pro64-support
[Top] [All Lists]

Re: Problem Collecting Trace Info

To: mpm@xxxxxxxxxxxxxxxxx (Michael Murphy)
Subject: Re: Problem Collecting Trace Info
From: Charles Hardnett <hardnett@xxxxxxxxxxxxx>
Date: Wed, 26 Jul 2000 09:44:50 -0400 (EDT)
Cc: pro64-support@xxxxxxxxxxx
In-reply-to: <200007252150.OAA53712@xxxxxxxxxxxxxxxxx>
References: <200007252150.OAA53712@xxxxxxxxxxxxxxxxx>
Reply-to: hardnett@xxxxxxxxxxxxx
Sender: owner-pro64-support@xxxxxxxxxxx
Thanks for all of the replies. I got help yesterday from Nenad at
SGI, and everything seems to be working pretty good. 

Very nice tracing system.

Take Care,

Charles

Michael Murphy writes:
 >      From: Charles Hardnett <hardnett@xxxxxxxxxxxxx>
 >      
 >      Hi,
 >      
 >      I have been trying to get the trace information out of the BE of the
 >      SGIcc compiler. I am using the following command:
 >      
 >       sgicc -Wb,-tt30 -clist -O3 -v test1.c   
 >      
 >      The contents of the test1.t file:
 >      
 >      !!! DevWarn during Loop Nest Optimizer: Count limit reached on the 
 > following DevWarn:
 >      !!! DevWarn during Loop Nest Optimizer: TODO: Tune LNO machine model 
 > parameters
 >      for IA-64 
 >      
 >      1. How can i tune the LNO parms to make this work under the NUE
 >      environment?
 >      
 > Ignore this devwarn.  LNO does work, this was just a note to
 > do more testing of the parameters we use.
 > 
 >      2. Is there documentation on the various outputs of the trace options?
 >      
 > Unfortunately we don't provide any outside of the source.
 > That's something we should include in the next release (IMO).
 > The format is -Wb,-tt<phase>:<num>
 > where phase is a number or name (see common/util/tracing.{h,c}).
 > E.g. -tt30 and -ttvdd are the same.
 > The trace numbers are what is used in the Get_Trace calls, e.g.
 > -ttexp:2 will cause (in the compiler):
 >      Get_Trace (TP_CGEXP, 2)
 > to return true.
 > There is also -Wb,-tr<phase> which will dump the ir after the phase,
 > and -Wb,-ts<phase> to dump the symbol table after the phase (some
 > phases ignore this).
 >      
 > -- Mike Murphy
 > -- mpm@xxxxxxx
 > -- quote of the day:
 > --  "We can have knowledge and not have wisdom.
 > --   But we cannot have wisdom without knowledge." (R.C. Sproul)
 > 

-- 
--------------------------------------------------------------------
Charles R. Hardnett www.spelman.edu/~hardnett 

<Prev in Thread] Current Thread [Next in Thread>