Workshop 4.2 -> 5.0 -> 6.0 slower and slower...

Workshop 4.2 -> 5.0 -> 6.0 slower and slower...

Post by Chris Chiap » Wed, 14 Jun 2000 04:00:00



Looked around on Dejanews and couldn't find anything about this...

We've been having problems with the abysmal slowness of the Sun compilers.
C in 4.2 and 5.0 isn't bad but C++ is slow enough that it makes Solaris
development extraordinarily painful in comparison to NT, on which M$'
compilers go quite a but faster.  Now, having demoed the new Workshop 6 (err
I mean Forte 6) compiler, the situation is looking even more grim.  Building
2 fairly straight forward C modules in debug mode:

      mod1  mod2
WS5   3.6s  1m26s        
WS6   7.6s  2m30s        

and     optimized:          
WS5   5.3s   2m6s                  
WS6    11s   4m2s      

Now, I can understand optimized compilation getting much slower, but
*debug*???  We're very keen to move to Workshop 6 because we've found
Workshop 5's de* to be extremely unstable, and the Workshop 6 debugging
interface is much nicer than 5's, but this continued slowdown isn't
encouraging us.  Does this match other people's experiences?  Any hints on
speeding up the process?

--

+------- ---  -- --  -


                                                    -  -- --  --- -------+

 
 
 

Workshop 4.2 -> 5.0 -> 6.0 slower and slower...

Post by Thomas Deh » Thu, 15 Jun 2000 04:00:00



> Looked around on Dejanews and couldn't find anything about this...

> We've been having problems with the abysmal slowness of the Sun compilers.
> C in 4.2 and 5.0 isn't bad but C++ is slow enough that it makes Solaris
> development extraordinarily painful in comparison to NT, on which M$'
> compilers go quite a but faster.
>  Now, having demoed the new Workshop 6 (err
> I mean Forte 6) compiler, the situation is
> looking even more grim. Building
> 2 fairly straight forward C modules in debug mode:

I am confused. Are you complaining about C++,
or about C?

Quote:>       mod1  mod2
> WS5   3.6s  1m26s
> WS6   7.6s  2m30s

> and optimized:
> WS5   5.3s   2m6s
> WS6    11s   4m2s

On my system, C++ 5.1 is about as fast as C++ 5.0.
Your testcases are unrealistically small, anyway.
What are your compiler options?

Quote:> Now, I can understand optimized compilation getting much slower, but
> *debug*???  We're very keen to move to
> Workshop 6 because we've found
> Workshop 5's de* to be
> extremely unstable, and the Workshop 6 debugging
> interface is much nicer than 5's, but this continued slowdown isn't
> encouraging us.  Does this match other people's experiences?  Any hints on
> speeding up the process?

Compile with +d.
Take care not to include unnecessary or unnecessarily
large header files. The template bloat required by
the "everything is in template form" ANSI/ISO standard is horrible.
Then, dbx 6.0 can debug code compiled with C++ 5.0 or
C++ 4.2.

Thomas

 
 
 

Workshop 4.2 -> 5.0 -> 6.0 slower and slower...

Post by Chris Chiap » Thu, 15 Jun 2000 04:00:00



>I am confused. Are you complaining about C++,
>or about C?

Well, C++ has always been a problem in 4.2 and 5.0, but with 6.0 even C has
gotten to the point of "irritatingly slow".

Quote:

>>       mod1  mod2
>> WS5   3.6s  1m26s
>> WS6   7.6s  2m30s

>> and optimized:
>> WS5   5.3s   2m6s
>> WS6    11s   4m2s

>On my system, C++ 5.1 is about as fast as C++ 5.0.

Hrmm?  What's 5.1?  

Quote:>Your testcases are unrealistically small, anyway.

Quibbling over test case size is silly, I simply chose these two as
representative...I'm seeing the ~2X slowdown across the board on all of our
source code.

Quote:>What are your compiler options?

The relevant ones would be
/opt/SUNWspro/WS6/bin/cc -mt -KPIC -xCC -g -v -c -o foo.o foo.c
about as vanilla as you can get...

Quote:>Compile with +d.

I believe this specifies dynamic linking, which is already the default,
correct?

Quote:>Take care not to include unnecessary or unnecessarily
>large header files. The template bloat required by
>the "everything is in template form" ANSI/ISO standard is horrible.
>Then, dbx 6.0 can debug code compiled with C++ 5.0 or
>C++ 4.2.

Yes, I'm aware of the generic "how to make compiles go faster" options.  The
point here isn't that our compiles take too long it's that moving to
workshop 6 nearly doubles our compile time.

--

+------- ---  -- --  -


                                                    -  -- --  --- -------+

 
 
 

Workshop 4.2 -> 5.0 -> 6.0 slower and slower...

Post by Thomas Deh » Sat, 17 Jun 2000 04:00:00




> On Wed, 14 Jun 2000 20:20:06 +0200, Thomas Dehn

> >I am confused. Are you complaining about C++,
> >or about C?
> Well, C++ has always been a problem in 4.2 and 5.0,
> but with 6.0 even C has
> gotten to the point of "irritatingly slow".

mmmh, that's new to me. Haven't compiled much C stuff recently.

Quote:> >>       mod1  mod2
> >> WS5   3.6s  1m26s
> >> WS6   7.6s  2m30s

> >> and optimized:
> >> WS5   5.3s   2m6s
> >> WS6    11s   4m2s

> >On my system, C++ 5.1 is about as fast as C++ 5.0.
> Hrmm?  What's 5.1?

The version number of the C++ compiler from WS6.

Quote:> >Your testcases are unrealistically small, anyway.
> Quibbling over test case size is silly, I simply chose these two as
> representative...I'm seeing the
> ~2X slowdown across the board on all of our
> source code.

> >What are your compiler options?
> The relevant ones would be
> /opt/SUNWspro/WS6/bin/cc -mt -KPIC -xCC -g -v -c -o foo.o foo.c
> about as vanilla as you can get...

What happens if you turn of -g?

Quote:> >Compile with +d.
> I believe this specifies dynamic linking,
> which is already the default,
> correct?

No, +d turns off inlining.
-Bdynamic would be dynamic linking.

Thomas

Sent via Deja.com http://www.deja.com/
Before you buy.

 
 
 

1. cable connect.slower&slower->dead

My cablemodem connection works for a few minutes, then it slows down
until it is completely gone. I am new to Linux so I have no idea why or
where the problem lies. Please help me.
I have a ppc 5500/225 with an ethernetcard from Asanth connected to my
cablemodem.
This works fine with MacOs
Sincerly
Fredrik

2. Comparing File contents:

3. 4.2->5.0->5.0.2->Comet ??

4. Grub to install from HD

5. Cvsup 4.2 >> to >> 4.5

6. Conditional ftp script?

7. PPP gets slower and slower

8. X-Free with ET4000/W32P

9. telnet logins running slower and slower...

10. Server getting slower and slower. Why?

11. logout gets slower and slower!

12. <><><> MOUNTING EXTENDED PARTITION <><><>

13. Wanted: <><><> Unix Specialist <><><>