[i386] : Rewrite x87 trigonometric patterns

Message ID CAFULd4YcPzn0yHFZgOkNiJzPv-08oHXeFAcfDHFCCZ=STtyUiw@mail.gmail.com
State New
Headers show
Series
  • [i386] : Rewrite x87 trigonometric patterns
Related show

Commit Message

Uros Bizjak Sept. 10, 2018, 7:19 p.m.
Hello!

Attached patch removes a bunch of sincos -> sin, cos splitters. These
are not necessary anymore since middle end handles the transformation
in a generic way.
Additionally, the patch removes unnecessary mixed-mode patters
(SFmode/DFmode input operand and XFmode outputs). These are not
needed, because all x87 float_extend RTXes degenerate to a plain move
(or a no-op move). The patch also includes a couple of cleanups with
no functional changes.

2018-09-10  Uros Bizjak  <ubizjak@gmail.com>

    * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
    (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
    (<sincos>mode2): New expander.
    (sincos_extend<mode>xf3_i387): Remove insn pattern.
    (sincos -> sin, cos splitters): Remove splitter patterns.
    (sincos<mode>3): Change operand 2 predicate to general_operand.
    Extend operand 2 to XFmode and generate sincosxf3 insn.
    (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
    Change operand 3 predicate to const1_operand.
    (fptan_extend<mode>xf4_i387): Remove insn pattern.
    (tanxf2): Update operands in the call to fptanxf4_i387.
    (tan<mode>2): Change operand 1 predicate to general_operand.
    Extend operand 1 to XFmode and generate tanxf3 insn.
    (atan2xf3): Rename from *fpatanxf3_i387.
    (fpatan_extend<mode>xf3_i387): Remove insn pattern.
    (atan2xf3): Remove expander.
    (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
    Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
    (atan<mode>2): Change operand 1 predicate to general_operand.
    Extend operand 1 to XFmode and generate atanxf3 insn.

Patch was bootstrapped and regression tested on x86_64-linux-gnu {,-m32}.

Committed to mainline SVN.

Uros.

Patch

Index: config/i386/i386.md
===================================================================
--- config/i386/i386.md	(revision 264185)
+++ config/i386/i386.md	(working copy)
@@ -15375,7 +15375,7 @@ 
 	[(UNSPEC_SIN "sin")
 	 (UNSPEC_COS "cos")])
 
-(define_insn "*<sincos>xf2_i387"
+(define_insn "<sincos>xf2"
   [(set (match_operand:XF 0 "register_operand" "=f")
 	(unspec:XF [(match_operand:XF 1 "register_operand" "0")]
 		   SINCOS))]
@@ -15386,25 +15386,23 @@ 
    (set_attr "znver1_decode" "vector")
    (set_attr "mode" "XF")])
 
-(define_insn "*<sincos>_extend<mode>xf2_i387"
-  [(set (match_operand:XF 0 "register_operand" "=f")
-	(unspec:XF [(float_extend:XF
-		      (match_operand:MODEF 1 "register_operand" "0"))]
-		   SINCOS))]
+(define_expand "<sincos><mode>2"
+  [(set (match_operand:MODEF 0 "register_operand")
+	(unspec:MODEF [(match_operand:MODEF 1 "general_operand")]
+		      SINCOS))]
   "TARGET_USE_FANCY_MATH_387
    && (!(SSE_FLOAT_MODE_P (<MODE>mode) && TARGET_SSE_MATH)
        || TARGET_MIX_SSE_I387)
    && flag_unsafe_math_optimizations"
-  "f<sincos>"
-  [(set_attr "type" "fpspc")
-   (set_attr "znver1_decode" "vector")
-   (set_attr "mode" "XF")])
+{
+  rtx op0 = gen_reg_rtx (XFmode);
+  rtx op1 = gen_reg_rtx (XFmode);
 
-;; When sincos pattern is defined, sin and cos builtin functions will be
-;; expanded to sincos pattern with one of its outputs left unused.
-;; CSE pass will figure out if two sincos patterns can be combined,
-;; otherwise sincos pattern will be split back to sin or cos pattern,
-;; depending on the unused output.
+  emit_insn (gen_extend<mode>xf2 (op1, operands[1]));
+  emit_insn (gen_<sincos>xf2 (op0, op1));
+  emit_insn (gen_truncxf<mode>2 (operands[0], op0));
+  DONE;
+})
 
 (define_insn "sincosxf3"
   [(set (match_operand:XF 0 "register_operand" "=f")
@@ -15419,70 +15417,10 @@ 
    (set_attr "znver1_decode" "vector")
    (set_attr "mode" "XF")])
 
-(define_split
-  [(set (match_operand:XF 0 "register_operand")
-	(unspec:XF [(match_operand:XF 2 "register_operand")]
-		   UNSPEC_SINCOS_COS))
-   (set (match_operand:XF 1 "register_operand")
-	(unspec:XF [(match_dup 2)] UNSPEC_SINCOS_SIN))]
-  "find_regno_note (insn, REG_UNUSED, REGNO (operands[0]))
-   && can_create_pseudo_p ()"
-  [(set (match_dup 1) (unspec:XF [(match_dup 2)] UNSPEC_SIN))])
-
-(define_split
-  [(set (match_operand:XF 0 "register_operand")
-	(unspec:XF [(match_operand:XF 2 "register_operand")]
-		   UNSPEC_SINCOS_COS))
-   (set (match_operand:XF 1 "register_operand")
-	(unspec:XF [(match_dup 2)] UNSPEC_SINCOS_SIN))]
-  "find_regno_note (insn, REG_UNUSED, REGNO (operands[1]))
-   && can_create_pseudo_p ()"
-  [(set (match_dup 0) (unspec:XF [(match_dup 2)] UNSPEC_COS))])
-
-(define_insn "sincos_extend<mode>xf3_i387"
-  [(set (match_operand:XF 0 "register_operand" "=f")
-	(unspec:XF [(float_extend:XF
-		      (match_operand:MODEF 2 "register_operand" "0"))]
-		   UNSPEC_SINCOS_COS))
-   (set (match_operand:XF 1 "register_operand" "=u")
-        (unspec:XF [(float_extend:XF (match_dup 2))] UNSPEC_SINCOS_SIN))]
-  "TARGET_USE_FANCY_MATH_387
-   && (!(SSE_FLOAT_MODE_P (<MODE>mode) && TARGET_SSE_MATH)
-       || TARGET_MIX_SSE_I387)
-   && flag_unsafe_math_optimizations"
-  "fsincos"
-  [(set_attr "type" "fpspc")
-   (set_attr "znver1_decode" "vector")
-   (set_attr "mode" "XF")])
-
-(define_split
-  [(set (match_operand:XF 0 "register_operand")
-	(unspec:XF [(float_extend:XF
-		      (match_operand:MODEF 2 "register_operand"))]
-		   UNSPEC_SINCOS_COS))
-   (set (match_operand:XF 1 "register_operand")
-	(unspec:XF [(float_extend:XF (match_dup 2))] UNSPEC_SINCOS_SIN))]
-  "find_regno_note (insn, REG_UNUSED, REGNO (operands[0]))
-   && can_create_pseudo_p ()"
-  [(set (match_dup 1)
-	(unspec:XF [(float_extend:XF (match_dup 2))] UNSPEC_SIN))])
-
-(define_split
-  [(set (match_operand:XF 0 "register_operand")
-	(unspec:XF [(float_extend:XF
-		      (match_operand:MODEF 2 "register_operand"))]
-		   UNSPEC_SINCOS_COS))
-   (set (match_operand:XF 1 "register_operand")
-	(unspec:XF [(float_extend:XF (match_dup 2))] UNSPEC_SINCOS_SIN))]
-  "find_regno_note (insn, REG_UNUSED, REGNO (operands[1]))
-   && can_create_pseudo_p ()"
-  [(set (match_dup 0)
-	(unspec:XF [(float_extend:XF (match_dup 2))] UNSPEC_COS))])
-
 (define_expand "sincos<mode>3"
   [(use (match_operand:MODEF 0 "register_operand"))
    (use (match_operand:MODEF 1 "register_operand"))
-   (use (match_operand:MODEF 2 "register_operand"))]
+   (use (match_operand:MODEF 2 "general_operand"))]
   "TARGET_USE_FANCY_MATH_387
    && (!(SSE_FLOAT_MODE_P (<MODE>mode) && TARGET_SSE_MATH)
        || TARGET_MIX_SSE_I387)
@@ -15490,8 +15428,10 @@ 
 {
   rtx op0 = gen_reg_rtx (XFmode);
   rtx op1 = gen_reg_rtx (XFmode);
+  rtx op2 = gen_reg_rtx (XFmode);
 
-  emit_insn (gen_sincos_extend<mode>xf3_i387 (op0, op1, operands[2]));
+  emit_insn (gen_extend<mode>xf2 (op2, operands[2]));
+  emit_insn (gen_sincosxf3 (op0, op1, op2));
   emit_insn (gen_truncxf<mode>2 (operands[0], op0));
   emit_insn (gen_truncxf<mode>2 (operands[1], op1));
   DONE;
@@ -15498,36 +15438,18 @@ 
 })
 
 (define_insn "fptanxf4_i387"
-  [(set (match_operand:XF 0 "register_operand" "=f")
-	(match_operand:XF 3 "const_double_operand" "F"))
+  [(set (match_operand:SF 0 "register_operand" "=f")
+	(match_operand:SF 3 "const1_operand"))
    (set (match_operand:XF 1 "register_operand" "=u")
         (unspec:XF [(match_operand:XF 2 "register_operand" "0")]
 		   UNSPEC_TAN))]
   "TARGET_USE_FANCY_MATH_387
-   && flag_unsafe_math_optimizations
-   && standard_80387_constant_p (operands[3]) == 2"
+   && flag_unsafe_math_optimizations"
   "fptan"
   [(set_attr "type" "fpspc")
    (set_attr "znver1_decode" "vector")
    (set_attr "mode" "XF")])
 
-(define_insn "fptan_extend<mode>xf4_i387"
-  [(set (match_operand:MODEF 0 "register_operand" "=f")
-	(match_operand:MODEF 3 "const_double_operand" "F"))
-   (set (match_operand:XF 1 "register_operand" "=u")
-        (unspec:XF [(float_extend:XF
-		      (match_operand:MODEF 2 "register_operand" "0"))]
-		   UNSPEC_TAN))]
-  "TARGET_USE_FANCY_MATH_387
-   && (!(SSE_FLOAT_MODE_P (<MODE>mode) && TARGET_SSE_MATH)
-       || TARGET_MIX_SSE_I387)
-   && flag_unsafe_math_optimizations
-   && standard_80387_constant_p (operands[3]) == 2"
-  "fptan"
-  [(set_attr "type" "fpspc")
-   (set_attr "znver1_decode" "vector")
-   (set_attr "mode" "XF")])
-
 (define_expand "tanxf2"
   [(use (match_operand:XF 0 "register_operand"))
    (use (match_operand:XF 1 "register_operand"))]
@@ -15534,16 +15456,15 @@ 
   "TARGET_USE_FANCY_MATH_387
    && flag_unsafe_math_optimizations"
 {
-  rtx one = gen_reg_rtx (XFmode);
-  rtx op2 = CONST1_RTX (XFmode); /* fld1 */
-
-  emit_insn (gen_fptanxf4_i387 (one, operands[0], operands[1], op2));
+  rtx one = gen_reg_rtx (SFmode);
+  emit_insn (gen_fptanxf4_i387 (one, operands[0], operands[1],
+				CONST1_RTX (SFmode)));
   DONE;
 })
 
 (define_expand "tan<mode>2"
   [(use (match_operand:MODEF 0 "register_operand"))
-   (use (match_operand:MODEF 1 "register_operand"))]
+   (use (match_operand:MODEF 1 "general_operand"))]
   "TARGET_USE_FANCY_MATH_387
    && (!(SSE_FLOAT_MODE_P (<MODE>mode) && TARGET_SSE_MATH)
        || TARGET_MIX_SSE_I387)
@@ -15550,17 +15471,15 @@ 
    && flag_unsafe_math_optimizations"
 {
   rtx op0 = gen_reg_rtx (XFmode);
+  rtx op1 = gen_reg_rtx (XFmode);
 
-  rtx one = gen_reg_rtx (<MODE>mode);
-  rtx op2 = CONST1_RTX (<MODE>mode); /* fld1 */
-
-  emit_insn (gen_fptan_extend<mode>xf4_i387 (one, op0,
-					     operands[1], op2));
+  emit_insn (gen_extend<mode>xf2 (op1, operands[1]));
+  emit_insn (gen_tanxf2 (op0, op1));
   emit_insn (gen_truncxf<mode>2 (operands[0], op0));
   DONE;
 })
 
-(define_insn "*fpatanxf3_i387"
+(define_insn "atan2xf3"
   [(set (match_operand:XF 0 "register_operand" "=f")
         (unspec:XF [(match_operand:XF 1 "register_operand" "0")
 	            (match_operand:XF 2 "register_operand" "u")]
@@ -15573,36 +15492,10 @@ 
    (set_attr "znver1_decode" "vector")
    (set_attr "mode" "XF")])
 
-(define_insn "fpatan_extend<mode>xf3_i387"
-  [(set (match_operand:XF 0 "register_operand" "=f")
-        (unspec:XF [(float_extend:XF
-		      (match_operand:MODEF 1 "register_operand" "0"))
-		    (float_extend:XF
-		      (match_operand:MODEF 2 "register_operand" "u"))]
-	           UNSPEC_FPATAN))
-   (clobber (match_scratch:XF 3 "=2"))]
-  "TARGET_USE_FANCY_MATH_387
-   && (!(SSE_FLOAT_MODE_P (<MODE>mode) && TARGET_SSE_MATH)
-       || TARGET_MIX_SSE_I387)
-   && flag_unsafe_math_optimizations"
-  "fpatan"
-  [(set_attr "type" "fpspc")
-   (set_attr "znver1_decode" "vector")
-   (set_attr "mode" "XF")])
-
-(define_expand "atan2xf3"
-  [(parallel [(set (match_operand:XF 0 "register_operand")
-		   (unspec:XF [(match_operand:XF 2 "register_operand")
-			       (match_operand:XF 1 "register_operand")]
-			      UNSPEC_FPATAN))
-	      (clobber (match_scratch:XF 3))])]
-  "TARGET_USE_FANCY_MATH_387
-   && flag_unsafe_math_optimizations")
-
 (define_expand "atan2<mode>3"
   [(use (match_operand:MODEF 0 "register_operand"))
-   (use (match_operand:MODEF 1 "register_operand"))
-   (use (match_operand:MODEF 2 "register_operand"))]
+   (use (match_operand:MODEF 1 "general_operand"))
+   (use (match_operand:MODEF 2 "general_operand"))]
   "TARGET_USE_FANCY_MATH_387
    && (!(SSE_FLOAT_MODE_P (<MODE>mode) && TARGET_SSE_MATH)
        || TARGET_MIX_SSE_I387)
@@ -15609,8 +15502,13 @@ 
    && flag_unsafe_math_optimizations"
 {
   rtx op0 = gen_reg_rtx (XFmode);
+  rtx op1 = gen_reg_rtx (XFmode);
+  rtx op2 = gen_reg_rtx (XFmode);
 
-  emit_insn (gen_fpatan_extend<mode>xf3_i387 (op0, operands[2], operands[1]));
+  emit_insn (gen_extend<mode>xf2 (op2, operands[2]));
+  emit_insn (gen_extend<mode>xf2 (op1, operands[1]));
+
+  emit_insn (gen_atan2xf3 (op0, op2, op1));
   emit_insn (gen_truncxf<mode>2 (operands[0], op0));
   DONE;
 })
@@ -15623,14 +15521,11 @@ 
 	      (clobber (match_scratch:XF 3))])]
   "TARGET_USE_FANCY_MATH_387
    && flag_unsafe_math_optimizations"
-{
-  operands[2] = gen_reg_rtx (XFmode);
-  emit_move_insn (operands[2], CONST1_RTX (XFmode));  /* fld1 */
-})
+  "operands[2] = force_reg (XFmode, CONST1_RTX (XFmode));")
 
 (define_expand "atan<mode>2"
   [(use (match_operand:MODEF 0 "register_operand"))
-   (use (match_operand:MODEF 1 "register_operand"))]
+   (use (match_operand:MODEF 1 "general_operand"))]
   "TARGET_USE_FANCY_MATH_387
    && (!(SSE_FLOAT_MODE_P (<MODE>mode) && TARGET_SSE_MATH)
        || TARGET_MIX_SSE_I387)
@@ -15637,11 +15532,10 @@ 
    && flag_unsafe_math_optimizations"
 {
   rtx op0 = gen_reg_rtx (XFmode);
+  rtx op1 = gen_reg_rtx (XFmode);
 
-  rtx op2 = gen_reg_rtx (<MODE>mode);
-  emit_move_insn (op2, CONST1_RTX (<MODE>mode));  /* fld1 */
-
-  emit_insn (gen_fpatan_extend<mode>xf3_i387 (op0, op2, operands[1]));
+  emit_insn (gen_extend<mode>xf2 (op1, operands[1]));
+  emit_insn (gen_atanxf2 (op0, op1));
   emit_insn (gen_truncxf<mode>2 (operands[0], op0));
   DONE;
 })
@@ -15664,7 +15558,7 @@ 
   for (i = 2; i < 6; i++)
     operands[i] = gen_reg_rtx (XFmode);
 
-  emit_move_insn (operands[3], CONST1_RTX (XFmode));  /* fld1 */
+  operands[3] = force_reg (XFmode, CONST1_RTX (XFmode));
 })
 
 (define_expand "asin<mode>2"
@@ -15702,7 +15596,7 @@ 
   for (i = 2; i < 6; i++)
     operands[i] = gen_reg_rtx (XFmode);
 
-  emit_move_insn (operands[3], CONST1_RTX (XFmode));  /* fld1 */
+  operands[3] = force_reg (XFmode, CONST1_RTX (XFmode));
 })
 
 (define_expand "acos<mode>2"