Paul Brook wrote:
Lines = 2863
Other 50 = 1.7% (These are branch targets)
Branch 80 = 2.8%
Scalar 1008 = 35.2%
Vector 1725 = 60.3%
The only argument I've heard so far in favour of MIMD is that it
would improve the performance for scalar workloads. For a graphics
chip, that doesn't appear to be a smart approach.
By your own numbers conventional 4-element SIMD gets <70% ALU utilization.
not necessarily because it's a static analysis.
percentage of instructions does not mean utilisation.
For example a control instruction is not used as often
as others, depending on the program structure.
To me it sounds like :
a GPU can use LIW instructions with 2 vector fields and one scalar/control
field.
It would fit the static analysis numbers.
However a dynamic analysis could uncover other more interesting trends.
That leaves quite a lot of scope for improvement, be it full MIMD or several
ALU threads controlled by a single dispatch unit (ala. Larabee SIMD with
software thread combining).
If no improvement was possible, we wouldn't be speaking here ;-)
Now the problem is to limit the scope enough to get a decent,
simple, flexible model, keeping in mind that 100% ALU usage is not possible.
90% or 95% is good enough, however, but the most difficult is the first %s ...
regards,
Paul
yg
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)
