[gcc/devel/c++-coroutines] i386: Optimize popcnt followed by zero/sign extension [PR91824]

[Iain D Sandoe]

https://gcc.gnu.org/g:b285bebe6ad1e9f6416f0eb6cb69edc44db7813c

commit b285bebe6ad1e9f6416f0eb6cb69edc44db7813c
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Thu Jan 30 09:39:05 2020 +0100

    i386: Optimize popcnt followed by zero/sign extension [PR91824]
    
    Like any other instruction with 32-bit GPR destination operand in 64-bit
    mode, popcntl also clears the upper 32 bits of the register (and other bits
    too, it can return only 0 to 32 inclusive).
    
    During combine, the zero or sign extensions of it show up as paradoxical
    subreg of the popcount & 63, there 63 is the smallest power of two - 1 mask
    that can represent all the 0 to 32 inclusive values.
    
    2020-01-30  Jakub Jelinek  <jakub@redhat.com>
    
    	PR target/91824
    	* config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
    	(*popcountsi2_zext_falsedep): New define_insn.
    
    	* gcc.target/i386/pr91824-1.c: New test.

Diff:
---
 gcc/ChangeLog                             |  6 ++++
 gcc/config/i386/i386.md                   | 54 +++++++++++++++++++++++++++++++
 gcc/testsuite/ChangeLog                   |  5 +++
 gcc/testsuite/gcc.target/i386/pr91824-1.c | 54 +++++++++++++++++++++++++++++++
 4 files changed, 119 insertions(+)

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index a36e732..c86b9c2 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,9 @@
+2020-01-30  Jakub Jelinek  <jakub@redhat.com>
+
+	PR target/91824
+	* config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
+	(*popcountsi2_zext_falsedep): New define_insn.
+
 2020-01-30  Dragan Mladjenovic  <dmladjenovic@wavecomp.com>
 
 	* config.in: Regenerated.
diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md
index f83b370..f5c8d55 100644
--- a/gcc/config/i386/i386.md
+++ b/gcc/config/i386/i386.md
@@ -14563,6 +14563,60 @@
    (set_attr "type" "bitmanip")
    (set_attr "mode" "<MODE>")])
 
+(define_insn_and_split "*popcountsi2_zext"
+  [(set (match_operand:DI 0 "register_operand" "=r")
+	(and:DI
+	  (subreg:DI
+	    (popcount:SI
+	      (match_operand:SI 1 "nonimmediate_operand" "rm")) 0)
+	  (const_int 63)))
+   (clobber (reg:CC FLAGS_REG))]
+  "TARGET_POPCNT && TARGET_64BIT"
+{
+#if TARGET_MACHO
+  return "popcnt\t{%1, %k0|%k0, %1}";
+#else
+  return "popcnt{l}\t{%1, %k0|%k0, %1}";
+#endif
+}
+  "&& TARGET_AVOID_FALSE_DEP_FOR_BMI && epilogue_completed
+   && optimize_function_for_speed_p (cfun)
+   && !reg_mentioned_p (operands[0], operands[1])"
+  [(parallel
+    [(set (match_dup 0)
+	  (and:DI (subreg:DI (popcount:SI (match_dup 1)) 0) (const_int 63)))
+     (unspec [(match_dup 0)] UNSPEC_INSN_FALSE_DEP)
+     (clobber (reg:CC FLAGS_REG))])]
+  "ix86_expand_clear (operands[0]);"
+  [(set_attr "prefix_rep" "1")
+   (set_attr "type" "bitmanip")
+   (set_attr "mode" "SI")])
+
+; False dependency happens when destination is only updated by tzcnt,
+; lzcnt or popcnt.  There is no false dependency when destination is
+; also used in source.
+(define_insn "*popcountsi2_zext_falsedep"
+  [(set (match_operand:DI 0 "register_operand" "=r")
+	(and:DI
+	  (subreg:DI
+	    (popcount:SI
+	      (match_operand:SI 1 "nonimmediate_operand" "rm")) 0)
+	  (const_int 63)))
+   (unspec [(match_operand:DI 2 "register_operand" "0")]
+	   UNSPEC_INSN_FALSE_DEP)
+   (clobber (reg:CC FLAGS_REG))]
+  "TARGET_POPCNT && TARGET_64BIT"
+{
+#if TARGET_MACHO
+  return "popcnt\t{%1, %k0|%k0, %1}";
+#else
+  return "popcnt{l}\t{%1, %k0|%k0, %1}";
+#endif
+}
+  [(set_attr "prefix_rep" "1")
+   (set_attr "type" "bitmanip")
+   (set_attr "mode" "SI")])
+
 (define_insn_and_split "*popcounthi2_1"
   [(set (match_operand:SI 0 "register_operand")
 	(popcount:SI
diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index 30e804b..9b36606 100644
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,8 @@
+2020-01-30  Jakub Jelinek  <jakub@redhat.com>
+
+	PR target/91824
+	* gcc.target/i386/pr91824-1.c: New test.
+
 2020-01-30  Bin Cheng  <bin.cheng@linux.alibaba.com>
 
 	* g++.dg/coroutines/co-await-syntax-09-convert.C: New test.
diff --git a/gcc/testsuite/gcc.target/i386/pr91824-1.c b/gcc/testsuite/gcc.target/i386/pr91824-1.c
new file mode 100644
index 0000000..0bb24e7
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr91824-1.c
@@ -0,0 +1,54 @@
+/* PR target/91824 */
+/* { dg-do compile { target lp64 } } */
+/* { dg-options "-O2 -mpopcnt" } */
+/* { dg-final { scan-assembler-not "cltq" } } */
+
+unsigned int foo (void);
+
+unsigned long
+f1 (unsigned int x)
+{
+  return __builtin_popcount (x);
+}
+
+unsigned long
+f2 (unsigned int x)
+{
+  return (unsigned) __builtin_popcount (x);
+}
+
+unsigned long
+f3 (unsigned int x)
+{
+  return __builtin_popcount (x) & 63ULL;
+}
+
+unsigned long
+f4 (unsigned int x)
+{
+  return __builtin_popcount (x) & 1023ULL;
+}
+
+unsigned long
+f5 (void)
+{
+  return __builtin_popcount (foo ());
+}
+
+unsigned long
+f6 (void)
+{
+  return (unsigned) __builtin_popcount (foo ());
+}
+
+unsigned long
+f7 (void)
+{
+  return __builtin_popcount (foo ()) & 63ULL;
+}
+
+unsigned long
+f8 (void)
+{
+  return __builtin_popcount (foo ()) & 1023ULL;
+}