public inbox for bfd@sourceware.org
 help / color / mirror / Atom feed
* Re: patch was: dwarf2 debug dependent upon GAS?
       [not found]           ` <19990302002736.G27552@rjlhome.sco.com>
@ 1999-03-11  0:02             ` Jason Merrill
  1999-03-11  7:41               ` Ian Lance Taylor
  0 siblings, 1 reply; 6+ messages in thread
From: Jason Merrill @ 1999-03-11  0:02 UTC (permalink / raw)
  To: Robert Lipe; +Cc: bfd

>>>>> Robert Lipe <robertlipe@usa.net> writes:

[ context: SCO as doesn't support referring to the section name as a
symbol, so we're changing the dwarf2 code in gcc to spit out a label to use
instead. ]

 >>  >> shortening the label names, so as to use less space in the symbol
 >>  >> table.
 >> 
 >>  > They're locals, so they don't go in the symbol table, right?
 >>  > They certainly don't appear in nm or even a 'strings -a' of the
 >>  > resulting dot-o.
 >> 
 >> Don't they?  I thought they would, since we are creating relocs
 >> against them.

 > Uuuh, no.  That's how we started this discussion.  The section name
 > was treated as a local label for the beginning of the section in this
 > compilation unit.   Therefore, they're completely local and don't need
 > to appear in the symbol table.

This is irrelevant to your patch, but on Linux the local symbols do show up
in the symbol table, as I remembered; I always filter nm output as a
result.  This may be a gas thing.  Any gas folk care to comment?

Jason

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: patch was: dwarf2 debug dependent upon GAS?
  1999-03-11  0:02             ` patch was: dwarf2 debug dependent upon GAS? Jason Merrill
@ 1999-03-11  7:41               ` Ian Lance Taylor
  1999-03-11 10:31                 ` Jason Merrill
  0 siblings, 1 reply; 6+ messages in thread
From: Ian Lance Taylor @ 1999-03-11  7:41 UTC (permalink / raw)
  To: jason; +Cc: robertlipe, bfd

   From: Jason Merrill <jason@cygnus.com>
   Date: 11 Mar 1999 00:02:35 -0800

   This is irrelevant to your patch, but on Linux the local symbols do show up
   in the symbol table, as I remembered; I always filter nm output as a
   result.  This may be a gas thing.  Any gas folk care to comment?

Yes, there is some bad interaction going on, I believe related to
debugging references to local symbols in non-standard sections.  The
assembler is not converting them into relocs against a section symbol
for some reason.  I've never gotten around to looking into it.

Ian

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: patch was: dwarf2 debug dependent upon GAS?
  1999-03-11  7:41               ` Ian Lance Taylor
@ 1999-03-11 10:31                 ` Jason Merrill
  1999-03-11 10:44                   ` Ian Lance Taylor
  0 siblings, 1 reply; 6+ messages in thread
From: Jason Merrill @ 1999-03-11 10:31 UTC (permalink / raw)
  To: Ian Lance Taylor; +Cc: robertlipe, bfd

>>>>> Ian Lance Taylor <ian@cygnus.com> writes:

 >    From: Jason Merrill <jason@cygnus.com>
 >    Date: 11 Mar 1999 00:02:35 -0800

 >    This is irrelevant to your patch, but on Linux the local symbols do
 >    show up in the symbol table, as I remembered; I always filter nm
 >    output as a result.  This may be a gas thing.  Any gas folk care to
 >    comment?

 > Yes, there is some bad interaction going on, I believe related to
 > debugging references to local symbols in non-standard sections.  The
 > assembler is not converting them into relocs against a section symbol
 > for some reason.  I've never gotten around to looking into it.

I assume that fixing this would dramatically improve link time for DWARF2?

Jason

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: patch was: dwarf2 debug dependent upon GAS?
  1999-03-11 10:31                 ` Jason Merrill
@ 1999-03-11 10:44                   ` Ian Lance Taylor
  1999-03-11 11:20                     ` Jason Merrill
  0 siblings, 1 reply; 6+ messages in thread
From: Ian Lance Taylor @ 1999-03-11 10:44 UTC (permalink / raw)
  To: jason; +Cc: robertlipe, bfd

   From: Jason Merrill <jason@cygnus.com>
   Date: 11 Mar 1999 10:31:35 -0800

    >    This is irrelevant to your patch, but on Linux the local symbols do
    >    show up in the symbol table, as I remembered; I always filter nm
    >    output as a result.  This may be a gas thing.  Any gas folk care to
    >    comment?

    > Yes, there is some bad interaction going on, I believe related to
    > debugging references to local symbols in non-standard sections.  The
    > assembler is not converting them into relocs against a section symbol
    > for some reason.  I've never gotten around to looking into it.

   I assume that fixing this would dramatically improve link time for DWARF2?

I would not expect to see a dramatic improvement, but maybe there is
something I am overlooking.  You can probably estimate the improvement
by linking with the --discard-locals option.

Ian

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: patch was: dwarf2 debug dependent upon GAS?
  1999-03-11 10:44                   ` Ian Lance Taylor
@ 1999-03-11 11:20                     ` Jason Merrill
  1999-03-11 18:02                       ` Ian Lance Taylor
  0 siblings, 1 reply; 6+ messages in thread
From: Jason Merrill @ 1999-03-11 11:20 UTC (permalink / raw)
  To: Ian Lance Taylor; +Cc: robertlipe, bfd

>>>>> Ian Lance Taylor <ian@cygnus.com> writes:

 >    I assume that fixing this would dramatically improve link time for
 >    DWARF2?

 > I would not expect to see a dramatic improvement, but maybe there is
 > something I am overlooking.

I was thinking that a bunch of relocs to the same symbol would resolve
faster than relocs to different symbols; perhaps that's not the case.

Jason

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: patch was: dwarf2 debug dependent upon GAS?
  1999-03-11 11:20                     ` Jason Merrill
@ 1999-03-11 18:02                       ` Ian Lance Taylor
  0 siblings, 0 replies; 6+ messages in thread
From: Ian Lance Taylor @ 1999-03-11 18:02 UTC (permalink / raw)
  To: jason; +Cc: robertlipe, bfd

   From: Jason Merrill <jason@cygnus.com>
   Date: 11 Mar 1999 11:20:22 -0800

    >    I assume that fixing this would dramatically improve link time for
    >    DWARF2?

    > I would not expect to see a dramatic improvement, but maybe there is
    > something I am overlooking.

   I was thinking that a bunch of relocs to the same symbol would resolve
   faster than relocs to different symbols; perhaps that's not the case.

No, it isn't.  Each reloc contains a symbol index.  When the linker
performs the relocations, it first computes a table of all the symbol
values; the linker must perform this operation anyhow in order to
generate the symbol table of the output file.  Then the linker loops
through the relocs, resolving them.  Calculating the value of a symbol
is just a table lookup.

Ian

^ permalink raw reply	[flat|nested] 6+ messages in thread

end of thread, other threads:[~1999-03-11 18:02 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
     [not found] <19990226012522.D17655@rjlhome.sco.com>
     [not found] ` <29507.920192219@hurl.cygnus.com>
     [not found]   ` <19990228160642.H20315@rjlhome.sco.com>
     [not found]     ` <u9u2w67y67.fsf@yorick.cygnus.com>
     [not found]       ` <19990301102454.Q20315@rjlhome.sco.com>
     [not found]         ` <u9hfs585cd.fsf@yorick.cygnus.com>
     [not found]           ` <19990302002736.G27552@rjlhome.sco.com>
1999-03-11  0:02             ` patch was: dwarf2 debug dependent upon GAS? Jason Merrill
1999-03-11  7:41               ` Ian Lance Taylor
1999-03-11 10:31                 ` Jason Merrill
1999-03-11 10:44                   ` Ian Lance Taylor
1999-03-11 11:20                     ` Jason Merrill
1999-03-11 18:02                       ` Ian Lance Taylor

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).