1 ; RUN: llc -march=amdgcn < %s -verify-machineinstrs | FileCheck -check-prefix=SI %s 2 ; RUN: llc -march=amdgcn -mcpu=tonga < %s -verify-machineinstrs | FileCheck -check-prefix=SI %s 3 4 ; If this occurs it is likely due to reordering and the restore was 5 ; originally supposed to happen before SI_END_CF. 6 7 ; SI: s_or_b64 exec, exec, [[SAVED:s\[[0-9]+:[0-9]+\]|[a-z]+]] 8 ; SI-NOT: v_readlane_b32 [[SAVED]] 9 10 define amdgpu_ps void @main() #0 { 11 main_body: 12 %tmp = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 16) 13 %tmp1 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 32) 14 %tmp2 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 80) 15 %tmp3 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 84) 16 %tmp4 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 88) 17 %tmp5 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 96) 18 %tmp6 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 100) 19 %tmp7 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 104) 20 %tmp8 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 112) 21 %tmp9 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 116) 22 %tmp10 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 120) 23 %tmp11 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 128) 24 %tmp12 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 132) 25 %tmp13 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 136) 26 %tmp14 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 144) 27 %tmp15 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 148) 28 %tmp16 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 152) 29 %tmp17 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 160) 30 %tmp18 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 164) 31 %tmp19 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 168) 32 %tmp20 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 176) 33 %tmp21 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 180) 34 %tmp22 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 184) 35 %tmp23 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 192) 36 %tmp24 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 196) 37 %tmp25 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 200) 38 %tmp26 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 208) 39 %tmp27 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 212) 40 %tmp28 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 216) 41 %tmp29 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 224) 42 %tmp30 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 228) 43 %tmp31 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 232) 44 %tmp32 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 240) 45 %tmp33 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 244) 46 %tmp34 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 248) 47 %tmp35 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 256) 48 %tmp36 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 260) 49 %tmp37 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 264) 50 %tmp38 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 272) 51 %tmp39 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 276) 52 %tmp40 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 280) 53 %tmp41 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 288) 54 %tmp42 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 292) 55 %tmp43 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 296) 56 %tmp44 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 304) 57 %tmp45 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 308) 58 %tmp46 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 312) 59 %tmp47 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 320) 60 %tmp48 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 324) 61 %tmp49 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 328) 62 %tmp50 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 336) 63 %tmp51 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 340) 64 %tmp52 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 344) 65 %tmp53 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 352) 66 %tmp54 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 356) 67 %tmp55 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 360) 68 %tmp56 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 368) 69 %tmp57 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 372) 70 %tmp58 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 376) 71 %tmp59 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 384) 72 %tmp60 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 388) 73 %tmp61 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 392) 74 %tmp62 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 400) 75 %tmp63 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 404) 76 %tmp64 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 408) 77 %tmp65 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 416) 78 %tmp66 = call float @llvm.SI.load.const.v4i32(<4 x i32> undef, i32 420) 79 br label %LOOP 80 81 LOOP: ; preds = %ENDIF2795, %main_body 82 %temp894.0 = phi float [ 0.000000e+00, %main_body ], [ %temp894.1, %ENDIF2795 ] 83 %temp18.0 = phi float [ undef, %main_body ], [ %temp18.1, %ENDIF2795 ] 84 %tid = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) 85 %tmp67 = icmp sgt i32 %tid, 4 86 br i1 %tmp67, label %ENDLOOP, label %ENDIF 87 88 ENDLOOP: ; preds = %ELSE2566, %LOOP 89 %one.sub.a.i = fsub float 1.000000e+00, %tmp 90 %one.sub.ac.i = fmul float %one.sub.a.i, undef 91 %result.i = fadd float fmul (float undef, float undef), %one.sub.ac.i 92 call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float undef, float %result.i, float undef, float 1.000000e+00, i1 true, i1 true) #0 93 ret void 94 95 ENDIF: ; preds = %LOOP 96 %tmp68 = fsub float %tmp2, undef 97 %tmp69 = fsub float %tmp3, undef 98 %tmp70 = fsub float %tmp4, undef 99 %tmp71 = fmul float %tmp68, 0.000000e+00 100 %tmp72 = fmul float %tmp69, undef 101 %tmp73 = fmul float %tmp70, undef 102 %tmp74 = fsub float %tmp6, undef 103 %tmp75 = fsub float %tmp7, undef 104 %tmp76 = fmul float %tmp74, undef 105 %tmp77 = fmul float %tmp75, 0.000000e+00 106 %tmp78 = call float @llvm.minnum.f32(float %tmp73, float %tmp77) 107 %tmp79 = call float @llvm.maxnum.f32(float %tmp71, float 0.000000e+00) 108 %tmp80 = call float @llvm.maxnum.f32(float %tmp72, float %tmp76) 109 %tmp81 = call float @llvm.maxnum.f32(float undef, float %tmp78) 110 %tmp82 = call float @llvm.minnum.f32(float %tmp79, float %tmp80) 111 %tmp83 = call float @llvm.minnum.f32(float %tmp82, float undef) 112 %tmp84 = fsub float %tmp14, undef 113 %tmp85 = fsub float %tmp15, undef 114 %tmp86 = fsub float %tmp16, undef 115 %tmp87 = fmul float %tmp84, undef 116 %tmp88 = fmul float %tmp85, undef 117 %tmp89 = fmul float %tmp86, undef 118 %tmp90 = fsub float %tmp17, undef 119 %tmp91 = fsub float %tmp18, undef 120 %tmp92 = fsub float %tmp19, undef 121 %tmp93 = fmul float %tmp90, 0.000000e+00 122 %tmp94 = fmul float %tmp91, undef 123 %tmp95 = fmul float %tmp92, undef 124 %tmp96 = call float @llvm.minnum.f32(float %tmp88, float %tmp94) 125 %tmp97 = call float @llvm.maxnum.f32(float %tmp87, float %tmp93) 126 %tmp98 = call float @llvm.maxnum.f32(float %tmp89, float %tmp95) 127 %tmp99 = call float @llvm.maxnum.f32(float undef, float %tmp96) 128 %tmp100 = call float @llvm.maxnum.f32(float %tmp99, float undef) 129 %tmp101 = call float @llvm.minnum.f32(float %tmp97, float undef) 130 %tmp102 = call float @llvm.minnum.f32(float %tmp101, float %tmp98) 131 %tmp103 = fsub float %tmp30, undef 132 %tmp104 = fsub float %tmp31, undef 133 %tmp105 = fmul float %tmp103, 0.000000e+00 134 %tmp106 = fmul float %tmp104, 0.000000e+00 135 %tmp107 = call float @llvm.minnum.f32(float undef, float %tmp105) 136 %tmp108 = call float @llvm.maxnum.f32(float undef, float %tmp106) 137 %tmp109 = call float @llvm.maxnum.f32(float undef, float %tmp107) 138 %tmp110 = call float @llvm.maxnum.f32(float %tmp109, float undef) 139 %tmp111 = call float @llvm.minnum.f32(float undef, float %tmp108) 140 %tmp112 = fsub float %tmp32, undef 141 %tmp113 = fsub float %tmp33, undef 142 %tmp114 = fsub float %tmp34, undef 143 %tmp115 = fmul float %tmp112, 0.000000e+00 144 %tmp116 = fmul float %tmp113, undef 145 %tmp117 = fmul float %tmp114, undef 146 %tmp118 = fsub float %tmp35, undef 147 %tmp119 = fsub float %tmp36, undef 148 %tmp120 = fsub float %tmp37, undef 149 %tmp121 = fmul float %tmp118, undef 150 %tmp122 = fmul float %tmp119, undef 151 %tmp123 = fmul float %tmp120, undef 152 %tmp124 = call float @llvm.minnum.f32(float %tmp115, float %tmp121) 153 %tmp125 = call float @llvm.minnum.f32(float %tmp116, float %tmp122) 154 %tmp126 = call float @llvm.minnum.f32(float %tmp117, float %tmp123) 155 %tmp127 = call float @llvm.maxnum.f32(float %tmp124, float %tmp125) 156 %tmp128 = call float @llvm.maxnum.f32(float %tmp127, float %tmp126) 157 %tmp129 = fsub float %tmp38, undef 158 %tmp130 = fsub float %tmp39, undef 159 %tmp131 = fsub float %tmp40, undef 160 %tmp132 = fmul float %tmp129, 0.000000e+00 161 %tmp133 = fmul float %tmp130, undef 162 %tmp134 = fmul float %tmp131, undef 163 %tmp135 = fsub float %tmp41, undef 164 %tmp136 = fsub float %tmp42, undef 165 %tmp137 = fsub float %tmp43, undef 166 %tmp138 = fmul float %tmp135, undef 167 %tmp139 = fmul float %tmp136, undef 168 %tmp140 = fmul float %tmp137, undef 169 %tmp141 = call float @llvm.minnum.f32(float %tmp132, float %tmp138) 170 %tmp142 = call float @llvm.minnum.f32(float %tmp133, float %tmp139) 171 %tmp143 = call float @llvm.minnum.f32(float %tmp134, float %tmp140) 172 %tmp144 = call float @llvm.maxnum.f32(float %tmp141, float %tmp142) 173 %tmp145 = call float @llvm.maxnum.f32(float %tmp144, float %tmp143) 174 %tmp146 = fsub float %tmp44, undef 175 %tmp147 = fsub float %tmp45, undef 176 %tmp148 = fsub float %tmp46, undef 177 %tmp149 = fmul float %tmp146, 0.000000e+00 178 %tmp150 = fmul float %tmp147, 0.000000e+00 179 %tmp151 = fmul float %tmp148, undef 180 %tmp152 = fsub float %tmp47, undef 181 %tmp153 = fsub float %tmp48, undef 182 %tmp154 = fsub float %tmp49, undef 183 %tmp155 = fmul float %tmp152, undef 184 %tmp156 = fmul float %tmp153, 0.000000e+00 185 %tmp157 = fmul float %tmp154, undef 186 %tmp158 = call float @llvm.minnum.f32(float %tmp149, float %tmp155) 187 %tmp159 = call float @llvm.minnum.f32(float %tmp150, float %tmp156) 188 %tmp160 = call float @llvm.minnum.f32(float %tmp151, float %tmp157) 189 %tmp161 = call float @llvm.maxnum.f32(float %tmp158, float %tmp159) 190 %tmp162 = call float @llvm.maxnum.f32(float %tmp161, float %tmp160) 191 %tmp163 = fsub float %tmp50, undef 192 %tmp164 = fsub float %tmp51, undef 193 %tmp165 = fsub float %tmp52, undef 194 %tmp166 = fmul float %tmp163, undef 195 %tmp167 = fmul float %tmp164, 0.000000e+00 196 %tmp168 = fmul float %tmp165, 0.000000e+00 197 %tmp169 = fsub float %tmp53, undef 198 %tmp170 = fsub float %tmp54, undef 199 %tmp171 = fsub float %tmp55, undef 200 %tmp172 = fdiv float 1.000000e+00, %temp18.0 201 %tmp173 = fmul float %tmp169, undef 202 %tmp174 = fmul float %tmp170, undef 203 %tmp175 = fmul float %tmp171, %tmp172 204 %tmp176 = call float @llvm.minnum.f32(float %tmp166, float %tmp173) 205 %tmp177 = call float @llvm.minnum.f32(float %tmp167, float %tmp174) 206 %tmp178 = call float @llvm.minnum.f32(float %tmp168, float %tmp175) 207 %tmp179 = call float @llvm.maxnum.f32(float %tmp176, float %tmp177) 208 %tmp180 = call float @llvm.maxnum.f32(float %tmp179, float %tmp178) 209 %tmp181 = fsub float %tmp62, undef 210 %tmp182 = fsub float %tmp63, undef 211 %tmp183 = fsub float %tmp64, undef 212 %tmp184 = fmul float %tmp181, 0.000000e+00 213 %tmp185 = fmul float %tmp182, undef 214 %tmp186 = fmul float %tmp183, undef 215 %tmp187 = fsub float %tmp65, undef 216 %tmp188 = fsub float %tmp66, undef 217 %tmp189 = fmul float %tmp187, undef 218 %tmp190 = fmul float %tmp188, undef 219 %tmp191 = call float @llvm.maxnum.f32(float %tmp184, float %tmp189) 220 %tmp192 = call float @llvm.maxnum.f32(float %tmp185, float %tmp190) 221 %tmp193 = call float @llvm.maxnum.f32(float %tmp186, float undef) 222 %tmp194 = call float @llvm.minnum.f32(float %tmp191, float %tmp192) 223 %tmp195 = call float @llvm.minnum.f32(float %tmp194, float %tmp193) 224 %.temp292.7 = select i1 undef, float %tmp162, float undef 225 %temp292.9 = select i1 false, float %tmp180, float %.temp292.7 226 %.temp292.9 = select i1 undef, float undef, float %temp292.9 227 %tmp196 = fcmp ogt float undef, 0.000000e+00 228 %tmp197 = fcmp olt float undef, %tmp195 229 %tmp198 = and i1 %tmp196, %tmp197 230 %tmp199 = fcmp olt float undef, %.temp292.9 231 %tmp200 = and i1 %tmp198, %tmp199 232 %temp292.11 = select i1 %tmp200, float undef, float %.temp292.9 233 %tid0 = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) 234 %cmp0 = icmp eq i32 %tid0, 0 235 br i1 %cmp0, label %IF2565, label %ELSE2566 236 237 IF2565: ; preds = %ENDIF 238 %tid1 = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) 239 %cmp1 = icmp eq i32 %tid1, 0 240 br i1 %cmp1, label %ENDIF2582, label %ELSE2584 241 242 ELSE2566: ; preds = %ENDIF 243 %tid2 = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) 244 %tidf = bitcast i32 %tid2 to float 245 %tmp201 = fcmp oeq float %temp292.11, %tidf 246 br i1 %tmp201, label %ENDLOOP, label %ELSE2593 247 248 ENDIF2564: ; preds = %ENDIF2594, %ENDIF2588 249 %temp894.1 = phi float [ undef, %ENDIF2588 ], [ %temp894.2, %ENDIF2594 ] 250 %temp18.1 = phi float [ %tmp218, %ENDIF2588 ], [ undef, %ENDIF2594 ] 251 %tmp202 = fsub float %tmp5, undef 252 %tmp203 = fmul float %tmp202, undef 253 %tmp204 = call float @llvm.maxnum.f32(float undef, float %tmp203) 254 %tmp205 = call float @llvm.minnum.f32(float %tmp204, float undef) 255 %tmp206 = call float @llvm.minnum.f32(float %tmp205, float undef) 256 %tmp207 = fcmp ogt float undef, 0.000000e+00 257 %tmp208 = fcmp olt float undef, 1.000000e+00 258 %tmp209 = and i1 %tmp207, %tmp208 259 %tid3 = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) 260 %tidf3 = bitcast i32 %tid3 to float 261 %tmp210 = fcmp olt float %tidf3, %tmp206 262 %tmp211 = and i1 %tmp209, %tmp210 263 br i1 %tmp211, label %ENDIF2795, label %ELSE2797 264 265 ELSE2584: ; preds = %IF2565 266 br label %ENDIF2582 267 268 ENDIF2582: ; preds = %ELSE2584, %IF2565 269 %tmp212 = fadd float %tmp1, undef 270 %tmp213 = fadd float 0.000000e+00, %tmp212 271 %floor = call float @llvm.floor.f32(float %tmp213) 272 %tmp214 = fsub float %tmp213, %floor 273 %tid4 = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) 274 %cmp4 = icmp eq i32 %tid4, 0 275 br i1 %cmp4, label %IF2589, label %ELSE2590 276 277 IF2589: ; preds = %ENDIF2582 278 br label %ENDIF2588 279 280 ELSE2590: ; preds = %ENDIF2582 281 br label %ENDIF2588 282 283 ENDIF2588: ; preds = %ELSE2590, %IF2589 284 %tmp215 = fsub float 1.000000e+00, %tmp214 285 %tmp216 = call float @llvm.sqrt.f32(float %tmp215) 286 %tmp217 = fmul float %tmp216, undef 287 %tmp218 = fadd float %tmp217, undef 288 br label %ENDIF2564 289 290 ELSE2593: ; preds = %ELSE2566 291 %tmp219 = fcmp oeq float %temp292.11, %tmp81 292 %tmp220 = fcmp olt float %tmp81, %tmp83 293 %tmp221 = and i1 %tmp219, %tmp220 294 br i1 %tmp221, label %ENDIF2594, label %ELSE2596 295 296 ELSE2596: ; preds = %ELSE2593 297 %tmp222 = fcmp oeq float %temp292.11, %tmp100 298 %tmp223 = fcmp olt float %tmp100, %tmp102 299 %tmp224 = and i1 %tmp222, %tmp223 300 br i1 %tmp224, label %ENDIF2594, label %ELSE2632 301 302 ENDIF2594: ; preds = %ELSE2788, %ELSE2785, %ELSE2782, %ELSE2779, %IF2775, %ELSE2761, %ELSE2758, %IF2757, %ELSE2704, %ELSE2686, %ELSE2671, %ELSE2668, %IF2667, %ELSE2632, %ELSE2596, %ELSE2593 303 %temp894.2 = phi float [ 0.000000e+00, %IF2667 ], [ 0.000000e+00, %ELSE2671 ], [ 0.000000e+00, %IF2757 ], [ 0.000000e+00, %ELSE2761 ], [ %temp894.0, %ELSE2758 ], [ 0.000000e+00, %IF2775 ], [ 0.000000e+00, %ELSE2779 ], [ 0.000000e+00, %ELSE2782 ], [ %.2848, %ELSE2788 ], [ 0.000000e+00, %ELSE2785 ], [ 0.000000e+00, %ELSE2593 ], [ 0.000000e+00, %ELSE2632 ], [ 0.000000e+00, %ELSE2704 ], [ 0.000000e+00, %ELSE2686 ], [ 0.000000e+00, %ELSE2668 ], [ 0.000000e+00, %ELSE2596 ] 304 %tmp225 = fmul float %temp894.2, undef 305 br label %ENDIF2564 306 307 ELSE2632: ; preds = %ELSE2596 308 br i1 undef, label %ENDIF2594, label %ELSE2650 309 310 ELSE2650: ; preds = %ELSE2632 311 %tmp226 = fcmp oeq float %temp292.11, %tmp110 312 %tmp227 = fcmp olt float %tmp110, %tmp111 313 %tmp228 = and i1 %tmp226, %tmp227 314 br i1 %tmp228, label %IF2667, label %ELSE2668 315 316 IF2667: ; preds = %ELSE2650 317 br i1 undef, label %ENDIF2594, label %ELSE2671 318 319 ELSE2668: ; preds = %ELSE2650 320 %tmp229 = fcmp oeq float %temp292.11, %tmp128 321 %tmp230 = fcmp olt float %tmp128, undef 322 %tmp231 = and i1 %tmp229, %tmp230 323 br i1 %tmp231, label %ENDIF2594, label %ELSE2686 324 325 ELSE2671: ; preds = %IF2667 326 br label %ENDIF2594 327 328 ELSE2686: ; preds = %ELSE2668 329 %tmp232 = fcmp oeq float %temp292.11, %tmp145 330 %tmp233 = fcmp olt float %tmp145, undef 331 %tmp234 = and i1 %tmp232, %tmp233 332 br i1 %tmp234, label %ENDIF2594, label %ELSE2704 333 334 ELSE2704: ; preds = %ELSE2686 335 %tmp235 = fcmp oeq float %temp292.11, %tmp180 336 %tmp236 = fcmp olt float %tmp180, undef 337 %tmp237 = and i1 %tmp235, %tmp236 338 br i1 %tmp237, label %ENDIF2594, label %ELSE2740 339 340 ELSE2740: ; preds = %ELSE2704 341 br i1 undef, label %IF2757, label %ELSE2758 342 343 IF2757: ; preds = %ELSE2740 344 br i1 undef, label %ENDIF2594, label %ELSE2761 345 346 ELSE2758: ; preds = %ELSE2740 347 br i1 undef, label %IF2775, label %ENDIF2594 348 349 ELSE2761: ; preds = %IF2757 350 br label %ENDIF2594 351 352 IF2775: ; preds = %ELSE2758 353 %tmp238 = fcmp olt float undef, undef 354 br i1 %tmp238, label %ENDIF2594, label %ELSE2779 355 356 ELSE2779: ; preds = %IF2775 357 br i1 undef, label %ENDIF2594, label %ELSE2782 358 359 ELSE2782: ; preds = %ELSE2779 360 br i1 undef, label %ENDIF2594, label %ELSE2785 361 362 ELSE2785: ; preds = %ELSE2782 363 %tmp239 = fcmp olt float undef, 0.000000e+00 364 br i1 %tmp239, label %ENDIF2594, label %ELSE2788 365 366 ELSE2788: ; preds = %ELSE2785 367 %tmp240 = fcmp olt float 0.000000e+00, undef 368 %.2848 = select i1 %tmp240, float -1.000000e+00, float 1.000000e+00 369 br label %ENDIF2594 370 371 ELSE2797: ; preds = %ENDIF2564 372 %tmp241 = fsub float %tmp8, undef 373 %tmp242 = fsub float %tmp9, undef 374 %tmp243 = fsub float %tmp10, undef 375 %tmp244 = fmul float %tmp241, undef 376 %tmp245 = fmul float %tmp242, undef 377 %tmp246 = fmul float %tmp243, undef 378 %tmp247 = fsub float %tmp11, undef 379 %tmp248 = fsub float %tmp12, undef 380 %tmp249 = fsub float %tmp13, undef 381 %tmp250 = fmul float %tmp247, undef 382 %tmp251 = fmul float %tmp248, undef 383 %tmp252 = fmul float %tmp249, undef 384 %tmp253 = call float @llvm.minnum.f32(float %tmp244, float %tmp250) 385 %tmp254 = call float @llvm.minnum.f32(float %tmp245, float %tmp251) 386 %tmp255 = call float @llvm.maxnum.f32(float %tmp246, float %tmp252) 387 %tmp256 = call float @llvm.maxnum.f32(float %tmp253, float %tmp254) 388 %tmp257 = call float @llvm.maxnum.f32(float %tmp256, float undef) 389 %tmp258 = call float @llvm.minnum.f32(float undef, float %tmp255) 390 %tmp259 = fcmp ogt float %tmp257, 0.000000e+00 391 %tmp260 = fcmp olt float %tmp257, 1.000000e+00 392 %tmp261 = and i1 %tmp259, %tmp260 393 %tmp262 = fcmp olt float %tmp257, %tmp258 394 %tmp263 = and i1 %tmp261, %tmp262 395 br i1 %tmp263, label %ENDIF2795, label %ELSE2800 396 397 ENDIF2795: ; preds = %ELSE2824, %ELSE2821, %ELSE2818, %ELSE2815, %ELSE2812, %ELSE2809, %ELSE2806, %ELSE2803, %ELSE2800, %ELSE2797, %ENDIF2564 398 br label %LOOP 399 400 ELSE2800: ; preds = %ELSE2797 401 br i1 undef, label %ENDIF2795, label %ELSE2803 402 403 ELSE2803: ; preds = %ELSE2800 404 %tmp264 = fsub float %tmp20, undef 405 %tmp265 = fsub float %tmp21, undef 406 %tmp266 = fsub float %tmp22, undef 407 %tmp267 = fmul float %tmp264, undef 408 %tmp268 = fmul float %tmp265, undef 409 %tmp269 = fmul float %tmp266, 0.000000e+00 410 %tmp270 = fsub float %tmp23, undef 411 %tmp271 = fsub float %tmp24, undef 412 %tmp272 = fsub float %tmp25, undef 413 %tmp273 = fmul float %tmp270, undef 414 %tmp274 = fmul float %tmp271, undef 415 %tmp275 = fmul float %tmp272, undef 416 %tmp276 = call float @llvm.minnum.f32(float %tmp267, float %tmp273) 417 %tmp277 = call float @llvm.maxnum.f32(float %tmp268, float %tmp274) 418 %tmp278 = call float @llvm.maxnum.f32(float %tmp269, float %tmp275) 419 %tmp279 = call float @llvm.maxnum.f32(float %tmp276, float undef) 420 %tmp280 = call float @llvm.maxnum.f32(float %tmp279, float undef) 421 %tmp281 = call float @llvm.minnum.f32(float undef, float %tmp277) 422 %tmp282 = call float @llvm.minnum.f32(float %tmp281, float %tmp278) 423 %tmp283 = fcmp ogt float %tmp280, 0.000000e+00 424 %tmp284 = fcmp olt float %tmp280, 1.000000e+00 425 %tmp285 = and i1 %tmp283, %tmp284 426 %tmp286 = fcmp olt float %tmp280, %tmp282 427 %tmp287 = and i1 %tmp285, %tmp286 428 br i1 %tmp287, label %ENDIF2795, label %ELSE2806 429 430 ELSE2806: ; preds = %ELSE2803 431 %tmp288 = fsub float %tmp26, undef 432 %tmp289 = fsub float %tmp27, undef 433 %tmp290 = fsub float %tmp28, undef 434 %tmp291 = fmul float %tmp288, undef 435 %tmp292 = fmul float %tmp289, 0.000000e+00 436 %tmp293 = fmul float %tmp290, undef 437 %tmp294 = fsub float %tmp29, undef 438 %tmp295 = fmul float %tmp294, undef 439 %tmp296 = call float @llvm.minnum.f32(float %tmp291, float %tmp295) 440 %tmp297 = call float @llvm.minnum.f32(float %tmp292, float undef) 441 %tmp298 = call float @llvm.maxnum.f32(float %tmp293, float undef) 442 %tmp299 = call float @llvm.maxnum.f32(float %tmp296, float %tmp297) 443 %tmp300 = call float @llvm.maxnum.f32(float %tmp299, float undef) 444 %tmp301 = call float @llvm.minnum.f32(float undef, float %tmp298) 445 %tmp302 = fcmp ogt float %tmp300, 0.000000e+00 446 %tmp303 = fcmp olt float %tmp300, 1.000000e+00 447 %tmp304 = and i1 %tmp302, %tmp303 448 %tmp305 = fcmp olt float %tmp300, %tmp301 449 %tmp306 = and i1 %tmp304, %tmp305 450 br i1 %tmp306, label %ENDIF2795, label %ELSE2809 451 452 ELSE2809: ; preds = %ELSE2806 453 br i1 undef, label %ENDIF2795, label %ELSE2812 454 455 ELSE2812: ; preds = %ELSE2809 456 br i1 undef, label %ENDIF2795, label %ELSE2815 457 458 ELSE2815: ; preds = %ELSE2812 459 br i1 undef, label %ENDIF2795, label %ELSE2818 460 461 ELSE2818: ; preds = %ELSE2815 462 br i1 undef, label %ENDIF2795, label %ELSE2821 463 464 ELSE2821: ; preds = %ELSE2818 465 %tmp307 = fsub float %tmp56, undef 466 %tmp308 = fsub float %tmp57, undef 467 %tmp309 = fsub float %tmp58, undef 468 %tmp310 = fmul float %tmp307, undef 469 %tmp311 = fmul float %tmp308, 0.000000e+00 470 %tmp312 = fmul float %tmp309, undef 471 %tmp313 = fsub float %tmp59, undef 472 %tmp314 = fsub float %tmp60, undef 473 %tmp315 = fsub float %tmp61, undef 474 %tmp316 = fmul float %tmp313, undef 475 %tmp317 = fmul float %tmp314, undef 476 %tmp318 = fmul float %tmp315, undef 477 %tmp319 = call float @llvm.maxnum.f32(float %tmp310, float %tmp316) 478 %tmp320 = call float @llvm.maxnum.f32(float %tmp311, float %tmp317) 479 %tmp321 = call float @llvm.maxnum.f32(float %tmp312, float %tmp318) 480 %tmp322 = call float @llvm.minnum.f32(float %tmp319, float %tmp320) 481 %tmp323 = call float @llvm.minnum.f32(float %tmp322, float %tmp321) 482 %tmp324 = fcmp ogt float undef, 0.000000e+00 483 %tmp325 = fcmp olt float undef, 1.000000e+00 484 %tmp326 = and i1 %tmp324, %tmp325 485 %tmp327 = fcmp olt float undef, %tmp323 486 %tmp328 = and i1 %tmp326, %tmp327 487 br i1 %tmp328, label %ENDIF2795, label %ELSE2824 488 489 ELSE2824: ; preds = %ELSE2821 490 %.2849 = select i1 undef, float 0.000000e+00, float 1.000000e+00 491 br label %ENDIF2795 492 } 493 494 declare float @llvm.floor.f32(float) #1 495 declare float @llvm.sqrt.f32(float) #1 496 declare float @llvm.minnum.f32(float, float) #1 497 declare float @llvm.maxnum.f32(float, float) #1 498 declare i32 @llvm.amdgcn.mbcnt.lo(i32, i32) #1 499 declare void @llvm.amdgcn.exp.f32(i32, i32, float, float, float, float, i1, i1) #0 500 declare float @llvm.SI.load.const.v4i32(<4 x i32>, i32) #1 501 502 attributes #0 = { nounwind } 503 attributes #1 = { nounwind readnone } 504
