Re: [RFA] Patch for THUMB skip_prologue code

[Michael Snyder <msnyder@redhat.com>]

Fred Fish wrote:
> 
> The current THUMB prologue skipping code does not stop until
> it reaches either the end of the function, or an address that
> is the function start address plus 40.  This patch fixes it.
> The gdb testsuite results for thumb code after applying this
> patch have the following diff:
> 
>   10694c10694
>   < FAIL: gdb.base/nodebug.exp: running to middle in runto
>   ---
>   > PASS: gdb.base/nodebug.exp: backtrace from middle in nodebug.exp
>   15907,15908c15907,15908
>   < # of expected passes        7469
>   < # of unexpected failures    104
>   ---
>   > # of expected passes        7470
>   > # of unexpected failures    103
> 
> -Fred

Cool -- did you measure a marked performance increase?

> 
> ============================================================
> 
> 2002-01-18  Fred Fish  <fnf@redhat.com>
> 
>         * arm-tdep.c (thumb_skip_prologue): Quit scanning prologue
>         when we have found all instructions we are looking for.
> 
> Index: arm-tdep.c
> ===================================================================
> RCS file: /cvs/src/src/gdb/arm-tdep.c,v
> retrieving revision 1.26
> diff -u -p -r1.26 arm-tdep.c
> --- arm-tdep.c  2002/01/09 18:07:48     1.26
> +++ arm-tdep.c  2002/01/18 23:26:14
> @@ -376,6 +376,10 @@ thumb_skip_prologue (CORE_ADDR pc, CORE_
>         {
>           findmask |= 2;  /* setting of r7 found */
>         }
> +      else if (findmask == (4+2+1))
> +       {
> +         break;        /* We have found one of each type of prologue instruction */
> +       }
>        else
>         continue;       /* something in the prolog that we don't care about or some
>                            instruction from outside the prolog scheduled here for optimization */