[Solved] Are multi-core processors really faster?

For everything that's not in any way related to PureBasic. General chat etc...
User avatar
Olliv
Enthusiast
Enthusiast
Posts: 542
Joined: Tue Sep 22, 2009 10:41 pm

Re: [Solved] Are multi-core processors really faster?

Post by Olliv »

Back to the future !

I apologize. I forgot a waiting delay before starting...

Hoping now initializing basis value is a little bit right...

Now (when basis value is right and first results are positive) if 'X + 1' is replaced with a recursive procedure call ( named rpc), we can grow the clock tickcount.

Code: Select all

Procedure rpc(n)
   If n
      rpc(n - 1)
   endif
endprocedure

Value0.Q = TickCount()
rpc(1000)
Value1.Q = TickCount()
If you get a "consistant" result (i.e. : 126870 ticks)
we maybe could use this result too to have a second referency, a second basis.

Code: Select all

basis2.Q = Value1 - Value0 - Basis
Then we can keep it respectively...

Code: Select all

;************************************************************
TickCountInitially.Q = TickCount()
Repeat
   Delay(0)
Until (TickCount() - TickCountInitially) > Basis2
One-thread-multiple-operation.
User avatar
tj1010
Enthusiast
Enthusiast
Posts: 621
Joined: Mon Feb 25, 2013 5:51 pm
Location: US or Estonia
Contact:

Re: [Solved] Are multi-core processors really faster?

Post by tj1010 »

Kernels maintain cores transparently and you just see mutex and TLS data. Even AAA games that swear they make special use of 64bit and multi-cores are just improving usage of the high level thread API.. Mostly with TLS collision and mutex scaling..

At least this is what I seen when I went digging in to binaries that advertised special consideration for multiple physical cores. There may be some tricks from "userland" to optimize what the kernel does when mapping.

Chip makers are who really dictate this with architecture. It's what caused research in to things like optical buses and SIMD "cells"
The truth hurts.
Post Reply