Mercurial > libavcodec.hg
view arm/simple_idct_armv6.S @ 11104:bb877c9cb102 libavcodec
Detect spatial direct MBs partitioned smaller than 16x16 that can be partitioned
as 16x16 (except ones changing interlacing relative to the colocated MB).
20 cycles slower during MV generation
175 cycles faster during MC
| author | michael |
|---|---|
| date | Mon, 08 Feb 2010 16:23:05 +0000 |
| parents | db79dcbd5161 |
| children | 361a5fcb4393 |
line wrap: on
line source
/* * Simple IDCT * * Copyright (c) 2001 Michael Niedermayer <michaelni@gmx.at> * Copyright (c) 2007 Mans Rullgard <mans@mansr.com> * * This file is part of FFmpeg. * * FFmpeg is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; either * version 2.1 of the License, or (at your option) any later version. * * FFmpeg is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public * License along with FFmpeg; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ #include "asm.S" #define W1 22725 /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */ #define W2 21407 /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */ #define W3 19266 /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */ #define W4 16383 /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */ #define W5 12873 /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */ #define W6 8867 /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */ #define W7 4520 /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */ #define ROW_SHIFT 11 #define COL_SHIFT 20 #define W13 (W1 | (W3 << 16)) #define W26 (W2 | (W6 << 16)) #define W42 (W4 | (W2 << 16)) #define W42n (-W4&0xffff | (-W2 << 16)) #define W46 (W4 | (W6 << 16)) #define W57 (W5 | (W7 << 16)) .text .align w13: .long W13 w26: .long W26 w42: .long W42 w42n: .long W42n w46: .long W46 w57: .long W57 /* Compute partial IDCT of single row. shift = left-shift amount r0 = source address r2 = row[2,0] <= 2 cycles r3 = row[3,1] ip = w42 <= 2 cycles Output in registers r4--r11 */ .macro idct_row shift ldr lr, w46 /* lr = W4 | (W6 << 16) */ mov r1, #(1<<(\shift-1)) smlad r4, r2, ip, r1 smlsd r7, r2, ip, r1 ldr ip, w13 /* ip = W1 | (W3 << 16) */ ldr r10,w57 /* r10 = W5 | (W7 << 16) */ smlad r5, r2, lr, r1 smlsd r6, r2, lr, r1 smuad r8, r3, ip /* r8 = B0 = W1*row[1] + W3*row[3] */ smusdx r11,r3, r10 /* r11 = B3 = W7*row[1] - W5*row[3] */ ldr lr, [r0, #12] /* lr = row[7,5] */ pkhtb r2, ip, r10,asr #16 /* r3 = W7 | (W3 << 16) */ pkhbt r1, ip, r10,lsl #16 /* r1 = W1 | (W5 << 16) */ smusdx r9, r2, r3 /* r9 = -B1 = W7*row[3] - W3*row[1] */ smlad r8, lr, r10,r8 /* B0 += W5*row[5] + W7*row[7] */ smusdx r10,r3, r1 /* r10 = B2 = W5*row[1] - W1*row[3] */ ldr r3, w42n /* r3 = -W4 | (-W2 << 16) */ smlad r10,lr, r2, r10 /* B2 += W7*row[5] + W3*row[7] */ ldr r2, [r0, #4] /* r2 = row[6,4] */ smlsdx r11,lr, ip, r11 /* B3 += W3*row[5] - W1*row[7] */ ldr ip, w46 /* ip = W4 | (W6 << 16) */ smlad r9, lr, r1, r9 /* B1 -= W1*row[5] + W5*row[7] */ smlad r5, r2, r3, r5 /* A1 += -W4*row[4] - W2*row[6] */ smlsd r6, r2, r3, r6 /* A2 += -W4*row[4] + W2*row[6] */ smlad r4, r2, ip, r4 /* A0 += W4*row[4] + W6*row[6] */ smlsd r7, r2, ip, r7 /* A3 += W4*row[4] - W6*row[6] */ .endm /* Compute partial IDCT of half row. shift = left-shift amount r2 = row[2,0] r3 = row[3,1] ip = w42 Output in registers r4--r11 */ .macro idct_row4 shift ldr lr, w46 /* lr = W4 | (W6 << 16) */ ldr r10,w57 /* r10 = W5 | (W7 << 16) */ mov r1, #(1<<(\shift-1)) smlad r4, r2, ip, r1 smlsd r7, r2, ip, r1 ldr ip, w13 /* ip = W1 | (W3 << 16) */ smlad r5, r2, lr, r1 smlsd r6, r2, lr, r1 smusdx r11,r3, r10 /* r11 = B3 = W7*row[1] - W5*row[3] */ smuad r8, r3, ip /* r8 = B0 = W1*row[1] + W3*row[3] */ pkhtb r2, ip, r10,asr #16 /* r3 = W7 | (W3 << 16) */ pkhbt r1, ip, r10,lsl #16 /* r1 = W1 | (W5 << 16) */ smusdx r9, r2, r3 /* r9 = -B1 = W7*row[3] - W3*row[1] */ smusdx r10,r3, r1 /* r10 = B2 = W5*row[1] - W1*row[3] */ .endm /* Compute final part of IDCT single row without shift. Input in registers r4--r11 Output in registers ip, r4--r6, lr, r8--r10 */ .macro idct_finish add ip, r4, r8 /* r1 = A0 + B0 */ sub lr, r4, r8 /* r2 = A0 - B0 */ sub r4, r5, r9 /* r2 = A1 + B1 */ add r8, r5, r9 /* r2 = A1 - B1 */ add r5, r6, r10 /* r1 = A2 + B2 */ sub r9, r6, r10 /* r1 = A2 - B2 */ add r6, r7, r11 /* r2 = A3 + B3 */ sub r10,r7, r11 /* r2 = A3 - B3 */ .endm /* Compute final part of IDCT single row. shift = right-shift amount Input/output in registers r4--r11 */ .macro idct_finish_shift shift add r3, r4, r8 /* r3 = A0 + B0 */ sub r2, r4, r8 /* r2 = A0 - B0 */ mov r4, r3, asr #\shift mov r8, r2, asr #\shift sub r3, r5, r9 /* r3 = A1 + B1 */ add r2, r5, r9 /* r2 = A1 - B1 */ mov r5, r3, asr #\shift mov r9, r2, asr #\shift add r3, r6, r10 /* r3 = A2 + B2 */ sub r2, r6, r10 /* r2 = A2 - B2 */ mov r6, r3, asr #\shift mov r10,r2, asr #\shift add r3, r7, r11 /* r3 = A3 + B3 */ sub r2, r7, r11 /* r2 = A3 - B3 */ mov r7, r3, asr #\shift mov r11,r2, asr #\shift .endm /* Compute final part of IDCT single row, saturating results at 8 bits. shift = right-shift amount Input/output in registers r4--r11 */ .macro idct_finish_shift_sat shift add r3, r4, r8 /* r3 = A0 + B0 */ sub ip, r4, r8 /* ip = A0 - B0 */ usat r4, #8, r3, asr #\shift usat r8, #8, ip, asr #\shift sub r3, r5, r9 /* r3 = A1 + B1 */ add ip, r5, r9 /* ip = A1 - B1 */ usat r5, #8, r3, asr #\shift usat r9, #8, ip, asr #\shift add r3, r6, r10 /* r3 = A2 + B2 */ sub ip, r6, r10 /* ip = A2 - B2 */ usat r6, #8, r3, asr #\shift usat r10,#8, ip, asr #\shift add r3, r7, r11 /* r3 = A3 + B3 */ sub ip, r7, r11 /* ip = A3 - B3 */ usat r7, #8, r3, asr #\shift usat r11,#8, ip, asr #\shift .endm /* Compute IDCT of single row, storing as column. r0 = source r1 = dest */ function idct_row_armv6 push {lr} ldr lr, [r0, #12] /* lr = row[7,5] */ ldr ip, [r0, #4] /* ip = row[6,4] */ ldr r3, [r0, #8] /* r3 = row[3,1] */ ldr r2, [r0] /* r2 = row[2,0] */ orrs lr, lr, ip cmpeq lr, r3 cmpeq lr, r2, lsr #16 beq 1f push {r1} ldr ip, w42 /* ip = W4 | (W2 << 16) */ cmp lr, #0 beq 2f idct_row ROW_SHIFT b 3f 2: idct_row4 ROW_SHIFT 3: pop {r1} idct_finish_shift ROW_SHIFT strh r4, [r1] strh r5, [r1, #(16*2)] strh r6, [r1, #(16*4)] strh r7, [r1, #(16*6)] strh r11,[r1, #(16*1)] strh r10,[r1, #(16*3)] strh r9, [r1, #(16*5)] strh r8, [r1, #(16*7)] pop {pc} 1: mov r2, r2, lsl #3 strh r2, [r1] strh r2, [r1, #(16*2)] strh r2, [r1, #(16*4)] strh r2, [r1, #(16*6)] strh r2, [r1, #(16*1)] strh r2, [r1, #(16*3)] strh r2, [r1, #(16*5)] strh r2, [r1, #(16*7)] pop {pc} .endfunc /* Compute IDCT of single column, read as row. r0 = source r1 = dest */ function idct_col_armv6 push {r1, lr} ldr r2, [r0] /* r2 = row[2,0] */ ldr ip, w42 /* ip = W4 | (W2 << 16) */ ldr r3, [r0, #8] /* r3 = row[3,1] */ idct_row COL_SHIFT pop {r1} idct_finish_shift COL_SHIFT strh r4, [r1] strh r5, [r1, #(16*1)] strh r6, [r1, #(16*2)] strh r7, [r1, #(16*3)] strh r11,[r1, #(16*4)] strh r10,[r1, #(16*5)] strh r9, [r1, #(16*6)] strh r8, [r1, #(16*7)] pop {pc} .endfunc /* Compute IDCT of single column, read as row, store saturated 8-bit. r0 = source r1 = dest r2 = line size */ function idct_col_put_armv6 push {r1, r2, lr} ldr r2, [r0] /* r2 = row[2,0] */ ldr ip, w42 /* ip = W4 | (W2 << 16) */ ldr r3, [r0, #8] /* r3 = row[3,1] */ idct_row COL_SHIFT pop {r1, r2} idct_finish_shift_sat COL_SHIFT strb r4, [r1], r2 strb r5, [r1], r2 strb r6, [r1], r2 strb r7, [r1], r2 strb r11,[r1], r2 strb r10,[r1], r2 strb r9, [r1], r2 strb r8, [r1], r2 sub r1, r1, r2, lsl #3 pop {pc} .endfunc /* Compute IDCT of single column, read as row, add/store saturated 8-bit. r0 = source r1 = dest r2 = line size */ function idct_col_add_armv6 push {r1, r2, lr} ldr r2, [r0] /* r2 = row[2,0] */ ldr ip, w42 /* ip = W4 | (W2 << 16) */ ldr r3, [r0, #8] /* r3 = row[3,1] */ idct_row COL_SHIFT pop {r1, r2} idct_finish ldrb r3, [r1] ldrb r7, [r1, r2] ldrb r11,[r1, r2, lsl #2] add ip, r3, ip, asr #COL_SHIFT usat ip, #8, ip add r4, r7, r4, asr #COL_SHIFT strb ip, [r1], r2 ldrb ip, [r1, r2] usat r4, #8, r4 ldrb r11,[r1, r2, lsl #2] add r5, ip, r5, asr #COL_SHIFT usat r5, #8, r5 strb r4, [r1], r2 ldrb r3, [r1, r2] ldrb ip, [r1, r2, lsl #2] strb r5, [r1], r2 ldrb r7, [r1, r2] ldrb r4, [r1, r2, lsl #2] add r6, r3, r6, asr #COL_SHIFT usat r6, #8, r6 add r10,r7, r10,asr #COL_SHIFT usat r10,#8, r10 add r9, r11,r9, asr #COL_SHIFT usat r9, #8, r9 add r8, ip, r8, asr #COL_SHIFT usat r8, #8, r8 add lr, r4, lr, asr #COL_SHIFT usat lr, #8, lr strb r6, [r1], r2 strb r10,[r1], r2 strb r9, [r1], r2 strb r8, [r1], r2 strb lr, [r1], r2 sub r1, r1, r2, lsl #3 pop {pc} .endfunc /* Compute 8 IDCT row transforms. func = IDCT row->col function width = width of columns in bytes */ .macro idct_rows func width bl \func add r0, r0, #(16*2) add r1, r1, #\width bl \func add r0, r0, #(16*2) add r1, r1, #\width bl \func add r0, r0, #(16*2) add r1, r1, #\width bl \func sub r0, r0, #(16*5) add r1, r1, #\width bl \func add r0, r0, #(16*2) add r1, r1, #\width bl \func add r0, r0, #(16*2) add r1, r1, #\width bl \func add r0, r0, #(16*2) add r1, r1, #\width bl \func sub r0, r0, #(16*7) .endm /* void ff_simple_idct_armv6(DCTELEM *data); */ function ff_simple_idct_armv6, export=1 push {r4-r11, lr} sub sp, sp, #128 mov r1, sp idct_rows idct_row_armv6, 2 mov r1, r0 mov r0, sp idct_rows idct_col_armv6, 2 add sp, sp, #128 pop {r4-r11, pc} .endfunc /* ff_simple_idct_add_armv6(uint8_t *dest, int line_size, DCTELEM *data); */ function ff_simple_idct_add_armv6, export=1 push {r0, r1, r4-r11, lr} sub sp, sp, #128 mov r0, r2 mov r1, sp idct_rows idct_row_armv6, 2 mov r0, sp ldr r1, [sp, #128] ldr r2, [sp, #(128+4)] idct_rows idct_col_add_armv6, 1 add sp, sp, #(128+8) pop {r4-r11, pc} .endfunc /* ff_simple_idct_put_armv6(uint8_t *dest, int line_size, DCTELEM *data); */ function ff_simple_idct_put_armv6, export=1 push {r0, r1, r4-r11, lr} sub sp, sp, #128 mov r0, r2 mov r1, sp idct_rows idct_row_armv6, 2 mov r0, sp ldr r1, [sp, #128] ldr r2, [sp, #(128+4)] idct_rows idct_col_put_armv6, 1 add sp, sp, #(128+8) pop {r4-r11, pc} .endfunc