[LLVMbugs] [PATCH] calculateFrameObjectOffsets

[Vladimir Prus]

Chris Lattner wrote:

> > I would not like to change it, especially since I haven't yet learned
> > howto run LLVM testsuite.
>
> I would suggest trying to run the LLVM test suite.  It should be pretty
> easy.  To try it out, go into:
>
> llvm/test/Programs/MultiSource/Benchmarks/Olden
>
> ... and type 'make'.  It should run through all of the programs in that
> directory and build them with the CBE, JIT, and LLC targets, run them all
> and make sure they work.

Ok, running now. What is relation between Makefile-based tests and QMTest? At 
least for me QMTest is more convenient? And also, what version of QMTest did 
you use?

> > What do you think, is this patch OK or I should try to remove as many of
> > conditionals as I can?
>
> I would prefer that you eliminated some of the obvious ones, like this
> one:
>
> +    if (StackGrowsDown) {
> +        if (FixedOff > Offset) Offset = FixedOff;
> +    } else {
> +        if (FixedOff > Offset) Offset = FixedOff;
> +    }

Ok, I've made a second pass through the function, tweaking things and adding 
comments and finally I like the result.

The patch also makes X86 and PowerPC backends to pass -4 as local area offset 
instead of 4 (not matter that PowerPC is not finished). The 
calculateFrameObjectOffsets is accordingly modified.


> If you send and updated patch, I would be happy to apply it for you.  For
> patches, please send them to the llvmbugs list:
> http://mail.cs.uiuc.edu/mailman/listinfo/llvmbugs

Ok.

- Volodya
-------------- next part --------------
A non-text attachment was scrubbed...
Name: PrologueEpilogue.diff
Type: text/x-diff
Size: 5173 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20040609/8f8fa25b/attachment.diff>