Hi!
The following peephole2 changes:
- addl (%rdi), %esi
+ xorl %eax, %eax
+ addl %esi, (%rdi)
setc %al
- movl %esi, (%rdi)
- movzbl %al, %eax
ret
on the testcase. *add<mode>3_cc_overflow_1, being an add{l,q} insn, is
commutative, so if TARGET_READ_MODIFY_WRITE we can replace
addl (%rdi), %esi; movl %esi, (%rdi)
with
addl %esi, (%rdi)
if %edi is dead after those two insns.
Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
2020-05-08 Jakub Jelinek <ja...@redhat.com>
PR target/94857
* config/i386/i386.md (peephole2 after *add<mode>3_cc_overflow_1): New
define_peephole2.
* gcc.target/i386/pr94857.c: New test.
--- gcc/config/i386/i386.md.jj 2020-05-06 20:03:38.489391717 +0200
+++ gcc/config/i386/i386.md 2020-05-07 19:59:45.650521607 +0200
@@ -6992,6 +6992,23 @@ (define_insn "*add<mode>3_cc_overflow_1"
[(set_attr "type" "alu")
(set_attr "mode" "<MODE>")])
+(define_peephole2
+ [(parallel [(set (reg:CCC FLAGS_REG)
+ (compare:CCC
+ (plus:SWI (match_operand:SWI 0 "general_reg_operand")
+ (match_operand:SWI 1 "memory_operand"))
+ (match_dup 0)))
+ (set (match_dup 0) (plus:SWI (match_dup 0) (match_dup 1)))])
+ (set (match_dup 1) (match_dup 0))]
+ "(TARGET_READ_MODIFY_WRITE || optimize_insn_for_size_p ())
+ && peep2_reg_dead_p (2, operands[0])
+ && !reg_overlap_mentioned_p (operands[0], operands[1])"
+ [(parallel [(set (reg:CCC FLAGS_REG)
+ (compare:CCC
+ (plus:SWI (match_dup 1) (match_dup 0))
+ (match_dup 1)))
+ (set (match_dup 1) (plus:SWI (match_dup 1) (match_dup 0)))])])
+
(define_insn "*addsi3_zext_cc_overflow_1"
[(set (reg:CCC FLAGS_REG)
(compare:CCC
--- gcc/testsuite/gcc.target/i386/pr94857.c.jj 2020-05-07 20:09:16.002907217
+0200
+++ gcc/testsuite/gcc.target/i386/pr94857.c 2020-05-07 20:08:46.533351730
+0200
@@ -0,0 +1,15 @@
+/* PR target/94857 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -mtune=skylake -masm=att" } */
+/* { dg-final { scan-assembler "\taddl\t%\[a-z0-9]\*, \\\(" } } */
+
+#ifndef __x86_64__
+__attribute__((regparm(2)))
+#endif
+int
+foo (unsigned *p, unsigned x)
+{
+ unsigned u = *p;
+ *p += x;
+ return u > *p;
+}
Jakub
