linux-kernel - Re: [PATCH v2] x86/lib: Optimize num_digits() and fix INT

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <ee26a04e-7b08-4400-abda-7bf81a67d560@formalgen.com>
Date: Wed, 21 Jan 2026 00:32:58 +0100
From: David Desobry <david.desobry@...malgen.com>
To: David Laight <david.laight.linux@...il.com>
Cc: tglx@...nel.org, mingo@...hat.com, bp@...en8.de,
 dave.hansen@...ux.intel.com, x86@...nel.org, hpa@...or.com,
 linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2] x86/lib: Optimize num_digits() and fix INT_MIN
 overflow

On 1/20/26 22:46, David Laight wrote:

> 
> Maybe better to write as:
> 	if (val < 0) {
> 		d = 1;
> 		v = -val;
> 	} else {
> 		d = 0;
> 		v = val;
> 	}
> The compiler will only generate one jump.

Agreed, I've adopted this in v3.

> 
>>   
>> -	while (val >= m) {
>> -		m *= 10;
>> -		d++;
>> +	switch (v) {
>> +	case 0 ... 9:
>> +		return d + 1;
>> +	case 10 ... 99:
>> +		return d + 2;
>> +	case 100 ... 999:
>> +		return d + 3;
>> +	case 1000 ... 9999:
>> +		return d + 4;
>> +	case 10000 ... 99999:
>> +		return d + 5;
>> +	case 100000 ... 999999:
>> +		return d + 6;
>> +	case 1000000 ... 9999999:
>> +		return d + 7;
>> +	case 10000000 ... 99999999:
>> +		return d + 8;
>> +	case 100000000 ... 999999999:
>> +		return d + 9;
>> +	default:
>> +		return d + 10;
> 
> clang generates something really horrid for that.
> 
> Either:
>      if (v <= 9) return d + 1;
>      if (v <= 99) return d + 2;
>      if (v <= 999) return d + 3;
>      if (v <= 9999) return d + 4;
>      if (v <= 99999) return d + 5;
>      if (v <= 999999) return d + 6;
>      if (v <= 9999999) return d + 7;
>      if (v <= 99999999) return d + 8;
>      if (v <= 999999999) return d + 9;
>      return d + 10;
> or:
>      if ((++d && v > 9) &&
>          (++d && v > 99) &&
>          (++d && v > 999) &&
>          (++d && v > 9999) &&
>          (++d && v > 99999) &&
>          (++d && v > 999999) &&
>          (++d && v > 9999999) &&
>          (++d && v > 99999999) &&
>          (++d && v > 999999999))
>          d++;
>      return d;
> generate better code.
> In particular it is almost certainly best to only have one taken branch.
> Dumping in a load of unlikely() might help that as well.
> (Neither compiler does the ++d inline, the add is done before the return.)

Good catch. I have replaced the switch statement with a linear if-chain 
in v3 to ensure better code generation for both GCC and Clang.

David