[gcc r12-7616] PR tree-optimization/98335: New peephole2 xorl; movb -> movzbl

[Roger Sayle <sayle@gcc.gnu.org>]

https://gcc.gnu.org/g:251ea6dfbdb4448875e41081682bb3aa451b5729

commit r12-7616-g251ea6dfbdb4448875e41081682bb3aa451b5729
Author: Roger Sayle <roger@nextmovesoftware.com>
Date:   Fri Mar 11 17:57:12 2022 +0000

    PR tree-optimization/98335: New peephole2 xorl;movb -> movzbl
    
    This patch is the backend piece of my proposed fix to PR tree-opt/98335,
    to allow C++ partial struct initialization to be as efficient/optimized
    as full struct initialization.
    
    With the middle-end patch just posted to gcc-patches, the test case
    in the PR compiles on x86_64-pc-linux-gnu with -O2 to:
    
            xorl    %eax, %eax
            movb    c(%rip), %al
            ret
    
    with this additional peephole2 (actually four peephole2s):
    
            movzbl  c(%rip), %eax
            ret
    
    2022-03-11  Roger Sayle  <roger@nextmovesoftware.com>
    
    gcc/ChangeLog
            PR tree-optimization/98335
            * config/i386/i386.md (peephole2): Eliminate redundant insv.
            Combine movl followed by movb.  Transform xorl followed by
            a suitable movb or movw into the equivalent movz[bw]l.
    
    gcc/testsuite/ChangeLog
            PR tree-optimization/98335
            * g++.target/i386/pr98335.C: New test case.
            * gcc.target/i386/pr98335.c: New test case.

Diff:
---
 gcc/config/i386/i386.md                 | 50 +++++++++++++++++++++++++++++++++
 gcc/testsuite/g++.target/i386/pr98335.C | 18 ++++++++++++
 gcc/testsuite/gcc.target/i386/pr98335.c | 12 ++++++++
 3 files changed, 80 insertions(+)

diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md
index d15170e8acf..c8fbf605e41 100644
--- a/gcc/config/i386/i386.md
+++ b/gcc/config/i386/i386.md
@@ -3180,6 +3180,38 @@
 			     (const_int 8))
 	   (subreg:SWI248 (match_dup 1) 0))])
 
+;; Eliminate redundant insv, e.g. xorl %eax,%eax; movb $0, %ah
+(define_peephole2
+  [(parallel [(set (match_operand:SWI48 0 "general_reg_operand")
+		   (const_int 0))
+	      (clobber (reg:CC FLAGS_REG))])
+   (set (zero_extract:SWI248 (match_operand:SWI248 1 "general_reg_operand")
+			     (const_int 8)
+			     (const_int 8))
+	(const_int 0))]
+  "REGNO (operands[0]) == REGNO (operands[1])"
+  [(parallel [(set (match_operand:SWI48 0 "general_reg_operand")
+		   (const_int 0))
+	      (clobber (reg:CC FLAGS_REG))])])
+
+;; Combine movl followed by movb.
+(define_peephole2
+  [(set (match_operand:SWI48 0 "general_reg_operand")
+	(match_operand:SWI48 1 "const_int_operand"))
+   (set (zero_extract:SWI248 (match_operand:SWI248 2 "general_reg_operand")
+			     (const_int 8)
+			     (const_int 8))
+	(match_operand:SWI248 3 "const_int_operand"))]
+  "REGNO (operands[0]) == REGNO (operands[2])"
+  [(set (match_operand:SWI48 0 "general_reg_operand")
+	(match_dup 4))]
+{
+  HOST_WIDE_INT tmp = INTVAL (operands[1]) & ~(HOST_WIDE_INT)0xff00;
+  tmp |= (INTVAL (operands[3]) & 0xff) << 8;
+  operands[4] = gen_int_mode (tmp, <SWI48:MODE>mode);
+})
+
+
 (define_code_iterator any_extract [sign_extract zero_extract])
 
 (define_insn "*insvqi_2"
@@ -4276,6 +4308,24 @@
   [(set_attr "isa" "*,avx512dq,avx512dq")
    (set_attr "type" "imovx,mskmov,mskmov")
    (set_attr "mode" "SI,QI,QI")])
+
+;; Transform xorl; mov[bw] (set strict_low_part) into movz[bw]l.
+(define_peephole2
+  [(parallel [(set (match_operand:SWI48 0 "general_reg_operand")
+		   (const_int 0))
+	      (clobber (reg:CC FLAGS_REG))])
+   (set (strict_low_part (match_operand:SWI12 1 "general_reg_operand"))
+	(match_operand:SWI12 2 "nonimmediate_operand"))]
+  "REGNO (operands[0]) == REGNO (operands[1])"
+  [(set (match_dup 0) (zero_extend:SWI48 (match_dup 2)))])
+
+;; Likewise, but preserving FLAGS_REG.
+(define_peephole2
+  [(set (match_operand:SWI48 0 "general_reg_operand") (const_int 0))
+   (set (strict_low_part (match_operand:SWI12 1 "general_reg_operand"))
+	(match_operand:SWI12 2 "nonimmediate_operand"))]
+  "REGNO (operands[0]) == REGNO (operands[1])"
+  [(set (match_dup 0) (zero_extend:SWI48 (match_dup 2)))])
 \f
 ;; Sign extension instructions
 
diff --git a/gcc/testsuite/g++.target/i386/pr98335.C b/gcc/testsuite/g++.target/i386/pr98335.C
new file mode 100644
index 00000000000..2581b83ab6a
--- /dev/null
+++ b/gcc/testsuite/g++.target/i386/pr98335.C
@@ -0,0 +1,18 @@
+/* { dg-do compile { target { ! ia32 } } } */
+/* { dg-options "-O2" } */
+
+struct Data {
+  char a;
+  int b;
+};
+
+char c;
+
+Data val(int idx) {
+  return { c };  // { dg-warning "extended initializer" "c++ 98"  { target { c++98_only } } }
+}
+
+/* { dg-final { scan-assembler "movzbl" } } */
+/* { dg-final { scan-assembler-not "xorl" } } */
+/* { dg-final { scan-assembler-not "movb" } } */
+
diff --git a/gcc/testsuite/gcc.target/i386/pr98335.c b/gcc/testsuite/gcc.target/i386/pr98335.c
new file mode 100644
index 00000000000..7fa7ad70dce
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr98335.c
@@ -0,0 +1,12 @@
+/* { dg-do compile } */
+/* { dg-options "-O2" } */
+union Data { char a; short b; };
+
+char c;
+
+void val(void) {
+  __asm__ __volatile__ ("" : : "r" ((union Data) { c } )); } 
+
+/* { dg-final { scan-assembler "movzbl" } } */
+/* { dg-final { scan-assembler-not "xorl" } } */
+/* { dg-final { scan-assembler-not "movb" } } */