out float4 a<>, out float4 b[]
Hello,
I am getting strange errors with the following test program:
kernel void foo(out float4 a<>, out float4 b[])
{
}
int main()
{
float4 a<20>;
float4 b<30>;
foo(a, b);
return 0;
}
If I run it in CPU emulation mode, I get the following error message:
% BRT_RUNTIME=CPU ./test2
Using CPU runtime
test2: cpukernel.cpp:385: virtual void brook::CPUKernel::Map(): Assertion `output_args->getDomainMax() - output_args->getDomainMin() == dmax - dmin' failed.
Aborted
Why do I get an error here? If b was not a scatter stream (i.e. specifying the parameter b as 'out float4 b<>'), then it is clear that the sizes of a and b would have to be the same. But if b is a scatter stream, then this shouldn't matter, should it?
In CPU mode, the error disappears if I give a and b the same sizes (declaring b as 'float4 b<20>'). But when I use the GPU, I still get an error even if a and b are of the same size:
% ./test2
Could not setup output target
Assertion failure: calcontext.cpp (1070): Could not setup output target
test2: calbase.hpp:92: void CALAssertImpl(const char*, int, const char*): Assertion `0' failed.
Aborted
Is this a hardware problem (I just upgraded to an HD3850 card)? Or a bug in brook+/cal? Will this be resolved in future releases?
Thanks for any help,
Ingo
