This is the mail archive of the
gdb@sourceware.org
mailing list for the GDB project.
Re: gdb 6.4 - more packets sent for same command, any change in gdb internal logic ?
- From: Daniel Jacobowitz <drow at false dot org>
- To: suresh ds <dssuresh66 at mail dot com>
- Cc: gdb at sourceware dot org
- Date: Mon, 30 Oct 2006 08:44:04 -0500
- Subject: Re: gdb 6.4 - more packets sent for same command, any change in gdb internal logic ?
- References: <20061030065355.96EAA1024D@ws1-3.us4.outblaze.com>
On Mon, Oct 30, 2006 at 01:53:55AM -0500, suresh ds wrote:
> One important observation here is, after "p1d", it repeats the memory
> requests twice (ie. 200790, 200794, 200798, and again 200790, 200794,
> 200798). Can you explain why ?
By default, GDB does not cache memory; every time it needs it, it
rereads it.
> Why it sends so many memory-read packets for the same image and at
> same location, leading to total no.of packets getting increased by a
> factor of 4 ??!! Because of these too many packets, more time is
> taken now to complete this command. Can you explain what change gdb
> 6.4 has undergone to achieve this and why it does this ? Is there any
> way to reduce these many packets (ie. any gdb command/option) to
> bring back it's 6.3 behaviour ?
You're seeing GDB analyze the prologue of the function in which you're
stepping. It should not normally have to do this for stepi; that may
be a bug in the MIPS port. But you can probably speed it up a lot by
saying "set trust-readonly-sections". I think you'll need GDB from CVS
for that to work; it was broken for several releases.
--
Daniel Jacobowitz
CodeSourcery