[llvm-commits] [llvm] r147090 - /llvm/trunk/lib/CodeGen/MachineBlockPlacement.cpp

[Jakub Staszak]

I have to agree with operator[] vs lookup(). We don't get any benefit here, so there is no point to make code less readable. In this case I'd have to revert the whole patch, because most of these operator[] won't work because of its lack of constantness.

- Kuba

On Dec 21, 2011, at 11:11 PM, Chandler Carruth wrote:

> On Wed, Dec 21, 2011 at 2:00 PM, Jakub Staszak <kubastaszak at gmail.com> wrote:
> operator[] inserts an object for key if it doesn't exist.
> 
> // X is empty
> if (X[0] == a) { }
> if (X[1] == b) { }
> if (X[2] == c) { }
> // X.size() = 3
> 
> This is the way that std::map works. I believe we want to be quite compatible here.
> 
> Certainly, but is it actually cheaper if you know ahead of time that no such insertion will occur? I would expect the lookup path to a hot path through the [] operators.
> 
> To be honest, I'm beginning to question the entire patch. No insertion ever happens with these [] uses, and they seem more readable to me than lookup... I don't feel strongly about the lookup, but I do feel strongly about the const thing. I'd really rather you strip back out all the const here.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20111221/9d3c2c6f/attachment.html>