Files
c3c/test/test_suite/builtins/shufflevector.c3t

90 lines
2.4 KiB
C

// #target: macos-x64
module test;
import std::io;
struct Matrix2x2
{
union
{
struct
{
float m00, m01;
float m10, m11;
}
struct
{
float[<2>] m0;
float[<2>] m1;
}
float[<4>] m;
}
}
fn float[<2>] apply1(Matrix2x2* mat, float[<2>] vec)
{
return float[<2>] {
mat.m00 * vec[0] + mat.m01 * vec[1],
mat.m10 * vec[0] + mat.m11 * vec[1],
};
}
fn float[<2>] apply2(Matrix2x2* mat, float[<2>] vec)
{
return float[<2>] {
mat.m0[0] * vec[0] + mat.m0[1] * vec[1],
mat.m1[0] * vec[0] + mat.m1[1] * vec[1],
};
}
fn float[<2>] apply3(Matrix2x2* mat, float[<2>] vec)
{
float[<2>] a = $$swizzle2(mat.m0, mat.m1, 0, 3);
float[<2>] b = $$swizzle2(mat.m0, mat.m1, 1, 2);
float[<2>] flip = $$swizzle(vec, 1, 0);
return a * vec + b * flip;
}
fn void main()
{
Matrix2x2 a = { 1, -3, 5, -7 };
io::printfn("1: %s", apply1(&a, float[<2>] { 11, 13 }));
io::printfn("2: %s", apply2(&a, float[<2>] { 11, 13 }));
io::printfn("3: %s", apply3(&a, float[<2>] { 11, 13 }));
}
/* #expect: test.ll
define double @test.apply3(ptr %0, double %1) #0 {
entry:
%vec = alloca <2 x float>, align 8
%a = alloca <2 x float>, align 8
%b = alloca <2 x float>, align 8
%flip = alloca <2 x float>, align 8
%taddr = alloca <2 x float>, align 8
store double %1, ptr %vec, align 8
%2 = load <2 x float>, ptr %0, align 16
%ptradd = getelementptr inbounds i8, ptr %0, i64 8
%3 = load <2 x float>, ptr %ptradd, align 8
%shuffle = shufflevector <2 x float> %2, <2 x float> %3, <2 x i32> <i32 0, i32 3>
store <2 x float> %shuffle, ptr %a, align 8
%4 = load <2 x float>, ptr %0, align 16
%ptradd1 = getelementptr inbounds i8, ptr %0, i64 8
%5 = load <2 x float>, ptr %ptradd1, align 8
%shuffle2 = shufflevector <2 x float> %4, <2 x float> %5, <2 x i32> <i32 1, i32 2>
store <2 x float> %shuffle2, ptr %b, align 8
%6 = load <2 x float>, ptr %vec, align 8
%shuffle3 = shufflevector <2 x float> %6, <2 x float> poison, <2 x i32> <i32 1, i32 0>
store <2 x float> %shuffle3, ptr %flip, align 8
%7 = load <2 x float>, ptr %a, align 8
%8 = load <2 x float>, ptr %vec, align 8
%fmul = fmul <2 x float> %7, %8
%9 = load <2 x float>, ptr %b, align 8
%10 = load <2 x float>, ptr %flip, align 8
%fmul4 = fmul <2 x float> %9, %10
%fadd = fadd <2 x float> %fmul, %fmul4
store <2 x float> %fadd, ptr %taddr, align 8
%11 = load double, ptr %taddr, align 8
ret double %11
}