Hello Peter,
On 2020-07-11 03:00, Peter Levart wrote:
On 7/11/20 9:31 AM, Peter Levart wrote:
- compile the two sets of benchmarks separately with separate output
directories and create separate benchmarks.jar files for them
Here's my attempt at a patch for separate benchmarks.jar files. The
minor complication, as I found out, was what to do when running the
micro benchmarks via "make test TEST=..." command when there are two
possible jar files to choose from. I opted for a separate tag in TEST
variable, so preview benchmark(s) are run as follows:
make test TEST="micro.preview: ..."
http://cr.openjdk.java.net/~plevart/jdk-dev/8248429_jmh_enable_preview/webrev.02/
This generally looks pretty good to me. I will defer to Claes on what
build steps are actually needed for the new jar file. I agree with Claes
that having a list of files in the makefile is probably easier to update
than moving files around. I would recommend something like this:
MICROBENCHMARK_PREVIEW_FILES := \
org/openjdk/bench/java/io/RecordDeserialization.java \
#
This variable can then be fed into EXCLUDE_FILES and INCLUDE_FILES
respectively. The formatting is our recommended way of building lists of
things in the makefiles where lines can easily be added or removed
without affecting any adjacent lines to minimize merge conflicts in the
future.
Otherwise, I would only ask that you try to shorten some lines a bit. We
aren't strictly enforcing 80 chars, but try to stay close when possible
for easier side by side reviews and 3-way merges in the future.
Specifically RunTests.gmk:682 (maybe an intermediate value?) and
BuildMicrobenchmark.gmk:197.
/Erik
WDYT?
Regards, Peter
