vpx/vp8/common/arm/neon/shortidct4x4llm_1_neon.asm
2010-05-18 11:58:33 -04:00

67 lines
1.6 KiB
NASM

;
; Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
;
; Use of this source code is governed by a BSD-style license and patent
; grant that can be found in the LICENSE file in the root of the source
; tree. All contributing project authors may be found in the AUTHORS
; file in the root of the source tree.
;
EXPORT |vp8_short_idct4x4llm_1_neon|
EXPORT |vp8_dc_only_idct_neon|
ARM
REQUIRE8
PRESERVE8
AREA ||.text||, CODE, READONLY, ALIGN=2
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;void vp8_short_idct4x4llm_1_c(short *input, short *output, int pitch);
; r0 short *input;
; r1 short *output;
; r2 int pitch;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
|vp8_short_idct4x4llm_1_neon| PROC
vld1.16 {d0[]}, [r0] ;load input[0]
add r3, r1, r2
add r12, r3, r2
vrshr.s16 d0, d0, #3
add r0, r12, r2
vst1.16 {d0}, [r1]
vst1.16 {d0}, [r3]
vst1.16 {d0}, [r12]
vst1.16 {d0}, [r0]
bx lr
ENDP
;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;void vp8_dc_only_idct_c(short input_dc, short *output, int pitch);
; r0 short input_dc;
; r1 short *output;
; r2 int pitch;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
|vp8_dc_only_idct_neon| PROC
vdup.16 d0, r0
add r3, r1, r2
add r12, r3, r2
vrshr.s16 d0, d0, #3
add r0, r12, r2
vst1.16 {d0}, [r1]
vst1.16 {d0}, [r3]
vst1.16 {d0}, [r12]
vst1.16 {d0}, [r0]
bx lr
ENDP
END