[cfe-commits] r59094 - in /cfe/trunk/lib/CodeGen: CGBuilder.h CodeGenFunction.h

[Daniel Dunbar]

In my opinion this really isn't a major issue. If we want to fix it, I
would advocate changing the IRBuilder to allow run time configuration
of that option (which I would like, but isn't a high priority).

In the end, it sometimes makes sense to have some static compilation
time choice between easier debugging/development and performance.
NDEBUG seems a reasonable choice for that option. Note that we use
this in other places in LLVM to not compile in debugging specific
code.

On Tue, Nov 11, 2008 at 8:50 PM, Chris Lattner <clattner at apple.com> wrote:
>
> On Nov 11, 2008, at 6:30 PM, Eli Friedman wrote:
>
>> On Tue, Nov 11, 2008 at 4:01 PM, Daniel Dunbar <daniel at zuster.org> wrote:
>>>
>>> Author: ddunbar
>>> Date: Tue Nov 11 18:01:12 2008
>>> New Revision: 59094
>>>
>>> URL: http://llvm.org/viewvc/llvm-project?rev=59094&view=rev
>>> Log:
>>> Disable generation of basic block names in NDEBUG mode.
>>>
>>> Revert to enabling generation of instruction names when not in NDEBUG
>>> mode.
>>
>> Is this really a good idea?  Normally NDEBUG doesn't change the
>> visible behavior of the program.
>
> This also breaks abi compatibility between .o files produced in debug and
> release builds.  I don't know how much that matters in practice.
>
> What case are instruction names useful for?

The code is more readable with names, in my opinion, esp. when wanting
to relate the output code with the part of IRGen that emitted it.
It is also substantially more editable.

I think the burden of proof goes the other direction, in what cases
are names not useful? :)

The only real answer is performance, but this doesn't motivate
ditching them altogether in my opinion.

 - Daniel