123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630 |
- ;
- ; Copyright (c) 2014 The WebM project authors. All Rights Reserved.
- ;
- ; Use of this source code is governed by a BSD-style license
- ; that can be found in the LICENSE file in the root of the source
- ; tree. An additional intellectual property rights grant can be found
- ; in the file PATENTS. All contributing project authors may
- ; be found in the AUTHORS file in the root of the source tree.
- ;
- EXPORT |vpx_v_predictor_4x4_neon|
- EXPORT |vpx_v_predictor_8x8_neon|
- EXPORT |vpx_v_predictor_16x16_neon|
- EXPORT |vpx_v_predictor_32x32_neon|
- EXPORT |vpx_h_predictor_4x4_neon|
- EXPORT |vpx_h_predictor_8x8_neon|
- EXPORT |vpx_h_predictor_16x16_neon|
- EXPORT |vpx_h_predictor_32x32_neon|
- EXPORT |vpx_tm_predictor_4x4_neon|
- EXPORT |vpx_tm_predictor_8x8_neon|
- EXPORT |vpx_tm_predictor_16x16_neon|
- EXPORT |vpx_tm_predictor_32x32_neon|
- ARM
- REQUIRE8
- PRESERVE8
- AREA ||.text||, CODE, READONLY, ALIGN=2
- ;void vpx_v_predictor_4x4_neon(uint8_t *dst, ptrdiff_t y_stride,
- ; const uint8_t *above,
- ; const uint8_t *left)
- ; r0 uint8_t *dst
- ; r1 ptrdiff_t y_stride
- ; r2 const uint8_t *above
- ; r3 const uint8_t *left
- |vpx_v_predictor_4x4_neon| PROC
- vld1.32 {d0[0]}, [r2]
- vst1.32 {d0[0]}, [r0], r1
- vst1.32 {d0[0]}, [r0], r1
- vst1.32 {d0[0]}, [r0], r1
- vst1.32 {d0[0]}, [r0], r1
- bx lr
- ENDP ; |vpx_v_predictor_4x4_neon|
- ;void vpx_v_predictor_8x8_neon(uint8_t *dst, ptrdiff_t y_stride,
- ; const uint8_t *above,
- ; const uint8_t *left)
- ; r0 uint8_t *dst
- ; r1 ptrdiff_t y_stride
- ; r2 const uint8_t *above
- ; r3 const uint8_t *left
- |vpx_v_predictor_8x8_neon| PROC
- vld1.8 {d0}, [r2]
- vst1.8 {d0}, [r0], r1
- vst1.8 {d0}, [r0], r1
- vst1.8 {d0}, [r0], r1
- vst1.8 {d0}, [r0], r1
- vst1.8 {d0}, [r0], r1
- vst1.8 {d0}, [r0], r1
- vst1.8 {d0}, [r0], r1
- vst1.8 {d0}, [r0], r1
- bx lr
- ENDP ; |vpx_v_predictor_8x8_neon|
- ;void vpx_v_predictor_16x16_neon(uint8_t *dst, ptrdiff_t y_stride,
- ; const uint8_t *above,
- ; const uint8_t *left)
- ; r0 uint8_t *dst
- ; r1 ptrdiff_t y_stride
- ; r2 const uint8_t *above
- ; r3 const uint8_t *left
- |vpx_v_predictor_16x16_neon| PROC
- vld1.8 {q0}, [r2]
- vst1.8 {q0}, [r0], r1
- vst1.8 {q0}, [r0], r1
- vst1.8 {q0}, [r0], r1
- vst1.8 {q0}, [r0], r1
- vst1.8 {q0}, [r0], r1
- vst1.8 {q0}, [r0], r1
- vst1.8 {q0}, [r0], r1
- vst1.8 {q0}, [r0], r1
- vst1.8 {q0}, [r0], r1
- vst1.8 {q0}, [r0], r1
- vst1.8 {q0}, [r0], r1
- vst1.8 {q0}, [r0], r1
- vst1.8 {q0}, [r0], r1
- vst1.8 {q0}, [r0], r1
- vst1.8 {q0}, [r0], r1
- vst1.8 {q0}, [r0], r1
- bx lr
- ENDP ; |vpx_v_predictor_16x16_neon|
- ;void vpx_v_predictor_32x32_neon(uint8_t *dst, ptrdiff_t y_stride,
- ; const uint8_t *above,
- ; const uint8_t *left)
- ; r0 uint8_t *dst
- ; r1 ptrdiff_t y_stride
- ; r2 const uint8_t *above
- ; r3 const uint8_t *left
- |vpx_v_predictor_32x32_neon| PROC
- vld1.8 {q0, q1}, [r2]
- mov r2, #2
- loop_v
- vst1.8 {q0, q1}, [r0], r1
- vst1.8 {q0, q1}, [r0], r1
- vst1.8 {q0, q1}, [r0], r1
- vst1.8 {q0, q1}, [r0], r1
- vst1.8 {q0, q1}, [r0], r1
- vst1.8 {q0, q1}, [r0], r1
- vst1.8 {q0, q1}, [r0], r1
- vst1.8 {q0, q1}, [r0], r1
- vst1.8 {q0, q1}, [r0], r1
- vst1.8 {q0, q1}, [r0], r1
- vst1.8 {q0, q1}, [r0], r1
- vst1.8 {q0, q1}, [r0], r1
- vst1.8 {q0, q1}, [r0], r1
- vst1.8 {q0, q1}, [r0], r1
- vst1.8 {q0, q1}, [r0], r1
- vst1.8 {q0, q1}, [r0], r1
- subs r2, r2, #1
- bgt loop_v
- bx lr
- ENDP ; |vpx_v_predictor_32x32_neon|
- ;void vpx_h_predictor_4x4_neon(uint8_t *dst, ptrdiff_t y_stride,
- ; const uint8_t *above,
- ; const uint8_t *left)
- ; r0 uint8_t *dst
- ; r1 ptrdiff_t y_stride
- ; r2 const uint8_t *above
- ; r3 const uint8_t *left
- |vpx_h_predictor_4x4_neon| PROC
- vld1.32 {d1[0]}, [r3]
- vdup.8 d0, d1[0]
- vst1.32 {d0[0]}, [r0], r1
- vdup.8 d0, d1[1]
- vst1.32 {d0[0]}, [r0], r1
- vdup.8 d0, d1[2]
- vst1.32 {d0[0]}, [r0], r1
- vdup.8 d0, d1[3]
- vst1.32 {d0[0]}, [r0], r1
- bx lr
- ENDP ; |vpx_h_predictor_4x4_neon|
- ;void vpx_h_predictor_8x8_neon(uint8_t *dst, ptrdiff_t y_stride,
- ; const uint8_t *above,
- ; const uint8_t *left)
- ; r0 uint8_t *dst
- ; r1 ptrdiff_t y_stride
- ; r2 const uint8_t *above
- ; r3 const uint8_t *left
- |vpx_h_predictor_8x8_neon| PROC
- vld1.64 {d1}, [r3]
- vdup.8 d0, d1[0]
- vst1.64 {d0}, [r0], r1
- vdup.8 d0, d1[1]
- vst1.64 {d0}, [r0], r1
- vdup.8 d0, d1[2]
- vst1.64 {d0}, [r0], r1
- vdup.8 d0, d1[3]
- vst1.64 {d0}, [r0], r1
- vdup.8 d0, d1[4]
- vst1.64 {d0}, [r0], r1
- vdup.8 d0, d1[5]
- vst1.64 {d0}, [r0], r1
- vdup.8 d0, d1[6]
- vst1.64 {d0}, [r0], r1
- vdup.8 d0, d1[7]
- vst1.64 {d0}, [r0], r1
- bx lr
- ENDP ; |vpx_h_predictor_8x8_neon|
- ;void vpx_h_predictor_16x16_neon(uint8_t *dst, ptrdiff_t y_stride,
- ; const uint8_t *above,
- ; const uint8_t *left)
- ; r0 uint8_t *dst
- ; r1 ptrdiff_t y_stride
- ; r2 const uint8_t *above
- ; r3 const uint8_t *left
- |vpx_h_predictor_16x16_neon| PROC
- vld1.8 {q1}, [r3]
- vdup.8 q0, d2[0]
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d2[1]
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d2[2]
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d2[3]
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d2[4]
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d2[5]
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d2[6]
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d2[7]
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d3[0]
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d3[1]
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d3[2]
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d3[3]
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d3[4]
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d3[5]
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d3[6]
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d3[7]
- vst1.8 {q0}, [r0], r1
- bx lr
- ENDP ; |vpx_h_predictor_16x16_neon|
- ;void vpx_h_predictor_32x32_neon(uint8_t *dst, ptrdiff_t y_stride,
- ; const uint8_t *above,
- ; const uint8_t *left)
- ; r0 uint8_t *dst
- ; r1 ptrdiff_t y_stride
- ; r2 const uint8_t *above
- ; r3 const uint8_t *left
- |vpx_h_predictor_32x32_neon| PROC
- sub r1, r1, #16
- mov r2, #2
- loop_h
- vld1.8 {q1}, [r3]!
- vdup.8 q0, d2[0]
- vst1.8 {q0}, [r0]!
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d2[1]
- vst1.8 {q0}, [r0]!
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d2[2]
- vst1.8 {q0}, [r0]!
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d2[3]
- vst1.8 {q0}, [r0]!
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d2[4]
- vst1.8 {q0}, [r0]!
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d2[5]
- vst1.8 {q0}, [r0]!
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d2[6]
- vst1.8 {q0}, [r0]!
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d2[7]
- vst1.8 {q0}, [r0]!
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d3[0]
- vst1.8 {q0}, [r0]!
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d3[1]
- vst1.8 {q0}, [r0]!
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d3[2]
- vst1.8 {q0}, [r0]!
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d3[3]
- vst1.8 {q0}, [r0]!
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d3[4]
- vst1.8 {q0}, [r0]!
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d3[5]
- vst1.8 {q0}, [r0]!
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d3[6]
- vst1.8 {q0}, [r0]!
- vst1.8 {q0}, [r0], r1
- vdup.8 q0, d3[7]
- vst1.8 {q0}, [r0]!
- vst1.8 {q0}, [r0], r1
- subs r2, r2, #1
- bgt loop_h
- bx lr
- ENDP ; |vpx_h_predictor_32x32_neon|
- ;void vpx_tm_predictor_4x4_neon (uint8_t *dst, ptrdiff_t y_stride,
- ; const uint8_t *above,
- ; const uint8_t *left)
- ; r0 uint8_t *dst
- ; r1 ptrdiff_t y_stride
- ; r2 const uint8_t *above
- ; r3 const uint8_t *left
- |vpx_tm_predictor_4x4_neon| PROC
- ; Load ytop_left = above[-1];
- sub r12, r2, #1
- vld1.u8 {d0[]}, [r12]
- ; Load above 4 pixels
- vld1.32 {d2[0]}, [r2]
- ; Compute above - ytop_left
- vsubl.u8 q3, d2, d0
- ; Load left row by row and compute left + (above - ytop_left)
- ; 1st row and 2nd row
- vld1.u8 {d2[]}, [r3]!
- vld1.u8 {d4[]}, [r3]!
- vmovl.u8 q1, d2
- vmovl.u8 q2, d4
- vadd.s16 q1, q1, q3
- vadd.s16 q2, q2, q3
- vqmovun.s16 d0, q1
- vqmovun.s16 d1, q2
- vst1.32 {d0[0]}, [r0], r1
- vst1.32 {d1[0]}, [r0], r1
- ; 3rd row and 4th row
- vld1.u8 {d2[]}, [r3]!
- vld1.u8 {d4[]}, [r3]
- vmovl.u8 q1, d2
- vmovl.u8 q2, d4
- vadd.s16 q1, q1, q3
- vadd.s16 q2, q2, q3
- vqmovun.s16 d0, q1
- vqmovun.s16 d1, q2
- vst1.32 {d0[0]}, [r0], r1
- vst1.32 {d1[0]}, [r0], r1
- bx lr
- ENDP ; |vpx_tm_predictor_4x4_neon|
- ;void vpx_tm_predictor_8x8_neon (uint8_t *dst, ptrdiff_t y_stride,
- ; const uint8_t *above,
- ; const uint8_t *left)
- ; r0 uint8_t *dst
- ; r1 ptrdiff_t y_stride
- ; r2 const uint8_t *above
- ; r3 const uint8_t *left
- |vpx_tm_predictor_8x8_neon| PROC
- ; Load ytop_left = above[-1];
- sub r12, r2, #1
- vld1.8 {d0[]}, [r12]
- ; preload 8 left
- vld1.8 {d30}, [r3]
- ; Load above 8 pixels
- vld1.64 {d2}, [r2]
- vmovl.u8 q10, d30
- ; Compute above - ytop_left
- vsubl.u8 q3, d2, d0
- ; Load left row by row and compute left + (above - ytop_left)
- ; 1st row and 2nd row
- vdup.16 q0, d20[0]
- vdup.16 q1, d20[1]
- vadd.s16 q0, q3, q0
- vadd.s16 q1, q3, q1
- ; 3rd row and 4th row
- vdup.16 q8, d20[2]
- vdup.16 q9, d20[3]
- vadd.s16 q8, q3, q8
- vadd.s16 q9, q3, q9
- vqmovun.s16 d0, q0
- vqmovun.s16 d1, q1
- vqmovun.s16 d2, q8
- vqmovun.s16 d3, q9
- vst1.64 {d0}, [r0], r1
- vst1.64 {d1}, [r0], r1
- vst1.64 {d2}, [r0], r1
- vst1.64 {d3}, [r0], r1
- ; 5th row and 6th row
- vdup.16 q0, d21[0]
- vdup.16 q1, d21[1]
- vadd.s16 q0, q3, q0
- vadd.s16 q1, q3, q1
- ; 7th row and 8th row
- vdup.16 q8, d21[2]
- vdup.16 q9, d21[3]
- vadd.s16 q8, q3, q8
- vadd.s16 q9, q3, q9
- vqmovun.s16 d0, q0
- vqmovun.s16 d1, q1
- vqmovun.s16 d2, q8
- vqmovun.s16 d3, q9
- vst1.64 {d0}, [r0], r1
- vst1.64 {d1}, [r0], r1
- vst1.64 {d2}, [r0], r1
- vst1.64 {d3}, [r0], r1
- bx lr
- ENDP ; |vpx_tm_predictor_8x8_neon|
- ;void vpx_tm_predictor_16x16_neon (uint8_t *dst, ptrdiff_t y_stride,
- ; const uint8_t *above,
- ; const uint8_t *left)
- ; r0 uint8_t *dst
- ; r1 ptrdiff_t y_stride
- ; r2 const uint8_t *above
- ; r3 const uint8_t *left
- |vpx_tm_predictor_16x16_neon| PROC
- ; Load ytop_left = above[-1];
- sub r12, r2, #1
- vld1.8 {d0[]}, [r12]
- ; Load above 8 pixels
- vld1.8 {q1}, [r2]
- ; preload 8 left into r12
- vld1.8 {d18}, [r3]!
- ; Compute above - ytop_left
- vsubl.u8 q2, d2, d0
- vsubl.u8 q3, d3, d0
- vmovl.u8 q10, d18
- ; Load left row by row and compute left + (above - ytop_left)
- ; Process 8 rows in each single loop and loop 2 times to process 16 rows.
- mov r2, #2
- loop_16x16_neon
- ; Process two rows.
- vdup.16 q0, d20[0]
- vdup.16 q8, d20[1]
- vadd.s16 q1, q0, q2
- vadd.s16 q0, q0, q3
- vadd.s16 q11, q8, q2
- vadd.s16 q8, q8, q3
- vqmovun.s16 d2, q1
- vqmovun.s16 d3, q0
- vqmovun.s16 d22, q11
- vqmovun.s16 d23, q8
- vdup.16 q0, d20[2] ; proload next 2 rows data
- vdup.16 q8, d20[3]
- vst1.64 {d2,d3}, [r0], r1
- vst1.64 {d22,d23}, [r0], r1
- ; Process two rows.
- vadd.s16 q1, q0, q2
- vadd.s16 q0, q0, q3
- vadd.s16 q11, q8, q2
- vadd.s16 q8, q8, q3
- vqmovun.s16 d2, q1
- vqmovun.s16 d3, q0
- vqmovun.s16 d22, q11
- vqmovun.s16 d23, q8
- vdup.16 q0, d21[0] ; proload next 2 rows data
- vdup.16 q8, d21[1]
- vst1.64 {d2,d3}, [r0], r1
- vst1.64 {d22,d23}, [r0], r1
- vadd.s16 q1, q0, q2
- vadd.s16 q0, q0, q3
- vadd.s16 q11, q8, q2
- vadd.s16 q8, q8, q3
- vqmovun.s16 d2, q1
- vqmovun.s16 d3, q0
- vqmovun.s16 d22, q11
- vqmovun.s16 d23, q8
- vdup.16 q0, d21[2] ; proload next 2 rows data
- vdup.16 q8, d21[3]
- vst1.64 {d2,d3}, [r0], r1
- vst1.64 {d22,d23}, [r0], r1
- vadd.s16 q1, q0, q2
- vadd.s16 q0, q0, q3
- vadd.s16 q11, q8, q2
- vadd.s16 q8, q8, q3
- vqmovun.s16 d2, q1
- vqmovun.s16 d3, q0
- vqmovun.s16 d22, q11
- vqmovun.s16 d23, q8
- vld1.8 {d18}, [r3]! ; preload 8 left into r12
- vmovl.u8 q10, d18
- vst1.64 {d2,d3}, [r0], r1
- vst1.64 {d22,d23}, [r0], r1
- subs r2, r2, #1
- bgt loop_16x16_neon
- bx lr
- ENDP ; |vpx_tm_predictor_16x16_neon|
- ;void vpx_tm_predictor_32x32_neon (uint8_t *dst, ptrdiff_t y_stride,
- ; const uint8_t *above,
- ; const uint8_t *left)
- ; r0 uint8_t *dst
- ; r1 ptrdiff_t y_stride
- ; r2 const uint8_t *above
- ; r3 const uint8_t *left
- |vpx_tm_predictor_32x32_neon| PROC
- ; Load ytop_left = above[-1];
- sub r12, r2, #1
- vld1.8 {d0[]}, [r12]
- ; Load above 32 pixels
- vld1.8 {q1}, [r2]!
- vld1.8 {q2}, [r2]
- ; preload 8 left pixels
- vld1.8 {d26}, [r3]!
- ; Compute above - ytop_left
- vsubl.u8 q8, d2, d0
- vsubl.u8 q9, d3, d0
- vsubl.u8 q10, d4, d0
- vsubl.u8 q11, d5, d0
- vmovl.u8 q3, d26
- ; Load left row by row and compute left + (above - ytop_left)
- ; Process 8 rows in each single loop and loop 4 times to process 32 rows.
- mov r2, #4
- loop_32x32_neon
- ; Process two rows.
- vdup.16 q0, d6[0]
- vdup.16 q2, d6[1]
- vadd.s16 q12, q0, q8
- vadd.s16 q13, q0, q9
- vadd.s16 q14, q0, q10
- vadd.s16 q15, q0, q11
- vqmovun.s16 d0, q12
- vqmovun.s16 d1, q13
- vadd.s16 q12, q2, q8
- vadd.s16 q13, q2, q9
- vqmovun.s16 d2, q14
- vqmovun.s16 d3, q15
- vadd.s16 q14, q2, q10
- vadd.s16 q15, q2, q11
- vst1.64 {d0-d3}, [r0], r1
- vqmovun.s16 d24, q12
- vqmovun.s16 d25, q13
- vqmovun.s16 d26, q14
- vqmovun.s16 d27, q15
- vdup.16 q1, d6[2]
- vdup.16 q2, d6[3]
- vst1.64 {d24-d27}, [r0], r1
- ; Process two rows.
- vadd.s16 q12, q1, q8
- vadd.s16 q13, q1, q9
- vadd.s16 q14, q1, q10
- vadd.s16 q15, q1, q11
- vqmovun.s16 d0, q12
- vqmovun.s16 d1, q13
- vadd.s16 q12, q2, q8
- vadd.s16 q13, q2, q9
- vqmovun.s16 d2, q14
- vqmovun.s16 d3, q15
- vadd.s16 q14, q2, q10
- vadd.s16 q15, q2, q11
- vst1.64 {d0-d3}, [r0], r1
- vqmovun.s16 d24, q12
- vqmovun.s16 d25, q13
- vqmovun.s16 d26, q14
- vqmovun.s16 d27, q15
- vdup.16 q0, d7[0]
- vdup.16 q2, d7[1]
- vst1.64 {d24-d27}, [r0], r1
- ; Process two rows.
- vadd.s16 q12, q0, q8
- vadd.s16 q13, q0, q9
- vadd.s16 q14, q0, q10
- vadd.s16 q15, q0, q11
- vqmovun.s16 d0, q12
- vqmovun.s16 d1, q13
- vadd.s16 q12, q2, q8
- vadd.s16 q13, q2, q9
- vqmovun.s16 d2, q14
- vqmovun.s16 d3, q15
- vadd.s16 q14, q2, q10
- vadd.s16 q15, q2, q11
- vst1.64 {d0-d3}, [r0], r1
- vqmovun.s16 d24, q12
- vqmovun.s16 d25, q13
- vqmovun.s16 d26, q14
- vqmovun.s16 d27, q15
- vdup.16 q0, d7[2]
- vdup.16 q2, d7[3]
- vst1.64 {d24-d27}, [r0], r1
- ; Process two rows.
- vadd.s16 q12, q0, q8
- vadd.s16 q13, q0, q9
- vadd.s16 q14, q0, q10
- vadd.s16 q15, q0, q11
- vqmovun.s16 d0, q12
- vqmovun.s16 d1, q13
- vadd.s16 q12, q2, q8
- vadd.s16 q13, q2, q9
- vqmovun.s16 d2, q14
- vqmovun.s16 d3, q15
- vadd.s16 q14, q2, q10
- vadd.s16 q15, q2, q11
- vst1.64 {d0-d3}, [r0], r1
- vqmovun.s16 d24, q12
- vqmovun.s16 d25, q13
- vld1.8 {d0}, [r3]! ; preload 8 left pixels
- vqmovun.s16 d26, q14
- vqmovun.s16 d27, q15
- vmovl.u8 q3, d0
- vst1.64 {d24-d27}, [r0], r1
- subs r2, r2, #1
- bgt loop_32x32_neon
- bx lr
- ENDP ; |vpx_tm_predictor_32x32_neon|
- END
|