Hi John,
On 5/7/2024 6:25 PM, John Hubbard wrote:
> On 5/7/24 6:21 PM, Reinette Chatre wrote:
>> Hi John,
> ...
>>
>> The following (what was in v1) looks good to me. What am I missing?
>>
>> diff --git a/tools/testing/selftests/resctrl/cmt_test.c b/tools/testing/selftests/resctrl/cmt_test.c
>> index a81f91222a89..05a241519ae8 100644
>> --- a/tools/testing/selftests/resctrl/cmt_test.c
>> +++ b/tools/testing/selftests/resctrl/cmt_test.c
>> @@ -40,11 +40,11 @@ static int show_results_info(unsigned long sum_llc_val, int no_of_bits,
>> int ret;
>> avg_llc_val = sum_llc_val / num_of_runs;
>> - avg_diff = (long)abs(cache_span - avg_llc_val);
>> + avg_diff = (long)(cache_span - avg_llc_val);
>
> This deletes the abs() call, because I incorrectly let clang's warning
> lead me to believe that the abs() call was a no-op. But both you and Ilpo
> pointed out that the math breaks if you do that.
To me the extra abs() was unnecessary anyway ...
>
>> diff_percent = ((float)cache_span - avg_llc_val) / cache_span * 100;
>> ret = platform && abs((int)diff_percent) > max_diff_percent &&
>> - abs(avg_diff) > max_diff;
>> + labs(avg_diff) > max_diff;
>
.. because it is repeated here.
> This hunk is OK.
>
>> ksft_print_msg("%s Check cache miss rate within %lu%%\n",
>> ret ? "Fail:" : "Pass:", max_diff_percent);
>>
>> Reinette
>
> thanks,
Reinette
On Tue, 7 May 2024, Reinette Chatre wrote:
> Hi John,
>
> On 5/7/2024 6:25 PM, John Hubbard wrote:
> > On 5/7/24 6:21 PM, Reinette Chatre wrote:
> >> Hi John,
> > ...
> >>
> >> The following (what was in v1) looks good to me. What am I missing?
> >>
> >> diff --git a/tools/testing/selftests/resctrl/cmt_test.c b/tools/testing/selftests/resctrl/cmt_test.c
> >> index a81f91222a89..05a241519ae8 100644
> >> --- a/tools/testing/selftests/resctrl/cmt_test.c
> >> +++ b/tools/testing/selftests/resctrl/cmt_test.c
> >> @@ -40,11 +40,11 @@ static int show_results_info(unsigned long sum_llc_val, int no_of_bits,
> >> int ret;
> >> avg_llc_val = sum_llc_val / num_of_runs;
> >> - avg_diff = (long)abs(cache_span - avg_llc_val);
> >> + avg_diff = (long)(cache_span - avg_llc_val);
> >
> > This deletes the abs() call, because I incorrectly let clang's warning
> > lead me to believe that the abs() call was a no-op. But both you and Ilpo
> > pointed out that the math breaks if you do that.
>
> To me the extra abs() was unnecessary anyway ...
>
> >
> >> diff_percent = ((float)cache_span - avg_llc_val) / cache_span * 100;
> >> ret = platform && abs((int)diff_percent) > max_diff_percent &&
> >> - abs(avg_diff) > max_diff;
> >> + labs(avg_diff) > max_diff;
> >
>
> .. because it is repeated here.
Yes, there are two *abs() calls in this function.
In this case is okay to remove the first one since it didn't remove
absolute value completely, whereas in the MBA/MBM cases v1 removed *abs()
call entirely which was wrong thing to do.
I explicitly noted in my v1 comment that this CMT change is okay but the
other two were not.
--
i.
On 5/8/24 1:00 AM, Ilpo Järvinen wrote:
> On Tue, 7 May 2024, Reinette Chatre wrote:
>
>> Hi John,
>>
>> On 5/7/2024 6:25 PM, John Hubbard wrote:
>>> On 5/7/24 6:21 PM, Reinette Chatre wrote:
>>>> Hi John,
>>> ...
> Yes, there are two *abs() calls in this function.
>
> In this case is okay to remove the first one since it didn't remove
> absolute value completely, whereas in the MBA/MBM cases v1 removed *abs()
> call entirely which was wrong thing to do.
>
> I explicitly noted in my v1 comment that this CMT change is okay but the
> other two were not.
>
Got it, thanks Ilpo (and Reinette), I'll send a v3 with it fixed up
accordingly, appreciate your patience here. :)
thanks,
--
John Hubbard
NVIDIA