It is too hard to check performance issue on Android/arm now since autophone doesn't run default try server parameters.
I sometimes fix cross compiling issue for Linux/arm and aarch64 since it is tier-3 platform. So if it runs this on arm cloud, Linux/arm has to raise to tier-1 or 2. -- Makoto On Fri, Dec 8, 2017 at 3:57 AM, Henri Sivonen <hsivo...@hsivonen.fi> wrote: > Considering that running microbenchmarks on phone hardware is > problematic (thermal throttling messes things up, normal shell > environment isn't available to the harness, limited number of phones > available to perfherder, etc.), considering that it's easier to run > GNU/Linux in the cloud than it is to run Android and considering that > Gecko internals should compile roughly the same way for GNU/Linux and > Android on aarch64 and armv7, is getting some of our tryserver > performance tests (such as the GTest platform microbenchmarks) running > on in the cloud on GNU/Linux on aarch64 and GNU/Linux armv7+neon on > any active todo list? > > Packet.net seems to provide bare-metal aarch64 in the cloud. > > I don't know who provides bare-metal armv7+neon in the cloud. > Scaleway's bare-metal armv7 lacks neon. Additionally, ARMv7 code > running on an ARMv8 core seems to have different enough microbenchmark > behavior that ARMv7 benchmarks really should be run on an ARMv7 core > to be representative of ARMv7 deployment. (Currently, of course, a > good proportion of our Android users run ARMv7 code on an ARMv8 CPU, > but, naively, I'm hoping that we'll ship Fennec on aarch64 soon enough > for that combination to stop being interesting for benchmarking soon > enough.) > > -- > Henri Sivonen > hsivo...@hsivonen.fi > https://hsivonen.fi/ > _______________________________________________ > dev-platform mailing list > dev-platform@lists.mozilla.org > https://lists.mozilla.org/listinfo/dev-platform _______________________________________________ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
