cuda - Does nvcc have an equivalent of fsingle-precision-constant? -


GCC allows to save the problem of manually adding every literal continuity. Is there any convenient equivalent in NVCC?

I believe the answer is new architecture Not with)

The old Open 64-based compiler supported a huge range of command line options (though many did nothing). Possible to pass -Xcompiler = -fshort-double to force all double precision calculations into single precision, the compiler is automatically done for two precise floating points automatically Sm_10, sm_11 and sm_12 targets which were the lack of original double exact support. The old compiler is still used for calculation 1.x target, so it is still possible.

The new cicc LVVM front end (which is used for the devices; 1.x calculation capacity) is too much, very small list of arguments, and they are not AFAIK documents surrounding the poking code generation and Shows the following options for controlling output:

  -information-line-information-device-c-kernel-params-are-restrict -restrict -disable-inlining- Wno-memory -space -disable-llc-opts -show-src -g -O3 -O0 -fmad = 0 -fmad = 1 -prec-div = 0 -prec-div = 1 -prec-sqrt = 0 -prec-Sqrt = 1 -ftz = 0 -ftz = 1 -m64 -m32   

So I think However, you are out of luck, if you really want to do , you may have the opportunity to write a Livmeela replaces double precision constants themselves floating in CIC emitted IR. It looks as if a person is asking at least one thing, and can give you a place to start.

Comments

Popular posts from this blog

Java - Error: no suitable method found for add(int, java.lang.String) -

java - JPA TypedQuery: Parameter value element did not match expected type -

c++ - static template member variable has internal linkage but is not defined -