I want a function f : usize -> u32 such that: f(4) = 0b1111 f(8) = 0b11111111 f(22) = 0x3FFFFF f(32) = 0xFFFFFFFF The last one is the hardest case. The canonical c/c++ implementation would be fn setbits(x: usize) -> u32 { (1 << x) - 1 } This will overflow for x=32, although unsigned overfl…

Your godbolt links do not match your function. They take two arguments.

pub fn setbits(x: u8) -> u32 { u32::MAX >> (32 - x) } Rust Playground A browser interface to the Rust compiler to experiment with the language

Oh oops, although the assembly is still different if you remove the first parameter and replace it with 1

Ah, thanks I should have thought of that, thats perfect

That doesn't work for x==0 though, either panicking in debug or leaving it as MAX in release.

[image] IndianBoy42: This will overflow for x=32 , although unsigned overflow is defined such that this works, and signed integer is UB till C++20, but will most likely work correctly I believe even the unsigned case is undefined in all versions of C++. The following program prints either 0 …

This is harder than I thought Considering that the x86 instruction, afaik, offer the semantics needed for this function (bits slide of the end and shifting by >= the number of bits in an integer always gives 0) is it worth adding a function that does this? I don't know what the name would be but pe…

Hard to say. Rust also doesn't include methods for more-common bitwise things like "test if bit is set" in its stdlib. (Or most of the BMI1/BMI2 stuff.) So unless there's an LLVM intrinsic or a instruction for it on a common processor that's not otherwise pattern-matched at codegen by LLVM, my in…

[image] IndianBoy42: Considering that the x86 instruction, afaik, offer the semantics needed for this function Should every arcane aspect of every CPU architecture be reflected in special Rust functions? Or should Rust focus on supporting those architectural features that are commonly support…

[image] IndianBoy42: Considering that the x86 instruction, afaik, offer the semantics needed for this function (bits slide of the end and shifting by >= the number of bits in an integer always gives 0) is it worth adding a function that does this? Actually x86 does not have this semantics. Sh…

How to make an integer with N bits set without overflow

s3bk August 2, 2021, 2:22pm 3

pub fn setbits(x: u8) -> u32 {
    u32::MAX >> (32 - x)
}

Topic		Replies	Views
Apparent arithmetic bug help	21	1516	April 12, 2020
Bitshift and overflowing behavior help	2	3048	December 17, 2015
Why do the non-panicking variants of left bit-shift work the way they do? help	11	3303	April 2, 2023
Disable overflow check	14	2316	July 11, 2023
>> operator panics in debug build, works in release build help	14	2505	March 16, 2016

How to make an integer with N bits set without overflow

Related topics