Matrix gives llvm error when accessed with non-64-bit int

[Up05]

Context

• Operating System & Odin Version:

Odin:   dev-2024-08-nightly:8359995
        OS:      CachyOS, Linux 6.10.6-2-cachyos
        CPU:     Quad-Core Processor (up to 1.4GHz)
        RAM:     3354 MiB
        Backend: LLVM 18.1.8

Expected Behavior

The matrix's element is returned

Current Behavior

LLVM CODE GEN FAILED FOR PROCEDURE: main.main
define internal void @main.main(ptr noalias nocapture nonnull %__.context_ptr) {
; [..]
entry:                                            ; preds = %decls
  store i32 1, ptr %_i32, align 4
  %3 = load i32, ptr %_i32, align 4
  %4 = load i32, ptr %_i32, align 4
  %5 = sext i32 %3 to i64
  %6 = sext i32 %4 to i64
  ; [..]

Use of instruction is not an instruction!
  %6 = add i64 %3, %5

Failure Information (for bugs)

It seems to me, that it is assumed, that matrix will only ever be accessed with an i64. (See: last line of LLVM error, i64)

Steps to Reproduce

mat : matrix[4, 4] i64 // whatever* sub-type

main :: proc() {
    _i32 : i32 = 1
    c := mat[_i32, _i32]
    fmt.println(c)
}   

Failure Logs

The rest of LLVM code seems unnecessary.

[pfiadDi]

(EDIT: works when m is not global)

Get the also a llvm error
Odin: dev-2024-08:d41028176
OS: Ubuntu 22.04.4 LTS, Linux 6.8.0-40-generic
CPU: 11th Gen Intel(R) Core(TM) i5-1135G7 @ 2.40GHz
RAM: 39897 MiB
Backend: LLVM 14.0.0

with this code:

package main

import "core:fmt"
import "core:math/rand"
import "core:strings"

l := [dynamic]i64{2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 7, 7, 7, 7, 7}

m: matrix[4, 4]i64
nL: i64 = 0
c: i64 = 0
main :: proc() {
	i: i64 = 0
	for 0 < 16 {
		nL = choose_number()
		switch i {
		case 0 ..< 4:
			m[0, c] = nL
		case 4 ..< 8:
			c = i - 4
			m[1, c] = nL
		case 8 ..< 12:
			c = i - 8
			m[2, c] = nL
		case 12 ..< 16:
			c = i - 12
			m[3, c] = nL
		}
	}

	fmt.println(m)
}

choose_number :: proc() -> i64 {
		if len(l) == 1 {return l[0]}
	randIndex := rand.int31_max(cast(i32)len(l))
	newLetter := l[randIndex]
	unordered_remove(&l, cast(int)randIndex)
	return newLetter
}

Here' the beginning of the error:
LLVM CODE GEN FAILED FOR PROCEDURE: main.main
define internal void @main.main(i8* noalias nocapture nonnull %__.context_ptr) {
decls:
%i = alloca i64, align 8
%0 = alloca { %..any*, i64 }, align 8
%1 = alloca [16 x i8], align 16
%2 = alloca %..any, align 8
br label %entry

entry: ; preds = %decls
%3 = bitcast i8* %__.context_ptr to %runtime.Context*
store i64 0, i64* %i, align 8
br label %for.loop

for.loop: ; preds = %switch.done, %entry
br i1 true, label %for.body, label %for.done

for.body: ; preds = %for.loop
%4 = bitcast %runtime.Context* %3 to i8*
%5 = call i64 @main.choose_number(i8* %4)
store i64 %5, i64* @main.nL, align 8
%6 = load i64, i64* %i, align 8
%7 = icmp sle i64 0, %6
%8 = icmp slt i64 %6, 4
%9 = zext i1 %7 to i8
%10 = zext i1 %8 to i8
%11 = and i8 %9, %10
%12 = trunc i8 %11 to i1
br i1 %12, label %switch.case.body, label %switch.case.next

...

[FourteenBrush]

This only seems to be caused when the matrix is a global variable, however, the following code runs fine:

package bug

mat: matrix[4, 4]i64

main :: proc() {
    c := mat[i32(0), i32(0)]
}