Every measurement repeats until a stated interval has passed, so a clock
that reads the same value twice does not slow the benchmark down, it stops
it returning at all.
CLOCK_MONOTONIC does not advance on every target. On qemu-intel64 it
reports success and stays at zero, while CLOCK_REALTIME advances normally,
and the benchmark spins in its first measurement with no output after the
heading.
Sample each candidate twice around a busy wait and take the first one whose
reading changes. Where none does, say so and skip the timing rather than
hang.
Assisted-by: Claude:claude-opus-5
Signed-off-by: Justin Hammond <justin@dynam.ac>
Neither is covered here, and both are overridable, so a machine or libc
implementation of either goes in unmeasured and unchecked.
memccpy is checked with the search character present, where the copy
stops just past it and the result points there, and absent, where the
whole length is copied and the result is NULL. stpncpy is checked
against every capacity from zero to four past the length, for the
content, the zero padding beyond the terminator, and the returned
pointer, which is the terminator when the string fits and one past the
end when it does not.
Both sweep all sixty four source and destination alignment pairs, and
both are added to the benchmark, which now covers nineteen functions.
Assisted-by: Claude:claude-opus-5
Signed-off-by: Justin Hammond <justin@dynam.ac>
strlcpy is the one function in this directory's reach that nothing here
covers, and a machine directory may override it like any other.
Sweep every source and destination alignment pair against sizes 1 to 64,
and for each of those every capacity from zero to one past the length.
Check the return value, which is the length of src whether or not the
copy fit, the truncation point, the content, that a capacity of zero
writes nothing at all, and that nothing lands past the terminator.
The alignment pairs are the point. An implementation that walks one of
the two pointers to a boundary and then copies a register at a time is
correct whenever the two agree, so a test that only ever passes matching
alignments says nothing about it.
The timing half is guarded. perf_gettime() is not a system call, so an
application reaches it only where the C library builds its own copy or
where the application and the kernel are one image; calling it
unconditionally leaves the test unbuildable on a kernel build, which is
where the correctness half is still wanted.
Assisted-by: Claude:claude-opus-5
Signed-off-by: Justin Hammond <justin@dynam.ac>
The existing speed checks time one call at one size, 128 bytes, with both
operands aligned. A machine implementation usually takes its wide path
only when the pointers satisfy some alignment condition, so that single
point reports the best case and says nothing about the rest of the input
space.
Measure the same functions across a size sweep and every source and
destination alignment pair instead, plus strlcpy. On rv64 the difference
this exposes is not marginal:
strcpy 32768 B s+0/d+0 2938.0 MB/s
strcpy 32768 B s+1/d+1 2942.0 MB/s
strcpy 32768 B s+1/d+2 626.0 MB/s
memcmp 32768 B s+0/d+0 412.4 MB/s
memcmp 32768 B s+1/d+2 41.0 MB/s
Two pointers misaligned by the same amount run at the aligned rate;
misaligned by different amounts they fall to a tenth of it. Neither
number is visible from an aligned measurement alone.
A function with no machine implementation reports the same rate at every
alignment, so the sweep also shows which of them a machine directory
actually covers.
Each result reports MB/s, which compares across machines, and cycles per
byte where perf_gettime() is reachable from an application, both from one
timed loop. strcat starts from an empty destination on each turn, since
appending to the last result would grow it without bound, so its figure
includes that store.
It sits behind TESTING_ARCH_LIBC_BENCH, default n, because a measurement
runs for a fixed interval and a full sweep takes about a minute.
Assisted-by: Claude:claude-opus-5
Signed-off-by: Justin Hammond <justin@dynam.ac>
test_strcpy(), test_strncpy() and test_stpcpy() applied the same offset
to the source and to the destination, so both pointers always shared the
same word congruence. Architecture optimized copy routines take a
different code path when the two offsets differ: a byte prologue to
align the destination, then either a byte fallback or a shift-merge loop
that recombines two source words per store. None of that was reached by
the test.
Vary the source and destination offsets independently over 0..7 in those
three tests, so both the equal congruence (aligned word copy) and the
unequal congruence (shift-merge) paths are covered, and report both
offsets on failure so a regression points at the offending combination.
Also drop the ARCH_TOOLCHAIN_GNU dependency from TESTING_ARCH_LIBC. The
test only uses standard C string functions and perf_gettime(), with no
GNU specific construct, so it builds with non GNU toolchains such as
TASKING as well.
Impact: test only, selected by CONFIG_TESTING_ARCH_LIBC (default n).
Dropping the ARCH_TOOLCHAIN_GNU dependency only widens the set of
toolchains that may select the test, no existing configuration changes.
Testing: built and ran sim:nsh on Linux x86_64 (Ubuntu 24.04,
gcc 13.3.0) with CONFIG_TESTING_ARCH_LIBC=y. strcpy, strncpy and
stpcpy report PASSED for all 64 offset combinations, and
"arch_libc_test Passed".
Assisted-by: Claude:claude-opus-5
Signed-off-by: zhangyuan29 <zhangyuan29@xiaomi.com>
The adjacent overlap case in test_memmove() placed the source at a fixed
g_buf1 + align + 64 and the destination one size further, so the
destination tail reached align + 64 + 2 * size. g_buf1 is only
TEST_BUF_SIZE + MAX_ALIGN (528) bytes, so the larger swept sizes ran off
the end: align=0 with size=255 writes up to offset 573, that is 46 bytes
past the object. AddressSanitizer aborted arch_libctest with a
global-buffer-overflow.
Start the adjacent layout at g_buf1 + align instead. The tail then
reaches align + 2 * size, which is at most 7 + 2 * 257 = 521 and stays
inside g_buf1 for every alignment and boundary size that is swept, while
still keeping source and destination exactly adjacent.
Impact: test only, selected by CONFIG_TESTING_ARCH_LIBC (default n).
Testing: built and ran sim:nsh on Linux x86_64 (Ubuntu 24.04,
gcc 13.3.0) with CONFIG_TESTING_ARCH_LIBC=y. memmove reports PASSED
with no sanitizer report, and "arch_libc_test Passed".
Assisted-by: Claude:claude-opus-5
Signed-off-by: dengwenqi <dengwenqi@xiaomi.com>
Add test_strchrnul() and speed_strchrnul(), selected by the new
CONFIG_TESTING_ARCH_LIBC_STRCHRNUL option, covering the hit, miss and
NUL cases.
Sweep alignment 0..7 and the boundary sizes {0, 1, 7, 8, 9, 15, 16, 17,
31, 32, 33, 63, 64, 65, 127, 128, 129, 255, 256, 257} in the scan
function tests (memcmp, memchr, strlen, strcmp, strchr, strncmp,
strnlen, strrchr) and in memmove. Those sizes sit on the 8 and 16 byte
chunk edges and on the sub-word tails, so vectorized (NEON/MVE) and
word-at-a-time implementations are stressed exactly at their alignment
and size boundaries instead of only at "nice" lengths. memmove is
additionally exercised across four overlap layouts: forward, backward,
contained and adjacent.
Impact: test only, selected by CONFIG_TESTING_ARCH_LIBC (default n).
Testing: built and ran qemu-armv7a:nsh (Cortex-A7, generic C
implementation) and sim:nsh on Linux x86_64 (Ubuntu 24.04, gcc 13.3.0)
with CONFIG_TESTING_ARCH_LIBC=y. All 16 enabled functions report
PASSED and "arch_libc_test Passed". These tests pass against the
generic C routines, which establishes the correctness baseline before
architecture optimized assembly is introduced.
Signed-off-by: anjiahao <anjiahao@xiaomi.com>
The arch_libc test only covered strcpy, so the architecture optimized
implementations of the remaining string and memory routines were never
exercised by the test suite.
Extend the test to also cover memcpy, memmove, memset, memcmp, memchr,
strlen, strcmp, strchr, strncmp, strnlen, strncpy, stpcpy, strcat and
strrchr:
* Every function gets a correctness test that sweeps the buffer
alignment and the transfer size and compares the result against the
expected value.
* Every function gets a speed test that reports the average cycle count
measured with perf_gettime().
* Every individual test is selected by its own
CONFIG_TESTING_ARCH_LIBC_<FUNC> option (default y), so a target can
drop the ones it does not need.
Impact: test only. Nothing is built unless CONFIG_TESTING_ARCH_LIBC
(default n) is selected, so no existing board configuration changes.
Testing: built and ran sim:nsh on Linux x86_64 (Ubuntu 24.04,
gcc 13.3.0) with CONFIG_TESTING_ARCH_LIBC=y. All 15 enabled functions
report PASSED and "arch_libc_test Passed".
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
Replace app-side includes of <debug.h> with <nuttx/debug.h> to use the
header from the NuttX tree explicitly after the header move.
Signed-off-by: Piyush Patle <piyushpatle228@gmail.com>
Skip UDP listen test when CONFIG_NET_UDP is disabled to prevent
test failures on configurations without UDP support.
Signed-off-by: tengshuangshuang <tengshuangshuang@xiaomi.com>
Correct date formatting in cache_test and arch_lib_test outputs to
ensure consistent date/time display format across test results.
Signed-off-by: tengshuangshuang <tengshuangshuang@xiaomi.com>
Specify architecture format in Kconfig to ensure proper configuration
handling for different architectures in testing modules.
Signed-off-by: tengshuangshuang <tengshuangshuang@xiaomi.com>
