IL simplification does not take advantage of distributive property

For the following source code,

struct AllocStruct {
    void * padding[0x30];
    char * buffers[0x10];
    int sizes[0x10];
};

int main() {

    struct AllocStruct* a = malloc(sizeof(struct AllocStruct));

    for (int i = 0; i < 0x10; i++) {
        a->buffers[i] = malloc(0x18);
        a->sizes[i] = 0x20;
    }
    return 0;
}

a.out.zip

When it gets compiled with gcc -O0 test.c on Ubuntu 22.04, the HLIL looks like this:

The IL simplification does not take advantage that

rax_5->padding[sx.q(i) + 0x30] = malloc(bytes: 0x18)

is equivalent to

rax_5->buffers[sx.q(i)] = malloc(bytes: 0x18)

, since padding[0x30] is the size of the padding member.

Start with the supplied test.c example and compile it with gcc -O0 on Ubuntu 22.04, then inspect the resulting HLIL and the a.out.zip reproduction. Trace the IL simplification responsible for the padding and buffers indexing, and verify that the distributive equivalence is recognized without changing other expressions.