[PATCH] D14547: [llvm-profdata] Add support for weighted merge of profile data

[Diego Novillo via llvm-commits]

On Tue, Nov 10, 2015 at 3:07 PM, David Li <davidxl at google.com> wrote:

>
> ================
> Comment at: test/tools/llvm-profdata/weight-sample.test:42
> @@ +41,3 @@
> +3- Bad merge: foo and bar profiles with too few weight values
> +RUN: not llvm-profdata merge --sample --text -weights=2
> %p/Inputs/weight-sample-bar.proftext %p/Inputs/weight-sample-foo.proftext
> -o %t.out 2>&1 | FileCheck %s --check-prefix=ERROR1
> +ERROR1: error: Profile weight must be specified for each input file.
> ----------------
> No need for such tests if the weight is always attached to the input file.
>

But then what happens if one of the files is specified without ":WEIGHT"?
Should we complain about it?  Should we assume MAX(other weights)?

I'm not sure I'm completely convinced that file1:W1 file2:W2 is better than
--weights=W1,W2.  What if ':' is part of the filename?  I know, seems like
a stretch.  I'm not actually arguing against it.  I mostly find it a bit
odd.


Diego.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20151110/dc45993d/attachment.html>