[Patch, i386]: Commited: macroize fmod and remainder x87 patterns

Václav Haisman V.Haisman@sh.cvut.cz
Sat Nov 25 16:03:00 GMT 2006 


Uros Bizjak wrote, On 25.11.2006 16:02:
> Hello!
> 
> This relatively boring patch macroizes fmod and remainder i387 patterns.
> 
> However, as fprem{,1}xf4_i387 patterns are inserted in a tight loop,
> implementation using float_extend patterns produce quite unoptimal code.
> To maintain optimal register live information, fprem pattern must be
> generated as:
> 
> gen_insn (gen_fpremxf4_i387 (op1, op2, op1, op2))
> 
> where op1 and op2 represent registers holding XFmode value, extended to
> XFmode before the loop.
> 
> When fpremxf4_i387 pattern is implemented as
> 
> (define_insn "fpremxf4_i387"
>  [(set (match_operand:XF 0 "register_operand" "=f")
>        (unspec:XF [(float_extend:XF (match_operand:DF 2
> "register_operand" "0"))
>                    (float_extend:XF (match_operand:DF 3
> "register_operand" "1"))]
>                   UNSPEC_FPREM_F))
>    ...
> 
> This pattern is generated from fmoddf3 expander as:
> 
> gen_insn(gen_fpremxf4_i387(op1, op2, operands[1], operands[2])
> 
> Unfortunatelly, in this case life analysis figures out that operands[2]
> operand dies inside the loop, and a couple of compensating fstp and flds
> are emitted inside the loop. So, current fpremxf4_i387 pattern (without
> inherent XFmode extension) produces better code (tighter fmod reduction
> loop).
> 
> Other than a straightforward DFmode and SFmode macroization, this patch
> relaxes operand constraints a bit. All input operands are moved through
> extend<mode>xf2 pattern, they can be constrainted ad "general_operands",
> to match extend<mode>xf2 input operands.
> 
> 2006-11-25  Uros Bizjak  <ubizjak@gmail.com>
> 
>        config/i386/i386.md (fpremxf4): Rename to fpremxf4_i387.
>        (fprem1xf4): Rename to fprem1xf4_i387.
> 
>        (fmodsf3, fmoddf3): Macroize patterns using X87MODEF12 mode macro.
>        Rename patterns to fmod<mode>3. Use general_operand operand
>        constraint for operands 1 and 2. Use SSE_FLOAT_MODE_P to disable
>        patterns for SSE math.
>        (remaindersf3, remainderdf3): Ditto.
> 
> Patch was regression tested on i686-pc-linux-gnu for c, c++ and fortran.
> At the end, it has distilled to quite trivial patch, so I guess it can
> be committed to SVN under obvious rule. (Patch also includes trivial
> rename of "sqrt<mode>xf2_i387" to "sqrt_extend<mode>xf2_i387",
> mistakenly left out of my previous patch.)
> 
> Uros.
> 
> 
> ------------------------------------------------------------------------
> 
> Index: config/i386/i386.md
> ===================================================================
> --- config/i386/i386.md	(revision 119188)
> +++ config/i386/i386.md	(working copy)
> @@ -15549,7 +15549,7 @@
>     (set_attr "mode" "XF")
>     (set_attr "athlon_decode" "direct")])
>  
> -(define_insn "sqrt<mode>xf2_i387"
> +(define_insn "sqrt_extend<mode>xf2_i387"
>    [(set (match_operand:XF 0 "register_operand" "=f")
>  	(sqrt:XF
>  	  (float_extend:XF
> @@ -15582,13 +15582,13 @@
>        rtx op0 = gen_reg_rtx (XFmode);
>        rtx op1 = force_reg (<MODE>mode, operands[1]);
>  
> -      emit_insn (gen_sqrt<mode>xf2_i387 (op0, op1));
> +      emit_insn (gen_sqrt_extend<mode>xf2_i387 (op0, op1));
>        emit_insn (gen_truncxf<mode>2_i387_noop_unspec (operands[0], op0));
>        DONE;
>     }
>  })
>  
> -(define_insn "fpremxf4"
> +(define_insn "fpremxf4_i387"
>    [(set (match_operand:XF 0 "register_operand" "=f")
>  	(unspec:XF [(match_operand:XF 2 "register_operand" "0")
>  		    (match_operand:XF 3 "register_operand" "1")]
> @@ -15603,73 +15603,49 @@
>    [(set_attr "type" "fpspc")
>     (set_attr "mode" "XF")])
>  
> -(define_expand "fmodsf3"
> -  [(use (match_operand:SF 0 "register_operand" ""))
> -   (use (match_operand:SF 1 "register_operand" ""))
> -   (use (match_operand:SF 2 "register_operand" ""))]
> -  "TARGET_USE_FANCY_MATH_387
> -   && (!TARGET_SSE_MATH || TARGET_MIX_SSE_I387)"
> +(define_expand "fmodxf3"
> +  [(use (match_operand:XF 0 "register_operand" ""))
> +   (use (match_operand:XF 1 "register_operand" ""))
> +   (use (match_operand:XF 2 "register_operand" ""))]
> +  "TARGET_USE_FANCY_MATH_387"
>  {
>    rtx label = gen_label_rtx ();
>  
> -  rtx op1 = gen_reg_rtx (XFmode);
> -  rtx op2 = gen_reg_rtx (XFmode);
> -
> -  emit_insn(gen_extendsfxf2 (op1, operands[1]));
> -  emit_insn(gen_extendsfxf2 (op2, operands[2]));
> -
>    emit_label (label);
>  
> -  emit_insn (gen_fpremxf4 (op1, op2, op1, op2));
> +  emit_insn (gen_fpremxf4_i387 (operands[1], operands[2],
> +				operands[1], operands[2]));
>    ix86_emit_fp_unordered_jump (label);
>  
> -  emit_insn (gen_truncxfsf2_i387_noop_unspec (operands[0], op1));
> +  emit_move_insn (operands[0], operands[1]);
>    DONE;
>  })
>  
> -(define_expand "fmoddf3"
> -  [(use (match_operand:DF 0 "register_operand" ""))
> -   (use (match_operand:DF 1 "register_operand" ""))
> -   (use (match_operand:DF 2 "register_operand" ""))]
> +(define_expand "fmod<mode>3"
> +  [(use (match_operand:X87MODEF12 0 "register_operand" ""))
> +   (use (match_operand:X87MODEF12 1 "general_operand" ""))
> +   (use (match_operand:X87MODEF12 2 "general_operand" ""))]
>    "TARGET_USE_FANCY_MATH_387
> -   && (!(TARGET_SSE2 && TARGET_SSE_MATH) || TARGET_MIX_SSE_I387)"
> +   && (!(SSE_FLOAT_MODE_P (<MODE>mode) && TARGET_SSE_MATH)
> +       || TARGET_MIX_SSE_I387)"
>  {
>    rtx label = gen_label_rtx ();
>  
>    rtx op1 = gen_reg_rtx (XFmode);
>    rtx op2 = gen_reg_rtx (XFmode);
>  
> -  emit_insn (gen_extenddfxf2 (op1, operands[1]));
> -  emit_insn (gen_extenddfxf2 (op2, operands[2]));
> +  emit_insn(gen_extend<mode>xf2 (op1, operands[1]));
> +  emit_insn(gen_extend<mode>xf2 (op2, operands[2]));
Missing space before the (.

>  
>    emit_label (label);
> -
> -  emit_insn (gen_fpremxf4 (op1, op2, op1, op2));
> +  emit_insn (gen_fpremxf4_i387 (op1, op2, op1, op2));
>    ix86_emit_fp_unordered_jump (label);
>  
> -  emit_insn (gen_truncxfdf2_i387_noop_unspec (operands[0], op1));
> +  emit_insn (gen_truncxf<mode>2_i387_noop_unspec (operands[0], op1));
>    DONE;
>  })
>  
> -(define_expand "fmodxf3"
> -  [(use (match_operand:XF 0 "register_operand" ""))
> -   (use (match_operand:XF 1 "register_operand" ""))
> -   (use (match_operand:XF 2 "register_operand" ""))]
> -  "TARGET_USE_FANCY_MATH_387"
> -{
> -  rtx label = gen_label_rtx ();
> -
> -  emit_label (label);
> -
> -  emit_insn (gen_fpremxf4 (operands[1], operands[2],
> -			   operands[1], operands[2]));
> -  ix86_emit_fp_unordered_jump (label);
> -
> -  emit_move_insn (operands[0], operands[1]);
> -  DONE;
> -})
> -
> -(define_insn "fprem1xf4"
> +(define_insn "fprem1xf4_i387"
>    [(set (match_operand:XF 0 "register_operand" "=f")
>  	(unspec:XF [(match_operand:XF 2 "register_operand" "0")
>  		    (match_operand:XF 3 "register_operand" "1")]
> @@ -15684,72 +15660,49 @@
>    [(set_attr "type" "fpspc")
>     (set_attr "mode" "XF")])
>  
> -(define_expand "remaindersf3"
> -  [(use (match_operand:SF 0 "register_operand" ""))
> -   (use (match_operand:SF 1 "register_operand" ""))
> -   (use (match_operand:SF 2 "register_operand" ""))]
> -  "TARGET_USE_FANCY_MATH_387
> -   && (!TARGET_SSE_MATH || TARGET_MIX_SSE_I387)"
> +(define_expand "remainderxf3"
> +  [(use (match_operand:XF 0 "register_operand" ""))
> +   (use (match_operand:XF 1 "register_operand" ""))
> +   (use (match_operand:XF 2 "register_operand" ""))]
> +  "TARGET_USE_FANCY_MATH_387"
>  {
>    rtx label = gen_label_rtx ();
>  
> -  rtx op1 = gen_reg_rtx (XFmode);
> -  rtx op2 = gen_reg_rtx (XFmode);
> -
> -  emit_insn(gen_extendsfxf2 (op1, operands[1]));
> -  emit_insn(gen_extendsfxf2 (op2, operands[2]));
> -
>    emit_label (label);
>  
> -  emit_insn (gen_fprem1xf4 (op1, op2, op1, op2));
> +  emit_insn (gen_fprem1xf4_i387 (operands[1], operands[2],
> +				 operands[1], operands[2]));
>    ix86_emit_fp_unordered_jump (label);
>  
> -  emit_insn (gen_truncxfsf2_i387_noop_unspec (operands[0], op1));
> +  emit_move_insn (operands[0], operands[1]);
>    DONE;
>  })
>  
> -(define_expand "remainderdf3"
> -  [(use (match_operand:DF 0 "register_operand" ""))
> -   (use (match_operand:DF 1 "register_operand" ""))
> -   (use (match_operand:DF 2 "register_operand" ""))]
> +(define_expand "remainder<mode>3"
> +  [(use (match_operand:X87MODEF12 0 "register_operand" ""))
> +   (use (match_operand:X87MODEF12 1 "general_operand" ""))
> +   (use (match_operand:X87MODEF12 2 "general_operand" ""))]
>    "TARGET_USE_FANCY_MATH_387
> -   && (!(TARGET_SSE2 && TARGET_SSE_MATH) || TARGET_MIX_SSE_I387)"
> +   && (!(SSE_FLOAT_MODE_P (<MODE>mode) && TARGET_SSE_MATH)
> +       || TARGET_MIX_SSE_I387)"
>  {
>    rtx label = gen_label_rtx ();
>  
>    rtx op1 = gen_reg_rtx (XFmode);
>    rtx op2 = gen_reg_rtx (XFmode);
>  
> -  emit_insn (gen_extenddfxf2 (op1, operands[1]));
> -  emit_insn (gen_extenddfxf2 (op2, operands[2]));
> +  emit_insn(gen_extend<mode>xf2 (op1, operands[1]));
> +  emit_insn(gen_extend<mode>xf2 (op2, operands[2]));
Ditto.

>  
>    emit_label (label);
>  
> -  emit_insn (gen_fprem1xf4 (op1, op2, op1, op2));
> +  emit_insn (gen_fprem1xf4_i387 (op1, op2, op1, op2));
>    ix86_emit_fp_unordered_jump (label);
>  
> -  emit_insn (gen_truncxfdf2_i387_noop_unspec (operands[0], op1));
> +  emit_insn (gen_truncxf<mode>2_i387_noop_unspec (operands[0], op1));
>    DONE;
>  })
>  
> -(define_expand "remainderxf3"
> -  [(use (match_operand:XF 0 "register_operand" ""))
> -   (use (match_operand:XF 1 "register_operand" ""))
> -   (use (match_operand:XF 2 "register_operand" ""))]
> -  "TARGET_USE_FANCY_MATH_387"
> -{
> -  rtx label = gen_label_rtx ();
> -
> -  emit_label (label);
> -
> -  emit_insn (gen_fprem1xf4 (operands[1], operands[2],
> -			    operands[1], operands[2]));
> -  ix86_emit_fp_unordered_jump (label);
> -
> -  emit_move_insn (operands[0], operands[1]);
> -  DONE;
> -})
> -
>  (define_insn "*sindf2"
>    [(set (match_operand:DF 0 "register_operand" "=f")
>  	(unspec:DF [(match_operand:DF 1 "register_operand" "0")] UNSPEC_SIN))]


--
VH

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 542 bytes
Desc: OpenPGP digital signature
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20061125/af8937c2/attachment.sig>

More information about the Gcc-patches mailing list