re PR target/82498 (Missed optimization for x86 rotate instruction)

	PR target/82498
	* config/i386/i386.md (*ashl<mode>3_mask_1,
	*<shift_insn><mode>3_mask_1, *<rotate_insn><mode>3_mask_1,
	*<btsc><mode>_mask_1, *btr<mode>_mask_1): New define_insn_and_split
	patterns.

	* gcc.target/i386/pr82498-1.c: New test.
	* gcc.target/i386/pr82498-2.c: New test.

From-SVN: r253695

5 files changed, 220 insertions, 0 deletions

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 92cdc5f..a9dbfce 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,11 @@
+2017-10-12  Jakub Jelinek  <jakub@redhat.com>
+
+	PR target/82498
+	* config/i386/i386.md (*ashl<mode>3_mask_1,
+	*<shift_insn><mode>3_mask_1, *<rotate_insn><mode>3_mask_1,
+	*<btsc><mode>_mask_1, *btr<mode>_mask_1): New define_insn_and_split
+	patterns.
+
 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
 
 	* profile-count.h (safe_scale_64bit): Fix GCC4.x path.
diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md
index 9e1f85f..2fa982c 100644
--- a/gcc/config/i386/i386.md
+++ b/gcc/config/i386/i386.md
@@ -10228,6 +10228,26 @@
       (clobber (reg:CC FLAGS_REG))])]
   "operands[2] = gen_lowpart (QImode, operands[2]);")
 
+(define_insn_and_split "*ashl<mode>3_mask_1"
+  [(set (match_operand:SWI48 0 "nonimmediate_operand")
+	(ashift:SWI48
+	  (match_operand:SWI48 1 "nonimmediate_operand")
+	  (and:QI
+	    (match_operand:QI 2 "register_operand")
+	    (match_operand:QI 3 "const_int_operand"))))
+   (clobber (reg:CC FLAGS_REG))]
+  "ix86_binary_operator_ok (ASHIFT, <MODE>mode, operands)
+   && (INTVAL (operands[3]) & (GET_MODE_BITSIZE (<MODE>mode)-1))
+      == GET_MODE_BITSIZE (<MODE>mode)-1
+   && can_create_pseudo_p ()"
+  "#"
+  "&& 1"
+  [(parallel
+     [(set (match_dup 0)
+	   (ashift:SWI48 (match_dup 1)
+			 (match_dup 2)))
+      (clobber (reg:CC FLAGS_REG))])])
+
 (define_insn "*bmi2_ashl<mode>3_1"
   [(set (match_operand:SWI48 0 "register_operand" "=r")
 	(ashift:SWI48 (match_operand:SWI48 1 "nonimmediate_operand" "rm")
@@ -10728,6 +10748,26 @@
       (clobber (reg:CC FLAGS_REG))])]
   "operands[2] = gen_lowpart (QImode, operands[2]);")
 
+(define_insn_and_split "*<shift_insn><mode>3_mask_1"
+  [(set (match_operand:SWI48 0 "nonimmediate_operand")
+	(any_shiftrt:SWI48
+	  (match_operand:SWI48 1 "nonimmediate_operand")
+	  (and:QI
+	    (match_operand:QI 2 "register_operand")
+	    (match_operand:QI 3 "const_int_operand"))))
+   (clobber (reg:CC FLAGS_REG))]
+  "ix86_binary_operator_ok (<CODE>, <MODE>mode, operands)
+   && (INTVAL (operands[3]) & (GET_MODE_BITSIZE (<MODE>mode)-1))
+      == GET_MODE_BITSIZE (<MODE>mode)-1
+   && can_create_pseudo_p ()"
+  "#"
+  "&& 1"
+  [(parallel
+     [(set (match_dup 0)
+	   (any_shiftrt:SWI48 (match_dup 1)
+			      (match_dup 2)))
+      (clobber (reg:CC FLAGS_REG))])])
+
 (define_insn_and_split "*<shift_insn><mode>3_doubleword"
   [(set (match_operand:DWI 0 "register_operand" "=&r")
 	(any_shiftrt:DWI (match_operand:DWI 1 "register_operand" "0")
@@ -11187,6 +11227,26 @@
       (clobber (reg:CC FLAGS_REG))])]
   "operands[2] = gen_lowpart (QImode, operands[2]);")
 
+(define_insn_and_split "*<rotate_insn><mode>3_mask_1"
+  [(set (match_operand:SWI48 0 "nonimmediate_operand")
+	(any_rotate:SWI48
+	  (match_operand:SWI48 1 "nonimmediate_operand")
+	  (and:QI
+	    (match_operand:QI 2 "register_operand")
+	    (match_operand:QI 3 "const_int_operand"))))
+   (clobber (reg:CC FLAGS_REG))]
+  "ix86_binary_operator_ok (<CODE>, <MODE>mode, operands)
+   && (INTVAL (operands[3]) & (GET_MODE_BITSIZE (<MODE>mode)-1))
+      == GET_MODE_BITSIZE (<MODE>mode)-1
+   && can_create_pseudo_p ()"
+  "#"
+  "&& 1"
+  [(parallel
+     [(set (match_dup 0)
+	   (any_rotate:SWI48 (match_dup 1)
+			     (match_dup 2)))
+      (clobber (reg:CC FLAGS_REG))])])
+
 ;; Implement rotation using two double-precision
 ;; shift instructions and a scratch register.
 
@@ -11494,6 +11554,30 @@
       (clobber (reg:CC FLAGS_REG))])]
   "operands[1] = gen_lowpart (QImode, operands[1]);")
 
+(define_insn_and_split "*<btsc><mode>_mask_1"
+  [(set (match_operand:SWI48 0 "register_operand")
+	(any_or:SWI48
+	  (ashift:SWI48
+	    (const_int 1)
+	    (and:QI
+	      (match_operand:QI 1 "register_operand")
+	      (match_operand:QI 2 "const_int_operand")))
+	  (match_operand:SWI48 3 "register_operand")))
+   (clobber (reg:CC FLAGS_REG))]
+  "TARGET_USE_BT
+   && (INTVAL (operands[2]) & (GET_MODE_BITSIZE (<MODE>mode)-1))
+      == GET_MODE_BITSIZE (<MODE>mode)-1
+   && can_create_pseudo_p ()"
+  "#"
+  "&& 1"
+  [(parallel
+     [(set (match_dup 0)
+	   (any_or:SWI48
+	     (ashift:SWI48 (const_int 1)
+			   (match_dup 1))
+	     (match_dup 3)))
+      (clobber (reg:CC FLAGS_REG))])])
+
 (define_insn "*btr<mode>"
   [(set (match_operand:SWI48 0 "register_operand" "=r")
 	(and:SWI48
@@ -11535,6 +11619,30 @@
       (clobber (reg:CC FLAGS_REG))])]
   "operands[1] = gen_lowpart (QImode, operands[1]);")
 
+(define_insn_and_split "*btr<mode>_mask_1"
+  [(set (match_operand:SWI48 0 "register_operand")
+	(and:SWI48
+	  (rotate:SWI48
+	    (const_int -2)
+	    (and:QI
+	      (match_operand:QI 1 "register_operand")
+	      (match_operand:QI 2 "const_int_operand")))
+	  (match_operand:SWI48 3 "register_operand")))
+   (clobber (reg:CC FLAGS_REG))]
+  "TARGET_USE_BT
+   && (INTVAL (operands[2]) & (GET_MODE_BITSIZE (<MODE>mode)-1))
+      == GET_MODE_BITSIZE (<MODE>mode)-1
+   && can_create_pseudo_p ()"
+  "#"
+  "&& 1"
+  [(parallel
+     [(set (match_dup 0)
+	   (and:SWI48
+	     (rotate:SWI48 (const_int -2)
+			   (match_dup 1))
+	     (match_dup 3)))
+      (clobber (reg:CC FLAGS_REG))])])
+
 ;; These instructions are never faster than the corresponding
 ;; and/ior/xor operations when using immediate operand, so with
 ;; 32-bit there's no point.  But in 64-bit, we can't hold the
diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index c804364..9e77fa6 100644
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,9 @@
+2017-10-12  Jakub Jelinek  <jakub@redhat.com>
+
+	PR target/82498
+	* gcc.target/i386/pr82498-1.c: New test.
+	* gcc.target/i386/pr82498-2.c: New test.
+
 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
 
 	* gcc.dg/predict-13.c: Update template for probaility change.
diff --git a/gcc/testsuite/gcc.target/i386/pr82498-1.c b/gcc/testsuite/gcc.target/i386/pr82498-1.c
new file mode 100644
index 0000000..78a6698
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr82498-1.c
@@ -0,0 +1,52 @@
+/* PR target/82498 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -mtune=generic -masm=att" } */
+/* { dg-final { scan-assembler-not {\mand[bwlq]\M} } } */
+
+unsigned
+f1 (unsigned x, unsigned char y)
+{
+  if (y == 0)
+    return x;
+  y &= __CHAR_BIT__ * __SIZEOF_INT__ - 1;
+  return (x << y) | (x >> (__CHAR_BIT__ * __SIZEOF_INT__ - y));
+}
+
+unsigned
+f2 (unsigned x, unsigned y)
+{
+  if (y == 0)
+    return x;
+  y &= __CHAR_BIT__ * __SIZEOF_INT__ - 1;
+  return (x << y) | (x >> (__CHAR_BIT__ * __SIZEOF_INT__ - y));
+}
+
+unsigned
+f3 (unsigned x, unsigned short y)
+{
+  if (y == 0)
+    return x;
+  y &= __CHAR_BIT__ * __SIZEOF_INT__ - 1;
+  return (x << y) | (x >> (__CHAR_BIT__ * __SIZEOF_INT__ - y));
+}
+
+unsigned
+f4 (unsigned x, unsigned char y)
+{
+  y &= __CHAR_BIT__ * __SIZEOF_INT__ - 1;
+  return (x << y) | (x >> (-y & (__CHAR_BIT__ * __SIZEOF_INT__ - 1)));
+}
+
+unsigned
+f5 (unsigned x, unsigned int y)
+{
+  y &= __CHAR_BIT__ * __SIZEOF_INT__ - 1;
+  return (x << y) | (x >> (-y & (__CHAR_BIT__ * __SIZEOF_INT__ - 1)));
+}
+
+unsigned
+f6 (unsigned x, unsigned short y)
+{
+  y &= __CHAR_BIT__ * __SIZEOF_INT__ - 1;
+  return (x << y) | (x >> (-y & (__CHAR_BIT__ * __SIZEOF_INT__ - 1)));
+}
diff --git a/gcc/testsuite/gcc.target/i386/pr82498-2.c b/gcc/testsuite/gcc.target/i386/pr82498-2.c
new file mode 100644
index 0000000..9e065ee
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr82498-2.c
@@ -0,0 +1,46 @@
+/* PR target/82498 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -mtune=generic -masm=att" } */
+/* { dg-final { scan-assembler-not {\mand[bwlq]\M} } } */
+
+int
+f1 (int x, unsigned char y)
+{
+  y &= __CHAR_BIT__ * __SIZEOF_INT__ - 1;
+  return x >> y;
+}
+
+unsigned
+f2 (unsigned x, unsigned char y)
+{
+  y &= __CHAR_BIT__ * __SIZEOF_INT__ - 1;
+  return x >> y;
+}
+
+unsigned
+f3 (unsigned x, unsigned char y)
+{
+  y &= __CHAR_BIT__ * __SIZEOF_INT__ - 1;
+  return x << y;
+}
+
+unsigned
+f4 (unsigned x, unsigned char y)
+{
+  y &= __CHAR_BIT__ * __SIZEOF_INT__ - 1;
+  return x | (1U << y);
+}
+
+unsigned
+f5 (unsigned x, unsigned char y)
+{
+  y &= __CHAR_BIT__ * __SIZEOF_INT__ - 1;
+  return x ^ (1U << y);
+}
+
+unsigned
+f6 (unsigned x, unsigned char y)
+{
+  y &= __CHAR_BIT__ * __SIZEOF_INT__ - 1;
+  return (x + 2) & ~(1U << y);
+}