Investigating the few remaining testsuite failures on Solaris 11/x86, I came
across
FAIL: gcc.target/i386/abi-2.c scan-assembler-times ymm0 1
In the assembler output, ymm0 appears twice:
vmovdqa .LC0, %ymm0
vmovdqa %ymm0, (%eax)
FAIL: gcc.target/i386/pr22076.c scan-assembler-times movq 3
movq appears only twice:
paddb .LC0, %mm0
movl 24(%esp), %eax
movq %mm0, 8(%esp)
FAIL: gcc.target/i386/pr22076.c scan-assembler-not movl
movl appears 5 times
The failures occur only for 32-bit.
Both failures seem to be due to i386.c (ix86_sol10_return_in_memory): unlike
return_in_memory_32, that function doesn't know about AVX yet. It seems like
the testcase needs to take the differing Solaris 2/x86 ABI for MMX, though
it's unusual that the 32-bit ABI should have changed between Solaris 9 and 10.
Perhaps Richard can shed some light on this?
Given this divergence between the generic and sol10 function, it seems unwise
to leave them separate. Instead, I'd rather have code (conditional on say
TARGET_SOLARIS or TARGET_SOLARIS10) in return_in_memory_32 to handle the
Solaris ABI requirements there. The only problem with such an approach is
that i386/vx-common.h uses ix86_sol10_return_in_memory, too, but would probably
be interested in every possible code conditional on TARGET_SOLARIS.
--
Summary: gcc.target/i386/abi-2.c and gcc.target/i386/pr22076.c
fail on 32-bit Solaris 10+/x86
Product: gcc
Version: 4.6.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: target
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: ro at gcc dot gnu dot org
GCC build triplet: i386-pc-solaris2.11
GCC host triplet: i386-pc-solaris2.11
GCC target triplet: i386-pc-solaris2.11
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44452
