Re: several messages

[Gerald Pfeifer <pfeifer at dbai dot tuwien dot ac dot at>]

On Tue, 10 Jun 2003, Theodore Papadopoulo wrote:
>> This indicates that the SPEC tester is not sufficient wrt. to
>> compile-time regressions.
> Let me make a guess:
>
>     - Gerald, your test code is C++.
>     - SPEC is mostly (if not all) C ?

Perfect guess!

> I bet that most of the slowdown reported is for C++, whereas most of
> the new code introduced in the compiled is tested for not introducing
> slowdowns against C code (ie SPEC+gcc itself).

That's may guess, indeed, else somebody else should have noticed a
slowdown by a factor of two.

> Perhaps, (if the above guess is valid), it is the proper time to add
> some specific C++ compile-time tests, and to start to measure the
> slowdowns also for C++.

I couldn't agree more, especially since I originally contributed the code
that's now in PR8361 years ago.  Diego, would you mind adding that?

On Tue, 10 Jun 2003, Andrew MacLeod wrote:
> FWIW, The very limited amount of compile time experience Ive got (an
> early cut at reducing compile time for out-of-SSA) indicated that
> tree-ssa requires more memory. If you have a smaller memory machine, its
> compile times shoot up quite a bit due to swapping. On machines where it
> can compile without swapping, compiles times are not so bad.

I can guarantee that the machine I used was not swapping.

Having under my desktop at my (former) work place, I would have (a)
heard the disk, and (b)  noticed that while reading mail in Pine (it's
a dual-PIII/1000, so that shouldn't have interfered much otherweise).

Gerald