microbenchmark
longqimin opened this issue · comments
i found roaring_bitmap_and_cardinality() benchmark slower than roaring_bitmap_and() + roaring_bitmap_get_cardinality(), is this reasonable?
~/s/t/CRoaring ((4fdf6ffb)|✔) $ ./build/microbenchmarks/bench
2023-04-05T15:51:48+08:00
Running ./build/microbenchmarks/bench
Run on (48 X 3000 MHz CPU s)
CPU Caches:
L1 Data 32 KiB (x48)
L1 Instruction 32 KiB (x48)
L2 Unified 4096 KiB (x48)
L3 Unified 16384 KiB (x48)
Load Average: 0.27, 1.42, 5.63
AVX-2 hardware: yes
AVX-512: supported by compiler
AVX-512 hardware: no
In RAM volume in MiB (estimated): 1.802828
benchmarking other files: You may pass is a data directory as a parameter.
data source: /home/longqimin00/stdb/thirdparty/CRoaring/benchmarks/realdata/census1881
number of bitmaps: 200
performance counters: No privileged access (sudo may help).
x64: detected
----------------------------------------------------------------------------
Benchmark Time CPU Iterations
----------------------------------------------------------------------------
SuccessiveIntersection 25549 ns 25491 ns 27321
SuccessiveIntersectionCardinality 28714 ns 28654 ns 24485
SuccessiveUnionCardinality 41574 ns 41460 ns 17880
SuccessiveDifferenceCardinality 34451 ns 34363 ns 20397
SuccessiveUnion 473435 ns 471382 ns 1488
TotalUnion 568666 ns 565881 ns 1313
TotalUnionHeap 1618086 ns 1610175 ns 431
RandomAccess 2545 ns 2539 ns 273106
ToArray 132780 ns 132244 ns 5209
IterateAll 1985539 ns 1977863 ns 356
ComputeCardinality 3466 ns 3452 ns 202821
No, it is not reasonable but make sure sure you are testing on a server configured for testing.
You are not benchmarking on a laptop, are you?
If you cannot resolve your issue, please open a new issue with a detailed description of your configuration (hardware, processor, operating system, compiler). The measurements must not be taken on a laptop.
Ideally, run your benchmarks on something like AWS so that we can reproduce your results.