Comments
Patch
===================================================================
@@ -1,6 +1,5 @@
/* { dg-do compile } */
/* { dg-require-effective-target ilp32 } */
-/* { dg-skip-if "no stack realignment" { *-*-darwin* } { "*" } { "" } } */
/* { dg-options "-fstack-usage -msse2 -mforce-drap" } */
typedef int __attribute__((vector_size(16))) vec;
===================================================================
@@ -8,11 +8,7 @@
Then check that this is the actual stack usage in the assembly file. */
#if defined(__i386__)
-# if defined (__MACH__)
-# define SIZE 232
-# else
-# define SIZE 248
-# endif
+# define SIZE 248
#elif defined(__x86_64__)
# define SIZE 356
#elif defined (__sparc__)
===================================================================
@@ -79,11 +79,10 @@
Failure to ensure this will lead to a crash in the system libraries
or dynamic loader. */
#undef STACK_BOUNDARY
-#define STACK_BOUNDARY 128
+#define STACK_BOUNDARY \
+ ((profile_flag || (TARGET_64BIT && ix86_abi == MS_ABI)) \
+ ? 128 : BITS_PER_WORD)
-#undef MAIN_STACK_BOUNDARY
-#define MAIN_STACK_BOUNDARY 128
-
/* Since we'll never want a stack boundary less aligned than 128 bits
we need the extra work here otherwise bits of gcc get very grumpy
when we ask for lower alignment. We could just reject values less
@@ -91,7 +90,7 @@
it's below the minimum. */
#undef PREFERRED_STACK_BOUNDARY
#define PREFERRED_STACK_BOUNDARY \
- MAX (STACK_BOUNDARY, ix86_preferred_stack_boundary)
+ MAX (128, ix86_preferred_stack_boundary)
/* We want -fPIC by default, unless we're using -static to compile for
the kernel or some such. */
@@ -0,0 +1,7 @@
+/* PR target/36502 */
+/* { dg-do compile { target { *-*-darwin* && ilp32 } } } */
+/* { dg-options "-O -fomit-frame-pointer -fno-pic -S" } */
+int a;
+void f() {a++;}
+/* { dg-final { scan-assembler-not "esp" } } */
+
The attached patch solves PR36502, eliminating the unnecessary stack operations, by defining STACK_BOUNDARY to BITS_PER_WORD except for profiling or the use of the MS_ABI at 64-bit and by replacing STACK_BOUNDARY with 128 in the MAX macro defining PREFERRED_STACK_BOUNDARY. Eliminating the redefinition of MAIN_STACK_BOUNDARY enables stack realignments so that the testcase for gcc.target/i386/stack-usage-realign.c can be run on darwin. The special adjustment for darwin in gcc/testsuite/gcc.dg/stack-usage-1.c can also be dropped. Finally, a new test case for PR36502, gcc.target/i386/pr36502.c, of unnecessary stack operations is added. Bootstrap and regression tested on x86_64-apple-darwin10 in conjunction with the proposed fix for PR45234 (http://gcc.gnu.org/ml/gcc-patches/2010-08/msg01916.html). http://gcc.gnu.org/ml/gcc-testresults/2010-09/msg00259.html Okay for gcc trunk? Jack 2010-08-31 H.J. Lu <hjl.tools@gmail.com> Jack Howarth <howarth@bromo.med.uc.edu> PR 36502/target * gcc/config/i386/darwin.h (STACK_BOUNDARY): Redefine as 128 for profiling or 64-bit MS_ABI and as BITS_PER_WORD otherwise. (MAIN_STACK_BOUNDARY): Don't redefine. (PREFERRED_STACK_BOUNDARY): Replace STACK_BOUNDARY with 128 in MAX macro. * gcc.target/i386/pr36502.c: New test. * gcc.target/i386/stack-usage-realign.c: Don't skip on darwin. * gcc/testsuite/gcc.dg/stack-usage-1.c: Use default on i386/Darwin.