Age | Commit message (Collapse) | Author |
|
Signed-off-by: Vishal Bhoj <vishal.bhoj@linaro.org>
Change-Id: I8aa1673488604aa72188b6925d37b983da984975
|
|
Change-Id: I6d4639b341e3121c44e7b49d00ada8b366093541
|
|
The benchmarks to run can be filtered with the `run.py` script or via the java
app itself. This allows to simplify the build system.
One can stil build an app without specific benchmarks by simply deleting the
benchmark files.
Change-Id: Ie81bcd795796baf4e7920faf9c46b67cf0fc6887
|
|
Change-Id: Ia9eb120606b149bc3c914cbe6f10b0b733b035c4
|
|
Change-Id: I71d7d06de39ae15b29192dd2697ee9731006336a
|
|
This option allows to not run some benchmarks. By default it is set to ignore
all deprecated benchmarks.
Multiple filter can be combined. For example:
./run.py --filter "deprecated/Factorial*" --filter-out "deprecated/FactorialFloat"
Change-Id: I3fb868677451520ae1a06019e2e92c72bf909510
|
|
- Move most existing benchmarks under the `deprecated` sub-directory. We are
not confident in the value of these benchmarks. They can be reinstated
individually if they are shown to be useful.
- Move `Intrinsics.java` under the `micro` sub-directory.
Change-Id: I2094e66cb38d4f59a114ac3fefbdc0c6f1aa2a49
|
|
Change-Id: I05016cffb0f9fa474d9193b4dc56bab5f0605695
|
|
Change-Id: I837dc25d6c24bf7430765a62374213b57a205806
|
|
Change-Id: I9464e1a4761869747ed0cd094e0deb19019f3faf
|
|
- Don't print "ns", which confuses the report tools.
- Reduce the number of iterations.
Change-Id: I2a907bd843e1420ab9ce382397389e6b59648590
|
|
Change-Id: I2f4123fcf462a102237e0a82b2f649e78f8ed113
|
|
Change-Id: I2e353932d5e467b51a1a712047c27a9bfad940c8
|
|
Multiple tests can be specified on the command line. For example:
java com.arm.microbench.RunBench BubbleSort Rotation
Change-Id: I83ad867bae5fa2abf298e9b909fb732c1a590f52
|
|
Change-Id: Id77394809a5e2ae7bd27e3e0e7eb985620e58883
|
|
The path to the built classes was incorrect.
Change-Id: I351e609b0c55cfc874043575b80ce15fe45b988d
|
|
Change-Id: I7392e795252882673eb8e353a87e9c9b62cb03b0
|
|
- use float("inf") to avoid errors running under Python 2.
- Optionally verify benchmark results with boolean verify().
- Pass --debug to java when run.py is verbose.
- Fix 'mode' argument ignored by run_adb.
- -f, --filter benchmarks by an optional glob.
Change-Id: I233503b82445eed71b32a42989f1684c72e73509
|
|
Includes infrastructure for microbenchmarking Java code along with some
small microbenchmarks.
Infrastructure for running perf is included.
Change-Id: I06abb2aa9da962bbd1a750dbaf27ea122e3915f0
|
|
|