You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm noting this down here because I discovered it by mistake and it might get lost otherwise.
I built the benchmarks with the standard command used for the baseline:
and used --cpu-mhz 16 by mistake. I then noticed that crc32 was 4 bytes bigger then expected. By removing --cpu-mhz 16 or by using --cpu-mhz 1 I got the expected size. I tried with different values between 1 and 16 and the expected result comes only with 1 MHz. cpu-mhz isn't used for the size benchmark so it shouldn't have an effect. It shouldn't affect size in any case.
At a first glance it only enters the benchmark in the standard benchmark body function: int benchmark (void) { return benchmark_body (LOCAL_SCALE_FACTOR * CPU_MHZ); }
like all the other benchmarks. I might have done something wrong but if someone else can reproduce this behaviour then we have a bug to investigate.
I tested this with embench 0.5/1.0/master and alsways got this behaviour.
The text was updated successfully, but these errors were encountered:
I'm noting this down here because I discovered it by mistake and it might get lost otherwise.
I built the benchmarks with the standard command used for the baseline:
./build_all.py --arch arm --chip cortex-m4 --board stm32f4-discovery --cc arm-none-eabi-gcc --cflags='-c -Os -ffunction-sections -march=armv7-m -mcpu=cortex-m4 -mfloat-abi=soft -mthumb' --ldflags='-Wl,-gc-sections -nostartfiles -nostdlib' --dummy-libs='crt0 libc libgcc libm' --cpu-mhz 16 --clean
and used
--cpu-mhz 16
by mistake. I then noticed that crc32 was 4 bytes bigger then expected. By removing--cpu-mhz 16
or by using--cpu-mhz 1
I got the expected size. I tried with different values between 1 and 16 and the expected result comes only with 1 MHz. cpu-mhz isn't used for the size benchmark so it shouldn't have an effect. It shouldn't affect size in any case.At a first glance it only enters the benchmark in the standard benchmark body function:
int benchmark (void) { return benchmark_body (LOCAL_SCALE_FACTOR * CPU_MHZ); }
like all the other benchmarks. I might have done something wrong but if someone else can reproduce this behaviour then we have a bug to investigate.
I tested this with embench 0.5/1.0/master and alsways got this behaviour.
The text was updated successfully, but these errors were encountered: