taichi: If condition must be int32 error

To reproduce:

import taichi as ti
ti.init(default_ip=ti.i64)

@ti.kernel
def test():
    for i in range(100):
        if i % 2 == 0:
            print(i)

test()

The above code reports error

taichi.lang.exception.TaichiTypeError: 
`if` conditions must be of type int32; found i64. Consider using `if x != 0` instead of `if x` for float values.

About this issue

Original URL
State: closed
Created 2 years ago
Comments: 16 (6 by maintainers)

Most upvoted comments

After I applied this modification,I can use four int64 to simulate float256 and draw this in 10^-55 precision 😃 Re:-0.6717536270481832370389902348813046305959745988723619134343491834092885898911496280 Im:0.4609205425545165168733304746281937721035176497122051662619296867035239619125442493 zoom:2.28E54 frame_00372 Thanks again for your efficient bug fix.

SuperWildFireFox on Jul 26, 2022

Here are the basic code: https://github.com/SuperWildFireFox/MandelbrotZoom/blob/main/FLOAT/FLOAT256.py

SuperWildFireFox on Jul 26, 2022

Maybe one or two days later, I will release my code.It is just an experimental program, but the efficiency is satisfactory to me.

SuperWildFireFox on Jul 26, 2022

I also tried this:

import taichi as ti
ti.init(default_ip=ti.i64)


@ti.kernel
def test():
    a = 20
    while ti.cast(a > 0, ti.i32):
        a -= 1


test()

still output: while conditions must be of type int32; found i64. Consider using while x != 0 instead of while x for float values.

SuperWildFireFox on Jul 24, 2022

and thank you for your quick reply.

SuperWildFireFox on Jul 24, 2022

import taichi as ti

ti.init(default_ip=ti.i64)


@ti.kernel
def test():
    a = 20
    while a > 0:
        a -= 1


test()

output:while conditions must be of type int32; found i64. Consider using while x != 0 instead of while x for float values.

SuperWildFireFox on Jul 24, 2022

it will still happened in while loop

SuperWildFireFox on Jul 24, 2022

Currently we use i32 to represent boolean and it is self-contained. The bug in this issue is due to wrong implementation, and can be easily fixed by #5487. i1 is only a nice-to-have for now, and we may plan it in some future version like v1.3 or so.

strongoier on Jul 21, 2022