[HIP] [HIPIFY] [FIX] cuModuleLoadDataEx -> hipModuleLoadDataEx

https://github.com/GPUOpen-ProfessionalCompute-Tools/HIP/issues/81 1. Do not use JIT options on HCC path, call hipModuleLoadData instead. 2. NVCC path is unchanged, to call cuModuleLoadDataEx with all options. 3. Get rid of manual hipification, based on #ifdef #else for NVCC/HIP. 4. Update documentation accordingly.
2017-05-19 17:22:14 +03:00
commit 270f643c9c
@@ -1915,7 +1915,7 @@ hipError_t hipModuleGetFunction(hipFunction_t *function, hipModule_t module, con
 * @brief returns device memory pointer and size of the kernel present in the module with symbol @p name
 *
 * @param [out] dptr
- * @param [out[ bytes
+ * @param [out] bytes
 * @param [in] hmod
 * @param [in] name
 *
@@ -1923,7 +1923,6 @@ hipError_t hipModuleGetFunction(hipFunction_t *function, hipModule_t module, con
 */
 hipError_t hipModuleGetGlobal(hipDeviceptr_t *dptr, size_t *bytes, hipModule_t hmod, const char *name);

-
 /**
 * @brief builds module from code object which resides in host memory. Image is pointer to that location.
 *
@@ -1934,11 +1933,23 @@ hipError_t hipModuleGetGlobal(hipDeviceptr_t *dptr, size_t *bytes, hipModule_t h
 */
 hipError_t hipModuleLoadData(hipModule_t *module, const void *image);

+/**
+* @brief builds module from code object which resides in host memory. Image is pointer to that location. Options are not used. hipModuleLoadData is called.
+*
+* @param [in] image
+* @param [out] module
+* @param [in] number of options
+* @param [in] options for JIT
+* @param [in] option values for JIT
+*
+* @returns hipSuccess, hipErrorNotInitialized, hipErrorOutOfMemory, hipErrorNotInitialized
+*/
+hipError_t hipModuleLoadDataEx(hipModule_t *module, const void *image, unsigned int numOptions, hipJitOption *options, void **optionValues);

 /**
 * @brief launches kernel f with launch parameters and shared memory on stream with arguments passed to kernelparams or extra
 *
- * @param [in[ f	 Kernel to launch.
+ * @param [in] f         Kernel to launch.
 * @param [in] gridDimX  X grid dimension specified as multiple of blockDimX.
 * @param [in] gridDimY  Y grid dimension specified as multiple of blockDimY.
 * @param [in] gridDimZ  Z grid dimension specified as multiple of blockDimZ.
@@ -1946,7 +1957,7 @@ hipError_t hipModuleLoadData(hipModule_t *module, const void *image);
 * @param [in] blockDimY Y grid dimension specified in work-items
 * @param [in] blockDimZ Z grid dimension specified in work-items
 * @param [in] sharedMemBytes Amount of dynamic shared memory to allocate for this kernel.  The kernel can access this with HIP_DYNAMIC_SHARED.
- * @param [in] stream Stream where the kernel should be dispatched.  May be 0, in which case th default stream is used with associated synchronization rules.
+ * @param [in] stream    Stream where the kernel should be dispatched.  May be 0, in which case th default stream is used with associated synchronization rules.
 * @param [in] kernelParams
 * @param [in] extra     Pointer to kernel arguments.   These are passed directly to the kernel and must be in the memory layout and alignment expected by the kernel.
 *
@@ -250,6 +250,30 @@ typedef enum hipDeviceAttribute_t {
    hipDeviceAttributeIsMultiGpuBoard,                      ///< Multiple GPU devices.
 } hipDeviceAttribute_t;

+/*
+* @brief hipJitOption
+* @enum
+* @ingroup Enumerations
+*/
+typedef enum hipJitOption {
+  hipJitOptionMaxRegisters = 0,
+  hipJitOptionThreadsPerBlock,
+  hipJitOptionWallTime,
+  hipJitOptionInfoLogBuffer,
+  hipJitOptionInfoLogBufferSizeBytes,
+  hipJitOptionErrorLogBuffer,
+  hipJitOptionErrorLogBufferSizeBytes,
+  hipJitOptionOptimizationLevel,
+  hipJitOptionTargetFromContext,
+  hipJitOptionTarget,
+  hipJitOptionFallbackStrategy,
+  hipJitOptionGenerateDebugInfo,
+  hipJitOptionLogVerbose,
+  hipJitOptionGenerateLineInfo,
+  hipJitOptionCacheMode,
+  hipJitOptionNumOptions
+} hipJitOption;
+
 /**
 *     @}
 */
@@ -54,26 +54,44 @@ hipMemcpyHostToHost
 #define hipFilterModePoint cudaFilterModePoint

 //! Flags that can be used with hipEventCreateWithFlags:
-#define hipEventDefault             cudaEventDefault
-#define hipEventBlockingSync        cudaEventBlockingSync
-#define hipEventDisableTiming       cudaEventDisableTiming
-#define hipEventInterprocess        cudaEventInterprocess
+#define hipEventDefault              cudaEventDefault
+#define hipEventBlockingSync         cudaEventBlockingSync
+#define hipEventDisableTiming        cudaEventDisableTiming
+#define hipEventInterprocess         cudaEventInterprocess
 #define hipEventDisableSystemRelease cudaEventDefault  /* no-op on CUDA platform */


-#define hipHostMallocDefault cudaHostAllocDefault
-#define hipHostMallocPortable cudaHostAllocPortable
-#define hipHostMallocMapped cudaHostAllocMapped
+#define hipHostMallocDefault       cudaHostAllocDefault
+#define hipHostMallocPortable      cudaHostAllocPortable
+#define hipHostMallocMapped        cudaHostAllocMapped
 #define hipHostMallocWriteCombined cudaHostAllocWriteCombined

 #define hipHostRegisterPortable cudaHostRegisterPortable
-#define hipHostRegisterMapped cudaHostRegisterMapped
+#define hipHostRegisterMapped   cudaHostRegisterMapped

 #define HIP_LAUNCH_PARAM_BUFFER_POINTER CU_LAUNCH_PARAM_BUFFER_POINTER
-#define HIP_LAUNCH_PARAM_BUFFER_SIZE     CU_LAUNCH_PARAM_BUFFER_SIZE
+#define HIP_LAUNCH_PARAM_BUFFER_SIZE    CU_LAUNCH_PARAM_BUFFER_SIZE
 #define HIP_LAUNCH_PARAM_END            CU_LAUNCH_PARAM_END
 #define hipLimitMallocHeapSize          cudaLimitMallocHeapSize
-#define hipIpcMemLazyEnablePeerAccess          cudaIpcMemLazyEnablePeerAccess
+#define hipIpcMemLazyEnablePeerAccess   cudaIpcMemLazyEnablePeerAccess
+
+// enum CUjit_option redefines
+#define hipJitOptionMaxRegisters            CU_JIT_MAX_REGISTERS
+#define hipJitOptionThreadsPerBlock         CU_JIT_THREADS_PER_BLOCK
+#define hipJitOptionWallTime                CU_JIT_WALL_TIME
+#define hipJitOptionInfoLogBuffer           CU_JIT_INFO_LOG_BUFFER
+#define hipJitOptionInfoLogBufferSizeBytes  CU_JIT_INFO_LOG_BUFFER_SIZE_BYTES
+#define hipJitOptionErrorLogBuffer          CU_JIT_ERROR_LOG_BUFFER
+#define hipJitOptionErrorLogBufferSizeBytes CU_JIT_ERROR_LOG_BUFFER_SIZE_BYTES
+#define hipJitOptionOptimizationLevel       CU_JIT_OPTIMIZATION_LEVEL
+#define hipJitOptionTargetFromContext       CU_JIT_TARGET_FROM_CUCONTEXT
+#define hipJitOptionTarget                  CU_JIT_TARGET
+#define hipJitOptionFallbackStrategy        CU_JIT_FALLBACK_STRATEGY
+#define hipJitOptionGenerateDebugInfo       CU_JIT_GENERATE_DEBUG_INFO
+#define hipJitOptionLogVerbose              CU_JIT_LOG_VERBOSE
+#define hipJitOptionGenerateLineInfo        CU_JIT_GENERATE_LINE_INFO
+#define hipJitOptionCacheMode               CU_JIT_CACHE_MODE
+#define hipJitOptionNumOptions              CU_JIT_NUM_OPTIONS

 typedef cudaEvent_t hipEvent_t;
 typedef cudaStream_t hipStream_t;
@@ -84,6 +102,7 @@ typedef cudaFuncCache hipFuncCache_t;
 typedef CUcontext hipCtx_t;
 typedef CUsharedconfig hipSharedMemConfig;
 typedef CUfunc_cache hipFuncCache;
+typedef CUjit_option hipJitOption;
 typedef CUdevice hipDevice_t;
 typedef CUmodule hipModule_t;
 typedef CUfunction hipFunction_t;
@@ -894,6 +913,11 @@ inline static hipError_t hipModuleLoadData(hipModule_t *module, const void *imag
    return hipCUResultTohipError(cuModuleLoadData(module, image));
 }

+inline static hipError_t hipModuleLoadDataEx(hipModule_t *module, const void *image, unsigned int numOptions, hipJitOption *options, void **optionValues)
+{
+  return hipCUResultTohipError(cuModuleLoadDataEx(module, image, numOptions, options, optionValues));
+}
+
 inline static hipError_t hipModuleLaunchKernel(hipFunction_t f,
      unsigned int gridDimX, unsigned int gridDimY, unsigned int gridDimZ,
      unsigned int blockDimX, unsigned int blockDimY, unsigned int blockDimZ,