| 1 |
.file "quant_h263.c" |
// **************************************************************************** |
| 2 |
|
// * |
| 3 |
|
// * XVID MPEG-4 VIDEO CODEC |
| 4 |
|
// * - IA64 h.263 quantization - |
| 5 |
|
// * |
| 6 |
|
// * Copyright(C) 2002 Christian Engel, Hans-Joachim Daniels |
| 7 |
|
// * |
| 8 |
|
// * This program is free software; you can redistribute it and/or modify it |
| 9 |
|
// * under the terms of the GNU General Public License as published by |
| 10 |
|
// * the Free Software Foundation; either version 2 of the License, or |
| 11 |
|
// * (at your option) any later version. |
| 12 |
|
// * |
| 13 |
|
// * This program is distributed in the hope that it will be useful, |
| 14 |
|
// * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 15 |
|
// * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 16 |
|
// * GNU General Public License for more details. |
| 17 |
|
// * |
| 18 |
|
// * You should have received a copy of the GNU General Public License |
| 19 |
|
// * along with this program; if not, write to the Free Software |
| 20 |
|
// * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
| 21 |
|
// * |
| 22 |
|
// * $Id$ |
| 23 |
|
// * |
| 24 |
|
// ***************************************************************************/ |
| 25 |
|
// |
| 26 |
|
// **************************************************************************** |
| 27 |
|
// * |
| 28 |
|
// * quant_h263_ia64.s, IA-64 h.263 quantization |
| 29 |
|
// * |
| 30 |
|
// * This version was implemented during an IA-64 practical training at |
| 31 |
|
// * the University of Karlsruhe (http://i44w3.info.uni-karlsruhe.de/) |
| 32 |
|
// * |
| 33 |
|
// **************************************************************************** |
| 34 |
|
|
| 35 |
|
// ***************************************************************************** |
| 36 |
|
// * |
| 37 |
|
// * functions quant_inter and dequant_inter have been softwarepipelined |
| 38 |
|
// * use was made of the pmpyshr2 instruction |
| 39 |
|
// * |
| 40 |
|
// * by Christian Engel and Hans-Joachim Daniels |
| 41 |
|
// * christian.engel@ira.uka.de hans-joachim.daniels@ira.uka.de |
| 42 |
|
// * |
| 43 |
|
// * This was made for the ia64 DivX laboratory (yes, it was really called |
| 44 |
|
// * this way, originally OpenDivX was intendet, but died shortly before our |
| 45 |
|
// * work started (you will probably already know ...)) |
| 46 |
|
// * at the Universitat Karlsruhe (TH) held between April and July 2002 |
| 47 |
|
// * http://www.info.uni-karlsruhe.de/~rubino/ia64p/ |
| 48 |
|
// * |
| 49 |
|
// *****************************************************************************/ |
| 50 |
|
|
| 51 |
|
.file "quant_h263_ia64.s" |
| 52 |
.pred.safe_across_calls p1-p5,p16-p63 |
.pred.safe_across_calls p1-p5,p16-p63 |
| 53 |
.section .rodata |
.section .rodata |
| 54 |
.align 4 |
.align 4 |
| 90 |
.global __divdi3# |
.global __divdi3# |
| 91 |
.text |
.text |
| 92 |
.align 16 |
.align 16 |
| 93 |
.global quant_intra_ia64# |
.global quant_h263_intra_ia64# |
| 94 |
.proc quant_intra_ia64# |
.proc quant_h263_intra_ia64# |
| 95 |
quant_intra_ia64: |
quant_h263_intra_ia64: |
| 96 |
.prologue //12, 37 |
.prologue |
| 97 |
.save ar.pfs, r38 |
.save ar.pfs, r38 |
| 98 |
alloc r38 = ar.pfs, 4, 3, 2, 0 |
alloc r38 = ar.pfs, 4, 3, 2, 0 |
| 99 |
adds r16 = -8, r12 |
adds r16 = -8, r12 |
| 122 |
ld4 r16 = [r16] |
ld4 r16 = [r16] |
| 123 |
;; |
;; |
| 124 |
setf.sig f2 = r16 |
setf.sig f2 = r16 |
| 125 |
(p6) br.cond.dptk .L4 |
(p6) br.cond.dptk .L8 |
| 126 |
extr r39 = r35, 1, 31 |
extr r39 = r35, 1, 31 |
| 127 |
sxt4 r40 = r35 |
sxt4 r40 = r35 |
| 128 |
;; |
;; |
| 129 |
add r39 = r39, r15 |
add r39 = r39, r15 |
| 130 |
br .L38 |
br .L21 |
| 131 |
;; |
;; |
| 132 |
.L4: |
.L8: |
| 133 |
extr r39 = r35, 1, 31 |
extr r39 = r35, 1, 31 |
| 134 |
sxt4 r40 = r35 |
sxt4 r40 = r35 |
| 135 |
;; |
;; |
| 136 |
sub r39 = r15, r39 |
sub r39 = r15, r39 |
| 137 |
;; |
;; |
| 138 |
.L38: |
.L21: |
| 139 |
sxt4 r39 = r39 |
sxt4 r39 = r39 |
| 140 |
br.call.sptk.many b0 = __divdi3# |
br.call.sptk.many b0 = __divdi3# |
| 141 |
;; |
;; |
| 142 |
addl r16 = 2, r0 |
addl r14 = 62, r0 |
| 143 |
st2 [r32] = r8 |
st2 [r32] = r8 |
| 144 |
addl r17 = 1, r0 |
addl r19 = 1, r0 |
|
;; |
|
|
add r14 = r33, r16 |
|
|
;; |
|
|
ld2 r15 = [r14] |
|
|
;; |
|
|
sxt2 r15 = r15 |
|
|
;; |
|
|
mov r14 = r15 |
|
|
;; |
|
|
cmp4.le p6, p7 = r0, r14 |
|
|
(p6) br.cond.dptk .L21 |
|
|
sub r14 = r0, r14 |
|
|
;; |
|
|
sxt2 r14 = r14 |
|
|
;; |
|
|
cmp4.le p6, p7 = r36, r14 |
|
|
;; |
|
|
(p7) add r14 = r32, r16 |
|
|
(p6) add r15 = r32, r16 |
|
|
(p6) setf.sig f6 = r14 |
|
|
;; |
|
|
(p7) st2 [r14] = r0 |
|
|
(p6) xma.l f6 = f6, f2, f0 |
|
|
;; |
|
|
(p6) getf.sig r14 = f6 |
|
|
;; |
|
|
(p6) extr r14 = r14, 16, 16 |
|
|
;; |
|
|
(p6) sub r14 = r0, r14 |
|
|
br .L39 |
|
|
;; |
|
|
.L21: |
|
|
cmp4.le p6, p7 = r36, r14 |
|
|
;; |
|
|
(p7) add r14 = r32, r16 |
|
|
(p6) setf.sig f6 = r15 |
|
|
;; |
|
|
(p7) st2 [r14] = r0 |
|
|
(p6) xma.l f6 = f6, f2, f0 |
|
|
(p6) add r15 = r32, r16 |
|
|
;; |
|
|
(p6) getf.sig r14 = f6 |
|
|
;; |
|
|
(p6) extr r14 = r14, 16, 16 |
|
|
.L39: |
|
|
//.pred.rel.mutex p6, p7 |
|
|
;; |
|
|
(p6) st2 [r15] = r14 |
|
|
adds r17 = 1, r17 |
|
|
;; |
|
|
cmp4.geu p6, p7 = 63, r17 |
|
|
(p7) br.cond.dptk .L16 |
|
|
addl r14 = 30, r0 |
|
| 145 |
;; |
;; |
| 146 |
mov ar.lc = r14 |
mov ar.lc = r14 |
| 147 |
;; |
;; |
| 148 |
.L37: |
.L20: |
| 149 |
dep.z r16 = r17, 1, 32 |
dep.z r17 = r19, 1, 32 |
|
;; |
|
|
add r14 = r16, r33 |
|
|
;; |
|
|
ld2 r15 = [r14] |
|
| 150 |
;; |
;; |
| 151 |
sxt2 r15 = r15 |
add r15 = r17, r33 |
| 152 |
;; |
adds r19 = 1, r19 |
|
mov r14 = r15 |
|
|
;; |
|
|
cmp4.le p6, p7 = r0, r14 |
|
|
(p6) br.cond.dptk .L27 |
|
|
sub r14 = r0, r14 |
|
|
;; |
|
|
sxt2 r14 = r14 |
|
|
;; |
|
|
cmp4.le p6, p7 = r36, r14 |
|
|
;; |
|
|
(p7) add r14 = r16, r32 |
|
|
(p6) add r15 = r16, r32 |
|
|
(p6) setf.sig f6 = r14 |
|
|
;; |
|
|
(p7) st2 [r14] = r0 |
|
|
(p6) xma.l f6 = f6, f2, f0 |
|
|
;; |
|
|
(p6) getf.sig r14 = f6 |
|
|
;; |
|
|
(p6) extr r14 = r14, 16, 16 |
|
|
;; |
|
|
(p6) sub r14 = r0, r14 |
|
|
br .L40 |
|
|
;; |
|
|
.L27: |
|
|
cmp4.le p6, p7 = r36, r14 |
|
|
;; |
|
|
(p7) add r14 = r16, r32 |
|
|
(p6) setf.sig f6 = r15 |
|
|
;; |
|
|
(p7) st2 [r14] = r0 |
|
|
(p6) xma.l f6 = f6, f2, f0 |
|
|
(p6) add r15 = r16, r32 |
|
|
;; |
|
|
(p6) getf.sig r14 = f6 |
|
|
;; |
|
|
(p6) extr r14 = r14, 16, 16 |
|
|
.L40: |
|
|
//.pred.rel.mutex p6, p7 |
|
|
;; |
|
|
(p6) st2 [r15] = r14 |
|
|
adds r14 = 1, r17 |
|
|
;; |
|
|
dep.z r16 = r14, 1, 32 |
|
|
;; |
|
|
add r15 = r16, r33 |
|
| 153 |
;; |
;; |
| 154 |
ld2 r14 = [r15] |
ld2 r14 = [r15] |
| 155 |
;; |
;; |
| 156 |
sxt2 r14 = r14 |
sxt2 r14 = r14 |
| 157 |
;; |
;; |
| 158 |
mov r15 = r14 |
mov r16 = r14 |
| 159 |
;; |
mov r18 = r14 |
|
cmp4.le p6, p7 = r0, r15 |
|
|
(p6) br.cond.dptk .L33 |
|
|
sub r14 = r0, r15 |
|
|
;; |
|
|
sxt2 r14 = r14 |
|
|
;; |
|
|
mov r15 = r14 |
|
|
;; |
|
|
cmp4.le p6, p7 = r36, r15 |
|
|
;; |
|
|
(p7) add r14 = r16, r32 |
|
|
(p6) setf.sig f6 = r15 |
|
| 160 |
;; |
;; |
| 161 |
(p7) st2 [r14] = r0 |
sub r15 = r0, r16 |
| 162 |
(p6) xma.l f6 = f6, f2, f0 |
cmp4.le p8, p9 = r36, r16 |
| 163 |
(p6) add r15 = r16, r32 |
cmp4.le p6, p7 = r0, r16 |
| 164 |
;; |
;; |
| 165 |
(p6) getf.sig r14 = f6 |
sxt2 r14 = r15 |
| 166 |
|
(p6) br.cond.dptk .L14 |
| 167 |
;; |
;; |
| 168 |
(p6) extr r14 = r14, 16, 16 |
mov r16 = r14 |
| 169 |
|
add r18 = r17, r32 |
| 170 |
;; |
;; |
| 171 |
(p6) sub r14 = r0, r14 |
setf.sig f6 = r16 |
| 172 |
br .L41 |
cmp4.le p6, p7 = r36, r16 |
| 173 |
.L33: |
mov r15 = r18 |
|
cmp4.le p6, p7 = r36, r15 |
|
| 174 |
;; |
;; |
| 175 |
(p7) add r14 = r16, r32 |
xma.l f6 = f6, f2, f0 |
| 176 |
(p6) add r15 = r16, r32 |
(p7) st2 [r18] = r0 |
|
(p6) setf.sig f6 = r14 |
|
| 177 |
;; |
;; |
| 178 |
(p7) st2 [r14] = r0 |
getf.sig r14 = f6 |
|
(p6) xma.l f6 = f6, f2, f0 |
|
| 179 |
;; |
;; |
| 180 |
(p6) getf.sig r14 = f6 |
extr r14 = r14, 16, 16 |
| 181 |
;; |
;; |
| 182 |
(p6) extr r14 = r14, 16, 16 |
sub r14 = r0, r14 |
|
.L41: |
|
|
//.pred.rel.mutex p6, p7 |
|
| 183 |
;; |
;; |
| 184 |
(p6) st2 [r15] = r14 |
(p6) st2 [r15] = r14 |
| 185 |
adds r17 = 2, r17 |
br .L12 |
| 186 |
br.cloop.sptk.few .L37 |
.L14: |
| 187 |
.L16: |
.pred.rel "mutex", p8, p9 |
| 188 |
|
setf.sig f6 = r18 |
| 189 |
|
add r16 = r17, r32 |
| 190 |
|
;; |
| 191 |
|
xma.l f6 = f6, f2, f0 |
| 192 |
|
mov r15 = r16 |
| 193 |
|
(p9) st2 [r16] = r0 |
| 194 |
|
;; |
| 195 |
|
getf.sig r14 = f6 |
| 196 |
|
;; |
| 197 |
|
extr r14 = r14, 16, 16 |
| 198 |
|
;; |
| 199 |
|
(p8) st2 [r15] = r14 |
| 200 |
|
.L12: |
| 201 |
|
br.cloop.sptk.few .L20 |
| 202 |
adds r18 = 24, r12 |
adds r18 = 24, r12 |
| 203 |
;; |
;; |
| 204 |
ld8 r19 = [r18], 8 |
ld8 r19 = [r18], 8 |
| 210 |
.restore sp |
.restore sp |
| 211 |
adds r12 = 32, r12 |
adds r12 = 32, r12 |
| 212 |
br.ret.sptk.many b0 |
br.ret.sptk.many b0 |
| 213 |
.endp quant_intra_ia64# |
.endp quant_h263_intra_ia64# |
| 214 |
.align 16 |
.common quant_h263_intra#,8,8 |
| 215 |
.global quant_inter_ia64# |
.common dequant_h263_intra#,8,8 |
|
.proc quant_inter_ia64# |
|
|
quant_inter_ia64: |
|
|
.prologue |
|
|
addl r14 = @ltoff(multipliers#), gp |
|
|
dep.z r15 = r34, 2, 32 |
|
|
.save ar.lc, r2 |
|
|
mov r2 = ar.lc |
|
|
;; |
|
|
.body |
|
|
ld8 r14 = [r14] |
|
|
extr.u r16 = r34, 1, 16 |
|
|
dep.z r17 = r34, 1, 15 |
|
|
;; |
|
|
add r15 = r15, r14 |
|
|
mov r18 = r16 |
|
|
mov r8 = r0 |
|
|
;; |
|
|
ld4 r15 = [r15] |
|
|
addl r14 = 31, r0 |
|
|
mov r19 = r0 |
|
|
;; |
|
|
setf.sig f6 = r15 |
|
|
mov ar.lc = r14 |
|
|
;; |
|
|
.L65: |
|
|
dep.z r16 = r19, 1, 32 |
|
|
;; |
|
|
add r14 = r16, r33 |
|
|
;; |
|
|
ld2 r15 = [r14] |
|
|
;; |
|
|
sxt2 r15 = r15 |
|
|
;; |
|
|
mov r14 = r15 |
|
|
;; |
|
|
cmp4.le p6, p7 = r0, r14 |
|
|
(p6) br.cond.dptk .L55 |
|
|
sub r14 = r0, r14 |
|
|
;; |
|
|
sub r14 = r14, r18 |
|
|
;; |
|
|
sxt2 r14 = r14 |
|
|
;; |
|
|
cmp4.le p6, p7 = r17, r14 |
|
|
;; |
|
|
(p7) add r14 = r16, r32 |
|
|
(p6) setf.sig f7 = r14 |
|
|
;; |
|
|
(p7) st2 [r14] = r0 |
|
|
(p6) add r16 = r16, r32 |
|
|
(p6) xma.l f7 = f7, f6, f0 |
|
|
;; |
|
|
(p6) getf.sig r14 = f7 |
|
|
;; |
|
|
(p6) extr r14 = r14, 16, 16 |
|
|
;; |
|
|
(p6) sub r15 = r0, r14 |
|
|
(p6) add r8 = r8, r14 |
|
|
;; |
|
|
(p6) st2 [r16] = r15 |
|
|
br .L53 |
|
|
.L55: |
|
|
sub r14 = r14, r18 |
|
|
;; |
|
|
sxt2 r14 = r14 |
|
|
;; |
|
|
cmp4.le p6, p7 = r17, r14 |
|
|
;; |
|
|
(p7) add r14 = r16, r32 |
|
|
(p6) add r15 = r16, r32 |
|
|
(p6) setf.sig f7 = r14 |
|
|
;; |
|
|
(p7) st2 [r14] = r0 |
|
|
(p6) xma.l f7 = f7, f6, f0 |
|
|
;; |
|
|
(p6) getf.sig r14 = f7 |
|
|
;; |
|
|
(p6) extr r14 = r14, 16, 16 |
|
|
;; |
|
|
(p6) st2 [r15] = r14 |
|
|
(p6) add r8 = r8, r14 |
|
|
.L53: |
|
|
adds r14 = 1, r19 |
|
|
;; |
|
|
dep.z r16 = r14, 1, 32 |
|
|
;; |
|
|
add r15 = r16, r33 |
|
|
;; |
|
|
ld2 r14 = [r15] |
|
|
;; |
|
|
sxt2 r14 = r14 |
|
|
;; |
|
|
cmp4.le p6, p7 = r0, r14 |
|
|
(p6) br.cond.dptk .L61 |
|
|
sub r14 = r0, r14 |
|
|
;; |
|
|
sub r14 = r14, r18 |
|
|
;; |
|
|
sxt2 r14 = r14 |
|
|
;; |
|
|
cmp4.le p6, p7 = r17, r14 |
|
|
;; |
|
|
(p7) add r14 = r16, r32 |
|
|
(p6) setf.sig f7 = r14 |
|
|
;; |
|
|
(p7) st2 [r14] = r0 |
|
|
(p6) add r16 = r16, r32 |
|
|
(p6) xma.l f7 = f7, f6, f0 |
|
|
;; |
|
|
(p6) getf.sig r14 = f7 |
|
|
;; |
|
|
(p6) extr r14 = r14, 16, 16 |
|
|
;; |
|
|
(p6) sub r15 = r0, r14 |
|
|
(p6) add r8 = r8, r14 |
|
|
;; |
|
|
(p6) st2 [r16] = r15 |
|
|
br .L59 |
|
|
.L61: |
|
|
sub r14 = r14, r18 |
|
|
;; |
|
|
sxt2 r14 = r14 |
|
|
;; |
|
|
cmp4.le p6, p7 = r17, r14 |
|
|
;; |
|
|
(p7) add r14 = r16, r32 |
|
|
(p6) add r15 = r16, r32 |
|
|
(p6) setf.sig f7 = r14 |
|
|
;; |
|
|
(p7) st2 [r14] = r0 |
|
|
(p6) xma.l f7 = f7, f6, f0 |
|
|
;; |
|
|
(p6) getf.sig r14 = f7 |
|
|
;; |
|
|
(p6) extr r14 = r14, 16, 16 |
|
|
;; |
|
|
(p6) st2 [r15] = r14 |
|
|
(p6) add r8 = r8, r14 |
|
|
.L59: |
|
|
adds r19 = 2, r19 |
|
|
br.cloop.sptk.few .L65 |
|
|
;; |
|
|
mov ar.lc = r2 |
|
|
br.ret.sptk.many b0 |
|
|
.endp quant_inter_ia64# |
|
|
.common quant_intra#,8,8 |
|
|
.common dequant_intra#,8,8 |
|
| 216 |
.align 16 |
.align 16 |
| 217 |
.global dequant_intra_ia64# |
.global dequant_h263_intra_ia64# |
| 218 |
.proc dequant_intra_ia64# |
.proc dequant_h263_intra_ia64# |
| 219 |
dequant_intra_ia64: |
dequant_h263_intra_ia64: |
| 220 |
.prologue |
.prologue |
| 221 |
ld2 r14 = [r33] |
ld2 r14 = [r33] |
| 222 |
andcm r15 = 1, r34 |
andcm r15 = 1, r34 |
| 244 |
cmp4.le p6, p7 = r16, r15 |
cmp4.le p6, p7 = r16, r15 |
| 245 |
;; |
;; |
| 246 |
(p7) st2 [r32] = r16 |
(p7) st2 [r32] = r16 |
| 247 |
(p7) br.cond.dptk .L68 |
(p7) br.cond.dptk .L32 |
| 248 |
addl r14 = 2047, r0 |
addl r14 = 2047, r0 |
| 249 |
;; |
;; |
| 250 |
cmp4.ge p6, p7 = r14, r15 |
cmp4.ge p6, p7 = r14, r15 |
| 251 |
;; |
;; |
| 252 |
(p7) st2 [r32] = r14 |
(p7) st2 [r32] = r14 |
| 253 |
.L68: |
.L32: |
| 254 |
addl r14 = 20, r0 |
addl r14 = 62, r0 |
| 255 |
addl r19 = 1, r0 |
addl r19 = 1, r0 |
| 256 |
addl r21 = 2048, r0 |
addl r22 = 2048, r0 |
| 257 |
addl r20 = -2048, r0 |
addl r21 = -2048, r0 |
| 258 |
addl r18 = 2047, r0 |
addl r20 = 2047, r0 |
| 259 |
;; |
;; |
| 260 |
mov ar.lc = r14 |
mov ar.lc = r14 |
| 261 |
;; |
;; |
| 262 |
.L110: |
.L56: |
| 263 |
dep.z r16 = r19, 1, 32 |
dep.z r16 = r19, 1, 32 |
| 264 |
;; |
;; |
| 265 |
add r14 = r16, r33 |
add r14 = r16, r33 |
| 266 |
|
add r17 = r16, r32 |
| 267 |
|
adds r19 = 1, r19 |
| 268 |
;; |
;; |
| 269 |
ld2 r15 = [r14] |
ld2 r15 = [r14] |
| 270 |
;; |
;; |
| 271 |
sxt2 r15 = r15 |
sxt2 r15 = r15 |
| 272 |
;; |
;; |
| 273 |
cmp4.ne p6, p7 = 0, r15 |
cmp4.ne p6, p7 = 0, r15 |
| 274 |
|
cmp4.le p8, p9 = r0, r15 |
| 275 |
;; |
;; |
| 276 |
(p7) add r14 = r16, r32 |
(p7) st2 [r17] = r0 |
| 277 |
;; |
(p7) br.cond.dpnt .L36 |
| 278 |
(p7) st2 [r14] = r0 |
add r18 = r16, r32 |
| 279 |
(p7) br.cond.dpnt .L92 |
sub r17 = r0, r15 |
| 280 |
cmp4.le p6, p7 = r0, r15 |
;; |
| 281 |
(p6) br.cond.dptk .L95 |
mov r14 = r18 |
| 282 |
sub r14 = r0, r15 |
(p8) br.cond.dptk .L40 |
| 283 |
add r17 = r16, r32 |
setf.sig f8 = r17 |
|
;; |
|
|
setf.sig f8 = r14 |
|
| 284 |
;; |
;; |
| 285 |
xma.l f8 = f6, f8, f7 |
xma.l f8 = f6, f8, f7 |
| 286 |
;; |
;; |
| 287 |
getf.sig r15 = f8 |
getf.sig r15 = f8 |
| 288 |
;; |
;; |
| 289 |
cmp4.lt p6, p7 = r21, r15 |
cmp4.lt p6, p7 = r22, r15 |
| 290 |
;; |
sub r16 = r0, r15 |
|
(p7) sub r14 = r0, r15 |
|
| 291 |
;; |
;; |
| 292 |
(p7) st2 [r17] = r14 |
(p7) st2 [r14] = r16 |
| 293 |
(p6) st2 [r17] = r20 |
(p6) st2 [r14] = r21 |
| 294 |
br .L92 |
br .L36 |
| 295 |
.L95: |
.L40: |
| 296 |
setf.sig f8 = r15 |
setf.sig f8 = r15 |
|
add r14 = r16, r32 |
|
| 297 |
;; |
;; |
| 298 |
xma.l f8 = f6, f8, f7 |
xma.l f8 = f6, f8, f7 |
| 299 |
;; |
;; |
| 300 |
getf.sig r15 = f8 |
getf.sig r15 = f8 |
| 301 |
;; |
;; |
| 302 |
cmp4.le p6, p7 = r18, r15 |
cmp4.le p6, p7 = r20, r15 |
|
;; |
|
|
(p6) mov r15 = r18 |
|
| 303 |
;; |
;; |
| 304 |
st2 [r14] = r15 |
(p6) mov r14 = r20 |
| 305 |
.L92: |
(p7) mov r14 = r15 |
|
adds r14 = 1, r19 |
|
|
;; |
|
|
dep.z r17 = r14, 1, 32 |
|
|
;; |
|
|
add r15 = r17, r33 |
|
|
;; |
|
|
ld2 r14 = [r15] |
|
|
;; |
|
|
sxt2 r14 = r14 |
|
|
;; |
|
|
mov r16 = r14 |
|
|
;; |
|
|
cmp4.ne p6, p7 = 0, r16 |
|
|
;; |
|
|
(p7) add r14 = r17, r32 |
|
|
;; |
|
|
(p7) st2 [r14] = r0 |
|
|
(p7) br.cond.dpnt .L98 |
|
|
cmp4.le p6, p7 = r0, r16 |
|
|
(p6) br.cond.dptk .L101 |
|
|
sub r14 = r0, r16 |
|
|
add r17 = r17, r32 |
|
|
;; |
|
|
setf.sig f8 = r14 |
|
|
;; |
|
|
xma.l f8 = f6, f8, f7 |
|
| 306 |
;; |
;; |
| 307 |
getf.sig r16 = f8 |
st2 [r18] = r14 |
| 308 |
|
.L36: |
| 309 |
|
br.cloop.sptk.few .L56 |
| 310 |
;; |
;; |
| 311 |
cmp4.lt p6, p7 = r21, r16 |
mov ar.lc = r2 |
| 312 |
;; |
br.ret.sptk.many b0 |
| 313 |
(p7) sub r14 = r0, r16 |
.endp dequant_h263_intra_ia64# |
| 314 |
;; |
|
| 315 |
(p7) st2 [r17] = r14 |
|
| 316 |
(p6) st2 [r17] = r20 |
|
| 317 |
br .L98 |
// uint32_t quant_h263_inter_ia64(int16_t *coeff, const int16_t *data, const uint32_t quant) |
| 318 |
.L101: |
|
| 319 |
setf.sig f8 = r16 |
|
| 320 |
add r14 = r17, r32 |
|
| 321 |
;; |
.common quant_h263_inter#,8,8 |
| 322 |
xma.l f8 = f6, f8, f7 |
.align 16 |
| 323 |
;; |
.global quant_h263_inter_ia64# |
| 324 |
getf.sig r16 = f8 |
.proc quant_h263_inter_ia64# |
| 325 |
;; |
quant_h263_inter_ia64: |
| 326 |
cmp4.le p6, p7 = r18, r16 |
|
| 327 |
;; |
|
| 328 |
(p6) mov r15 = r18 |
//******************************************************* |
| 329 |
(p7) mov r15 = r16 |
//* * |
| 330 |
;; |
//* const uint32_t mult = multipliers[quant]; * |
| 331 |
st2 [r14] = r15 |
//* const uint16_t quant_m_2 = quant << 1; * |
| 332 |
.L98: |
//* const uint16_t quant_d_2 = quant >> 1; * |
| 333 |
adds r14 = 2, r19 |
//* int sum = 0; * |
| 334 |
;; |
//* uint32_t i; * |
| 335 |
dep.z r17 = r14, 1, 32 |
//* int16_t acLevel,acL; * |
| 336 |
;; |
//* * |
| 337 |
add r15 = r17, r33 |
//*******************************************************/ |
| 338 |
;; |
|
| 339 |
ld2 r14 = [r15] |
|
| 340 |
;; |
|
| 341 |
sxt2 r14 = r14 |
LL=3 // LL = load latency |
| 342 |
;; |
//if LL is changed, you'll also have to change the .pred.rel... parts below! |
| 343 |
mov r16 = r14 |
.prologue |
| 344 |
;; |
addl r14 = @ltoff(multipliers#), gp |
| 345 |
cmp4.ne p6, p7 = 0, r16 |
dep.z r15 = r34, 2, 32 |
| 346 |
;; |
.save ar.lc, r2 |
| 347 |
(p7) add r14 = r17, r32 |
mov r2 = ar.lc |
|
;; |
|
|
(p7) st2 [r14] = r0 |
|
|
(p7) br.cond.dpnt .L104 |
|
|
cmp4.le p6, p7 = r0, r16 |
|
|
(p6) br.cond.dptk .L107 |
|
|
sub r14 = r0, r16 |
|
|
add r17 = r17, r32 |
|
|
;; |
|
|
setf.sig f8 = r14 |
|
|
;; |
|
|
xma.l f8 = f6, f8, f7 |
|
|
;; |
|
|
getf.sig r16 = f8 |
|
|
;; |
|
|
cmp4.lt p6, p7 = r21, r16 |
|
|
;; |
|
|
(p7) sub r14 = r0, r16 |
|
| 348 |
;; |
;; |
| 349 |
(p7) st2 [r17] = r14 |
.body |
| 350 |
(p6) st2 [r17] = r20 |
alloc r9=ar.pfs,0,24,0,24 |
| 351 |
br .L104 |
mov r17 = ar.ec |
| 352 |
.L107: |
mov r10 = pr |
| 353 |
setf.sig f8 = r16 |
ld8 r14 = [r14] |
| 354 |
add r14 = r17, r32 |
extr.u r16 = r34, 1, 16 //r16 = quant_d_2 |
| 355 |
|
dep.z r20 = r34, 1, 15 //r20 = quant_m_2 |
| 356 |
;; |
;; |
| 357 |
xma.l f8 = f6, f8, f7 |
add r15 = r15, r14 |
| 358 |
|
mov r21 = r16 //r21 = quant_d_2 |
| 359 |
|
mov r8 = r0 //r8 = sum = 0 |
| 360 |
|
mov pr.rot = 0 //p16-p63 = 0 |
| 361 |
;; |
;; |
| 362 |
getf.sig r16 = f8 |
ld4 r15 = [r15] |
| 363 |
|
addl r14 = 63, r0 |
| 364 |
|
mov pr.rot = 1 << 16 //p16=1 |
| 365 |
;; |
;; |
| 366 |
cmp4.le p6, p7 = r18, r16 |
mov ar.lc = r14 |
| 367 |
|
mov ar.ec = LL+9 |
| 368 |
|
mov r29 = r15 |
| 369 |
;; |
;; |
| 370 |
(p6) mov r15 = r18 |
mov r15 = r33 //r15 = data |
| 371 |
(p7) mov r15 = r16 |
mov r18 = r32 //r18 = coeff |
| 372 |
;; |
;; |
| 373 |
st2 [r14] = r15 |
|
| 374 |
.L104: |
|
| 375 |
adds r19 = 3, r19 |
.rotr ac1[LL+3], ac2[8], ac3[2] |
| 376 |
br.cloop.sptk.few .L110 |
.rotp p[LL+9], cmp1[8], cmp1neg[8],cmp2[5], cmp2neg[2] |
| 377 |
|
|
| 378 |
|
|
| 379 |
|
|
| 380 |
|
//******************************************************************************* |
| 381 |
|
//* * |
| 382 |
|
//* for (i = 0; i < 64; i++) { * |
| 383 |
|
//* acL=acLevel = data[i]; * |
| 384 |
|
//* acLevel = ((acLevel < 0)?-acLevel:acLevel) - quant_d_2; * |
| 385 |
|
//* if (acLevel < quant_m_2){ * |
| 386 |
|
//* acLevel = 0; * |
| 387 |
|
//* } * |
| 388 |
|
//* acLevel = (acLevel * mult) >> SCALEBITS; * |
| 389 |
|
//* sum += acLevel; * |
| 390 |
|
//* coeff[i] = ((acL < 0)?-acLevel:acLevel); * |
| 391 |
|
//* } * |
| 392 |
|
//* * |
| 393 |
|
//*******************************************************************************/ |
| 394 |
|
|
| 395 |
|
|
| 396 |
|
|
| 397 |
|
.explicit |
| 398 |
|
.L58: |
| 399 |
|
.pred.rel "clear", p29, p37 |
| 400 |
|
.pred.rel "mutex", p29, p37 |
| 401 |
|
|
| 402 |
|
//pipeline stage |
| 403 |
|
{.mmi |
| 404 |
|
(p[0]) ld2 ac1[0] = [r15],2 // 0 acL=acLevel = data[i]; |
| 405 |
|
(p[LL+1]) sub ac2[0] = r0, ac1[LL+1] // LL+1 ac2=-acLevel |
| 406 |
|
(p[LL]) sxt2 ac1[LL] = ac1[LL] // LL |
| 407 |
|
} |
| 408 |
|
{.mmi |
| 409 |
|
(p[LL+1]) cmp4.le cmp1[0], cmp1neg[0] = r0, ac1[LL+1] // LL+1 cmp1 = (0<=acLevel) ; cmp1neg = !(0<=acLevel) |
| 410 |
|
(p[LL+4]) cmp4.le cmp2[0], cmp2neg[0] = r20, ac2[3] // LL+4 cmp2 = (quant_m_2 < acLevel) ; cmp2neg = !(quant_m_2 < acLevel) |
| 411 |
|
(cmp1[1]) sub ac2[1] = ac1[LL+2], r21 // LL+2 acLevel = acLevel - quant_d_2; |
| 412 |
|
} |
| 413 |
|
{.mmi |
| 414 |
|
(cmp2neg[1]) mov ac2[4] = r0 // LL+5 if (acLevel < quant_m_2) acLevel=0; |
| 415 |
|
(cmp1neg[1]) sub ac2[1] = ac2[1], r21 // LL+2 acLevel = ac2 - quant_d_2; |
| 416 |
|
(p[LL+3]) sxt2 ac2[2] = ac2[2] // LL+3 |
| 417 |
|
} |
| 418 |
|
{.mmi |
| 419 |
|
.pred.rel "mutex", p34, p42 |
| 420 |
|
(cmp1[6]) mov ac3[0] = ac2[6] // LL+7 ac3 = acLevel; |
| 421 |
|
(cmp1neg[6]) sub ac3[0] = r0, ac2[6] // LL+7 ac3 = -acLevel; |
| 422 |
|
(p[LL+6]) pmpyshr2.u ac2[5] = r29, ac2[5], 16 // LL+6 acLevel = (acLevel * mult) >> SCALEBITS; |
| 423 |
|
} |
| 424 |
|
{.mib |
| 425 |
|
(p[LL+8]) st2 [r18] = ac3[1] , 2 // LL+8 coeff[i] = ac3; |
| 426 |
|
(cmp2[4]) add r8 = r8, ac2[7] // LL+8 sum += acLevel; |
| 427 |
|
br.ctop.sptk.few .L58 |
| 428 |
|
;; |
| 429 |
|
} |
| 430 |
|
|
| 431 |
|
.pred.rel "clear", p29, p37 |
| 432 |
|
.default |
| 433 |
|
mov ar.ec = r17 |
| 434 |
;; |
;; |
| 435 |
mov ar.lc = r2 |
mov ar.lc = r2 |
| 436 |
|
mov pr = r10, -1 |
| 437 |
|
mov ar.pfs = r9 |
| 438 |
br.ret.sptk.many b0 |
br.ret.sptk.many b0 |
| 439 |
.endp dequant_intra_ia64# |
.endp quant_h263_inter_ia64# |
| 440 |
.common quant_inter#,8,8 |
|
| 441 |
.common dequant_inter#,8,8 |
|
| 442 |
|
|
| 443 |
|
|
| 444 |
|
|
| 445 |
|
|
| 446 |
|
|
| 447 |
|
// void dequant_h263_inter_ia64(int16_t *data, const int16_t *coeff, const uint32_t quant) |
| 448 |
|
|
| 449 |
|
.common dequant_h263_inter#,8,8 |
| 450 |
.align 16 |
.align 16 |
| 451 |
.global dequant_inter_ia64# |
.global dequant_h263_inter_ia64# |
| 452 |
.proc dequant_inter_ia64# |
.proc dequant_h263_inter_ia64# |
| 453 |
dequant_inter_ia64: |
dequant_h263_inter_ia64: |
| 454 |
|
|
| 455 |
|
//*********************************************************************** |
| 456 |
|
//* * |
| 457 |
|
//* const uint16_t quant_m_2 = quant << 1; * |
| 458 |
|
//* const uint16_t quant_add = (quant & 1 ? quant : quant - 1); * |
| 459 |
|
//* uint32_t i; * |
| 460 |
|
//* * |
| 461 |
|
//*********************************************************************** |
| 462 |
|
|
| 463 |
|
|
| 464 |
|
|
| 465 |
|
|
| 466 |
.prologue |
.prologue |
| 467 |
andcm r14 = 1, r34 |
andcm r14 = 1, r34 |
| 468 |
dep.z r15 = r34, 1, 15 |
dep.z r29 = r34, 1, 15 |
| 469 |
|
alloc r9=ar.pfs,0,32,0,32 |
| 470 |
.save ar.lc, r2 |
.save ar.lc, r2 |
| 471 |
mov r2 = ar.lc |
mov r2 = ar.lc |
| 472 |
;; |
;; |
| 473 |
.body |
.body |
| 474 |
sub r34 = r34, r14 |
sub r15 = r34, r14 // r15 = quant |
| 475 |
setf.sig f6 = r15 |
addl r14 = 63, r0 |
| 476 |
mov r19 = r0 |
addl r21 = -2048, r0 |
| 477 |
addl r14 = 31, r0 |
addl r20 = 2047, r0 |
| 478 |
addl r18 = -2048, r0 |
mov r16 = ar.ec |
| 479 |
addl r17 = 2047, r0 |
mov r17 = pr |
| 480 |
;; |
;; |
| 481 |
zxt2 r34 = r34 |
zxt2 r15 = r15 |
| 482 |
mov ar.lc = r14 |
mov ar.lc = r14 |
| 483 |
|
mov pr.rot = 0 |
| 484 |
;; |
;; |
| 485 |
.L122: |
adds r14 = 0, r33 // r14 = coeff |
| 486 |
dep.z r16 = r19, 1, 32 |
mov r18 = r32 // r18 = data |
| 487 |
;; |
mov ar.ec = LL+10 |
| 488 |
add r14 = r16, r33 |
mov pr.rot = 1 << 16 |
| 489 |
;; |
;; |
| 490 |
ld2 r15 = [r14] |
|
| 491 |
;; |
//******************************************************************************* |
| 492 |
sxt2 r15 = r15 |
//* * |
| 493 |
;; |
//*for (i = 0; i < 64; i++) { * |
| 494 |
mov r14 = r15 |
//* int16_t acLevel = coeff[i]; * |
| 495 |
;; |
//* * |
| 496 |
cmp4.ne p6, p7 = 0, r14 |
//* if (acLevel == 0) * |
| 497 |
;; |
//* { * |
| 498 |
(p7) add r14 = r16, r32 |
//* data[i] = 0; * |
| 499 |
;; |
//* } * |
| 500 |
(p7) st2 [r14] = r0 |
//* else if (acLevel < 0) * |
| 501 |
(p7) br.cond.dpnt .L112 |
//* { * |
| 502 |
cmp4.le p6, p7 = r0, r14 |
//* acLevel = acLevel * quant_m_2 - quant_add; * |
| 503 |
(p6) br.cond.dptk .L115 |
//* data[i] = (acLevel >= -2048 ? acLevel : -2048); * |
| 504 |
setf.sig f7 = r14 |
//* } * |
| 505 |
add r15 = r16, r32 |
//* else // if (acLevel > 0) * |
| 506 |
;; |
//* { * |
| 507 |
xma.l f7 = f7, f6, f0 |
//* acLevel = acLevel * quant_m_2 + quant_add; * |
| 508 |
;; |
//* data[i] = (acLevel <= 2047 ? acLevel : 2047); * |
| 509 |
getf.sig r14 = f7 |
//* } * |
| 510 |
;; |
//* } * |
| 511 |
sub r14 = r14, r34 |
//* * |
| 512 |
;; |
//*******************************************************************************/ |
| 513 |
sxt2 r14 = r14 |
|
| 514 |
;; |
|
| 515 |
cmp4.le p6, p7 = r18, r14 |
|
| 516 |
;; |
LL=2 // LL := load latency |
| 517 |
(p7) mov r14 = r18 |
//if LL is changed, you'll also have to change the .pred.rel... parts below! |
| 518 |
br .L123 |
|
| 519 |
.L115: |
|
| 520 |
setf.sig f8 = r15 |
.rotr ac1[LL+10], x[5], y1[3], y2[3] |
| 521 |
setf.sig f7 = r34 |
.rotp p[LL+10] , cmp1neg[8], cmp2[5], cmp2neg[5],cmp3[2], cmp3neg[2] |
| 522 |
;; |
|
| 523 |
xma.l f8 = f8, f6, f7 |
.explicit |
| 524 |
add r15 = r16, r32 |
//pipeline stage |
| 525 |
;; |
|
| 526 |
getf.sig r14 = f8 |
.L60: |
| 527 |
;; |
.pred.rel "clear", p36 |
| 528 |
sxt2 r14 = r14 |
.pred.rel "mutex", p47, p49 |
| 529 |
;; |
.pred.rel "mutex", p46, p48 |
| 530 |
cmp4.le p6, p7 = r17, r14 |
.pred.rel "mutex", p40, p45 |
| 531 |
;; |
.pred.rel "mutex", p39, p44 |
| 532 |
(p6) mov r14 = r17 |
.pred.rel "mutex", p38, p43 |
| 533 |
;; |
.pred.rel "mutex", p37, p42 |
| 534 |
.L123: |
.pred.rel "mutex", p36, p41 |
| 535 |
st2 [r15] = r14 |
{.mmi |
| 536 |
.L112: |
(p[0])ld2 ac1[0] = [r14] ,2 // 0 acLevel = coeff[i]; |
| 537 |
adds r14 = 1, r19 |
(p[LL+1])cmp4.ne p6, cmp1neg[0] = 0, ac1[LL+1] // LL+1 |
| 538 |
;; |
(p[LL])sxt2 ac1[LL] = ac1[LL] // LL |
| 539 |
dep.z r16 = r14, 1, 32 |
|
| 540 |
;; |
} |
| 541 |
add r15 = r16, r33 |
{.mmi |
| 542 |
;; |
(p[LL+1])cmp4.le cmp2[0], cmp2neg[0] = r0, ac1[LL+1] // LL+1 |
| 543 |
ld2 r14 = [r15] |
(cmp2[1]) mov x[0] = r20 // LL+2 |
| 544 |
;; |
(p[LL+2])pmpyshr2.u ac1[LL+2] = r29, ac1[LL+2], 0 // LL+2 |
| 545 |
sxt2 r14 = r14 |
} |
| 546 |
;; |
{.mmi |
| 547 |
mov r15 = r14 |
(cmp2neg[1]) mov x[0] = r21 // LL+2 |
| 548 |
;; |
(cmp2[2]) add ac1[LL+3] = ac1[LL+3], r15 // LL+3 |
| 549 |
cmp4.ne p6, p7 = 0, r15 |
(cmp2neg[2]) sub ac1[LL+3] = ac1[LL+3], r15 // LL+3 |
| 550 |
;; |
|
| 551 |
(p7) add r14 = r16, r32 |
} |
| 552 |
;; |
{.mmi |
| 553 |
(p7) st2 [r14] = r0 |
(cmp2neg[4]) mov y1[0] = ac1[LL+5] // LL+5 |
| 554 |
(p7) br.cond.dpnt .L117 |
(cmp2neg[4]) mov y2[0] = x[3] // LL+5 |
| 555 |
cmp4.le p6, p7 = r0, r15 |
(p[LL+4])sxt2 ac1[LL+4] = ac1[LL+4] // LL+4 |
| 556 |
(p6) br.cond.dptk .L120 |
} |
| 557 |
setf.sig f8 = r15 |
{.mmi |
| 558 |
;; |
(cmp2[4]) mov y1[0] = x[3] // LL+5 |
| 559 |
xma.l f8 = f8, f6, f0 |
(cmp2[4]) mov y2[0] = ac1[LL+5] // LL+5 |
| 560 |
add r15 = r16, r32 |
(p[LL+6])cmp4.le cmp3[0], cmp3neg[0] = x[4], ac1[LL+6] // LL+6 |
| 561 |
;; |
} |
| 562 |
getf.sig r14 = f8 |
{.mmi |
| 563 |
;; |
(cmp3[1]) mov ac1[LL+7] = y1[2] // LL+7 |
| 564 |
sub r14 = r14, r34 |
(cmp3neg[1]) mov ac1[LL+7] = y2[2] // LL+7 |
| 565 |
;; |
(cmp1neg[7]) mov ac1[LL+8] = r0 // LL+8 |
| 566 |
sxt2 r14 = r14 |
} |
| 567 |
;; |
{.mbb |
| 568 |
cmp4.le p6, p7 = r18, r14 |
(p[LL+9])st2 [r18] = ac1[LL+9] ,2 // LL+9 |
| 569 |
;; |
nop.b 0x0 |
| 570 |
(p7) mov r14 = r18 |
br.ctop.sptk.few .L60 |
| 571 |
br .L124 |
;; |
| 572 |
;; |
} |
| 573 |
.L120: |
.pred.rel "clear", p36 |
| 574 |
setf.sig f7 = r14 |
.default |
| 575 |
setf.sig f8 = r34 |
mov ar.lc = r2 |
| 576 |
add r15 = r16, r32 |
mov ar.pfs = r9 |
| 577 |
;; |
mov ar.ec = r16 |
| 578 |
xma.l f7 = f7, f6, f8 |
mov pr = r17, -1 |
|
;; |
|
|
getf.sig r14 = f7 |
|
|
;; |
|
|
sxt2 r14 = r14 |
|
|
;; |
|
|
cmp4.le p6, p7 = r17, r14 |
|
|
;; |
|
|
(p6) mov r14 = r17 |
|
|
;; |
|
|
.L124: |
|
|
st2 [r15] = r14 |
|
|
.L117: |
|
|
adds r19 = 2, r19 |
|
|
br.cloop.sptk.few .L122 |
|
| 579 |
;; |
;; |
| 580 |
mov ar.lc = r2 |
mov ar.lc = r2 |
| 581 |
br.ret.sptk.many b0 |
br.ret.sptk.many b0 |
| 582 |
.endp dequant_inter_ia64# |
.endp dequant_h263_inter_ia64# |
| 583 |
.ident "GCC: (GNU) 2.96 20000731 (Red Hat Linux 7.1 2.96-85)" |
.ident "GCC: (GNU) 2.96 20000731 (Red Hat Linux 7.1 2.96-85)" |
Parent Directory
| 
Revision Log
| 
Patch