[PATCH] D142556: [DebugInfo] Merge partially matching chains of inlined locations

Mon Feb 13 13:16:41 PST 2023

dblaikie added a comment.

In D142556#4119818 <https://reviews.llvm.org/D142556#4119818>, @dstenb wrote:

> In D142556#4119696 <https://reviews.llvm.org/D142556#4119696>, @dblaikie wrote:
>
>> I'm a bit confused by all these reports - at least with my `bloaty` locally, "+100%" would mean the section doubled in size (100% growth) but it shows "+100%" even for a section that shrunk (.debug_line_str) slightly... Confusing.
>>
>> but the overall numbers still sound OK.
>
> Yes, I thought that the output format looked odd. That was with `bloaty` built from main (52948c), which we can compare to v1.1 (for a small example binary):
>
>   $ diff -u <(./bloaty-1.1 with.out -- without.out) <(./bloaty-52948c1 with.out -- without.out)
>   --- /dev/fd/63	2023-02-11 01:25:09.785684383 +0100
>   +++ /dev/fd/62	2023-02-11 01:25:09.785684383 +0100
>   @@ -1,8 +1,8 @@
>        FILE SIZE        VM SIZE    
>     --------------  -------------- 
>   -   +11%     +26  [ = ]       0    .debug_info
>   -  +6.5%     +15  [ = ]       0    .debug_abbrev
>   -  +6.7%      +8  [ = ]       0    .debug_addr
>   -  +0.1%      +6  [ = ]       0    [Unmapped]
>   -  +0.5%      +1  [ = ]       0    .debug_line
>   -  +0.3%     +56  [ = ]       0    TOTAL
>   +  +115%     +26  [ = ]       0    .debug_info
>   +  +109%     +15  [ = ]       0    .debug_abbrev
>   +  +114%      +8  [ = ]       0    .debug_addr
>   +  +111%      +6  [ = ]       0    .debug_str_offsets
>   +  +101%      +1  [ = ]       0    .debug_line
>   +  +100% +17.1Ki  +100% +2.61Ki    TOTAL
>
> Although they find the same number of bytes, v1.1 reports an +11% increase for .debug_info for this example compared to +115% when using bloaty from main. Perhaps it is broken on main?
>
> I don't have access to run v1.1 on the binaries in question now, but if we look at the ratios for the `"#bytes"` diff lines from llvm-dwarfdump in one of my earlier comments we can see that the ratios are smaller than how I interpreted bloaty's output, similar to this small example.

*nod* All a bit confusing, unfortunately - but I /think/ the numbers sound OK.

compile-time-tracker results?

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D142556/new/

https://reviews.llvm.org/D142556