We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
There was an error while loading. Please reload this page.
1 parent 519c59c commit f940c2cCopy full SHA for f940c2c
benchmarks/scripts/indexing_and_static_ndranges.jl
@@ -18,7 +18,16 @@ the performance of vector vs array operations in a broadcast
18
setting where linear indexing is allowed.
19
20
# Summary:
21
- - Pointwise
+ - On the CPU:
22
+ static NDRanges do not play an important role,
23
+ but linear indexing is 2x faster than cartesian
24
+ indexing.
25
+ - On the GPU:
26
+ static NDRanges DO play an important role,
27
+ but we could (alternatively) see an improvement
28
+ by using linear indexing. Supporting StaticNDRanges
29
+ also impacts non-pointwise kernels, and yields
30
+ nearly the same benefit as linear indexing.
31
32
# References:
33
- https://github.com/CliMA/ClimaCore.jl/issues/1889
0 commit comments