Chris Theis <Christian.Theis_at_cern.ch> sent on November 26th, 2010:
|------------------------------------------------------------------------------------------------------------------------------|
|"[..]
|
|
|
|[..]
|
|[..] The benchmarks GPU vs. CPU usually shown are based on single
precision!" |
|------------------------------------------------------------------------------------------------------------------------------|
Sender: owner-fluka-discuss_at_mi.infn.it
Hi,
The benchmarks do not always even use the faster F.P.U. of the
C.P.U.'s F.P.U.s. However, this merely makes realistic speedup factors
smaller instead of proving that no speedup would exist.
|------------------------------------------------------------------------------------------------------------------------------|
|"The probably more important point is that different vendors have
allowed themselves a bit of freedom with |
|respect to the implementation of the IEEE floating point standard. It
is not guaranteed that the |
|GPU that you are running your program on will yield the same result as
the one of your colleague because it's from a different|
|vendor. In terms of debugging& quality control this is a very
important |
|aspect that is commonly overlooked!
|
|
|
|[..]"
|
|------------------------------------------------------------------------------------------------------------------------------|
It is true that Fluka developers like reproducibility when answering
debugging queries. However, the I.E.E.E. floating-point standard is
not the only floating-point policy available and not adhering to it
often results in faster simulations. Not adhering to the
I.E.E.E. floating-point standard is not the same as being
mathematically sloppier.
Regards,
Paul Colin
Received on Sat Nov 27 2010 - 13:23:22 CET
This archive was generated by hypermail 2.2.0 : Sat Nov 27 2010 - 13:23:29 CET