HI group,
I've encountered a problem compiling for the avx512skx target using the
reduce_add function.
The following sample never ends compiling using ispc 1.9.2 on Linux:
export uniform int select_eq (
const uniform int count,
const uniform int * const uniform values,
const uniform int operand,
uniform float * const uniform mask
)
{
uniform int n = 0;
foreach (i = 0 ... count)
{
int m = (int)(values[i] == operand);
n += reduce_add (m); // (*)
mask[i] = m; // (**)
}
return n;
}
The command line used is:
ispc reduce_add.ispc -o reduce_add.o --target=avx512skx-i32x16 -O3
Note that in -O0, the sample compiles fine. Anything equal or above -O1
keeps the compiler hanged.
Other targets like sse2, avx1 and avx2 are fine.
Additionally, if you either comment the reduce_add line (*) or the
following line (**) the sample compiles fine.
I've worked it around specifically for avx512 by splitting the foreach loop
in 2, but this is a bit nasty.
Thank you in advance for your insights!
Best regards,
Ben
--
You received this message because you are subscribed to the Google Groups
"Intel SPMD Program Compiler Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
