Commit f0cdc3c0 authored by Cianciosa, Mark's avatar Cianciosa, Mark
Browse files

Separate out cuda init and max_concurrency and make these static methods to... 

Separate out cuda init and max_concurrency and make these static methods to void errors when initalizing cuda.
parent 80186db4

CMakeLists.txt

+5 −5
Original line number Diff line number Diff line
@@ -108,11 +108,11 @@ register_sanitizer_option (float-divide-by-zero ON) 


target_compile_options (sanitizer

                        INTERFACE

                        $<$<CONFIG:Sanitized>:-fsanitize-trap=all>

                        $<$<AND:$<CONFIG:Sanitized>,$<PLATFORM_ID:Darwin>>:-fsanitize-trap=all>

)

target_link_options (sanitizer

                     INTERFACE

                     $<$<CONFIG:Sanitized>:-fsanitize-trap=all>

                     $<$<AND:$<CONFIG:Sanitized>,$<PLATFORM_ID:Darwin>>:-fsanitize-trap=all>

)



#-------------------------------------------------------------------------------

graph_driver/xrays.cpp

+1 −1
Original line number Diff line number Diff line
@@ -40,7 +40,7 @@ int main(int argc, const char * argv[]) { 
    const size_t num_steps = num_times/sub_steps;

    const size_t num_rays = 100000;



    std::vector<std::thread> threads(std::max(std::min(static_cast<unsigned int> (jit::context<base, use_safe_math>::max_concurrency),

    std::vector<std::thread> threads(std::max(std::min(static_cast<unsigned int> (jit::context<base, use_safe_math>::max_concurrency()),

                                                       static_cast<unsigned int> (num_rays)),

                                              static_cast<unsigned int> (1)));

graph_framework/cuda_context.hpp

+8 −4
Original line number Diff line number Diff line
@@ -16,8 +16,12 @@ 
#include "node.hpp"



namespace gpu {

//  Initialize the cuda driver.

    static const CUresult init = cuInit(0);

//------------------------------------------------------------------------------

///   @brief Initalize cuda.

//------------------------------------------------------------------------------

    static void init() {

        cuInit(0);

    }



//------------------------------------------------------------------------------

///  @brief Class representing a cuda gpu context.
@@ -74,7 +78,7 @@ namespace gpu { 
///  @params[in] result Result code of the operation.

///  @params[in] name   Name of the operation.

//------------------------------------------------------------------------------

        static void check_error_async(CUresult result,

        void check_error_async(CUresult result,

                               const std::string &name) {

            check_error(result, name);

#ifndef NDEBUG

graph_framework/jit.hpp

+20 −2
Original line number Diff line number Diff line
@@ -18,12 +18,24 @@ 
#ifdef USE_METAL

#define START_GPU @autoreleasepool {

#define END_GPU }

#elif defined (USE_CUDA)

#define START_GPU jit::init();

#define END_GPU

#else

#define START_GPU

#define END_GPU

#endif



namespace jit {

//------------------------------------------------------------------------------

///   @brief Initalize the gpu\_context.

//------------------------------------------------------------------------------

    static void init() {

#ifdef USE_CUDA

        gpu::init();

#endif

    }



//------------------------------------------------------------------------------

///  @brief Class for JIT compile of the GPU kernels.

//------------------------------------------------------------------------------
@@ -52,8 +64,14 @@ namespace jit { 
        gpu_context_type gpu_context;



    public:

///  The maximum available concurrency.

        static inline size_t max_concurrency = gpu_context_type::max_concurrency();

//------------------------------------------------------------------------------

///  @brief Get the maximum number of concurrent instances.

///

///  @returns The maximum available concurrency.

//------------------------------------------------------------------------------

        static size_t max_concurrency() {

            return gpu_context_type::max_concurrency();

        }



//------------------------------------------------------------------------------

///  @brief Construct a jit context object.