Re: [PATCH v2 3/3] perf record: adapt affinity to machines with #CPUs > 1K

[Alexey Budankov]

On 25.11.2019 14:27, Alexey Budankov wrote:
> 
> On 25.11.2019 14:21, Jiri Olsa wrote:
>> On Mon, Nov 25, 2019 at 02:13:20PM +0300, Alexey Budankov wrote:
>>> On 25.11.2019 12:42, Jiri Olsa wrote:
>>>> On Mon, Nov 25, 2019 at 09:08:57AM +0300, Alexey Budankov wrote:
>>>>
>>>> SNIP
>>>>
>>>>> -static void perf_mmap__setup_affinity_mask(struct mmap *map, struct mmap_params *mp)
>>>>> +static int perf_mmap__setup_affinity_mask(struct mmap *map, struct mmap_params *mp)
>>>>>  {
>>>>> -	CPU_ZERO(&map->affinity_mask);
>>>>> +	map->affinity_mask.nbits = cpu__max_cpu();
>>>>> +	map->affinity_mask.bits = bitmap_alloc(map->affinity_mask.nbits);
>>>>> +	if (!map->affinity_mask.bits)
>>>>> +		return -1;
>>>>> +
>>>>>  	if (mp->affinity == PERF_AFFINITY_NODE && cpu__max_node() > 1)
>>>>>  		build_node_mask(cpu__get_node(map->core.cpu), &map->affinity_mask);
>>>>>  	else if (mp->affinity == PERF_AFFINITY_CPU)
>>>>> -		CPU_SET(map->core.cpu, &map->affinity_mask);
>>>>> +		set_bit(map->core.cpu, map->affinity_mask.bits);
>>>>> +
>>>>> +	return 0;
>>>>>  }
>>>>>  
>>>>> +#define MASK_SIZE 1023
>>>>>  int mmap__mmap(struct mmap *map, struct mmap_params *mp, int fd, int cpu)
>>>>>  {
>>>>> +	char mask[MASK_SIZE + 1] = {0};
>>>>
>>>> does this need to be initialized?
>>>
>>> This is to make sure the message is zero terminated for vfprintf call()
>>
>> hum AFAICS it's used only in bitmap_scnprintf, which should
>> terminate the string properly
> 
> If vfprintf() explicitly terminates output buffer with zero then
> the initialization above can be avoided.

Well, bitmap_scnprintf() returns the length of resulting string and 
the length can be used to make zero termination in the buffer and 
avoid mask initialization in the beginning.

~Alexey
 
> 
>>
>> jirka
>>
> 
> ~Alexey
>