psim failures with -O1

Chris Johns chrisj at rtems.org
Tue Jan 29 08:54:04 UTC 2019


On 29/1/19 7:48 pm, Sebastian Huber wrote:
> On 29/01/2019 09:14, Chris Johns wrote:
>> Hi,
>>
>> I thought it may be a good idea to set the optimization level for the psim BSP
>> to `-O1`. The xilinx_zynq_a9_qemu has this and it is useful.
>>
>> The change causes the number of test failures to jump from around 20 to over 230.
>>
>> Is this expected?
> 
> Yes, this is a compiler bug/feature. GCC generates a floating point instruction
> in a function which is called in ISR context. This leads to an exception.

OK.

> 
> On SPARC using -O0 to run the tests is also problematic due to the much higher
> stack usage.
> 

OK.

Thanks
Chris



More information about the devel mailing list