aarch64 A55: chrRangeFromJpeg8_1920_c: 28836.2 (1.00x) chrRangeFromJpeg8_1920_neon: 5312.6 (5.43x) 5313.9 (5.43x) chrRangeToJpeg8_1920_c: 44196.2 (1.00x) chrRangeToJpeg8_1920_neon: 6034.6 (7.32x) 5551.3 (7.96x) lumRangeFromJpeg8_1920_c: 15388.5 (1.00x) lumRangeFromJpeg8_1920_neon: 3150.7 (4.88x) 3152.3 (4.88x) lumRangeToJpeg8_1920_c: 23069.7 (1.00x) lumRangeToJpeg8_1920_neon: 3873.2 (5.96x) 3628.7 (6.36x) aarch64 A76: chrRangeFromJpeg8_1920_c: 6334.7 (1.00x) chrRangeFromJpeg8_1920_neon: 2264.5 (2.80x) 2344.5 (2.70x) chrRangeToJpeg8_1920_c: 11474.5 (1.00x) chrRangeToJpeg8_1920_neon: 2646.5 (4.34x) 2824.2 (4.06x) lumRangeFromJpeg8_1920_c: 4453.2 (1.00x) lumRangeFromJpeg8_1920_neon: 1104.8 (4.03x) 1104.5 (4.03x) lumRangeToJpeg8_1920_c: 6645.0 (1.00x) lumRangeToJpeg8_1920_neon: 1310.5 (5.07x) 1329.8 (5.00x)
Unified · Split
| @@ -20,12 +20,8 @@ · 19 unmodified lines | |||
| 20 | 20 | ||
| 21 | #include "libavutil/aarch64/asm.S" | 21 | #include "libavutil/aarch64/asm.S" |
| 22 | 22 | ||
| 23 | .macro lumConvertRange name, max, mult, offset, shift | 23 | .macro lumConvertRange name, fromto, mult, offset, shift |
| 24 | function ff_\name, export=1 | 24 | function ff_\name, export=1 |
| 25 | .if \max != 0 | ||
| 26 | mov w3, #\max | ||
| 27 | dup v24.8h, w3 | ||
| 28 | .endif | ||
| 29 | mov w3, #\mult | 25 | mov w3, #\mult |
| 30 | dup v25.4s, w3 | 26 | dup v25.4s, w3 |
| 31 | movz w3, #(\offset & 0xffff) | 27 | movz w3, #(\offset & 0xffff) |
| @@ -33,17 +29,19 @@ function ff_\name, export=1 · 1 unmodified line | |||
| 33 | dup v26.4s, w3 | 29 | dup v26.4s, w3 |
| 34 | 1: | 30 | 1: |
| 35 | ld1 {v0.8h}, [x0] | 31 | ld1 {v0.8h}, [x0] |
| 36 | .if \max != 0 | ||
| 37 | smin v0.8h, v0.8h, v24.8h | ||
| 38 | .endif | ||
| 39 | mov v16.16b, v26.16b | 32 | mov v16.16b, v26.16b |
| 40 | mov v18.16b, v26.16b | 33 | mov v18.16b, v26.16b |
| 41 | sxtl v20.4s, v0.4h | 34 | sxtl v20.4s, v0.4h |
| 42 | sxtl2 v22.4s, v0.8h | 35 | sxtl2 v22.4s, v0.8h |
| 43 | mla v16.4s, v20.4s, v25.4s | 36 | mla v16.4s, v20.4s, v25.4s |
| 44 | mla v18.4s, v22.4s, v25.4s | 37 | mla v18.4s, v22.4s, v25.4s |
| 38 | .ifc \fromto, To | ||
| 39 | sqshrn v0.4h, v16.4s, #\shift | ||
| 40 | sqshrn2 v0.8h, v18.4s, #\shift | ||
| 41 | .else | ||
| 45 | shrn v0.4h, v16.4s, #\shift | 42 | shrn v0.4h, v16.4s, #\shift |
| 46 | shrn2 v0.8h, v18.4s, #\shift | 43 | shrn2 v0.8h, v18.4s, #\shift |
| 44 | .endif | ||
| 47 | subs w1, w1, #8 | 45 | subs w1, w1, #8 |
| 48 | st1 {v0.8h}, [x0], #16 | 46 | st1 {v0.8h}, [x0], #16 |
| 49 | b.gt 1b | 47 | b.gt 1b |
| @@ -51,12 +49,8 @@ function ff_\name, export=1 · 1 unmodified line | |||
| 51 | endfunc | 49 | endfunc |
| 52 | .endm | 50 | .endm |
| 53 | 51 | ||
| 54 | .macro chrConvertRange name, max, mult, offset, shift | 52 | .macro chrConvertRange name, fromto, mult, offset, shift |
| 55 | function ff_\name, export=1 | 53 | function ff_\name, export=1 |
| 56 | .if \max != 0 | ||
| 57 | mov w3, #\max | ||
| 58 | dup v24.8h, w3 | ||
| 59 | .endif | ||
| 60 | mov w3, #\mult | 54 | mov w3, #\mult |
| 61 | dup v25.4s, w3 | 55 | dup v25.4s, w3 |
| 62 | movz w3, #(\offset & 0xffff) | 56 | movz w3, #(\offset & 0xffff) |
| @@ -65,10 +59,6 @@ function ff_\name, export=1 · 2 unmodified lines | |||
| 65 | 1: | 59 | 1: |
| 66 | ld1 {v0.8h}, [x0] | 60 | ld1 {v0.8h}, [x0] |
| 67 | ld1 {v1.8h}, [x1] | 61 | ld1 {v1.8h}, [x1] |
| 68 | .if \max != 0 | ||
| 69 | smin v0.8h, v0.8h, v24.8h | ||
| 70 | smin v1.8h, v1.8h, v24.8h | ||
| 71 | .endif | ||
| 72 | mov v16.16b, v26.16b | 62 | mov v16.16b, v26.16b |
| 73 | mov v17.16b, v26.16b | 63 | mov v17.16b, v26.16b |
| 74 | mov v18.16b, v26.16b | 64 | mov v18.16b, v26.16b |
| @@ -81,10 +71,17 @@ function ff_\name, export=1 · 6 unmodified lines | |||
| 81 | mla v17.4s, v21.4s, v25.4s | 71 | mla v17.4s, v21.4s, v25.4s |
| 82 | mla v18.4s, v22.4s, v25.4s | 72 | mla v18.4s, v22.4s, v25.4s |
| 83 | mla v19.4s, v23.4s, v25.4s | 73 | mla v19.4s, v23.4s, v25.4s |
| 74 | .ifc \fromto, To | ||
| 75 | sqshrn v0.4h, v16.4s, #\shift | ||
| 76 | sqshrn v1.4h, v17.4s, #\shift | ||
| 77 | sqshrn2 v0.8h, v18.4s, #\shift | ||
| 78 | sqshrn2 v1.8h, v19.4s, #\shift | ||
| 79 | .else | ||
| 84 | shrn v0.4h, v16.4s, #\shift | 80 | shrn v0.4h, v16.4s, #\shift |
| 85 | shrn v1.4h, v17.4s, #\shift | 81 | shrn v1.4h, v17.4s, #\shift |
| 86 | shrn2 v0.8h, v18.4s, #\shift | 82 | shrn2 v0.8h, v18.4s, #\shift |
| 87 | shrn2 v1.8h, v19.4s, #\shift | 83 | shrn2 v1.8h, v19.4s, #\shift |
| 84 | .endif | ||
| 88 | subs w2, w2, #8 | 85 | subs w2, w2, #8 |
| 89 | st1 {v0.8h}, [x0], #16 | 86 | st1 {v0.8h}, [x0], #16 |
| 90 | st1 {v1.8h}, [x1], #16 | 87 | st1 {v1.8h}, [x1], #16 |
| @@ -93,7 +90,7 @@ function ff_\name, export=1 · 2 unmodified lines | |||
| 93 | endfunc | 90 | endfunc |
| 94 | .endm | 91 | .endm |
| 95 | 92 | ||
| 96 | lumConvertRange lumRangeToJpeg_neon, 30189, 19077, -39057361, 14 | 93 | lumConvertRange lumRangeToJpeg_neon, To, 19077, -39057361, 14 |
| 97 | chrConvertRange chrRangeToJpeg_neon, 30775, 4663, -9289992, 12 | 94 | chrConvertRange chrRangeToJpeg_neon, To, 4663, -9289992, 12 |
| 98 | lumConvertRange lumRangeFromJpeg_neon, 0, 14071, 33561947, 14 | 95 | lumConvertRange lumRangeFromJpeg_neon, From, 14071, 33561947, 14 |
| 99 | chrConvertRange chrRangeFromJpeg_neon, 0, 1799, 4081085, 11 | 96 | chrConvertRange chrRangeFromJpeg_neon, From, 1799, 4081085, 11 |
| @@ -225,10 +225,6 @@ void ff_chrRangeToJpeg_neon(int16_t *dstU, int16_t *dstV, int width); · 224 unmodified lines | |||
| 225 | 225 | ||
| 226 | av_cold void ff_sws_init_range_convert_aarch64(SwsInternal *c) | 226 | av_cold void ff_sws_init_range_convert_aarch64(SwsInternal *c) |
| 227 | { | 227 | { |
| 228 | /* This code is currently disabled because of changes in the base | ||
| 229 | * implementation of these functions. This code should be enabled | ||
| 230 | * again once those changes are ported to this architecture. */ | ||
| 231 | #if 0 | ||
| 232 | int cpu_flags = av_get_cpu_flags(); | 228 | int cpu_flags = av_get_cpu_flags(); |
| 233 | 229 | ||
| 234 | if (have_neon(cpu_flags)) { | 230 | if (have_neon(cpu_flags)) { |
| @@ -242,7 +238,6 @@ av_cold void ff_sws_init_range_convert_aarch64(SwsInternal *c) · 7 unmodified lines | |||
| 242 | } | 238 | } |
| 243 | } | 239 | } |
| 244 | } | 240 | } |
| 245 | #endif | ||
| 246 | } | 241 | } |
| 247 | 242 | ||
| 248 | av_cold void ff_sws_init_swscale_aarch64(SwsInternal *c) | 243 | av_cold void ff_sws_init_swscale_aarch64(SwsInternal *c) |