This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] powerpc: Fix usage of elision transient failure adapt param
- From: Adhemerval Zanella <adhemerval dot zanella at linaro dot org>
- To: libc-alpha at sourceware dot org, Tulio Magno Quites Machado Filho <tuliom at linux dot vnet dot ibm dot com>
- Date: Thu, 20 Aug 2015 16:22:15 -0300
- Subject: Re: [PATCH] powerpc: Fix usage of elision transient failure adapt param
- Authentication-results: sourceware.org; auth=none
- References: <55D625B9 dot 9060700 at linux dot vnet dot ibm dot com> <55D627C6 dot 3060408 at linaro dot org>
However this clash with Tulio update from his rave condition fix [1].
You will need to coordinate with him to either drop this patch or
push his modification without this fix.
[1] https://sourceware.org/ml/libc-alpha/2015-08/msg00858.html
On 20-08-2015 16:17, Adhemerval Zanella wrote:
> LGTM, thanks for checking on that.
>
> On 20-08-2015 16:08, Paul E. Murphy wrote:
>> The skip_lock_out_of_tbegin_retries adaptive parameter was
>> not being used correctly, nor as described. This prevents
>> a fallback for all users of the lock if a transient abort
>> occurs within the accepted number of retries.
>>
>> 2015-08-20 Paul E. Murphy <murphyp@linux.vnet.ibm.com>
>> * sysdeps/powerpc/nptl/elide.h(__elide_lock): Fix
>> usage of .skip_lock_out_of_tbegin_retries
>> * sysdeps/unix/sysv/linux/powerpc/elision-lock.c (__lll_lock_elision):
>> Likewise, and respect a value of try_tbegin <= 0
>> ---
>> sysdeps/powerpc/nptl/elide.h | 10 +++++-----
>> sysdeps/unix/sysv/linux/powerpc/elision-lock.c | 15 ++++++---------
>> 2 files changed, 11 insertions(+), 14 deletions(-)
>>
>> diff --git a/sysdeps/powerpc/nptl/elide.h b/sysdeps/powerpc/nptl/elide.h
>> index 389f5a5..38e1565 100644
>> --- a/sysdeps/powerpc/nptl/elide.h
>> +++ b/sysdeps/powerpc/nptl/elide.h
>> @@ -55,14 +55,14 @@ __elide_lock (uint8_t *adapt_count, int is_lock_free)
>> *adapt_count = __elision_aconf.skip_lock_internal_abort;
>> break;
>> }
>> - /* Same logic as above, but for a number of temporary failures in a
>> - a row. */
>> - else if (__elision_aconf.skip_lock_out_of_tbegin_retries > 0
>> - && __elision_aconf.try_tbegin > 0)
>> - *adapt_count = __elision_aconf.skip_lock_out_of_tbegin_retries;
>> }
>> }
>>
>> + /* Fall back to locks for a bit if retries have been exceeded */
>> + if (__elision_aconf.skip_lock_out_of_tbegin_retries > 0
>> + && __elision_aconf.try_tbegin > 0)
>> + *adapt_count = __elision_aconf.skip_lock_out_of_tbegin_retries;
>> +
>> return false;
>> }
>>
>> diff --git a/sysdeps/unix/sysv/linux/powerpc/elision-lock.c b/sysdeps/unix/sysv/linux/powerpc/elision-lock.c
>> index 7f9bcc2..c6731ca 100644
>> --- a/sysdeps/unix/sysv/linux/powerpc/elision-lock.c
>> +++ b/sysdeps/unix/sysv/linux/powerpc/elision-lock.c
>> @@ -72,8 +72,7 @@ __lll_lock_elision (int *lock, short *adapt_count, EXTRAARG int pshared)
>> goto use_lock;
>> }
>>
>> - int try_begin = aconf.try_tbegin;
>> - while (1)
>> + for (int i = aconf.try_tbegin; i > 0; i--)
>> {
>> if (__builtin_tbegin (0))
>> {
>> @@ -87,21 +86,19 @@ __lll_lock_elision (int *lock, short *adapt_count, EXTRAARG int pshared)
>> /* A persistent failure indicates that a retry will probably
>> result in another failure. Use normal locking now and
>> for the next couple of calls. */
>> - if (try_begin-- <= 0
>> - || _TEXASRU_FAILURE_PERSISTENT (__builtin_get_texasru ()))
>> + if (_TEXASRU_FAILURE_PERSISTENT (__builtin_get_texasru ()))
>> {
>> if (aconf.skip_lock_internal_abort > 0)
>> *adapt_count = aconf.skip_lock_internal_abort;
>> goto use_lock;
>> }
>> - /* Same logic as above, but for for a number of temporary failures
>> - in a row. */
>> - else if (aconf.skip_lock_out_of_tbegin_retries > 0
>> - && aconf.try_tbegin > 0)
>> - *adapt_count = aconf.skip_lock_out_of_tbegin_retries;
>> }
>> }
>>
>> + /* Fall back to locks for a bit if retries have been exhausted */
>> + if (aconf.try_tbegin > 0 && aconf.skip_lock_out_of_tbegin_retries > 0)
>> + *adapt_count = aconf.skip_lock_out_of_tbegin_retries;
>> +
>> use_lock:
>> return LLL_LOCK ((*lock), pshared);
>> }
>>