Proper way to load structs in LLVM - llvm

Question

I have been experimenting with MCJIT in LLVM and have come across a problem when trying to load struct types.
The relevant struct I am trying to load can be seen here:
typedef struct {
float x;
float y;
} cfloat;
The generated LLVM code can be seen below.
--LLVM IR--
; ModuleID = 'Test_Module'
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
%cfloat = type { float, float }
declare float #cabs(%cfloat)
declare float #cabsptr(%cfloat*)
define i32 #test(i32 %argc, i8** %args) {
entry:
%argloc0 = getelementptr i8*, i8** %args, i64 0
%argloc1 = getelementptr i8*, i8** %args, i64 1
%arg0 = load i8*, i8** %argloc0
%iptr = bitcast i8* %arg0 to %cfloat*
%arg1 = load i8*, i8** %argloc1
%optr = bitcast i8* %arg1 to float*
%ival = load %cfloat, %cfloat* %iptr
%oval = call float #cabs(%cfloat %ival)
%oval2 = call float #cabsptr(%cfloat* %iptr)
store float %oval, float* %optr
ret i32 %argc
}
The relevant structs and functions:
extern "C" float cabs(cfloat val)
{
printf("%f, %f\n", val.x, val.y);
return sqrtf(val.x * val.x + val.y * val.y);
}
extern "C" float cabsptr(cfloat *val)
{
printf("%f, %f\n", val->x, val->y);
return sqrtf(val->x * val->x + val->y * val->y);
}
When I pass in cfloat val = {300, -400}; to the function
cabs prints 300.0000, 0.0000.
cabsptr prints 300.0000, -400.0000.
I'd expect both the functions to print the input values.
Is this a bug I am encountering or did I miss something when generating the LLVM function?
The full source code to reproduce this issue can be seen here:
https://gist.github.com/pavanky/4439f1a6c2ca182058c19187aa8215a7
EDIT:
Defining my custom type to be a VectorType of <2 x float> seems to be a work around for me. But I'd like to know a proper solution if available.