public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
* [PATCH] PR/41822 pdp11: fix wrong code for and operation
@ 2010-10-29 22:44 Paul Koning
  2010-10-29 23:08 ` Paolo Bonzini
  0 siblings, 1 reply; 4+ messages in thread
From: Paul Koning @ 2010-10-29 22:44 UTC (permalink / raw)
  To: gcc-patches

PDP11 has a bic instruction, which is and not.  The code generation for that was wrong. Fixed by borrowing, with changes, from vax.md.

Tested by test build, inspection of generated assembly code.  Committed.

ChangeLog:

2010-10-29  Paul Koning  <ni1d@arrl.net>

	PR/41822
	* config/pdp11/pdp11.md (andhi3, andqi3): Fix wrong code error.

Index: config/pdp11/pdp11.md
===================================================================
--- config/pdp11/pdp11.md	(revision 166060)
+++ config/pdp11/pdp11.md	(working copy)
@@ -43,6 +43,11 @@
 ;; HI is 16 bit
 ;; QI is 8 bit 
 
+;; Integer modes supported on the PDP11, with a mapping from machine mode
+;; to mnemonic suffix.  SImode and DImode always are special cases.
+(define_mode_iterator PDPint [QI HI])
+(define_mode_attr  isfx [(QI "b") (HI "")])
+
 ;;- See file "rtl.def" for documentation on define_insn, match_*, et. al.
 
 ;;- cpp macro #define NOTICE_UPDATE_CC in file tm.h handles condition code
@@ -809,71 +814,41 @@
 ;;;;- and instructions
 ;; Bit-and on the pdp (like on the VAX) is done with a clear-bits insn.
 
-(define_insn "andsi3"
-  [(set (match_operand:SI 0 "general_operand" "=r,r,o,o,r,r,r,o,o,o")
-        (and:SI (match_operand:SI 1 "general_operand" "%0,0,0,0,0,0,0,0,0,0")
-                (not:SI (match_operand:SI 2 "general_operand" "r,o,r,o,I,J,K,I,J,K"))))]
+(define_expand "and<mode>3"
+  [(set (match_operand:PDPint 0 "general_operand" "")
+	(and:PDPint (not:PDPint (match_operand:PDPint 1 "general_operand" ""))
+		   (match_operand:PDPint 2 "general_operand" "")))]
   ""
-  "*
-{ /* Here we trust that operands don't overlap 
+  "
+{
+  rtx op1 = operands[1];
 
-     or is lateoperands the low word?? - looks like it! */
+  /* If there is a constant argument, complement that one.
+     Similarly, if one of the inputs is the same as the output,
+     complement the other input.  */
+  if ((CONST_INT_P (operands[2]) && ! CONST_INT_P (op1)) ||
+      rtx_equal_p (operands[0], operands[1]))
+    {
+      operands[1] = operands[2];
+      operands[2] = op1;
+      op1 = operands[1];
+    }
 
-  rtx lateoperands[3];
-  
-  lateoperands[0] = operands[0];
-
-  if (REG_P (operands[0]))
-    operands[0] = gen_rtx_REG (HImode, REGNO (operands[0]) + 1);
+  if (CONST_INT_P (op1))
+    operands[1] = GEN_INT (~INTVAL (op1));
   else
-    operands[0] = adjust_address (operands[0], HImode, 2);
-  
-  if (! CONSTANT_P(operands[2]))
-  {
-    lateoperands[2] = operands[2];
+    operands[1] = expand_unop (<MODE>mode, one_cmpl_optab, op1, 0, 1);
+}")
 
-    if (REG_P (operands[2]))
-      operands[2] = gen_rtx_REG (HImode, REGNO (operands[2]) + 1);
-    else
-      operands[2] = adjust_address (operands[2], HImode, 2);
-
-    output_asm_insn (\"bic %2, %0\", operands);
-    output_asm_insn (\"bic %2, %0\", lateoperands);
-    return \"\";
-  }
-
-  lateoperands[2] = GEN_INT ((INTVAL (operands[2]) >> 16) & 0xffff);
-  operands[2] = GEN_INT (INTVAL (operands[2]) & 0xffff);
-  
-  /* these have different lengths, so we should have 
-     different constraints! */
-  if (INTVAL(operands[2]))
-    output_asm_insn (\"bic %2, %0\", operands);
-
-  if (INTVAL(lateoperands[2]))
-    output_asm_insn (\"bic %2, %0\", lateoperands);
-
-  return \"\";
-}"
-  [(set_attr "length" "4,8,8,12,4,4,8,6,6,12")])
-
-;; FIXME This definition is wrong, PR/41822
-(define_insn "andhi3"
-  [(set (match_operand:HI 0 "general_operand" "=rR,rR,Q,Q")
-	(and:HI (match_operand:HI 1 "general_operand" "0,0,0,0")
-		(not:HI (match_operand:HI 2 "general_operand" "rR,Qi,rR,Qi"))))]
+(define_insn "*and<mode>"
+  [(set (match_operand:PDPint 0 "general_operand" "=rR,rR,Q,Q")
+	(and:PDPint
+	     (not: PDPint (match_operand:PDPint 1 "general_operand" "rR,Qi,rR,Qi"))
+	     (match_operand:PDPint 2 "general_operand" "0,0,0,0")))]
   ""
-  "bic %2, %0"
+  "bic<PDPint:isfx> %1, %0"
   [(set_attr "length" "2,4,4,6")])
 
-(define_insn "andqi3"
-  [(set (match_operand:QI 0 "general_operand" "=rR,rR,Q,Q")
-	(and:QI (match_operand:QI 1 "general_operand" "0,0,0,0")
-		(not:QI (match_operand:QI 2 "general_operand" "rR,Qi,rR,Qi"))))]
-  ""
-  "bicb %2, %0"
-  [(set_attr "length" "2,4,4,6")])
-
 ;;- Bit set (inclusive or) instructions
 (define_insn "iorsi3"
   [(set (match_operand:SI 0 "general_operand" "=r,r,o,o,r,r,r,o,o,o")

^ permalink raw reply	[flat|nested] 4+ messages in thread
end of thread, other threads:[~2010-10-29 20:43 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2010-10-29 22:44 [PATCH] PR/41822 pdp11: fix wrong code for and operation Paul Koning
2010-10-29 23:08 ` Paolo Bonzini
2010-10-29 23:14   ` Paul Koning
2010-10-29 23:40   ` Paul Koning

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).