Nicolas Boulay wrote:
http://www.realworldtech.com/forums/index.cfm?action=detail&id=103164&threadid=102997&roomid=2
An interresting discussion about the Tomasulo algorithm to do OoO. The
point of view of Linus is that pushing at some point the performance,
out of order became easier than in-order.
this is absolutely dependent on the design tools :
OOO was not possible, for example, when the only
mean to design a chip was to draw the lines by hand
on a huuuuuge sheet of cellophane.
And the chip surface was precious at that time.
EDA companies appeared in the mid-80s and 10 years later,
OOO was definitely adopted by Intel, who could afford huge dies.
Maybe out of order execution could be use in the shader. The most
interresting is asyncronous memory access to hide the latency better
than SMP.
What do you think of it?
KISS your MIPS :-)
But you would design your own OOO CPU if you wanted to.
Regards,
Nicolas
YG
--
http://ygdes.com / http://yasep.org
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)
