Hi!

On the following testcase, RTL ifcvt sees then_bb
(note 7 6 8 3 [bb 3] NOTE_INSN_BASIC_BLOCK)
(insn 8 7 9 3 (set (mem/c:SI (symbol_ref:DI ("b") [flags 0x2]  <var_decl 
0x7fdccf5b0cf0 b>) [1 b+0 S4 A32])
        (const_int 1 [0x1])) "pr103908.c":6:7 81 {*movsi_internal}
     (nil))
(jump_insn 9 8 13 3 (parallel [
            (asm_operands/v ("# insn 1") ("") 0 []
                 []
                 [
                    (label_ref:DI 21)
                ] pr103908.c:7)
            (clobber (reg:CC 17 flags))
        ]) "pr103908.c":7:5 -1
     (expr_list:REG_UNUSED (reg:CC 17 flags)
        (nil))
 -> 21)
and similarly else_bb (just with a different asm_operands template).
It checks that those basic blocks have a single successor and
uses last_active_insn which intentionally skips over JUMP_INSNs, sees
both basic blocks contain the same set and merges them (or if the
sets are different, attempts some other noce optimization).
But we can't assume that the jump, even when it has only a single successor,
has no side-effects.

The following patch fixes it by checking specially for asm goto, but
I wonder if it wouldn't be better to
  if (JUMP_P (BB_END (test_bb)) && !onlyjump_p (BB_END (test_bb)))
    return false;
and give up on any other hypothetical single successor jumps.

The patch below has been bootstrapped/regtested on x86_64-linux and
i686-linux, I could bootstrap the !onlyjump_p version too...

2022-01-05  Jakub Jelinek  <ja...@redhat.com>

        PR rtl-optimization/103908
        * ifcvt.c (bb_valid_for_noce_process_p): Punt on bbs ending with
        asm goto.

        * gcc.target/i386/pr103908.c: New test.

--- gcc/ifcvt.c.jj      2022-01-03 10:40:42.919140216 +0100
+++ gcc/ifcvt.c 2022-01-04 23:50:40.962529052 +0100
@@ -3064,6 +3064,13 @@ bb_valid_for_noce_process_p (basic_block
 
   if (!insn_valid_noce_process_p (last_insn, cc))
     return false;
+
+  /* Punt for blocks ending with asm goto, last_active_insn
+     ignores JUMP_INSNs.  */
+  if (JUMP_P (BB_END (test_bb))
+      && asm_noperands (PATTERN (BB_END (test_bb))) >= 0)
+    return false;
+
   last_set = single_set (last_insn);
 
   rtx x = SET_DEST (last_set);
--- gcc/testsuite/gcc.target/i386/pr103908.c.jj 2022-01-04 23:58:02.992341275 
+0100
+++ gcc/testsuite/gcc.target/i386/pr103908.c    2022-01-04 23:57:39.399671539 
+0100
@@ -0,0 +1,24 @@
+/* PR rtl-optimization/103908 */
+/* { dg-do compile } */
+/* { dg-options "-O1 -fdisable-tree-cselim -fno-tree-sink" } */
+/* { dg-final { scan-assembler "# insn 1" } } */
+/* { dg-final { scan-assembler "# insn 2" } } */
+
+int a, b;
+
+void
+foo (void)
+{
+  if (a)
+    {
+      b = 1;
+      asm goto ("# insn 1" : : : : lab1);
+    lab1:;
+    }
+  else
+    {
+      b = 1;
+      asm goto ("# insn 2" : : : : lab2);
+    lab2:;
+    }
+}

        Jakub

Reply via email to