procyberian/gcc
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

rs6000: Fix default alignment ABI break caused by MMA base support

Browse source 
As part of the MMA base support, we incremented BIGGEST_ALIGNMENT in
order to align the __vector_pair and __vector_quad types to 256 and 512
bytes respectively.  This had the unintended effect of changing the
default alignment used by __attribute__ ((__aligned__)) which causes
an ABI break because of some dodgy code in GLIBC's struct pthread.
The fix is to revert the BIGGEST_ALIGNMENT change and to force the
alignment on the type itself rather than the mode used by the type.

2020-11-06  Peter Bergner  <bergner@linux.ibm.com>

gcc/
	* config/rs6000/rs6000.h (BIGGEST_ALIGNMENT): Revert previous commit
	so as not to break the ABI.
	* config/rs6000/rs6000-call.c (rs6000_init_builtins): Set the ABI
	mandated alignment for __vector_pair and __vector_quad types.

gcc/testsuite/
	* gcc.target/powerpc/mma-alignment.c: New test.
This commit is contained in:
Peter Bergner 2020-11-06 17:00:49 -06:00
parent 659ba632e4
commit a37b5bcf15
3 changed files with 47 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
gcc/config/rs6000/rs6000-call.c
View file

@ -13191,12 +13191,14 @@ rs6000_init_builtins (void)
if (TARGET_EXTRA_BUILTINS)
{
vector_pair_type_node = make_unsigned_type (256);
SET_TYPE_ALIGN (vector_pair_type_node, 256);
SET_TYPE_MODE (vector_pair_type_node, POImode);
layout_type (vector_pair_type_node);
lang_hooks.types.register_builtin_type (vector_pair_type_node,
"__vector_pair");
vector_quad_type_node = make_unsigned_type (512);
SET_TYPE_ALIGN (vector_quad_type_node, 512);
SET_TYPE_MODE (vector_quad_type_node, PXImode);
layout_type (vector_quad_type_node);
lang_hooks.types.register_builtin_type (vector_quad_type_node,

6
gcc/config/rs6000/rs6000.h
View file

@ -776,8 +776,10 @@ extern unsigned rs6000_pointer_size;
/* Allocation boundary (in *bits*) for the code of a function. */
#define FUNCTION_BOUNDARY 32
/* No data type wants to be aligned rounder than this. */
#define BIGGEST_ALIGNMENT (TARGET_MMA ? 512 : 128)
/* No data type is required to be aligned rounder than this. Warning, if
BIGGEST_ALIGNMENT is changed, then this may be an ABI break. An example
of where this can break an ABI is in GLIBC's struct _Unwind_Exception. */
#define BIGGEST_ALIGNMENT 128
/* Alignment of field after `int : 0' in a structure. */
#define EMPTY_FIELD_BOUNDARY 32

41
gcc/testsuite/gcc.target/powerpc/mma-alignment.c Normal file
View file

@ -0,0 +1,41 @@
/* { dg-do run } */
/* { dg-require-effective-target hard_float } */
/* { dg-options "-O2 -mhard-float" } */
#include <stdlib.h>
/* The MMA types below are enabled for pre-power10 compiles, because the
built-ins that use them must always be initialized in case the user has
a target attribute or pragma on a function that uses the MMA built-ins.
Since the test below doesn't need any other MMA support, we can enable
this test case on basically any cpu that has hard floating point
registers. */
struct
{
int __attribute__ ((__aligned__)) ivar;
__vector_pair pair;
__vector_quad quad;
} s;
int
main (void)
{
/* Verify default alignment is 16-byte aligned (BIGGEST_ALIGNMENT).
This may change in the future, but that is an ABI break, so this
hardcoded test case is here to be a noisy FAIL as a warning, in
case the ABI change was unintended and unwanted. An example of where
this can break an ABI is in glibc's struct _Unwind_Exception. */
if (__alignof__ (s.ivar) != 16)
abort ();
/* Verify __vector_pair types are 32-byte aligned. */
if (__alignof__ (s.pair) != 32)
abort ();
/* Verify __vector_quad types are 64-byte aligned. */
if (__alignof__ (s.quad) != 64)
abort ();
return 0;
}