[RFA] V850 patch

[RFA] V850 patch

[Martin M. Hunt]

I'm submitting this for Jim Wilson.  It is a revised patch 
of the one submitted by Miles Bader for GDB PR #870.

I have verified it applies cleanly to the current CVS head and 
seems to fix the problems it claims. There are no testsuite regressions.

2003-01-08  Miles Bader  <miles@gnu.org>
	    Jim Wilson  <wilson@redhat.com>
	
	* v850-tdep.c (v850_scan_prologue): Handle six byte instructions.
	Correctly sign-extend insn2.  Use insn2 as well as insn to recognize
	insns where appropriate (some insns only differ in the second 16-bit
	word).  Handle mov imm5,r12 and mov imm32,r12.

Index: v850-tdep.c
===================================================================
RCS file: /cvs/cvsfiles/devo/gdb/v850-tdep.c,v
retrieving revision 2.33.6.1
diff -p -r2.33.6.1 v850-tdep.c
*** v850-tdep.c	2002/08/08 07:56:41	2.33.6.1
--- v850-tdep.c	2003/01/08 20:06:06
*************** v850_scan_prologue (CORE_ADDR pc, struct
*** 627,643 ****
  
        insn = read_memory_unsigned_integer (current_pc, 2);
        current_pc += 2;
!       if ((insn & 0x0780) >= 0x0600)	/* Four byte instruction? */
  	{
  	  insn2 = read_memory_unsigned_integer (current_pc, 2);
  	  current_pc += 2;
  	}
  
!       if ((insn & 0xffc0) == ((10 << 11) | 0x0780) && !regsave_func_p)
  	{			/* jarl <func>,10 */
! 	  long low_disp = insn2 & ~(long) 1;
! 	  long disp = (((((insn & 0x3f) << 16) + low_disp)
! 			& ~(long) 1) ^ 0x00200000) - 0x00200000;
  
  	  save_pc = current_pc;
  	  save_end = prologue_end;
--- 627,654 ----
  
        insn = read_memory_unsigned_integer (current_pc, 2);
        current_pc += 2;
!       if ((insn & 0xffe0) == 0x0620)		/* Six byte instruction? */
  	{
+ 	  insn2 = read_memory_unsigned_integer (current_pc, 4);
+ 	  current_pc += 4;
+ 	}
+       else if ((insn & 0x0780) >= 0x0600)	/* Four byte instruction? */
+ 	{
  	  insn2 = read_memory_unsigned_integer (current_pc, 2);
  	  current_pc += 2;
+ 
+ 	  /* Most uses of insn2 below interpret it as a sign-extended
+ 	     16-bit field, so sign-extend it here.  */
+ 	  insn2 = (insn2 ^ 0x8000) - 0x8000;
  	}
  
!       if ((insn & 0xffc0) == ((10 << 11) | 0x0780)
! 	  && (insn2 & 1) == 0
! 	  && !regsave_func_p)
  	{			/* jarl <func>,10 */
! 	  long low_disp = insn2 & 0xfffe;
! 	  long disp = (((((insn & 0x3f) << 16) | low_disp) ^ 0x00200000)
! 		       - 0x00200000);
  
  	  save_pc = current_pc;
  	  save_end = prologue_end;
*************** v850_scan_prologue (CORE_ADDR pc, struct
*** 676,683 ****
  #endif
  	  continue;
  	}
!       else if ((insn & 0xffc0) == 0x0780)	/* prepare list2,imm5 */
! 	{
  	  handle_prepare (insn, insn2, &current_pc, pi, &pifsr);
  	  continue;
  	}
--- 687,695 ----
  #endif
  	  continue;
  	}
!       else if ((insn & 0xffc0) == 0x0780
! 	       && ((insn2 & 0x1f) == 0x01 || (insn2 & 0x03) == 0x03))
! 	{			/* prepare list2,imm5 */
  	  handle_prepare (insn, insn2, &current_pc, pi, &pifsr);
  	  continue;
  	}
*************** v850_scan_prologue (CORE_ADDR pc, struct
*** 706,712 ****
  #endif
  	  continue;
  	}
!       else if ((insn & 0x07c0) == 0x0780	/* jarl or jr */
  	       || (insn & 0xffe0) == 0x0060	/* jmp */
  	       || (insn & 0x0780) == 0x0580)	/* branch */
  	{
--- 718,725 ----
  #endif
  	  continue;
  	}
!       else if (((insn & 0x07c0) == 0x0780	/* jarl or jr */
! 		&& (insn2 & 1) == 0)
  	       || (insn & 0xffe0) == 0x0060	/* jmp */
  	       || (insn & 0x0780) == 0x0580)	/* branch */
  	{
*************** v850_scan_prologue (CORE_ADDR pc, struct
*** 726,731 ****
--- 739,748 ----
  	  pi->framereg = E_FP_RAW_REGNUM;
  	}
  
+       else if ((insn & 0xffe0) == ((E_R12_REGNUM << 11) | 0x0200))	/* mov imm5,r12 */
+ 	r12_tmp = ((insn & 0x1f) ^ 0x10) - 0x10;
+       else if (insn == (0x0620 | E_R12_REGNUM))				/* mov imm32,r12 */
+ 	r12_tmp = insn2;
        else if (insn == ((E_R12_REGNUM << 11) | 0x0640 | E_R0_REGNUM))	/* movhi hi(const),r0,r12 */
  	r12_tmp = insn2 << 16;
        else if (insn == ((E_R12_REGNUM << 11) | 0x0620 | E_R12_REGNUM))	/* movea lo(const),r12,r12 */

Re: [RFA] V850 patch

[Andrew Cagney]

> I'm submitting this for Jim Wilson.  It is a revised patch 
> of the one submitted by Miles Bader for GDB PR #870.
> 
> I have verified it applies cleanly to the current CVS head and 
> seems to fix the problems it claims. There are no testsuite regressions.
> 
> 2003-01-08  Miles Bader  <miles@gnu.org>
> 	    Jim Wilson  <wilson@redhat.com>
> 	
> 	* v850-tdep.c (v850_scan_prologue): Handle six byte instructions.
> 	Correctly sign-extend insn2.  Use insn2 as well as insn to recognize
> 	insns where appropriate (some insns only differ in the second 16-bit
> 	word).  Handle mov imm5,r12 and mov imm32,r12.
> 
FYI, while Jim has an assignment. Miles doesn't :-(  Do you know how 
`revised' it is?

Andrew

Re: [RFA] V850 patch

[Martin M. Hunt]

On Thu, 2003-02-20 at 05:02, Andrew Cagney wrote:
> > I'm submitting this for Jim Wilson.  It is a revised patch 
> > of the one submitted by Miles Bader for GDB PR #870.
> > 
> > I have verified it applies cleanly to the current CVS head and 
> > seems to fix the problems it claims. There are no testsuite regressions.
> > 
> > 2003-01-08  Miles Bader  <miles@gnu.org>
> > 	    Jim Wilson  <wilson@redhat.com>
> > 	
> > 	* v850-tdep.c (v850_scan_prologue): Handle six byte instructions.
> > 	Correctly sign-extend insn2.  Use insn2 as well as insn to recognize
> > 	insns where appropriate (some insns only differ in the second 16-bit
> > 	word).  Handle mov imm5,r12 and mov imm32,r12.
> > 
> FYI, while Jim has an assignment. Miles doesn't :-(  Do you know how 
> `revised' it is?

It's about 95% changed.  Jim said he started with the patch Miles sent
but there were numerous problems and he had to rewrite it.  I only see
one or two similar lines in the patches.

If Miles Bader doesn't have a copyright assignment, we have a problem.  

> head gnu-nat.c
/* Interface GDB to the GNU Hurd.
   Copyright 1992, 1995, 1996, 1997, 1998, 1999, 2000, 2001
   Free Software Foundation, Inc.

   This file is part of GDB.

   Written by Miles Bader <miles@gnu.ai.mit.edu>

Martin

Re: [RFA] V850 patch

[Andrew Cagney]

Yes, we have a problem, it's being addressed :-(

Re: [RFA] V850 patch

[Andrew Cagney]

Martin,

I've you've the original C file that caused the problem you can 
probably, with a little work, end up with the needed code.

Andrew