It has been reported that units are not used consistently in the document. Specifically
Unify performance unit TFLOPS or GFLOPS.
However, I cannot currently find any reference to FLOPS anywhere in the document. Possibly, the user refers to the submission Top500 website.
Solution
Scan the document for use of units
Unify units (also prefixes)
GFLOPS etc.
It is not unlikely that we introduces text that uses performance, or energy efficiency, units in the document.
My suggestion would be:
performance in Gflop/s
energy efficiency in Gflop/(sW)
Note that HPL uses Gflops (in the version I know), Top500.org uses, PFlop/s and GFlops/watts.
Personally, I am strongly opposed to GFLOPS (or any capitalization thereof) due to the ambiguity with Floating Point OperationS. And therefore I am opposed to GFLOPS/W and even more so GFLOPS/watt. While the form Gflop/(sW) is uncommon, I think it is basically the only correct form (that does not use an exponent, e.g. as per NIST).
Problem statement
It has been reported that units are not used consistently in the document. Specifically
However, I cannot currently find any reference to FLOPS anywhere in the document. Possibly, the user refers to the submission Top500 website.
Solution
GFLOPS etc.
It is not unlikely that we introduces text that uses performance, or energy efficiency, units in the document. My suggestion would be:
Gflop/s
Gflop/(sW)
Note that HPL uses
Gflops
(in the version I know), Top500.org uses,PFlop/s
andGFlops/watts
. Personally, I am strongly opposed toGFLOPS
(or any capitalization thereof) due to the ambiguity with Floating Point OperationS. And therefore I am opposed toGFLOPS/W
and even more soGFLOPS/watt
. While the formGflop/(sW)
is uncommon, I think it is basically the only correct form (that does not use an exponent, e.g. as per NIST).Maybe we don't have to go there.
Related issue
93