Most changes to Python require a NEWS entry.

Please add it using the blurb_it web app or the blurb command-line tool.

@oda-gitso Did you benchmark whether this is faster? If so, can you show the results.

@eendebakpt I did not benchmark the differences. What I did was look at the assembly for a simple program such as

#include <stdio.h>

int 
main(void) 
{
    int i = 0;

    for (i = 0; i < 3; i++)
        printf("%d ", i);
    printf("\n");

    for (i = 3; i > 0; i--)
        printf("%d ", i);
    printf("\n");

    return 0;
}

versus one where the second for begins with for (; i > 0; i--).

On my system (Ubuntu 20.04 LTS, x86-64), the assembly of both programs is identical except for an additional instruction for the case that i = 3 is present. Here is a snippet with i = 3 versus without. The rest of the assembly, not shown, is identical across both versions.

.L2:                        .L2:
    cmpl    $2, -4(%rbp)        cmpl    $2, -4(%rbp)   
    jle .L3                     jle .L3
    movl    $10, %edi           movl    $10, %edi
    call    putchar@PLT         call    putchar@PLT
    movl    $3, -4(%rbp)        jmp .L4
    jmp .L4

I do not think a benchmark is necessary in this instance. It is very clear to me that removing i = wordshift can only make it more efficient. It definitely won't make it any less.

By the way, I do not think I am able to add the skip news and skip issue labels. Are you able to?

@mdickinson @vstinner It says you both have touched this function in the past. I was wondering if you would be able to review this change? Thanks!

The policy stated in the developer's guide is:

Almost all changes made to the code base deserve an entry in Misc/NEWS.d. [...] Changes that affect documentation only generally do not require a NEWS entry.

On that basis, this change probably should have a news entry (though I agree it doesn't need an issue). But honestly, I'd expect the change to make no observable difference with an optimised build.

What compiler flags did you use when doing your assembly comparison?

Most changes to Python require a NEWS entry.

Please add it using the blurb_it web app or the blurb command-line tool.

Most changes to Python require a NEWS entry.

Please add it using the blurb_it web app or the blurb command-line tool.

@mdickinson @vstinner Thanks to you both for getting back to me.

I have just added a news entry.

Could you also add the skip issue label?

But honestly, I'd expect the change to make no observable difference with an optimised build.

I agree. There might be some behind-the-scenes optimization that I do not know about. However, my thinking is that (1) it will not make it slower (2) it may (depending on system, compiler, flags, etc.) make it faster.

What compiler flags did you use when doing your assembly comparison?

gcc -S -o some_name.s some_name.c.

Thanks for your time!

Is there a go-to issue number for news entries that don't have an issue?

Hmm, interesting; looks like it's hard to have news but no issue. (The same was true in the bad old Roundup days, but there it was clearer that everything needed an issue.) On balance, we could probably get away without a news entry here.

@mdickinson As a workaround, I just used the pull request number as the issue number as well. I think this should be okay.

On balance, we could probably get away without a news entry here.

@mdickinson Just letting you know that I have opted to delete the news entry, as suggested by Victor. I hope this is okay with you too! Thanks!

Merged. Thank you for the contribution, @oda-gitso!

Thanks a lot @mdickinson, @vstinner!