Parent Directory
| 
Revision Log
Revision 244 - (view) (download)
| 1 : | ia64p | 244 | |
| 2 : | .file "interpolate8x8_ia64.s" | ||
| 3 : | .pred.safe_across_calls p1-p5,p16-p63 | ||
| 4 : | .text | ||
| 5 : | .align 16 | ||
| 6 : | .global interpolate8x8_halfpel_h_ia64# | ||
| 7 : | .proc interpolate8x8_halfpel_h_ia64# | ||
| 8 : | interpolate8x8_halfpel_h_ia64: | ||
| 9 : | LL=3 | ||
| 10 : | SL=1 | ||
| 11 : | SL2=1 | ||
| 12 : | OL=1 | ||
| 13 : | OL2=1 | ||
| 14 : | AVL=1 | ||
| 15 : | AL=1 | ||
| 16 : | STL=3 | ||
| 17 : | |||
| 18 : | alloc r9=ar.pfs,4, 60,0,64 | ||
| 19 : | |||
| 20 : | mov r20 = ar.lc | ||
| 21 : | mov r21 = pr | ||
| 22 : | |||
| 23 : | dep.z r22 = r33,3,3 // rshift of src | ||
| 24 : | |||
| 25 : | and r14 = -8,r33 // align src | ||
| 26 : | mov r15 = r32 // get dest | ||
| 27 : | mov r16 = r34 // stride | ||
| 28 : | sub r17 = 1,r35 // 1-rounding | ||
| 29 : | ;; | ||
| 30 : | |||
| 31 : | add r18 = 8,r14 | ||
| 32 : | mux1 r17 = r17, @brcst // broadcast 1-rounding | ||
| 33 : | |||
| 34 : | sub r24 = 64,r22 // lshift of src | ||
| 35 : | add r26 = 8,r22 // rshift of src+1 | ||
| 36 : | sub r27 = 56,r22 // lshift of src+1 | ||
| 37 : | |||
| 38 : | mov ar.lc = 7 // loopcounter | ||
| 39 : | mov ar.ec = LL + SL +OL + AVL + AL + STL // sum of latencies | ||
| 40 : | mov pr.rot = 1 << 16 // init pr regs for sw-pipeling | ||
| 41 : | |||
| 42 : | ;; | ||
| 43 : | .rotr ald1[LL+1],ald2[LL+1],shru1[SL+1],shl1[SL+1],shru2[SL+1],shl2[SL+1],or1[OL+1],or2[OL+1+AL],add1[AL+1],avg[AVL+1] | ||
| 44 : | .rotp aldp[LL], sh1p[SL], or1p[OL], addp[AL], pavg1p[AVL],stp[STL] | ||
| 45 : | |||
| 46 : | |||
| 47 : | loop_interpolate: | ||
| 48 : | (aldp[0]) ld8 ald1[0] = [r14],r16 // load aligned src | ||
| 49 : | (aldp[0]) ld8 ald2[0] = [r18],r16 // and aligned src+8 | ||
| 50 : | |||
| 51 : | (sh1p[0]) shr.u shru1[0] = ald1[LL],r22 // get src | ||
| 52 : | (sh1p[0]) shl shl1[0] = ald2[LL],r27 | ||
| 53 : | (sh1p[0]) shr.u shru2[0] = ald1[LL],r26 // get src+1 | ||
| 54 : | (sh1p[0]) shl shl2[0] = ald2[LL],r24 | ||
| 55 : | |||
| 56 : | (or1p[0]) or or1[0] = shru1[SL],shl2[SL] // merge things | ||
| 57 : | (or1p[0]) or or2[0] = shru2[SL],shl1[SL] | ||
| 58 : | |||
| 59 : | (addp[0]) padd1.uus add1[0] = or1[OL],r17 // add 1-rounding | ||
| 60 : | |||
| 61 : | (pavg1p[0]) pavg1 avg[0] = add1[AL],or2[OL+AL] // parallel average | ||
| 62 : | |||
| 63 : | (stp[0]) st8 [r15] = avg[AVL] // store results | ||
| 64 : | (stp[0]) add r15 = r15,r16 | ||
| 65 : | |||
| 66 : | |||
| 67 : | |||
| 68 : | |||
| 69 : | br.ctop.sptk.few loop_interpolate | ||
| 70 : | ;; | ||
| 71 : | mov ar.lc = r20 | ||
| 72 : | mov pr = r21,-1 | ||
| 73 : | br.ret.sptk.many b0 | ||
| 74 : | .endp interpolate8x8_halfpel_h_ia64# | ||
| 75 : | |||
| 76 : | .align 16 | ||
| 77 : | .global interpolate8x8_halfpel_v_ia64# | ||
| 78 : | .proc interpolate8x8_halfpel_v_ia64# | ||
| 79 : | interpolate8x8_halfpel_v_ia64: | ||
| 80 : | LL=3 | ||
| 81 : | SL=1 | ||
| 82 : | SL2=1 | ||
| 83 : | OL=1 | ||
| 84 : | OL2=1 | ||
| 85 : | AVL=1 | ||
| 86 : | AL=1 | ||
| 87 : | STL=3 | ||
| 88 : | |||
| 89 : | alloc r9=ar.pfs,4, 60,0,64 | ||
| 90 : | |||
| 91 : | mov r20 = ar.lc | ||
| 92 : | mov r21 = pr | ||
| 93 : | |||
| 94 : | dep.z r22 = r33,3,3 | ||
| 95 : | |||
| 96 : | and r14 = -8,r33 | ||
| 97 : | mov r15 = r32 | ||
| 98 : | mov r16 = r34 | ||
| 99 : | sub r17 = 1,r35 | ||
| 100 : | ;; | ||
| 101 : | |||
| 102 : | add r18 = 8,r14 | ||
| 103 : | add r19 = r14,r16 // src + stride | ||
| 104 : | mux1 r17 = r17, @brcst | ||
| 105 : | |||
| 106 : | sub r24 = 64,r22 | ||
| 107 : | ;; | ||
| 108 : | add r26 = 8,r19 // src + stride + 8 | ||
| 109 : | |||
| 110 : | mov ar.lc = 7 | ||
| 111 : | mov ar.ec = LL + SL +OL + AVL + AL + STL | ||
| 112 : | mov pr.rot = 1 << 16 | ||
| 113 : | |||
| 114 : | ;; | ||
| 115 : | .rotr ald1[LL+1],ald2[LL+1],ald3[LL+1],ald4[LL+1],shru1[SL+1],shl1[SL+1],shru2[SL+1],shl2[SL+1],or1[OL+1],or2[OL+1+AL],add1[AL+1],avg[AVL+1] | ||
| 116 : | .rotp aldp[LL], sh1p[SL], or1p[OL], addp[AL], pavg1p[AVL],stp[STL] | ||
| 117 : | |||
| 118 : | |||
| 119 : | loop_interpolate2: | ||
| 120 : | (aldp[0]) ld8 ald1[0] = [r14],r16 | ||
| 121 : | (aldp[0]) ld8 ald2[0] = [r18],r16 | ||
| 122 : | (aldp[0]) ld8 ald3[0] = [r19],r16 | ||
| 123 : | (aldp[0]) ld8 ald4[0] = [r26],r16 | ||
| 124 : | |||
| 125 : | (sh1p[0]) shr.u shru1[0] = ald1[LL],r22 | ||
| 126 : | (sh1p[0]) shl shl1[0] = ald2[LL],r24 | ||
| 127 : | (sh1p[0]) shr.u shru2[0] = ald3[LL],r22 | ||
| 128 : | (sh1p[0]) shl shl2[0] = ald4[LL],r24 | ||
| 129 : | |||
| 130 : | (or1p[0]) or or1[0] = shru1[SL],shl1[SL] | ||
| 131 : | (or1p[0]) or or2[0] = shru2[SL],shl2[SL] | ||
| 132 : | |||
| 133 : | (addp[0]) padd1.uus add1[0] = or1[OL],r17 | ||
| 134 : | |||
| 135 : | (pavg1p[0]) pavg1 avg[0] = add1[AL],or2[OL+AL] | ||
| 136 : | |||
| 137 : | (stp[0]) st8 [r15] = avg[AVL] | ||
| 138 : | (stp[0]) add r15 = r15,r16 | ||
| 139 : | |||
| 140 : | |||
| 141 : | |||
| 142 : | |||
| 143 : | br.ctop.sptk.few loop_interpolate2 | ||
| 144 : | ;; | ||
| 145 : | mov ar.lc = r20 | ||
| 146 : | mov pr = r21,-1 | ||
| 147 : | br.ret.sptk.many b0 | ||
| 148 : | .endp interpolate8x8_halfpel_v_ia64# | ||
| 149 : | |||
| 150 : | .align 16 | ||
| 151 : | .global interpolate8x8_halfpel_hv_ia64# | ||
| 152 : | .proc interpolate8x8_halfpel_hv_ia64# | ||
| 153 : | interpolate8x8_halfpel_hv_ia64: | ||
| 154 : | LL=3 | ||
| 155 : | SL=1 | ||
| 156 : | SL2=1 | ||
| 157 : | OL=1 | ||
| 158 : | OL2=1 | ||
| 159 : | AVL=1 | ||
| 160 : | AL=1 | ||
| 161 : | STL=3 | ||
| 162 : | |||
| 163 : | alloc r9=ar.pfs,4, 60,0,64 | ||
| 164 : | |||
| 165 : | mov r20 = ar.lc | ||
| 166 : | mov r21 = pr | ||
| 167 : | |||
| 168 : | dep.z r22 = r33,3,3 | ||
| 169 : | |||
| 170 : | and r14 = -8,r33 | ||
| 171 : | mov r15 = r32 | ||
| 172 : | mov r16 = r34 | ||
| 173 : | sub r17 = 1,r35 | ||
| 174 : | ;; | ||
| 175 : | |||
| 176 : | add r18 = 8,r14 | ||
| 177 : | add r19 = r14,r16 | ||
| 178 : | mux1 r17 = r17, @brcst | ||
| 179 : | |||
| 180 : | add r27 = 8,r22 | ||
| 181 : | sub r28 = 56,r22 | ||
| 182 : | sub r24 = 64,r22 | ||
| 183 : | ;; | ||
| 184 : | add r26 = 8,r19 | ||
| 185 : | |||
| 186 : | mov ar.lc = 7 | ||
| 187 : | mov ar.ec = LL + SL +OL + 2*AVL + AL + STL | ||
| 188 : | mov pr.rot = 1 << 16 | ||
| 189 : | |||
| 190 : | ;; | ||
| 191 : | .rotr ald1[LL+1],ald2[LL+1],ald3[LL+1],ald4[LL+1],shru1[SL+1],shl1[SL+1],shru2[SL+1],shl2[SL+1],shl3[SL+1],shru3[SL+1],shl4[SL+1],shru4[SL+1],or1[OL+1],or2[OL+1+AL],or3[OL+AL+1],or4[OL+AL+1],add1[AL+1],avg[AVL+1],avg1[AVL+1],avg2[AVL+1] | ||
| 192 : | .rotp aldp[LL], sh1p[SL], or1p[OL], addp[AL],pavg1p[AVL],pavg2p[AVL],stp[STL] | ||
| 193 : | |||
| 194 : | |||
| 195 : | loop_interpolate3: | ||
| 196 : | (aldp[0]) ld8 ald1[0] = [r14],r16 | ||
| 197 : | (aldp[0]) ld8 ald2[0] = [r18],r16 | ||
| 198 : | (aldp[0]) ld8 ald3[0] = [r19],r16 | ||
| 199 : | (aldp[0]) ld8 ald4[0] = [r26],r16 | ||
| 200 : | |||
| 201 : | (sh1p[0]) shr.u shru1[0] = ald1[LL],r22 | ||
| 202 : | (sh1p[0]) shl shl1[0] = ald2[LL],r24 | ||
| 203 : | (sh1p[0]) shr.u shru2[0] = ald3[LL],r22 | ||
| 204 : | (sh1p[0]) shl shl2[0] = ald4[LL],r24 | ||
| 205 : | (sh1p[0]) shr.u shru3[0] = ald1[LL],r27 | ||
| 206 : | (sh1p[0]) shl shl3[0] = ald2[LL],r28 | ||
| 207 : | (sh1p[0]) shr.u shru4[0] = ald3[LL],r27 | ||
| 208 : | (sh1p[0]) shl shl4[0] = ald4[LL],r28 | ||
| 209 : | |||
| 210 : | |||
| 211 : | (or1p[0]) or or1[0] = shru1[SL],shl1[SL] | ||
| 212 : | (or1p[0]) or or2[0] = shru2[SL],shl2[SL] | ||
| 213 : | (or1p[0]) or or3[0] = shru3[SL],shl3[SL] | ||
| 214 : | (or1p[0]) or or4[0] = shru4[SL],shl4[SL] | ||
| 215 : | |||
| 216 : | (addp[0]) padd1.uus add1[0] = or1[OL],r17 | ||
| 217 : | |||
| 218 : | (pavg1p[0]) pavg1 avg[0] = add1[AL],or2[OL+AL] | ||
| 219 : | (pavg1p[0]) pavg1 avg1[0] = or3[OL+AL],or4[OL+AL] | ||
| 220 : | |||
| 221 : | (pavg2p[0]) pavg1 avg2[0] = avg[AVL],avg1[AVL] | ||
| 222 : | |||
| 223 : | (stp[0]) st8 [r15] = avg2[AVL] | ||
| 224 : | (stp[0]) add r15 = r15,r16 | ||
| 225 : | |||
| 226 : | |||
| 227 : | |||
| 228 : | |||
| 229 : | br.ctop.sptk.few loop_interpolate3 | ||
| 230 : | ;; | ||
| 231 : | mov ar.lc = r20 | ||
| 232 : | mov pr = r21,-1 | ||
| 233 : | br.ret.sptk.many b0 | ||
| 234 : | .endp interpolate8x8_halfpel_hv_ia64# | ||
| 235 : | |||
| 236 : |
| No admin address has been configured | ViewVC Help |
| Powered by ViewVC 1.0.4 |