Open monofuel opened 2 months ago
trying to compute cache[cacheIndex] += cache[cacheIndex + i] from a gpu kernel gives the following error from hipcc:
cache[cacheIndex] += cache[cacheIndex + i]
test_dot_product.nim.cpp:356:6: error: no matching function for call to 'pluseq___pureZstrutils_u2431'
pluseq___pureZstrutils_u2431 looks like a helper function from the compiler, however this is not available for use as a device function.
pluseq___pureZstrutils_u2431
static N_INLINE(void, pluseq___pureZstrutils_u2431)(NF& x_p0, NF y_p1);
looks like nim includes a bunch of functions that are supposed to get inlined, wonder how to handle these...
__device__
trying to compute
cache[cacheIndex] += cache[cacheIndex + i]
from a gpu kernel gives the following error from hipcc:pluseq___pureZstrutils_u2431
looks like a helper function from the compiler, however this is not available for use as a device function.looks like nim includes a bunch of functions that are supposed to get inlined, wonder how to handle these...
__device__
versions of these functions?