Pentium friendly ffs code

[Jan Hubicka]

Hi
This patch adds emiting of bfs replacement code for Pentium according to the
Intel Optimizing Manual. It makes gcc to beat vc in yet another benchmark :)

Fri Oct 1 06:47:45 CEST 1999  Jan Hubicka  <hubicka@freesoft.cz>
	* i386.md (ffs expander): Emit Pentium friendly code for
	TARGET_PENTIUM.

*** i386.mc.old	Thu Sep 30 20:33:58 1999
--- i386.md	Fri Oct  1 11:19:54 1999
***************
*** 6778,6783 ****
--- 6834,6897 ----
  				const0_rtx),
  		    tmp,
  		    out)));
+       emit_insn (gen_addsi3 (out, out, const1_rtx));
+       emit_move_insn (operands[0], out);
+     }
+ 
+   /* Pentium bsf instruction is extremly slow.  Following code is recommended by
+      the Optimizing Manual as resonable replacement:
+            TEST    EAX,EAX
+ 	   JZ      SHORT BS2
+ 	   XOR     ECX,ECX
+ 	   MOV     DWORD PTR [TEMP+4],ECX
+ 	   SUB     ECX,EAX
+ 	   AND     EAX,ECX
+ 	   MOV     DWORD PTR [TEMP],EAX
+ 	   FILD    QWORD PTR [TEMP]
+ 	   FSTP    QWORD PTR [TEMP]
+ 	   WAIT    ; WAIT only needed for compatibility with
+ 	           ; earlier processors
+ 	   MOV     ECX, DWORD PTR [TEMP+4]
+ 	   SHR     ECX,20
+ 	   SUB     ECX,3FFH
+ 	   TEST    EAX,EAX       ; clear zero flag
+        BS2:
+      Following piece of code expand ffs to similar beast.
+        */
+ 
+   else if (TARGET_PENTIUM && !optimize_size && TARGET_80387)
+     {
+       rtx label = gen_label_rtx ();
+       rtx jump, lo, hi;
+       rtx mem = assign_386_stack_local (DImode, 0);
+       rtx fptmp = gen_reg_rtx (DFmode);
+       split_di (&mem, 1, &lo, &hi);
+ 
+       emit_move_insn (out, const0_rtx);
+       emit_insn (gen_cmpsi_0 (in, const0_rtx));
+ 
+       jump = gen_rtx_REG (CCNOmode, FLAGS_REG);
+       jump = gen_rtx_EQ (VOIDmode, jump, const0_rtx);
+       jump = gen_rtx_IF_THEN_ELSE (VOIDmode, jump,
+ 			          gen_rtx_LABEL_REF (VOIDmode, label),
+ 			          pc_rtx);
+       jump = emit_jump_insn (gen_rtx_SET (VOIDmode, pc_rtx, jump));
+       JUMP_LABEL (jump) = label;
+ 
+       emit_move_insn (hi, out);
+       emit_insn (gen_subsi3 (out, out, in));
+       emit_insn (gen_andsi3 (out, out, in));
+       emit_move_insn (lo, out);
+       emit_insn (gen_floatdidf2 (fptmp,mem));
+       emit_move_insn (gen_rtx_MEM (DFmode, XEXP (mem, 0)), fptmp);
+       emit_move_insn (out, hi);
+       emit_insn (gen_lshrsi3 (out, out, GEN_INT (20)));
+       emit_insn (gen_subsi3 (out, out, GEN_INT (0x3fe)));
+ 
+       emit_label (label);
+       LABEL_NUSES (label) = 1;
+ 
+       emit_move_insn (operands[0], out);
      }
    else
      {