On Sat, Mar 23, 2013 at 02:11:37PM -0400, Peter Colberg wrote:
> If this transformation needs to be done at launch time anyway, we
> could convert __local and __constant kernel arguments to automatic
> arrays.
To be precise, the LLVM IR of an OpenCL kernel such as
__kernel void func(__global uint *a,
__local uint *b,
__local uint *c,
__constant uint *d,
__constant uint *e)
{
would be transformed to the IR equivalent of
__constant uint d[d_size];
__constant uint e[e_size];
__kernel void func(__global uint *a)
{
__local uint b[b_size];
__local uint c[c_size];
Pointers to the __constant arrays would be obtained using
cuModuleGetGlobal() and initialised using cuMemcpyDtoD().
Peter
------------------------------------------------------------------------------
Everyone hates slow websites. So do we.
Make your web apps faster with AppDynamics
Download AppDynamics Lite for free today:
http://p.sf.net/sfu/appdyn_d2d_mar
_______________________________________________
pocl-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/pocl-devel
