RFC: Optimization to write_register_bytes()

Fernando Nasser fnasser@cygnus.com
Fri May 5 01:03:00 GMT 2000


Andrew Cagney wrote:
> 
(...) Register fix plan skipped
> 
> PS: Fernando, sorry to hit you with this as a response.  You're simply
> the latest in a long list of people (me included) that have had to
> wrestle with the current register-cache design.  For as long as I can
> remember people have been comming to grief with this code (just changing
> REGISTER_NAMES[] to REGISTER_NAME() was a radical breakthrough :-).  I
> would really like to set out something that is going to finally fix the
> problem.

No problem.  I agree with you that all this code needs a redesign and I
believe the answer is some abstraction as you described.  It is a good thing to add it to the TODO
list so we eventually get someone to implement it.

Unfortunately I won't be able to work on this at this time, but I still want/need to have the MMX
registers available as soon as possible.  And adding the few lines below would allow me to do that
very easily.

Fernando




Index: findvar.c
===================================================================
RCS file: /cvs/src/src/gdb/findvar.c,v
retrieving revision 1.9
diff -c -p -r1.9 findvar.c
*** findvar.c   2000/04/27 15:33:01     1.9
--- findvar.c   2000/05/05 07:58:51
*************** write_register_bytes (myregstart, myaddr
*** 829,840 ****
    int regno;
  
    target_prepare_to_store ();
  
    /* Scan through the registers updating any that are covered by the range
       myregstart<=>myregend using write_register_gen, which does nice things
       like handling threads, and avoiding updates when the new and old contents
!      are the same.  */
  
    for (regno = 0; regno < NUM_REGS; regno++)
      {
        int regstart, regend;
--- 829,846 ----
    int regno;
  
    target_prepare_to_store ();
+   
  
    /* Scan through the registers updating any that are covered by the range
       myregstart<=>myregend using write_register_gen, which does nice things
       like handling threads, and avoiding updates when the new and old contents
!      are the same.
  
+      We only update a contiguous range of registers, stopping when the
+      byte at myregend is written.  This avoids writting to aliased registers
+      twice (note that aliased register numbers must be higher than the
+      real registers they share memory with).  */
+ 
    for (regno = 0; regno < NUM_REGS; regno++)
      {
        int regstart, regend;
*************** write_register_bytes (myregstart, myaddr
*** 848,854 ****
  
        /* Is this register completely within the range the user is writing?  */
        else if (myregstart <= regstart && regend <= myregend)
!       write_register_gen (regno, myaddr + (regstart - myregstart));
  
        /* The register partially overlaps the range being written.  */
        else
--- 854,864 ----
  
        /* Is this register completely within the range the user is writing?  */
        else if (myregstart <= regstart && regend <= myregend)
!         {
!         write_register_gen (regno, myaddr + (regstart - myregstart));
!         if (regend >= myregend)
!           break;
!         }
  
        /* The register partially overlaps the range being written.  */
        else
*************** write_register_bytes (myregstart, myaddr
*** 868,873 ****
--- 878,886 ----
                  overlapend - overlapstart);
  
          target_store_registers (regno);
+ 
+         if (regend >= myregend)
+           break;
        }
      }
  }


-- 
Fernando Nasser
Cygnus Solutions (a Red Hat company)    E-Mail:  fnasser@cygnus.com
2323 Yonge Street, Suite #300           Tel:  416-482-2661 ext. 311
Toronto, Ontario   M4P 2C9              Fax:  416-482-6299


More information about the Gdb-patches mailing list