Hi,
this patch adds slightly more simplification in try_combine(), under
-fexpensive-optimizations, by calling subst() to both i2 and i1.
Supplied is an ARM testcase where this resulted in one less redundant
ZERO_EXTEND insn under ARMv6 and above.
Bootstrapped and tested on i686 and x86_64, cross-tested on ARM-Linux
using QEMU, all with no regressions. Okay for trunk?
Thanks,
Chung-Lin
2011-03-16 Chung-Lin Tang <clt...@codesourcery.com>
* combine.c (try_combine): Do simplification only call of
subst() on i2 even when i1 is present. Update comments.
testsuite/
* gcc.target/arm/unsigned-extend-1.c: New testcase.
Index: combine.c
===================================================================
--- combine.c (revision 170996)
+++ combine.c (working copy)
@@ -3063,7 +3063,7 @@
/* It is possible that the source of I2 or I1 may be performing
an unneeded operation, such as a ZERO_EXTEND of something
that is known to have the high part zero. Handle that case
- by letting subst look at the innermost one of them.
+ by letting subst look at the inner insns.
Another way to do this would be to have a function that tries
to simplify a single insn instead of merging two or more
@@ -3088,11 +3088,9 @@
subst_low_luid = DF_INSN_LUID (i1);
i1src = subst (i1src, pc_rtx, pc_rtx, 0, 0);
}
- else
- {
- subst_low_luid = DF_INSN_LUID (i2);
- i2src = subst (i2src, pc_rtx, pc_rtx, 0, 0);
- }
+
+ subst_low_luid = DF_INSN_LUID (i2);
+ i2src = subst (i2src, pc_rtx, pc_rtx, 0, 0);
}
n_occurrences = 0; /* `subst' counts here */
Index: testsuite/gcc.target/arm/unsigned-extend-1.c
===================================================================
--- testsuite/gcc.target/arm/unsigned-extend-1.c (revision 0)
+++ testsuite/gcc.target/arm/unsigned-extend-1.c (revision 0)
@@ -0,0 +1,9 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -march=armv6" } */
+
+unsigned char foo (unsigned char c)
+{
+ return (c >= '0') && (c <= '9');
+}
+
+/* { dg-final { scan-assembler-not "uxtb" } } */
