I have an error building OpenCL generated ptx.
The lines of code:
The nvidia ptx isa document says that b32 is not a valid type for the cvt instruction.
Is the following the correct replacement for those two lines of code?
st.local.b32 [ptx_ptr+0], var;
I have changed the names of the registers to protect my code.
I think it is but I am just checking before writing a program to patch the faulty ptx.
Note: the ptxas error is:
ptxas ptx_code.ptx, line 112; error : Arguments mismatch for instruction ‘cvt’