Both THNN and THCUNN define the accfloat/accdouble typedef via ffi. The behavior of this seems to be lua version/luajit version dependent (saw test failures with luajit) and it's probably a good idea to get rid of it (as C doesn't let you redefine a typedef in any case).
Since we no longer support 'real' in the header declaration, it makes sense to just define the gsub on accreal; if we want to add back real support we'll have to do something more complicated.
with nn.
Both THNN and THCUNN define the accfloat/accdouble typedef via ffi. The behavior of this seems to be lua version/luajit version dependent (saw test failures with luajit) and it's probably a good idea to get rid of it (as C doesn't let you redefine a typedef in any case).
Since we no longer support 'real' in the header declaration, it makes sense to just define the gsub on accreal; if we want to add back real support we'll have to do something more complicated.