Re: [PATCH] stdlib: reorganize stdlib Makefile routines by functionality

[Adhemerval Zanella Netto <adhemerval.zanella@linaro.org>]

On 01/04/24 10:17, H.J. Lu wrote:
> On Mon, Apr 1, 2024 at 6:13 AM Florian Weimer <fweimer@redhat.com> wrote:
>>
>> * abush wang:
>>
>>> This is test:
>>> ```
>>> uint64_t getnsecs() {
>>>     uint32_t lo, hi;
>>>     __asm__ __volatile__ (
>>>         "rdtsc" : "=a"(lo), "=d"(hi)
>>>     );
>>>     return ((uint64_t)hi << 32) | lo;
>>> }
>>>
>>> int main() {
>>>     const int num_iterations = 1;
>>>     uint64_t start, end, total_time = 0;
>>>
>>>     start = getnsecs();
>>>     for (int i = 0; i < num_iterations; i++) {
>>>         (void) lrand48();
>>>     }
>>>     end = getnsecs();
>>>     total_time += (end - start);
>>>
>>>     printf("Average time for lrand48: %lu cycles\n", total_time / num_iterations);
>>>     return 0;
>>> }
>>> ```
>>> before:
>>> Average time for lrand48: 21418 cycles
>>>
>>> after:
>>> Average time for lrand48: 9892 cycles
>>
>> Do you see this on x86-64?  So this isn't a displacement range issue?
>>
>> It could be that this is a random performance change due to code
>> alignment, and not actually caused by the direct call distance.
>>
> 
> I have a linker patch to control section layout:
> 
> https://patchwork.sourceware.org/project/binutils/list/?series=29973
> 
> It can
> 
> 1. Reduce gaps between text sections.
> 2. Put hot text sections close to each other.
> 
> If it can solve this issue, we should add this feature to ld.
> 

Another possibility, if this is related to a displacement range due some
ISA limitation; would to move the lrand entrypoint to the same TU (at
least the one that are simple wrapper that ended up being tail calls).