#2525
100%

Ran 28 Nov 2020 11:45AM UTC

Jobs 1

Files 77

Run time 6s

Badge

Embed ▾

pending completion

Build # #2525

Build Type

push

other

Committed by

ben-manes

Commit Message

Improved simulator runtime for large traces

Unfortunately, the elegant approach of partitioning a stream of events
into batches using the Iterator view comes with a hidden performance
cost. Instead of picking off events one-by-one, large internal buffers
are created and grown (SpinedBuffer) before the first element is
provided to the consumer. The maximum chunk size is 1M elements, which
can cause a long pause while the trace is read into memory before it
is processed by the simulator. For the DS1 trace this adds 10 seconds
of startup time for the simulator and doubles the time for the trace
rewriter.

The forEach operation pulls on-demand, where our smaller I/O buffers
and concurrent prefetching keep the policies fed with a lower startup
overhead. This also greatly reduces the memory footprint to allow for
more easily analyzing large caches. However this coding structure is
slightly less pleasant.

Coverage Stats

6226 of 6681 relevant lines covered (93.19%)

0.93 hits per line

Jobs

ID	Job ID	Ran	Files	Coverage
1	#2525.1	28 Nov 2020 11:45AM UTC	0	93.19

Source Files on build #2525

Detailed source file information is not available for this build.

ben-manes / caffeine / #2525
100%

README BADGES
x

Markdown

Textile

RDoc

HTML

Rst

Jobs

Source Files on build #2525

ben-manes / caffeine / #2525 100%

README BADGES x

Markdown

Textile

RDoc

HTML

Rst

Jobs

Source Files on build #2525

ben-manes / caffeine / #2525
100%

README BADGES
x