* [RS6000] Fragile testcase breaks with -frename-registers
@ 2016-05-03 14:35 Alan Modra
2016-05-09 12:25 ` Segher Boessenkool
0 siblings, 1 reply; 2+ messages in thread
From: Alan Modra @ 2016-05-03 14:35 UTC (permalink / raw)
To: gcc-patches; +Cc: David Edelsohn, Segher Boessenkool
See comments added below. Bootstrapped and regression tested
powerpc64le-linux. OK to apply?
PR testsuite/70826
* gcc.target/powerpc/savres.c: Compile with -fno-rename-registers.
diff --git a/gcc/testsuite/gcc.target/powerpc/savres.c b/gcc/testsuite/gcc.target/powerpc/savres.c
index f10c99a..9c4c627 100644
--- a/gcc/testsuite/gcc.target/powerpc/savres.c
+++ b/gcc/testsuite/gcc.target/powerpc/savres.c
@@ -1,5 +1,5 @@
/* { dg-do run } */
-/* { dg-options "-fno-inline -fomit-frame-pointer" } */
+/* { dg-options "-fno-inline -fomit-frame-pointer -fno-rename-registers" } */
/* { dg-additional-options "-mdynamic-no-pic" { target *-*-darwin* } } */
/* -fno-inline -maltivec -m32/-m64 -mmultiple/no-multiple -Os/-O2. */
@@ -10,6 +10,13 @@
#define SET_GPR(R,V) SET (long, R, V)
#define SET_FPR(R,V) SET (double, R, V)
#define SET_VR(R,V) SET (__attribute__ ((vector_size (16))) int, R, V)
+/* There doesn't seem to be a way of letting gcc know that cr2, cr3
+ and cr4 are being used, and therefore should not be touched by
+ gcc. Unlike gpr, fpr and vr we can't do something like
+ register __attribute__ ((__mode__ ("__CC__"))) int cr2 __asm__ ("cr2");
+ This makes the test somewhat fragile, dependent on gcc not using
+ any of cr2, cr3 and cr4 in main(), and is why -fno-rename-registers
+ is required. */
#define SET_CR(R,V) __asm__ __volatile__ ("mtcrf %0,%1" : : "n" (1<<(7-R)), "r" (V<<(4*(7-R))) : "cr" #R)
#define TRASH_GPR(R) SET_GPR (R, 0)
#define TRASH_FPR(R) SET_FPR (R, 0)
@@ -1053,7 +1060,18 @@ void ws_0 (void)
abort ();
}
-int main (void)
+/* We'd like to compile main with
+ __attribute__ ((__optimize__ ("fixed-cr2,fixed-cr3,fixed-cr4")))
+ but that doesn't do anything currently. Obviously we don't want to
+ compile the whole file with -ffixed-cr2 -ffixed-cr3 -ffixed-cr4 as
+ that would also tell gcc not to save/restore cr, and we're trying
+ to check that the above functions do save/restore cr.
+ __attribute__ ((__optimize__ ("no-rename-registers,omit-frame-pointer")))
+ works, but it seems odd to need omit-frame-pointer and raises the
+ question of whether darwin would need -mdynamic-no-pic.
+ So for now use -fno-rename-registers over the whole test. */
+int
+main (void)
{
INIT_REGS;
USE_ALL_CR;
--
Alan Modra
Australia Development Lab, IBM
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [RS6000] Fragile testcase breaks with -frename-registers
2016-05-03 14:35 [RS6000] Fragile testcase breaks with -frename-registers Alan Modra
@ 2016-05-09 12:25 ` Segher Boessenkool
0 siblings, 0 replies; 2+ messages in thread
From: Segher Boessenkool @ 2016-05-09 12:25 UTC (permalink / raw)
To: Alan Modra; +Cc: gcc-patches, David Edelsohn
On Wed, May 04, 2016 at 12:05:45AM +0930, Alan Modra wrote:
> See comments added below. Bootstrapped and regression tested
> powerpc64le-linux. OK to apply?
>
> PR testsuite/70826
> * gcc.target/powerpc/savres.c: Compile with -fno-rename-registers.
Okay for trunk. Thanks!
Segher
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2016-05-09 12:25 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-05-03 14:35 [RS6000] Fragile testcase breaks with -frename-registers Alan Modra
2016-05-09 12:25 ` Segher Boessenkool
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).