[bug] llvm ICmpInst assert fail when using if statement with non 32bit int value

[Dinngger]

Describe the bug
llvm ICmpInst assert fail when using if statement with non 32bit value.

To Reproduce

import taichi as ti

ti.init(arch=ti.gpu)

@ti.kernel
def my_test(x: ti.u8): # any non 32bit int type.
    if x:
        print(1)

my_test(1)

Log/Screenshots

$ python3 ti_icmp.py 
[Taichi] version 1.6.0, llvm 15.0.4, commit f1c6fbbd, linux, python 3.8.10
[Taichi] Starting on arch=cuda
python3: /home/dev/.cache/ti-build-cache/llvm15/include/llvm/IR/Instructions.h:1191: void llvm::ICmpInst::AssertOK(): Assertion `getOperand(0)->getType() == getOperand(1)->getType() && "Both operands to ICmp instruction are not of the same type!"' failed.
Aborted (core dumped)

Additional comments
Both cpu and cuda backend raise this error. Vunkan works will with this kernel.

[lin-hitonami]

It seems that this bug is already fixed in the latest master. You can install the nightly build (by pip install -i https://pypi.taichi.graphics/simple/ taichi-nightly) or wait for the next release. If the problem persists even after installing the nightly build, please contact us again.