Messages in this thread | | | Date | Wed, 18 Mar 1998 15:08:11 +0000 (GMT) | From | Chris Evans <> | Subject | Re: BogoMIPS revisitied...MMX |
| |
On Tue, 17 Mar 1998, George Bonser wrote:
> Here is an interesting array: > > INTEL P-60: 23.96 BMPS
> > 4 instructions. Why is it that my P100 spits out something like > > 39.19 BogoMIPS and my P233MMX spits out 466.15 bogomips? I know > > that the Pentium had some issue with this, being that a standard > > 486 executed *that* particular loop faster. > > > > So, why such the HUGE difference?
I'm surprised no one has noted yet that the original pentiums had an _awful_ branch prediction bug, they would predict which way a brach was likely to occur, then go and act as if the brach were going the other way!!!!
This was fixed in the MMX chips.
The difference sticks out like a sore thumb in a tight bogomips loop since every 3 or 4 instructions the pentium guesses a branch wrong and all sorts of things stall because of it.
Chris
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu
| |