
702 uxtl v26.4s, v0.4h702 uxtl v26.4s, v0.4h
703 sxtl v27.4s, v28.4H703 sxtl v27.4s, v28.4h
704 uxtl2 v0.4s, v0.8h704 uxtl2 v0.4s, v0.8h
705 mul v5.4s, v26.4s, v27.4s705 mul v5.4s, v26.4s, v27.4s
706 sxtl2 v28.4s, v28.8H706 sxtl2 v28.4s, v28.8h
707 uxtl v26.4s, v1.4h707 uxtl v26.4s, v1.4h
708 mul v6.4s, v0.4s, v28.4s708 mul v6.4s, v0.4s, v28.4s
709 709
710 sxtl v27.4s, v29.4H710 sxtl v27.4s, v29.4h
711 uxtl2 v0.4s, v1.8h711 uxtl2 v0.4s, v1.8h
712 mla v5.4s, v27.4s, v26.4s712 mla v5.4s, v27.4s, v26.4s
713 sxtl2 v28.4s, v29.8H713 sxtl2 v28.4s, v29.8h
714 uxtl v26.4s, v2.4h714 uxtl v26.4s, v2.4h
715 mla v6.4s, v28.4s, v0.4s715 mla v6.4s, v28.4s, v0.4s
716 716
717 sxtl v27.4s, v30.4H717 sxtl v27.4s, v30.4h
718 uxtl2 v0.4s, v2.8h718 uxtl2 v0.4s, v2.8h
719 mla v5.4s, v27.4s, v26.4s719 mla v5.4s, v27.4s, v26.4s
720 sxtl2 v28.4s, v30.8H720 sxtl2 v28.4s, v30.8h
721 uxtl v26.4s, v3.4h721 uxtl v26.4s, v3.4h
722 mla v6.4s, v28.4s, v0.4s722 mla v6.4s, v28.4s, v0.4s
723 723
724 sxtl v27.4s, v31.4H724 sxtl v27.4s, v31.4h
725 uxtl2 v0.4s, v3.8h725 uxtl2 v0.4s, v3.8h
726 mla v5.4s, v27.4s, v26.4s726 mla v5.4s, v27.4s, v26.4s
727 sxtl2 v28.4s, v31.8H727 sxtl2 v28.4s, v31.8h
728 sub w2, w2, #8728 sub w2, w2, #8
729 mla v6.4s, v28.4s, v0.4s729 mla v6.4s, v28.4s, v0.4s
730 730
@@ -775,31 +775,31 @@ function ff_hscale16to15_4_neon_asm, export=1 · 44 unmodified lines
775 ld4 {v28.8h, v29.8h, v30.8h, v31.8h}, [x4], #64775 ld4 {v28.8h, v29.8h, v30.8h, v31.8h}, [x4], #64
776 776
777 uxtl v26.4s, v0.4h777 uxtl v26.4s, v0.4h
778 sxtl v27.4s, v28.4H778 sxtl v27.4s, v28.4h
779 uxtl2 v0.4s, v0.8h779 uxtl2 v0.4s, v0.8h
780 mul v5.4s, v26.4s, v27.4s780 mul v5.4s, v26.4s, v27.4s
781 sxtl2 v28.4s, v28.8H781 sxtl2 v28.4s, v28.8h
782 uxtl v26.4s, v1.4h782 uxtl v26.4s, v1.4h
783 mul v6.4s, v0.4s, v28.4s783 mul v6.4s, v0.4s, v28.4s
784 784
785 sxtl v27.4s, v29.4H785 sxtl v27.4s, v29.4h
786 uxtl2 v0.4s, v1.8h786 uxtl2 v0.4s, v1.8h
787 mla v5.4s, v26.4s, v27.4s787 mla v5.4s, v26.4s, v27.4s
788 sxtl2 v28.4s, v29.8H788 sxtl2 v28.4s, v29.8h
789 uxtl v26.4s, v2.4h789 uxtl v26.4s, v2.4h
790 mla v6.4s, v0.4s, v28.4s790 mla v6.4s, v0.4s, v28.4s
791 791
792 sxtl v27.4s, v30.4H792 sxtl v27.4s, v30.4h
793 uxtl2 v0.4s, v2.8h793 uxtl2 v0.4s, v2.8h
794 mla v5.4s, v26.4s, v27.4s794 mla v5.4s, v26.4s, v27.4s
795 sxtl2 v28.4s, v30.8H795 sxtl2 v28.4s, v30.8h
796 uxtl v26.4s, v3.4h796 uxtl v26.4s, v3.4h
797 mla v6.4s, v0.4s, v28.4s797 mla v6.4s, v0.4s, v28.4s
798 798
799 sxtl v27.4s, v31.4H799 sxtl v27.4s, v31.4h
800 uxtl2 v0.4s, v3.8h800 uxtl2 v0.4s, v3.8h
801 mla v5.4s, v26.4s, v27.4s801 mla v5.4s, v26.4s, v27.4s
802 sxtl2 v28.4s, v31.8H802 sxtl2 v28.4s, v31.8h
803 subs w2, w2, #8803 subs w2, w2, #8
804 mla v6.4s, v0.4s, v28.4s804 mla v6.4s, v0.4s, v28.4s
805 805
@@ -807,7 +807,7 @@ function ff_hscale16to15_4_neon_asm, export=1 · 1 unmodified line
807 sshl v6.4s, v6.4s, v17.4s807 sshl v6.4s, v6.4s, v17.4s
808 smin v5.4s, v5.4s, v18.4s808 smin v5.4s, v5.4s, v18.4s
809 smin v6.4s, v6.4s, v18.4s809 smin v6.4s, v6.4s, v18.4s
810 xtn v5.4h, v5.4S810 xtn v5.4h, v5.4s
811 xtn2 v5.8h, v6.4s811 xtn2 v5.8h, v6.4s
812 812
813 st1 {v5.8h}, [x1], #16813 st1 {v5.8h}, [x1], #16
@@ -826,7 +826,7 @@ function ff_hscale16to15_4_neon_asm, export=1 · 12 unmodified lines
826 uxtl v0.4s, v0.4h826 uxtl v0.4s, v0.4h
827 sxtl v31.4s, v31.4h827 sxtl v31.4s, v31.4h
828 mul v5.4s, v0.4s, v31.4s828 mul v5.4s, v0.4s, v31.4s
829 addv s0, v5.4S829 addv s0, v5.4s
830 sshl v0.4s, v0.4s, v17.4s830 sshl v0.4s, v0.4s, v17.4s
831 smin v0.4s, v0.4s, v18.4s831 smin v0.4s, v0.4s, v18.4s
832 st1 {v0.h}[0], [x1], #2832 st1 {v0.h}[0], [x1], #2
@@ -865,58 +865,58 @@ function ff_hscale16to15_X8_neon_asm, export=1 · 32 unmodified lines
865 add x12, x16, x7 // filter1 = filter0 + filterSize*2865 add x12, x16, x7 // filter1 = filter0 + filterSize*2
866 add x13, x12, x7 // filter2 = filter1 + filterSize*2866 add x13, x12, x7 // filter2 = filter1 + filterSize*2
867 add x4, x13, x7 // filter3 = filter2 + filterSize*2867 add x4, x13, x7 // filter3 = filter2 + filterSize*2
868 movi v0.2D, #0 // val sum part 1 (for dst[0])868 movi v0.2d, #0 // val sum part 1 (for dst[0])
869 movi v1.2D, #0 // val sum part 2 (for dst[1])869 movi v1.2d, #0 // val sum part 2 (for dst[1])
870 movi v2.2D, #0 // val sum part 3 (for dst[2])870 movi v2.2d, #0 // val sum part 3 (for dst[2])
871 movi v3.2D, #0 // val sum part 4 (for dst[3])871 movi v3.2d, #0 // val sum part 4 (for dst[3])
872 add x17, x3, w8, UXTW // srcp + filterPos[0]872 add x17, x3, w8, UXTW // srcp + filterPos[0]
873 add x8, x3, w10, UXTW // srcp + filterPos[1]873 add x8, x3, w10, UXTW // srcp + filterPos[1]
874 add x10, x3, w11, UXTW // srcp + filterPos[2]874 add x10, x3, w11, UXTW // srcp + filterPos[2]
875 add x11, x3, w9, UXTW // srcp + filterPos[3]875 add x11, x3, w9, UXTW // srcp + filterPos[3]
876 mov w15, w6 // filterSize counter876 mov w15, w6 // filterSize counter
8772: ld1 {v4.8H}, [x17], #16 // srcp[filterPos[0] + {0..7}]8772: ld1 {v4.8h}, [x17], #16 // srcp[filterPos[0] + {0..7}]
878 ld1 {v5.8H}, [x16], #16 // load 8x16-bit filter values, part 1878 ld1 {v5.8h}, [x16], #16 // load 8x16-bit filter values, part 1
879 ld1 {v6.8H}, [x8], #16 // srcp[filterPos[1] + {0..7}]879 ld1 {v6.8h}, [x8], #16 // srcp[filterPos[1] + {0..7}]
880 ld1 {v7.8H}, [x12], #16 // load 8x16-bit at filter+filterSize880 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 sign881 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 size882 sxtl v25.4s, v5.4h // extend filter lower half to 32 bits to match srcp size
883 uxtl2 v4.4s, v4.8h // extend srcp upper half to 32 bits883 uxtl2 v4.4s, v4.8h // extend srcp upper half to 32 bits
884 mla v0.4s, v24.4s, v25.4s // multiply accumulate lower half of v4 * v5884 mla v0.4s, v24.4s, v25.4s // multiply accumulate lower half of v4 * v5
885 sxtl2 v5.4s, v5.8h // extend filter upper half to 32 bits885 sxtl2 v5.4s, v5.8h // extend filter upper half to 32 bits
886 uxtl v26.4s, v6.4h // extend srcp lower half to 32 bits886 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 * v5887 mla v0.4s, v4.4s, v5.4s // multiply accumulate upper half of v4 * v5
888 sxtl v27.4s, v7.4H // exted filter lower half888 sxtl v27.4s, v7.4h // exted filter lower half
889 uxtl2 v6.4s, v6.8H // extend srcp upper half889 uxtl2 v6.4s, v6.8h // extend srcp upper half
890 sxtl2 v7.4s, v7.8h // extend filter upper half890 sxtl2 v7.4s, v7.8h // extend filter upper half
891 ld1 {v16.8H}, [x10], #16 // srcp[filterPos[2] + {0..7}]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}]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*filterSize893 ld1 {v17.8h}, [x13], #16 // load 8x16-bit at filter+2*filterSize
894 uxtl v22.4s, v16.4H // extend srcp lower half894 uxtl v22.4s, v16.4h // extend srcp lower half
895 sxtl v23.4s, v17.4H // extend filter lower half895 sxtl v23.4s, v17.4h // extend filter lower half
896 uxtl2 v16.4s, v16.8H // extend srcp upper half896 uxtl2 v16.4s, v16.8h // extend srcp upper half
897 sxtl2 v17.4s, v17.8h // extend filter upper half897 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}]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}]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}]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}]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*filterSize902 ld1 {v19.8h}, [x4], #16 // load 8x16-bit at filter+3*filterSize
903 subs w15, w15, #8 // j -= 8: processed 8/filterSize903 subs w15, w15, #8 // j -= 8: processed 8/filterSize
904 uxtl v28.4s, v18.4H // extend srcp lower half904 uxtl v28.4s, v18.4h // extend srcp lower half
905 sxtl v29.4s, v19.4H // extend filter lower half905 sxtl v29.4s, v19.4h // extend filter lower half
906 uxtl2 v18.4s, v18.8H // extend srcp upper half906 uxtl2 v18.4s, v18.8h // extend srcp upper half
907 sxtl2 v19.4s, v19.8h // extend filter upper half907 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}]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}]909 mla v3.4s, v18.4s, v19.4s // v3 accumulates srcp[filterPos[3] + {4..7}] * filter[{4..7}]
910 b.gt 2b // inner loop if filterSize not consumed completely910 b.gt 2b // inner loop if filterSize not consumed completely
911 addp v0.4S, v0.4S, v1.4S // part01 horizontal pair adding911 addp v0.4s, v0.4s, v1.4s // part01 horizontal pair adding
912 addp v2.4S, v2.4S, v3.4S // part23 horizontal pair adding912 addp v2.4s, v2.4s, v3.4s // part23 horizontal pair adding
913 addp v0.4S, v0.4S, v2.4S // part0123 horizontal pair adding913 addp v0.4s, v0.4s, v2.4s // part0123 horizontal pair adding
914 subs w2, w2, #4 // dstW -= 4914 subs w2, w2, #4 // dstW -= 4
915 sshl v0.4s, v0.4s, v21.4s // shift right (effectively rigth, as shift is negative); overflow expected915 sshl v0.4s, v0.4s, v21.4s // shift right (effectively rigth, as shift is negative); overflow expected
916 smin v0.4s, v0.4s, v20.4s // apply min (do not use sqshl)916 smin v0.4s, v0.4s, v20.4s // apply min (do not use sqshl)
917 xtn v0.4h, v0.4s // narrow down to 16 bits917 xtn v0.4h, v0.4s // narrow down to 16 bits
918 918
919 st1 {v0.4H}, [x1], #8 // write to destination part0123919 st1 {v0.4h}, [x1], #8 // write to destination part0123
920 b.gt 1b // loop until end of line920 b.gt 1b // loop until end of line
921 ret921 ret
922endfunc922endfunc
@@ -1108,31 +1108,31 @@ function ff_hscale16to19_4_neon_asm, export=1 · 185 unmodified lines
1108 // Extending to 32 bits is necessary, as unit16_t values can't1108 // Extending to 32 bits is necessary, as unit16_t values can't
1109 // be represented as int16_t without type promotion.1109 // be represented as int16_t without type promotion.
1110 uxtl v26.4s, v0.4h1110 uxtl v26.4s, v0.4h
1111 sxtl v27.4s, v28.4H1111 sxtl v27.4s, v28.4h
1112 uxtl2 v0.4s, v0.8h1112 uxtl2 v0.4s, v0.8h
1113 mul v5.4s, v26.4s, v27.4s1113 mul v5.4s, v26.4s, v27.4s
1114 sxtl2 v28.4s, v28.8H1114 sxtl2 v28.4s, v28.8h
1115 uxtl v26.4s, v1.4h1115 uxtl v26.4s, v1.4h
1116 mul v6.4s, v0.4s, v28.4s1116 mul v6.4s, v0.4s, v28.4s
1117 1117
1118 sxtl v27.4s, v29.4H1118 sxtl v27.4s, v29.4h
1119 uxtl2 v0.4s, v1.8h1119 uxtl2 v0.4s, v1.8h
1120 mla v5.4s, v27.4s, v26.4s1120 mla v5.4s, v27.4s, v26.4s
1121 sxtl2 v28.4s, v29.8H1121 sxtl2 v28.4s, v29.8h
1122 uxtl v26.4s, v2.4h1122 uxtl v26.4s, v2.4h
1123 mla v6.4s, v28.4s, v0.4s1123 mla v6.4s, v28.4s, v0.4s
1124 1124
1125 sxtl v27.4s, v30.4H1125 sxtl v27.4s, v30.4h
1126 uxtl2 v0.4s, v2.8h1126 uxtl2 v0.4s, v2.8h
1127 mla v5.4s, v27.4s, v26.4s1127 mla v5.4s, v27.4s, v26.4s
1128 sxtl2 v28.4s, v30.8H1128 sxtl2 v28.4s, v30.8h
1129 uxtl v26.4s, v3.4h1129 uxtl v26.4s, v3.4h
1130 mla v6.4s, v28.4s, v0.4s1130 mla v6.4s, v28.4s, v0.4s
1131 1131
1132 sxtl v27.4s, v31.4H1132 sxtl v27.4s, v31.4h
1133 uxtl2 v0.4s, v3.8h1133 uxtl2 v0.4s, v3.8h
1134 mla v5.4s, v27.4s, v26.4s1134 mla v5.4s, v27.4s, v26.4s
1135 sxtl2 v28.4s, v31.8H1135 sxtl2 v28.4s, v31.8h
1136 sub w2, w2, #81136 sub w2, w2, #8
1137 mla v6.4s, v28.4s, v0.4s1137 mla v6.4s, v28.4s, v0.4s
1138 1138
@@ -1181,31 +1181,31 @@ function ff_hscale16to19_4_neon_asm, export=1 · 42 unmodified lines
1181 ld4 {v28.8h, v29.8h, v30.8h, v31.8h}, [x4], #641181 ld4 {v28.8h, v29.8h, v30.8h, v31.8h}, [x4], #64
1182 1182
1183 uxtl v26.4s, v0.4h1183 uxtl v26.4s, v0.4h
1184 sxtl v27.4s, v28.4H1184 sxtl v27.4s, v28.4h
1185 uxtl2 v0.4s, v0.8h1185 uxtl2 v0.4s, v0.8h
1186 mul v5.4s, v26.4s, v27.4s1186 mul v5.4s, v26.4s, v27.4s
1187 sxtl2 v28.4s, v28.8H1187 sxtl2 v28.4s, v28.8h
1188 uxtl v26.4s, v1.4h1188 uxtl v26.4s, v1.4h
1189 mul v6.4s, v0.4s, v28.4s1189 mul v6.4s, v0.4s, v28.4s
1190 1190
1191 sxtl v27.4s, v29.4H1191 sxtl v27.4s, v29.4h
1192 uxtl2 v0.4s, v1.8h1192 uxtl2 v0.4s, v1.8h
1193 mla v5.4s, v26.4s, v27.4s1193 mla v5.4s, v26.4s, v27.4s
1194 sxtl2 v28.4s, v29.8H1194 sxtl2 v28.4s, v29.8h
1195 uxtl v26.4s, v2.4h1195 uxtl v26.4s, v2.4h
1196 mla v6.4s, v0.4s, v28.4s1196 mla v6.4s, v0.4s, v28.4s
1197 1197
1198 sxtl v27.4s, v30.4H1198 sxtl v27.4s, v30.4h
1199 uxtl2 v0.4s, v2.8h1199 uxtl2 v0.4s, v2.8h
1200 mla v5.4s, v26.4s, v27.4s1200 mla v5.4s, v26.4s, v27.4s
1201 sxtl2 v28.4s, v30.8H1201 sxtl2 v28.4s, v30.8h
1202 uxtl v26.4s, v3.4h1202 uxtl v26.4s, v3.4h
1203 mla v6.4s, v0.4s, v28.4s1203 mla v6.4s, v0.4s, v28.4s
1204 1204
1205 sxtl v27.4s, v31.4H1205 sxtl v27.4s, v31.4h
1206 uxtl2 v0.4s, v3.8h1206 uxtl2 v0.4s, v3.8h
1207 mla v5.4s, v26.4s, v27.4s1207 mla v5.4s, v26.4s, v27.4s
1208 sxtl2 v28.4s, v31.8H1208 sxtl2 v28.4s, v31.8h
1209 subs w2, w2, #81209 subs w2, w2, #8
1210 mla v6.4s, v0.4s, v28.4s1210 mla v6.4s, v0.4s, v28.4s
1211 1211
@@ -1232,7 +1232,7 @@ function ff_hscale16to19_4_neon_asm, export=1 · 20 unmodified lines
1232 sxtl v31.4s, v31.4h1232 sxtl v31.4s, v31.4h
1233 subs w2, w2, #11233 subs w2, w2, #1
1234 mul v5.4s, v0.4s, v31.4s1234 mul v5.4s, v0.4s, v31.4s
1235 addv s0, v5.4S1235 addv s0, v5.4s
1236 sshl v0.4s, v0.4s, v17.4s1236 sshl v0.4s, v0.4s, v17.4s
1237 smin v0.4s, v0.4s, v18.4s1237 smin v0.4s, v0.4s, v18.4s
1238 st1 {v0.s}[0], [x1], #41238 st1 {v0.s}[0], [x1], #4
@@ -1270,52 +1270,52 @@ function ff_hscale16to19_X8_neon_asm, export=1 · 31 unmodified lines
1270 add x13, x12, x7 // filter2 = filter1 + filterSize*21270 add x13, x12, x7 // filter2 = filter1 + filterSize*2
1271 lsl w10, w10, #11271 lsl w10, w10, #1
1272 add x4, x13, x7 // filter3 = filter2 + filterSize*21272 add x4, x13, x7 // filter3 = filter2 + filterSize*2
1273 movi v0.2D, #0 // val sum part 1 (for dst[0])1273 movi v0.2d, #0 // val sum part 1 (for dst[0])
1274 movi v1.2D, #0 // val sum part 2 (for dst[1])1274 movi v1.2d, #0 // val sum part 2 (for dst[1])
1275 movi v2.2D, #0 // val sum part 3 (for dst[2])1275 movi v2.2d, #0 // val sum part 3 (for dst[2])
1276 movi v3.2D, #0 // val sum part 4 (for dst[3])1276 movi v3.2d, #0 // val sum part 4 (for dst[3])
1277 add x17, x3, w8, UXTW // srcp + filterPos[0]1277 add x17, x3, w8, UXTW // srcp + filterPos[0]
1278 add x8, x3, w10, UXTW // srcp + filterPos[1]1278 add x8, x3, w10, UXTW // srcp + filterPos[1]
1279 add x10, x3, w11, UXTW // srcp + filterPos[2]1279 add x10, x3, w11, UXTW // srcp + filterPos[2]
1280 add x11, x3, w9, UXTW // srcp + filterPos[3]1280 add x11, x3, w9, UXTW // srcp + filterPos[3]
1281 mov w15, w6 // filterSize counter1281 mov w15, w6 // filterSize counter
12822: ld1 {v4.8H}, [x17], #16 // srcp[filterPos[0] + {0..7}]12822: ld1 {v4.8h}, [x17], #16 // srcp[filterPos[0] + {0..7}]
1283 ld1 {v5.8H}, [x16], #16 // load 8x16-bit filter values, part 11283 ld1 {v5.8h}, [x16], #16 // load 8x16-bit filter values, part 1
1284 ld1 {v6.8H}, [x8], #16 // srcp[filterPos[1] + {0..7}]1284 ld1 {v6.8h}, [x8], #16 // srcp[filterPos[1] + {0..7}]
1285 ld1 {v7.8H}, [x12], #16 // load 8x16-bit at filter+filterSize1285 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 sign1286 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 size1287 sxtl v25.4s, v5.4h // extend filter lower half to 32 bits to match srcp size
1288 uxtl2 v4.4s, v4.8h // extend srcp upper half to 32 bits1288 uxtl2 v4.4s, v4.8h // extend srcp upper half to 32 bits
1289 mla v0.4s, v24.4s, v25.4s // multiply accumulate lower half of v4 * v51289 mla v0.4s, v24.4s, v25.4s // multiply accumulate lower half of v4 * v5
1290 sxtl2 v5.4s, v5.8h // extend filter upper half to 32 bits1290 sxtl2 v5.4s, v5.8h // extend filter upper half to 32 bits
1291 uxtl v26.4s, v6.4h // extend srcp lower half to 32 bits1291 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 * v51292 mla v0.4s, v4.4s, v5.4s // multiply accumulate upper half of v4 * v5
1293 sxtl v27.4s, v7.4H // exted filter lower half1293 sxtl v27.4s, v7.4h // exted filter lower half
1294 uxtl2 v6.4s, v6.8H // extend srcp upper half1294 uxtl2 v6.4s, v6.8h // extend srcp upper half
1295 sxtl2 v7.4s, v7.8h // extend filter upper half1295 sxtl2 v7.4s, v7.8h // extend filter upper half
1296 ld1 {v16.8H}, [x10], #16 // srcp[filterPos[2] + {0..7}]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}]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*filterSize1298 ld1 {v17.8h}, [x13], #16 // load 8x16-bit at filter+2*filterSize
1299 uxtl v22.4s, v16.4H // extend srcp lower half1299 uxtl v22.4s, v16.4h // extend srcp lower half
1300 sxtl v23.4s, v17.4H // extend filter lower half1300 sxtl v23.4s, v17.4h // extend filter lower half
1301 uxtl2 v16.4s, v16.8H // extend srcp upper half1301 uxtl2 v16.4s, v16.8h // extend srcp upper half
1302 sxtl2 v17.4s, v17.8h // extend filter upper half1302 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}]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}]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}]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}]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*filterSize1307 ld1 {v19.8h}, [x4], #16 // load 8x16-bit at filter+3*filterSize
1308 subs w15, w15, #8 // j -= 8: processed 8/filterSize1308 subs w15, w15, #8 // j -= 8: processed 8/filterSize
1309 uxtl v28.4s, v18.4H // extend srcp lower half1309 uxtl v28.4s, v18.4h // extend srcp lower half
1310 sxtl v29.4s, v19.4H // extend filter lower half1310 sxtl v29.4s, v19.4h // extend filter lower half
1311 uxtl2 v18.4s, v18.8H // extend srcp upper half1311 uxtl2 v18.4s, v18.8h // extend srcp upper half
1312 sxtl2 v19.4s, v19.8h // extend filter upper half1312 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}]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}]1314 mla v3.4s, v18.4s, v19.4s // v3 accumulates srcp[filterPos[3] + {4..7}] * filter[{4..7}]
1315 b.gt 2b // inner loop if filterSize not consumed completely1315 b.gt 2b // inner loop if filterSize not consumed completely
1316 addp v0.4S, v0.4S, v1.4S // part01 horizontal pair adding1316 addp v0.4s, v0.4s, v1.4s // part01 horizontal pair adding
1317 addp v2.4S, v2.4S, v3.4S // part23 horizontal pair adding1317 addp v2.4s, v2.4s, v3.4s // part23 horizontal pair adding
1318 addp v0.4S, v0.4S, v2.4S // part0123 horizontal pair adding1318 addp v0.4s, v0.4s, v2.4s // part0123 horizontal pair adding
1319 subs w2, w2, #4 // dstW -= 41319 subs w2, w2, #4 // dstW -= 4
1320 sshl v0.4s, v0.4s, v21.4s // shift right (effectively rigth, as shift is negative); overflow expected1320 sshl v0.4s, v0.4s, v21.4s // shift right (effectively rigth, as shift is negative); overflow expected
1321 smin v0.4s, v0.4s, v20.4s // apply min (do not use sqshl)1321 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
29// x5 - const uint8_t *dither,29// x5 - const uint8_t *dither,
30// w6 - int offset30// w6 - int offset
31 31
32 ld1 {v0.8B}, [x5] // load 8x8-bit dither32 ld1 {v0.8b}, [x5] // load 8x8-bit dither
33 and w6, w6, #733 and w6, w6, #7
34 cbz w6, 1f // check if offsetting present34 cbz w6, 1f // check if offsetting present
35 ext v0.8B, v0.8B, v0.8B, #3 // honor offsetting which can be 0 or 3 only35 ext v0.8b, v0.8b, v0.8b, #3 // honor offsetting which can be 0 or 3 only
361: uxtl v0.8H, v0.8B // extend dither to 16-bit361: 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)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)38 ushll2 v2.4s, v0.8h, #12 // extend dither to 32-bit with left shift by 12 (part 2)
39 cmp w1, #8 // if filterSize == 8, branch to specialized version39 cmp w1, #8 // if filterSize == 8, branch to specialized version
40 b.eq 6f40 b.eq 6f
41 cmp w1, #4 // if filterSize == 4, branch to specialized version41 cmp w1, #4 // if filterSize == 4, branch to specialized version
@@ -48,8 +48,8 @@ function ff_yuv2planeX_8_neon, export=1 · 6 unmodified lines
48 mov x7, #0 // i = 048 mov x7, #0 // i = 0
49 tbnz w1, #0, 4f // if filterSize % 2 != 0 branch to specialized version49 tbnz w1, #0, 4f // if filterSize % 2 != 0 branch to specialized version
50// fs % 2 == 050// fs % 2 == 0
512: mov v3.16B, v1.16B // initialize accumulator part 1 with dithering value512: mov v3.16b, v1.16b // initialize accumulator part 1 with dithering value
52 mov v4.16B, v2.16B // initialize accumulator part 2 with dithering value52 mov v4.16b, v2.16b // initialize accumulator part 2 with dithering value
53 mov w8, w1 // tmpfilterSize = filterSize53 mov w8, w1 // tmpfilterSize = filterSize
54 mov x9, x2 // srcp = src54 mov x9, x2 // srcp = src
55 mov x10, x0 // filterp = filter55 mov x10, x0 // filterp = filter
@@ -57,12 +57,12 @@ function ff_yuv2planeX_8_neon, export=1 · 1 unmodified line
57 ldr s7, [x10], #4 // read 2x16-bit coeff X and Y at filter[j] and filter[j+1]57 ldr s7, [x10], #4 // read 2x16-bit coeff X and Y at filter[j] and filter[j+1]
58 add x11, x11, x7, lsl #1 // &src[j ][i]58 add x11, x11, x7, lsl #1 // &src[j ][i]
59 add x12, x12, x7, lsl #1 // &src[j+1][i]59 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,H60 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,P61 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} * X62 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} * X63 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} * Y64 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} * Y65 smlal2 v4.4s, v6.8h, v7.h[1] // val1 += {M,N,O,P} * Y
66 subs w8, w8, #2 // tmpfilterSize -= 266 subs w8, w8, #2 // tmpfilterSize -= 2
67 b.gt 3b // loop until filterSize consumed67 b.gt 3b // loop until filterSize consumed
68 68
@@ -77,17 +77,17 @@ function ff_yuv2planeX_8_neon, export=1 · 8 unmodified lines
77 77
78// If filter size is odd (most likely == 1), then use this section.78// If filter size is odd (most likely == 1), then use this section.
79// fs % 2 != 079// fs % 2 != 0
804: mov v3.16B, v1.16B // initialize accumulator part 1 with dithering value804: mov v3.16b, v1.16b // initialize accumulator part 1 with dithering value
81 mov v4.16B, v2.16B // initialize accumulator part 2 with dithering value81 mov v4.16b, v2.16b // initialize accumulator part 2 with dithering value
82 mov w8, w1 // tmpfilterSize = filterSize82 mov w8, w1 // tmpfilterSize = filterSize
83 mov x9, x2 // srcp = src83 mov x9, x2 // srcp = src
84 mov x10, x0 // filterp = filter84 mov x10, x0 // filterp = filter
855: ldr x11, [x9], #8 // get 1 pointer: src[j]855: ldr x11, [x9], #8 // get 1 pointer: src[j]
86 ldr h6, [x10], #2 // read 1 16 bit coeff X at filter[j]86 ldr h6, [x10], #2 // read 1 16 bit coeff X at filter[j]
87 add x11, x11, x7, lsl #1 // &src[j ][i]87 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,H88 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} * X89 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} * X90 smlal2 v4.4s, v5.8h, v6.h[0] // val1 += {E,F,G,H} * X
91 subs w8, w8, #1 // tmpfilterSize -= 291 subs w8, w8, #1 // tmpfilterSize -= 2
92 b.gt 5b // loop until filterSize consumed92 b.gt 5b // loop until filterSize consumed
93 93
@@ -107,36 +107,36 @@ function ff_yuv2planeX_8_neon, export=1 · 13 unmodified lines
107 ldp x12, x13, [x2, #48] // load 2 pointers: src[j+6] and src[j+7]107 ldp x12, x13, [x2, #48] // load 2 pointers: src[j+6] and src[j+7]
108 108
109 // load 8x16-bit values for filter[j], where j=0..7109 // load 8x16-bit values for filter[j], where j=0..7
110 ld1 {v6.8H}, [x0]110 ld1 {v6.8h}, [x0]
1117:1117:
112 mov v3.16B, v1.16B // initialize accumulator part 1 with dithering value112 mov v3.16b, v1.16b // initialize accumulator part 1 with dithering value
113 mov v4.16B, v2.16B // initialize accumulator part 2 with dithering value113 mov v4.16b, v2.16b // initialize accumulator part 2 with dithering value
114 114
115 ld1 {v24.8H}, [x5], #16 // load 8x16-bit values for src[j + 0][i + {0..7}]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}]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}]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}]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}]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}]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}]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}]122 ld1 {v31.8h}, [x13], #16 // load 8x16-bit values for src[j + 7][i + {0..7}]
123 123
124 smlal v3.4S, v24.4H, v6.H[0] // val0 += src[0][i + {0..3}] * filter[0]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]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]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]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]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]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]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]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]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]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]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]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]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]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]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]139 smlal2 v4.4s, v31.8h, v6.h[7] // val1 += src[7][i + {4..7}] * filter[7]
140 140
141 sqshrun v3.4h, v3.4s, #16 // clip16(val0>>16)141 sqshrun v3.4h, v3.4s, #16 // clip16(val0>>16)
142 sqshrun2 v3.8h, v4.4s, #16 // clip16(val1>>16)142 sqshrun2 v3.8h, v4.4s, #16 // clip16(val1>>16)
@@ -151,24 +151,24 @@ function ff_yuv2planeX_8_neon, export=1 · 8 unmodified lines
151 ldp x7, x9, [x2, #16] // load 2 pointers: src[j+2] and src[j+3]151 ldp x7, x9, [x2, #16] // load 2 pointers: src[j+2] and src[j+3]
152 152
153 // load 4x16-bit values for filter[j], where j=0..3 and replicated across lanes153 // 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]
1559:1559:
156 mov v3.16B, v1.16B // initialize accumulator part 1 with dithering value156 mov v3.16b, v1.16b // initialize accumulator part 1 with dithering value
157 mov v4.16B, v2.16B // initialize accumulator part 2 with dithering value157 mov v4.16b, v2.16b // initialize accumulator part 2 with dithering value
158 158
159 ld1 {v24.8H}, [x5], #16 // load 8x16-bit values for src[j + 0][i + {0..7}]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}]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}]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}]162 ld1 {v27.8h}, [x9], #16 // load 8x16-bit values for src[j + 3][i + {0..7}]
163 163
164 smlal v3.4S, v24.4H, v6.H[0] // val0 += src[0][i + {0..3}] * filter[0]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]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]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]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]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]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]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]171 smlal2 v4.4s, v27.8h, v6.h[3] // val1 += src[3][i + {4..7}] * filter[3]
172 172
173 sqshrun v3.4h, v3.4s, #16 // clip16(val0>>16)173 sqshrun v3.4h, v3.4s, #16 // clip16(val0>>16)
174 sqshrun2 v3.8h, v4.4s, #16 // clip16(val1>>16)174 sqshrun2 v3.8h, v4.4s, #16 // clip16(val1>>16)
@@ -184,16 +184,16 @@ function ff_yuv2planeX_8_neon, export=1 · 9 unmodified lines
184 // load 2x16-bit values for filter[j], where j=0..1 and replicated across lanes184 // load 2x16-bit values for filter[j], where j=0..1 and replicated across lanes
185 ldr s6, [x0]185 ldr s6, [x0]
18611:18611:
187 mov v3.16B, v1.16B // initialize accumulator part 1 with dithering value187 mov v3.16b, v1.16b // initialize accumulator part 1 with dithering value
188 mov v4.16B, v2.16B // initialize accumulator part 2 with dithering value188 mov v4.16b, v2.16b // initialize accumulator part 2 with dithering value
189 189
190 ld1 {v24.8H}, [x5], #16 // load 8x16-bit values for src[j + 0][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}]191 ld1 {v25.8h}, [x6], #16 // load 8x16-bit values for src[j + 1][i + {0..7}]
192 192
193 smlal v3.4S, v24.4H, v6.H[0] // val0 += src[0][i + {0..3}] * filter[0]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]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]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]196 smlal2 v4.4s, v25.8h, v6.h[1] // val1 += src[1][i + {4..7}] * filter[1]
197 197
198 sqshrun v3.4h, v3.4s, #16 // clip16(val0>>16)198 sqshrun v3.4h, v3.4s, #16 // clip16(val0>>16)
199 sqshrun2 v3.8h, v4.4s, #16 // clip16(val1>>16)199 sqshrun2 v3.8h, v4.4s, #16 // clip16(val1>>16)
@@ -210,11 +210,11 @@ function ff_yuv2plane1_8_neon, export=1 · 10 unmodified lines
210// w2 - int dstW,210// w2 - int dstW,
211// x3 - const uint8_t *dither,211// x3 - const uint8_t *dither,
212// w4 - int offset212// w4 - int offset
213 ld1 {v0.8B}, [x3] // load 8x8-bit dither213 ld1 {v0.8b}, [x3] // load 8x8-bit dither
214 and w4, w4, #7214 and w4, w4, #7
215 cbz w4, 1f // check if offsetting present215 cbz w4, 1f // check if offsetting present
216 ext v0.8B, v0.8B, v0.8B, #3 // honor offsetting which can be 0 or 3 only216 ext v0.8b, v0.8b, v0.8b, #3 // honor offsetting which can be 0 or 3 only
2171: uxtl v0.8H, v0.8B // extend dither to 32-bit2171: uxtl v0.8h, v0.8b // extend dither to 32-bit
218 uxtl v1.4s, v0.4h218 uxtl v1.4s, v0.4h
219 uxtl2 v2.4s, v0.8h219 uxtl2 v2.4s, v0.8h
2202:2202:
libswscale/aarch64/yuv2rgb_neon.S+58−58
@@ -33,9 +33,9 @@ · 32 unmodified lines
33.macro load_args_nv1233.macro load_args_nv12
34 ldr x8, [sp] // table34 ldr x8, [sp] // table
35 load_yoff_ycoeff 8, 16 // y_offset, y_coeff35 load_yoff_ycoeff 8, 16 // y_offset, y_coeff
36 ld1 {v1.1D}, [x8]36 ld1 {v1.1d}, [x8]
37 dup v0.8H, w1037 dup v0.8h, w10
38 dup v3.8H, w938 dup v3.8h, w9
39 sub w3, w3, w0, lsl #2 // w3 = linesize - width * 4 (padding)39 sub w3, w3, w0, lsl #2 // w3 = linesize - width * 4 (padding)
40 sub w5, w5, w0 // w5 = linesizeY - width (paddingY)40 sub w5, w5, w0 // w5 = linesizeY - width (paddingY)
41 sub w7, w7, w0 // w7 = linesizeC - width (paddingC)41 sub w7, w7, w0 // w7 = linesizeC - width (paddingC)
@@ -51,9 +51,9 @@ · 9 unmodified lines
51 ldr w14, [sp, #8] // linesizeV51 ldr w14, [sp, #8] // linesizeV
52 ldr x8, [sp, #16] // table52 ldr x8, [sp, #16] // table
53 load_yoff_ycoeff 24, 32 // y_offset, y_coeff53 load_yoff_ycoeff 24, 32 // y_offset, y_coeff
54 ld1 {v1.1D}, [x8]54 ld1 {v1.1d}, [x8]
55 dup v0.8H, w1055 dup v0.8h, w10
56 dup v3.8H, w956 dup v3.8h, w9
57 sub w3, w3, w0, lsl #2 // w3 = linesize - width * 4 (padding)57 sub w3, w3, w0, lsl #2 // w3 = linesize - width * 4 (padding)
58 sub w5, w5, w0 // w5 = linesizeY - width (paddingY)58 sub w5, w5, w0 // w5 = linesizeY - width (paddingY)
59 sub w7, w7, w0, lsr #1 // w7 = linesizeU - width / 2 (paddingU)59 sub w7, w7, w0, lsr #1 // w7 = linesizeU - width / 2 (paddingU)
@@ -67,9 +67,9 @@ · 7 unmodified lines
67 ldr w14, [sp, #8] // linesizeV67 ldr w14, [sp, #8] // linesizeV
68 ldr x8, [sp, #16] // table68 ldr x8, [sp, #16] // table
69 load_yoff_ycoeff 24, 32 // y_offset, y_coeff69 load_yoff_ycoeff 24, 32 // y_offset, y_coeff
70 ld1 {v1.1D}, [x8]70 ld1 {v1.1d}, [x8]
71 dup v0.8H, w1071 dup v0.8h, w10
72 dup v3.8H, w972 dup v3.8h, w9
73 sub w3, w3, w0, lsl #2 // w3 = linesize - width * 4 (padding)73 sub w3, w3, w0, lsl #2 // w3 = linesize - width * 4 (padding)
74 sub w5, w5, w0 // w5 = linesizeY - width (paddingY)74 sub w5, w5, w0 // w5 = linesizeY - width (paddingY)
75 sub w7, w7, w0, lsr #1 // w7 = linesizeU - width / 2 (paddingU)75 sub w7, w7, w0, lsr #1 // w7 = linesizeU - width / 2 (paddingU)
@@ -77,22 +77,22 @@ · 1 unmodified line
77.endm77.endm
78 78
79.macro load_chroma_nv1279.macro load_chroma_nv12
80 ld2 {v16.8B, v17.8B}, [x6], #1680 ld2 {v16.8b, v17.8b}, [x6], #16
81 ushll v18.8H, v16.8B, #381 ushll v18.8h, v16.8b, #3
82 ushll v19.8H, v17.8B, #382 ushll v19.8h, v17.8b, #3
83.endm83.endm
84 84
85.macro load_chroma_nv2185.macro load_chroma_nv21
86 ld2 {v16.8B, v17.8B}, [x6], #1686 ld2 {v16.8b, v17.8b}, [x6], #16
87 ushll v19.8H, v16.8B, #387 ushll v19.8h, v16.8b, #3
88 ushll v18.8H, v17.8B, #388 ushll v18.8h, v17.8b, #3
89.endm89.endm
90 90
91.macro load_chroma_yuv420p91.macro load_chroma_yuv420p
92 ld1 {v16.8B}, [ x6], #892 ld1 {v16.8b}, [ x6], #8
93 ld1 {v17.8B}, [x13], #893 ld1 {v17.8b}, [x13], #8
94 ushll v18.8H, v16.8B, #394 ushll v18.8h, v16.8b, #3
95 ushll v19.8H, v17.8B, #395 ushll v19.8h, v17.8b, #3
96.endm96.endm
97 97
98.macro load_chroma_yuv422p98.macro load_chroma_yuv422p
@@ -123,18 +123,18 @@ · 24 unmodified lines
123.endm123.endm
124 124
125.macro compute_rgba r1 g1 b1 a1 r2 g2 b2 a2125.macro compute_rgba r1 g1 b1 a1 r2 g2 b2 a2
126 add v20.8H, v26.8H, v20.8H // Y1 + R1126 add v20.8h, v26.8h, v20.8h // Y1 + R1
127 add v21.8H, v27.8H, v21.8H // Y2 + R2127 add v21.8h, v27.8h, v21.8h // Y2 + R2
128 add v22.8H, v26.8H, v22.8H // Y1 + G1128 add v22.8h, v26.8h, v22.8h // Y1 + G1
129 add v23.8H, v27.8H, v23.8H // Y2 + G2129 add v23.8h, v27.8h, v23.8h // Y2 + G2
130 add v24.8H, v26.8H, v24.8H // Y1 + B1130 add v24.8h, v26.8h, v24.8h // Y1 + B1
131 add v25.8H, v27.8H, v25.8H // Y2 + B2131 add v25.8h, v27.8h, v25.8h // Y2 + B2
132 sqrshrun \r1, v20.8H, #1 // clip_u8((Y1 + R1) >> 1)132 sqrshrun \r1, v20.8h, #1 // clip_u8((Y1 + R1) >> 1)
133 sqrshrun \r2, v21.8H, #1 // clip_u8((Y2 + R1) >> 1)133 sqrshrun \r2, v21.8h, #1 // clip_u8((Y2 + R1) >> 1)
134 sqrshrun \g1, v22.8H, #1 // clip_u8((Y1 + G1) >> 1)134 sqrshrun \g1, v22.8h, #1 // clip_u8((Y1 + G1) >> 1)
135 sqrshrun \g2, v23.8H, #1 // clip_u8((Y2 + G1) >> 1)135 sqrshrun \g2, v23.8h, #1 // clip_u8((Y2 + G1) >> 1)
136 sqrshrun \b1, v24.8H, #1 // clip_u8((Y1 + B1) >> 1)136 sqrshrun \b1, v24.8h, #1 // clip_u8((Y1 + B1) >> 1)
137 sqrshrun \b2, v25.8H, #1 // clip_u8((Y2 + B1) >> 1)137 sqrshrun \b2, v25.8h, #1 // clip_u8((Y2 + B1) >> 1)
138 movi \a1, #255138 movi \a1, #255
139 movi \a2, #255139 movi \a2, #255
140.endm140.endm
@@ -146,47 +146,47 @@ function ff_\ifmt\()_to_\ofmt\()_neon, export=1 · 5 unmodified lines
1461:1461:
147 mov w8, w0 // w8 = width147 mov w8, w0 // w8 = width
1482:1482:
149 movi v5.8H, #4, lsl #8 // 128 * (1<<3)149 movi v5.8h, #4, lsl #8 // 128 * (1<<3)
150 load_chroma_\ifmt150 load_chroma_\ifmt
151 sub v18.8H, v18.8H, v5.8H // U*(1<<3) - 128*(1<<3)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)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)153 sqdmulh v20.8h, v19.8h, v1.h[0] // V * v2r (R)
154 sqdmulh v22.8H, v18.8H, v1.H[1] // U * u2g154 sqdmulh v22.8h, v18.8h, v1.h[1] // U * u2g
155 sqdmulh v19.8H, v19.8H, v1.H[2] // V * v2g155 sqdmulh v19.8h, v19.8h, v1.h[2] // V * v2g
156 add v22.8H, v22.8H, v19.8H // U * u2g + V * v2g (G)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)157 sqdmulh v24.8h, v18.8h, v1.h[3] // U * u2b (B)
158 zip2 v21.8H, v20.8H, v20.8H // R2158 zip2 v21.8h, v20.8h, v20.8h // R2
159 zip1 v20.8H, v20.8H, v20.8H // R1159 zip1 v20.8h, v20.8h, v20.8h // R1
160 zip2 v23.8H, v22.8H, v22.8H // G2160 zip2 v23.8h, v22.8h, v22.8h // G2
161 zip1 v22.8H, v22.8H, v22.8H // G1161 zip1 v22.8h, v22.8h, v22.8h // G1
162 zip2 v25.8H, v24.8H, v24.8H // B2162 zip2 v25.8h, v24.8h, v24.8h // B2
163 zip1 v24.8H, v24.8H, v24.8H // B1163 zip1 v24.8h, v24.8h, v24.8h // B1
164 ld1 {v2.16B}, [x4], #16 // load luma164 ld1 {v2.16b}, [x4], #16 // load luma
165 ushll v26.8H, v2.8B, #3 // Y1*(1<<3)165 ushll v26.8h, v2.8b, #3 // Y1*(1<<3)
166 ushll2 v27.8H, v2.16B, #3 // Y2*(1<<3)166 ushll2 v27.8h, v2.16b, #3 // Y2*(1<<3)
167 sub v26.8H, v26.8H, v3.8H // Y1*(1<<3) - y_offset167 sub v26.8h, v26.8h, v3.8h // Y1*(1<<3) - y_offset
168 sub v27.8H, v27.8H, v3.8H // Y2*(1<<3) - y_offset168 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) >> 15169 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) >> 15170 sqdmulh v27.8h, v27.8h, v0.8h // ((Y2*(1<<3) - y_offset) * y_coeff) >> 15
171 171
172.ifc \ofmt,argb // 1 2 3 0172.ifc \ofmt,argb // 1 2 3 0
173 compute_rgba v5.8B,v6.8B,v7.8B,v4.8B, v17.8B,v18.8B,v19.8B,v16.8B173 compute_rgba v5.8b,v6.8b,v7.8b,v4.8b, v17.8b,v18.8b,v19.8b,v16.8b
174.endif174.endif
175 175
176.ifc \ofmt,rgba // 0 1 2 3176.ifc \ofmt,rgba // 0 1 2 3
177 compute_rgba v4.8B,v5.8B,v6.8B,v7.8B, v16.8B,v17.8B,v18.8B,v19.8B177 compute_rgba v4.8b,v5.8b,v6.8b,v7.8b, v16.8b,v17.8b,v18.8b,v19.8b
178.endif178.endif
179 179
180.ifc \ofmt,abgr // 3 2 1 0180.ifc \ofmt,abgr // 3 2 1 0
181 compute_rgba v7.8B,v6.8B,v5.8B,v4.8B, v19.8B,v18.8B,v17.8B,v16.8B181 compute_rgba v7.8b,v6.8b,v5.8b,v4.8b, v19.8b,v18.8b,v17.8b,v16.8b
182.endif182.endif
183 183
184.ifc \ofmt,bgra // 2 1 0 3184.ifc \ofmt,bgra // 2 1 0 3
185 compute_rgba v6.8B,v5.8B,v4.8B,v7.8B, v18.8B,v17.8B,v16.8B,v19.8B185 compute_rgba v6.8b,v5.8b,v4.8b,v7.8b, v18.8b,v17.8b,v16.8b,v19.8b
186.endif186.endif
187 187
188 st4 { v4.8B, v5.8B, v6.8B, v7.8B}, [x2], #32188 st4 { v4.8b, v5.8b, v6.8b, v7.8b}, [x2], #32
189 st4 {v16.8B,v17.8B,v18.8B,v19.8B}, [x2], #32189 st4 {v16.8b,v17.8b,v18.8b,v19.8b}, [x2], #32
190 subs w8, w8, #16 // width -= 16190 subs w8, w8, #16 // width -= 16
191 b.gt 2b191 b.gt 2b
192 add x2, x2, w3, SXTW // dst += padding192 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