When fp-pidbench was originally written SVE hardware was not widely
available so it was useful to run it in emulation and the default number
of loops was set very low, running for less than a second on actual
hardware. Now that SVE hardware is reasonably available it is very much
less interesting to use emulation, bump the default number of loops up to
even out a bit of the noise on real systems. On the machine I have to hand
this now takes about 15s which is still a toy microbenchmark but perhaps a
bit more useful.
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Will Deacon <will@kernel.org>
Some applications use SVE intermittently, one common case being where SVE
is used during statup (eg, by ld.so) but then rarely if ever during the
main application runtime. Add a repeat of the no SVE loop after we've done
the SVE loops to fp-pidbench to capture results for that.
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Will Deacon <will@kernel.org>
There are a number of freestanding static executables used in floating
point testing that have no runtime at all. These all define the main entry
point as:
.globl _start
function _start
_start:
but clang's integrated assembler complains that:
error: symbol '_start' is already defined
due to having both a label and function directive. Remove the label to
allow building with clang.
No functional change.
Signed-off-by: Mark Brown <broonie@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Link: https://lore.kernel.org/r/20230111-arm64-kselftest-clang-v1-2-89c69d377727@kernel.org
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Since it's likely to be useful for performance work with SVE let's have a
pidbench that gives us some numbers for consideration. In order to ensure
that we test exactly the scenario we want this is written in assembly - if
system libraries use SVE this would stop us exercising the case where the
process has never used SVE.
We exercise three cases:
- Never having used SVE.
- Having used SVE once.
- Using SVE after each syscall.
by spinning running getpid() for a fixed number of iterations with the
time measured using CNTVCT_EL0 reported on the console. This is obviously
a totally unrealistic benchmark which will show the extremes of any
performance variation but equally given the potential gotchas with use of
FP instructions by system libraries it's good to have some concrete code
shared to make it easier to compare notes on results.
Testing over multiple SVE vector lengths will need to be done with vlset
currently, the test could be extended to iterate over all of them if
desired.
Signed-off-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211202165107.1075259-1-broonie@kernel.org
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
