mitsuba3 python path integrator dump

crashdump_mit3_path.txt

python customplugins/path_integrator_refined.py 2>&1 crashdump.txt


Critical Dr.Jit compiler failure: jit_llvm_compile(): parsing failed. Please see the LLVM IR and error message below:

define void @drjit_32c09595272bc74a586d2fe8f7e4726e(i64 %start, i64 %end, i8** noalias %params) #0 {
entry:
    %callables = load i8**, i8*** @callables
    %buffer = alloca i8, i32 736, align 32

    br label %body

body:
    %index = phi i64 [ %index_next, %suffix ], [ %start, %entry ]
    %f1_p1 = getelementptr inbounds i8*, i8** %params, i32 3
    %f1_p2 = load i8*, i8** %f1_p1, align 8, !alias.scope !2
    %f1_p3 = bitcast i8* %f1_p2 to float*
    %f1_0 = load float, float* %f1_p3, align 4, !alias.scope !2
    %f1_1 = insertelement <8 x float> undef, float %f1_0, i32 0
    %f1 = shufflevector <8 x float> %f1_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f2_p1 = getelementptr inbounds i8*, i8** %params, i32 4
    %f2_p2 = load i8*, i8** %f2_p1, align 8, !alias.scope !2
    %f2_p3 = bitcast i8* %f2_p2 to float*
    %f2_0 = load float, float* %f2_p3, align 4, !alias.scope !2
    %f2_1 = insertelement <8 x float> undef, float %f2_0, i32 0
    %f2 = shufflevector <8 x float> %f2_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f3_p1 = getelementptr inbounds i8*, i8** %params, i32 5
    %f3_p2 = load i8*, i8** %f3_p1, align 8, !alias.scope !2
    %f3_p3 = bitcast i8* %f3_p2 to float*
    %f3_0 = load float, float* %f3_p3, align 4, !alias.scope !2
    %f3_1 = insertelement <8 x float> undef, float %f3_0, i32 0
    %f3 = shufflevector <8 x float> %f3_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f4_p1 = getelementptr inbounds i8*, i8** %params, i32 6
    %f4_p2 = load i8*, i8** %f4_p1, align 8, !alias.scope !2
    %f4_p3 = bitcast i8* %f4_p2 to float*
    %f4_0 = load float, float* %f4_p3, align 4, !alias.scope !2
    %f4_1 = insertelement <8 x float> undef, float %f4_0, i32 0
    %f4 = shufflevector <8 x float> %f4_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f5_p1 = getelementptr inbounds i8*, i8** %params, i32 7
    %f5_p2 = load i8*, i8** %f5_p1, align 8, !alias.scope !2
    %f5_p3 = bitcast i8* %f5_p2 to float*
    %f5_0 = load float, float* %f5_p3, align 4, !alias.scope !2
    %f5_1 = insertelement <8 x float> undef, float %f5_0, i32 0
    %f5 = shufflevector <8 x float> %f5_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f6_p1 = getelementptr inbounds i8*, i8** %params, i32 8
    %f6_p2 = load i8*, i8** %f6_p1, align 8, !alias.scope !2
    %f6_p3 = bitcast i8* %f6_p2 to float*
    %f6_0 = load float, float* %f6_p3, align 4, !alias.scope !2
    %f6_1 = insertelement <8 x float> undef, float %f6_0, i32 0
    %f6 = shufflevector <8 x float> %f6_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f7_p1 = getelementptr inbounds i8*, i8** %params, i32 9
    %f7_p2 = load i8*, i8** %f7_p1, align 8, !alias.scope !2
    %f7_p3 = bitcast i8* %f7_p2 to float*
    %f7_0 = load float, float* %f7_p3, align 4, !alias.scope !2
    %f7_1 = insertelement <8 x float> undef, float %f7_0, i32 0
    %f7 = shufflevector <8 x float> %f7_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f8_p1 = getelementptr inbounds i8*, i8** %params, i32 10
    %f8_p2 = load i8*, i8** %f8_p1, align 8, !alias.scope !2
    %f8_p3 = bitcast i8* %f8_p2 to float*
    %f8_0 = load float, float* %f8_p3, align 4, !alias.scope !2
    %f8_1 = insertelement <8 x float> undef, float %f8_0, i32 0
    %f8 = shufflevector <8 x float> %f8_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f9_p1 = getelementptr inbounds i8*, i8** %params, i32 11
    %f9_p2 = load i8*, i8** %f9_p1, align 8, !alias.scope !2
    %f9_p3 = bitcast i8* %f9_p2 to float*
    %f9_0 = load float, float* %f9_p3, align 4, !alias.scope !2
    %f9_1 = insertelement <8 x float> undef, float %f9_0, i32 0
    %f9 = shufflevector <8 x float> %f9_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f10_p1 = getelementptr inbounds i8*, i8** %params, i32 12
    %f10_p2 = load i8*, i8** %f10_p1, align 8, !alias.scope !2
    %f10_p3 = bitcast i8* %f10_p2 to float*
    %f10_0 = load float, float* %f10_p3, align 4, !alias.scope !2
    %f10_1 = insertelement <8 x float> undef, float %f10_0, i32 0
    %f10 = shufflevector <8 x float> %f10_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f11_p1 = getelementptr inbounds i8*, i8** %params, i32 13
    %f11_p2 = load i8*, i8** %f11_p1, align 8, !alias.scope !2
    %f11_p3 = bitcast i8* %f11_p2 to float*
    %f11_0 = load float, float* %f11_p3, align 4, !alias.scope !2
    %f11_1 = insertelement <8 x float> undef, float %f11_0, i32 0
    %f11 = shufflevector <8 x float> %f11_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f12_p1 = getelementptr inbounds i8*, i8** %params, i32 14
    %f12_p2 = load i8*, i8** %f12_p1, align 8, !alias.scope !2
    %f12_p3 = bitcast i8* %f12_p2 to float*
    %f12_0 = load float, float* %f12_p3, align 4, !alias.scope !2
    %f12_1 = insertelement <8 x float> undef, float %f12_0, i32 0
    %f12 = shufflevector <8 x float> %f12_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f13_p1 = getelementptr inbounds i8*, i8** %params, i32 15
    %f13_p2 = load i8*, i8** %f13_p1, align 8, !alias.scope !2
    %f13_p3 = bitcast i8* %f13_p2 to float*
    %f13_0 = load float, float* %f13_p3, align 4, !alias.scope !2
    %f13_1 = insertelement <8 x float> undef, float %f13_0, i32 0
    %f13 = shufflevector <8 x float> %f13_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f14_p1 = getelementptr inbounds i8*, i8** %params, i32 16
    %f14_p2 = load i8*, i8** %f14_p1, align 8, !alias.scope !2
    %f14_p3 = bitcast i8* %f14_p2 to float*
    %f14_0 = load float, float* %f14_p3, align 4, !alias.scope !2
    %f14_1 = insertelement <8 x float> undef, float %f14_0, i32 0
    %f14 = shufflevector <8 x float> %f14_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f15_p1 = getelementptr inbounds i8*, i8** %params, i32 17
    %f15_p2 = load i8*, i8** %f15_p1, align 8, !alias.scope !2
    %f15_p3 = bitcast i8* %f15_p2 to float*
    %f15_0 = load float, float* %f15_p3, align 4, !alias.scope !2
    %f15_1 = insertelement <8 x float> undef, float %f15_0, i32 0
    %f15 = shufflevector <8 x float> %f15_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f16_p1 = getelementptr inbounds i8*, i8** %params, i32 18
    %f16_p2 = load i8*, i8** %f16_p1, align 8, !alias.scope !2
    %f16_p3 = bitcast i8* %f16_p2 to float*
    %f16_0 = load float, float* %f16_p3, align 4, !alias.scope !2
    %f16_1 = insertelement <8 x float> undef, float %f16_0, i32 0
    %f16 = shufflevector <8 x float> %f16_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f17_p1 = getelementptr inbounds i8*, i8** %params, i32 19
    %f17_p2 = load i8*, i8** %f17_p1, align 8, !alias.scope !2
    %f17_p3 = bitcast i8* %f17_p2 to float*
    %f17_0 = load float, float* %f17_p3, align 4, !alias.scope !2
    %f17_1 = insertelement <8 x float> undef, float %f17_0, i32 0
    %f17 = shufflevector <8 x float> %f17_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f18_p1 = getelementptr inbounds i8*, i8** %params, i32 20
    %f18_p2 = load i8*, i8** %f18_p1, align 8, !alias.scope !2
    %f18_p3 = bitcast i8* %f18_p2 to float*
    %f18_0 = load float, float* %f18_p3, align 4, !alias.scope !2
    %f18_1 = insertelement <8 x float> undef, float %f18_0, i32 0
    %f18 = shufflevector <8 x float> %f18_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f19_p1 = getelementptr inbounds i8*, i8** %params, i32 21
    %f19_p2 = load i8*, i8** %f19_p1, align 8, !alias.scope !2
    %f19_p3 = bitcast i8* %f19_p2 to float*
    %f19_0 = load float, float* %f19_p3, align 4, !alias.scope !2
    %f19_1 = insertelement <8 x float> undef, float %f19_0, i32 0
    %f19 = shufflevector <8 x float> %f19_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f20_p1 = getelementptr inbounds i8*, i8** %params, i32 22
    %f20_p2 = load i8*, i8** %f20_p1, align 8, !alias.scope !2
    %f20_p3 = bitcast i8* %f20_p2 to float*
    %f20_0 = load float, float* %f20_p3, align 4, !alias.scope !2
    %f20_1 = insertelement <8 x float> undef, float %f20_0, i32 0
    %f20 = shufflevector <8 x float> %f20_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f21_p1 = getelementptr inbounds i8*, i8** %params, i32 23
    %f21_p2 = load i8*, i8** %f21_p1, align 8, !alias.scope !2
    %f21_p3 = bitcast i8* %f21_p2 to float*
    %f21_0 = load float, float* %f21_p3, align 4, !alias.scope !2
    %f21_1 = insertelement <8 x float> undef, float %f21_0, i32 0
    %f21 = shufflevector <8 x float> %f21_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f22_p1 = getelementptr inbounds i8*, i8** %params, i32 24
    %f22_p2 = load i8*, i8** %f22_p1, align 8, !alias.scope !2
    %f22_p3 = bitcast i8* %f22_p2 to float*
    %f22_0 = load float, float* %f22_p3, align 4, !alias.scope !2
    %f22_1 = insertelement <8 x float> undef, float %f22_0, i32 0
    %f22 = shufflevector <8 x float> %f22_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f23_p1 = getelementptr inbounds i8*, i8** %params, i32 25
    %f23_p2 = load i8*, i8** %f23_p1, align 8, !alias.scope !2
    %f23_p3 = bitcast i8* %f23_p2 to float*
    %f23_0 = load float, float* %f23_p3, align 4, !alias.scope !2
    %f23_1 = insertelement <8 x float> undef, float %f23_0, i32 0
    %f23 = shufflevector <8 x float> %f23_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f24_p1 = getelementptr inbounds i8*, i8** %params, i32 26
    %f24_p2 = load i8*, i8** %f24_p1, align 8, !alias.scope !2
    %f24_p3 = bitcast i8* %f24_p2 to float*
    %f24_0 = load float, float* %f24_p3, align 4, !alias.scope !2
    %f24_1 = insertelement <8 x float> undef, float %f24_0, i32 0
    %f24 = shufflevector <8 x float> %f24_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f25_p1 = getelementptr inbounds i8*, i8** %params, i32 27
    %f25_p2 = load i8*, i8** %f25_p1, align 8, !alias.scope !2
    %f25_p3 = bitcast i8* %f25_p2 to float*
    %f25_0 = load float, float* %f25_p3, align 4, !alias.scope !2
    %f25_1 = insertelement <8 x float> undef, float %f25_0, i32 0
    %f25 = shufflevector <8 x float> %f25_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f26_p1 = getelementptr inbounds i8*, i8** %params, i32 28
    %f26_p2 = load i8*, i8** %f26_p1, align 8, !alias.scope !2
    %f26_p3 = bitcast i8* %f26_p2 to float*
    %f26_0 = load float, float* %f26_p3, align 4, !alias.scope !2
    %f26_1 = insertelement <8 x float> undef, float %f26_0, i32 0
    %f26 = shufflevector <8 x float> %f26_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f27_p1 = getelementptr inbounds i8*, i8** %params, i32 29
    %f27_p2 = load i8*, i8** %f27_p1, align 8, !alias.scope !2
    %f27_p3 = bitcast i8* %f27_p2 to float*
    %f27_0 = load float, float* %f27_p3, align 4, !alias.scope !2
    %f27_1 = insertelement <8 x float> undef, float %f27_0, i32 0
    %f27 = shufflevector <8 x float> %f27_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f28_p1 = getelementptr inbounds i8*, i8** %params, i32 30
    %f28_p2 = load i8*, i8** %f28_p1, align 8, !alias.scope !2
    %f28_p3 = bitcast i8* %f28_p2 to float*
    %f28_0 = load float, float* %f28_p3, align 4, !alias.scope !2
    %f28_1 = insertelement <8 x float> undef, float %f28_0, i32 0
    %f28 = shufflevector <8 x float> %f28_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f29_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f29 = shufflevector <8 x float> %f29_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f30_1 = insertelement <8 x float> undef, float 0x3ff0000000000000, i32 0
    %f30 = shufflevector <8 x float> %f30_1, <8 x float> undef, <8 x i32> zeroinitializer
    %r31_0 = trunc i64 %index to i32
    %r31_1 = insertelement <8 x i32> undef, i32 %r31_0, i32 0
    %r31_2 = shufflevector <8 x i32> %r31_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r31 = add <8 x i32> %r31_2, <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
    %r32_p1 = getelementptr inbounds i8*, i8** %params, i32 31
    %r32_p2 = load i8*, i8** %r32_p1, align 8, !alias.scope !2
    %r32_p3 = bitcast i8* %r32_p2 to i32*
    %r32_0 = load i32, i32* %r32_p3, align 4, !alias.scope !2
    %r32_1 = insertelement <8 x i32> undef, i32 %r32_0, i32 0
    %r32 = shufflevector <8 x i32> %r32_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r33_1 = insertelement <8 x i32> undef, i32 2654435769, i32 0
    %r33 = shufflevector <8 x i32> %r33_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r34_1 = insertelement <8 x i32> undef, i32 4, i32 0
    %r34 = shufflevector <8 x i32> %r34_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r35 = shl <8 x i32> %r31, %r34
    %r36_1 = insertelement <8 x i32> undef, i32 2738958700, i32 0
    %r36 = shufflevector <8 x i32> %r36_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r37 = add <8 x i32> %r35, %r36
    %r38 = add <8 x i32> %r31, %r33
    %r39 = xor <8 x i32> %r37, %r38
    %r40_1 = insertelement <8 x i32> undef, i32 5, i32 0
    %r40 = shufflevector <8 x i32> %r40_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r41 = lshr <8 x i32> %r31, %r40
    %r42_1 = insertelement <8 x i32> undef, i32 3355524772, i32 0
    %r42 = shufflevector <8 x i32> %r42_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r43 = add <8 x i32> %r41, %r42
    %r44 = xor <8 x i32> %r39, %r43
    %r45 = add <8 x i32> %r32, %r44
    %r46 = shl <8 x i32> %r45, %r34
    %r47_1 = insertelement <8 x i32> undef, i32 2911926141, i32 0
    %r47 = shufflevector <8 x i32> %r47_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r48 = add <8 x i32> %r46, %r47
    %r49 = add <8 x i32> %r45, %r33
    %r50 = xor <8 x i32> %r48, %r49
    %r51 = lshr <8 x i32> %r45, %r40
    %r52_1 = insertelement <8 x i32> undef, i32 2123724318, i32 0
    %r52 = shufflevector <8 x i32> %r52_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r53 = add <8 x i32> %r51, %r52
    %r54 = xor <8 x i32> %r50, %r53
    %r55 = add <8 x i32> %r31, %r54
    %r56_1 = insertelement <8 x i32> undef, i32 1013904242, i32 0
    %r56 = shufflevector <8 x i32> %r56_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r57 = shl <8 x i32> %r55, %r34
    %r58 = add <8 x i32> %r57, %r36
    %r59 = add <8 x i32> %r55, %r56
    %r60 = xor <8 x i32> %r58, %r59
    %r61 = lshr <8 x i32> %r55, %r40
    %r62 = add <8 x i32> %r61, %r42
    %r63 = xor <8 x i32> %r60, %r62
    %r64 = add <8 x i32> %r45, %r63
    %r65 = shl <8 x i32> %r64, %r34
    %r66 = add <8 x i32> %r65, %r47
    %r67 = add <8 x i32> %r64, %r56
    %r68 = xor <8 x i32> %r66, %r67
    %r69 = lshr <8 x i32> %r64, %r40
    %r70 = add <8 x i32> %r69, %r52
    %r71 = xor <8 x i32> %r68, %r70
    %r72 = add <8 x i32> %r55, %r71
    %r73_1 = insertelement <8 x i32> undef, i32 3668340011, i32 0
    %r73 = shufflevector <8 x i32> %r73_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r74 = shl <8 x i32> %r72, %r34
    %r75 = add <8 x i32> %r74, %r36
    %r76 = add <8 x i32> %r72, %r73
    %r77 = xor <8 x i32> %r75, %r76
    %r78 = lshr <8 x i32> %r72, %r40
    %r79 = add <8 x i32> %r78, %r42
    %r80 = xor <8 x i32> %r77, %r79
    %r81 = add <8 x i32> %r64, %r80
    %r82 = shl <8 x i32> %r81, %r34
    %r83 = add <8 x i32> %r82, %r47
    %r84 = add <8 x i32> %r81, %r73
    %r85 = xor <8 x i32> %r83, %r84
    %r86 = lshr <8 x i32> %r81, %r40
    %r87 = add <8 x i32> %r86, %r52
    %r88 = xor <8 x i32> %r85, %r87
    %r89 = add <8 x i32> %r72, %r88
    %r90_1 = insertelement <8 x i32> undef, i32 2027808484, i32 0
    %r90 = shufflevector <8 x i32> %r90_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r91 = shl <8 x i32> %r89, %r34
    %r92 = add <8 x i32> %r91, %r36
    %r93 = add <8 x i32> %r89, %r90
    %r94 = xor <8 x i32> %r92, %r93
    %r95 = lshr <8 x i32> %r89, %r40
    %r96 = add <8 x i32> %r95, %r42
    %r97 = xor <8 x i32> %r94, %r96
    %r98 = add <8 x i32> %r81, %r97
    %r99 = shl <8 x i32> %r98, %r34
    %r100 = add <8 x i32> %r99, %r47
    %r101 = add <8 x i32> %r98, %r90
    %r102 = xor <8 x i32> %r100, %r101
    %r103 = lshr <8 x i32> %r98, %r40
    %r104 = add <8 x i32> %r103, %r52
    %r105 = xor <8 x i32> %r102, %r104
    %r106 = add <8 x i32> %r89, %r105
    %rd107 = zext <8 x i32> %r98 to <8 x i64>
    %rd108 = zext <8 x i32> %r106 to <8 x i64>
    %rd109_1 = insertelement <8 x i64> undef, i64 0, i32 0
    %rd109 = shufflevector <8 x i64> %rd109_1, <8 x i64> undef, <8 x i32> zeroinitializer
    %rd110_1 = insertelement <8 x i64> undef, i64 1, i32 0
    %rd110 = shufflevector <8 x i64> %rd110_1, <8 x i64> undef, <8 x i32> zeroinitializer
    %rd111 = shl <8 x i64> %rd108, %rd110
    %rd112 = or <8 x i64> %rd111, %rd110
    %rd113_1 = insertelement <8 x i64> undef, i64 6364136223846793005, i32 0
    %rd113 = shufflevector <8 x i64> %rd113_1, <8 x i64> undef, <8 x i32> zeroinitializer
    %rd114_0 = mul <8 x i64> %rd109, %rd113
    %rd114 = add <8 x i64> %rd114_0, %rd112
    %rd115 = add <8 x i64> %rd114, %rd107
    %rd116_0 = mul <8 x i64> %rd115, %rd113
    %rd116 = add <8 x i64> %rd116_0, %rd112
    %r117_p1 = getelementptr inbounds i8*, i8** %params, i32 32
    %r117_p2 = load i8*, i8** %r117_p1, align 8, !alias.scope !2
    %r117_p3 = bitcast i8* %r117_p2 to i32*
    %r117_0 = load i32, i32* %r117_p3, align 4, !alias.scope !2
    %r117_1 = insertelement <8 x i32> undef, i32 %r117_0, i32 0
    %r117 = shufflevector <8 x i32> %r117_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r118 = lshr <8 x i32> %r31, %r117
    %r119_1 = insertelement <8 x i32> undef, i32 8, i32 0
    %r119 = shufflevector <8 x i32> %r119_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r120 = lshr <8 x i32> %r118, %r119
    %r121_1 = insertelement <8 x i32> undef, i32 256, i32 0
    %r121 = shufflevector <8 x i32> %r121_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r122 = sub <8 x i32> zeroinitializer, %r120
    %r123_0 = mul <8 x i32> %r121, %r122
    %r123 = add <8 x i32> %r123_0, %r118
    %f124 = uitofp <8 x i32> %r123 to <8 x float>
    %f125 = uitofp <8 x i32> %r120 to <8 x float>
    %rd126_0 = mul <8 x i64> %rd116, %rd113
    %rd126 = add <8 x i64> %rd126_0, %rd112
    %rd127_1 = insertelement <8 x i64> undef, i64 18, i32 0
    %rd127 = shufflevector <8 x i64> %rd127_1, <8 x i64> undef, <8 x i32> zeroinitializer
    %rd128 = lshr <8 x i64> %rd116, %rd127
    %rd129 = xor <8 x i64> %rd128, %rd116
    %rd130_1 = insertelement <8 x i64> undef, i64 27, i32 0
    %rd130 = shufflevector <8 x i64> %rd130_1, <8 x i64> undef, <8 x i32> zeroinitializer
    %rd131 = lshr <8 x i64> %rd129, %rd130
    %r132 = trunc <8 x i64> %rd131 to <8 x i32>
    %rd133_1 = insertelement <8 x i64> undef, i64 59, i32 0
    %rd133 = shufflevector <8 x i64> %rd133_1, <8 x i64> undef, <8 x i32> zeroinitializer
    %rd134 = lshr <8 x i64> %rd116, %rd133
    %r135 = trunc <8 x i64> %rd134 to <8 x i32>
    %r136 = lshr <8 x i32> %r132, %r135
    %r137 = bitcast <8 x i32> %r135 to <8 x i32>
    %r138 = sub <8 x i32> zeroinitializer, %r137
    %r139_1 = insertelement <8 x i32> undef, i32 31, i32 0
    %r139 = shufflevector <8 x i32> %r139_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r140 = and <8 x i32> %r138, %r139
    %r141 = bitcast <8 x i32> %r140 to <8 x i32>
    %r142 = shl <8 x i32> %r132, %r141
    %r143 = or <8 x i32> %r136, %r142
    %r144_1 = insertelement <8 x i32> undef, i32 9, i32 0
    %r144 = shufflevector <8 x i32> %r144_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r145 = lshr <8 x i32> %r143, %r144
    %r146_1 = insertelement <8 x i32> undef, i32 1065353216, i32 0
    %r146 = shufflevector <8 x i32> %r146_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r147 = or <8 x i32> %r145, %r146
    %f148 = bitcast <8 x i32> %r147 to <8 x float>
    %f149 = fsub <8 x float> %f148, %f30
    %rd150_0 = mul <8 x i64> %rd126, %rd113
    %rd150 = add <8 x i64> %rd150_0, %rd112
    %rd151 = lshr <8 x i64> %rd126, %rd127
    %rd152 = xor <8 x i64> %rd151, %rd126
    %rd153 = lshr <8 x i64> %rd152, %rd130
    %r154 = trunc <8 x i64> %rd153 to <8 x i32>
    %rd155 = lshr <8 x i64> %rd126, %rd133
    %r156 = trunc <8 x i64> %rd155 to <8 x i32>
    %r157 = lshr <8 x i32> %r154, %r156
    %r158 = bitcast <8 x i32> %r156 to <8 x i32>
    %r159 = sub <8 x i32> zeroinitializer, %r158
    %r160 = and <8 x i32> %r159, %r139
    %r161 = bitcast <8 x i32> %r160 to <8 x i32>
    %r162 = shl <8 x i32> %r154, %r161
    %r163 = or <8 x i32> %r157, %r162
    %r164 = lshr <8 x i32> %r163, %r144
    %r165 = or <8 x i32> %r164, %r146
    %f166 = bitcast <8 x i32> %r165 to <8 x float>
    %f167 = fsub <8 x float> %f166, %f30
    %f168 = fadd <8 x float> %f124, %f149
    %f169 = fadd <8 x float> %f125, %f167
    %f170_1 = insertelement <8 x float> undef, float 0x3f70000000000000, i32 0
    %f170 = shufflevector <8 x float> %f170_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f171_1 = insertelement <8 x float> undef, float 0x8000000000000000, i32 0
    %f171 = shufflevector <8 x float> %f171_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f172 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f168, <8 x float> %f170, <8 x float> %f171)
    %f173 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f169, <8 x float> %f170, <8 x float> %f171)
    %f174 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f13, <8 x float> %f172, <8 x float> %f25)
    %f175 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f14, <8 x float> %f172, <8 x float> %f26)
    %f176 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f15, <8 x float> %f172, <8 x float> %f27)
    %f177 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f16, <8 x float> %f172, <8 x float> %f28)
    %f178 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f17, <8 x float> %f173, <8 x float> %f174)
    %f179 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f18, <8 x float> %f173, <8 x float> %f175)
    %f180 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f19, <8 x float> %f173, <8 x float> %f176)
    %f181 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f20, <8 x float> %f173, <8 x float> %f177)
    %f182 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f21, <8 x float> %f29, <8 x float> %f178)
    %f183 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f22, <8 x float> %f29, <8 x float> %f179)
    %f184 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f23, <8 x float> %f29, <8 x float> %f180)
    %f185 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f24, <8 x float> %f29, <8 x float> %f181)
    %f186 = fdiv <8 x float> %f30, %f185
    %f187 = fmul <8 x float> %f182, %f186
    %f188 = fmul <8 x float> %f183, %f186
    %f189 = fmul <8 x float> %f184, %f186
    %f190 = fmul <8 x float> %f187, %f187
    %f191 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f188, <8 x float> %f188, <8 x float> %f190)
    %f192 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f189, <8 x float> %f189, <8 x float> %f191)
    %f193 = fdiv <8 x float> %f30, %f192
    %f194 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> %f193)
    %f195 = fmul <8 x float> %f187, %f194
    %f196 = fmul <8 x float> %f188, %f194
    %f197 = fmul <8 x float> %f189, %f194
    %f198 = fmul <8 x float> %f1, %f195
    %f199 = fmul <8 x float> %f2, %f195
    %f200 = fmul <8 x float> %f3, %f195
    %f201 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f4, <8 x float> %f196, <8 x float> %f198)
    %f202 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f5, <8 x float> %f196, <8 x float> %f199)
    %f203 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f6, <8 x float> %f196, <8 x float> %f200)
    %f204 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f7, <8 x float> %f197, <8 x float> %f201)
    %f205 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f8, <8 x float> %f197, <8 x float> %f202)
    %f206 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f9, <8 x float> %f197, <8 x float> %f203)
    %f207 = fdiv <8 x float> %f30, %f197
    %f208_1 = insertelement <8 x float> undef, float 0x3f50624de0000000, i32 0
    %f208 = shufflevector <8 x float> %f208_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f209 = fmul <8 x float> %f208, %f207
    %f210_1 = insertelement <8 x float> undef, float 0x4059000000000000, i32 0
    %f210 = shufflevector <8 x float> %f210_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f211 = fmul <8 x float> %f210, %f207
    %f212 = fmul <8 x float> %f204, %f209
    %f213 = fmul <8 x float> %f205, %f209
    %f214 = fmul <8 x float> %f206, %f209
    %f215 = fadd <8 x float> %f10, %f212
    %f216 = fadd <8 x float> %f11, %f213
    %f217 = fadd <8 x float> %f12, %f214
    %f218 = fsub <8 x float> %f211, %f209
    %r219_1 = insertelement <8 x i32> undef, i32 0, i32 0
    %r219 = shufflevector <8 x i32> %r219_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %f220_1 = insertelement <8 x float> undef, float 0x7ff0000000000000, i32 0
    %f220 = shufflevector <8 x float> %f220_1, <8 x float> undef, <8 x i32> zeroinitializer
    %p221_1 = insertelement <8 x i1> undef, i1 1, i32 0
    %p221 = shufflevector <8 x i1> %p221_1, <8 x i1> undef, <8 x i32> zeroinitializer
    %r222_1 = insertelement <8 x i32> undef, i32 0, i32 0
    %r222 = shufflevector <8 x i32> %r222_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %f223_1 = insertelement <8 x float> undef, float 0x3ff0000000000000, i32 0
    %f223 = shufflevector <8 x float> %f223_1, <8 x float> undef, <8 x i32> zeroinitializer
    %p224_1 = insertelement <8 x i1> undef, i1 1, i32 0
    %p224 = shufflevector <8 x i1> %p224_1, <8 x i1> undef, <8 x i32> zeroinitializer
    %f225_1 = insertelement <8 x float> undef, float 0x3ff0000000000000, i32 0
    %f225 = shufflevector <8 x float> %f225_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f226_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f226 = shufflevector <8 x float> %f226_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f227_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f227 = shufflevector <8 x float> %f227_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f228_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f228 = shufflevector <8 x float> %f228_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f229_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f229 = shufflevector <8 x float> %f229_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f230_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f230 = shufflevector <8 x float> %f230_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f231_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f231 = shufflevector <8 x float> %f231_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f232_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f232 = shufflevector <8 x float> %f232_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f233_1 = insertelement <8 x float> undef, float 0x3ff0000000000000, i32 0
    %f233 = shufflevector <8 x float> %f233_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f234_1 = insertelement <8 x float> undef, float 0x3ff0000000000000, i32 0
    %f234 = shufflevector <8 x float> %f234_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f235_1 = insertelement <8 x float> undef, float 0x3ff0000000000000, i32 0
    %f235 = shufflevector <8 x float> %f235_1, <8 x float> undef, <8 x i32> zeroinitializer
    br label %l_236_start

l_236_start:
    br label %l_236_cond

l_236_cond: ; Loop (Path Tracer)
    %rd237 = phi <8 x i64> [ %rd237_final, %l_236_tail ], [ %rd150, %l_236_start ]
    %f238 = phi <8 x float> [ %f238_final, %l_236_tail ], [ %f215, %l_236_start ]
    %f239 = phi <8 x float> [ %f239_final, %l_236_tail ], [ %f216, %l_236_start ]
    %f240 = phi <8 x float> [ %f240_final, %l_236_tail ], [ %f217, %l_236_start ]
    %f241 = phi <8 x float> [ %f241_final, %l_236_tail ], [ %f204, %l_236_start ]
    %f242 = phi <8 x float> [ %f242_final, %l_236_tail ], [ %f205, %l_236_start ]
    %f243 = phi <8 x float> [ %f243_final, %l_236_tail ], [ %f206, %l_236_start ]
    %f244 = phi <8 x float> [ %f244_final, %l_236_tail ], [ %f218, %l_236_start ]
    %f245 = phi <8 x float> [ %f245_final, %l_236_tail ], [ %f233, %l_236_start ]
    %f246 = phi <8 x float> [ %f246_final, %l_236_tail ], [ %f234, %l_236_start ]
    %f247 = phi <8 x float> [ %f247_final, %l_236_tail ], [ %f235, %l_236_start ]
    %f248 = phi <8 x float> [ %f248_final, %l_236_tail ], [ %f230, %l_236_start ]
    %f249 = phi <8 x float> [ %f249_final, %l_236_tail ], [ %f231, %l_236_start ]
    %f250 = phi <8 x float> [ %f250_final, %l_236_tail ], [ %f232, %l_236_start ]
    %f251 = phi <8 x float> [ %f251_final, %l_236_tail ], [ %f223, %l_236_start ]
    %r252 = phi <8 x i32> [ %r252_final, %l_236_tail ], [ %r222, %l_236_start ]
    %f253 = phi <8 x float> [ %f253_final, %l_236_tail ], [ %f220, %l_236_start ]
    %f254 = phi <8 x float> [ %f254_final, %l_236_tail ], [ %f226, %l_236_start ]
    %f255 = phi <8 x float> [ %f255_final, %l_236_tail ], [ %f227, %l_236_start ]
    %f256 = phi <8 x float> [ %f256_final, %l_236_tail ], [ %f228, %l_236_start ]
    %f257 = phi <8 x float> [ %f257_final, %l_236_tail ], [ %f225, %l_236_start ]
    %p258 = phi <8 x i1> [ %p258_final, %l_236_tail ], [ %p224, %l_236_start ]
    %p259 = phi <8 x i1> [ %p259_final, %l_236_tail ], [ %p221, %l_236_start ]
    %p236 = call i1 @llvm.experimental.vector.reduce.or.v8i1(<8 x i1> %p259)
    br i1 %p236, label %l_236_body, label %l_236_done

l_236_body:
    %rd261 = phi <8 x i64> [ %rd237, %l_236_cond ]
    %f262 = phi <8 x float> [ %f238, %l_236_cond ]
    %f263 = phi <8 x float> [ %f239, %l_236_cond ]
    %f264 = phi <8 x float> [ %f240, %l_236_cond ]
    %f265 = phi <8 x float> [ %f241, %l_236_cond ]
    %f266 = phi <8 x float> [ %f242, %l_236_cond ]
    %f267 = phi <8 x float> [ %f243, %l_236_cond ]
    %f268 = phi <8 x float> [ %f244, %l_236_cond ]
    %f269 = phi <8 x float> [ %f245, %l_236_cond ]
    %f270 = phi <8 x float> [ %f246, %l_236_cond ]
    %f271 = phi <8 x float> [ %f247, %l_236_cond ]
    %f272 = phi <8 x float> [ %f248, %l_236_cond ]
    %f273 = phi <8 x float> [ %f249, %l_236_cond ]
    %f274 = phi <8 x float> [ %f250, %l_236_cond ]
    %f275 = phi <8 x float> [ %f251, %l_236_cond ]
    %r276 = phi <8 x i32> [ %r252, %l_236_cond ]
    %f277 = phi <8 x float> [ %f254, %l_236_cond ]
    %f278 = phi <8 x float> [ %f255, %l_236_cond ]
    %f279 = phi <8 x float> [ %f256, %l_236_cond ]
    %f280 = phi <8 x float> [ %f257, %l_236_cond ]
    %p281 = phi <8 x i1> [ %p258, %l_236_cond ]
    %p282 = phi <8 x i1> [ %p259, %l_236_cond ]
    %r283_0 = trunc i64 %index to i32
    %r283_1 = insertelement <8 x i32> undef, i32 %r283_0, i32 0
    %r283_2 = shufflevector <8 x i32> %r283_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r283 = add <8 x i32> %r283_2, <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
    %p284_0 = trunc i64 %end to i32
    %p284_1 = insertelement <8 x i32> undef, i32 %p284_0, i32 0
    %p284_2 = shufflevector <8 x i32> %p284_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %p284 = icmp ult <8 x i32> %r283, %p284_2
    %p285 = and <8 x i1> %p284, %p259
    %p286_1 = insertelement <8 x i1> undef, i1 0, i32 0
    %p286 = shufflevector <8 x i1> %p286_1, <8 x i1> undef, <8 x i32> zeroinitializer
    %rd287_p1 = getelementptr inbounds i8*, i8** %params, i32 33
    %rd287 = load i8*, i8** %rd287_p1, align 8, !alias.scope !2
    %rd288_p1 = getelementptr inbounds i8*, i8** %params, i32 34
    %rd288 = load i8*, i8** %rd288_p1, align 8, !alias.scope !2
    %r289_1 = insertelement <8 x i32> undef, i32 0, i32 0
    %r289 = shufflevector <8 x i32> %r289_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %f290_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f290 = shufflevector <8 x float> %f290_1, <8 x float> undef, <8 x i32> zeroinitializer
    %r291_1 = insertelement <8 x i32> undef, i32 4294967295, i32 0
    %r291 = shufflevector <8 x i32> %r291_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r292_1 = insertelement <8 x i32> undef, i32 0, i32 0
    %r292 = shufflevector <8 x i32> %r292_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r293 = select <8 x i1> %p285, <8 x i32> %r291, <8 x i32> zeroinitializer
    %f294_1 = insertelement <8 x float> undef, float 0x7ff0000000000000, i32 0
    %f294 = shufflevector <8 x float> %f294_1, <8 x float> undef, <8 x i32> zeroinitializer
    %r295_1 = insertelement <8 x i32> undef, i32 4294967295, i32 0
    %r295 = shufflevector <8 x i32> %r295_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %rd296_p1 = getelementptr inbounds i8*, i8** %params, i32 35
    %rd296 = load i8*, i8** %rd296_p1, align 8, !alias.scope !2
    %f297_1 = insertelement <8 x float> undef, float 0x3ff0000000000000, i32 0
    %f297 = shufflevector <8 x float> %f297_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f298_1 = insertelement <8 x float> undef, float 0xbff0000000000000, i32 0
    %f298 = shufflevector <8 x float> %f298_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f299 = fneg <8 x float> %f265
    %f300 = fneg <8 x float> %f266
    %f301 = fneg <8 x float> %f267
    %p302 = xor <8 x i1> %p281, <i1 1, i1 1, i1 1, i1 1, i1 1, i1 1, i1 1, i1 1>
    %f303 = fmul <8 x float> %f280, %f280
    %p304 = fcmp ogt <8 x float> %f303, %f290
    %r305_1 = insertelement <8 x i32> undef, i32 1, i32 0
    %r305 = shufflevector <8 x i32> %r305_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r306 = add <8 x i32> %r276, %r305
    %r307_1 = insertelement <8 x i32> undef, i32 6, i32 0
    %r307 = shufflevector <8 x i32> %r307_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %p308 = icmp ult <8 x i32> %r306, %r307
    %r309_1 = insertelement <8 x i32> undef, i32 30, i32 0
    %r309 = shufflevector <8 x i32> %r309_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %rd310_1 = insertelement <8 x i64> undef, i64 6364136223846793005, i32 0
    %rd310 = shufflevector <8 x i64> %rd310_1, <8 x i64> undef, <8 x i32> zeroinitializer
    %rd311_0 = mul <8 x i64> %rd261, %rd310
    %rd311 = add <8 x i64> %rd311_0, %rd112
    %rd312_1 = insertelement <8 x i64> undef, i64 18, i32 0
    %rd312 = shufflevector <8 x i64> %rd312_1, <8 x i64> undef, <8 x i32> zeroinitializer
    %rd313 = lshr <8 x i64> %rd261, %rd312
    %rd314 = xor <8 x i64> %rd313, %rd261
    %rd315_1 = insertelement <8 x i64> undef, i64 27, i32 0
    %rd315 = shufflevector <8 x i64> %rd315_1, <8 x i64> undef, <8 x i32> zeroinitializer
    %rd316 = lshr <8 x i64> %rd314, %rd315
    %r317 = trunc <8 x i64> %rd316 to <8 x i32>
    %rd318_1 = insertelement <8 x i64> undef, i64 59, i32 0
    %rd318 = shufflevector <8 x i64> %rd318_1, <8 x i64> undef, <8 x i32> zeroinitializer
    %rd319 = lshr <8 x i64> %rd261, %rd318
    %r320 = trunc <8 x i64> %rd319 to <8 x i32>
    %r321 = lshr <8 x i32> %r317, %r320
    %r322 = bitcast <8 x i32> %r320 to <8 x i32>
    %r323 = sub <8 x i32> zeroinitializer, %r322
    %r324_1 = insertelement <8 x i32> undef, i32 31, i32 0
    %r324 = shufflevector <8 x i32> %r324_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r325 = and <8 x i32> %r323, %r324
    %r326 = bitcast <8 x i32> %r325 to <8 x i32>
    %r327 = shl <8 x i32> %r317, %r326
    %r328 = or <8 x i32> %r321, %r327
    %r329_1 = insertelement <8 x i32> undef, i32 9, i32 0
    %r329 = shufflevector <8 x i32> %r329_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r330 = lshr <8 x i32> %r328, %r329
    %r331_1 = insertelement <8 x i32> undef, i32 1065353216, i32 0
    %r331 = shufflevector <8 x i32> %r331_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r332 = or <8 x i32> %r330, %r331
    %f333 = bitcast <8 x i32> %r332 to <8 x float>
    %f334 = fsub <8 x float> %f333, %f297
    %rd335_p1 = getelementptr inbounds i8*, i8** %params, i32 36
    %rd335 = load i8*, i8** %rd335_p1, align 8, !alias.scope !2
    %f336_1 = insertelement <8 x float> undef, float 0x7fffffffe0000000, i32 0
    %f336 = shufflevector <8 x float> %f336_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f337_1 = insertelement <8 x float> undef, float 0x3f17700000000000, i32 0
    %f337 = shufflevector <8 x float> %f337_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f338_1 = insertelement <8 x float> undef, float 0x3feff8ad00000000, i32 0
    %f338 = shufflevector <8 x float> %f338_1, <8 x float> undef, <8 x i32> zeroinitializer
    %rd339_p1 = getelementptr inbounds i8*, i8** %params, i32 37
    %rd339 = load i8*, i8** %rd339_p1, align 8, !alias.scope !2
    %f340_1 = insertelement <8 x float> undef, float 0x3fee666660000000, i32 0
    %f340 = shufflevector <8 x float> %f340_1, <8 x float> undef, <8 x i32> zeroinitializer
    %r341_1 = insertelement <8 x i32> undef, i32 5, i32 0
    %r341 = shufflevector <8 x i32> %r341_1, <8 x i32> undef, <8 x i32> zeroinitializer

    ; -------- Ray trace -------
    %u342_in_0_0 = getelementptr inbounds i8, i8* %buffer, i32 0
    %u342_in_0_1 = bitcast i8* %u342_in_0_0 to <8 x i32> *
    store <8 x i32> %r293, <8 x i32>* %u342_in_0_1, align 32
    %u342_in_1_0 = getelementptr inbounds i8, i8* %buffer, i32 32
    %u342_in_1_1 = bitcast i8* %u342_in_1_0 to <8 x float> *
    store <8 x float> %f262, <8 x float>* %u342_in_1_1, align 32
    %u342_in_2_0 = getelementptr inbounds i8, i8* %buffer, i32 64
    %u342_in_2_1 = bitcast i8* %u342_in_2_0 to <8 x float> *
    store <8 x float> %f263, <8 x float>* %u342_in_2_1, align 32
    %u342_in_3_0 = getelementptr inbounds i8, i8* %buffer, i32 96
    %u342_in_3_1 = bitcast i8* %u342_in_3_0 to <8 x float> *
    store <8 x float> %f264, <8 x float>* %u342_in_3_1, align 32
    %u342_in_4_0 = getelementptr inbounds i8, i8* %buffer, i32 128
    %u342_in_4_1 = bitcast i8* %u342_in_4_0 to <8 x float> *
    store <8 x float> %f290, <8 x float>* %u342_in_4_1, align 32
    %u342_in_5_0 = getelementptr inbounds i8, i8* %buffer, i32 160
    %u342_in_5_1 = bitcast i8* %u342_in_5_0 to <8 x float> *
    store <8 x float> %f265, <8 x float>* %u342_in_5_1, align 32
    %u342_in_6_0 = getelementptr inbounds i8, i8* %buffer, i32 192
    %u342_in_6_1 = bitcast i8* %u342_in_6_0 to <8 x float> *
    store <8 x float> %f266, <8 x float>* %u342_in_6_1, align 32
    %u342_in_7_0 = getelementptr inbounds i8, i8* %buffer, i32 224
    %u342_in_7_1 = bitcast i8* %u342_in_7_0 to <8 x float> *
    store <8 x float> %f267, <8 x float>* %u342_in_7_1, align 32
    %u342_in_8_0 = getelementptr inbounds i8, i8* %buffer, i32 256
    %u342_in_8_1 = bitcast i8* %u342_in_8_0 to <8 x float> *
    store <8 x float> %f229, <8 x float>* %u342_in_8_1, align 32
    %u342_in_9_0 = getelementptr inbounds i8, i8* %buffer, i32 288
    %u342_in_9_1 = bitcast i8* %u342_in_9_0 to <8 x float> *
    store <8 x float> %f268, <8 x float>* %u342_in_9_1, align 32
    %u342_in_10_0 = getelementptr inbounds i8, i8* %buffer, i32 320
    %u342_in_10_1 = bitcast i8* %u342_in_10_0 to <8 x i32> *
    store <8 x i32> %r289, <8 x i32>* %u342_in_10_1, align 32
    %u342_in_11_0 = getelementptr inbounds i8, i8* %buffer, i32 352
    %u342_in_11_1 = bitcast i8* %u342_in_11_0 to <8 x i32> *
    store <8 x i32> %r289, <8 x i32>* %u342_in_11_1, align 32
    %u342_in_12_0 = getelementptr inbounds i8, i8* %buffer, i32 384
    %u342_in_12_1 = bitcast i8* %u342_in_12_0 to <8 x i32> *
    store <8 x i32> %r289, <8 x i32>* %u342_in_12_1, align 32
    %u342_in_geomid_0 = getelementptr inbounds i8, i8* %buffer, i32 608
    %u342_in_geomid_1 = bitcast i8* %u342_in_geomid_0 to <8 x i32> *
    store <8 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1>, <8 x i32>* %u342_in_geomid_1, align 32
    %u342_in_ctx_0 = getelementptr inbounds i8, i8* %buffer, i32 672
    %u342_in_ctx_1 = bitcast i8* %u342_in_ctx_0 to <6 x i32> *
    store <6 x i32> <i32 0, i32 0, i32 0, i32 0, i32 -1, i32 0>, <6 x i32>* %u342_in_ctx_1, align 4
    %u342_func = bitcast i8* %rd287 to void (i8*, i8*, i8*, i8*)*
    call void %u342_func(i8* %u342_in_0_0, i8* %rd288, i8* %u342_in_ctx_0, i8* %u342_in_1_0)
    %u342_out_0_0 = getelementptr inbounds i8, i8* %buffer, i32 288
    %u342_out_0_1 = bitcast i8* %u342_out_0_0 to <8 x float> *
    %u342_out_0 = load <8 x float>, <8 x float>* %u342_out_0_1, align 32
    %u342_out_1_0 = getelementptr inbounds i8, i8* %buffer, i32 512
    %u342_out_1_1 = bitcast i8* %u342_out_1_0 to <8 x float> *
    %u342_out_1 = load <8 x float>, <8 x float>* %u342_out_1_1, align 32
    %u342_out_2_0 = getelementptr inbounds i8, i8* %buffer, i32 544
    %u342_out_2_1 = bitcast i8* %u342_out_2_0 to <8 x float> *
    %u342_out_2 = load <8 x float>, <8 x float>* %u342_out_2_1, align 32
    %u342_out_3_0 = getelementptr inbounds i8, i8* %buffer, i32 576
    %u342_out_3_1 = bitcast i8* %u342_out_3_0 to <8 x i32> *
    %u342_out_3 = load <8 x i32>, <8 x i32>* %u342_out_3_1, align 32
    %u342_out_4_0 = getelementptr inbounds i8, i8* %buffer, i32 608
    %u342_out_4_1 = bitcast i8* %u342_out_4_0 to <8 x i32> *
    %u342_out_4 = load <8 x i32>, <8 x i32>* %u342_out_4_1, align 32
    %u342_out_5_0 = getelementptr inbounds i8, i8* %buffer, i32 640
    %u342_out_5_1 = bitcast i8* %u342_out_5_0 to <8 x i32> *
    %u342_out_5 = load <8 x i32>, <8 x i32>* %u342_out_5_1, align 32
    ; -------------------

    %f343 = bitcast <8 x float> %u342_out_0 to <8 x float>
    %f344 = bitcast <8 x float> %u342_out_1 to <8 x float>
    %f345 = bitcast <8 x float> %u342_out_2 to <8 x float>
    %r346 = bitcast <8 x i32> %u342_out_3 to <8 x i32>
    %r347 = bitcast <8 x i32> %u342_out_4 to <8 x i32>
    %r348 = bitcast <8 x i32> %u342_out_5 to <8 x i32>
    %p349 = fcmp one <8 x float> %f343, %f268
    %f350 = select <8 x i1> %p349, <8 x float> %f343, <8 x float> %f294
    %p351 = icmp ne <8 x i32> %r348, %r295
    %p352 = and <8 x i1> %p349, %p351
    %r353 = select <8 x i1> %p352, <8 x i32> %r348, <8 x i32> %r347
    %p354 = and <8 x i1> %p349, %p285
    %r355_0 = bitcast i8* %rd296 to i32*
    %r355_1 = getelementptr i32, i32* %r355_0, <8 x i32> %r353
    %r355 = call <8 x i32> @llvm.masked.gather.v8i32(<8 x i32*> %r355_1, i32 4, <8 x i1> %p354, <8 x i32> zeroinitializer)
    %r356 = select <8 x i1> %p352, <8 x i32> %r355, <8 x i32> zeroinitializer
    %p357 = xor <8 x i1> %p352, <i1 1, i1 1, i1 1, i1 1, i1 1, i1 1, i1 1, i1 1>
    %r358 = select <8 x i1> %p357, <8 x i32> %r355, <8 x i32> zeroinitializer
    %p359 = fcmp one <8 x float> %f350, %f294
    %p360 = icmp eq <8 x i32> %r356, %r289
    %r361 = select <8 x i1> %p360, <8 x i32> %r358, <8 x i32> %r356
    %p362 = icmp ne <8 x i32> %r361, %r289
    %p363 = and <8 x i1> %p359, %p362
    %p364 = and <8 x i1> %p363, %p285
    %f365 = bitcast <8 x float> %f350 to <8 x float>
    %rd366_p1 = getelementptr inbounds i8*, i8** %params, i32 38
    %rd366 = load i8*, i8** %rd366_p1, align 8, !alias.scope !2
    %rd367_p1 = getelementptr inbounds i8*, i8** %params, i32 39
    %rd367 = load i8*, i8** %rd367_p1, align 8, !alias.scope !2
    br label %l368_start

l368_start:
    ; VCall: mitsuba::Shape::compute_surface_interaction()
    ;  - target 1 = @func_bac3cacf8aebcc87b13f4c43e6cdd822;
    ;  - target 2 = @func_570d1d9103e40ae9598d8f351c2e5c2a;
    ;  - target 3 = @func_e17fd895e12cc8bec8c4e55075a069b3;
    ;  - target 4 = @func_dcfca7b5b523c613a93412a4c52b1f7a;

    %u368_self_ptr_0 = bitcast i8* %rd366 to i64*
    %u368_self_ptr = getelementptr i64, i64* %u368_self_ptr_0, <8 x i32> %r361
    %u368_self_combined = call <8 x i64> @llvm.masked.gather.v8i64(<8 x i64*> %u368_self_ptr, i32 8, <8 x i1> %p364, <8 x i64> zeroinitializer)
    %u368_self_initial = trunc <8 x i64> %u368_self_combined to <8 x i32>
    %u368_offset_1 = lshr <8 x i64> %u368_self_combined, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
    %u368_offset = trunc <8 x i64> %u368_offset_1 to <8 x i32>
    %u368_in_0_0 = getelementptr inbounds i8, i8* %buffer, i32 0
    %u368_in_0_1 = bitcast i8* %u368_in_0_0 to <8 x float> *
    store <8 x float> %f262, <8 x float>* %u368_in_0_1, align 32
    %u368_in_1_0 = getelementptr inbounds i8, i8* %buffer, i32 32
    %u368_in_1_1 = bitcast i8* %u368_in_1_0 to <8 x float> *
    store <8 x float> %f263, <8 x float>* %u368_in_1_1, align 32
    %u368_in_2_0 = getelementptr inbounds i8, i8* %buffer, i32 64
    %u368_in_2_1 = bitcast i8* %u368_in_2_0 to <8 x float> *
    store <8 x float> %f264, <8 x float>* %u368_in_2_1, align 32
    %u368_in_3_0 = getelementptr inbounds i8, i8* %buffer, i32 96
    %u368_in_3_1 = bitcast i8* %u368_in_3_0 to <8 x float> *
    store <8 x float> %f265, <8 x float>* %u368_in_3_1, align 32
    %u368_in_4_0 = getelementptr inbounds i8, i8* %buffer, i32 128
    %u368_in_4_1 = bitcast i8* %u368_in_4_0 to <8 x float> *
    store <8 x float> %f266, <8 x float>* %u368_in_4_1, align 32
    %u368_in_5_0 = getelementptr inbounds i8, i8* %buffer, i32 160
    %u368_in_5_1 = bitcast i8* %u368_in_5_0 to <8 x float> *
    store <8 x float> %f267, <8 x float>* %u368_in_5_1, align 32
    %u368_in_6_0 = getelementptr inbounds i8, i8* %buffer, i32 192
    %u368_in_6_1 = bitcast i8* %u368_in_6_0 to <8 x float> *
    store <8 x float> %f350, <8 x float>* %u368_in_6_1, align 32
    %u368_in_7_0 = getelementptr inbounds i8, i8* %buffer, i32 224
    %u368_in_7_1 = bitcast i8* %u368_in_7_0 to <8 x float> *
    store <8 x float> %f344, <8 x float>* %u368_in_7_1, align 32
    %u368_in_8_0 = getelementptr inbounds i8, i8* %buffer, i32 256
    %u368_in_8_1 = bitcast i8* %u368_in_8_0 to <8 x float> *
    store <8 x float> %f345, <8 x float>* %u368_in_8_1, align 32
    %u368_in_9_0 = getelementptr inbounds i8, i8* %buffer, i32 288
    %u368_in_9_1 = bitcast i8* %u368_in_9_0 to <8 x i32> *
    store <8 x i32> %r346, <8 x i32>* %u368_in_9_1, align 32
    %u368_out = getelementptr i8, i8* %buffer, i32 320
    %u368_tmp_0_0 = getelementptr inbounds i8, i8* %u368_out, i64 0
    %u368_tmp_0_1 = bitcast i8* %u368_tmp_0_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u368_tmp_0_1, align 32
    %u368_tmp_2_0 = getelementptr inbounds i8, i8* %u368_out, i64 32
    %u368_tmp_2_1 = bitcast i8* %u368_tmp_2_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u368_tmp_2_1, align 32
    %u368_tmp_3_0 = getelementptr inbounds i8, i8* %u368_out, i64 64
    %u368_tmp_3_1 = bitcast i8* %u368_tmp_3_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u368_tmp_3_1, align 32
    %u368_tmp_4_0 = getelementptr inbounds i8, i8* %u368_out, i64 96
    %u368_tmp_4_1 = bitcast i8* %u368_tmp_4_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u368_tmp_4_1, align 32
    %u368_tmp_5_0 = getelementptr inbounds i8, i8* %u368_out, i64 128
    %u368_tmp_5_1 = bitcast i8* %u368_tmp_5_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u368_tmp_5_1, align 32
    %u368_tmp_6_0 = getelementptr inbounds i8, i8* %u368_out, i64 160
    %u368_tmp_6_1 = bitcast i8* %u368_tmp_6_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u368_tmp_6_1, align 32
    %u368_tmp_7_0 = getelementptr inbounds i8, i8* %u368_out, i64 192
    %u368_tmp_7_1 = bitcast i8* %u368_tmp_7_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u368_tmp_7_1, align 32
    %u368_tmp_8_0 = getelementptr inbounds i8, i8* %u368_out, i64 224
    %u368_tmp_8_1 = bitcast i8* %u368_tmp_8_0 to <8 x i32> *
    store <8 x i32> zeroinitializer, <8 x i32>* %u368_tmp_8_1, align 32
    %u368_tmp_13_0 = getelementptr inbounds i8, i8* %u368_out, i64 256
    %u368_tmp_13_1 = bitcast i8* %u368_tmp_13_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u368_tmp_13_1, align 32
    %u368_tmp_14_0 = getelementptr inbounds i8, i8* %u368_out, i64 288
    %u368_tmp_14_1 = bitcast i8* %u368_tmp_14_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u368_tmp_14_1, align 32
    %u368_tmp_15_0 = getelementptr inbounds i8, i8* %u368_out, i64 320
    %u368_tmp_15_1 = bitcast i8* %u368_tmp_15_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u368_tmp_15_1, align 32
    %u368_tmp_16_0 = getelementptr inbounds i8, i8* %u368_out, i64 352
    %u368_tmp_16_1 = bitcast i8* %u368_tmp_16_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u368_tmp_16_1, align 32
    %u368_tmp_17_0 = getelementptr inbounds i8, i8* %u368_out, i64 384
    %u368_tmp_17_1 = bitcast i8* %u368_tmp_17_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u368_tmp_17_1, align 32
    br label %l368_check

l368_check:
    %u368_self = phi <8 x i32> [ %u368_self_initial, %l368_start ], [ %u368_self_next, %l368_call ]
    %u368_next = call i32 @llvm.experimental.vector.reduce.umax.v8i32(<8 x i32> %u368_self)
    %u368_valid = icmp ne i32 %u368_next, 0
    br i1 %u368_valid, label %l368_call, label %l368_end

l368_call:
    %u368_bcast_0 = insertelement <8 x i32> undef, i32 %u368_next, i32 0
    %u368_bcast = shufflevector <8 x i32> %u368_bcast_0, <8 x i32> undef, <8 x i32> zeroinitializer
    %u368_active = icmp eq <8 x i32> %u368_self, %u368_bcast
    %u368_func_0 = getelementptr inbounds i8*, i8** %callables, i32 %u368_next
    %u368_func_1 = load i8*, i8** %u368_func_0
    %u368_func = bitcast i8* %u368_func_1 to void (<8 x i1>, <8 x i32>, i8*, i8*, <8 x i32>)*
    call void %u368_func(<8 x i1> %u368_active, <8 x i32> %r361, i8* %buffer, i8* %rd367, <8 x i32> %u368_offset)
    %u368_self_next = select <8 x i1> %u368_active, <8 x i32> zeroinitializer, <8 x i32> %u368_self
    br label %l368_check

l368_end:
    %u368_out_0_0 = getelementptr inbounds i8, i8* %u368_out, i64 0
    %u368_out_0_1 = bitcast i8* %u368_out_0_0 to <8 x float> *
    %f381 = load <8 x float>, <8 x float>* %u368_out_0_1, align 32
    %u368_out_2_0 = getelementptr inbounds i8, i8* %u368_out, i64 32
    %u368_out_2_1 = bitcast i8* %u368_out_2_0 to <8 x float> *
    %f370 = load <8 x float>, <8 x float>* %u368_out_2_1, align 32
    %u368_out_3_0 = getelementptr inbounds i8, i8* %u368_out, i64 64
    %u368_out_3_1 = bitcast i8* %u368_out_3_0 to <8 x float> *
    %f371 = load <8 x float>, <8 x float>* %u368_out_3_1, align 32
    %u368_out_4_0 = getelementptr inbounds i8, i8* %u368_out, i64 96
    %u368_out_4_1 = bitcast i8* %u368_out_4_0 to <8 x float> *
    %f372 = load <8 x float>, <8 x float>* %u368_out_4_1, align 32
    %u368_out_5_0 = getelementptr inbounds i8, i8* %u368_out, i64 128
    %u368_out_5_1 = bitcast i8* %u368_out_5_0 to <8 x float> *
    %f373 = load <8 x float>, <8 x float>* %u368_out_5_1, align 32
    %u368_out_6_0 = getelementptr inbounds i8, i8* %u368_out, i64 160
    %u368_out_6_1 = bitcast i8* %u368_out_6_0 to <8 x float> *
    %f374 = load <8 x float>, <8 x float>* %u368_out_6_1, align 32
    %u368_out_7_0 = getelementptr inbounds i8, i8* %u368_out, i64 192
    %u368_out_7_1 = bitcast i8* %u368_out_7_0 to <8 x float> *
    %f375 = load <8 x float>, <8 x float>* %u368_out_7_1, align 32
    %u368_out_8_0 = getelementptr inbounds i8, i8* %u368_out, i64 224
    %u368_out_8_1 = bitcast i8* %u368_out_8_0 to <8 x i32> *
    %r376 = load <8 x i32>, <8 x i32>* %u368_out_8_1, align 32
    %u368_out_13_0 = getelementptr inbounds i8, i8* %u368_out, i64 256
    %u368_out_13_1 = bitcast i8* %u368_out_13_0 to <8 x float> *
    %f382 = load <8 x float>, <8 x float>* %u368_out_13_1, align 32
    %u368_out_14_0 = getelementptr inbounds i8, i8* %u368_out, i64 288
    %u368_out_14_1 = bitcast i8* %u368_out_14_0 to <8 x float> *
    %f380 = load <8 x float>, <8 x float>* %u368_out_14_1, align 32
    %u368_out_15_0 = getelementptr inbounds i8, i8* %u368_out, i64 320
    %u368_out_15_1 = bitcast i8* %u368_out_15_0 to <8 x float> *
    %f379 = load <8 x float>, <8 x float>* %u368_out_15_1, align 32
    %u368_out_16_0 = getelementptr inbounds i8, i8* %u368_out, i64 352
    %u368_out_16_1 = bitcast i8* %u368_out_16_0 to <8 x float> *
    %f378 = load <8 x float>, <8 x float>* %u368_out_16_1, align 32
    %u368_out_17_0 = getelementptr inbounds i8, i8* %u368_out, i64 384
    %u368_out_17_1 = bitcast i8* %u368_out_17_0 to <8 x float> *
    %f377 = load <8 x float>, <8 x float>* %u368_out_17_1, align 32
    br label %l368_done

l368_done:
    %f369 = select <8 x i1> %p364, <8 x float> %f365, <8 x float> zeroinitializer
    %p383 = xor <8 x i1> %p359, <i1 1, i1 1, i1 1, i1 1, i1 1, i1 1, i1 1, i1 1>
    %f384 = select <8 x i1> %p383, <8 x float> %f294, <8 x float> %f369
    %p385 = fcmp one <8 x float> %f384, %f294
    %p386 = and <8 x i1> %p359, %p385
    %p387 = xor <8 x i1> %p386, <i1 1, i1 1, i1 1, i1 1, i1 1, i1 1, i1 1, i1 1>
    %r388 = select <8 x i1> %p387, <8 x i32> %r289, <8 x i32> %r376
    %f389 = fmul <8 x float> %f377, %f380
    %f390 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f378, <8 x float> %f381, <8 x float> %f389)
    %f391 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f379, <8 x float> %f382, <8 x float> %f390)
    %f392 = fneg <8 x float> %f391
    %f393 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f377, <8 x float> %f392, <8 x float> %f380)
    %f394 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f378, <8 x float> %f392, <8 x float> %f381)
    %f395 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f379, <8 x float> %f392, <8 x float> %f382)
    %f396 = fmul <8 x float> %f393, %f393
    %f397 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f394, <8 x float> %f394, <8 x float> %f396)
    %f398 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f395, <8 x float> %f395, <8 x float> %f397)
    %f399 = fdiv <8 x float> %f297, %f398
    %f400 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> %f399)
    %f401 = fmul <8 x float> %f393, %f400
    %f402 = fmul <8 x float> %f394, %f400
    %f403 = fmul <8 x float> %f395, %f400
    %p404 = fcmp oeq <8 x float> %f380, %f290
    %p405 = fcmp oeq <8 x float> %f381, %f290
    %p406 = fcmp oeq <8 x float> %f382, %f290
    %p407 = and <8 x i1> %p404, %p405
    %p408 = and <8 x i1> %p407, %p406
    %p409 = fcmp oge <8 x float> %f379, %f290
    %f410 = select <8 x i1> %p409, <8 x float> %f297, <8 x float> %f298
    %f411 = fadd <8 x float> %f410, %f379
    %f412 = fdiv <8 x float> %f297, %f411
    %f413 = fneg <8 x float> %f412
    %f414 = fmul <8 x float> %f377, %f378
    %f415 = fmul <8 x float> %f414, %f413
    %f416 = fmul <8 x float> %f377, %f377
    %f417 = fmul <8 x float> %f416, %f413
    %f418 = fneg <8 x float> %f417
    %f419 = select <8 x i1> %p409, <8 x float> %f417, <8 x float> %f418
    %f420 = fadd <8 x float> %f419, %f297
    %f421 = fneg <8 x float> %f415
    %f422 = select <8 x i1> %p409, <8 x float> %f415, <8 x float> %f421
    %f423 = fneg <8 x float> %f377
    %f424 = select <8 x i1> %p409, <8 x float> %f423, <8 x float> %f377
    %f425 = select <8 x i1> %p408, <8 x float> %f420, <8 x float> %f401
    %f426 = select <8 x i1> %p408, <8 x float> %f422, <8 x float> %f402
    %f427 = select <8 x i1> %p408, <8 x float> %f424, <8 x float> %f403
    %f428 = fmul <8 x float> %f379, %f426
    %f429 = fmul <8 x float> %f377, %f427
    %f430 = fmul <8 x float> %f378, %f425
    %f431 = fneg <8 x float> %f428
    %f432 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f378, <8 x float> %f427, <8 x float> %f431)
    %f433 = fneg <8 x float> %f429
    %f434 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f379, <8 x float> %f425, <8 x float> %f433)
    %f435 = fneg <8 x float> %f430
    %f436 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f377, <8 x float> %f426, <8 x float> %f435)
    %f437 = fmul <8 x float> %f299, %f425
    %f438 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f300, <8 x float> %f426, <8 x float> %f437)
    %f439 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f301, <8 x float> %f427, <8 x float> %f438)
    %f440 = fmul <8 x float> %f299, %f432
    %f441 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f300, <8 x float> %f434, <8 x float> %f440)
    %f442 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f301, <8 x float> %f436, <8 x float> %f441)
    %f443 = fmul <8 x float> %f299, %f377
    %f444 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f300, <8 x float> %f378, <8 x float> %f443)
    %f445 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f301, <8 x float> %f379, <8 x float> %f444)
    %f446 = select <8 x i1> %p386, <8 x float> %f439, <8 x float> %f299
    %f447 = select <8 x i1> %p386, <8 x float> %f442, <8 x float> %f300
    %f448 = select <8 x i1> %p386, <8 x float> %f445, <8 x float> %f301
    %p449 = and <8 x i1> %p385, %p282
    %p450 = icmp ne <8 x i32> %r388, %r289
    %f451 = fsub <8 x float> %f370, %f277
    %f452 = fsub <8 x float> %f371, %f278
    %f453 = fsub <8 x float> %f372, %f279
    %f454 = fmul <8 x float> %f451, %f451
    %f455 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f452, <8 x float> %f452, <8 x float> %f454)
    %f456 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f453, <8 x float> %f453, <8 x float> %f455)
    %f457 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> %f456)
    %f458 = fdiv <8 x float> %f297, %f457
    %f459 = fmul <8 x float> %f451, %f458
    %f460 = fmul <8 x float> %f452, %f458
    %f461 = fmul <8 x float> %f453, %f458
    %f462 = fneg <8 x float> %f446
    %f463 = fneg <8 x float> %f447
    %f464 = fneg <8 x float> %f448
    %f465 = select <8 x i1> %p385, <8 x float> %f459, <8 x float> %f462
    %f466 = select <8 x i1> %p385, <8 x float> %f460, <8 x float> %f463
    %f467 = select <8 x i1> %p385, <8 x float> %f461, <8 x float> %f464
    %p468 = and <8 x i1> %p450, %p285
    %rd469_p1 = getelementptr inbounds i8*, i8** %params, i32 40
    %rd469 = load i8*, i8** %rd469_p1, align 8, !alias.scope !2
    %r470_0 = bitcast i8* %rd469 to i32*
    %r470_1 = getelementptr i32, i32* %r470_0, <8 x i32> %r388
    %r470 = call <8 x i32> @llvm.masked.gather.v8i32(<8 x i32*> %r470_1, i32 4, <8 x i1> %p468, <8 x i32> zeroinitializer)
    %p471 = icmp ne <8 x i32> %r470, %r289
    %p472 = and <8 x i1> %p302, %p471
    %p473 = and <8 x i1> %p472, %p285
    %rd474_p1 = getelementptr inbounds i8*, i8** %params, i32 41
    %rd474 = load i8*, i8** %rd474_p1, align 8, !alias.scope !2
    %rd475_p1 = getelementptr inbounds i8*, i8** %params, i32 42
    %rd475 = load i8*, i8** %rd475_p1, align 8, !alias.scope !2
    br label %l476_start

l476_start:
    ; VCall: mitsuba::Emitter::pdf_direction()
    ;  - target 1 = @func_bf888b7a279d5208759c322da4df8aa9;

    %u476_self_ptr_0 = bitcast i8* %rd474 to i64*
    %u476_self_ptr = getelementptr i64, i64* %u476_self_ptr_0, <8 x i32> %r470
    %u476_self_combined = call <8 x i64> @llvm.masked.gather.v8i64(<8 x i64*> %u476_self_ptr, i32 8, <8 x i1> %p473, <8 x i64> zeroinitializer)
    %u476_self_initial = trunc <8 x i64> %u476_self_combined to <8 x i32>
    %u476_offset_1 = lshr <8 x i64> %u476_self_combined, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
    %u476_offset = trunc <8 x i64> %u476_offset_1 to <8 x i32>
    %u476_in_0_0 = getelementptr inbounds i8, i8* %buffer, i32 0
    %u476_in_0_1 = bitcast i8* %u476_in_0_0 to <8 x float> *
    store <8 x float> %f377, <8 x float>* %u476_in_0_1, align 32
    %u476_in_1_0 = getelementptr inbounds i8, i8* %buffer, i32 32
    %u476_in_1_1 = bitcast i8* %u476_in_1_0 to <8 x float> *
    store <8 x float> %f378, <8 x float>* %u476_in_1_1, align 32
    %u476_in_2_0 = getelementptr inbounds i8, i8* %buffer, i32 64
    %u476_in_2_1 = bitcast i8* %u476_in_2_0 to <8 x float> *
    store <8 x float> %f379, <8 x float>* %u476_in_2_1, align 32
    %u476_in_3_0 = getelementptr inbounds i8, i8* %buffer, i32 96
    %u476_in_3_1 = bitcast i8* %u476_in_3_0 to <8 x float> *
    store <8 x float> %f465, <8 x float>* %u476_in_3_1, align 32
    %u476_in_4_0 = getelementptr inbounds i8, i8* %buffer, i32 128
    %u476_in_4_1 = bitcast i8* %u476_in_4_0 to <8 x float> *
    store <8 x float> %f466, <8 x float>* %u476_in_4_1, align 32
    %u476_in_5_0 = getelementptr inbounds i8, i8* %buffer, i32 160
    %u476_in_5_1 = bitcast i8* %u476_in_5_0 to <8 x float> *
    store <8 x float> %f467, <8 x float>* %u476_in_5_1, align 32
    %u476_in_6_0 = getelementptr inbounds i8, i8* %buffer, i32 192
    %u476_in_6_1 = bitcast i8* %u476_in_6_0 to <8 x float> *
    store <8 x float> %f457, <8 x float>* %u476_in_6_1, align 32
    %u476_out = getelementptr i8, i8* %buffer, i32 224
    %u476_tmp_0_0 = getelementptr inbounds i8, i8* %u476_out, i64 0
    %u476_tmp_0_1 = bitcast i8* %u476_tmp_0_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u476_tmp_0_1, align 32
    br label %l476_check

l476_check:
    %u476_self = phi <8 x i32> [ %u476_self_initial, %l476_start ], [ %u476_self_next, %l476_call ]
    %u476_next = call i32 @llvm.experimental.vector.reduce.umax.v8i32(<8 x i32> %u476_self)
    %u476_valid = icmp ne i32 %u476_next, 0
    br i1 %u476_valid, label %l476_call, label %l476_end

l476_call:
    %u476_bcast_0 = insertelement <8 x i32> undef, i32 %u476_next, i32 0
    %u476_bcast = shufflevector <8 x i32> %u476_bcast_0, <8 x i32> undef, <8 x i32> zeroinitializer
    %u476_active = icmp eq <8 x i32> %u476_self, %u476_bcast
    %u476_func_0 = getelementptr inbounds i8*, i8** %callables, i32 %u476_next
    %u476_func_1 = load i8*, i8** %u476_func_0
    %u476_func = bitcast i8* %u476_func_1 to void (<8 x i1>, i8*, i8*, <8 x i32>)*
    call void %u476_func(<8 x i1> %u476_active, i8* %buffer, i8* %rd475, <8 x i32> %u476_offset)
    %u476_self_next = select <8 x i1> %u476_active, <8 x i32> zeroinitializer, <8 x i32> %u476_self
    br label %l476_check

l476_end:
    %u476_out_0_0 = getelementptr inbounds i8, i8* %u476_out, i64 0
    %u476_out_0_1 = bitcast i8* %u476_out_0_0 to <8 x float> *
    %f477 = load <8 x float>, <8 x float>* %u476_out_0_1, align 32
    br label %l476_done

l476_done:
    %f478 = fmul <8 x float> %f477, %f477
    %f479 = fadd <8 x float> %f303, %f478
    %f480 = fdiv <8 x float> %f303, %f479
    %f481 = select <8 x i1> %p304, <8 x float> %f480, <8 x float> zeroinitializer
    %p482 = and <8 x i1> %p304, %p471
    %p483 = and <8 x i1> %p482, %p285
    %rd484_p1 = getelementptr inbounds i8*, i8** %params, i32 43
    %rd484 = load i8*, i8** %rd484_p1, align 8, !alias.scope !2
    %rd485_p1 = getelementptr inbounds i8*, i8** %params, i32 44
    %rd485 = load i8*, i8** %rd485_p1, align 8, !alias.scope !2
    br label %l486_start

l486_start:
    ; VCall: mitsuba::Emitter::eval()
    ;  - target 1 = @func_972f6c492a7ec6284cf60b520b91aba3;

    %u486_self_ptr_0 = bitcast i8* %rd484 to i64*
    %u486_self_ptr = getelementptr i64, i64* %u486_self_ptr_0, <8 x i32> %r470
    %u486_self_combined = call <8 x i64> @llvm.masked.gather.v8i64(<8 x i64*> %u486_self_ptr, i32 8, <8 x i1> %p483, <8 x i64> zeroinitializer)
    %u486_self_initial = trunc <8 x i64> %u486_self_combined to <8 x i32>
    %u486_offset_1 = lshr <8 x i64> %u486_self_combined, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
    %u486_offset = trunc <8 x i64> %u486_offset_1 to <8 x i32>
    %u486_in_0_0 = getelementptr inbounds i8, i8* %buffer, i32 0
    %u486_in_0_1 = bitcast i8* %u486_in_0_0 to <8 x float> *
    store <8 x float> %f377, <8 x float>* %u486_in_0_1, align 32
    %u486_in_1_0 = getelementptr inbounds i8, i8* %buffer, i32 32
    %u486_in_1_1 = bitcast i8* %u486_in_1_0 to <8 x float> *
    store <8 x float> %f378, <8 x float>* %u486_in_1_1, align 32
    %u486_in_2_0 = getelementptr inbounds i8, i8* %buffer, i32 64
    %u486_in_2_1 = bitcast i8* %u486_in_2_0 to <8 x float> *
    store <8 x float> %f379, <8 x float>* %u486_in_2_1, align 32
    %u486_in_3_0 = getelementptr inbounds i8, i8* %buffer, i32 96
    %u486_in_3_1 = bitcast i8* %u486_in_3_0 to <8 x float> *
    store <8 x float> %f448, <8 x float>* %u486_in_3_1, align 32
    %u486_in_4_0 = getelementptr inbounds i8, i8* %buffer, i32 128
    %u486_in_4_1 = bitcast i8* %u486_in_4_0 to <8 x i32> *
    store <8 x i32> %r346, <8 x i32>* %u486_in_4_1, align 32
    %u486_out = getelementptr i8, i8* %buffer, i32 160
    %u486_tmp_0_0 = getelementptr inbounds i8, i8* %u486_out, i64 0
    %u486_tmp_0_1 = bitcast i8* %u486_tmp_0_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u486_tmp_0_1, align 32
    %u486_tmp_1_0 = getelementptr inbounds i8, i8* %u486_out, i64 32
    %u486_tmp_1_1 = bitcast i8* %u486_tmp_1_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u486_tmp_1_1, align 32
    %u486_tmp_2_0 = getelementptr inbounds i8, i8* %u486_out, i64 64
    %u486_tmp_2_1 = bitcast i8* %u486_tmp_2_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u486_tmp_2_1, align 32
    br label %l486_check

l486_check:
    %u486_self = phi <8 x i32> [ %u486_self_initial, %l486_start ], [ %u486_self_next, %l486_call ]
    %u486_next = call i32 @llvm.experimental.vector.reduce.umax.v8i32(<8 x i32> %u486_self)
    %u486_valid = icmp ne i32 %u486_next, 0
    br i1 %u486_valid, label %l486_call, label %l486_end

l486_call:
    %u486_bcast_0 = insertelement <8 x i32> undef, i32 %u486_next, i32 0
    %u486_bcast = shufflevector <8 x i32> %u486_bcast_0, <8 x i32> undef, <8 x i32> zeroinitializer
    %u486_active = icmp eq <8 x i32> %u486_self, %u486_bcast
    %u486_func_0 = getelementptr inbounds i8*, i8** %callables, i32 %u486_next
    %u486_func_1 = load i8*, i8** %u486_func_0
    %u486_func = bitcast i8* %u486_func_1 to void (<8 x i1>, i8*, i8*, <8 x i32>)*
    call void %u486_func(<8 x i1> %u486_active, i8* %buffer, i8* %rd485, <8 x i32> %u486_offset)
    %u486_self_next = select <8 x i1> %u486_active, <8 x i32> zeroinitializer, <8 x i32> %u486_self
    br label %l486_check

l486_end:
    %u486_out_0_0 = getelementptr inbounds i8, i8* %u486_out, i64 0
    %u486_out_0_1 = bitcast i8* %u486_out_0_0 to <8 x float> *
    %f487 = load <8 x float>, <8 x float>* %u486_out_0_1, align 32
    %u486_out_1_0 = getelementptr inbounds i8, i8* %u486_out, i64 32
    %u486_out_1_1 = bitcast i8* %u486_out_1_0 to <8 x float> *
    %f488 = load <8 x float>, <8 x float>* %u486_out_1_1, align 32
    %u486_out_2_0 = getelementptr inbounds i8, i8* %u486_out, i64 64
    %u486_out_2_1 = bitcast i8* %u486_out_2_0 to <8 x float> *
    %f489 = load <8 x float>, <8 x float>* %u486_out_2_1, align 32
    br label %l486_done

l486_done:
    %f490 = fmul <8 x float> %f487, %f481
    %f491 = fmul <8 x float> %f488, %f481
    %f492 = fmul <8 x float> %f489, %f481
    %f493 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f269, <8 x float> %f490, <8 x float> %f272)
    %f494 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f270, <8 x float> %f491, <8 x float> %f273)
    %f495 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f271, <8 x float> %f492, <8 x float> %f274)
    %p496 = and <8 x i1> %p308, %p385
    %rd497_p1 = getelementptr inbounds i8*, i8** %params, i32 45
    %rd497 = load i8*, i8** %rd497_p1, align 8, !alias.scope !2
    %r498_0 = bitcast i8* %rd497 to i32*
    %r498_1 = getelementptr i32, i32* %r498_0, <8 x i32> %r388
    %r498 = call <8 x i32> @llvm.masked.gather.v8i32(<8 x i32*> %r498_1, i32 4, <8 x i1> %p468, <8 x i32> zeroinitializer)
    %p499 = icmp ne <8 x i32> %r498, %r289
    %p500 = and <8 x i1> %p499, %p285
    %rd501_p1 = getelementptr inbounds i8*, i8** %params, i32 46
    %rd501 = load i8*, i8** %rd501_p1, align 8, !alias.scope !2
    %r502_0 = bitcast i8* %rd501 to i32*
    %r502_1 = getelementptr i32, i32* %r502_0, <8 x i32> %r498
    %r502 = call <8 x i32> @llvm.masked.gather.v8i32(<8 x i32*> %r502_1, i32 4, <8 x i1> %p500, <8 x i32> zeroinitializer)
    %r503 = and <8 x i32> %r502, %r309
    %p504 = icmp ne <8 x i32> %r503, %r289
    %p505 = and <8 x i1> %p496, %p504
    %rd506 = select <8 x i1> %p505, <8 x i64> %rd311, <8 x i64> %rd261
    %rd507_0 = mul <8 x i64> %rd506, %rd310
    %rd507 = add <8 x i64> %rd507_0, %rd112
    %rd508 = select <8 x i1> %p505, <8 x i64> %rd507, <8 x i64> %rd506
    %rd509 = lshr <8 x i64> %rd506, %rd312
    %rd510 = xor <8 x i64> %rd509, %rd506
    %rd511 = lshr <8 x i64> %rd510, %rd315
    %r512 = trunc <8 x i64> %rd511 to <8 x i32>
    %rd513 = lshr <8 x i64> %rd506, %rd318
    %r514 = trunc <8 x i64> %rd513 to <8 x i32>
    %r515 = lshr <8 x i32> %r512, %r514
    %r516 = bitcast <8 x i32> %r514 to <8 x i32>
    %r517 = sub <8 x i32> zeroinitializer, %r516
    %r518 = and <8 x i32> %r517, %r324
    %r519 = bitcast <8 x i32> %r518 to <8 x i32>
    %r520 = shl <8 x i32> %r512, %r519
    %r521 = or <8 x i32> %r515, %r520
    %r522 = lshr <8 x i32> %r521, %r329
    %r523 = or <8 x i32> %r522, %r331
    %f524 = bitcast <8 x i32> %r523 to <8 x float>
    %f525 = fsub <8 x float> %f524, %f297
    %p526 = and <8 x i1> %p505, %p285
    %r527_0 = bitcast i8* %rd335 to i32*
    %r527_1 = getelementptr i32, i32* %r527_0, <8 x i32> %r289
    %r527 = call <8 x i32> @llvm.masked.gather.v8i32(<8 x i32*> %r527_1, i32 4, <8 x i1> %p526, <8 x i32> zeroinitializer)
    %p528 = icmp ne <8 x i32> %r527, %r289
    %p529 = and <8 x i1> %p505, %p528
    %p530 = and <8 x i1> %p529, %p285
    %rd531_p1 = getelementptr inbounds i8*, i8** %params, i32 47
    %rd531 = load i8*, i8** %rd531_p1, align 8, !alias.scope !2
    %rd532_p1 = getelementptr inbounds i8*, i8** %params, i32 48
    %rd532 = load i8*, i8** %rd532_p1, align 8, !alias.scope !2
    br label %l533_start

l533_start:
    ; VCall: mitsuba::Emitter::sample_direction()
    ;  - target 1 = @func_19ee132d0bd6c9e577b7250c22c38555;

    %u533_self_ptr_0 = bitcast i8* %rd531 to i64*
    %u533_self_ptr = getelementptr i64, i64* %u533_self_ptr_0, <8 x i32> %r527
    %u533_self_combined = call <8 x i64> @llvm.masked.gather.v8i64(<8 x i64*> %u533_self_ptr, i32 8, <8 x i1> %p530, <8 x i64> zeroinitializer)
    %u533_self_initial = trunc <8 x i64> %u533_self_combined to <8 x i32>
    %u533_offset_1 = lshr <8 x i64> %u533_self_combined, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
    %u533_offset = trunc <8 x i64> %u533_offset_1 to <8 x i32>
    %u533_in_0_0 = getelementptr inbounds i8, i8* %buffer, i32 0
    %u533_in_0_1 = bitcast i8* %u533_in_0_0 to <8 x float> *
    store <8 x float> %f370, <8 x float>* %u533_in_0_1, align 32
    %u533_in_1_0 = getelementptr inbounds i8, i8* %buffer, i32 32
    %u533_in_1_1 = bitcast i8* %u533_in_1_0 to <8 x float> *
    store <8 x float> %f371, <8 x float>* %u533_in_1_1, align 32
    %u533_in_2_0 = getelementptr inbounds i8, i8* %buffer, i32 64
    %u533_in_2_1 = bitcast i8* %u533_in_2_0 to <8 x float> *
    store <8 x float> %f372, <8 x float>* %u533_in_2_1, align 32
    %u533_in_3_0 = getelementptr inbounds i8, i8* %buffer, i32 96
    %u533_in_3_1 = bitcast i8* %u533_in_3_0 to <8 x float> *
    store <8 x float> %f334, <8 x float>* %u533_in_3_1, align 32
    %u533_in_4_0 = getelementptr inbounds i8, i8* %buffer, i32 128
    %u533_in_4_1 = bitcast i8* %u533_in_4_0 to <8 x float> *
    store <8 x float> %f525, <8 x float>* %u533_in_4_1, align 32
    %u533_out = getelementptr i8, i8* %buffer, i32 160
    %u533_tmp_0_0 = getelementptr inbounds i8, i8* %u533_out, i64 0
    %u533_tmp_0_1 = bitcast i8* %u533_tmp_0_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u533_tmp_0_1, align 32
    %u533_tmp_1_0 = getelementptr inbounds i8, i8* %u533_out, i64 32
    %u533_tmp_1_1 = bitcast i8* %u533_tmp_1_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u533_tmp_1_1, align 32
    %u533_tmp_2_0 = getelementptr inbounds i8, i8* %u533_out, i64 64
    %u533_tmp_2_1 = bitcast i8* %u533_tmp_2_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u533_tmp_2_1, align 32
    %u533_tmp_8_0 = getelementptr inbounds i8, i8* %u533_out, i64 96
    %u533_tmp_8_1 = bitcast i8* %u533_tmp_8_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u533_tmp_8_1, align 32
    %u533_tmp_9_0 = getelementptr inbounds i8, i8* %u533_out, i64 128
    %u533_tmp_9_1 = bitcast i8* %u533_tmp_9_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u533_tmp_9_1, align 32
    %u533_tmp_10_0 = getelementptr inbounds i8, i8* %u533_out, i64 160
    %u533_tmp_10_1 = bitcast i8* %u533_tmp_10_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u533_tmp_10_1, align 32
    %u533_tmp_11_0 = getelementptr inbounds i8, i8* %u533_out, i64 192
    %u533_tmp_11_1 = bitcast i8* %u533_tmp_11_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u533_tmp_11_1, align 32
    %u533_tmp_14_0 = getelementptr inbounds i8, i8* %u533_out, i64 224
    %u533_tmp_14_1 = bitcast i8* %u533_tmp_14_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u533_tmp_14_1, align 32
    %u533_tmp_15_0 = getelementptr inbounds i8, i8* %u533_out, i64 256
    %u533_tmp_15_1 = bitcast i8* %u533_tmp_15_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u533_tmp_15_1, align 32
    %u533_tmp_16_0 = getelementptr inbounds i8, i8* %u533_out, i64 288
    %u533_tmp_16_1 = bitcast i8* %u533_tmp_16_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u533_tmp_16_1, align 32
    br label %l533_check

l533_check:
    %u533_self = phi <8 x i32> [ %u533_self_initial, %l533_start ], [ %u533_self_next, %l533_call ]
    %u533_next = call i32 @llvm.experimental.vector.reduce.umax.v8i32(<8 x i32> %u533_self)
    %u533_valid = icmp ne i32 %u533_next, 0
    br i1 %u533_valid, label %l533_call, label %l533_end

l533_call:
    %u533_bcast_0 = insertelement <8 x i32> undef, i32 %u533_next, i32 0
    %u533_bcast = shufflevector <8 x i32> %u533_bcast_0, <8 x i32> undef, <8 x i32> zeroinitializer
    %u533_active = icmp eq <8 x i32> %u533_self, %u533_bcast
    %u533_func_0 = getelementptr inbounds i8*, i8** %callables, i32 %u533_next
    %u533_func_1 = load i8*, i8** %u533_func_0
    %u533_func = bitcast i8* %u533_func_1 to void (<8 x i1>, <8 x i32>, i8*, i8*, <8 x i32>)*
    call void %u533_func(<8 x i1> %u533_active, <8 x i32> %r527, i8* %buffer, i8* %rd532, <8 x i32> %u533_offset)
    %u533_self_next = select <8 x i1> %u533_active, <8 x i32> zeroinitializer, <8 x i32> %u533_self
    br label %l533_check

l533_end:
    %u533_out_0_0 = getelementptr inbounds i8, i8* %u533_out, i64 0
    %u533_out_0_1 = bitcast i8* %u533_out_0_0 to <8 x float> *
    %f534 = load <8 x float>, <8 x float>* %u533_out_0_1, align 32
    %u533_out_1_0 = getelementptr inbounds i8, i8* %u533_out, i64 32
    %u533_out_1_1 = bitcast i8* %u533_out_1_0 to <8 x float> *
    %f535 = load <8 x float>, <8 x float>* %u533_out_1_1, align 32
    %u533_out_2_0 = getelementptr inbounds i8, i8* %u533_out, i64 64
    %u533_out_2_1 = bitcast i8* %u533_out_2_0 to <8 x float> *
    %f536 = load <8 x float>, <8 x float>* %u533_out_2_1, align 32
    %u533_out_8_0 = getelementptr inbounds i8, i8* %u533_out, i64 96
    %u533_out_8_1 = bitcast i8* %u533_out_8_0 to <8 x float> *
    %f537 = load <8 x float>, <8 x float>* %u533_out_8_1, align 32
    %u533_out_9_0 = getelementptr inbounds i8, i8* %u533_out, i64 128
    %u533_out_9_1 = bitcast i8* %u533_out_9_0 to <8 x float> *
    %f538 = load <8 x float>, <8 x float>* %u533_out_9_1, align 32
    %u533_out_10_0 = getelementptr inbounds i8, i8* %u533_out, i64 160
    %u533_out_10_1 = bitcast i8* %u533_out_10_0 to <8 x float> *
    %f539 = load <8 x float>, <8 x float>* %u533_out_10_1, align 32
    %u533_out_11_0 = getelementptr inbounds i8, i8* %u533_out, i64 192
    %u533_out_11_1 = bitcast i8* %u533_out_11_0 to <8 x float> *
    %f540 = load <8 x float>, <8 x float>* %u533_out_11_1, align 32
    %u533_out_14_0 = getelementptr inbounds i8, i8* %u533_out, i64 224
    %u533_out_14_1 = bitcast i8* %u533_out_14_0 to <8 x float> *
    %f541 = load <8 x float>, <8 x float>* %u533_out_14_1, align 32
    %u533_out_15_0 = getelementptr inbounds i8, i8* %u533_out, i64 256
    %u533_out_15_1 = bitcast i8* %u533_out_15_0 to <8 x float> *
    %f542 = load <8 x float>, <8 x float>* %u533_out_15_1, align 32
    %u533_out_16_0 = getelementptr inbounds i8, i8* %u533_out, i64 288
    %u533_out_16_1 = bitcast i8* %u533_out_16_0 to <8 x float> *
    %f543 = load <8 x float>, <8 x float>* %u533_out_16_1, align 32
    br label %l533_done

l533_done:
    %p544 = fcmp one <8 x float> %f537, %f290
    %p545 = and <8 x i1> %p505, %p544
    %f546 = fsub <8 x float> %f534, %f370
    %f547 = fsub <8 x float> %f535, %f371
    %f548 = fsub <8 x float> %f536, %f372
    %f549_0 = bitcast <8 x float> %f370 to <8 x i32>
    %f549_1 = bitcast <8 x float> %f336 to <8 x i32>
    %f549_2 = and <8 x i32> %f549_0, %f549_1
    %f549 = bitcast <8 x i32> %f549_2 to <8 x float>
    %f550_0 = bitcast <8 x float> %f371 to <8 x i32>
    %f550_1 = bitcast <8 x float> %f336 to <8 x i32>
    %f550_2 = and <8 x i32> %f550_0, %f550_1
    %f550 = bitcast <8 x i32> %f550_2 to <8 x float>
    %f551_0 = bitcast <8 x float> %f372 to <8 x i32>
    %f551_1 = bitcast <8 x float> %f336 to <8 x i32>
    %f551_2 = and <8 x i32> %f551_0, %f551_1
    %f551 = bitcast <8 x i32> %f551_2 to <8 x float>
    %f552 = call <8 x float> @llvm.maxnum.v8f32(<8 x float> %f549, <8 x float> %f550)
    %f553 = call <8 x float> @llvm.maxnum.v8f32(<8 x float> %f552, <8 x float> %f551)
    %f554 = fadd <8 x float> %f297, %f553
    %f555 = fmul <8 x float> %f554, %f337
    %f556 = fmul <8 x float> %f373, %f546
    %f557 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f374, <8 x float> %f547, <8 x float> %f556)
    %f558 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f375, <8 x float> %f548, <8 x float> %f557)
    %p559 = fcmp oge <8 x float> %f558, %f290
    %f560 = fneg <8 x float> %f555
    %f561 = select <8 x i1> %p559, <8 x float> %f555, <8 x float> %f560
    %f562 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f561, <8 x float> %f373, <8 x float> %f370)
    %f563 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f561, <8 x float> %f374, <8 x float> %f371)
    %f564 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f561, <8 x float> %f375, <8 x float> %f372)
    %f565 = fsub <8 x float> %f534, %f562
    %f566 = fsub <8 x float> %f535, %f563
    %f567 = fsub <8 x float> %f536, %f564
    %f568 = fmul <8 x float> %f565, %f565
    %f569 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f566, <8 x float> %f566, <8 x float> %f568)
    %f570 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f567, <8 x float> %f567, <8 x float> %f569)
    %f571 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> %f570)
    %f572 = fdiv <8 x float> %f297, %f571
    %f573 = fmul <8 x float> %f565, %f572
    %f574 = fmul <8 x float> %f566, %f572
    %f575 = fmul <8 x float> %f567, %f572
    %f576 = fmul <8 x float> %f571, %f338
    %p577 = and <8 x i1> %p545, %p285
    %r578 = select <8 x i1> %p577, <8 x i32> %r291, <8 x i32> zeroinitializer

    ; -------- Ray test -------
    %u579_in_0_0 = getelementptr inbounds i8, i8* %buffer, i32 0
    %u579_in_0_1 = bitcast i8* %u579_in_0_0 to <8 x i32> *
    store <8 x i32> %r578, <8 x i32>* %u579_in_0_1, align 32
    %u579_in_1_0 = getelementptr inbounds i8, i8* %buffer, i32 32
    %u579_in_1_1 = bitcast i8* %u579_in_1_0 to <8 x float> *
    store <8 x float> %f562, <8 x float>* %u579_in_1_1, align 32
    %u579_in_2_0 = getelementptr inbounds i8, i8* %buffer, i32 64
    %u579_in_2_1 = bitcast i8* %u579_in_2_0 to <8 x float> *
    store <8 x float> %f563, <8 x float>* %u579_in_2_1, align 32
    %u579_in_3_0 = getelementptr inbounds i8, i8* %buffer, i32 96
    %u579_in_3_1 = bitcast i8* %u579_in_3_0 to <8 x float> *
    store <8 x float> %f564, <8 x float>* %u579_in_3_1, align 32
    %u579_in_4_0 = getelementptr inbounds i8, i8* %buffer, i32 128
    %u579_in_4_1 = bitcast i8* %u579_in_4_0 to <8 x float> *
    store <8 x float> %f290, <8 x float>* %u579_in_4_1, align 32
    %u579_in_5_0 = getelementptr inbounds i8, i8* %buffer, i32 160
    %u579_in_5_1 = bitcast i8* %u579_in_5_0 to <8 x float> *
    store <8 x float> %f573, <8 x float>* %u579_in_5_1, align 32
    %u579_in_6_0 = getelementptr inbounds i8, i8* %buffer, i32 192
    %u579_in_6_1 = bitcast i8* %u579_in_6_0 to <8 x float> *
    store <8 x float> %f574, <8 x float>* %u579_in_6_1, align 32
    %u579_in_7_0 = getelementptr inbounds i8, i8* %buffer, i32 224
    %u579_in_7_1 = bitcast i8* %u579_in_7_0 to <8 x float> *
    store <8 x float> %f575, <8 x float>* %u579_in_7_1, align 32
    %u579_in_8_0 = getelementptr inbounds i8, i8* %buffer, i32 256
    %u579_in_8_1 = bitcast i8* %u579_in_8_0 to <8 x float> *
    store <8 x float> %f229, <8 x float>* %u579_in_8_1, align 32
    %u579_in_9_0 = getelementptr inbounds i8, i8* %buffer, i32 288
    %u579_in_9_1 = bitcast i8* %u579_in_9_0 to <8 x float> *
    store <8 x float> %f576, <8 x float>* %u579_in_9_1, align 32
    %u579_in_10_0 = getelementptr inbounds i8, i8* %buffer, i32 320
    %u579_in_10_1 = bitcast i8* %u579_in_10_0 to <8 x i32> *
    store <8 x i32> %r289, <8 x i32>* %u579_in_10_1, align 32
    %u579_in_11_0 = getelementptr inbounds i8, i8* %buffer, i32 352
    %u579_in_11_1 = bitcast i8* %u579_in_11_0 to <8 x i32> *
    store <8 x i32> %r289, <8 x i32>* %u579_in_11_1, align 32
    %u579_in_12_0 = getelementptr inbounds i8, i8* %buffer, i32 384
    %u579_in_12_1 = bitcast i8* %u579_in_12_0 to <8 x i32> *
    store <8 x i32> %r289, <8 x i32>* %u579_in_12_1, align 32
    %u579_in_ctx_0 = getelementptr inbounds i8, i8* %buffer, i32 416
    %u579_in_ctx_1 = bitcast i8* %u579_in_ctx_0 to <6 x i32> *
    store <6 x i32> <i32 0, i32 0, i32 0, i32 0, i32 -1, i32 0>, <6 x i32>* %u579_in_ctx_1, align 4
    %u579_func = bitcast i8* %rd339 to void (i8*, i8*, i8*, i8*)*
    call void %u579_func(i8* %u579_in_0_0, i8* %rd288, i8* %u579_in_ctx_0, i8* %u579_in_1_0)
    %u579_out_0_0 = getelementptr inbounds i8, i8* %buffer, i32 288
    %u579_out_0_1 = bitcast i8* %u579_out_0_0 to <8 x float> *
    %u579_out_0 = load <8 x float>, <8 x float>* %u579_out_0_1, align 32
    ; -------------------

    %f580 = bitcast <8 x float> %u579_out_0 to <8 x float>
    %p581 = fcmp one <8 x float> %f580, %f576
    %p582 = and <8 x i1> %p545, %p581
    %f583 = select <8 x i1> %p582, <8 x float> %f290, <8 x float> %f541
    %f584 = select <8 x i1> %p582, <8 x float> %f290, <8 x float> %f542
    %f585 = select <8 x i1> %p582, <8 x float> %f290, <8 x float> %f543
    %f586 = select <8 x i1> %p582, <8 x float> %f290, <8 x float> %f537
    %p587 = fcmp one <8 x float> %f586, %f290
    %p588 = and <8 x i1> %p505, %p587
    %f589 = fmul <8 x float> %f538, %f377
    %f590 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f539, <8 x float> %f378, <8 x float> %f589)
    %f591 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f540, <8 x float> %f379, <8 x float> %f590)
    %p592 = and <8 x i1> %p588, %p499
    %p593 = and <8 x i1> %p592, %p285
    %rd594_p1 = getelementptr inbounds i8*, i8** %params, i32 49
    %rd594 = load i8*, i8** %rd594_p1, align 8, !alias.scope !2
    %rd595_p1 = getelementptr inbounds i8*, i8** %params, i32 50
    %rd595 = load i8*, i8** %rd595_p1, align 8, !alias.scope !2
    br label %l596_start

l596_start:
    ; VCall: mitsuba::BSDF::eval_pdf()
    ;  - target 1 = @func_e9d5179176fa4aaf3f8d2cb1ff4d2c84;
    ;  - target 2 = @func_9d53124ccfd48ba20820a9ee3e480cd9;
    ;  - target 3 = @func_4448b4670e928f754bf33368168c3f24;

    %u596_self_ptr_0 = bitcast i8* %rd594 to i64*
    %u596_self_ptr = getelementptr i64, i64* %u596_self_ptr_0, <8 x i32> %r498
    %u596_self_combined = call <8 x i64> @llvm.masked.gather.v8i64(<8 x i64*> %u596_self_ptr, i32 8, <8 x i1> %p593, <8 x i64> zeroinitializer)
    %u596_self_initial = trunc <8 x i64> %u596_self_combined to <8 x i32>
    %u596_offset_1 = lshr <8 x i64> %u596_self_combined, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
    %u596_offset = trunc <8 x i64> %u596_offset_1 to <8 x i32>
    %u596_in_0_0 = getelementptr inbounds i8, i8* %buffer, i32 0
    %u596_in_0_1 = bitcast i8* %u596_in_0_0 to <8 x float> *
    store <8 x float> %f370, <8 x float>* %u596_in_0_1, align 32
    %u596_in_1_0 = getelementptr inbounds i8, i8* %buffer, i32 32
    %u596_in_1_1 = bitcast i8* %u596_in_1_0 to <8 x float> *
    store <8 x float> %f371, <8 x float>* %u596_in_1_1, align 32
    %u596_in_2_0 = getelementptr inbounds i8, i8* %buffer, i32 64
    %u596_in_2_1 = bitcast i8* %u596_in_2_0 to <8 x float> *
    store <8 x float> %f372, <8 x float>* %u596_in_2_1, align 32
    %u596_in_3_0 = getelementptr inbounds i8, i8* %buffer, i32 96
    %u596_in_3_1 = bitcast i8* %u596_in_3_0 to <8 x float> *
    store <8 x float> %f377, <8 x float>* %u596_in_3_1, align 32
    %u596_in_4_0 = getelementptr inbounds i8, i8* %buffer, i32 128
    %u596_in_4_1 = bitcast i8* %u596_in_4_0 to <8 x float> *
    store <8 x float> %f378, <8 x float>* %u596_in_4_1, align 32
    %u596_in_5_0 = getelementptr inbounds i8, i8* %buffer, i32 160
    %u596_in_5_1 = bitcast i8* %u596_in_5_0 to <8 x float> *
    store <8 x float> %f379, <8 x float>* %u596_in_5_1, align 32
    %u596_in_6_0 = getelementptr inbounds i8, i8* %buffer, i32 192
    %u596_in_6_1 = bitcast i8* %u596_in_6_0 to <8 x float> *
    store <8 x float> %f448, <8 x float>* %u596_in_6_1, align 32
    %u596_in_7_0 = getelementptr inbounds i8, i8* %buffer, i32 224
    %u596_in_7_1 = bitcast i8* %u596_in_7_0 to <8 x i32> *
    store <8 x i32> %r346, <8 x i32>* %u596_in_7_1, align 32
    %u596_in_8_0 = getelementptr inbounds i8, i8* %buffer, i32 256
    %u596_in_8_1 = bitcast i8* %u596_in_8_0 to <8 x float> *
    store <8 x float> %f591, <8 x float>* %u596_in_8_1, align 32
    %u596_out = getelementptr i8, i8* %buffer, i32 288
    %u596_tmp_0_0 = getelementptr inbounds i8, i8* %u596_out, i64 0
    %u596_tmp_0_1 = bitcast i8* %u596_tmp_0_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u596_tmp_0_1, align 32
    %u596_tmp_1_0 = getelementptr inbounds i8, i8* %u596_out, i64 32
    %u596_tmp_1_1 = bitcast i8* %u596_tmp_1_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u596_tmp_1_1, align 32
    %u596_tmp_2_0 = getelementptr inbounds i8, i8* %u596_out, i64 64
    %u596_tmp_2_1 = bitcast i8* %u596_tmp_2_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u596_tmp_2_1, align 32
    %u596_tmp_3_0 = getelementptr inbounds i8, i8* %u596_out, i64 96
    %u596_tmp_3_1 = bitcast i8* %u596_tmp_3_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u596_tmp_3_1, align 32
    br label %l596_check

l596_check:
    %u596_self = phi <8 x i32> [ %u596_self_initial, %l596_start ], [ %u596_self_next, %l596_call ]
    %u596_next = call i32 @llvm.experimental.vector.reduce.umax.v8i32(<8 x i32> %u596_self)
    %u596_valid = icmp ne i32 %u596_next, 0
    br i1 %u596_valid, label %l596_call, label %l596_end

l596_call:
    %u596_bcast_0 = insertelement <8 x i32> undef, i32 %u596_next, i32 0
    %u596_bcast = shufflevector <8 x i32> %u596_bcast_0, <8 x i32> undef, <8 x i32> zeroinitializer
    %u596_active = icmp eq <8 x i32> %u596_self, %u596_bcast
    %u596_func_0 = getelementptr inbounds i8*, i8** %callables, i32 %u596_next
    %u596_func_1 = load i8*, i8** %u596_func_0
    %u596_func = bitcast i8* %u596_func_1 to void (<8 x i1>, i8*, i8*, <8 x i32>)*
    call void %u596_func(<8 x i1> %u596_active, i8* %buffer, i8* %rd595, <8 x i32> %u596_offset)
    %u596_self_next = select <8 x i1> %u596_active, <8 x i32> zeroinitializer, <8 x i32> %u596_self
    br label %l596_check

l596_end:
    %u596_out_0_0 = getelementptr inbounds i8, i8* %u596_out, i64 0
    %u596_out_0_1 = bitcast i8* %u596_out_0_0 to <8 x float> *
    %f597 = load <8 x float>, <8 x float>* %u596_out_0_1, align 32
    %u596_out_1_0 = getelementptr inbounds i8, i8* %u596_out, i64 32
    %u596_out_1_1 = bitcast i8* %u596_out_1_0 to <8 x float> *
    %f598 = load <8 x float>, <8 x float>* %u596_out_1_1, align 32
    %u596_out_2_0 = getelementptr inbounds i8, i8* %u596_out, i64 64
    %u596_out_2_1 = bitcast i8* %u596_out_2_0 to <8 x float> *
    %f599 = load <8 x float>, <8 x float>* %u596_out_2_1, align 32
    %u596_out_3_0 = getelementptr inbounds i8, i8* %u596_out, i64 96
    %u596_out_3_1 = bitcast i8* %u596_out_3_0 to <8 x float> *
    %f600 = load <8 x float>, <8 x float>* %u596_out_3_1, align 32
    br label %l596_done

l596_done:
    %f601 = fmul <8 x float> %f586, %f586
    %f602 = fmul <8 x float> %f600, %f600
    %p603 = fcmp ogt <8 x float> %f601, %f290
    %f604 = fadd <8 x float> %f601, %f602
    %f605 = fdiv <8 x float> %f601, %f604
    %f606 = select <8 x i1> %p603, <8 x float> %f605, <8 x float> zeroinitializer
    %f607 = fmul <8 x float> %f583, %f597
    %f608 = fmul <8 x float> %f584, %f598
    %f609 = fmul <8 x float> %f585, %f599
    %f610 = fmul <8 x float> %f607, %f606
    %f611 = fmul <8 x float> %f608, %f606
    %f612 = fmul <8 x float> %f609, %f606
    %f613 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f269, <8 x float> %f610, <8 x float> %f493)
    %f614 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f270, <8 x float> %f611, <8 x float> %f494)
    %f615 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f271, <8 x float> %f612, <8 x float> %f495)
    %f616 = select <8 x i1> %p588, <8 x float> %f613, <8 x float> %f493
    %f617 = select <8 x i1> %p588, <8 x float> %f614, <8 x float> %f494
    %f618 = select <8 x i1> %p588, <8 x float> %f615, <8 x float> %f495
    %rd619_0 = mul <8 x i64> %rd508, %rd310
    %rd619 = add <8 x i64> %rd619_0, %rd112
    %rd620 = select <8 x i1> %p449, <8 x i64> %rd619, <8 x i64> %rd508
    %rd621 = lshr <8 x i64> %rd508, %rd312
    %rd622 = xor <8 x i64> %rd621, %rd508
    %rd623 = lshr <8 x i64> %rd622, %rd315
    %r624 = trunc <8 x i64> %rd623 to <8 x i32>
    %rd625 = lshr <8 x i64> %rd508, %rd318
    %r626 = trunc <8 x i64> %rd625 to <8 x i32>
    %r627 = lshr <8 x i32> %r624, %r626
    %r628 = bitcast <8 x i32> %r626 to <8 x i32>
    %r629 = sub <8 x i32> zeroinitializer, %r628
    %r630 = and <8 x i32> %r629, %r324
    %r631 = bitcast <8 x i32> %r630 to <8 x i32>
    %r632 = shl <8 x i32> %r624, %r631
    %r633 = or <8 x i32> %r627, %r632
    %r634 = lshr <8 x i32> %r633, %r329
    %r635 = or <8 x i32> %r634, %r331
    %f636 = bitcast <8 x i32> %r635 to <8 x float>
    %f637 = fsub <8 x float> %f636, %f297
    %rd638_0 = mul <8 x i64> %rd620, %rd310
    %rd638 = add <8 x i64> %rd638_0, %rd112
    %rd639 = select <8 x i1> %p449, <8 x i64> %rd638, <8 x i64> %rd620
    %rd640 = lshr <8 x i64> %rd620, %rd312
    %rd641 = xor <8 x i64> %rd640, %rd620
    %rd642 = lshr <8 x i64> %rd641, %rd315
    %r643 = trunc <8 x i64> %rd642 to <8 x i32>
    %rd644 = lshr <8 x i64> %rd620, %rd318
    %r645 = trunc <8 x i64> %rd644 to <8 x i32>
    %r646 = lshr <8 x i32> %r643, %r645
    %r647 = bitcast <8 x i32> %r645 to <8 x i32>
    %r648 = sub <8 x i32> zeroinitializer, %r647
    %r649 = and <8 x i32> %r648, %r324
    %r650 = bitcast <8 x i32> %r649 to <8 x i32>
    %r651 = shl <8 x i32> %r643, %r650
    %r652 = or <8 x i32> %r646, %r651
    %r653 = lshr <8 x i32> %r652, %r329
    %r654 = or <8 x i32> %r653, %r331
    %f655 = bitcast <8 x i32> %r654 to <8 x float>
    %f656 = fsub <8 x float> %f655, %f297
    %rd657_0 = mul <8 x i64> %rd639, %rd310
    %rd657 = add <8 x i64> %rd657_0, %rd112
    %rd658 = select <8 x i1> %p449, <8 x i64> %rd657, <8 x i64> %rd639
    %rd659 = lshr <8 x i64> %rd639, %rd312
    %rd660 = xor <8 x i64> %rd659, %rd639
    %rd661 = lshr <8 x i64> %rd660, %rd315
    %r662 = trunc <8 x i64> %rd661 to <8 x i32>
    %rd663 = lshr <8 x i64> %rd639, %rd318
    %r664 = trunc <8 x i64> %rd663 to <8 x i32>
    %r665 = lshr <8 x i32> %r662, %r664
    %r666 = bitcast <8 x i32> %r664 to <8 x i32>
    %r667 = sub <8 x i32> zeroinitializer, %r666
    %r668 = and <8 x i32> %r667, %r324
    %r669 = bitcast <8 x i32> %r668 to <8 x i32>
    %r670 = shl <8 x i32> %r662, %r669
    %r671 = or <8 x i32> %r665, %r670
    %r672 = lshr <8 x i32> %r671, %r329
    %r673 = or <8 x i32> %r672, %r331
    %f674 = bitcast <8 x i32> %r673 to <8 x float>
    %f675 = fsub <8 x float> %f674, %f297
    %p676 = and <8 x i1> %p449, %p499
    %p677 = and <8 x i1> %p676, %p285
    %rd678_p1 = getelementptr inbounds i8*, i8** %params, i32 51
    %rd678 = load i8*, i8** %rd678_p1, align 8, !alias.scope !2
    %rd679_p1 = getelementptr inbounds i8*, i8** %params, i32 52
    %rd679 = load i8*, i8** %rd679_p1, align 8, !alias.scope !2
    br label %l680_start

l680_start:
    ; VCall: mitsuba::BSDF::sample()
    ;  - target 1 = @func_60a1a348aca7bdb47e851f6cedce38a2;
    ;  - target 2 = @func_e6fff8d6b222436f71518443ce068358;
    ;  - target 3 = @func_48d4a1fb2376a0011da8c81578897b84;

    %u680_self_ptr_0 = bitcast i8* %rd678 to i64*
    %u680_self_ptr = getelementptr i64, i64* %u680_self_ptr_0, <8 x i32> %r498
    %u680_self_combined = call <8 x i64> @llvm.masked.gather.v8i64(<8 x i64*> %u680_self_ptr, i32 8, <8 x i1> %p677, <8 x i64> zeroinitializer)
    %u680_self_initial = trunc <8 x i64> %u680_self_combined to <8 x i32>
    %u680_offset_1 = lshr <8 x i64> %u680_self_combined, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
    %u680_offset = trunc <8 x i64> %u680_offset_1 to <8 x i32>
    %u680_in_0_0 = getelementptr inbounds i8, i8* %buffer, i32 0
    %u680_in_0_1 = bitcast i8* %u680_in_0_0 to <8 x float> *
    store <8 x float> %f370, <8 x float>* %u680_in_0_1, align 32
    %u680_in_1_0 = getelementptr inbounds i8, i8* %buffer, i32 32
    %u680_in_1_1 = bitcast i8* %u680_in_1_0 to <8 x float> *
    store <8 x float> %f371, <8 x float>* %u680_in_1_1, align 32
    %u680_in_2_0 = getelementptr inbounds i8, i8* %buffer, i32 64
    %u680_in_2_1 = bitcast i8* %u680_in_2_0 to <8 x float> *
    store <8 x float> %f372, <8 x float>* %u680_in_2_1, align 32
    %u680_in_3_0 = getelementptr inbounds i8, i8* %buffer, i32 96
    %u680_in_3_1 = bitcast i8* %u680_in_3_0 to <8 x float> *
    store <8 x float> %f377, <8 x float>* %u680_in_3_1, align 32
    %u680_in_4_0 = getelementptr inbounds i8, i8* %buffer, i32 128
    %u680_in_4_1 = bitcast i8* %u680_in_4_0 to <8 x float> *
    store <8 x float> %f378, <8 x float>* %u680_in_4_1, align 32
    %u680_in_5_0 = getelementptr inbounds i8, i8* %buffer, i32 160
    %u680_in_5_1 = bitcast i8* %u680_in_5_0 to <8 x float> *
    store <8 x float> %f379, <8 x float>* %u680_in_5_1, align 32
    %u680_in_6_0 = getelementptr inbounds i8, i8* %buffer, i32 192
    %u680_in_6_1 = bitcast i8* %u680_in_6_0 to <8 x float> *
    store <8 x float> %f446, <8 x float>* %u680_in_6_1, align 32
    %u680_in_7_0 = getelementptr inbounds i8, i8* %buffer, i32 224
    %u680_in_7_1 = bitcast i8* %u680_in_7_0 to <8 x float> *
    store <8 x float> %f447, <8 x float>* %u680_in_7_1, align 32
    %u680_in_8_0 = getelementptr inbounds i8, i8* %buffer, i32 256
    %u680_in_8_1 = bitcast i8* %u680_in_8_0 to <8 x float> *
    store <8 x float> %f448, <8 x float>* %u680_in_8_1, align 32
    %u680_in_9_0 = getelementptr inbounds i8, i8* %buffer, i32 288
    %u680_in_9_1 = bitcast i8* %u680_in_9_0 to <8 x i32> *
    store <8 x i32> %r346, <8 x i32>* %u680_in_9_1, align 32
    %u680_in_10_0 = getelementptr inbounds i8, i8* %buffer, i32 320
    %u680_in_10_1 = bitcast i8* %u680_in_10_0 to <8 x float> *
    store <8 x float> %f637, <8 x float>* %u680_in_10_1, align 32
    %u680_in_11_0 = getelementptr inbounds i8, i8* %buffer, i32 352
    %u680_in_11_1 = bitcast i8* %u680_in_11_0 to <8 x float> *
    store <8 x float> %f656, <8 x float>* %u680_in_11_1, align 32
    %u680_in_12_0 = getelementptr inbounds i8, i8* %buffer, i32 384
    %u680_in_12_1 = bitcast i8* %u680_in_12_0 to <8 x float> *
    store <8 x float> %f675, <8 x float>* %u680_in_12_1, align 32
    %u680_out = getelementptr i8, i8* %buffer, i32 416
    %u680_tmp_0_0 = getelementptr inbounds i8, i8* %u680_out, i64 0
    %u680_tmp_0_1 = bitcast i8* %u680_tmp_0_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u680_tmp_0_1, align 32
    %u680_tmp_1_0 = getelementptr inbounds i8, i8* %u680_out, i64 32
    %u680_tmp_1_1 = bitcast i8* %u680_tmp_1_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u680_tmp_1_1, align 32
    %u680_tmp_2_0 = getelementptr inbounds i8, i8* %u680_out, i64 64
    %u680_tmp_2_1 = bitcast i8* %u680_tmp_2_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u680_tmp_2_1, align 32
    %u680_tmp_3_0 = getelementptr inbounds i8, i8* %u680_out, i64 96
    %u680_tmp_3_1 = bitcast i8* %u680_tmp_3_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u680_tmp_3_1, align 32
    %u680_tmp_4_0 = getelementptr inbounds i8, i8* %u680_out, i64 128
    %u680_tmp_4_1 = bitcast i8* %u680_tmp_4_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u680_tmp_4_1, align 32
    %u680_tmp_5_0 = getelementptr inbounds i8, i8* %u680_out, i64 160
    %u680_tmp_5_1 = bitcast i8* %u680_tmp_5_0 to <8 x i32> *
    store <8 x i32> zeroinitializer, <8 x i32>* %u680_tmp_5_1, align 32
    %u680_tmp_7_0 = getelementptr inbounds i8, i8* %u680_out, i64 192
    %u680_tmp_7_1 = bitcast i8* %u680_tmp_7_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u680_tmp_7_1, align 32
    %u680_tmp_8_0 = getelementptr inbounds i8, i8* %u680_out, i64 224
    %u680_tmp_8_1 = bitcast i8* %u680_tmp_8_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u680_tmp_8_1, align 32
    %u680_tmp_9_0 = getelementptr inbounds i8, i8* %u680_out, i64 256
    %u680_tmp_9_1 = bitcast i8* %u680_tmp_9_0 to <8 x float> *
    store <8 x float> zeroinitializer, <8 x float>* %u680_tmp_9_1, align 32
    br label %l680_check

l680_check:
    %u680_self = phi <8 x i32> [ %u680_self_initial, %l680_start ], [ %u680_self_next, %l680_call ]
    %u680_next = call i32 @llvm.experimental.vector.reduce.umax.v8i32(<8 x i32> %u680_self)
    %u680_valid = icmp ne i32 %u680_next, 0
    br i1 %u680_valid, label %l680_call, label %l680_end

l680_call:
    %u680_bcast_0 = insertelement <8 x i32> undef, i32 %u680_next, i32 0
    %u680_bcast = shufflevector <8 x i32> %u680_bcast_0, <8 x i32> undef, <8 x i32> zeroinitializer
    %u680_active = icmp eq <8 x i32> %u680_self, %u680_bcast
    %u680_func_0 = getelementptr inbounds i8*, i8** %callables, i32 %u680_next
    %u680_func_1 = load i8*, i8** %u680_func_0
    %u680_func = bitcast i8* %u680_func_1 to void (<8 x i1>, i8*, i8*, <8 x i32>)*
    call void %u680_func(<8 x i1> %u680_active, i8* %buffer, i8* %rd679, <8 x i32> %u680_offset)
    %u680_self_next = select <8 x i1> %u680_active, <8 x i32> zeroinitializer, <8 x i32> %u680_self
    br label %l680_check

l680_end:
    %u680_out_0_0 = getelementptr inbounds i8, i8* %u680_out, i64 0
    %u680_out_0_1 = bitcast i8* %u680_out_0_0 to <8 x float> *
    %f681 = load <8 x float>, <8 x float>* %u680_out_0_1, align 32
    %u680_out_1_0 = getelementptr inbounds i8, i8* %u680_out, i64 32
    %u680_out_1_1 = bitcast i8* %u680_out_1_0 to <8 x float> *
    %f682 = load <8 x float>, <8 x float>* %u680_out_1_1, align 32
    %u680_out_2_0 = getelementptr inbounds i8, i8* %u680_out, i64 64
    %u680_out_2_1 = bitcast i8* %u680_out_2_0 to <8 x float> *
    %f683 = load <8 x float>, <8 x float>* %u680_out_2_1, align 32
    %u680_out_3_0 = getelementptr inbounds i8, i8* %u680_out, i64 96
    %u680_out_3_1 = bitcast i8* %u680_out_3_0 to <8 x float> *
    %f684 = load <8 x float>, <8 x float>* %u680_out_3_1, align 32
    %u680_out_4_0 = getelementptr inbounds i8, i8* %u680_out, i64 128
    %u680_out_4_1 = bitcast i8* %u680_out_4_0 to <8 x float> *
    %f685 = load <8 x float>, <8 x float>* %u680_out_4_1, align 32
    %u680_out_5_0 = getelementptr inbounds i8, i8* %u680_out, i64 160
    %u680_out_5_1 = bitcast i8* %u680_out_5_0 to <8 x i32> *
    %r686 = load <8 x i32>, <8 x i32>* %u680_out_5_1, align 32
    %u680_out_7_0 = getelementptr inbounds i8, i8* %u680_out, i64 192
    %u680_out_7_1 = bitcast i8* %u680_out_7_0 to <8 x float> *
    %f687 = load <8 x float>, <8 x float>* %u680_out_7_1, align 32
    %u680_out_8_0 = getelementptr inbounds i8, i8* %u680_out, i64 224
    %u680_out_8_1 = bitcast i8* %u680_out_8_0 to <8 x float> *
    %f688 = load <8 x float>, <8 x float>* %u680_out_8_1, align 32
    %u680_out_9_0 = getelementptr inbounds i8, i8* %u680_out, i64 256
    %u680_out_9_1 = bitcast i8* %u680_out_9_0 to <8 x float> *
    %f689 = load <8 x float>, <8 x float>* %u680_out_9_1, align 32
    br label %l680_done

l680_done:
    %f690 = fmul <8 x float> %f425, %f681
    %f691 = fmul <8 x float> %f426, %f681
    %f692 = fmul <8 x float> %f427, %f681
    %f693 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f432, <8 x float> %f682, <8 x float> %f690)
    %f694 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f434, <8 x float> %f682, <8 x float> %f691)
    %f695 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f436, <8 x float> %f682, <8 x float> %f692)
    %f696 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f377, <8 x float> %f683, <8 x float> %f693)
    %f697 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f378, <8 x float> %f683, <8 x float> %f694)
    %f698 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f379, <8 x float> %f683, <8 x float> %f695)
    %f699 = fmul <8 x float> %f373, %f696
    %f700 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f374, <8 x float> %f697, <8 x float> %f699)
    %f701 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f375, <8 x float> %f698, <8 x float> %f700)
    %p702 = fcmp oge <8 x float> %f701, %f290
    %f703 = select <8 x i1> %p702, <8 x float> %f555, <8 x float> %f560
    %f704 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f703, <8 x float> %f373, <8 x float> %f370)
    %f705 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f703, <8 x float> %f374, <8 x float> %f371)
    %f706 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f703, <8 x float> %f375, <8 x float> %f372)
    %f707_1 = insertelement <8 x float> undef, float 0x47efffffe0000000, i32 0
    %f707 = shufflevector <8 x float> %f707_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f708 = fmul <8 x float> %f269, %f687
    %f709 = fmul <8 x float> %f270, %f688
    %f710 = fmul <8 x float> %f271, %f689
    %f711 = fmul <8 x float> %f275, %f685
    %r712_1 = insertelement <8 x i32> undef, i32 97, i32 0
    %r712 = shufflevector <8 x i32> %r712_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r713 = and <8 x i32> %r686, %r712
    %p714 = icmp ne <8 x i32> %r713, %r289
    %r715 = select <8 x i1> %p385, <8 x i32> %r306, <8 x i32> %r276
    %f716 = call <8 x float> @llvm.maxnum.v8f32(<8 x float> %f708, <8 x float> %f709)
    %f717 = call <8 x float> @llvm.maxnum.v8f32(<8 x float> %f716, <8 x float> %f710)
    %f718 = fmul <8 x float> %f711, %f711
    %f719 = fmul <8 x float> %f717, %f718
    %f720 = call <8 x float> @llvm.minnum.v8f32(<8 x float> %f719, <8 x float> %f340)
    %p721 = icmp uge <8 x i32> %r715, %r341
    %rd722_0 = mul <8 x i64> %rd658, %rd310
    %rd722 = add <8 x i64> %rd722_0, %rd112
    %rd723 = lshr <8 x i64> %rd658, %rd312
    %rd724 = xor <8 x i64> %rd723, %rd658
    %rd725 = lshr <8 x i64> %rd724, %rd315
    %r726 = trunc <8 x i64> %rd725 to <8 x i32>
    %rd727 = lshr <8 x i64> %rd658, %rd318
    %r728 = trunc <8 x i64> %rd727 to <8 x i32>
    %r729 = lshr <8 x i32> %r726, %r728
    %r730 = bitcast <8 x i32> %r728 to <8 x i32>
    %r731 = sub <8 x i32> zeroinitializer, %r730
    %r732 = and <8 x i32> %r731, %r324
    %r733 = bitcast <8 x i32> %r732 to <8 x i32>
    %r734 = shl <8 x i32> %r726, %r733
    %r735 = or <8 x i32> %r729, %r734
    %r736 = lshr <8 x i32> %r735, %r329
    %r737 = or <8 x i32> %r736, %r331
    %f738 = bitcast <8 x i32> %r737 to <8 x float>
    %f739 = fsub <8 x float> %f738, %f297
    %p740 = fcmp olt <8 x float> %f739, %f720
    %f741 = fdiv <8 x float> %f297, %f720
    %f742 = fmul <8 x float> %f708, %f741
    %f743 = fmul <8 x float> %f709, %f741
    %f744 = fmul <8 x float> %f710, %f741
    %f745 = select <8 x i1> %p721, <8 x float> %f742, <8 x float> %f708
    %f746 = select <8 x i1> %p721, <8 x float> %f743, <8 x float> %f709
    %f747 = select <8 x i1> %p721, <8 x float> %f744, <8 x float> %f710
    %p748 = xor <8 x i1> %p721, <i1 1, i1 1, i1 1, i1 1, i1 1, i1 1, i1 1, i1 1>
    %p749 = or <8 x i1> %p748, %p740
    %p750 = and <8 x i1> %p496, %p749
    %p751 = fcmp one <8 x float> %f717, %f290
    %p752 = and <8 x i1> %p750, %p751
    br label %l_236_tail

l_236_tail:
    %rd237_final = select <8 x i1> %p259, <8 x i64> %rd722, <8 x i64> %rd237
    %f238_final = select <8 x i1> %p259, <8 x float> %f704, <8 x float> %f238
    %f239_final = select <8 x i1> %p259, <8 x float> %f705, <8 x float> %f239
    %f240_final = select <8 x i1> %p259, <8 x float> %f706, <8 x float> %f240
    %f241_final = select <8 x i1> %p259, <8 x float> %f696, <8 x float> %f241
    %f242_final = select <8 x i1> %p259, <8 x float> %f697, <8 x float> %f242
    %f243_final = select <8 x i1> %p259, <8 x float> %f698, <8 x float> %f243
    %f244_final = select <8 x i1> %p259, <8 x float> %f707, <8 x float> %f244
    %f245_final = select <8 x i1> %p259, <8 x float> %f745, <8 x float> %f245
    %f246_final = select <8 x i1> %p259, <8 x float> %f746, <8 x float> %f246
    %f247_final = select <8 x i1> %p259, <8 x float> %f747, <8 x float> %f247
    %f248_final = select <8 x i1> %p259, <8 x float> %f616, <8 x float> %f248
    %f249_final = select <8 x i1> %p259, <8 x float> %f617, <8 x float> %f249
    %f250_final = select <8 x i1> %p259, <8 x float> %f618, <8 x float> %f250
    %f251_final = select <8 x i1> %p259, <8 x float> %f711, <8 x float> %f251
    %r252_final = select <8 x i1> %p259, <8 x i32> %r715, <8 x i32> %r252
    %f254_final = select <8 x i1> %p259, <8 x float> %f370, <8 x float> %f254
    %f255_final = select <8 x i1> %p259, <8 x float> %f371, <8 x float> %f255
    %f256_final = select <8 x i1> %p259, <8 x float> %f372, <8 x float> %f256
    %f257_final = select <8 x i1> %p259, <8 x float> %f684, <8 x float> %f257
    %p258_final = select <8 x i1> %p259, <8 x i1> %p714, <8 x i1> %p258
    %p259_final = select <8 x i1> %p259, <8 x i1> %p752, <8 x i1> %p259
    br label %l_236_cond;

l_236_done:
    %f754 = bitcast <8 x float> %f248 to <8 x float>
    %f755 = bitcast <8 x float> %f249 to <8 x float>
    %f756 = bitcast <8 x float> %f250 to <8 x float>
    %f757 = bitcast <8 x float> %f253 to <8 x float>
    %p758 = fcmp one <8 x float> %f757, %f220
    %f759 = select <8 x i1> %p758, <8 x float> %f757, <8 x float> zeroinitializer
    %f760_1 = insertelement <8 x float> undef, float 0xbfe0000000000000, i32 0
    %f760 = shufflevector <8 x float> %f760_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f761 = fadd <8 x float> %f168, %f760
    %f762 = fadd <8 x float> %f169, %f760
    %f763 = fsub <8 x float> %f761, %f30
    %f764 = fsub <8 x float> %f762, %f30
    %f765 = fadd <8 x float> %f761, %f30
    %f766 = fadd <8 x float> %f762, %f30
    %f767 = call <8 x float> @llvm.ceil.v8f32(<8 x float> %f763)
    %r768 = fptosi <8 x float> %f767 to <8 x i32>
    %f769 = call <8 x float> @llvm.ceil.v8f32(<8 x float> %f764)
    %r770 = fptosi <8 x float> %f769 to <8 x i32>
    %r771_1 = insertelement <8 x i32> undef, i32 0, i32 0
    %r771 = shufflevector <8 x i32> %r771_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r772_0 = icmp sgt <8 x i32> %r768, %r771
    %r772 = select <8 x i1> %r772_0, <8 x i32> %r768, <8 x i32> %r771
    %r773_0 = icmp sgt <8 x i32> %r770, %r771
    %r773 = select <8 x i1> %r773_0, <8 x i32> %r770, <8 x i32> %r771
    %r774 = bitcast <8 x i32> %r772 to <8 x i32>
    %r775 = bitcast <8 x i32> %r773 to <8 x i32>
    %f776 = call <8 x float> @llvm.floor.v8f32(<8 x float> %f765)
    %r777 = fptosi <8 x float> %f776 to <8 x i32>
    %f778 = call <8 x float> @llvm.floor.v8f32(<8 x float> %f766)
    %r779 = fptosi <8 x float> %f778 to <8 x i32>
    %r780_1 = insertelement <8 x i32> undef, i32 255, i32 0
    %r780 = shufflevector <8 x i32> %r780_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r781_0 = icmp slt <8 x i32> %r777, %r780
    %r781 = select <8 x i1> %r781_0, <8 x i32> %r777, <8 x i32> %r780
    %r782_0 = icmp slt <8 x i32> %r779, %r780
    %r782 = select <8 x i1> %r782_0, <8 x i32> %r779, <8 x i32> %r780
    %r783 = bitcast <8 x i32> %r781 to <8 x i32>
    %r784 = bitcast <8 x i32> %r782 to <8 x i32>
    %r785_0 = mul <8 x i32> %r775, %r121
    %r785 = add <8 x i32> %r785_0, %r774
    %r786 = mul <8 x i32> %r785, %r40
    %p787 = icmp ule <8 x i32> %r774, %r783
    %p788 = icmp ule <8 x i32> %r775, %r784
    %p789 = and <8 x i1> %p787, %p788
    %f790 = uitofp <8 x i32> %r774 to <8 x float>
    %f791 = uitofp <8 x i32> %r775 to <8 x float>
    %f792 = fsub <8 x float> %f790, %f761
    %f793 = fsub <8 x float> %f791, %f762
    br label %l_794_start

l_794_start:
    br label %l_794_cond

l_794_cond: ; Loop (ImageBlock::put() [1])
    %r795 = phi <8 x i32> [ %r795_final, %l_794_tail ], [ %r219, %l_794_start ]
    %r796 = phi <8 x i32> [ %r796_final, %l_794_tail ], [ %r786, %l_794_start ]
    %r797_1 = insertelement <8 x i32> undef, i32 2, i32 0
    %r797 = shufflevector <8 x i32> %r797_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %p798 = icmp ult <8 x i32> %r795, %r797
    %p794 = call i1 @llvm.experimental.vector.reduce.or.v8i1(<8 x i1> %p798)
    br i1 %p794, label %l_794_body, label %l_794_done

l_794_body:
    %r800 = phi <8 x i32> [ %r795, %l_794_cond ]
    %r801 = phi <8 x i32> [ %r796, %l_794_cond ]
    %r802_0 = trunc i64 %index to i32
    %r802_1 = insertelement <8 x i32> undef, i32 %r802_0, i32 0
    %r802_2 = shufflevector <8 x i32> %r802_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r802 = add <8 x i32> %r802_2, <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
    %p803_0 = trunc i64 %end to i32
    %p803_1 = insertelement <8 x i32> undef, i32 %p803_0, i32 0
    %p803_2 = shufflevector <8 x i32> %p803_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %p803 = icmp ult <8 x i32> %r802, %p803_2
    %p804 = and <8 x i1> %p803, %p798
    %f805 = uitofp <8 x i32> %r800 to <8 x float>
    %f806 = fadd <8 x float> %f793, %f805
    %f807_1 = insertelement <8 x float> undef, float 0x7fffffffe0000000, i32 0
    %f807 = shufflevector <8 x float> %f807_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f808_0 = bitcast <8 x float> %f806 to <8 x i32>
    %f808_1 = bitcast <8 x float> %f807 to <8 x i32>
    %f808_2 = and <8 x i32> %f808_0, %f808_1
    %f808 = bitcast <8 x i32> %f808_2 to <8 x float>
    %f809_1 = insertelement <8 x float> undef, float 0x3ff0000000000000, i32 0
    %f809 = shufflevector <8 x float> %f809_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f810 = fsub <8 x float> %f809, %f808
    %f811_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f811 = shufflevector <8 x float> %f811_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f812 = call <8 x float> @llvm.maxnum.v8f32(<8 x float> %f811, <8 x float> %f810)
    %r813 = add <8 x i32> %r775, %r800
    %p814 = icmp ule <8 x i32> %r813, %r784
    %p815 = and <8 x i1> %p789, %p814
    %r816_1 = insertelement <8 x i32> undef, i32 0, i32 0
    %r816 = shufflevector <8 x i32> %r816_1, <8 x i32> undef, <8 x i32> zeroinitializer
    br label %l_817_start

l_817_start:
    br label %l_817_cond

l_817_cond: ; Loop (ImageBlock::put() [2])
    %r818 = phi <8 x i32> [ %r818_final, %l_817_tail ], [ %r816, %l_817_start ]
    %r819 = phi <8 x i32> [ %r819_final, %l_817_tail ], [ %r801, %l_817_start ]
    %r820_1 = insertelement <8 x i32> undef, i32 2, i32 0
    %r820 = shufflevector <8 x i32> %r820_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %p821 = icmp ult <8 x i32> %r818, %r820
    %p817 = call i1 @llvm.experimental.vector.reduce.or.v8i1(<8 x i1> %p821)
    br i1 %p817, label %l_817_body, label %l_817_done

l_817_body:
    %r823 = phi <8 x i32> [ %r818, %l_817_cond ]
    %r824 = phi <8 x i32> [ %r819, %l_817_cond ]
    %p825 = and <8 x i1> %p804, %p821
    %f826 = uitofp <8 x i32> %r823 to <8 x float>
    %f827 = fadd <8 x float> %f792, %f826
    %f828_1 = insertelement <8 x float> undef, float 0x7fffffffe0000000, i32 0
    %f828 = shufflevector <8 x float> %f828_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f829_0 = bitcast <8 x float> %f827 to <8 x i32>
    %f829_1 = bitcast <8 x float> %f828 to <8 x i32>
    %f829_2 = and <8 x i32> %f829_0, %f829_1
    %f829 = bitcast <8 x i32> %f829_2 to <8 x float>
    %f830_1 = insertelement <8 x float> undef, float 0x3ff0000000000000, i32 0
    %f830 = shufflevector <8 x float> %f830_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f831 = fsub <8 x float> %f830, %f829
    %f832_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f832 = shufflevector <8 x float> %f832_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f833 = call <8 x float> @llvm.maxnum.v8f32(<8 x float> %f832, <8 x float> %f831)
    %f834 = fmul <8 x float> %f833, %f812
    %r835 = add <8 x i32> %r774, %r823
    %p836 = icmp ule <8 x i32> %r835, %r783
    %p837 = and <8 x i1> %p815, %p836
    %f838 = fmul <8 x float> %f754, %f834
    %rd839_p1 = getelementptr inbounds i8*, i8** %params, i32 53
    %rd839 = load i8*, i8** %rd839_p1, align 8, !alias.scope !2
    %p840 = and <8 x i1> %p837, %p825
    %u841_0 = bitcast i8* %rd839 to float*
    %u841_1 = getelementptr float, float* %u841_0, <8 x i32> %r824
    call void @ek.scatter_add.v8f32(<8 x float*> %u841_1, <8 x float> %f838, <8 x i1> %p840)
    %r842_1 = insertelement <8 x i32> undef, i32 1, i32 0
    %r842 = shufflevector <8 x i32> %r842_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r843 = add <8 x i32> %r824, %r842
    %f844 = fmul <8 x float> %f755, %f834
    %u845_0 = bitcast i8* %rd839 to float*
    %u845_1 = getelementptr float, float* %u845_0, <8 x i32> %r843
    call void @ek.scatter_add.v8f32(<8 x float*> %u845_1, <8 x float> %f844, <8 x i1> %p840)
    %r846 = add <8 x i32> %r843, %r842
    %f847 = fmul <8 x float> %f756, %f834
    %u848_0 = bitcast i8* %rd839 to float*
    %u848_1 = getelementptr float, float* %u848_0, <8 x i32> %r846
    call void @ek.scatter_add.v8f32(<8 x float*> %u848_1, <8 x float> %f847, <8 x i1> %p840)
    %r849 = add <8 x i32> %r846, %r842
    %u850_0 = bitcast i8* %rd839 to float*
    %u850_1 = getelementptr float, float* %u850_0, <8 x i32> %r849
    call void @ek.scatter_add.v8f32(<8 x float*> %u850_1, <8 x float> %f834, <8 x i1> %p840)
    %r851 = add <8 x i32> %r849, %r842
    %f852 = fmul <8 x float> %f759, %f834
    %u853_0 = bitcast i8* %rd839 to float*
    %u853_1 = getelementptr float, float* %u853_0, <8 x i32> %r851
    call void @ek.scatter_add.v8f32(<8 x float*> %u853_1, <8 x float> %f852, <8 x i1> %p840)
    %r854 = add <8 x i32> %r851, %r842
    %r855 = add <8 x i32> %r823, %r842
    br label %l_817_tail

l_817_tail:
    %r818_final = select <8 x i1> %p821, <8 x i32> %r855, <8 x i32> %r818
    %r819_final = select <8 x i1> %p821, <8 x i32> %r854, <8 x i32> %r819
    br label %l_817_cond;

l_817_done:
    %r858 = bitcast <8 x i32> %r819 to <8 x i32>
    %r859_1 = insertelement <8 x i32> undef, i32 1, i32 0
    %r859 = shufflevector <8 x i32> %r859_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r860 = add <8 x i32> %r800, %r859
    %r861_1 = insertelement <8 x i32> undef, i32 1270, i32 0
    %r861 = shufflevector <8 x i32> %r861_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r862 = add <8 x i32> %r858, %r861
    br label %l_794_tail

l_794_tail:
    %r795_final = select <8 x i1> %p798, <8 x i32> %r860, <8 x i32> %r795
    %r796_final = select <8 x i1> %p798, <8 x i32> %r862, <8 x i32> %r796
    br label %l_794_cond;

l_794_done:

    br label %suffix

suffix:
    %index_next = add i64 %index, 8
    %cond = icmp uge i64 %index_next, %end
    br i1 %cond, label %done, label %body, !llvm.loop !2

done:
    ret void
}

define void @func_bac3cacf8aebcc87b13f4c43e6cdd822(<8 x i1> %mask, <8 x i32> %self, i8* noalias %params, i8* noalias %data, <8 x i32> %offsets) #0 {
entry:
    ; VCall: mitsuba::Shape::compute_surface_interaction()
    %f1_p1 = getelementptr inbounds i8, i8* %data, i32 12
    %f1_p2 = getelementptr inbounds i8, i8* %f1_p1, <8 x i32> %offsets
    %f1_p3 = bitcast <8 x i8*> %f1_p2 to <8 x float*>
    %f1 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f1_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f2_p1 = getelementptr inbounds i8, i8* %data, i32 0
    %f2_p2 = getelementptr inbounds i8, i8* %f2_p1, <8 x i32> %offsets
    %f2_p3 = bitcast <8 x i8*> %f2_p2 to <8 x float*>
    %f2 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f2_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f3_p1 = getelementptr inbounds i8, i8* %data, i32 8
    %f3_p2 = getelementptr inbounds i8, i8* %f3_p1, <8 x i32> %offsets
    %f3_p3 = bitcast <8 x i8*> %f3_p2 to <8 x float*>
    %f3 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f3_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f4_p1 = getelementptr inbounds i8, i8* %data, i32 4
    %f4_p2 = getelementptr inbounds i8, i8* %f4_p1, <8 x i32> %offsets
    %f4_p3 = bitcast <8 x i8*> %f4_p2 to <8 x float*>
    %f4 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f4_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f5_p1 = getelementptr inbounds i8, i8* %data, i32 72
    %f5_p2 = getelementptr inbounds i8, i8* %f5_p1, <8 x i32> %offsets
    %f5_p3 = bitcast <8 x i8*> %f5_p2 to <8 x float*>
    %f5 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f5_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f6_p1 = getelementptr inbounds i8, i8* %data, i32 84
    %f6_p2 = getelementptr inbounds i8, i8* %f6_p1, <8 x i32> %offsets
    %f6_p3 = bitcast <8 x i8*> %f6_p2 to <8 x float*>
    %f6 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f6_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f7_p1 = getelementptr inbounds i8, i8* %data, i32 96
    %f7_p2 = getelementptr inbounds i8, i8* %f7_p1, <8 x i32> %offsets
    %f7_p3 = bitcast <8 x i8*> %f7_p2 to <8 x float*>
    %f7 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f7_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f8_p1 = getelementptr inbounds i8, i8* %data, i32 68
    %f8_p2 = getelementptr inbounds i8, i8* %f8_p1, <8 x i32> %offsets
    %f8_p3 = bitcast <8 x i8*> %f8_p2 to <8 x float*>
    %f8 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f8_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f9_p1 = getelementptr inbounds i8, i8* %data, i32 80
    %f9_p2 = getelementptr inbounds i8, i8* %f9_p1, <8 x i32> %offsets
    %f9_p3 = bitcast <8 x i8*> %f9_p2 to <8 x float*>
    %f9 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f9_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f10_p1 = getelementptr inbounds i8, i8* %data, i32 92
    %f10_p2 = getelementptr inbounds i8, i8* %f10_p1, <8 x i32> %offsets
    %f10_p3 = bitcast <8 x i8*> %f10_p2 to <8 x float*>
    %f10 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f10_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f11_p1 = getelementptr inbounds i8, i8* %data, i32 64
    %f11_p2 = getelementptr inbounds i8, i8* %f11_p1, <8 x i32> %offsets
    %f11_p3 = bitcast <8 x i8*> %f11_p2 to <8 x float*>
    %f11 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f11_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f12_p1 = getelementptr inbounds i8, i8* %data, i32 76
    %f12_p2 = getelementptr inbounds i8, i8* %f12_p1, <8 x i32> %offsets
    %f12_p3 = bitcast <8 x i8*> %f12_p2 to <8 x float*>
    %f12 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f12_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f13_p1 = getelementptr inbounds i8, i8* %data, i32 88
    %f13_p2 = getelementptr inbounds i8, i8* %f13_p1, <8 x i32> %offsets
    %f13_p3 = bitcast <8 x i8*> %f13_p2 to <8 x float*>
    %f13 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f13_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f14_p1 = getelementptr inbounds i8, i8* %data, i32 40
    %f14_p2 = getelementptr inbounds i8, i8* %f14_p1, <8 x i32> %offsets
    %f14_p3 = bitcast <8 x i8*> %f14_p2 to <8 x float*>
    %f14 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f14_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f15_p1 = getelementptr inbounds i8, i8* %data, i32 24
    %f15_p2 = getelementptr inbounds i8, i8* %f15_p1, <8 x i32> %offsets
    %f15_p3 = bitcast <8 x i8*> %f15_p2 to <8 x float*>
    %f15 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f15_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f16_p1 = getelementptr inbounds i8, i8* %data, i32 36
    %f16_p2 = getelementptr inbounds i8, i8* %f16_p1, <8 x i32> %offsets
    %f16_p3 = bitcast <8 x i8*> %f16_p2 to <8 x float*>
    %f16 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f16_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f17_p1 = getelementptr inbounds i8, i8* %data, i32 20
    %f17_p2 = getelementptr inbounds i8, i8* %f17_p1, <8 x i32> %offsets
    %f17_p3 = bitcast <8 x i8*> %f17_p2 to <8 x float*>
    %f17 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f17_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f18_p1 = getelementptr inbounds i8, i8* %data, i32 32
    %f18_p2 = getelementptr inbounds i8, i8* %f18_p1, <8 x i32> %offsets
    %f18_p3 = bitcast <8 x i8*> %f18_p2 to <8 x float*>
    %f18 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f18_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f19_p1 = getelementptr inbounds i8, i8* %data, i32 16
    %f19_p2 = getelementptr inbounds i8, i8* %f19_p1, <8 x i32> %offsets
    %f19_p3 = bitcast <8 x i8*> %f19_p2 to <8 x float*>
    %f19 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f19_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f20_p1 = getelementptr inbounds i8, i8* %data, i32 44
    %f20_p2 = getelementptr inbounds i8, i8* %f20_p1, <8 x i32> %offsets
    %f20_p3 = bitcast <8 x i8*> %f20_p2 to <8 x float*>
    %f20 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f20_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f21_p1 = getelementptr inbounds i8, i8* %data, i32 28
    %f21_p2 = getelementptr inbounds i8, i8* %f21_p1, <8 x i32> %offsets
    %f21_p3 = bitcast <8 x i8*> %f21_p2 to <8 x float*>
    %f21 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f21_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f22_i0 = getelementptr inbounds i8, i8* %params, i64 0
    %f22_i1 = bitcast i8* %f22_i0 to <8 x float>*
    %f22 = load <8 x float>, <8 x float>* %f22_i1, align 32
    %f23_i0 = getelementptr inbounds i8, i8* %params, i64 32
    %f23_i1 = bitcast i8* %f23_i0 to <8 x float>*
    %f23 = load <8 x float>, <8 x float>* %f23_i1, align 32
    %f24_i0 = getelementptr inbounds i8, i8* %params, i64 64
    %f24_i1 = bitcast i8* %f24_i0 to <8 x float>*
    %f24 = load <8 x float>, <8 x float>* %f24_i1, align 32
    %f25_i0 = getelementptr inbounds i8, i8* %params, i64 96
    %f25_i1 = bitcast i8* %f25_i0 to <8 x float>*
    %f25 = load <8 x float>, <8 x float>* %f25_i1, align 32
    %f26_i0 = getelementptr inbounds i8, i8* %params, i64 128
    %f26_i1 = bitcast i8* %f26_i0 to <8 x float>*
    %f26 = load <8 x float>, <8 x float>* %f26_i1, align 32
    %f27_i0 = getelementptr inbounds i8, i8* %params, i64 160
    %f27_i1 = bitcast i8* %f27_i0 to <8 x float>*
    %f27 = load <8 x float>, <8 x float>* %f27_i1, align 32
    %f28_i0 = getelementptr inbounds i8, i8* %params, i64 192
    %f28_i1 = bitcast i8* %f28_i0 to <8 x float>*
    %f28 = load <8 x float>, <8 x float>* %f28_i1, align 32
    %r29 = bitcast <8 x i32> %self to <8 x i32>
    %f30_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f30 = shufflevector <8 x float> %f30_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f31 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f25, <8 x float> %f28, <8 x float> %f22)
    %f32 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f26, <8 x float> %f28, <8 x float> %f23)
    %f33 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f27, <8 x float> %f28, <8 x float> %f24)
    %f34 = fsub <8 x float> %f31, %f2
    %f35 = fsub <8 x float> %f32, %f3
    %f36 = fsub <8 x float> %f33, %f4
    %f37 = fmul <8 x float> %f34, %f34
    %f38 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f35, <8 x float> %f35, <8 x float> %f37)
    %f39 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f36, <8 x float> %f36, <8 x float> %f38)
    %f40_1 = insertelement <8 x float> undef, float 0x3ff0000000000000, i32 0
    %f40 = shufflevector <8 x float> %f40_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f41 = fdiv <8 x float> %f40, %f39
    %f42 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> %f41)
    %f43 = fmul <8 x float> %f34, %f42
    %f44 = fmul <8 x float> %f35, %f42
    %f45 = fmul <8 x float> %f36, %f42
    %f46 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f43, <8 x float> %f1, <8 x float> %f2)
    %f47 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f44, <8 x float> %f1, <8 x float> %f3)
    %f48 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f45, <8 x float> %f1, <8 x float> %f4)
    %f49 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f14, <8 x float> %f46, <8 x float> %f20)
    %f50 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f15, <8 x float> %f46, <8 x float> %f21)
    %f51 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f16, <8 x float> %f47, <8 x float> %f49)
    %f52 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f17, <8 x float> %f47, <8 x float> %f50)
    %f53 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f18, <8 x float> %f48, <8 x float> %f51)
    %f54 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f19, <8 x float> %f48, <8 x float> %f52)
    %f55_1 = insertelement <8 x float> undef, float 0x401921fb60000000, i32 0
    %f55 = shufflevector <8 x float> %f55_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f56 = fneg <8 x float> %f54
    %f57 = fmul <8 x float> %f5, %f56
    %f58 = fmul <8 x float> %f6, %f56
    %f59 = fmul <8 x float> %f7, %f56
    %f60 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f8, <8 x float> %f53, <8 x float> %f57)
    %f61 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f9, <8 x float> %f53, <8 x float> %f58)
    %f62 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f10, <8 x float> %f53, <8 x float> %f59)
    %f63 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f11, <8 x float> %f30, <8 x float> %f60)
    %f64 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f12, <8 x float> %f30, <8 x float> %f61)
    %f65 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f13, <8 x float> %f30, <8 x float> %f62)
    %f66 = fmul <8 x float> %f63, %f55
    %f67 = fmul <8 x float> %f64, %f55
    %f68 = fmul <8 x float> %f65, %f55
    %out_0_0 = getelementptr inbounds i8, i8* %params, i64 320
    %out_0_1 = bitcast i8* %out_0_0 to <8 x float>*
    %out_0_2 = load <8 x float>, <8 x float>* %out_0_1, align 32
    %out_0_3 = select <8 x i1> %mask, <8 x float> %f67, <8 x float> %out_0_2
    store <8 x float> %out_0_3, <8 x float>* %out_0_1, align 32
    %out_2_0 = getelementptr inbounds i8, i8* %params, i64 352
    %out_2_1 = bitcast i8* %out_2_0 to <8 x float>*
    %out_2_2 = load <8 x float>, <8 x float>* %out_2_1, align 32
    %out_2_3 = select <8 x i1> %mask, <8 x float> %f46, <8 x float> %out_2_2
    store <8 x float> %out_2_3, <8 x float>* %out_2_1, align 32
    %out_3_0 = getelementptr inbounds i8, i8* %params, i64 384
    %out_3_1 = bitcast i8* %out_3_0 to <8 x float>*
    %out_3_2 = load <8 x float>, <8 x float>* %out_3_1, align 32
    %out_3_3 = select <8 x i1> %mask, <8 x float> %f47, <8 x float> %out_3_2
    store <8 x float> %out_3_3, <8 x float>* %out_3_1, align 32
    %out_4_0 = getelementptr inbounds i8, i8* %params, i64 416
    %out_4_1 = bitcast i8* %out_4_0 to <8 x float>*
    %out_4_2 = load <8 x float>, <8 x float>* %out_4_1, align 32
    %out_4_3 = select <8 x i1> %mask, <8 x float> %f48, <8 x float> %out_4_2
    store <8 x float> %out_4_3, <8 x float>* %out_4_1, align 32
    %out_5_0 = getelementptr inbounds i8, i8* %params, i64 448
    %out_5_1 = bitcast i8* %out_5_0 to <8 x float>*
    %out_5_2 = load <8 x float>, <8 x float>* %out_5_1, align 32
    %out_5_3 = select <8 x i1> %mask, <8 x float> %f43, <8 x float> %out_5_2
    store <8 x float> %out_5_3, <8 x float>* %out_5_1, align 32
    %out_6_0 = getelementptr inbounds i8, i8* %params, i64 480
    %out_6_1 = bitcast i8* %out_6_0 to <8 x float>*
    %out_6_2 = load <8 x float>, <8 x float>* %out_6_1, align 32
    %out_6_3 = select <8 x i1> %mask, <8 x float> %f44, <8 x float> %out_6_2
    store <8 x float> %out_6_3, <8 x float>* %out_6_1, align 32
    %out_7_0 = getelementptr inbounds i8, i8* %params, i64 512
    %out_7_1 = bitcast i8* %out_7_0 to <8 x float>*
    %out_7_2 = load <8 x float>, <8 x float>* %out_7_1, align 32
    %out_7_3 = select <8 x i1> %mask, <8 x float> %f45, <8 x float> %out_7_2
    store <8 x float> %out_7_3, <8 x float>* %out_7_1, align 32
    %out_8_0 = getelementptr inbounds i8, i8* %params, i64 544
    %out_8_1 = bitcast i8* %out_8_0 to <8 x i32>*
    %out_8_2 = load <8 x i32>, <8 x i32>* %out_8_1, align 32
    %out_8_3 = select <8 x i1> %mask, <8 x i32> %r29, <8 x i32> %out_8_2
    store <8 x i32> %out_8_3, <8 x i32>* %out_8_1, align 32
    %out_13_0 = getelementptr inbounds i8, i8* %params, i64 576
    %out_13_1 = bitcast i8* %out_13_0 to <8 x float>*
    %out_13_2 = load <8 x float>, <8 x float>* %out_13_1, align 32
    %out_13_3 = select <8 x i1> %mask, <8 x float> %f68, <8 x float> %out_13_2
    store <8 x float> %out_13_3, <8 x float>* %out_13_1, align 32
    %out_14_0 = getelementptr inbounds i8, i8* %params, i64 608
    %out_14_1 = bitcast i8* %out_14_0 to <8 x float>*
    %out_14_2 = load <8 x float>, <8 x float>* %out_14_1, align 32
    %out_14_3 = select <8 x i1> %mask, <8 x float> %f66, <8 x float> %out_14_2
    store <8 x float> %out_14_3, <8 x float>* %out_14_1, align 32
    %out_15_0 = getelementptr inbounds i8, i8* %params, i64 640
    %out_15_1 = bitcast i8* %out_15_0 to <8 x float>*
    %out_15_2 = load <8 x float>, <8 x float>* %out_15_1, align 32
    %out_15_3 = select <8 x i1> %mask, <8 x float> %f45, <8 x float> %out_15_2
    store <8 x float> %out_15_3, <8 x float>* %out_15_1, align 32
    %out_16_0 = getelementptr inbounds i8, i8* %params, i64 672
    %out_16_1 = bitcast i8* %out_16_0 to <8 x float>*
    %out_16_2 = load <8 x float>, <8 x float>* %out_16_1, align 32
    %out_16_3 = select <8 x i1> %mask, <8 x float> %f44, <8 x float> %out_16_2
    store <8 x float> %out_16_3, <8 x float>* %out_16_1, align 32
    %out_17_0 = getelementptr inbounds i8, i8* %params, i64 704
    %out_17_1 = bitcast i8* %out_17_0 to <8 x float>*
    %out_17_2 = load <8 x float>, <8 x float>* %out_17_1, align 32
    %out_17_3 = select <8 x i1> %mask, <8 x float> %f43, <8 x float> %out_17_2
    store <8 x float> %out_17_3, <8 x float>* %out_17_1, align 32
    ret void;
}

define void @func_570d1d9103e40ae9598d8f351c2e5c2a(<8 x i1> %mask, <8 x i32> %self, i8* noalias %params, i8* noalias %data, <8 x i32> %offsets) #0 {
entry:
    ; VCall: mitsuba::Shape::compute_surface_interaction()
    %f1_i0 = getelementptr inbounds i8, i8* %params, i64 224
    %f1_i1 = bitcast i8* %f1_i0 to <8 x float>*
    %f1 = load <8 x float>, <8 x float>* %f1_i1, align 32
    %f2_i0 = getelementptr inbounds i8, i8* %params, i64 256
    %f2_i1 = bitcast i8* %f2_i0 to <8 x float>*
    %f2 = load <8 x float>, <8 x float>* %f2_i1, align 32
    %r3_i0 = getelementptr inbounds i8, i8* %params, i64 288
    %r3_i1 = bitcast i8* %r3_i0 to <8 x i32>*
    %r3 = load <8 x i32>, <8 x i32>* %r3_i1, align 32
    %f4_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f4 = shufflevector <8 x float> %f4_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f5_1 = insertelement <8 x float> undef, float 0x3ff0000000000000, i32 0
    %f5 = shufflevector <8 x float> %f5_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f6_1 = insertelement <8 x float> undef, float 0xbff0000000000000, i32 0
    %f6 = shufflevector <8 x float> %f6_1, <8 x float> undef, <8 x i32> zeroinitializer
    %r7 = bitcast <8 x i32> %self to <8 x i32>
    %p8 = bitcast <8 x i1> %mask to <8 x i1>
    %r9_1 = insertelement <8 x i32> undef, i32 3, i32 0
    %r9 = shufflevector <8 x i32> %r9_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r10 = mul <8 x i32> %r3, %r9
    %r11_1 = insertelement <8 x i32> undef, i32 1, i32 0
    %r11 = shufflevector <8 x i32> %r11_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r12 = add <8 x i32> %r10, %r11
    %r13_1 = insertelement <8 x i32> undef, i32 2, i32 0
    %r13 = shufflevector <8 x i32> %r13_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r14 = add <8 x i32> %r10, %r13
    %rd15_p1 = getelementptr inbounds i8, i8* %data, i32 8
    %rd15_p2 = getelementptr inbounds i8, i8* %rd15_p1, <8 x i32> %offsets
    %rd15_p3 = bitcast <8 x i8*> %rd15_p2 to <8 x i64*>
    %rd15_p4 = call <8 x i64> @llvm.masked.gather.v8i64(<8 x i64*> %rd15_p3, i32 8, <8 x i1> %mask, <8 x i64> zeroinitializer)
    %rd15 = inttoptr <8 x i64> %rd15_p4 to <8 x i8*>
    %r16_0 = bitcast <8 x i8*> %rd15 to <8 x i32*>
    %r16_1 = getelementptr i32, <8 x i32*> %r16_0, <8 x i32> %r10
    %r16 = call <8 x i32> @llvm.masked.gather.v8i32(<8 x i32*> %r16_1, i32 4, <8 x i1> %p8, <8 x i32> zeroinitializer)
    %r17_0 = bitcast <8 x i8*> %rd15 to <8 x i32*>
    %r17_1 = getelementptr i32, <8 x i32*> %r17_0, <8 x i32> %r12
    %r17 = call <8 x i32> @llvm.masked.gather.v8i32(<8 x i32*> %r17_1, i32 4, <8 x i1> %p8, <8 x i32> zeroinitializer)
    %r18_0 = bitcast <8 x i8*> %rd15 to <8 x i32*>
    %r18_1 = getelementptr i32, <8 x i32*> %r18_0, <8 x i32> %r14
    %r18 = call <8 x i32> @llvm.masked.gather.v8i32(<8 x i32*> %r18_1, i32 4, <8 x i1> %p8, <8 x i32> zeroinitializer)
    %r19 = mul <8 x i32> %r16, %r9
    %r20 = add <8 x i32> %r19, %r11
    %r21 = add <8 x i32> %r19, %r13
    %rd22_p1 = getelementptr inbounds i8, i8* %data, i32 0
    %rd22_p2 = getelementptr inbounds i8, i8* %rd22_p1, <8 x i32> %offsets
    %rd22_p3 = bitcast <8 x i8*> %rd22_p2 to <8 x i64*>
    %rd22_p4 = call <8 x i64> @llvm.masked.gather.v8i64(<8 x i64*> %rd22_p3, i32 8, <8 x i1> %mask, <8 x i64> zeroinitializer)
    %rd22 = inttoptr <8 x i64> %rd22_p4 to <8 x i8*>
    %f23_0 = bitcast <8 x i8*> %rd22 to <8 x float*>
    %f23_1 = getelementptr float, <8 x float*> %f23_0, <8 x i32> %r19
    %f23 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f23_1, i32 4, <8 x i1> %p8, <8 x float> zeroinitializer)
    %f24_0 = bitcast <8 x i8*> %rd22 to <8 x float*>
    %f24_1 = getelementptr float, <8 x float*> %f24_0, <8 x i32> %r20
    %f24 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f24_1, i32 4, <8 x i1> %p8, <8 x float> zeroinitializer)
    %f25_0 = bitcast <8 x i8*> %rd22 to <8 x float*>
    %f25_1 = getelementptr float, <8 x float*> %f25_0, <8 x i32> %r21
    %f25 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f25_1, i32 4, <8 x i1> %p8, <8 x float> zeroinitializer)
    %r26 = mul <8 x i32> %r17, %r9
    %r27 = add <8 x i32> %r26, %r11
    %r28 = add <8 x i32> %r26, %r13
    %f29_0 = bitcast <8 x i8*> %rd22 to <8 x float*>
    %f29_1 = getelementptr float, <8 x float*> %f29_0, <8 x i32> %r26
    %f29 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f29_1, i32 4, <8 x i1> %p8, <8 x float> zeroinitializer)
    %f30_0 = bitcast <8 x i8*> %rd22 to <8 x float*>
    %f30_1 = getelementptr float, <8 x float*> %f30_0, <8 x i32> %r27
    %f30 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f30_1, i32 4, <8 x i1> %p8, <8 x float> zeroinitializer)
    %f31_0 = bitcast <8 x i8*> %rd22 to <8 x float*>
    %f31_1 = getelementptr float, <8 x float*> %f31_0, <8 x i32> %r28
    %f31 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f31_1, i32 4, <8 x i1> %p8, <8 x float> zeroinitializer)
    %r32 = mul <8 x i32> %r18, %r9
    %r33 = add <8 x i32> %r32, %r11
    %r34 = add <8 x i32> %r32, %r13
    %f35_0 = bitcast <8 x i8*> %rd22 to <8 x float*>
    %f35_1 = getelementptr float, <8 x float*> %f35_0, <8 x i32> %r32
    %f35 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f35_1, i32 4, <8 x i1> %p8, <8 x float> zeroinitializer)
    %f36_0 = bitcast <8 x i8*> %rd22 to <8 x float*>
    %f36_1 = getelementptr float, <8 x float*> %f36_0, <8 x i32> %r33
    %f36 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f36_1, i32 4, <8 x i1> %p8, <8 x float> zeroinitializer)
    %f37_0 = bitcast <8 x i8*> %rd22 to <8 x float*>
    %f37_1 = getelementptr float, <8 x float*> %f37_0, <8 x i32> %r34
    %f37 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f37_1, i32 4, <8 x i1> %p8, <8 x float> zeroinitializer)
    %f38 = fsub <8 x float> %f5, %f1
    %f39 = fsub <8 x float> %f38, %f2
    %f40 = fsub <8 x float> %f29, %f23
    %f41 = fsub <8 x float> %f30, %f24
    %f42 = fsub <8 x float> %f31, %f25
    %f43 = fsub <8 x float> %f35, %f23
    %f44 = fsub <8 x float> %f36, %f24
    %f45 = fsub <8 x float> %f37, %f25
    %f46 = fmul <8 x float> %f35, %f2
    %f47 = fmul <8 x float> %f36, %f2
    %f48 = fmul <8 x float> %f37, %f2
    %f49 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f29, <8 x float> %f1, <8 x float> %f46)
    %f50 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f30, <8 x float> %f1, <8 x float> %f47)
    %f51 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f31, <8 x float> %f1, <8 x float> %f48)
    %f52 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f23, <8 x float> %f39, <8 x float> %f49)
    %f53 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f24, <8 x float> %f39, <8 x float> %f50)
    %f54 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f25, <8 x float> %f39, <8 x float> %f51)
    %f55 = fmul <8 x float> %f42, %f44
    %f56 = fmul <8 x float> %f40, %f45
    %f57 = fmul <8 x float> %f41, %f43
    %f58 = fneg <8 x float> %f55
    %f59 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f41, <8 x float> %f45, <8 x float> %f58)
    %f60 = fneg <8 x float> %f56
    %f61 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f42, <8 x float> %f43, <8 x float> %f60)
    %f62 = fneg <8 x float> %f57
    %f63 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f40, <8 x float> %f44, <8 x float> %f62)
    %f64 = fmul <8 x float> %f59, %f59
    %f65 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f61, <8 x float> %f61, <8 x float> %f64)
    %f66 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f63, <8 x float> %f63, <8 x float> %f65)
    %f67 = fdiv <8 x float> %f5, %f66
    %f68 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> %f67)
    %f69 = fmul <8 x float> %f59, %f68
    %f70 = fmul <8 x float> %f61, %f68
    %f71 = fmul <8 x float> %f63, %f68
    %p72 = fcmp oge <8 x float> %f71, %f4
    %f73 = select <8 x i1> %p72, <8 x float> %f5, <8 x float> %f6
    %f74 = fadd <8 x float> %f73, %f71
    %f75 = fdiv <8 x float> %f5, %f74
    %f76 = fneg <8 x float> %f75
    %f77 = fmul <8 x float> %f69, %f70
    %f78 = fmul <8 x float> %f77, %f76
    %f79 = fmul <8 x float> %f69, %f69
    %f80 = fmul <8 x float> %f79, %f76
    %f81 = fneg <8 x float> %f80
    %f82 = select <8 x i1> %p72, <8 x float> %f80, <8 x float> %f81
    %f83 = fadd <8 x float> %f82, %f5
    %f84 = fneg <8 x float> %f78
    %f85 = select <8 x i1> %p72, <8 x float> %f78, <8 x float> %f84
    %f86 = fneg <8 x float> %f69
    %f87 = select <8 x i1> %p72, <8 x float> %f86, <8 x float> %f69
    %rd88_p1 = getelementptr inbounds i8, i8* %data, i32 16
    %rd88_p2 = getelementptr inbounds i8, i8* %rd88_p1, <8 x i32> %offsets
    %rd88_p3 = bitcast <8 x i8*> %rd88_p2 to <8 x i64*>
    %rd88_p4 = call <8 x i64> @llvm.masked.gather.v8i64(<8 x i64*> %rd88_p3, i32 8, <8 x i1> %mask, <8 x i64> zeroinitializer)
    %rd88 = inttoptr <8 x i64> %rd88_p4 to <8 x i8*>
    %f89_0 = bitcast <8 x i8*> %rd88 to <8 x float*>
    %f89_1 = getelementptr float, <8 x float*> %f89_0, <8 x i32> %r19
    %f89 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f89_1, i32 4, <8 x i1> %p8, <8 x float> zeroinitializer)
    %f90_0 = bitcast <8 x i8*> %rd88 to <8 x float*>
    %f90_1 = getelementptr float, <8 x float*> %f90_0, <8 x i32> %r20
    %f90 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f90_1, i32 4, <8 x i1> %p8, <8 x float> zeroinitializer)
    %f91_0 = bitcast <8 x i8*> %rd88 to <8 x float*>
    %f91_1 = getelementptr float, <8 x float*> %f91_0, <8 x i32> %r21
    %f91 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f91_1, i32 4, <8 x i1> %p8, <8 x float> zeroinitializer)
    %f92_0 = bitcast <8 x i8*> %rd88 to <8 x float*>
    %f92_1 = getelementptr float, <8 x float*> %f92_0, <8 x i32> %r26
    %f92 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f92_1, i32 4, <8 x i1> %p8, <8 x float> zeroinitializer)
    %f93_0 = bitcast <8 x i8*> %rd88 to <8 x float*>
    %f93_1 = getelementptr float, <8 x float*> %f93_0, <8 x i32> %r27
    %f93 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f93_1, i32 4, <8 x i1> %p8, <8 x float> zeroinitializer)
    %f94_0 = bitcast <8 x i8*> %rd88 to <8 x float*>
    %f94_1 = getelementptr float, <8 x float*> %f94_0, <8 x i32> %r28
    %f94 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f94_1, i32 4, <8 x i1> %p8, <8 x float> zeroinitializer)
    %f95_0 = bitcast <8 x i8*> %rd88 to <8 x float*>
    %f95_1 = getelementptr float, <8 x float*> %f95_0, <8 x i32> %r32
    %f95 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f95_1, i32 4, <8 x i1> %p8, <8 x float> zeroinitializer)
    %f96_0 = bitcast <8 x i8*> %rd88 to <8 x float*>
    %f96_1 = getelementptr float, <8 x float*> %f96_0, <8 x i32> %r33
    %f96 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f96_1, i32 4, <8 x i1> %p8, <8 x float> zeroinitializer)
    %f97_0 = bitcast <8 x i8*> %rd88 to <8 x float*>
    %f97_1 = getelementptr float, <8 x float*> %f97_0, <8 x i32> %r34
    %f97 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f97_1, i32 4, <8 x i1> %p8, <8 x float> zeroinitializer)
    %f98 = fmul <8 x float> %f89, %f39
    %f99 = fmul <8 x float> %f90, %f39
    %f100 = fmul <8 x float> %f91, %f39
    %f101 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f92, <8 x float> %f1, <8 x float> %f98)
    %f102 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f93, <8 x float> %f1, <8 x float> %f99)
    %f103 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f94, <8 x float> %f1, <8 x float> %f100)
    %f104 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f95, <8 x float> %f2, <8 x float> %f101)
    %f105 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f96, <8 x float> %f2, <8 x float> %f102)
    %f106 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f97, <8 x float> %f2, <8 x float> %f103)
    %f107 = fmul <8 x float> %f104, %f104
    %f108 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f105, <8 x float> %f105, <8 x float> %f107)
    %f109 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f106, <8 x float> %f106, <8 x float> %f108)
    %f110 = fdiv <8 x float> %f5, %f109
    %f111 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> %f110)
    %f112 = fmul <8 x float> %f104, %f111
    %f113 = fmul <8 x float> %f105, %f111
    %f114 = fmul <8 x float> %f106, %f111
    %out_0_0 = getelementptr inbounds i8, i8* %params, i64 320
    %out_0_1 = bitcast i8* %out_0_0 to <8 x float>*
    %out_0_2 = load <8 x float>, <8 x float>* %out_0_1, align 32
    %out_0_3 = select <8 x i1> %mask, <8 x float> %f85, <8 x float> %out_0_2
    store <8 x float> %out_0_3, <8 x float>* %out_0_1, align 32
    %out_2_0 = getelementptr inbounds i8, i8* %params, i64 352
    %out_2_1 = bitcast i8* %out_2_0 to <8 x float>*
    %out_2_2 = load <8 x float>, <8 x float>* %out_2_1, align 32
    %out_2_3 = select <8 x i1> %mask, <8 x float> %f52, <8 x float> %out_2_2
    store <8 x float> %out_2_3, <8 x float>* %out_2_1, align 32
    %out_3_0 = getelementptr inbounds i8, i8* %params, i64 384
    %out_3_1 = bitcast i8* %out_3_0 to <8 x float>*
    %out_3_2 = load <8 x float>, <8 x float>* %out_3_1, align 32
    %out_3_3 = select <8 x i1> %mask, <8 x float> %f53, <8 x float> %out_3_2
    store <8 x float> %out_3_3, <8 x float>* %out_3_1, align 32
    %out_4_0 = getelementptr inbounds i8, i8* %params, i64 416
    %out_4_1 = bitcast i8* %out_4_0 to <8 x float>*
    %out_4_2 = load <8 x float>, <8 x float>* %out_4_1, align 32
    %out_4_3 = select <8 x i1> %mask, <8 x float> %f54, <8 x float> %out_4_2
    store <8 x float> %out_4_3, <8 x float>* %out_4_1, align 32
    %out_5_0 = getelementptr inbounds i8, i8* %params, i64 448
    %out_5_1 = bitcast i8* %out_5_0 to <8 x float>*
    %out_5_2 = load <8 x float>, <8 x float>* %out_5_1, align 32
    %out_5_3 = select <8 x i1> %mask, <8 x float> %f69, <8 x float> %out_5_2
    store <8 x float> %out_5_3, <8 x float>* %out_5_1, align 32
    %out_6_0 = getelementptr inbounds i8, i8* %params, i64 480
    %out_6_1 = bitcast i8* %out_6_0 to <8 x float>*
    %out_6_2 = load <8 x float>, <8 x float>* %out_6_1, align 32
    %out_6_3 = select <8 x i1> %mask, <8 x float> %f70, <8 x float> %out_6_2
    store <8 x float> %out_6_3, <8 x float>* %out_6_1, align 32
    %out_7_0 = getelementptr inbounds i8, i8* %params, i64 512
    %out_7_1 = bitcast i8* %out_7_0 to <8 x float>*
    %out_7_2 = load <8 x float>, <8 x float>* %out_7_1, align 32
    %out_7_3 = select <8 x i1> %mask, <8 x float> %f71, <8 x float> %out_7_2
    store <8 x float> %out_7_3, <8 x float>* %out_7_1, align 32
    %out_8_0 = getelementptr inbounds i8, i8* %params, i64 544
    %out_8_1 = bitcast i8* %out_8_0 to <8 x i32>*
    %out_8_2 = load <8 x i32>, <8 x i32>* %out_8_1, align 32
    %out_8_3 = select <8 x i1> %mask, <8 x i32> %r7, <8 x i32> %out_8_2
    store <8 x i32> %out_8_3, <8 x i32>* %out_8_1, align 32
    %out_13_0 = getelementptr inbounds i8, i8* %params, i64 576
    %out_13_1 = bitcast i8* %out_13_0 to <8 x float>*
    %out_13_2 = load <8 x float>, <8 x float>* %out_13_1, align 32
    %out_13_3 = select <8 x i1> %mask, <8 x float> %f87, <8 x float> %out_13_2
    store <8 x float> %out_13_3, <8 x float>* %out_13_1, align 32
    %out_14_0 = getelementptr inbounds i8, i8* %params, i64 608
    %out_14_1 = bitcast i8* %out_14_0 to <8 x float>*
    %out_14_2 = load <8 x float>, <8 x float>* %out_14_1, align 32
    %out_14_3 = select <8 x i1> %mask, <8 x float> %f83, <8 x float> %out_14_2
    store <8 x float> %out_14_3, <8 x float>* %out_14_1, align 32
    %out_15_0 = getelementptr inbounds i8, i8* %params, i64 640
    %out_15_1 = bitcast i8* %out_15_0 to <8 x float>*
    %out_15_2 = load <8 x float>, <8 x float>* %out_15_1, align 32
    %out_15_3 = select <8 x i1> %mask, <8 x float> %f114, <8 x float> %out_15_2
    store <8 x float> %out_15_3, <8 x float>* %out_15_1, align 32
    %out_16_0 = getelementptr inbounds i8, i8* %params, i64 672
    %out_16_1 = bitcast i8* %out_16_0 to <8 x float>*
    %out_16_2 = load <8 x float>, <8 x float>* %out_16_1, align 32
    %out_16_3 = select <8 x i1> %mask, <8 x float> %f113, <8 x float> %out_16_2
    store <8 x float> %out_16_3, <8 x float>* %out_16_1, align 32
    %out_17_0 = getelementptr inbounds i8, i8* %params, i64 704
    %out_17_1 = bitcast i8* %out_17_0 to <8 x float>*
    %out_17_2 = load <8 x float>, <8 x float>* %out_17_1, align 32
    %out_17_3 = select <8 x i1> %mask, <8 x float> %f112, <8 x float> %out_17_2
    store <8 x float> %out_17_3, <8 x float>* %out_17_1, align 32
    ret void;
}

define void @func_e17fd895e12cc8bec8c4e55075a069b3(<8 x i1> %mask, <8 x i32> %self, i8* noalias %params, i8* noalias %data, <8 x i32> %offsets) #0 {
entry:
    ; VCall: mitsuba::Shape::compute_surface_interaction()
    %f1_p1 = getelementptr inbounds i8, i8* %data, i32 12
    %f1_p2 = getelementptr inbounds i8, i8* %f1_p1, <8 x i32> %offsets
    %f1_p3 = bitcast <8 x i8*> %f1_p2 to <8 x float*>
    %f1 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f1_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f2_p1 = getelementptr inbounds i8, i8* %data, i32 0
    %f2_p2 = getelementptr inbounds i8, i8* %f2_p1, <8 x i32> %offsets
    %f2_p3 = bitcast <8 x i8*> %f2_p2 to <8 x float*>
    %f2 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f2_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f3_p1 = getelementptr inbounds i8, i8* %data, i32 8
    %f3_p2 = getelementptr inbounds i8, i8* %f3_p1, <8 x i32> %offsets
    %f3_p3 = bitcast <8 x i8*> %f3_p2 to <8 x float*>
    %f3 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f3_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f4_p1 = getelementptr inbounds i8, i8* %data, i32 4
    %f4_p2 = getelementptr inbounds i8, i8* %f4_p1, <8 x i32> %offsets
    %f4_p3 = bitcast <8 x i8*> %f4_p2 to <8 x float*>
    %f4 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f4_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f5_p1 = getelementptr inbounds i8, i8* %data, i32 72
    %f5_p2 = getelementptr inbounds i8, i8* %f5_p1, <8 x i32> %offsets
    %f5_p3 = bitcast <8 x i8*> %f5_p2 to <8 x float*>
    %f5 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f5_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f6_p1 = getelementptr inbounds i8, i8* %data, i32 84
    %f6_p2 = getelementptr inbounds i8, i8* %f6_p1, <8 x i32> %offsets
    %f6_p3 = bitcast <8 x i8*> %f6_p2 to <8 x float*>
    %f6 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f6_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f7_p1 = getelementptr inbounds i8, i8* %data, i32 96
    %f7_p2 = getelementptr inbounds i8, i8* %f7_p1, <8 x i32> %offsets
    %f7_p3 = bitcast <8 x i8*> %f7_p2 to <8 x float*>
    %f7 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f7_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f8_p1 = getelementptr inbounds i8, i8* %data, i32 68
    %f8_p2 = getelementptr inbounds i8, i8* %f8_p1, <8 x i32> %offsets
    %f8_p3 = bitcast <8 x i8*> %f8_p2 to <8 x float*>
    %f8 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f8_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f9_p1 = getelementptr inbounds i8, i8* %data, i32 80
    %f9_p2 = getelementptr inbounds i8, i8* %f9_p1, <8 x i32> %offsets
    %f9_p3 = bitcast <8 x i8*> %f9_p2 to <8 x float*>
    %f9 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f9_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f10_p1 = getelementptr inbounds i8, i8* %data, i32 92
    %f10_p2 = getelementptr inbounds i8, i8* %f10_p1, <8 x i32> %offsets
    %f10_p3 = bitcast <8 x i8*> %f10_p2 to <8 x float*>
    %f10 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f10_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f11_p1 = getelementptr inbounds i8, i8* %data, i32 64
    %f11_p2 = getelementptr inbounds i8, i8* %f11_p1, <8 x i32> %offsets
    %f11_p3 = bitcast <8 x i8*> %f11_p2 to <8 x float*>
    %f11 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f11_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f12_p1 = getelementptr inbounds i8, i8* %data, i32 76
    %f12_p2 = getelementptr inbounds i8, i8* %f12_p1, <8 x i32> %offsets
    %f12_p3 = bitcast <8 x i8*> %f12_p2 to <8 x float*>
    %f12 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f12_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f13_p1 = getelementptr inbounds i8, i8* %data, i32 88
    %f13_p2 = getelementptr inbounds i8, i8* %f13_p1, <8 x i32> %offsets
    %f13_p3 = bitcast <8 x i8*> %f13_p2 to <8 x float*>
    %f13 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f13_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f14_p1 = getelementptr inbounds i8, i8* %data, i32 40
    %f14_p2 = getelementptr inbounds i8, i8* %f14_p1, <8 x i32> %offsets
    %f14_p3 = bitcast <8 x i8*> %f14_p2 to <8 x float*>
    %f14 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f14_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f15_p1 = getelementptr inbounds i8, i8* %data, i32 24
    %f15_p2 = getelementptr inbounds i8, i8* %f15_p1, <8 x i32> %offsets
    %f15_p3 = bitcast <8 x i8*> %f15_p2 to <8 x float*>
    %f15 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f15_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f16_p1 = getelementptr inbounds i8, i8* %data, i32 36
    %f16_p2 = getelementptr inbounds i8, i8* %f16_p1, <8 x i32> %offsets
    %f16_p3 = bitcast <8 x i8*> %f16_p2 to <8 x float*>
    %f16 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f16_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f17_p1 = getelementptr inbounds i8, i8* %data, i32 20
    %f17_p2 = getelementptr inbounds i8, i8* %f17_p1, <8 x i32> %offsets
    %f17_p3 = bitcast <8 x i8*> %f17_p2 to <8 x float*>
    %f17 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f17_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f18_p1 = getelementptr inbounds i8, i8* %data, i32 32
    %f18_p2 = getelementptr inbounds i8, i8* %f18_p1, <8 x i32> %offsets
    %f18_p3 = bitcast <8 x i8*> %f18_p2 to <8 x float*>
    %f18 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f18_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f19_p1 = getelementptr inbounds i8, i8* %data, i32 16
    %f19_p2 = getelementptr inbounds i8, i8* %f19_p1, <8 x i32> %offsets
    %f19_p3 = bitcast <8 x i8*> %f19_p2 to <8 x float*>
    %f19 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f19_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f20_p1 = getelementptr inbounds i8, i8* %data, i32 44
    %f20_p2 = getelementptr inbounds i8, i8* %f20_p1, <8 x i32> %offsets
    %f20_p3 = bitcast <8 x i8*> %f20_p2 to <8 x float*>
    %f20 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f20_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f21_p1 = getelementptr inbounds i8, i8* %data, i32 28
    %f21_p2 = getelementptr inbounds i8, i8* %f21_p1, <8 x i32> %offsets
    %f21_p3 = bitcast <8 x i8*> %f21_p2 to <8 x float*>
    %f21 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f21_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f22_i0 = getelementptr inbounds i8, i8* %params, i64 0
    %f22_i1 = bitcast i8* %f22_i0 to <8 x float>*
    %f22 = load <8 x float>, <8 x float>* %f22_i1, align 32
    %f23_i0 = getelementptr inbounds i8, i8* %params, i64 32
    %f23_i1 = bitcast i8* %f23_i0 to <8 x float>*
    %f23 = load <8 x float>, <8 x float>* %f23_i1, align 32
    %f24_i0 = getelementptr inbounds i8, i8* %params, i64 64
    %f24_i1 = bitcast i8* %f24_i0 to <8 x float>*
    %f24 = load <8 x float>, <8 x float>* %f24_i1, align 32
    %f25_i0 = getelementptr inbounds i8, i8* %params, i64 96
    %f25_i1 = bitcast i8* %f25_i0 to <8 x float>*
    %f25 = load <8 x float>, <8 x float>* %f25_i1, align 32
    %f26_i0 = getelementptr inbounds i8, i8* %params, i64 128
    %f26_i1 = bitcast i8* %f26_i0 to <8 x float>*
    %f26 = load <8 x float>, <8 x float>* %f26_i1, align 32
    %f27_i0 = getelementptr inbounds i8, i8* %params, i64 160
    %f27_i1 = bitcast i8* %f27_i0 to <8 x float>*
    %f27 = load <8 x float>, <8 x float>* %f27_i1, align 32
    %f28_i0 = getelementptr inbounds i8, i8* %params, i64 192
    %f28_i1 = bitcast i8* %f28_i0 to <8 x float>*
    %f28 = load <8 x float>, <8 x float>* %f28_i1, align 32
    %f29_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f29 = shufflevector <8 x float> %f29_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f30 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f25, <8 x float> %f28, <8 x float> %f22)
    %f31 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f26, <8 x float> %f28, <8 x float> %f23)
    %f32 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f27, <8 x float> %f28, <8 x float> %f24)
    %f33_1 = insertelement <8 x float> undef, float 0x3ff0000000000000, i32 0
    %f33 = shufflevector <8 x float> %f33_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f34_1 = insertelement <8 x float> undef, float 0x401921fb60000000, i32 0
    %f34 = shufflevector <8 x float> %f34_1, <8 x float> undef, <8 x i32> zeroinitializer
    %r35 = bitcast <8 x i32> %self to <8 x i32>
    %f36 = fsub <8 x float> %f30, %f2
    %f37 = fsub <8 x float> %f31, %f3
    %f38 = fsub <8 x float> %f32, %f4
    %f39 = fmul <8 x float> %f36, %f36
    %f40 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f37, <8 x float> %f37, <8 x float> %f39)
    %f41 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f38, <8 x float> %f38, <8 x float> %f40)
    %f42 = fdiv <8 x float> %f33, %f41
    %f43 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> %f42)
    %f44 = fmul <8 x float> %f36, %f43
    %f45 = fmul <8 x float> %f37, %f43
    %f46 = fmul <8 x float> %f38, %f43
    %f47 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f44, <8 x float> %f1, <8 x float> %f2)
    %f48 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f45, <8 x float> %f1, <8 x float> %f3)
    %f49 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f46, <8 x float> %f1, <8 x float> %f4)
    %f50 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f14, <8 x float> %f47, <8 x float> %f20)
    %f51 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f15, <8 x float> %f47, <8 x float> %f21)
    %f52 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f16, <8 x float> %f48, <8 x float> %f50)
    %f53 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f17, <8 x float> %f48, <8 x float> %f51)
    %f54 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f18, <8 x float> %f49, <8 x float> %f52)
    %f55 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f19, <8 x float> %f49, <8 x float> %f53)
    %f56 = fneg <8 x float> %f55
    %f57 = fmul <8 x float> %f5, %f56
    %f58 = fmul <8 x float> %f6, %f56
    %f59 = fmul <8 x float> %f7, %f56
    %f60 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f8, <8 x float> %f54, <8 x float> %f57)
    %f61 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f9, <8 x float> %f54, <8 x float> %f58)
    %f62 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f10, <8 x float> %f54, <8 x float> %f59)
    %f63 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f11, <8 x float> %f29, <8 x float> %f60)
    %f64 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f12, <8 x float> %f29, <8 x float> %f61)
    %f65 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f13, <8 x float> %f29, <8 x float> %f62)
    %f66 = fmul <8 x float> %f63, %f34
    %f67 = fmul <8 x float> %f64, %f34
    %f68 = fmul <8 x float> %f65, %f34
    %out_0_0 = getelementptr inbounds i8, i8* %params, i64 320
    %out_0_1 = bitcast i8* %out_0_0 to <8 x float>*
    %out_0_2 = load <8 x float>, <8 x float>* %out_0_1, align 32
    %out_0_3 = select <8 x i1> %mask, <8 x float> %f67, <8 x float> %out_0_2
    store <8 x float> %out_0_3, <8 x float>* %out_0_1, align 32
    %out_2_0 = getelementptr inbounds i8, i8* %params, i64 352
    %out_2_1 = bitcast i8* %out_2_0 to <8 x float>*
    %out_2_2 = load <8 x float>, <8 x float>* %out_2_1, align 32
    %out_2_3 = select <8 x i1> %mask, <8 x float> %f47, <8 x float> %out_2_2
    store <8 x float> %out_2_3, <8 x float>* %out_2_1, align 32
    %out_3_0 = getelementptr inbounds i8, i8* %params, i64 384
    %out_3_1 = bitcast i8* %out_3_0 to <8 x float>*
    %out_3_2 = load <8 x float>, <8 x float>* %out_3_1, align 32
    %out_3_3 = select <8 x i1> %mask, <8 x float> %f48, <8 x float> %out_3_2
    store <8 x float> %out_3_3, <8 x float>* %out_3_1, align 32
    %out_4_0 = getelementptr inbounds i8, i8* %params, i64 416
    %out_4_1 = bitcast i8* %out_4_0 to <8 x float>*
    %out_4_2 = load <8 x float>, <8 x float>* %out_4_1, align 32
    %out_4_3 = select <8 x i1> %mask, <8 x float> %f49, <8 x float> %out_4_2
    store <8 x float> %out_4_3, <8 x float>* %out_4_1, align 32
    %out_5_0 = getelementptr inbounds i8, i8* %params, i64 448
    %out_5_1 = bitcast i8* %out_5_0 to <8 x float>*
    %out_5_2 = load <8 x float>, <8 x float>* %out_5_1, align 32
    %out_5_3 = select <8 x i1> %mask, <8 x float> %f44, <8 x float> %out_5_2
    store <8 x float> %out_5_3, <8 x float>* %out_5_1, align 32
    %out_6_0 = getelementptr inbounds i8, i8* %params, i64 480
    %out_6_1 = bitcast i8* %out_6_0 to <8 x float>*
    %out_6_2 = load <8 x float>, <8 x float>* %out_6_1, align 32
    %out_6_3 = select <8 x i1> %mask, <8 x float> %f45, <8 x float> %out_6_2
    store <8 x float> %out_6_3, <8 x float>* %out_6_1, align 32
    %out_7_0 = getelementptr inbounds i8, i8* %params, i64 512
    %out_7_1 = bitcast i8* %out_7_0 to <8 x float>*
    %out_7_2 = load <8 x float>, <8 x float>* %out_7_1, align 32
    %out_7_3 = select <8 x i1> %mask, <8 x float> %f46, <8 x float> %out_7_2
    store <8 x float> %out_7_3, <8 x float>* %out_7_1, align 32
    %out_8_0 = getelementptr inbounds i8, i8* %params, i64 544
    %out_8_1 = bitcast i8* %out_8_0 to <8 x i32>*
    %out_8_2 = load <8 x i32>, <8 x i32>* %out_8_1, align 32
    %out_8_3 = select <8 x i1> %mask, <8 x i32> %r35, <8 x i32> %out_8_2
    store <8 x i32> %out_8_3, <8 x i32>* %out_8_1, align 32
    %out_13_0 = getelementptr inbounds i8, i8* %params, i64 576
    %out_13_1 = bitcast i8* %out_13_0 to <8 x float>*
    %out_13_2 = load <8 x float>, <8 x float>* %out_13_1, align 32
    %out_13_3 = select <8 x i1> %mask, <8 x float> %f68, <8 x float> %out_13_2
    store <8 x float> %out_13_3, <8 x float>* %out_13_1, align 32
    %out_14_0 = getelementptr inbounds i8, i8* %params, i64 608
    %out_14_1 = bitcast i8* %out_14_0 to <8 x float>*
    %out_14_2 = load <8 x float>, <8 x float>* %out_14_1, align 32
    %out_14_3 = select <8 x i1> %mask, <8 x float> %f66, <8 x float> %out_14_2
    store <8 x float> %out_14_3, <8 x float>* %out_14_1, align 32
    %out_15_0 = getelementptr inbounds i8, i8* %params, i64 640
    %out_15_1 = bitcast i8* %out_15_0 to <8 x float>*
    %out_15_2 = load <8 x float>, <8 x float>* %out_15_1, align 32
    %out_15_3 = select <8 x i1> %mask, <8 x float> %f46, <8 x float> %out_15_2
    store <8 x float> %out_15_3, <8 x float>* %out_15_1, align 32
    %out_16_0 = getelementptr inbounds i8, i8* %params, i64 672
    %out_16_1 = bitcast i8* %out_16_0 to <8 x float>*
    %out_16_2 = load <8 x float>, <8 x float>* %out_16_1, align 32
    %out_16_3 = select <8 x i1> %mask, <8 x float> %f45, <8 x float> %out_16_2
    store <8 x float> %out_16_3, <8 x float>* %out_16_1, align 32
    %out_17_0 = getelementptr inbounds i8, i8* %params, i64 704
    %out_17_1 = bitcast i8* %out_17_0 to <8 x float>*
    %out_17_2 = load <8 x float>, <8 x float>* %out_17_1, align 32
    %out_17_3 = select <8 x i1> %mask, <8 x float> %f44, <8 x float> %out_17_2
    store <8 x float> %out_17_3, <8 x float>* %out_17_1, align 32
    ret void;
}

define void @func_dcfca7b5b523c613a93412a4c52b1f7a(<8 x i1> %mask, <8 x i32> %self, i8* noalias %params, i8* noalias %data, <8 x i32> %offsets) #0 {
entry:
    ; VCall: mitsuba::Shape::compute_surface_interaction()
    %f1_i0 = getelementptr inbounds i8, i8* %params, i64 224
    %f1_i1 = bitcast i8* %f1_i0 to <8 x float>*
    %f1 = load <8 x float>, <8 x float>* %f1_i1, align 32
    %f2_i0 = getelementptr inbounds i8, i8* %params, i64 256
    %f2_i1 = bitcast i8* %f2_i0 to <8 x float>*
    %f2 = load <8 x float>, <8 x float>* %f2_i1, align 32
    %r3_i0 = getelementptr inbounds i8, i8* %params, i64 288
    %r3_i1 = bitcast i8* %r3_i0 to <8 x i32>*
    %r3 = load <8 x i32>, <8 x i32>* %r3_i1, align 32
    %f4_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f4 = shufflevector <8 x float> %f4_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f5_1 = insertelement <8 x float> undef, float 0x3ff0000000000000, i32 0
    %f5 = shufflevector <8 x float> %f5_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f6_1 = insertelement <8 x float> undef, float 0xbff0000000000000, i32 0
    %f6 = shufflevector <8 x float> %f6_1, <8 x float> undef, <8 x i32> zeroinitializer
    %p7 = bitcast <8 x i1> %mask to <8 x i1>
    %r8_1 = insertelement <8 x i32> undef, i32 3, i32 0
    %r8 = shufflevector <8 x i32> %r8_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r9 = mul <8 x i32> %r3, %r8
    %r10_1 = insertelement <8 x i32> undef, i32 1, i32 0
    %r10 = shufflevector <8 x i32> %r10_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r11 = add <8 x i32> %r9, %r10
    %r12_1 = insertelement <8 x i32> undef, i32 2, i32 0
    %r12 = shufflevector <8 x i32> %r12_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r13 = add <8 x i32> %r9, %r12
    %f14 = fsub <8 x float> %f5, %f1
    %f15 = fsub <8 x float> %f14, %f2
    %r16 = bitcast <8 x i32> %self to <8 x i32>
    %rd17_p1 = getelementptr inbounds i8, i8* %data, i32 8
    %rd17_p2 = getelementptr inbounds i8, i8* %rd17_p1, <8 x i32> %offsets
    %rd17_p3 = bitcast <8 x i8*> %rd17_p2 to <8 x i64*>
    %rd17_p4 = call <8 x i64> @llvm.masked.gather.v8i64(<8 x i64*> %rd17_p3, i32 8, <8 x i1> %mask, <8 x i64> zeroinitializer)
    %rd17 = inttoptr <8 x i64> %rd17_p4 to <8 x i8*>
    %r18_0 = bitcast <8 x i8*> %rd17 to <8 x i32*>
    %r18_1 = getelementptr i32, <8 x i32*> %r18_0, <8 x i32> %r9
    %r18 = call <8 x i32> @llvm.masked.gather.v8i32(<8 x i32*> %r18_1, i32 4, <8 x i1> %p7, <8 x i32> zeroinitializer)
    %r19_0 = bitcast <8 x i8*> %rd17 to <8 x i32*>
    %r19_1 = getelementptr i32, <8 x i32*> %r19_0, <8 x i32> %r11
    %r19 = call <8 x i32> @llvm.masked.gather.v8i32(<8 x i32*> %r19_1, i32 4, <8 x i1> %p7, <8 x i32> zeroinitializer)
    %r20_0 = bitcast <8 x i8*> %rd17 to <8 x i32*>
    %r20_1 = getelementptr i32, <8 x i32*> %r20_0, <8 x i32> %r13
    %r20 = call <8 x i32> @llvm.masked.gather.v8i32(<8 x i32*> %r20_1, i32 4, <8 x i1> %p7, <8 x i32> zeroinitializer)
    %r21 = mul <8 x i32> %r18, %r8
    %r22 = add <8 x i32> %r21, %r10
    %r23 = add <8 x i32> %r21, %r12
    %rd24_p1 = getelementptr inbounds i8, i8* %data, i32 0
    %rd24_p2 = getelementptr inbounds i8, i8* %rd24_p1, <8 x i32> %offsets
    %rd24_p3 = bitcast <8 x i8*> %rd24_p2 to <8 x i64*>
    %rd24_p4 = call <8 x i64> @llvm.masked.gather.v8i64(<8 x i64*> %rd24_p3, i32 8, <8 x i1> %mask, <8 x i64> zeroinitializer)
    %rd24 = inttoptr <8 x i64> %rd24_p4 to <8 x i8*>
    %f25_0 = bitcast <8 x i8*> %rd24 to <8 x float*>
    %f25_1 = getelementptr float, <8 x float*> %f25_0, <8 x i32> %r21
    %f25 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f25_1, i32 4, <8 x i1> %p7, <8 x float> zeroinitializer)
    %f26_0 = bitcast <8 x i8*> %rd24 to <8 x float*>
    %f26_1 = getelementptr float, <8 x float*> %f26_0, <8 x i32> %r22
    %f26 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f26_1, i32 4, <8 x i1> %p7, <8 x float> zeroinitializer)
    %f27_0 = bitcast <8 x i8*> %rd24 to <8 x float*>
    %f27_1 = getelementptr float, <8 x float*> %f27_0, <8 x i32> %r23
    %f27 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f27_1, i32 4, <8 x i1> %p7, <8 x float> zeroinitializer)
    %r28 = mul <8 x i32> %r19, %r8
    %r29 = add <8 x i32> %r28, %r10
    %r30 = add <8 x i32> %r28, %r12
    %f31_0 = bitcast <8 x i8*> %rd24 to <8 x float*>
    %f31_1 = getelementptr float, <8 x float*> %f31_0, <8 x i32> %r28
    %f31 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f31_1, i32 4, <8 x i1> %p7, <8 x float> zeroinitializer)
    %f32_0 = bitcast <8 x i8*> %rd24 to <8 x float*>
    %f32_1 = getelementptr float, <8 x float*> %f32_0, <8 x i32> %r29
    %f32 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f32_1, i32 4, <8 x i1> %p7, <8 x float> zeroinitializer)
    %f33_0 = bitcast <8 x i8*> %rd24 to <8 x float*>
    %f33_1 = getelementptr float, <8 x float*> %f33_0, <8 x i32> %r30
    %f33 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f33_1, i32 4, <8 x i1> %p7, <8 x float> zeroinitializer)
    %r34 = mul <8 x i32> %r20, %r8
    %r35 = add <8 x i32> %r34, %r10
    %r36 = add <8 x i32> %r34, %r12
    %f37_0 = bitcast <8 x i8*> %rd24 to <8 x float*>
    %f37_1 = getelementptr float, <8 x float*> %f37_0, <8 x i32> %r34
    %f37 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f37_1, i32 4, <8 x i1> %p7, <8 x float> zeroinitializer)
    %f38_0 = bitcast <8 x i8*> %rd24 to <8 x float*>
    %f38_1 = getelementptr float, <8 x float*> %f38_0, <8 x i32> %r35
    %f38 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f38_1, i32 4, <8 x i1> %p7, <8 x float> zeroinitializer)
    %f39_0 = bitcast <8 x i8*> %rd24 to <8 x float*>
    %f39_1 = getelementptr float, <8 x float*> %f39_0, <8 x i32> %r36
    %f39 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f39_1, i32 4, <8 x i1> %p7, <8 x float> zeroinitializer)
    %f40 = fsub <8 x float> %f31, %f25
    %f41 = fsub <8 x float> %f32, %f26
    %f42 = fsub <8 x float> %f33, %f27
    %f43 = fsub <8 x float> %f37, %f25
    %f44 = fsub <8 x float> %f38, %f26
    %f45 = fsub <8 x float> %f39, %f27
    %f46 = fmul <8 x float> %f37, %f2
    %f47 = fmul <8 x float> %f38, %f2
    %f48 = fmul <8 x float> %f39, %f2
    %f49 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f31, <8 x float> %f1, <8 x float> %f46)
    %f50 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f32, <8 x float> %f1, <8 x float> %f47)
    %f51 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f33, <8 x float> %f1, <8 x float> %f48)
    %f52 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f25, <8 x float> %f15, <8 x float> %f49)
    %f53 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f26, <8 x float> %f15, <8 x float> %f50)
    %f54 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f27, <8 x float> %f15, <8 x float> %f51)
    %f55 = fmul <8 x float> %f42, %f44
    %f56 = fmul <8 x float> %f40, %f45
    %f57 = fmul <8 x float> %f41, %f43
    %f58 = fneg <8 x float> %f55
    %f59 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f41, <8 x float> %f45, <8 x float> %f58)
    %f60 = fneg <8 x float> %f56
    %f61 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f42, <8 x float> %f43, <8 x float> %f60)
    %f62 = fneg <8 x float> %f57
    %f63 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f40, <8 x float> %f44, <8 x float> %f62)
    %f64 = fmul <8 x float> %f59, %f59
    %f65 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f61, <8 x float> %f61, <8 x float> %f64)
    %f66 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f63, <8 x float> %f63, <8 x float> %f65)
    %f67 = fdiv <8 x float> %f5, %f66
    %f68 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> %f67)
    %f69 = fmul <8 x float> %f59, %f68
    %f70 = fmul <8 x float> %f61, %f68
    %f71 = fmul <8 x float> %f63, %f68
    %p72 = fcmp oge <8 x float> %f71, %f4
    %f73 = select <8 x i1> %p72, <8 x float> %f5, <8 x float> %f6
    %f74 = fadd <8 x float> %f73, %f71
    %f75 = fdiv <8 x float> %f5, %f74
    %f76 = fneg <8 x float> %f75
    %f77 = fmul <8 x float> %f69, %f70
    %f78 = fmul <8 x float> %f77, %f76
    %f79 = fmul <8 x float> %f69, %f69
    %f80 = fmul <8 x float> %f79, %f76
    %f81 = fneg <8 x float> %f80
    %f82 = select <8 x i1> %p72, <8 x float> %f80, <8 x float> %f81
    %f83 = fadd <8 x float> %f82, %f5
    %f84 = fneg <8 x float> %f78
    %f85 = select <8 x i1> %p72, <8 x float> %f78, <8 x float> %f84
    %f86 = fneg <8 x float> %f69
    %f87 = select <8 x i1> %p72, <8 x float> %f86, <8 x float> %f69
    %rd88_p1 = getelementptr inbounds i8, i8* %data, i32 16
    %rd88_p2 = getelementptr inbounds i8, i8* %rd88_p1, <8 x i32> %offsets
    %rd88_p3 = bitcast <8 x i8*> %rd88_p2 to <8 x i64*>
    %rd88_p4 = call <8 x i64> @llvm.masked.gather.v8i64(<8 x i64*> %rd88_p3, i32 8, <8 x i1> %mask, <8 x i64> zeroinitializer)
    %rd88 = inttoptr <8 x i64> %rd88_p4 to <8 x i8*>
    %f89_0 = bitcast <8 x i8*> %rd88 to <8 x float*>
    %f89_1 = getelementptr float, <8 x float*> %f89_0, <8 x i32> %r21
    %f89 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f89_1, i32 4, <8 x i1> %p7, <8 x float> zeroinitializer)
    %f90_0 = bitcast <8 x i8*> %rd88 to <8 x float*>
    %f90_1 = getelementptr float, <8 x float*> %f90_0, <8 x i32> %r22
    %f90 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f90_1, i32 4, <8 x i1> %p7, <8 x float> zeroinitializer)
    %f91_0 = bitcast <8 x i8*> %rd88 to <8 x float*>
    %f91_1 = getelementptr float, <8 x float*> %f91_0, <8 x i32> %r23
    %f91 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f91_1, i32 4, <8 x i1> %p7, <8 x float> zeroinitializer)
    %f92_0 = bitcast <8 x i8*> %rd88 to <8 x float*>
    %f92_1 = getelementptr float, <8 x float*> %f92_0, <8 x i32> %r28
    %f92 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f92_1, i32 4, <8 x i1> %p7, <8 x float> zeroinitializer)
    %f93_0 = bitcast <8 x i8*> %rd88 to <8 x float*>
    %f93_1 = getelementptr float, <8 x float*> %f93_0, <8 x i32> %r29
    %f93 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f93_1, i32 4, <8 x i1> %p7, <8 x float> zeroinitializer)
    %f94_0 = bitcast <8 x i8*> %rd88 to <8 x float*>
    %f94_1 = getelementptr float, <8 x float*> %f94_0, <8 x i32> %r30
    %f94 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f94_1, i32 4, <8 x i1> %p7, <8 x float> zeroinitializer)
    %f95_0 = bitcast <8 x i8*> %rd88 to <8 x float*>
    %f95_1 = getelementptr float, <8 x float*> %f95_0, <8 x i32> %r34
    %f95 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f95_1, i32 4, <8 x i1> %p7, <8 x float> zeroinitializer)
    %f96_0 = bitcast <8 x i8*> %rd88 to <8 x float*>
    %f96_1 = getelementptr float, <8 x float*> %f96_0, <8 x i32> %r35
    %f96 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f96_1, i32 4, <8 x i1> %p7, <8 x float> zeroinitializer)
    %f97_0 = bitcast <8 x i8*> %rd88 to <8 x float*>
    %f97_1 = getelementptr float, <8 x float*> %f97_0, <8 x i32> %r36
    %f97 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f97_1, i32 4, <8 x i1> %p7, <8 x float> zeroinitializer)
    %f98 = fmul <8 x float> %f89, %f15
    %f99 = fmul <8 x float> %f90, %f15
    %f100 = fmul <8 x float> %f91, %f15
    %f101 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f92, <8 x float> %f1, <8 x float> %f98)
    %f102 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f93, <8 x float> %f1, <8 x float> %f99)
    %f103 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f94, <8 x float> %f1, <8 x float> %f100)
    %f104 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f95, <8 x float> %f2, <8 x float> %f101)
    %f105 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f96, <8 x float> %f2, <8 x float> %f102)
    %f106 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f97, <8 x float> %f2, <8 x float> %f103)
    %f107 = fmul <8 x float> %f104, %f104
    %f108 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f105, <8 x float> %f105, <8 x float> %f107)
    %f109 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f106, <8 x float> %f106, <8 x float> %f108)
    %f110 = fdiv <8 x float> %f5, %f109
    %f111 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> %f110)
    %f112 = fmul <8 x float> %f104, %f111
    %f113 = fmul <8 x float> %f105, %f111
    %f114 = fmul <8 x float> %f106, %f111
    %out_0_0 = getelementptr inbounds i8, i8* %params, i64 320
    %out_0_1 = bitcast i8* %out_0_0 to <8 x float>*
    %out_0_2 = load <8 x float>, <8 x float>* %out_0_1, align 32
    %out_0_3 = select <8 x i1> %mask, <8 x float> %f85, <8 x float> %out_0_2
    store <8 x float> %out_0_3, <8 x float>* %out_0_1, align 32
    %out_2_0 = getelementptr inbounds i8, i8* %params, i64 352
    %out_2_1 = bitcast i8* %out_2_0 to <8 x float>*
    %out_2_2 = load <8 x float>, <8 x float>* %out_2_1, align 32
    %out_2_3 = select <8 x i1> %mask, <8 x float> %f52, <8 x float> %out_2_2
    store <8 x float> %out_2_3, <8 x float>* %out_2_1, align 32
    %out_3_0 = getelementptr inbounds i8, i8* %params, i64 384
    %out_3_1 = bitcast i8* %out_3_0 to <8 x float>*
    %out_3_2 = load <8 x float>, <8 x float>* %out_3_1, align 32
    %out_3_3 = select <8 x i1> %mask, <8 x float> %f53, <8 x float> %out_3_2
    store <8 x float> %out_3_3, <8 x float>* %out_3_1, align 32
    %out_4_0 = getelementptr inbounds i8, i8* %params, i64 416
    %out_4_1 = bitcast i8* %out_4_0 to <8 x float>*
    %out_4_2 = load <8 x float>, <8 x float>* %out_4_1, align 32
    %out_4_3 = select <8 x i1> %mask, <8 x float> %f54, <8 x float> %out_4_2
    store <8 x float> %out_4_3, <8 x float>* %out_4_1, align 32
    %out_5_0 = getelementptr inbounds i8, i8* %params, i64 448
    %out_5_1 = bitcast i8* %out_5_0 to <8 x float>*
    %out_5_2 = load <8 x float>, <8 x float>* %out_5_1, align 32
    %out_5_3 = select <8 x i1> %mask, <8 x float> %f69, <8 x float> %out_5_2
    store <8 x float> %out_5_3, <8 x float>* %out_5_1, align 32
    %out_6_0 = getelementptr inbounds i8, i8* %params, i64 480
    %out_6_1 = bitcast i8* %out_6_0 to <8 x float>*
    %out_6_2 = load <8 x float>, <8 x float>* %out_6_1, align 32
    %out_6_3 = select <8 x i1> %mask, <8 x float> %f70, <8 x float> %out_6_2
    store <8 x float> %out_6_3, <8 x float>* %out_6_1, align 32
    %out_7_0 = getelementptr inbounds i8, i8* %params, i64 512
    %out_7_1 = bitcast i8* %out_7_0 to <8 x float>*
    %out_7_2 = load <8 x float>, <8 x float>* %out_7_1, align 32
    %out_7_3 = select <8 x i1> %mask, <8 x float> %f71, <8 x float> %out_7_2
    store <8 x float> %out_7_3, <8 x float>* %out_7_1, align 32
    %out_8_0 = getelementptr inbounds i8, i8* %params, i64 544
    %out_8_1 = bitcast i8* %out_8_0 to <8 x i32>*
    %out_8_2 = load <8 x i32>, <8 x i32>* %out_8_1, align 32
    %out_8_3 = select <8 x i1> %mask, <8 x i32> %r16, <8 x i32> %out_8_2
    store <8 x i32> %out_8_3, <8 x i32>* %out_8_1, align 32
    %out_13_0 = getelementptr inbounds i8, i8* %params, i64 576
    %out_13_1 = bitcast i8* %out_13_0 to <8 x float>*
    %out_13_2 = load <8 x float>, <8 x float>* %out_13_1, align 32
    %out_13_3 = select <8 x i1> %mask, <8 x float> %f87, <8 x float> %out_13_2
    store <8 x float> %out_13_3, <8 x float>* %out_13_1, align 32
    %out_14_0 = getelementptr inbounds i8, i8* %params, i64 608
    %out_14_1 = bitcast i8* %out_14_0 to <8 x float>*
    %out_14_2 = load <8 x float>, <8 x float>* %out_14_1, align 32
    %out_14_3 = select <8 x i1> %mask, <8 x float> %f83, <8 x float> %out_14_2
    store <8 x float> %out_14_3, <8 x float>* %out_14_1, align 32
    %out_15_0 = getelementptr inbounds i8, i8* %params, i64 640
    %out_15_1 = bitcast i8* %out_15_0 to <8 x float>*
    %out_15_2 = load <8 x float>, <8 x float>* %out_15_1, align 32
    %out_15_3 = select <8 x i1> %mask, <8 x float> %f114, <8 x float> %out_15_2
    store <8 x float> %out_15_3, <8 x float>* %out_15_1, align 32
    %out_16_0 = getelementptr inbounds i8, i8* %params, i64 672
    %out_16_1 = bitcast i8* %out_16_0 to <8 x float>*
    %out_16_2 = load <8 x float>, <8 x float>* %out_16_1, align 32
    %out_16_3 = select <8 x i1> %mask, <8 x float> %f113, <8 x float> %out_16_2
    store <8 x float> %out_16_3, <8 x float>* %out_16_1, align 32
    %out_17_0 = getelementptr inbounds i8, i8* %params, i64 704
    %out_17_1 = bitcast i8* %out_17_0 to <8 x float>*
    %out_17_2 = load <8 x float>, <8 x float>* %out_17_1, align 32
    %out_17_3 = select <8 x i1> %mask, <8 x float> %f112, <8 x float> %out_17_2
    store <8 x float> %out_17_3, <8 x float>* %out_17_1, align 32
    ret void;
}

define void @func_bf888b7a279d5208759c322da4df8aa9(<8 x i1> %mask, i8* noalias %params, i8* noalias %data, <8 x i32> %offsets) #0 {
entry:
    ; VCall: mitsuba::Emitter::pdf_direction()
    %f1_p1 = getelementptr inbounds i8, i8* %data, i32 0
    %f1_p2 = getelementptr inbounds i8, i8* %f1_p1, <8 x i32> %offsets
    %f1_p3 = bitcast <8 x i8*> %f1_p2 to <8 x float*>
    %f1 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f1_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f2_i0 = getelementptr inbounds i8, i8* %params, i64 0
    %f2_i1 = bitcast i8* %f2_i0 to <8 x float>*
    %f2 = load <8 x float>, <8 x float>* %f2_i1, align 32
    %f3_i0 = getelementptr inbounds i8, i8* %params, i64 32
    %f3_i1 = bitcast i8* %f3_i0 to <8 x float>*
    %f3 = load <8 x float>, <8 x float>* %f3_i1, align 32
    %f4_i0 = getelementptr inbounds i8, i8* %params, i64 64
    %f4_i1 = bitcast i8* %f4_i0 to <8 x float>*
    %f4 = load <8 x float>, <8 x float>* %f4_i1, align 32
    %f5_i0 = getelementptr inbounds i8, i8* %params, i64 96
    %f5_i1 = bitcast i8* %f5_i0 to <8 x float>*
    %f5 = load <8 x float>, <8 x float>* %f5_i1, align 32
    %f6_i0 = getelementptr inbounds i8, i8* %params, i64 128
    %f6_i1 = bitcast i8* %f6_i0 to <8 x float>*
    %f6 = load <8 x float>, <8 x float>* %f6_i1, align 32
    %f7_i0 = getelementptr inbounds i8, i8* %params, i64 160
    %f7_i1 = bitcast i8* %f7_i0 to <8 x float>*
    %f7 = load <8 x float>, <8 x float>* %f7_i1, align 32
    %f8_i0 = getelementptr inbounds i8, i8* %params, i64 192
    %f8_i1 = bitcast i8* %f8_i0 to <8 x float>*
    %f8 = load <8 x float>, <8 x float>* %f8_i1, align 32
    %f9 = fmul <8 x float> %f5, %f2
    %f10 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f6, <8 x float> %f3, <8 x float> %f9)
    %f11 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f7, <8 x float> %f4, <8 x float> %f10)
    %f12_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f12 = shufflevector <8 x float> %f12_1, <8 x float> undef, <8 x i32> zeroinitializer
    %p13 = fcmp olt <8 x float> %f11, %f12
    %f14_1 = insertelement <8 x float> undef, float 0x7fffffffe0000000, i32 0
    %f14 = shufflevector <8 x float> %f14_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f15_0 = bitcast <8 x float> %f11 to <8 x i32>
    %f15_1 = bitcast <8 x float> %f14 to <8 x i32>
    %f15_2 = and <8 x i32> %f15_0, %f15_1
    %f15 = bitcast <8 x i32> %f15_2 to <8 x float>
    %p16 = fcmp one <8 x float> %f15, %f12
    %f17 = fmul <8 x float> %f8, %f8
    %f18 = fdiv <8 x float> %f17, %f15
    %f19 = select <8 x i1> %p16, <8 x float> %f18, <8 x float> zeroinitializer
    %f20 = fmul <8 x float> %f1, %f19
    %f21 = select <8 x i1> %p13, <8 x float> %f20, <8 x float> zeroinitializer
    %out_0_0 = getelementptr inbounds i8, i8* %params, i64 224
    %out_0_1 = bitcast i8* %out_0_0 to <8 x float>*
    %out_0_2 = load <8 x float>, <8 x float>* %out_0_1, align 32
    %out_0_3 = select <8 x i1> %mask, <8 x float> %f21, <8 x float> %out_0_2
    store <8 x float> %out_0_3, <8 x float>* %out_0_1, align 32
    ret void;
}

define void @func_972f6c492a7ec6284cf60b520b91aba3(<8 x i1> %mask, i8* noalias %params, i8* noalias %data, <8 x i32> %offsets) #0 {
entry:
    ; VCall: mitsuba::Emitter::eval()
    %f1_p1 = getelementptr inbounds i8, i8* %data, i32 0
    %f1_p2 = getelementptr inbounds i8, i8* %f1_p1, <8 x i32> %offsets
    %f1_p3 = bitcast <8 x i8*> %f1_p2 to <8 x float*>
    %f1 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f1_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f2_p1 = getelementptr inbounds i8, i8* %data, i32 4
    %f2_p2 = getelementptr inbounds i8, i8* %f2_p1, <8 x i32> %offsets
    %f2_p3 = bitcast <8 x i8*> %f2_p2 to <8 x float*>
    %f2 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f2_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f3_p1 = getelementptr inbounds i8, i8* %data, i32 8
    %f3_p2 = getelementptr inbounds i8, i8* %f3_p1, <8 x i32> %offsets
    %f3_p3 = bitcast <8 x i8*> %f3_p2 to <8 x float*>
    %f3 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f3_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f4_i0 = getelementptr inbounds i8, i8* %params, i64 96
    %f4_i1 = bitcast i8* %f4_i0 to <8 x float>*
    %f4 = load <8 x float>, <8 x float>* %f4_i1, align 32
    %f5_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f5 = shufflevector <8 x float> %f5_1, <8 x float> undef, <8 x i32> zeroinitializer
    %p6 = fcmp ogt <8 x float> %f4, %f5
    %f7 = select <8 x i1> %p6, <8 x float> %f1, <8 x float> zeroinitializer
    %f8 = select <8 x i1> %p6, <8 x float> %f2, <8 x float> zeroinitializer
    %f9 = select <8 x i1> %p6, <8 x float> %f3, <8 x float> zeroinitializer
    %out_0_0 = getelementptr inbounds i8, i8* %params, i64 160
    %out_0_1 = bitcast i8* %out_0_0 to <8 x float>*
    %out_0_2 = load <8 x float>, <8 x float>* %out_0_1, align 32
    %out_0_3 = select <8 x i1> %mask, <8 x float> %f7, <8 x float> %out_0_2
    store <8 x float> %out_0_3, <8 x float>* %out_0_1, align 32
    %out_1_0 = getelementptr inbounds i8, i8* %params, i64 192
    %out_1_1 = bitcast i8* %out_1_0 to <8 x float>*
    %out_1_2 = load <8 x float>, <8 x float>* %out_1_1, align 32
    %out_1_3 = select <8 x i1> %mask, <8 x float> %f8, <8 x float> %out_1_2
    store <8 x float> %out_1_3, <8 x float>* %out_1_1, align 32
    %out_2_0 = getelementptr inbounds i8, i8* %params, i64 224
    %out_2_1 = bitcast i8* %out_2_0 to <8 x float>*
    %out_2_2 = load <8 x float>, <8 x float>* %out_2_1, align 32
    %out_2_3 = select <8 x i1> %mask, <8 x float> %f9, <8 x float> %out_2_2
    store <8 x float> %out_2_3, <8 x float>* %out_2_1, align 32
    ret void;
}

define void @func_19ee132d0bd6c9e577b7250c22c38555(<8 x i1> %mask, <8 x i32> %self, i8* noalias %params, i8* noalias %data, <8 x i32> %offsets) #0 {
entry:
    ; VCall: mitsuba::Emitter::sample_direction()
    %f1_p1 = getelementptr inbounds i8, i8* %data, i32 48
    %f1_p2 = getelementptr inbounds i8, i8* %f1_p1, <8 x i32> %offsets
    %f1_p3 = bitcast <8 x i8*> %f1_p2 to <8 x float*>
    %f1 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f1_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f2_p1 = getelementptr inbounds i8, i8* %data, i32 52
    %f2_p2 = getelementptr inbounds i8, i8* %f2_p1, <8 x i32> %offsets
    %f2_p3 = bitcast <8 x i8*> %f2_p2 to <8 x float*>
    %f2 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f2_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f3_p1 = getelementptr inbounds i8, i8* %data, i32 56
    %f3_p2 = getelementptr inbounds i8, i8* %f3_p1, <8 x i32> %offsets
    %f3_p3 = bitcast <8 x i8*> %f3_p2 to <8 x float*>
    %f3 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f3_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f4_p1 = getelementptr inbounds i8, i8* %data, i32 24
    %f4_p2 = getelementptr inbounds i8, i8* %f4_p1, <8 x i32> %offsets
    %f4_p3 = bitcast <8 x i8*> %f4_p2 to <8 x float*>
    %f4 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f4_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f5_p1 = getelementptr inbounds i8, i8* %data, i32 28
    %f5_p2 = getelementptr inbounds i8, i8* %f5_p1, <8 x i32> %offsets
    %f5_p3 = bitcast <8 x i8*> %f5_p2 to <8 x float*>
    %f5 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f5_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f6_i0 = getelementptr inbounds i8, i8* %params, i64 96
    %f6_i1 = bitcast i8* %f6_i0 to <8 x float>*
    %f6 = load <8 x float>, <8 x float>* %f6_i1, align 32
    %f7_i0 = getelementptr inbounds i8, i8* %params, i64 0
    %f7_i1 = bitcast i8* %f7_i0 to <8 x float>*
    %f7 = load <8 x float>, <8 x float>* %f7_i1, align 32
    %f8_i0 = getelementptr inbounds i8, i8* %params, i64 32
    %f8_i1 = bitcast i8* %f8_i0 to <8 x float>*
    %f8 = load <8 x float>, <8 x float>* %f8_i1, align 32
    %f9_i0 = getelementptr inbounds i8, i8* %params, i64 64
    %f9_i1 = bitcast i8* %f9_i0 to <8 x float>*
    %f9 = load <8 x float>, <8 x float>* %f9_i1, align 32
    %f10_i0 = getelementptr inbounds i8, i8* %params, i64 128
    %f10_i1 = bitcast i8* %f10_i0 to <8 x float>*
    %f10 = load <8 x float>, <8 x float>* %f10_i1, align 32
    %p11 = bitcast <8 x i1> %mask to <8 x i1>
    %r12_1 = insertelement <8 x i32> undef, i32 0, i32 0
    %r12 = shufflevector <8 x i32> %r12_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %f13_1 = insertelement <8 x float> undef, float 0x7ff0000000000000, i32 0
    %f13 = shufflevector <8 x float> %f13_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f14 = fmul <8 x float> %f10, %f4
    %r15_1 = insertelement <8 x i32> undef, i32 1, i32 0
    %r15 = shufflevector <8 x i32> %r15_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %rd16_p1 = getelementptr inbounds i8, i8* %data, i32 16
    %rd16_p2 = getelementptr inbounds i8, i8* %rd16_p1, <8 x i32> %offsets
    %rd16_p3 = bitcast <8 x i8*> %rd16_p2 to <8 x i64*>
    %rd16_p4 = call <8 x i64> @llvm.masked.gather.v8i64(<8 x i64*> %rd16_p3, i32 8, <8 x i1> %mask, <8 x i64> zeroinitializer)
    %rd16 = inttoptr <8 x i64> %rd16_p4 to <8 x i8*>
    %f17_0 = bitcast <8 x i8*> %rd16 to <8 x float*>
    %f17_1 = getelementptr float, <8 x float*> %f17_0, <8 x i32> %r12
    %f17 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f17_1, i32 4, <8 x i1> %p11, <8 x float> zeroinitializer)
    %p18 = fcmp olt <8 x float> %f17, %f14
    %r19 = select <8 x i1> %p18, <8 x i32> %r15, <8 x i32> zeroinitializer
    %rd20_p1 = getelementptr inbounds i8, i8* %data, i32 32
    %rd20_p2 = getelementptr inbounds i8, i8* %rd20_p1, <8 x i32> %offsets
    %rd20_p3 = bitcast <8 x i8*> %rd20_p2 to <8 x i64*>
    %rd20_p4 = call <8 x i64> @llvm.masked.gather.v8i64(<8 x i64*> %rd20_p3, i32 8, <8 x i1> %mask, <8 x i64> zeroinitializer)
    %rd20 = inttoptr <8 x i64> %rd20_p4 to <8 x i8*>
    %f21_0 = bitcast <8 x i8*> %rd20 to <8 x float*>
    %f21_1 = getelementptr float, <8 x float*> %f21_0, <8 x i32> %r19
    %f21 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f21_1, i32 4, <8 x i1> %p11, <8 x float> zeroinitializer)
    %f22 = fmul <8 x float> %f21, %f5
    %r23 = sub <8 x i32> %r19, %r15
    %p24 = icmp ugt <8 x i32> %r19, %r12
    %p25 = and <8 x i1> %p24, %p11
    %f26_0 = bitcast <8 x i8*> %rd16 to <8 x float*>
    %f26_1 = getelementptr float, <8 x float*> %f26_0, <8 x i32> %r23
    %f26 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f26_1, i32 4, <8 x i1> %p25, <8 x float> zeroinitializer)
    %f27 = fmul <8 x float> %f26, %f5
    %f28 = fsub <8 x float> %f10, %f27
    %f29 = fdiv <8 x float> %f28, %f22
    %r30_1 = insertelement <8 x i32> undef, i32 3, i32 0
    %r30 = shufflevector <8 x i32> %r30_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r31 = mul <8 x i32> %r19, %r30
    %r32 = add <8 x i32> %r31, %r15
    %r33_1 = insertelement <8 x i32> undef, i32 2, i32 0
    %r33 = shufflevector <8 x i32> %r33_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r34 = add <8 x i32> %r31, %r33
    %rd35_p1 = getelementptr inbounds i8, i8* %data, i32 8
    %rd35_p2 = getelementptr inbounds i8, i8* %rd35_p1, <8 x i32> %offsets
    %rd35_p3 = bitcast <8 x i8*> %rd35_p2 to <8 x i64*>
    %rd35_p4 = call <8 x i64> @llvm.masked.gather.v8i64(<8 x i64*> %rd35_p3, i32 8, <8 x i1> %mask, <8 x i64> zeroinitializer)
    %rd35 = inttoptr <8 x i64> %rd35_p4 to <8 x i8*>
    %r36_0 = bitcast <8 x i8*> %rd35 to <8 x i32*>
    %r36_1 = getelementptr i32, <8 x i32*> %r36_0, <8 x i32> %r31
    %r36 = call <8 x i32> @llvm.masked.gather.v8i32(<8 x i32*> %r36_1, i32 4, <8 x i1> %p11, <8 x i32> zeroinitializer)
    %r37_0 = bitcast <8 x i8*> %rd35 to <8 x i32*>
    %r37_1 = getelementptr i32, <8 x i32*> %r37_0, <8 x i32> %r32
    %r37 = call <8 x i32> @llvm.masked.gather.v8i32(<8 x i32*> %r37_1, i32 4, <8 x i1> %p11, <8 x i32> zeroinitializer)
    %r38_0 = bitcast <8 x i8*> %rd35 to <8 x i32*>
    %r38_1 = getelementptr i32, <8 x i32*> %r38_0, <8 x i32> %r34
    %r38 = call <8 x i32> @llvm.masked.gather.v8i32(<8 x i32*> %r38_1, i32 4, <8 x i1> %p11, <8 x i32> zeroinitializer)
    %r39 = mul <8 x i32> %r36, %r30
    %r40 = add <8 x i32> %r39, %r15
    %r41 = add <8 x i32> %r39, %r33
    %rd42_p1 = getelementptr inbounds i8, i8* %data, i32 0
    %rd42_p2 = getelementptr inbounds i8, i8* %rd42_p1, <8 x i32> %offsets
    %rd42_p3 = bitcast <8 x i8*> %rd42_p2 to <8 x i64*>
    %rd42_p4 = call <8 x i64> @llvm.masked.gather.v8i64(<8 x i64*> %rd42_p3, i32 8, <8 x i1> %mask, <8 x i64> zeroinitializer)
    %rd42 = inttoptr <8 x i64> %rd42_p4 to <8 x i8*>
    %f43_0 = bitcast <8 x i8*> %rd42 to <8 x float*>
    %f43_1 = getelementptr float, <8 x float*> %f43_0, <8 x i32> %r39
    %f43 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f43_1, i32 4, <8 x i1> %p11, <8 x float> zeroinitializer)
    %f44_0 = bitcast <8 x i8*> %rd42 to <8 x float*>
    %f44_1 = getelementptr float, <8 x float*> %f44_0, <8 x i32> %r40
    %f44 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f44_1, i32 4, <8 x i1> %p11, <8 x float> zeroinitializer)
    %f45_0 = bitcast <8 x i8*> %rd42 to <8 x float*>
    %f45_1 = getelementptr float, <8 x float*> %f45_0, <8 x i32> %r41
    %f45 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f45_1, i32 4, <8 x i1> %p11, <8 x float> zeroinitializer)
    %r46 = mul <8 x i32> %r37, %r30
    %r47 = add <8 x i32> %r46, %r15
    %r48 = add <8 x i32> %r46, %r33
    %f49_0 = bitcast <8 x i8*> %rd42 to <8 x float*>
    %f49_1 = getelementptr float, <8 x float*> %f49_0, <8 x i32> %r46
    %f49 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f49_1, i32 4, <8 x i1> %p11, <8 x float> zeroinitializer)
    %f50_0 = bitcast <8 x i8*> %rd42 to <8 x float*>
    %f50_1 = getelementptr float, <8 x float*> %f50_0, <8 x i32> %r47
    %f50 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f50_1, i32 4, <8 x i1> %p11, <8 x float> zeroinitializer)
    %f51_0 = bitcast <8 x i8*> %rd42 to <8 x float*>
    %f51_1 = getelementptr float, <8 x float*> %f51_0, <8 x i32> %r48
    %f51 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f51_1, i32 4, <8 x i1> %p11, <8 x float> zeroinitializer)
    %r52 = mul <8 x i32> %r38, %r30
    %r53 = add <8 x i32> %r52, %r15
    %r54 = add <8 x i32> %r52, %r33
    %f55_0 = bitcast <8 x i8*> %rd42 to <8 x float*>
    %f55_1 = getelementptr float, <8 x float*> %f55_0, <8 x i32> %r52
    %f55 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f55_1, i32 4, <8 x i1> %p11, <8 x float> zeroinitializer)
    %f56_0 = bitcast <8 x i8*> %rd42 to <8 x float*>
    %f56_1 = getelementptr float, <8 x float*> %f56_0, <8 x i32> %r53
    %f56 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f56_1, i32 4, <8 x i1> %p11, <8 x float> zeroinitializer)
    %f57_0 = bitcast <8 x i8*> %rd42 to <8 x float*>
    %f57_1 = getelementptr float, <8 x float*> %f57_0, <8 x i32> %r54
    %f57 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f57_1, i32 4, <8 x i1> %p11, <8 x float> zeroinitializer)
    %f58 = fsub <8 x float> %f49, %f43
    %f59 = fsub <8 x float> %f50, %f44
    %f60 = fsub <8 x float> %f51, %f45
    %f61 = fsub <8 x float> %f55, %f43
    %f62 = fsub <8 x float> %f56, %f44
    %f63 = fsub <8 x float> %f57, %f45
    %f64_1 = insertelement <8 x float> undef, float 0x3ff0000000000000, i32 0
    %f64 = shufflevector <8 x float> %f64_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f65 = fsub <8 x float> %f64, %f6
    %f66_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f66 = shufflevector <8 x float> %f66_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f67 = call <8 x float> @llvm.maxnum.v8f32(<8 x float> %f65, <8 x float> %f66)
    %f68 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> %f67)
    %f69 = fsub <8 x float> %f64, %f68
    %f70 = fmul <8 x float> %f68, %f29
    %f71 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f61, <8 x float> %f70, <8 x float> %f43)
    %f72 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f62, <8 x float> %f70, <8 x float> %f44)
    %f73 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f63, <8 x float> %f70, <8 x float> %f45)
    %f74 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f58, <8 x float> %f69, <8 x float> %f71)
    %f75 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f59, <8 x float> %f69, <8 x float> %f72)
    %f76 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f60, <8 x float> %f69, <8 x float> %f73)
    %rd77_p1 = getelementptr inbounds i8, i8* %data, i32 40
    %rd77_p2 = getelementptr inbounds i8, i8* %rd77_p1, <8 x i32> %offsets
    %rd77_p3 = bitcast <8 x i8*> %rd77_p2 to <8 x i64*>
    %rd77_p4 = call <8 x i64> @llvm.masked.gather.v8i64(<8 x i64*> %rd77_p3, i32 8, <8 x i1> %mask, <8 x i64> zeroinitializer)
    %rd77 = inttoptr <8 x i64> %rd77_p4 to <8 x i8*>
    %f78_0 = bitcast <8 x i8*> %rd77 to <8 x float*>
    %f78_1 = getelementptr float, <8 x float*> %f78_0, <8 x i32> %r39
    %f78 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f78_1, i32 4, <8 x i1> %p11, <8 x float> zeroinitializer)
    %f79_0 = bitcast <8 x i8*> %rd77 to <8 x float*>
    %f79_1 = getelementptr float, <8 x float*> %f79_0, <8 x i32> %r40
    %f79 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f79_1, i32 4, <8 x i1> %p11, <8 x float> zeroinitializer)
    %f80_0 = bitcast <8 x i8*> %rd77 to <8 x float*>
    %f80_1 = getelementptr float, <8 x float*> %f80_0, <8 x i32> %r41
    %f80 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f80_1, i32 4, <8 x i1> %p11, <8 x float> zeroinitializer)
    %f81_0 = bitcast <8 x i8*> %rd77 to <8 x float*>
    %f81_1 = getelementptr float, <8 x float*> %f81_0, <8 x i32> %r46
    %f81 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f81_1, i32 4, <8 x i1> %p11, <8 x float> zeroinitializer)
    %f82_0 = bitcast <8 x i8*> %rd77 to <8 x float*>
    %f82_1 = getelementptr float, <8 x float*> %f82_0, <8 x i32> %r47
    %f82 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f82_1, i32 4, <8 x i1> %p11, <8 x float> zeroinitializer)
    %f83_0 = bitcast <8 x i8*> %rd77 to <8 x float*>
    %f83_1 = getelementptr float, <8 x float*> %f83_0, <8 x i32> %r48
    %f83 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f83_1, i32 4, <8 x i1> %p11, <8 x float> zeroinitializer)
    %f84_0 = bitcast <8 x i8*> %rd77 to <8 x float*>
    %f84_1 = getelementptr float, <8 x float*> %f84_0, <8 x i32> %r52
    %f84 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f84_1, i32 4, <8 x i1> %p11, <8 x float> zeroinitializer)
    %f85_0 = bitcast <8 x i8*> %rd77 to <8 x float*>
    %f85_1 = getelementptr float, <8 x float*> %f85_0, <8 x i32> %r53
    %f85 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f85_1, i32 4, <8 x i1> %p11, <8 x float> zeroinitializer)
    %f86_0 = bitcast <8 x i8*> %rd77 to <8 x float*>
    %f86_1 = getelementptr float, <8 x float*> %f86_0, <8 x i32> %r54
    %f86 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f86_1, i32 4, <8 x i1> %p11, <8 x float> zeroinitializer)
    %f87 = fsub <8 x float> %f64, %f69
    %f88 = fsub <8 x float> %f87, %f70
    %f89 = fmul <8 x float> %f84, %f70
    %f90 = fmul <8 x float> %f85, %f70
    %f91 = fmul <8 x float> %f86, %f70
    %f92 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f81, <8 x float> %f69, <8 x float> %f89)
    %f93 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f82, <8 x float> %f69, <8 x float> %f90)
    %f94 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f83, <8 x float> %f69, <8 x float> %f91)
    %f95 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f78, <8 x float> %f88, <8 x float> %f92)
    %f96 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f79, <8 x float> %f88, <8 x float> %f93)
    %f97 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f80, <8 x float> %f88, <8 x float> %f94)
    %f98 = fmul <8 x float> %f95, %f95
    %f99 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f96, <8 x float> %f96, <8 x float> %f98)
    %f100 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f97, <8 x float> %f97, <8 x float> %f99)
    %f101 = fdiv <8 x float> %f64, %f100
    %f102 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> %f101)
    %f103 = fmul <8 x float> %f95, %f102
    %f104 = fmul <8 x float> %f96, %f102
    %f105 = fmul <8 x float> %f97, %f102
    %f106 = fsub <8 x float> %f74, %f7
    %f107 = fsub <8 x float> %f75, %f8
    %f108 = fsub <8 x float> %f76, %f9
    %f109 = fmul <8 x float> %f106, %f106
    %f110 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f107, <8 x float> %f107, <8 x float> %f109)
    %f111 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f108, <8 x float> %f108, <8 x float> %f110)
    %f112 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> %f111)
    %f113 = fdiv <8 x float> %f64, %f112
    %f114 = fmul <8 x float> %f106, %f113
    %f115 = fmul <8 x float> %f107, %f113
    %f116 = fmul <8 x float> %f108, %f113
    %f117 = fmul <8 x float> %f114, %f103
    %f118 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f115, <8 x float> %f104, <8 x float> %f117)
    %f119 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f116, <8 x float> %f105, <8 x float> %f118)
    %f120_1 = insertelement <8 x float> undef, float 0x7fffffffe0000000, i32 0
    %f120 = shufflevector <8 x float> %f120_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f121_0 = bitcast <8 x float> %f119 to <8 x i32>
    %f121_1 = bitcast <8 x float> %f120 to <8 x i32>
    %f121_2 = and <8 x i32> %f121_0, %f121_1
    %f121 = bitcast <8 x i32> %f121_2 to <8 x float>
    %f122 = fdiv <8 x float> %f111, %f121
    %f123_0 = bitcast <8 x float> %f122 to <8 x i32>
    %f123_1 = bitcast <8 x float> %f120 to <8 x i32>
    %f123_2 = and <8 x i32> %f123_0, %f123_1
    %f123 = bitcast <8 x i32> %f123_2 to <8 x float>
    %p124 = fcmp olt <8 x float> %f123, %f13
    %f125 = select <8 x i1> %p124, <8 x float> %f122, <8 x float> zeroinitializer
    %f126 = fmul <8 x float> %f5, %f125
    %p127 = fcmp olt <8 x float> %f119, %f66
    %p128 = fcmp one <8 x float> %f126, %f66
    %p129 = and <8 x i1> %p127, %p128
    %f130 = fdiv <8 x float> %f64, %f126
    %f131 = fmul <8 x float> %f1, %f130
    %f132 = fmul <8 x float> %f2, %f130
    %f133 = fmul <8 x float> %f3, %f130
    %f134 = select <8 x i1> %p129, <8 x float> %f131, <8 x float> zeroinitializer
    %f135 = select <8 x i1> %p129, <8 x float> %f132, <8 x float> zeroinitializer
    %f136 = select <8 x i1> %p129, <8 x float> %f133, <8 x float> zeroinitializer
    %out_0_0 = getelementptr inbounds i8, i8* %params, i64 160
    %out_0_1 = bitcast i8* %out_0_0 to <8 x float>*
    %out_0_2 = load <8 x float>, <8 x float>* %out_0_1, align 32
    %out_0_3 = select <8 x i1> %mask, <8 x float> %f74, <8 x float> %out_0_2
    store <8 x float> %out_0_3, <8 x float>* %out_0_1, align 32
    %out_1_0 = getelementptr inbounds i8, i8* %params, i64 192
    %out_1_1 = bitcast i8* %out_1_0 to <8 x float>*
    %out_1_2 = load <8 x float>, <8 x float>* %out_1_1, align 32
    %out_1_3 = select <8 x i1> %mask, <8 x float> %f75, <8 x float> %out_1_2
    store <8 x float> %out_1_3, <8 x float>* %out_1_1, align 32
    %out_2_0 = getelementptr inbounds i8, i8* %params, i64 224
    %out_2_1 = bitcast i8* %out_2_0 to <8 x float>*
    %out_2_2 = load <8 x float>, <8 x float>* %out_2_1, align 32
    %out_2_3 = select <8 x i1> %mask, <8 x float> %f76, <8 x float> %out_2_2
    store <8 x float> %out_2_3, <8 x float>* %out_2_1, align 32
    %out_8_0 = getelementptr inbounds i8, i8* %params, i64 256
    %out_8_1 = bitcast i8* %out_8_0 to <8 x float>*
    %out_8_2 = load <8 x float>, <8 x float>* %out_8_1, align 32
    %out_8_3 = select <8 x i1> %mask, <8 x float> %f126, <8 x float> %out_8_2
    store <8 x float> %out_8_3, <8 x float>* %out_8_1, align 32
    %out_9_0 = getelementptr inbounds i8, i8* %params, i64 288
    %out_9_1 = bitcast i8* %out_9_0 to <8 x float>*
    %out_9_2 = load <8 x float>, <8 x float>* %out_9_1, align 32
    %out_9_3 = select <8 x i1> %mask, <8 x float> %f114, <8 x float> %out_9_2
    store <8 x float> %out_9_3, <8 x float>* %out_9_1, align 32
    %out_10_0 = getelementptr inbounds i8, i8* %params, i64 320
    %out_10_1 = bitcast i8* %out_10_0 to <8 x float>*
    %out_10_2 = load <8 x float>, <8 x float>* %out_10_1, align 32
    %out_10_3 = select <8 x i1> %mask, <8 x float> %f115, <8 x float> %out_10_2
    store <8 x float> %out_10_3, <8 x float>* %out_10_1, align 32
    %out_11_0 = getelementptr inbounds i8, i8* %params, i64 352
    %out_11_1 = bitcast i8* %out_11_0 to <8 x float>*
    %out_11_2 = load <8 x float>, <8 x float>* %out_11_1, align 32
    %out_11_3 = select <8 x i1> %mask, <8 x float> %f116, <8 x float> %out_11_2
    store <8 x float> %out_11_3, <8 x float>* %out_11_1, align 32
    %out_14_0 = getelementptr inbounds i8, i8* %params, i64 384
    %out_14_1 = bitcast i8* %out_14_0 to <8 x float>*
    %out_14_2 = load <8 x float>, <8 x float>* %out_14_1, align 32
    %out_14_3 = select <8 x i1> %mask, <8 x float> %f134, <8 x float> %out_14_2
    store <8 x float> %out_14_3, <8 x float>* %out_14_1, align 32
    %out_15_0 = getelementptr inbounds i8, i8* %params, i64 416
    %out_15_1 = bitcast i8* %out_15_0 to <8 x float>*
    %out_15_2 = load <8 x float>, <8 x float>* %out_15_1, align 32
    %out_15_3 = select <8 x i1> %mask, <8 x float> %f135, <8 x float> %out_15_2
    store <8 x float> %out_15_3, <8 x float>* %out_15_1, align 32
    %out_16_0 = getelementptr inbounds i8, i8* %params, i64 448
    %out_16_1 = bitcast i8* %out_16_0 to <8 x float>*
    %out_16_2 = load <8 x float>, <8 x float>* %out_16_1, align 32
    %out_16_3 = select <8 x i1> %mask, <8 x float> %f136, <8 x float> %out_16_2
    store <8 x float> %out_16_3, <8 x float>* %out_16_1, align 32
    ret void;
}

define void @func_e9d5179176fa4aaf3f8d2cb1ff4d2c84(<8 x i1> %mask, i8* noalias %params, i8* noalias %data, <8 x i32> %offsets) #0 {
entry:
    ; VCall: mitsuba::BSDF::eval_pdf()
    %f1_p1 = getelementptr inbounds i8, i8* %data, i32 0
    %f1_p2 = getelementptr inbounds i8, i8* %f1_p1, <8 x i32> %offsets
    %f1_p3 = bitcast <8 x i8*> %f1_p2 to <8 x float*>
    %f1 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f1_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f2_p1 = getelementptr inbounds i8, i8* %data, i32 4
    %f2_p2 = getelementptr inbounds i8, i8* %f2_p1, <8 x i32> %offsets
    %f2_p3 = bitcast <8 x i8*> %f2_p2 to <8 x float*>
    %f2 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f2_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f3_p1 = getelementptr inbounds i8, i8* %data, i32 8
    %f3_p2 = getelementptr inbounds i8, i8* %f3_p1, <8 x i32> %offsets
    %f3_p3 = bitcast <8 x i8*> %f3_p2 to <8 x float*>
    %f3 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f3_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f4_i0 = getelementptr inbounds i8, i8* %params, i64 192
    %f4_i1 = bitcast i8* %f4_i0 to <8 x float>*
    %f4 = load <8 x float>, <8 x float>* %f4_i1, align 32
    %f5_i0 = getelementptr inbounds i8, i8* %params, i64 256
    %f5_i1 = bitcast i8* %f5_i0 to <8 x float>*
    %f5 = load <8 x float>, <8 x float>* %f5_i1, align 32
    %f6_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f6 = shufflevector <8 x float> %f6_1, <8 x float> undef, <8 x i32> zeroinitializer
    %p7 = fcmp ogt <8 x float> %f4, %f6
    %p8 = fcmp ogt <8 x float> %f5, %f6
    %p9 = and <8 x i1> %p7, %p8
    %f10_1 = insertelement <8 x float> undef, float 0x3fd45f3060000000, i32 0
    %f10 = shufflevector <8 x float> %f10_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f11 = fmul <8 x float> %f1, %f10
    %f12 = fmul <8 x float> %f2, %f10
    %f13 = fmul <8 x float> %f3, %f10
    %f14 = fmul <8 x float> %f11, %f5
    %f15 = fmul <8 x float> %f12, %f5
    %f16 = fmul <8 x float> %f13, %f5
    %f17 = fmul <8 x float> %f10, %f5
    %f18 = select <8 x i1> %p9, <8 x float> %f14, <8 x float> zeroinitializer
    %f19 = select <8 x i1> %p9, <8 x float> %f15, <8 x float> zeroinitializer
    %f20 = select <8 x i1> %p9, <8 x float> %f16, <8 x float> zeroinitializer
    %f21 = select <8 x i1> %p9, <8 x float> %f17, <8 x float> zeroinitializer
    %out_0_0 = getelementptr inbounds i8, i8* %params, i64 288
    %out_0_1 = bitcast i8* %out_0_0 to <8 x float>*
    %out_0_2 = load <8 x float>, <8 x float>* %out_0_1, align 32
    %out_0_3 = select <8 x i1> %mask, <8 x float> %f18, <8 x float> %out_0_2
    store <8 x float> %out_0_3, <8 x float>* %out_0_1, align 32
    %out_1_0 = getelementptr inbounds i8, i8* %params, i64 320
    %out_1_1 = bitcast i8* %out_1_0 to <8 x float>*
    %out_1_2 = load <8 x float>, <8 x float>* %out_1_1, align 32
    %out_1_3 = select <8 x i1> %mask, <8 x float> %f19, <8 x float> %out_1_2
    store <8 x float> %out_1_3, <8 x float>* %out_1_1, align 32
    %out_2_0 = getelementptr inbounds i8, i8* %params, i64 352
    %out_2_1 = bitcast i8* %out_2_0 to <8 x float>*
    %out_2_2 = load <8 x float>, <8 x float>* %out_2_1, align 32
    %out_2_3 = select <8 x i1> %mask, <8 x float> %f20, <8 x float> %out_2_2
    store <8 x float> %out_2_3, <8 x float>* %out_2_1, align 32
    %out_3_0 = getelementptr inbounds i8, i8* %params, i64 384
    %out_3_1 = bitcast i8* %out_3_0 to <8 x float>*
    %out_3_2 = load <8 x float>, <8 x float>* %out_3_1, align 32
    %out_3_3 = select <8 x i1> %mask, <8 x float> %f21, <8 x float> %out_3_2
    store <8 x float> %out_3_3, <8 x float>* %out_3_1, align 32
    ret void;
}

define void @func_9d53124ccfd48ba20820a9ee3e480cd9(<8 x i1> %mask, i8* noalias %params, i8* noalias %data, <8 x i32> %offsets) #0 {
entry:
    ; VCall: mitsuba::BSDF::eval_pdf()
    %f1_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f1 = shufflevector <8 x float> %f1_1, <8 x float> undef, <8 x i32> zeroinitializer
    %out_0_0 = getelementptr inbounds i8, i8* %params, i64 288
    %out_0_1 = bitcast i8* %out_0_0 to <8 x float>*
    %out_0_2 = load <8 x float>, <8 x float>* %out_0_1, align 32
    %out_0_3 = select <8 x i1> %mask, <8 x float> %f1, <8 x float> %out_0_2
    store <8 x float> %out_0_3, <8 x float>* %out_0_1, align 32
    %out_1_0 = getelementptr inbounds i8, i8* %params, i64 320
    %out_1_1 = bitcast i8* %out_1_0 to <8 x float>*
    %out_1_2 = load <8 x float>, <8 x float>* %out_1_1, align 32
    %out_1_3 = select <8 x i1> %mask, <8 x float> %f1, <8 x float> %out_1_2
    store <8 x float> %out_1_3, <8 x float>* %out_1_1, align 32
    %out_2_0 = getelementptr inbounds i8, i8* %params, i64 352
    %out_2_1 = bitcast i8* %out_2_0 to <8 x float>*
    %out_2_2 = load <8 x float>, <8 x float>* %out_2_1, align 32
    %out_2_3 = select <8 x i1> %mask, <8 x float> %f1, <8 x float> %out_2_2
    store <8 x float> %out_2_3, <8 x float>* %out_2_1, align 32
    %out_3_0 = getelementptr inbounds i8, i8* %params, i64 384
    %out_3_1 = bitcast i8* %out_3_0 to <8 x float>*
    %out_3_2 = load <8 x float>, <8 x float>* %out_3_1, align 32
    %out_3_3 = select <8 x i1> %mask, <8 x float> %f1, <8 x float> %out_3_2
    store <8 x float> %out_3_3, <8 x float>* %out_3_1, align 32
    ret void;
}

define void @func_4448b4670e928f754bf33368168c3f24(<8 x i1> %mask, i8* noalias %params, i8* noalias %data, <8 x i32> %offsets) #0 {
entry:
    ; VCall: mitsuba::BSDF::eval_pdf()
    %f1_p1 = getelementptr inbounds i8, i8* %data, i32 0
    %f1_p2 = getelementptr inbounds i8, i8* %f1_p1, <8 x i32> %offsets
    %f1_p3 = bitcast <8 x i8*> %f1_p2 to <8 x float*>
    %f1 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f1_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f2_p1 = getelementptr inbounds i8, i8* %data, i32 4
    %f2_p2 = getelementptr inbounds i8, i8* %f2_p1, <8 x i32> %offsets
    %f2_p3 = bitcast <8 x i8*> %f2_p2 to <8 x float*>
    %f2 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f2_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f3_p1 = getelementptr inbounds i8, i8* %data, i32 8
    %f3_p2 = getelementptr inbounds i8, i8* %f3_p1, <8 x i32> %offsets
    %f3_p3 = bitcast <8 x i8*> %f3_p2 to <8 x float*>
    %f3 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f3_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f4_i0 = getelementptr inbounds i8, i8* %params, i64 192
    %f4_i1 = bitcast i8* %f4_i0 to <8 x float>*
    %f4 = load <8 x float>, <8 x float>* %f4_i1, align 32
    %f5_i0 = getelementptr inbounds i8, i8* %params, i64 256
    %f5_i1 = bitcast i8* %f5_i0 to <8 x float>*
    %f5 = load <8 x float>, <8 x float>* %f5_i1, align 32
    %f6_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f6 = shufflevector <8 x float> %f6_1, <8 x float> undef, <8 x i32> zeroinitializer
    %p7 = fcmp ogt <8 x float> %f4, %f6
    %p8 = fcmp ogt <8 x float> %f5, %f6
    %p9 = and <8 x i1> %p7, %p8
    %f10_1 = insertelement <8 x float> undef, float 0x3fd45f3060000000, i32 0
    %f10 = shufflevector <8 x float> %f10_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f11 = fmul <8 x float> %f10, %f5
    %f12 = select <8 x i1> %p9, <8 x float> %f11, <8 x float> zeroinitializer
    %f13 = fmul <8 x float> %f1, %f10
    %f14 = fmul <8 x float> %f2, %f10
    %f15 = fmul <8 x float> %f3, %f10
    %f16 = fmul <8 x float> %f13, %f5
    %f17 = fmul <8 x float> %f14, %f5
    %f18 = fmul <8 x float> %f15, %f5
    %f19 = select <8 x i1> %p9, <8 x float> %f16, <8 x float> zeroinitializer
    %f20 = select <8 x i1> %p9, <8 x float> %f17, <8 x float> zeroinitializer
    %f21 = select <8 x i1> %p9, <8 x float> %f18, <8 x float> zeroinitializer
    %out_0_0 = getelementptr inbounds i8, i8* %params, i64 288
    %out_0_1 = bitcast i8* %out_0_0 to <8 x float>*
    %out_0_2 = load <8 x float>, <8 x float>* %out_0_1, align 32
    %out_0_3 = select <8 x i1> %mask, <8 x float> %f19, <8 x float> %out_0_2
    store <8 x float> %out_0_3, <8 x float>* %out_0_1, align 32
    %out_1_0 = getelementptr inbounds i8, i8* %params, i64 320
    %out_1_1 = bitcast i8* %out_1_0 to <8 x float>*
    %out_1_2 = load <8 x float>, <8 x float>* %out_1_1, align 32
    %out_1_3 = select <8 x i1> %mask, <8 x float> %f20, <8 x float> %out_1_2
    store <8 x float> %out_1_3, <8 x float>* %out_1_1, align 32
    %out_2_0 = getelementptr inbounds i8, i8* %params, i64 352
    %out_2_1 = bitcast i8* %out_2_0 to <8 x float>*
    %out_2_2 = load <8 x float>, <8 x float>* %out_2_1, align 32
    %out_2_3 = select <8 x i1> %mask, <8 x float> %f21, <8 x float> %out_2_2
    store <8 x float> %out_2_3, <8 x float>* %out_2_1, align 32
    %out_3_0 = getelementptr inbounds i8, i8* %params, i64 384
    %out_3_1 = bitcast i8* %out_3_0 to <8 x float>*
    %out_3_2 = load <8 x float>, <8 x float>* %out_3_1, align 32
    %out_3_3 = select <8 x i1> %mask, <8 x float> %f12, <8 x float> %out_3_2
    store <8 x float> %out_3_3, <8 x float>* %out_3_1, align 32
    ret void;
}

define void @func_60a1a348aca7bdb47e851f6cedce38a2(<8 x i1> %mask, i8* noalias %params, i8* noalias %data, <8 x i32> %offsets) #0 {
entry:
    ; VCall: mitsuba::BSDF::sample()
    %f1_p1 = getelementptr inbounds i8, i8* %data, i32 0
    %f1_p2 = getelementptr inbounds i8, i8* %f1_p1, <8 x i32> %offsets
    %f1_p3 = bitcast <8 x i8*> %f1_p2 to <8 x float*>
    %f1 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f1_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f2_p1 = getelementptr inbounds i8, i8* %data, i32 4
    %f2_p2 = getelementptr inbounds i8, i8* %f2_p1, <8 x i32> %offsets
    %f2_p3 = bitcast <8 x i8*> %f2_p2 to <8 x float*>
    %f2 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f2_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f3_p1 = getelementptr inbounds i8, i8* %data, i32 8
    %f3_p2 = getelementptr inbounds i8, i8* %f3_p1, <8 x i32> %offsets
    %f3_p3 = bitcast <8 x i8*> %f3_p2 to <8 x float*>
    %f3 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f3_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f4_i0 = getelementptr inbounds i8, i8* %params, i64 256
    %f4_i1 = bitcast i8* %f4_i0 to <8 x float>*
    %f4 = load <8 x float>, <8 x float>* %f4_i1, align 32
    %f5_i0 = getelementptr inbounds i8, i8* %params, i64 352
    %f5_i1 = bitcast i8* %f5_i0 to <8 x float>*
    %f5 = load <8 x float>, <8 x float>* %f5_i1, align 32
    %f6_i0 = getelementptr inbounds i8, i8* %params, i64 384
    %f6_i1 = bitcast i8* %f6_i0 to <8 x float>*
    %f6 = load <8 x float>, <8 x float>* %f6_i1, align 32
    %f7_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f7 = shufflevector <8 x float> %f7_1, <8 x float> undef, <8 x i32> zeroinitializer
    %p8 = fcmp ogt <8 x float> %f4, %f7
    %f9_1 = insertelement <8 x float> undef, float 0x4000000000000000, i32 0
    %f9 = shufflevector <8 x float> %f9_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f10_1 = insertelement <8 x float> undef, float 0xbff0000000000000, i32 0
    %f10 = shufflevector <8 x float> %f10_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f11 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f9, <8 x float> %f5, <8 x float> %f10)
    %f12 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f9, <8 x float> %f6, <8 x float> %f10)
    %p13 = fcmp oeq <8 x float> %f11, %f7
    %p14 = fcmp oeq <8 x float> %f12, %f7
    %p15 = and <8 x i1> %p13, %p14
    %f16_1 = insertelement <8 x float> undef, float 0x7fffffffe0000000, i32 0
    %f16 = shufflevector <8 x float> %f16_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f17_0 = bitcast <8 x float> %f11 to <8 x i32>
    %f17_1 = bitcast <8 x float> %f16 to <8 x i32>
    %f17_2 = and <8 x i32> %f17_0, %f17_1
    %f17 = bitcast <8 x i32> %f17_2 to <8 x float>
    %f18_0 = bitcast <8 x float> %f12 to <8 x i32>
    %f18_1 = bitcast <8 x float> %f16 to <8 x i32>
    %f18_2 = and <8 x i32> %f18_0, %f18_1
    %f18 = bitcast <8 x i32> %f18_2 to <8 x float>
    %p19 = fcmp olt <8 x float> %f17, %f18
    %f20 = select <8 x i1> %p19, <8 x float> %f12, <8 x float> %f11
    %f21 = select <8 x i1> %p19, <8 x float> %f11, <8 x float> %f12
    %f22_1 = insertelement <8 x float> undef, float 0x3fe921fb60000000, i32 0
    %f22 = shufflevector <8 x float> %f22_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f23 = fmul <8 x float> %f22, %f21
    %f24 = fdiv <8 x float> %f23, %f20
    %f25_1 = insertelement <8 x float> undef, float 0x3ff921fb60000000, i32 0
    %f25 = shufflevector <8 x float> %f25_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f26 = fsub <8 x float> %f25, %f24
    %f27 = select <8 x i1> %p19, <8 x float> %f26, <8 x float> %f24
    %f28 = select <8 x i1> %p15, <8 x float> %f7, <8 x float> %f27
    %f29_0 = bitcast <8 x float> %f28 to <8 x i32>
    %f29_1 = bitcast <8 x float> %f16 to <8 x i32>
    %f29_2 = and <8 x i32> %f29_0, %f29_1
    %f29 = bitcast <8 x i32> %f29_2 to <8 x float>
    %f30_1 = insertelement <8 x float> undef, float 0x3ff45f3060000000, i32 0
    %f30 = shufflevector <8 x float> %f30_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f31 = fmul <8 x float> %f29, %f30
    %r32 = fptosi <8 x float> %f31 to <8 x i32>
    %r33_1 = insertelement <8 x i32> undef, i32 1, i32 0
    %r33 = shufflevector <8 x i32> %r33_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r34 = add <8 x i32> %r32, %r33
    %r35_1 = insertelement <8 x i32> undef, i32 4294967294, i32 0
    %r35 = shufflevector <8 x i32> %r35_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r36 = and <8 x i32> %r34, %r35
    %f37 = sitofp <8 x i32> %r36 to <8 x float>
    %r38_1 = insertelement <8 x i32> undef, i32 29, i32 0
    %r38 = shufflevector <8 x i32> %r38_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r39 = shl <8 x i32> %r36, %r38
    %f40 = bitcast <8 x i32> %r39 to <8 x float>
    %f41_0 = bitcast <8 x float> %f40 to <8 x i32>
    %f41_1 = bitcast <8 x float> %f28 to <8 x i32>
    %f41_2 = xor <8 x i32> %f41_0, %f41_1
    %f41 = bitcast <8 x i32> %f41_2 to <8 x float>
    %r42_1 = insertelement <8 x i32> undef, i32 2, i32 0
    %r42 = shufflevector <8 x i32> %r42_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r43 = sub <8 x i32> %r36, %r42
    %r44 = xor <8 x i32> %r43, <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1>
    %r45 = shl <8 x i32> %r44, %r38
    %f46 = bitcast <8 x i32> %r45 to <8 x float>
    %f47_1 = insertelement <8 x float> undef, float 0x3fe9200000000000, i32 0
    %f47 = shufflevector <8 x float> %f47_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f48 = fmul <8 x float> %f37, %f47
    %f49 = fsub <8 x float> %f29, %f48
    %f50_1 = insertelement <8 x float> undef, float 0x3f2fb40000000000, i32 0
    %f50 = shufflevector <8 x float> %f50_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f51 = fmul <8 x float> %f37, %f50
    %f52 = fsub <8 x float> %f49, %f51
    %f53_1 = insertelement <8 x float> undef, float 0x3e64442d20000000, i32 0
    %f53 = shufflevector <8 x float> %f53_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f54 = fmul <8 x float> %f37, %f53
    %f55 = fsub <8 x float> %f52, %f54
    %f56 = fmul <8 x float> %f55, %f55
    %f57_1 = insertelement <8 x float> undef, float 0x7ff0000000000000, i32 0
    %f57 = shufflevector <8 x float> %f57_1, <8 x float> undef, <8 x i32> zeroinitializer
    %p58 = fcmp oeq <8 x float> %f29, %f57
    %f59_0 = sext <8 x i1> %p58 to <8 x i32>
    %f59_1 = bitcast <8 x float> %f56 to <8 x i32>
    %f59_2 = or <8 x i32> %f59_0, %f59_1
    %f59 = bitcast <8 x i32> %f59_2 to <8 x float>
    %f60_1 = insertelement <8 x float> undef, float 0xbfc5555460000000, i32 0
    %f60 = shufflevector <8 x float> %f60_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f61_1 = insertelement <8 x float> undef, float 0x3f811073c0000000, i32 0
    %f61 = shufflevector <8 x float> %f61_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f62_1 = insertelement <8 x float> undef, float 0xbf29943f20000000, i32 0
    %f62 = shufflevector <8 x float> %f62_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f63 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f59, <8 x float> %f61, <8 x float> %f60)
    %f64 = fmul <8 x float> %f59, %f59
    %f65 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f64, <8 x float> %f62, <8 x float> %f63)
    %f66 = fmul <8 x float> %f65, %f59
    %f67_1 = insertelement <8 x float> undef, float 0x3fa55554a0000000, i32 0
    %f67 = shufflevector <8 x float> %f67_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f68_1 = insertelement <8 x float> undef, float 0xbf56c0c340000000, i32 0
    %f68 = shufflevector <8 x float> %f68_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f69_1 = insertelement <8 x float> undef, float 0x3ef99eb9c0000000, i32 0
    %f69 = shufflevector <8 x float> %f69_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f70 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f59, <8 x float> %f68, <8 x float> %f67)
    %f71 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f64, <8 x float> %f69, <8 x float> %f70)
    %f72 = fmul <8 x float> %f71, %f59
    %f73 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f66, <8 x float> %f55, <8 x float> %f55)
    %f74_1 = insertelement <8 x float> undef, float 0xbfe0000000000000, i32 0
    %f74 = shufflevector <8 x float> %f74_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f75_1 = insertelement <8 x float> undef, float 0x3ff0000000000000, i32 0
    %f75 = shufflevector <8 x float> %f75_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f76 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f59, <8 x float> %f74, <8 x float> %f75)
    %f77 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f72, <8 x float> %f59, <8 x float> %f76)
    %r78 = and <8 x i32> %r36, %r42
    %r79_1 = insertelement <8 x i32> undef, i32 0, i32 0
    %r79 = shufflevector <8 x i32> %r79_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %p80 = icmp eq <8 x i32> %r78, %r79
    %f81 = select <8 x i1> %p80, <8 x float> %f73, <8 x float> %f77
    %f82_1 = insertelement <8 x float> undef, float 0x8000000000000000, i32 0
    %f82 = shufflevector <8 x float> %f82_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f83_0 = bitcast <8 x float> %f82 to <8 x i32>
    %f83_1 = bitcast <8 x float> %f41 to <8 x i32>
    %f83_2 = and <8 x i32> %f83_0, %f83_1
    %f83 = bitcast <8 x i32> %f83_2 to <8 x float>
    %f84_0 = bitcast <8 x float> %f81 to <8 x i32>
    %f84_1 = bitcast <8 x float> %f83 to <8 x i32>
    %f84_2 = xor <8 x i32> %f84_0, %f84_1
    %f84 = bitcast <8 x i32> %f84_2 to <8 x float>
    %f85 = select <8 x i1> %p80, <8 x float> %f77, <8 x float> %f73
    %f86_0 = bitcast <8 x float> %f82 to <8 x i32>
    %f86_1 = bitcast <8 x float> %f46 to <8 x i32>
    %f86_2 = and <8 x i32> %f86_0, %f86_1
    %f86 = bitcast <8 x i32> %f86_2 to <8 x float>
    %f87_0 = bitcast <8 x float> %f85 to <8 x i32>
    %f87_1 = bitcast <8 x float> %f86 to <8 x i32>
    %f87_2 = xor <8 x i32> %f87_0, %f87_1
    %f87 = bitcast <8 x i32> %f87_2 to <8 x float>
    %f88 = fmul <8 x float> %f20, %f87
    %f89 = fmul <8 x float> %f20, %f84
    %f90 = fmul <8 x float> %f88, %f88
    %f91 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f89, <8 x float> %f89, <8 x float> %f90)
    %f92 = fsub <8 x float> %f75, %f91
    %f93 = call <8 x float> @llvm.maxnum.v8f32(<8 x float> %f92, <8 x float> %f7)
    %f94 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> %f93)
    %f95_1 = insertelement <8 x float> undef, float 0x3fd45f3060000000, i32 0
    %f95 = shufflevector <8 x float> %f95_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f96 = fmul <8 x float> %f95, %f94
    %r97_1 = insertelement <8 x i32> undef, i32 2, i32 0
    %r97 = shufflevector <8 x i32> %r97_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %p98 = fcmp ogt <8 x float> %f96, %f7
    %p99 = and <8 x i1> %p8, %p98
    %f100 = select <8 x i1> %p99, <8 x float> %f1, <8 x float> zeroinitializer
    %f101 = select <8 x i1> %p99, <8 x float> %f2, <8 x float> zeroinitializer
    %f102 = select <8 x i1> %p99, <8 x float> %f3, <8 x float> zeroinitializer
    %out_0_0 = getelementptr inbounds i8, i8* %params, i64 416
    %out_0_1 = bitcast i8* %out_0_0 to <8 x float>*
    %out_0_2 = load <8 x float>, <8 x float>* %out_0_1, align 32
    %out_0_3 = select <8 x i1> %mask, <8 x float> %f88, <8 x float> %out_0_2
    store <8 x float> %out_0_3, <8 x float>* %out_0_1, align 32
    %out_1_0 = getelementptr inbounds i8, i8* %params, i64 448
    %out_1_1 = bitcast i8* %out_1_0 to <8 x float>*
    %out_1_2 = load <8 x float>, <8 x float>* %out_1_1, align 32
    %out_1_3 = select <8 x i1> %mask, <8 x float> %f89, <8 x float> %out_1_2
    store <8 x float> %out_1_3, <8 x float>* %out_1_1, align 32
    %out_2_0 = getelementptr inbounds i8, i8* %params, i64 480
    %out_2_1 = bitcast i8* %out_2_0 to <8 x float>*
    %out_2_2 = load <8 x float>, <8 x float>* %out_2_1, align 32
    %out_2_3 = select <8 x i1> %mask, <8 x float> %f94, <8 x float> %out_2_2
    store <8 x float> %out_2_3, <8 x float>* %out_2_1, align 32
    %out_3_0 = getelementptr inbounds i8, i8* %params, i64 512
    %out_3_1 = bitcast i8* %out_3_0 to <8 x float>*
    %out_3_2 = load <8 x float>, <8 x float>* %out_3_1, align 32
    %out_3_3 = select <8 x i1> %mask, <8 x float> %f96, <8 x float> %out_3_2
    store <8 x float> %out_3_3, <8 x float>* %out_3_1, align 32
    %out_4_0 = getelementptr inbounds i8, i8* %params, i64 544
    %out_4_1 = bitcast i8* %out_4_0 to <8 x float>*
    %out_4_2 = load <8 x float>, <8 x float>* %out_4_1, align 32
    %out_4_3 = select <8 x i1> %mask, <8 x float> %f75, <8 x float> %out_4_2
    store <8 x float> %out_4_3, <8 x float>* %out_4_1, align 32
    %out_5_0 = getelementptr inbounds i8, i8* %params, i64 576
    %out_5_1 = bitcast i8* %out_5_0 to <8 x i32>*
    %out_5_2 = load <8 x i32>, <8 x i32>* %out_5_1, align 32
    %out_5_3 = select <8 x i1> %mask, <8 x i32> %r97, <8 x i32> %out_5_2
    store <8 x i32> %out_5_3, <8 x i32>* %out_5_1, align 32
    %out_7_0 = getelementptr inbounds i8, i8* %params, i64 608
    %out_7_1 = bitcast i8* %out_7_0 to <8 x float>*
    %out_7_2 = load <8 x float>, <8 x float>* %out_7_1, align 32
    %out_7_3 = select <8 x i1> %mask, <8 x float> %f100, <8 x float> %out_7_2
    store <8 x float> %out_7_3, <8 x float>* %out_7_1, align 32
    %out_8_0 = getelementptr inbounds i8, i8* %params, i64 640
    %out_8_1 = bitcast i8* %out_8_0 to <8 x float>*
    %out_8_2 = load <8 x float>, <8 x float>* %out_8_1, align 32
    %out_8_3 = select <8 x i1> %mask, <8 x float> %f101, <8 x float> %out_8_2
    store <8 x float> %out_8_3, <8 x float>* %out_8_1, align 32
    %out_9_0 = getelementptr inbounds i8, i8* %params, i64 672
    %out_9_1 = bitcast i8* %out_9_0 to <8 x float>*
    %out_9_2 = load <8 x float>, <8 x float>* %out_9_1, align 32
    %out_9_3 = select <8 x i1> %mask, <8 x float> %f102, <8 x float> %out_9_2
    store <8 x float> %out_9_3, <8 x float>* %out_9_1, align 32
    ret void;
}

define void @func_e6fff8d6b222436f71518443ce068358(<8 x i1> %mask, i8* noalias %params, i8* noalias %data, <8 x i32> %offsets) #0 {
entry:
    ; VCall: mitsuba::BSDF::sample()
    %f1_i0 = getelementptr inbounds i8, i8* %params, i64 256
    %f1_i1 = bitcast i8* %f1_i0 to <8 x float>*
    %f1 = load <8 x float>, <8 x float>* %f1_i1, align 32
    %f2_i0 = getelementptr inbounds i8, i8* %params, i64 192
    %f2_i1 = bitcast i8* %f2_i0 to <8 x float>*
    %f2 = load <8 x float>, <8 x float>* %f2_i1, align 32
    %f3_i0 = getelementptr inbounds i8, i8* %params, i64 224
    %f3_i1 = bitcast i8* %f3_i0 to <8 x float>*
    %f3 = load <8 x float>, <8 x float>* %f3_i1, align 32
    %f4_i0 = getelementptr inbounds i8, i8* %params, i64 320
    %f4_i1 = bitcast i8* %f4_i0 to <8 x float>*
    %f4 = load <8 x float>, <8 x float>* %f4_i1, align 32
    %f5_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f5 = shufflevector <8 x float> %f5_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f6_1 = insertelement <8 x float> undef, float 0x7fffffffe0000000, i32 0
    %f6 = shufflevector <8 x float> %f6_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f7_1 = insertelement <8 x float> undef, float 0x3ff0000000000000, i32 0
    %f7 = shufflevector <8 x float> %f7_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f8_1 = insertelement <8 x float> undef, float 0x3ff8112280000000, i32 0
    %f8 = shufflevector <8 x float> %f8_1, <8 x float> undef, <8 x i32> zeroinitializer
    %p9 = fcmp oge <8 x float> %f1, %f5
    %f10_1 = insertelement <8 x float> undef, float 0x3fe5462520000000, i32 0
    %f10 = shufflevector <8 x float> %f10_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f11 = select <8 x i1> %p9, <8 x float> %f8, <8 x float> %f10
    %f12 = select <8 x i1> %p9, <8 x float> %f10, <8 x float> %f8
    %f13 = fneg <8 x float> %f1
    %f14 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f1, <8 x float> %f13, <8 x float> %f7)
    %f15 = fmul <8 x float> %f12, %f12
    %f16 = fneg <8 x float> %f15
    %f17 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f14, <8 x float> %f16, <8 x float> %f7)
    %f18_0 = bitcast <8 x float> %f1 to <8 x i32>
    %f18_1 = bitcast <8 x float> %f6 to <8 x i32>
    %f18_2 = and <8 x i32> %f18_0, %f18_1
    %f18 = bitcast <8 x i32> %f18_2 to <8 x float>
    %f19 = call <8 x float> @llvm.maxnum.v8f32(<8 x float> %f17, <8 x float> %f5)
    %f20 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> %f19)
    %p21 = fcmp oeq <8 x float> %f18, %f5
    %f22 = fneg <8 x float> %f20
    %f23 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f11, <8 x float> %f22, <8 x float> %f18)
    %f24 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f11, <8 x float> %f20, <8 x float> %f18)
    %f25 = fdiv <8 x float> %f23, %f24
    %f26 = fneg <8 x float> %f18
    %f27 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f11, <8 x float> %f26, <8 x float> %f20)
    %f28 = call <8 x float> @llvm.fma.v8f32(<8 x float> %f11, <8 x float> %f18, <8 x float> %f20)
    %f29 = fdiv <8 x float> %f27, %f28
    %f30 = fmul <8 x float> %f25, %f25
    %f31 = fmul <8 x float> %f29, %f29
    %f32 = fadd <8 x float> %f30, %f31
    %f33_1 = insertelement <8 x float> undef, float 0x3fe0000000000000, i32 0
    %f33 = shufflevector <8 x float> %f33_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f34 = fmul <8 x float> %f33, %f32
    %f35 = select <8 x i1> %p21, <8 x float> %f7, <8 x float> %f34
    %f36 = select <8 x i1> %p9, <8 x float> %f22, <8 x float> %f20
    %f37 = fsub <8 x float> %f7, %f35
    %p38 = fcmp ole <8 x float> %f4, %f35
    %f39 = select <8 x i1> %p38, <8 x float> %f35, <8 x float> %f37
    %p40 = xor <8 x i1> %p38, <i1 1, i1 1, i1 1, i1 1, i1 1, i1 1, i1 1, i1 1>
    %r41_1 = insertelement <8 x i32> undef, i32 32, i32 0
    %r41 = shufflevector <8 x i32> %r41_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r42_1 = insertelement <8 x i32> undef, i32 64, i32 0
    %r42 = shufflevector <8 x i32> %r42_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %r43 = select <8 x i1> %p38, <8 x i32> %r41, <8 x i32> %r42
    %f44 = fneg <8 x float> %f2
    %f45 = fneg <8 x float> %f3
    %f46 = fneg <8 x float> %f12
    %f47 = fmul <8 x float> %f46, %f2
    %f48 = fmul <8 x float> %f46, %f3
    %f49 = select <8 x i1> %p38, <8 x float> %f44, <8 x float> %f47
    %f50 = select <8 x i1> %p38, <8 x float> %f45, <8 x float> %f48
    %f51 = select <8 x i1> %p38, <8 x float> %f1, <8 x float> %f36
    %f52 = select <8 x i1> %p38, <8 x float> %f7, <8 x float> %f11
    %f53 = select <8 x i1> %p40, <8 x float> %f15, <8 x float> %f7
    %out_0_0 = getelementptr inbounds i8, i8* %params, i64 416
    %out_0_1 = bitcast i8* %out_0_0 to <8 x float>*
    %out_0_2 = load <8 x float>, <8 x float>* %out_0_1, align 32
    %out_0_3 = select <8 x i1> %mask, <8 x float> %f49, <8 x float> %out_0_2
    store <8 x float> %out_0_3, <8 x float>* %out_0_1, align 32
    %out_1_0 = getelementptr inbounds i8, i8* %params, i64 448
    %out_1_1 = bitcast i8* %out_1_0 to <8 x float>*
    %out_1_2 = load <8 x float>, <8 x float>* %out_1_1, align 32
    %out_1_3 = select <8 x i1> %mask, <8 x float> %f50, <8 x float> %out_1_2
    store <8 x float> %out_1_3, <8 x float>* %out_1_1, align 32
    %out_2_0 = getelementptr inbounds i8, i8* %params, i64 480
    %out_2_1 = bitcast i8* %out_2_0 to <8 x float>*
    %out_2_2 = load <8 x float>, <8 x float>* %out_2_1, align 32
    %out_2_3 = select <8 x i1> %mask, <8 x float> %f51, <8 x float> %out_2_2
    store <8 x float> %out_2_3, <8 x float>* %out_2_1, align 32
    %out_3_0 = getelementptr inbounds i8, i8* %params, i64 512
    %out_3_1 = bitcast i8* %out_3_0 to <8 x float>*
    %out_3_2 = load <8 x float>, <8 x float>* %out_3_1, align 32
    %out_3_3 = select <8 x i1> %mask, <8 x float> %f39, <8 x float> %out_3_2
    store <8 x float> %out_3_3, <8 x float>* %out_3_1, align 32
    %out_4_0 = getelementptr inbounds i8, i8* %params, i64 544
    %out_4_1 = bitcast i8* %out_4_0 to <8 x float>*
    %out_4_2 = load <8 x float>, <8 x float>* %out_4_1, align 32
    %out_4_3 = select <8 x i1> %mask, <8 x float> %f52, <8 x float> %out_4_2
    store <8 x float> %out_4_3, <8 x float>* %out_4_1, align 32
    %out_5_0 = getelementptr inbounds i8, i8* %params, i64 576
    %out_5_1 = bitcast i8* %out_5_0 to <8 x i32>*
    %out_5_2 = load <8 x i32>, <8 x i32>* %out_5_1, align 32
    %out_5_3 = select <8 x i1> %mask, <8 x i32> %r43, <8 x i32> %out_5_2
    store <8 x i32> %out_5_3, <8 x i32>* %out_5_1, align 32
    %out_7_0 = getelementptr inbounds i8, i8* %params, i64 608
    %out_7_1 = bitcast i8* %out_7_0 to <8 x float>*
    %out_7_2 = load <8 x float>, <8 x float>* %out_7_1, align 32
    %out_7_3 = select <8 x i1> %mask, <8 x float> %f53, <8 x float> %out_7_2
    store <8 x float> %out_7_3, <8 x float>* %out_7_1, align 32
    %out_8_0 = getelementptr inbounds i8, i8* %params, i64 640
    %out_8_1 = bitcast i8* %out_8_0 to <8 x float>*
    %out_8_2 = load <8 x float>, <8 x float>* %out_8_1, align 32
    %out_8_3 = select <8 x i1> %mask, <8 x float> %f53, <8 x float> %out_8_2
    store <8 x float> %out_8_3, <8 x float>* %out_8_1, align 32
    %out_9_0 = getelementptr inbounds i8, i8* %params, i64 672
    %out_9_1 = bitcast i8* %out_9_0 to <8 x float>*
    %out_9_2 = load <8 x float>, <8 x float>* %out_9_1, align 32
    %out_9_3 = select <8 x i1> %mask, <8 x float> %f53, <8 x float> %out_9_2
    store <8 x float> %out_9_3, <8 x float>* %out_9_1, align 32
    ret void;
}

define void @func_48d4a1fb2376a0011da8c81578897b84(<8 x i1> %mask, i8* noalias %params, i8* noalias %data, <8 x i32> %offsets) #0 {
entry:
    ; VCall: mitsuba::BSDF::sample()
    %f1_p1 = getelementptr inbounds i8, i8* %data, i32 0
    %f1_p2 = getelementptr inbounds i8, i8* %f1_p1, <8 x i32> %offsets
    %f1_p3 = bitcast <8 x i8*> %f1_p2 to <8 x float*>
    %f1 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f1_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f2_p1 = getelementptr inbounds i8, i8* %data, i32 4
    %f2_p2 = getelementptr inbounds i8, i8* %f2_p1, <8 x i32> %offsets
    %f2_p3 = bitcast <8 x i8*> %f2_p2 to <8 x float*>
    %f2 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f2_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f3_p1 = getelementptr inbounds i8, i8* %data, i32 8
    %f3_p2 = getelementptr inbounds i8, i8* %f3_p1, <8 x i32> %offsets
    %f3_p3 = bitcast <8 x i8*> %f3_p2 to <8 x float*>
    %f3 = call <8 x float> @llvm.masked.gather.v8f32(<8 x float*> %f3_p3, i32 4, <8 x i1> %mask, <8 x float> zeroinitializer)
    %f4_i0 = getelementptr inbounds i8, i8* %params, i64 256
    %f4_i1 = bitcast i8* %f4_i0 to <8 x float>*
    %f4 = load <8 x float>, <8 x float>* %f4_i1, align 32
    %f5_i0 = getelementptr inbounds i8, i8* %params, i64 192
    %f5_i1 = bitcast i8* %f5_i0 to <8 x float>*
    %f5 = load <8 x float>, <8 x float>* %f5_i1, align 32
    %f6_i0 = getelementptr inbounds i8, i8* %params, i64 224
    %f6_i1 = bitcast i8* %f6_i0 to <8 x float>*
    %f6 = load <8 x float>, <8 x float>* %f6_i1, align 32
    %f7_1 = insertelement <8 x float> undef, float 0x0, i32 0
    %f7 = shufflevector <8 x float> %f7_1, <8 x float> undef, <8 x i32> zeroinitializer
    %p8 = fcmp ogt <8 x float> %f4, %f7
    %f9_1 = insertelement <8 x float> undef, float 0x4000000000000000, i32 0
    %f9 = shufflevector <8 x float> %f9_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f10_1 = insertelement <8 x float> undef, float 0x3ff0000000000000, i32 0
    %f10 = shufflevector <8 x float> %f10_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f11_1 = insertelement <8 x float> undef, float 0x3fe0000000000000, i32 0
    %f11 = shufflevector <8 x float> %f11_1, <8 x float> undef, <8 x i32> zeroinitializer
    %r12_1 = insertelement <8 x i32> undef, i32 32, i32 0
    %r12 = shufflevector <8 x i32> %r12_1, <8 x i32> undef, <8 x i32> zeroinitializer
    %f13 = fneg <8 x float> %f5
    %f14 = fneg <8 x float> %f6
    %f15 = fmul <8 x float> %f4, %f4
    %f16 = fsub <8 x float> %f10, %f15
    %f17 = fmul <8 x float> %f16, %f16
    %f18 = fmul <8 x float> %f2, %f2
    %f19 = fmul <8 x float> %f3, %f3
    %f20 = fsub <8 x float> %f18, %f19
    %f21 = fsub <8 x float> %f20, %f16
    %f22 = fmul <8 x float> %f21, %f21
    %f23_1 = insertelement <8 x float> undef, float 0x4010000000000000, i32 0
    %f23 = shufflevector <8 x float> %f23_1, <8 x float> undef, <8 x i32> zeroinitializer
    %f24 = fmul <8 x float> %f23, %f3
    %f25 = fmul <8 x float> %f24, %f3
    %f26 = fmul <8 x float> %f25, %f2
    %f27 = fmul <8 x float> %f26, %f2
    %f28 = fadd <8 x float> %f22, %f27
    %f29 = call <8 x float> @llvm.maxnum.v8f32(<8 x float> %f28, <8 x float> %f7)
    %f30 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> %f29)
    %f31 = fadd <8 x float> %f30, %f21
    %f32 = fmul <8 x float> %f11, %f31
    %f33 = call <8 x float> @llvm.maxnum.v8f32(<8 x float> %f32, <8 x float> %f7)
    %f34 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> %f33)
    %f35 = fadd <8 x float> %f30, %f15
    %f36 = fmul <8 x float> %f9, %f4
    %f37 = fmul <8 x float> %f36, %f34
    %f38 = fsub <8 x float> %f35, %f37
    %f39 = fadd <8 x float> %f35, %f37
    %f40 = fdiv <8 x float> %f38, %f39
    %f41 = fmul <8 x float> %f30, %f15
    %f42 = fadd <8 x float> %f41, %f17
    %f43 = fmul <8 x float> %f37, %f16
    %f44 = fsub <8 x float> %f42, %f43
    %f45 = fmul <8 x float> %f40, %f44
    %f46 = fadd <8 x float> %f42, %f43
    %f47 = fdiv <8 x float> %f45, %f46
    %f48 = fadd <8 x float> %f40, %f47
    %f49 = fmul <8 x float> %f11, %f48
    %f50 = fmul <8 x float> %f1, %f49
    %f51 = select <8 x i1> %p8, <8 x float> %f50, <8 x float> zeroinitializer
    %out_0_0 = getelementptr inbounds i8, i8* %params, i64 416
    %out_0_1 = bitcast i8* %out_0_0 to <8 x float>*
    %out_0_2 = load <8 x float>, <8 x float>* %out_0_1, align 32
    %out_0_3 = select <8 x i1> %mask, <8 x float> %f13, <8 x float> %out_0_2
    store <8 x float> %out_0_3, <8 x float>* %out_0_1, align 32
    %out_1_0 = getelementptr inbounds i8, i8* %params, i64 448
    %out_1_1 = bitcast i8* %out_1_0 to <8 x float>*
    %out_1_2 = load <8 x float>, <8 x float>* %out_1_1, align 32
    %out_1_3 = select <8 x i1> %mask, <8 x float> %f14, <8 x float> %out_1_2
    store <8 x float> %out_1_3, <8 x float>* %out_1_1, align 32
    %out_2_0 = getelementptr inbounds i8, i8* %params, i64 480
    %out_2_1 = bitcast i8* %out_2_0 to <8 x float>*
    %out_2_2 = load <8 x float>, <8 x float>* %out_2_1, align 32
    %out_2_3 = select <8 x i1> %mask, <8 x float> %f4, <8 x float> %out_2_2
    store <8 x float> %out_2_3, <8 x float>* %out_2_1, align 32
    %out_3_0 = getelementptr inbounds i8, i8* %params, i64 512
    %out_3_1 = bitcast i8* %out_3_0 to <8 x float>*
    %out_3_2 = load <8 x float>, <8 x float>* %out_3_1, align 32
    %out_3_3 = select <8 x i1> %mask, <8 x float> %f10, <8 x float> %out_3_2
    store <8 x float> %out_3_3, <8 x float>* %out_3_1, align 32
    %out_4_0 = getelementptr inbounds i8, i8* %params, i64 544
    %out_4_1 = bitcast i8* %out_4_0 to <8 x float>*
    %out_4_2 = load <8 x float>, <8 x float>* %out_4_1, align 32
    %out_4_3 = select <8 x i1> %mask, <8 x float> %f10, <8 x float> %out_4_2
    store <8 x float> %out_4_3, <8 x float>* %out_4_1, align 32
    %out_5_0 = getelementptr inbounds i8, i8* %params, i64 576
    %out_5_1 = bitcast i8* %out_5_0 to <8 x i32>*
    %out_5_2 = load <8 x i32>, <8 x i32>* %out_5_1, align 32
    %out_5_3 = select <8 x i1> %mask, <8 x i32> %r12, <8 x i32> %out_5_2
    store <8 x i32> %out_5_3, <8 x i32>* %out_5_1, align 32
    %out_7_0 = getelementptr inbounds i8, i8* %params, i64 608
    %out_7_1 = bitcast i8* %out_7_0 to <8 x float>*
    %out_7_2 = load <8 x float>, <8 x float>* %out_7_1, align 32
    %out_7_3 = select <8 x i1> %mask, <8 x float> %f51, <8 x float> %out_7_2
    store <8 x float> %out_7_3, <8 x float>* %out_7_1, align 32
    %out_8_0 = getelementptr inbounds i8, i8* %params, i64 640
    %out_8_1 = bitcast i8* %out_8_0 to <8 x float>*
    %out_8_2 = load <8 x float>, <8 x float>* %out_8_1, align 32
    %out_8_3 = select <8 x i1> %mask, <8 x float> %f51, <8 x float> %out_8_2
    store <8 x float> %out_8_3, <8 x float>* %out_8_1, align 32
    %out_9_0 = getelementptr inbounds i8, i8* %params, i64 672
    %out_9_1 = bitcast i8* %out_9_0 to <8 x float>*
    %out_9_2 = load <8 x float>, <8 x float>* %out_9_1, align 32
    %out_9_3 = select <8 x i1> %mask, <8 x float> %f51, <8 x float> %out_9_2
    store <8 x float> %out_9_3, <8 x float>* %out_9_1, align 32
    ret void;
}

declare <8 x float> @llvm.fma.v8f32(<8 x float>, <8 x float>, <8 x float>)

declare <8 x float> @llvm.sqrt.v8f32(<8 x float>)

declare i1 @llvm.experimental.vector.reduce.or.v8i1(<8 x i1>)

declare <8 x i32> @llvm.masked.gather.v8i32(<8 x i32*>, i32, <8 x i1>, <8 x i32>)

declare <8 x float> @llvm.masked.gather.v8f32(<8 x float*>, i32, <8 x i1>, <8 x float>)

declare <8 x i64> @llvm.masked.gather.v8i64(<8 x i64*>, i32, <8 x i1>, <8 x i64>)

declare i32 @llvm.experimental.vector.reduce.umax.v8i32(<8 x i32>)

@callables = internal local_unnamed_addr global i8** null, align 8

define void @set_callables(i8** %ptr) local_unnamed_addr #0 {
    store i8** %ptr, i8*** @callables
    ret void
}

declare <8 x float> @llvm.maxnum.v8f32(<8 x float>, <8 x float>)

declare <8 x float> @llvm.minnum.v8f32(<8 x float>, <8 x float>)

declare <8 x float> @llvm.ceil.v8f32(<8 x float>)

declare <8 x float> @llvm.floor.v8f32(<8 x float>)

define internal void @ek.scatter_add.v8f32(<8 x float*> %ptrs, <8 x float> %value, <8 x i1> %active_in) #0 {
L0:
   br label %L1

L1:
   %index = phi i32 [ 0, %L0 ], [ %index_next, %L3 ]
   %active = phi <8 x i1> [ %active_in, %L0 ], [ %active_next_2, %L3 ]
   %active_i = extractelement <8 x i1> %active, i32 %index
   br i1 %active_i, label %L2, label %L3

L2:
   %ptr_0 = extractelement <8 x float *> %ptrs, i32 %index
   %ptr_1 = insertelement <8 x float *> undef, float* %ptr_0, i32 0
   %ptr_2 = shufflevector <8 x float *> %ptr_1, <8 x float *> undef, <8 x i32> zeroinitializer
   %ptr_eq = icmp eq <8 x float *> %ptrs, %ptr_2
   %active_cur = and <8 x i1> %ptr_eq, %active
   %value_cur = select <8 x i1> %active_cur, <8 x float> %value, <8 x float> zeroinitializer
   %sum = call reassoc float @llvm.experimental.vector.reduce.v2.fadd.f32.v8f32(float 0.0, <8 x float> %value_cur)
   atomicrmw fadd float* %ptr_0, float %sum monotonic
   %active_next = xor <8 x i1> %active, %active_cur
   %active_red = call i1 @llvm.experimental.vector.reduce.or.v8i1(<8 x i1> %active_next)
   br i1 %active_red, label %L3, label %L4

L3:
   %active_next_2 = phi <8 x i1> [ %active, %L1 ], [ %active_next, %L2 ]
   %index_next = add nuw nsw i32 %index, 1
   %cond_2 = icmp eq i32 %index_next, 8
   br i1 %cond_2, label %L4, label %L1

L4:
   ret void
}

declare float @llvm.experimental.vector.reduce.v2.fadd.f32.v8f32(float, <8 x float>)

!0 = !{!0}
!1 = !{!1, !0}
!2 = !{!1}
!3 = !{!"llvm.loop.unroll.disable", !"llvm.loop.vectorize.enable", i1 0}

attributes #0 = { norecurse nounwind "frame-pointer"="none" "no-builtins" "no-stack-arg-probe" "target-cpu"="skylake" "target-features"="-vzeroupper,-avx512pf,-tsxldtrk,+cx16,+sahf,-tbm,-avx512ifma,-sha,+crc32,-fma4,-vpclmulqdq,+prfchw,+bmi2,-cldemote,+fsgsbase,-avx512bf16,-amx-tile,-uintr,-gfni,+popcnt,-ptwrite,+aes,-avx512bitalg,-movdiri,-widekl,+xsaves,-avx512er,-avxvnni,-avx512fp16,-avx512vnni,-amx-bf16,-avx512vpopcntdq,-pconfig,-clwb,-avx512f,+xsavec,-clzero,-pku,+mmx,-lwp,-rdpid,-xop,+rdseed,-waitpkg,-kl,-movdir64b,-sse4a,-avx512bw,+clflushopt,+xsave,-avx512vbmi2,+64bit,-avx512vl,-serialize,-hreset,+invpcid,-avx512cd,+avx,-vaes,-amx-int8,+cx8,+fma,-rtm,+bmi,-enqcmd,+rdrnd,-mwaitx,+sse4.1,+sse4.2,+avx2,+fxsr,-wbnoinvd,+sse,+lzcnt,+pclmul,-rdpru,+f16c,+ssse3,+sgx,-prefetchwt1,+cmov,-avx512vbmi,-shstk,+movbe,-avx512vp2intersect,+xsaveopt,-avx512dq,+sse2,+adx,+sse3" }

drjit_32c09595272bc74a586d2fe8f7e4726e:469:31: error: use of undefined value '%f253_final'
    %f253 = phi <8 x float> [ %f253_final, %l_236_tail ], [ %f220, %l_236_start ]
                              ^

Aborted (core dumped)