[RFA/alpha] Add handling of FP control insn in software-single step (second try)

[RFA/alpha] Add handling of FP control insn in software-single step (second try)

[Joel Brobecker]

[-- Attachment #1: Type: text/plain, Size: 816 bytes --]

This is a rework of:

        http://sources.redhat.com/ml/gdb-patches/2005-05/msg00405.html

Hopefully this patch will be more acceptable.

2005-05-17  Joel Brobecker  <brobecker@adacore.com>

        * alpha-tdep.c (fp_register_zero_p): New function.
        (fp_register_sign_bit): New function.
        (alpha_next_pc): Add support for floating-point control instructions.

Tested on alpha-tru64 5.1a with no regression. I'll try to contribute
an asm test asap.

One remark: For zero_mask in fp_register_zero_p(), Richard suggested
to use "((LONGEST) 1 << 63) - 1", but that triggered an overflow warning
from GCC. So I reverted to using the xor binary operator. Daniel said
that many things would break in GDB with 128-bit types, so I think the
xor is quite acceptable for now.

OK to commit?

Thanks,
-- 
Joel

[-- Attachment #2: alpha-tdep.c.diff --]
[-- Type: text/plain, Size: 2795 bytes --]

Index: alpha-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/alpha-tdep.c,v
retrieving revision 1.148
diff -u -p -r1.148 alpha-tdep.c
--- alpha-tdep.c	17 Feb 2005 13:49:48 -0000	1.148
+++ alpha-tdep.c	20 May 2005 01:14:42 -0000
@@ -1335,6 +1335,30 @@ alpha_fill_fp_regs (int regno, void *f0_
 }
 
 \f
+
+/* Return nonzero if the G_floating register value in REG is equal to
+   zero for FP control instructions.  */
+   
+static int
+fp_register_zero_p (LONGEST reg)
+{
+  /* Check that all bits except the sign bit are zero.  */
+  const LONGEST zero_mask = ((LONGEST) 1 << 63) ^ -1;
+
+  return ((reg & zero_mask) == 0);
+}
+
+/* Return the value of the sign bit for the G_floating register
+   value held in REG.  */
+
+static int
+fp_register_sign_bit (LONGEST reg)
+{
+  const LONGEST sign_mask = (LONGEST) 1 << 63;
+
+  return ((reg & sign_mask) != 0);
+}
+
 /* alpha_software_single_step() is called just before we want to resume
    the inferior, if we want to single-step it but there is no hardware
    or kernel single-step support (NetBSD on Alpha, for example).  We find
@@ -1350,6 +1374,7 @@ alpha_next_pc (CORE_ADDR pc)
   unsigned int op;
   int offset;
   LONGEST rav;
+  char reg[8];
 
   insn = alpha_read_insn (pc);
 
@@ -1379,7 +1404,9 @@ alpha_next_pc (CORE_ADDR pc)
 	}
 
       /* Need to determine if branch is taken; read RA.  */
-      rav = (LONGEST) read_register ((insn >> 21) & 0x1f);
+      regcache_cooked_read (current_regcache, (insn >> 21) & 0x1f, reg);
+      rav = extract_signed_integer (reg, 8);
+
       switch (op)
 	{
 	case 0x38:		/* BLBC */
@@ -1415,7 +1442,32 @@ alpha_next_pc (CORE_ADDR pc)
 	    goto branch_taken;
 	  break;
 
-	/* ??? Missing floating-point branches.  */
+        /* Floating point branches.  */
+        
+        case 0x31:              /* FBEQ */
+          if (fp_register_zero_p (rav))
+            goto branch_taken;
+          break;
+        case 0x36:              /* FBGE */
+          if (fp_register_sign_bit (rav) == 0 || fp_register_zero_p (rav))
+            goto branch_taken;
+          break;
+        case 0x37:              /* FBGT */
+          if (fp_register_sign_bit (rav) == 0 && ! fp_register_zero_p (rav))
+            goto branch_taken;
+          break;
+        case 0x33:              /* FBLE */
+          if (fp_register_sign_bit (rav) == 1 || fp_register_zero_p (rav))
+            goto branch_taken;
+          break;
+        case 0x32:              /* FBLT */
+          if (fp_register_sign_bit (rav) == 1 && ! fp_register_zero_p (rav))
+            goto branch_taken;
+          break;
+        case 0x35:              /* FBNE */
+          if (! fp_register_zero_p (rav))
+            goto branch_taken;
+          break;
 	}
     }
 

Re: [RFA/alpha] Add handling of FP control insn in software-single step (second try)

[Richard Henderson]

On Fri, May 20, 2005 at 11:20:55AM +1000, Joel Brobecker wrote:
>         * alpha-tdep.c (fp_register_zero_p): New function.
>         (fp_register_sign_bit): New function.
>         (alpha_next_pc): Add support for floating-point control instructions.

Much better, thanks.  Though I don't have approval rights.  ;-)


r~

Re: [RFA/alpha] Add handling of FP control insn in software-single step (second try)

[Daniel Jacobowitz]

On Fri, May 20, 2005 at 11:20:55AM +1000, Joel Brobecker wrote:
> This is a rework of:
> 
>         http://sources.redhat.com/ml/gdb-patches/2005-05/msg00405.html
> 
> Hopefully this patch will be more acceptable.
> 
> 2005-05-17  Joel Brobecker  <brobecker@adacore.com>
> 
>         * alpha-tdep.c (fp_register_zero_p): New function.
>         (fp_register_sign_bit): New function.
>         (alpha_next_pc): Add support for floating-point control instructions.
> 
> Tested on alpha-tru64 5.1a with no regression. I'll try to contribute
> an asm test asap.
> 
> One remark: For zero_mask in fp_register_zero_p(), Richard suggested
> to use "((LONGEST) 1 << 63) - 1", but that triggered an overflow warning
> from GCC. So I reverted to using the xor binary operator. Daniel said
> that many things would break in GDB with 128-bit types, so I think the
> xor is quite acceptable for now.
> 
> OK to commit?

I was just waiting for Richard to OK the actual implementation.  This
is fine, thank you.

-- 
Daniel Jacobowitz
CodeSourcery, LLC

Re: [RFA/alpha] Add handling of FP control insn in software-single step (second try)

[Joel Brobecker]

> > 2005-05-17  Joel Brobecker  <brobecker@adacore.com>
> > 
> >         * alpha-tdep.c (fp_register_zero_p): New function.
> >         (fp_register_sign_bit): New function.
> >         (alpha_next_pc): Add support for floating-point control instructions.
> I was just waiting for Richard to OK the actual implementation.  This
> is fine, thank you.

Thanks. Checked in.

-- 
Joel