Commit Graph

16 Commits

Author SHA1 Message Date
Sebastien Bourdeauducq 47e88dfcbe Revert "test: temporarily disable test_async_throughput"
This reverts commit f0289d49ab.
2020-09-03 14:19:55 +08:00
Sebastien Bourdeauducq f0289d49ab test: temporarily disable test_async_throughput
M-Labs/artiq-zynq#104
2020-09-01 17:49:40 +08:00
pca006132 69f0699ebd test: improved test_performance
1. Added tests for small payload.
2. Added statistics.
2020-08-28 14:58:34 +08:00
pca006132 cfddc13294 test: fixed test_performance
Added more tests and use normal rpc instead of async rpc.

Async RPC does not represent the real throughput which is limited by the
hardware and the network. Normal RPC which requires a response from the
remote is closer to real usecases.
2020-08-26 14:17:06 +08:00
Sebastien Bourdeauducq 709026d945 test: relax device_to_host_rate 2020-07-30 17:46:22 +08:00
Sebastien Bourdeauducq 85b5a04acf test: print transfer rates in MiB/s 2020-07-07 17:28:47 +08:00
Sebastien Bourdeauducq 4cc9bd33ce test: only test_rpc_timing actually requires ARTIQ_LOW_LATENCY 2019-04-24 11:22:07 +08:00
Robert Jördens 90c9fa446f test: add array transfer test
200 kB/s, more than a factor of 10 slower than the bare string transfer

Signed-off-by: Robert Jördens <rj@quartiq.de>
2019-01-28 14:30:44 +00:00
Sebastien Bourdeauducq ba6094c3e5 test: relax network transfer rates
Due to lower Kasli system clock frequency.
2018-08-27 16:47:48 +08:00
whitequark e285fe0d56 test: tighten required TransferTest timings.
smoltcp performs significantly better with LTO.
2018-08-12 20:17:37 +00:00
whitequark 2648b1b7a1 firmware: migrate to Rust 1.28.0.
This also updates / is a prerequisite for updating smoltcp.

Rationale for changes made:
  * compiler_builtins is now shipped in the rust prefix.
  * rustc's libpanic_unwind no longer works for us because it
    has a hard dependency on Box (and it's a horrible hack);
    fortunately, we only ever needed a personality function
    from it.
  * panic and oom handlers are now set in a completely different
    way.
  * allocators are quite different (and finally stable).
  * NLL caused internal compiler errors in runtime, so code using
    NLL was rewritten to not rely on it and it was turned off.
2018-08-12 19:17:45 +00:00
Sebastien Bourdeauducq 423929a125 test: relax min transfer rates from 2MB/s to 1.9MB/s 2018-07-09 18:00:24 +08:00
Sebastien Bourdeauducq 67625fe912 test: check kernel overhead credibility 2018-01-28 01:02:03 +08:00
Sebastien Bourdeauducq e8ed3475ea test: add kernel overhead test (#407) 2018-01-28 01:00:59 +08:00
Robert Jördens a44f8282dc test_performance: relax network speed to 2 MB/s
At QUARTIQ I am getting 2.4/2.3 MB/s and with single switch at M-Labs we
apparently regularly met 2.2/2.2 MB/s. But with the current multiple
switches and one of them being a problematic switch that triggered #837
it looks like it is a tad slower.

http://buildbot.m-labs.hk/builders/artiq/builds/1818/steps/python_coverage_1/logs/stdio
2017-12-06 17:21:41 +01:00
whitequark 7dfe7d8c93 test: verify no network performance regression from current 2.2 MB/s. 2017-10-30 04:25:58 +00:00