On 11/17/2011 03:47 PM, Alan Modra wrote:
BTW, we're left with dead stores into oldval's stack slot. Is it
legal for the first two parameters of __atomic_compare_exchange to
alias? I'm guessing that the stores might disappear if we tell gcc
that they can't alias.
No they cannot alias.
This is part of work in progress getting locking back into shape on
powerpc. (If we ever were in shape, which is doubtful.) Using the
ia64 version means we have a needless sync at the start of
gomp_mutex_lock (courtesy of __sync_val_compare_and_swap) and a
needless isync at the end of
On 11/17/2011 01:56 AM, Alan Modra wrote:
This is part of work in progress getting locking back into shape on
powerpc. (If we ever were in shape, which is doubtful.) Using the
ia64 version means we have a needless sync at the start of
gomp_mutex_lock (courtesy of __sync_val_compare_and_swap)
On Thu, Nov 17, 2011 at 02:56:31PM -1000, Richard Henderson wrote:
On 11/17/2011 01:56 AM, Alan Modra wrote:
This is part of work in progress getting locking back into shape on
powerpc. (If we ever were in shape, which is doubtful.) Using the
ia64 version means we have a needless sync at