Currently, to compute the 'item' size in a ring buffer, we have
`SIZE32_OF`.
Several issues with this:
- `SIZE32_OF` only works on variables, not types, due to an extra
parenthesis pair. Indeed, `sizeof((int))` is not valid C, whereas
`sizeof((my_var))` is.
- `SIZE32_OF` is not a proper public API
- `SIZE32_OF` rounds down if the argument size is not a multiple
of 4 bytes.
Thus, we introduce a proper `RING_BUF_ITEM_SIZEOF`, fixing the
aforementioned issues.
Signed-off-by: Henri Xavier <datacomos@huawei.com>
Add a bunch of missing "zephyr/" prefixes to #include statements in
various test and test framework files.
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
In order to bring consistency in-tree, migrate all tests to the new
prefix <zephyr/...>. Note that the conversion has been scripted, refer
to #45388 for more details.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
Test were executed on single CPU only and with qemu_cortex_a9
excluded. Removing those limitations after fixes.
Signed-off-by: Krzysztof Chruscinski <krzysztof.chruscinski@nordicsemi.no>
Make it possible to "finish" with fewer bytes than what was "claimed".
This was possible before on the get side, but the put side was
cummulative wrt finish. The revamp made it cummulative on both sides.
Turns out that existing users rely on the opposite behavior which is
more logical and useful. So make both sides that way.
Adjust documentation, test case and users accordingly.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
The core ring buffer code presumes a 32-bit wrap-around behavior when its
head and tail indices cross the maximum value range. Make sure this is
covered by the test suite, for both 32- and 64-bit targets.
While at it, reduce stack requirement of the modified test.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
This code is rather hairy. When I look at it I don't like the way it
stares back at me.
First, the rewind business looks fishy. It has to die.
And we don't have to rely on modulus either. Not even for non-power-of-2
buffers. Let's kill that distinction too and make all sizes always
"high performance".
The code is now entirely relying only on simple ALU operations (add,
sub and compare).
The key assumption: 32-bit values do wrap around after max range has
been reached. No saturation. All architectures supported by Zephyr
do that.
Some stats:
lib/os/ring_buffer.c: 62 insertions(+), 124 deletions(-)
ring_buffer.c.obj before after diff
----------------------------------------------
frdm_k64f 1224 1136 -88
m2gl025_miv 2485 2079 -406
mps2_an385 1228 1132 -96
mps2_an521 1228 1132 -96
native_posix 1546 1496 -50
native_posix_64 1598 1595 -3
nsim_hs_mpuv6 1252 1192 -60
nsim_hs_smp 1252 1192 -60
nsim_sem 1252 1192 -60
qemu_arc_em 1324 1192 -132
qemu_arc_hs6x 1824 1620 -204
qemu_arc_hs 1252 1192 -60
qemu_cortex_a53_smp 2154 1888 -266
qemu_cortex_a53 2154 1888 -266
qemu_cortex_a9 1938 1792 -146
Before (qemu_cortex_a53):
START - test_ringbuffer_performance
1 byte put-get, avg cycles: 52
4 byte put-get, avg cycles: 47
1 byte put claim-finish, avg cycles: 39
5 byte put claim-finish, avg cycles: 41
5 byte get claim-finish, avg cycles: 52
PASS - test_ringbuffer_performance in 0.8 seconds
After (qemu_cortex_a53):
START - test_ringbuffer_performance
1 byte put-get, avg cycles: 34
4 byte put-get, avg cycles: 41
1 byte put claim-finish, avg cycles: 27
5 byte put claim-finish, avg cycles: 29
5 byte get claim-finish, avg cycles: 29
PASS - test_ringbuffer_performance in 0.4 seconds
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Conceptually, ring_buf_item_put() and ring_buf_item_get() are specialized
versions of ring_buf_put() and ring_buf_get(). Make it so to rationalize
the code to open the way for more optimizations.
This means we need specialized wrappers on top of ring_buf_init()
accordingly, given that the core machinery is now common and byte based.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Commit 1e46bb3bb5 ("lib: os: ring_buffer: Allow using full buffer
capacity") changed the capacity test condition but left the comment
relating to the previous test condition in place.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
The tests from tests/lib/ringbuffer fails on all nrf platforms due
to a timeout. Increasing the timeout value solves the issue.
Signed-off-by: Maciej Perkowski <Maciej.Perkowski@nordicsemi.no>
The stress ('scuse me, "zstress") cases here are all written to
exercise reader/writer threads at different priority combinations.
That's defeated if the threads are allowed to run on different CPUs
(because being "low" priority doesn't matter if you have a spare CPU
to run on).
There is also extensive use of stack buffers to pass data through the
ring buffer zero copy implementation, which runs afoul of the
KERNEL_COHERENCE rules on intel_adsp platforms (where stack memory is
incoherent between CPUs and can't be shared like that).
Fix both issues by just setting CONFIG_MP_NUM_CPUS=1
Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
Updated test to use stress testing framework instead of implementing
own framework.
Signed-off-by: Krzysztof Chruscinski <krzysztof.chruscinski@nordicsemi.no>
tests/lib/ringbuffer is now finally fixed with commit 15e834a687
("linker: __data_region_start equal to __data_start"), so we can enable
it back.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Evgeniy Paltsev <PaltsevEvgeniy@gmail.com>
Counter was not reset between various tests which may hide the fact
that number of preemptions is less than expected.
Signed-off-by: Krzysztof Chruscinski <krzysztof.chruscinski@nordicsemi.no>
Stress test that was validating thread safeness was configured to use
higher sys tick rate and targeting qemu_x86 platform. However, it was
also build for other platforms and was failing. Fixing it by extracting
stress tests to separate configuration that is run only on qemu_x86.
Modified the test to fail when number of preemptions is less than
expected only on qemu_x86.
Signed-off-by: Krzysztof Chruscinski <krzysztof.chruscinski@nordicsemi.no>
