Global array and global index

[blackgeorge-boom]

#include <stdio.h>
#include <stdlib.h>


int nx[6];
int ny[6];
int nz[6];
int lt;

int main()
{
    FILE *fp;

    fscanf(fp, "%d%d", &nx[lt], &ny[lt]);
}

[blackgeorge-boom]

The first important difference is that AArch64 spills the value of lt.

AArch64:

32B	  %1:gpr64temp = ADRP target-flags(aarch64-page) @lt, debug-location !30; main.c:14:28
48B	  %14:gpr64 = LDRSWui killed %1:gpr64temp, target-flags(aarch64-pageoff, aarch64-nc) @lt, debug-location !30 :: (dereferenceable load 4 from @lt); main.c:14:28
56B	  STRXui %14:gpr64, %stack.1, 0 :: (store 8 into %stack.1)

X86:

32B	  %1:gr64_nosp = MOVSX64rm32 $rip, 1, $noreg, @lt, $noreg, debug-location !30 :: (dereferenceable load 4 from @lt); main.c:14:28

This introduces maybe the other changes to the register allocator.

AArch64:

• fp, lt <-> spilled
• nx + lt <-> x20
• ny + lt <-> x19

X86:

• nx +lt, fp <-> spilled
• lt <-> r15
• ny + lt <-> rbx