Community
cancel
Showing results for 
Search instead for 
Did you mean: 
matthieu_darbois
New Contributor III
572 Views

popcnt latency/throughput in 64bits

Jump to solution
Hi,
I could find these values for 32bits operations in the last Optimization Reference Manual but the counter part for 64bits operations is not there.
Does anyone know where I can find this ?
Reagrds,
Matthieu
0 Kudos
1 Solution
Max_L
Employee
572 Views
Hello Matthieu - POPCNT has the same 3-cycle latency / 1-cycle throughputfor both 32-bit and 64-bit forms.

Regards,
-Max

View solution in original post

4 Replies
Aubrey_W_
New Contributor I
572 Views
Hi Mattheiu,

POPCNT is an SSE4 instruction. You might try looking in the Intel SSE4 Programming Reference here:
http://software.intel.com/en-us/articles/45nm-next-generation-intel-coret-2-processor-family-penryn-...

AVX reference can be found at http://software.intel.com/en-us/avx/

I hope that helps.

==
Aubrey W.
Intel Software Network Support

matthieu_darbois
New Contributor III
572 Views
Hi,
This was not the kind of information I was looking for. I've used this instruction for a while now and thus, I know that it is SSE4.2 or SSE4a on AMD. In fact, just have to look for the POPCNT flag in CPUID...
What I am looking for is the performance of this instruction. As I said in the first post, the "Intel 64 and IA-32ArchitecturesOptimization Reference Manual" (Order Number: 248966-020) lists in Appendix C almost all throughput and latency values for SIMD extensions.

However, I can't find these values for POPCNT in 64 bits flavour.

Regards,
Matthieu
Max_L
Employee
573 Views
Hello Matthieu - POPCNT has the same 3-cycle latency / 1-cycle throughputfor both 32-bit and 64-bit forms.

Regards,
-Max

View solution in original post

matthieu_darbois
New Contributor III
572 Views
Hi Max,
Thanks for this answer.
Is there anyway to post comments/ request additional information like this to be put in the reference manual ?
Regards,
Matthieu
Reply