LLVM ERROR: Unable to expand fixed point multiplication.

[sukkomon]

Expected Behavior:

The program should compile successfully and allow the multiplication of two Fixed16_16 fixed-point numbers.

Actual Behavior:

Compilation fails with the following error:
LLVM ERROR: Unable to expand fixed point multiplication.

System Information:

Odin: dev-2024-11:e6475fec4
OS: macOS Sonoma 14.6.1 (build: 23G93, kernel: 23.6.0)
CPU: Apple M3 Pro
RAM: 18432 MiB
Backend: LLVM 18.1.8

Code to Reproduce:

import fixed "core:math/fixed"

main :: proc() {
    x := fixed.Fixed16_16{}
    y := fixed.Fixed16_16{}
    product := fixed.mul(x, y)
}

Additional Notes:

This issue seems related to the use of intrinsics.fixed_point_mul in the fixed.mul procedure within core:math/fixed. The backend reports LLVM 18.1.8, which might lack support for this intrinsic or may require implementation in Odin's backend.

[Kelimion]

I'm unable to replicate it on Win 10 Pro with LLVM 18.1.8 with the latest commit. This may be peculiar to macOS's LLVM, even though it's the same version.

package main

import "core:fmt"
import "core:math/fixed"

main :: proc() {
	x := fixed.Fixed16_16{}
	y := fixed.Fixed16_16{}
	product := fixed.mul(x, y)

	fmt.println(product) // The program compiles and prints `Fixed16_16{i = 0}`
}