Source AIR SHA256: f97705b88f5491057df2117aef457ffca84e11020574cd57742b0670db4484d8
Line numbers refer to the complete original metal-objdump text. Static IR; original shader not executed.

583: define internal fastcc <2 x float> @_ZL25mssFaceShapePinCoordinateDv2_fPU11MTLconstantK16MSSFaceThinPointi18MSSFaceShapeParamsS_(<2 x float> %0, %struct.MSSFaceThinPoint addrspace(2)* nocapture readonly %1, i32 %2, float %3, float %4, float %5, float %6, float %7, float %8, float %9, float %10, float %11, float %12, <2 x float> %13) unnamed_addr #2 {
584:   %15 = alloca [17 x <3 x i32>], align 16
585:   %16 = alloca [16 x <2 x i32>], align 8
586:   %17 = alloca [4 x <2 x i32>], align 8
587:   %18 = alloca [6 x <3 x i32>], align 16
588:   %19 = alloca [6 x <2 x i32>], align 8
589:   %20 = alloca [3 x <2 x i32>], align 8
590:   %21 = alloca [11 x i32], align 4
591:   %22 = alloca [11 x <2 x float>], align 8
592:   %23 = add nsw i32 %2, 75
593:   %24 = sext i32 %23 to i64
594:   %25 = getelementptr inbounds %struct.MSSFaceThinPoint, %struct.MSSFaceThinPoint addrspace(2)* %1, i64 %24, i32 0
595:   %26 = load <2 x float>, <2 x float> addrspace(2)* %25, align 8, !tbaa !62
596:   %27 = add nsw i32 %2, 59
597:   %28 = sext i32 %27 to i64
598:   %29 = getelementptr inbounds %struct.MSSFaceThinPoint, %struct.MSSFaceThinPoint addrspace(2)* %1, i64 %28, i32 0
599:   %30 = load <2 x float>, <2 x float> addrspace(2)* %29, align 8, !tbaa !62
600:   %31 = add nsw i32 %2, 67
601:   %32 = sext i32 %31 to i64
602:   %33 = getelementptr inbounds %struct.MSSFaceThinPoint, %struct.MSSFaceThinPoint addrspace(2)* %1, i64 %32, i32 0
603:   %34 = load <2 x float>, <2 x float> addrspace(2)* %33, align 8, !tbaa !62
604:   %35 = add nsw i32 %2, 74
605:   %36 = sext i32 %35 to i64
606:   %37 = getelementptr inbounds %struct.MSSFaceThinPoint, %struct.MSSFaceThinPoint addrspace(2)* %1, i64 %36, i32 0
607:   %38 = load <2 x float>, <2 x float> addrspace(2)* %37, align 8, !tbaa !62
608:   %39 = add nsw i32 %2, 60
609:   %40 = sext i32 %39 to i64
610:   %41 = getelementptr inbounds %struct.MSSFaceThinPoint, %struct.MSSFaceThinPoint addrspace(2)* %1, i64 %40, i32 0
611:   %42 = load <2 x float>, <2 x float> addrspace(2)* %41, align 8, !tbaa !62
612:   %43 = fadd fast <2 x float> %42, %38
613:   %44 = fmul fast <2 x float> %43, <float 5.000000e-01, float 5.000000e-01>
614:   %45 = fsub fast <2 x float> %26, %30
615:   %46 = fmul fast <2 x float> %45, %13
616:   %47 = tail call fast float @air.dot.v2f32(<2 x float> %46, <2 x float> %46) #3
617:   %48 = tail call fast float @air.fast_sqrt.f32(float %47) #3
618:   %49 = fsub fast <2 x float> %34, %44
619:   %50 = fmul fast <2 x float> %49, %13
620:   %51 = tail call fast float @air.dot.v2f32(<2 x float> %50, <2 x float> %50) #3
621:   %52 = tail call fast float @air.fast_sqrt.f32(float %51) #3
622:   %53 = fmul fast float %52, 2.000000e+00
623:   %54 = tail call fast float @air.fast_fmax.f32(float %48, float %53) #3
624:   %55 = tail call fast float @air.fast_fmax.f32(float %54, float 1.000000e+00) #3
625:   %56 = fdiv fast float 2.800000e+02, %55
626:   %57 = tail call fast float @air.fast_clamp.f32(float %56, float 1.000000e+00, float 2.250000e+00) #3
627:   %58 = fmul fast float %57, %3
628:   %59 = fmul fast float %57, %4
629:   %60 = fmul fast float %57, %5
630:   %61 = fmul fast float %57, %6
631:   %62 = fmul fast float %57, %7
632:   %63 = fmul fast float %57, %8
633:   %64 = fmul fast float %57, %9
634:   %65 = tail call fast float @air.fast_fabs.f32(float %58) #3
635:   %66 = fcmp fast ogt float %65, 0x3EB0C6F7A0000000

899: 251:                                              ; preds = %212, %200
900:   %252 = phi <2 x float> [ %248, %212 ], [ %201, %200 ]
901:   %253 = tail call fast float @air.fast_fabs.f32(float %61) #3
902:   %254 = fcmp fast ogt float %253, 0x3EB0C6F7A0000000
903:   br i1 %254, label %255, label %309
904:
905: 255:                                              ; preds = %251
906:   %256 = bitcast [6 x <3 x i32>]* %18 to i8*
907:   call void @llvm.lifetime.start.p0i8(i64 96, i8* nonnull %256) #8
908:   %257 = getelementptr inbounds [6 x <3 x i32>], [6 x <3 x i32>]* %18, i64 0, i64 0
909:   store <3 x i32> <i32 72, i32 71, i32 29>, <3 x i32>* %257, align 16
910:   %258 = getelementptr inbounds [6 x <3 x i32>], [6 x <3 x i32>]* %18, i64 0, i64 1
911:   store <3 x i32> <i32 62, i32 63, i32 29>, <3 x i32>* %258, align 16
912:   %259 = getelementptr inbounds [6 x <3 x i32>], [6 x <3 x i32>]* %18, i64 0, i64 2
913:   store <3 x i32> <i32 71, i32 70, i32 29>, <3 x i32>* %259, align 16
914:   %260 = getelementptr inbounds [6 x <3 x i32>], [6 x <3 x i32>]* %18, i64 0, i64 3
915:   store <3 x i32> <i32 63, i32 64, i32 29>, <3 x i32>* %260, align 16
916:   %261 = getelementptr inbounds [6 x <3 x i32>], [6 x <3 x i32>]* %18, i64 0, i64 4
917:   store <3 x i32> <i32 70, i32 71, i32 29>, <3 x i32>* %261, align 16
918:   %262 = getelementptr inbounds [6 x <3 x i32>], [6 x <3 x i32>]* %18, i64 0, i64 5
919:   store <3 x i32> <i32 64, i32 63, i32 29>, <3 x i32>* %262, align 16
920:   %263 = insertelement <2 x float> undef, float %61, i64 0
921:   %264 = shufflevector <2 x float> %263, <2 x float> undef, <2 x i32> zeroinitializer
922:   br label %266
923:
924: 265:                                              ; preds = %305
925:   call void @llvm.lifetime.end.p0i8(i64 96, i8* nonnull %256) #8
926:   br label %309
927:
928: 266:                                              ; preds = %305, %255
929:   %267 = phi <2 x float> [ %252, %255 ], [ %306, %305 ]
930:   %268 = phi i32 [ 0, %255 ], [ %307, %305 ]
931:   %269 = zext i32 %268 to i64
932:   %270 = getelementptr inbounds [6 x <3 x i32>], [6 x <3 x i32>]* %18, i64 0, i64 %269
933:   %271 = load <3 x i32>, <3 x i32>* %270, align 16
934:   %272 = extractelement <3 x i32> %271, i64 0
935:   %273 = add nsw i32 %272, %2
936:   %274 = sext i32 %273 to i64
937:   %275 = getelementptr inbounds %struct.MSSFaceThinPoint, %struct.MSSFaceThinPoint addrspace(2)* %1, i64 %274, i32 0
938:   %276 = load <2 x float>, <2 x float> addrspace(2)* %275, align 8, !tbaa !62
939:   %277 = extractelement <3 x i32> %271, i64 1
940:   %278 = add nsw i32 %277, %2
941:   %279 = sext i32 %278 to i64
942:   %280 = getelementptr inbounds %struct.MSSFaceThinPoint, %struct.MSSFaceThinPoint addrspace(2)* %1, i64 %279, i32 0
943:   %281 = load <2 x float>, <2 x float> addrspace(2)* %280, align 8, !tbaa !62
944:   %282 = fsub fast <2 x float> %276, %281
945:   %283 = tail call fast float @air.dot.v2f32(<2 x float> %282, <2 x float> %282) #3
946:   %284 = tail call fast float @air.fast_sqrt.f32(float %283) #3
947:   %285 = fcmp fast ugt float %284, 0x3F1A36E2E0000000
948:   br i1 %285, label %286, label %305
949:
950: 286:                                              ; preds = %266
951:   %287 = extractelement <3 x i32> %271, i64 2
952:   %288 = add nsw i32 %287, %2
953:   %289 = sext i32 %288 to i64
954:   %290 = getelementptr inbounds %struct.MSSFaceThinPoint, %struct.MSSFaceThinPoint addrspace(2)* %1, i64 %289, i32 0
955:   %291 = load <2 x float>, <2 x float> addrspace(2)* %290, align 8, !tbaa !62
956:   %292 = fsub fast <2 x float> %291, %276
957:   %293 = fsub fast <2 x float> %267, %276
958:   %294 = tail call fast float @air.dot.v2f32(<2 x float> %293, <2 x float> %293) #3
959:   %295 = tail call fast float @air.fast_sqrt.f32(float %294) #3
960:   %296 = fdiv fast float %295, %284
961:   %297 = fsub fast float 1.000000e+00, %296
962:   %298 = tail call fast float @air.fast_clamp.f32(float %297, float 0.000000e+00, float 1.000000e+00) #3
963:   %299 = insertelement <2 x float> undef, float %298, i64 0
964:   %300 = shufflevector <2 x float> %299, <2 x float> undef, <2 x i32> zeroinitializer
965:   %301 = fmul fast <2 x float> %292, %264
966:   %302 = fmul fast <2 x float> %301, %300
967:   %303 = fsub fast <2 x float> %267, %302
968:   %304 = tail call fast <2 x float> @air.fast_clamp.v2f32(<2 x float> %303, <2 x float> zeroinitializer, <2 x float> <float 1.000000e+00, float 1.000000e+00>) #3
969:   br label %305
970:
971: 305:                                              ; preds = %286, %266
972:   %306 = phi <2 x float> [ %304, %286 ], [ %267, %266 ]
973:   %307 = add nuw nsw i32 %268, 1
974:   %308 = icmp eq i32 %307, 6
975:   br i1 %308, label %265, label %266, !llvm.loop !67
976:
977: 309:                                              ; preds = %265, %251
978:   %310 = phi <2 x float> [ %306, %265 ], [ %252, %251 ]
979:   %311 = tail call fast float @air.fast_fabs.f32(float %62) #3
980:   %312 = fcmp fast ogt float %311, 0x3EB0C6F7A0000000
981:   br i1 %312, label %313, label %372
982:
983: 313:                                              ; preds = %309
984:   %314 = bitcast [6 x <2 x i32>]* %19 to i8*
985:   call void @llvm.lifetime.start.p0i8(i64 48, i8* nonnull %314) #8
986:   %315 = getelementptr inbounds [6 x <2 x i32>], [6 x <2 x i32>]* %19, i64 0, i64 0

1127: 422:                                              ; preds = %383, %372
1128:   %423 = phi <2 x float> [ %419, %383 ], [ %373, %372 ]
1129:   %424 = tail call fast float @air.fast_fabs.f32(float %64) #3
1130:   %425 = fcmp fast ogt float %424, 0x3EB0C6F7A0000000
1131:   br i1 %425, label %426, label %585
1132:
1133: 426:                                              ; preds = %422
1134:   %427 = bitcast [11 x i32]* %21 to i8*
1135:   call void @llvm.lifetime.start.p0i8(i64 44, i8* nonnull %427) #8
1136:   %428 = getelementptr inbounds [11 x i32], [11 x i32]* %21, i64 0, i64 0
1137:   store i32 62, i32* %428, align 4
1138:   %429 = getelementptr inbounds [11 x i32], [11 x i32]* %21, i64 0, i64 1
1139:   store i32 63, i32* %429, align 4
1140:   %430 = getelementptr inbounds [11 x i32], [11 x i32]* %21, i64 0, i64 2
1141:   store i32 64, i32* %430, align 4
1142:   %431 = getelementptr inbounds [11 x i32], [11 x i32]* %21, i64 0, i64 3
1143:   store i32 65, i32* %431, align 4
1144:   %432 = getelementptr inbounds [11 x i32], [11 x i32]* %21, i64 0, i64 4
1145:   store i32 66, i32* %432, align 4
1146:   %433 = getelementptr inbounds [11 x i32], [11 x i32]* %21, i64 0, i64 5
1147:   store i32 67, i32* %433, align 4
1148:   %434 = getelementptr inbounds [11 x i32], [11 x i32]* %21, i64 0, i64 6
1149:   store i32 68, i32* %434, align 4
1150:   %435 = getelementptr inbounds [11 x i32], [11 x i32]* %21, i64 0, i64 7
1151:   store i32 69, i32* %435, align 4
1152:   %436 = getelementptr inbounds [11 x i32], [11 x i32]* %21, i64 0, i64 8
1153:   store i32 70, i32* %436, align 4
1154:   %437 = getelementptr inbounds [11 x i32], [11 x i32]* %21, i64 0, i64 9
1155:   store i32 71, i32* %437, align 4
1156:   %438 = getelementptr inbounds [11 x i32], [11 x i32]* %21, i64 0, i64 10
1157:   store i32 72, i32* %438, align 4
1158:   %439 = add nsw i32 %2, 46
1159:   %440 = sext i32 %439 to i64
1160:   %441 = getelementptr inbounds %struct.MSSFaceThinPoint, %struct.MSSFaceThinPoint addrspace(2)* %1, i64 %440, i32 0
1161:   %442 = load <2 x float>, <2 x float> addrspace(2)* %441, align 8, !tbaa !62
1162:   %443 = add nsw i32 %2, 48
1163:   %444 = sext i32 %443 to i64
1164:   %445 = getelementptr inbounds %struct.MSSFaceThinPoint, %struct.MSSFaceThinPoint addrspace(2)* %1, i64 %444, i32 0
1165:   %446 = load <2 x float>, <2 x float> addrspace(2)* %445, align 8, !tbaa !62
1166:   %447 = bitcast [11 x <2 x float>]* %22 to i8*
1167:   call void @llvm.lifetime.start.p0i8(i64 88, i8* nonnull %447) #8
1168:   br label %492
1169:
1170: 448:                                              ; preds = %492
1171:   %449 = fsub fast <2 x float> %442, %446
1172:   %450 = extractelement <2 x float> %423, i64 1
1173:   %451 = extractelement <2 x float> %423, i64 0
1174:   %452 = getelementptr inbounds [11 x <2 x float>], [11 x <2 x float>]* %22, i64 0, i64 10
1175:   %453 = load <2 x float>, <2 x float>* %452, align 8, !tbaa !62
1176:   br label %454
1177:
1178: 454:                                              ; preds = %484, %448
1179:   %455 = phi <2 x float> [ %453, %448 ], [ %460, %484 ]
1180:   %456 = phi i8 [ 0, %448 ], [ %485, %484 ]
1181:   %457 = phi i32 [ 0, %448 ], [ %486, %484 ]
1182:   %458 = zext i32 %457 to i64
1183:   %459 = getelementptr inbounds [11 x <2 x float>], [11 x <2 x float>]* %22, i64 0, i64 %458
1184:   %460 = load <2 x float>, <2 x float>* %459, align 8, !tbaa !62
1185:   %461 = extractelement <2 x float> %455, i64 1
1186:   %462 = extractelement <2 x float> %460, i64 1
1187:   %463 = fcmp fast ule float %462, %450
1188:   %464 = fcmp fast ogt float %461, %450
1189:   %465 = xor i1 %463, %464
1190:   br i1 %465, label %484, label %466
1191:
1192: 466:                                              ; preds = %454
1193:   %467 = fsub fast float %461, %462
1194:   %468 = tail call fast float @air.fast_fabs.f32(float %467) #3
1195:   %469 = fcmp fast ugt float %468, 0x3EB0C6F7A0000000
1196:   %470 = fcmp fast olt float %467, 0.000000e+00
1197:   %471 = select fast i1 %470, float 0xBEB0C6F7A0000000, float 0x3EB0C6F7A0000000
1198:   %472 = select i1 %469, float %467, float %471
1199:   %473 = extractelement <2 x float> %455, i64 0
1200:   %474 = extractelement <2 x float> %460, i64 0
1201:   %475 = fsub fast float %473, %474
1202:   %476 = fsub fast float %450, %462
1203:   %477 = fmul fast float %475, %476
1204:   %478 = fdiv fast float %477, %472
1205:   %479 = fadd fast float %478, %474
1206:   %480 = fcmp fast olt float %451, %479
1207:   br i1 %480, label %481, label %484
1208:
1209: 481:                                              ; preds = %466
1210:   %482 = and i8 %456, 1
1211:   %483 = xor i8 %482, 1
1212:   br label %484
1213:
1214: 484:                                              ; preds = %481, %466, %454
1215:   %485 = phi i8 [ %483, %481 ], [ %456, %466 ], [ %456, %454 ]
1216:   %486 = add nuw nsw i32 %457, 1
1217:   %487 = icmp eq i32 %486, 11
1218:   br i1 %487, label %488, label %454, !llvm.loop !70
1219:
1220: 488:                                              ; preds = %484
1221:   %489 = getelementptr inbounds [11 x <2 x float>], [11 x <2 x float>]* %22, i64 0, i64 0
1222:   %490 = and i8 %485, 1
1223:   %491 = load <2 x float>, <2 x float>* %489, align 8, !tbaa !62
1224:   br label %506
1225:
1226: 492:                                              ; preds = %492, %426
1227:   %493 = phi i32 [ 0, %426 ], [ %502, %492 ]
1228:   %494 = zext i32 %493 to i64
1229:   %495 = getelementptr inbounds [11 x i32], [11 x i32]* %21, i64 0, i64 %494
1230:   %496 = load i32, i32* %495, align 4, !tbaa !53
1231:   %497 = add nsw i32 %496, %2
1232:   %498 = sext i32 %497 to i64
1233:   %499 = getelementptr inbounds %struct.MSSFaceThinPoint, %struct.MSSFaceThinPoint addrspace(2)* %1, i64 %498, i32 0
1234:   %500 = load <2 x float>, <2 x float> addrspace(2)* %499, align 8, !tbaa !62
1235:   %501 = getelementptr inbounds [11 x <2 x float>], [11 x <2 x float>]* %22, i64 0, i64 %494
1236:   store <2 x float> %500, <2 x float>* %501, align 8, !tbaa !62
1237:   %502 = add nuw nsw i32 %493, 1
1238:   %503 = icmp eq i32 %502, 11
1239:   br i1 %503, label %448, label %492, !llvm.loop !71
1240:
1241: 504:                                              ; preds = %531
1242:   %505 = icmp eq i8 %490, 0
1243:   br i1 %505, label %559, label %540
1244:
1245: 506:                                              ; preds = %531, %488
1246:   %507 = phi i32 [ 0, %488 ], [ %513, %531 ]
1247:   %508 = phi <2 x float> [ %491, %488 ], [ %538, %531 ]
1248:   %509 = phi float [ 1.000000e+01, %488 ], [ %537, %531 ]
1249:   %510 = zext i32 %507 to i64
1250:   %511 = getelementptr inbounds [11 x <2 x float>], [11 x <2 x float>]* %22, i64 0, i64 %510
1251:   %512 = load <2 x float>, <2 x float>* %511, align 8, !tbaa !62
1252:   %513 = add nuw nsw i32 %507, 1
1253:   %514 = trunc i32 %513 to i8
1254:   %515 = urem i8 %514, 11
1255:   %516 = zext i8 %515 to i64
1256:   %517 = getelementptr inbounds [11 x <2 x float>], [11 x <2 x float>]* %22, i64 0, i64 %516
1257:   %518 = load <2 x float>, <2 x float>* %517, align 8, !tbaa !62
1258:   %519 = fsub fast <2 x float> %518, %512
1259:   %520 = tail call fast float @air.dot.v2f32(<2 x float> %519, <2 x float> %519) #3
1260:   %521 = fcmp fast ugt float %520, 0x3EB0C6F7A0000000
1261:   br i1 %521, label %522, label %531
1262:
1263: 522:                                              ; preds = %506
1264:   %523 = fsub fast <2 x float> %423, %512
1265:   %524 = tail call fast float @air.dot.v2f32(<2 x float> %523, <2 x float> %519) #3
1266:   %525 = fdiv fast float %524, %520
1267:   %526 = tail call fast float @air.fast_clamp.f32(float %525, float 0.000000e+00, float 1.000000e+00) #3
1268:   %527 = insertelement <2 x float> undef, float %526, i64 0
1269:   %528 = shufflevector <2 x float> %527, <2 x float> undef, <2 x i32> zeroinitializer
1270:   %529 = fmul fast <2 x float> %528, %519
1271:   %530 = fadd fast <2 x float> %529, %512
1272:   br label %531
1273:
1274: 531:                                              ; preds = %522, %506
1275:   %532 = phi <2 x float> [ %530, %522 ], [ %512, %506 ]
1276:   %533 = fsub fast <2 x float> %423, %532
1277:   %534 = tail call fast float @air.dot.v2f32(<2 x float> %533, <2 x float> %533) #3
1278:   %535 = tail call fast float @air.fast_sqrt.f32(float %534) #3
1279:   %536 = fcmp fast olt float %535, %509
1280:   %537 = select i1 %536, float %535, float %509
1281:   %538 = select i1 %536, <2 x float> %532, <2 x float> %508
1282:   %539 = icmp eq i32 %513, 11
1283:   br i1 %539, label %504, label %506, !llvm.loop !72
1284:
1285: 540:                                              ; preds = %504
1286:   %541 = fsub fast <2 x float> <float -0.000000e+00, float -0.000000e+00>, %449
1287:   %542 = tail call fast float @air.dot.v2f32(<2 x float> %541, <2 x float> %541) #3
1288:   %543 = tail call fast float @air.fast_sqrt.f32(float %542) #3
1289:   %544 = fcmp fast ugt float %543, 0x3F1A36E2E0000000
1290:   br i1 %544, label %545, label %583
1291:
1292: 545:                                              ; preds = %540
1293:   %546 = tail call fast float @air.dot.v2f32(<2 x float> zeroinitializer, <2 x float> zeroinitializer) #3
1294:   %547 = tail call fast float @air.fast_sqrt.f32(float %546) #3
1295:   %548 = fdiv fast float %547, %543
1296:   %549 = fsub fast float 1.000000e+00, %548
1297:   %550 = tail call fast float @air.fast_clamp.f32(float %549, float 0.000000e+00, float 1.000000e+00) #3
1298:   %551 = insertelement <2 x float> undef, float %550, i64 0
1299:   %552 = shufflevector <2 x float> %551, <2 x float> undef, <2 x i32> zeroinitializer
1300:   %553 = insertelement <2 x float> undef, float %64, i64 0
1301:   %554 = shufflevector <2 x float> %553, <2 x float> undef, <2 x i32> zeroinitializer
1302:   %555 = fmul fast <2 x float> %449, %554
1303:   %556 = fmul fast <2 x float> %555, %552
1304:   %557 = fsub fast <2 x float> %423, %556
1305:   %558 = tail call fast <2 x float> @air.fast_clamp.v2f32(<2 x float> %557, <2 x float> zeroinitializer, <2 x float> <float 1.000000e+00, float 1.000000e+00>) #3
1306:   br label %583
1307:
1308: 559:                                              ; preds = %504
1309:   %560 = tail call fast float @air.dot.v2f32(<2 x float> %449, <2 x float> %449) #3
1310:   %561 = tail call fast float @air.fast_sqrt.f32(float %560) #3
1311:   %562 = fcmp fast olt float %537, %561
1312:   br i1 %562, label %563, label %583
1313:
1314: 563:                                              ; preds = %559
1315:   %564 = fsub fast <2 x float> <float -0.000000e+00, float -0.000000e+00>, %449
1316:   %565 = tail call fast float @air.dot.v2f32(<2 x float> %564, <2 x float> %564) #3
1317:   %566 = tail call fast float @air.fast_sqrt.f32(float %565) #3
1318:   %567 = fcmp fast ugt float %566, 0x3F1A36E2E0000000
1319:   br i1 %567, label %568, label %583
1320:
1321: 568:                                              ; preds = %563
1322:   %569 = fsub fast <2 x float> %423, %538
1323:   %570 = tail call fast float @air.dot.v2f32(<2 x float> %569, <2 x float> %569) #3
1324:   %571 = tail call fast float @air.fast_sqrt.f32(float %570) #3
1325:   %572 = fdiv fast float %571, %566
1326:   %573 = fsub fast float 1.000000e+00, %572
1327:   %574 = tail call fast float @air.fast_clamp.f32(float %573, float 0.000000e+00, float 1.000000e+00) #3
1328:   %575 = insertelement <2 x float> undef, float %574, i64 0
1329:   %576 = shufflevector <2 x float> %575, <2 x float> undef, <2 x i32> zeroinitializer
1330:   %577 = insertelement <2 x float> undef, float %64, i64 0
1331:   %578 = shufflevector <2 x float> %577, <2 x float> undef, <2 x i32> zeroinitializer
1332:   %579 = fmul fast <2 x float> %449, %578
1333:   %580 = fmul fast <2 x float> %579, %576
1334:   %581 = fsub fast <2 x float> %423, %580
1335:   %582 = tail call fast <2 x float> @air.fast_clamp.v2f32(<2 x float> %581, <2 x float> zeroinitializer, <2 x float> <float 1.000000e+00, float 1.000000e+00>) #3
1336:   br label %583
1337:
1338: 583:                                              ; preds = %568, %563, %559, %545, %540
1339:   %584 = phi <2 x float> [ %423, %559 ], [ %558, %545 ], [ %423, %540 ], [ %582, %568 ], [ %423, %563 ]
1340:   call void @llvm.lifetime.end.p0i8(i64 88, i8* nonnull %447) #8
1341:   call void @llvm.lifetime.end.p0i8(i64 44, i8* nonnull %427) #8
1342:   br label %585
1343:
1344: 585:                                              ; preds = %583, %422
1345:   %586 = phi <2 x float> [ %584, %583 ], [ %423, %422 ]
1346:   %587 = tail call fast float @air.fast_fabs.f32(float %10) #3
1347:   %588 = fcmp fast ogt float %587, 0x3EB0C6F7A0000000
1348:   br i1 %588, label %589, label %613

4634: !27 = !{i32 8, !"air.buffer", !"air.location_index", i32 7, i32 1, !"air.read", !"air.struct_type_info", !28, !"air.arg_type_size", i32 40, !"air.arg_type_align_size", i32 4, !"air.arg_type_name", !"MSSFaceShapeParams", !"air.arg_name", !"faceShapeParams"}
4635: !28 = !{i32 0, i32 4, i32 0, !"float", !"shoulian", i32 4, i32 4, i32 0, !"float", !"zhailian", i32 8, i32 4, i32 0, !"float", !"vlian", i32 12, i32 4, i32 0, !"float", !"xiaegu", i32 16, i32 4, i32 0, !"float", !"quangu", i32 20, i32 4, i32 0, !"float", !"xiaba", i32 24, i32 4, i32 0, !"float", !"duanlian", i32 28, i32 4, i32 0, !"float", !"xiaotou", i32 32, i32 4, i32 0, !"float", !"xiaotoufanwei", i32 36, i32 4, i32 0, !"float", !"dayan"}