[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Simulavr-devel] Re: Simulavr / clock
From: |
Theodore A. Roth |
Subject: |
Re: [Simulavr-devel] Re: Simulavr / clock |
Date: |
Wed, 11 Feb 2004 09:30:45 -0800 (PST) |
On Wed, 11 Feb 2004, Keith Gudger wrote:
> Ted:
>
> Thanks for the info and the patch. I applied it instead to the debug
> output in avrcore.c:
>
> @@ -893,8 +893,8 @@
> result = opi->func (core, opcode, opi->arg1, opi->arg2);
>
> if (global_debug_inst_output)
> - fprintf (stderr, "0x%06x (0x%06x) : 0x%04x : %s\n", pc, pc * 2,
> - opcode, global_opcode_name[result]);
> + fprintf (stderr, "0x%06x (0x%06x) : 0x%04x : %s \t: clock =
> %lld\n", pc, pc * 2,
> + opcode, global_opcode_name[result], avr_core_CK_get
> (core));
>
> return result;
> }
>
> The clock information is very useful when benchmarking various ways to
> perform the same problem, especially when they were written in C.
Ah yes, that seems like a sensible place to print out the clock count.
This and the output from handling a BREAK opcode should give you plenty
of information for tuning code.
>
> I'm going to look into the "named registers" which get returned with an
> "info all-registers". Could I put the cpu clock into some such register
> so that gdb can find it and print it with "p $r21" style?
I'm not sure that this should go into gdb. When you are running gdb
using avarice/jtagice as a back end, there is no way that I know of to
get clock information. Also, gdb with avarice is slow enough as it is
and something like this could slow it down even more.
Ted Roth