NVCC improvements.

- Complete translation tables for cudaError <-> hipError_t.
- Remove some odd errors that were not correctly translated or not used.
- Add HIPCHECK_API to test infrastructure.  Used for negative testing
  an API ; if a mismatch occurs it shows the expected return error
  code.  Can also print a warning rather than error.
- Enable hipMemoryAllocate on NV system, and review error coded.
- Add hipErrorName to nvcc.

Change-Id: I680427dcf32a5796d5913cf9e7f3b4c6f6b91599

Conflicts:
	tests/src/CMakeLists.txt

Bug fixes and improved docs for hipFree and hipHostFree.

    - Passing NULL pointer initialized runtime and return hipSuccess
      (not an error like before).
    - add negative test for this. (hipMemoryAllocate, improved)
    - Match NVCC errors for invalid pointers, add to test.
    - Update hipFree and hipHostFree docs.
    - hipGetDevicePointer always set *devicePointer=NULL, even for
      invalid flags.
    - Gate shared memory usage on specific HCC work-week.

Change-Id: I533b4fd3280a3d6cdbf05eb768976f0c7506c012

src/device_util.cpp

@@ -27,6 +27,12 @@ THE SOFTWARE.
 using namespace hc::precise_math;
 #endif
 
+#if __hcc_workweek__ > 16186
+#define USE_DYNAMIC_SHARED 1
+#else
+#define USE_DYNAMIC_SHARED 0
+#endif
+
 #define HIP_SQRT_2 1.41421356237
 
 #define __hip_erfinva3 -0.140543331
@@ -606,6 +612,11 @@ __device__ float __hip_y1f(float x)
     return ret;
 }
 
+#if __hcc_workweek__ > 16186
+#define USE_DYNAMIC_SHARED 1
+#else
+#define USE_DYNAMIC_SHARED 0
+#endif
 
 __device__ float acosf(float x)
 {
@@ -1634,10 +1645,12 @@ __host__ __device__ int max(int arg1, int arg2)
   return (int)(hc::precise_math::fmax((float)arg1, (float)arg2));
 }
 
+#if USE_DYNAMIC_SHARED
 __device__ __attribute__((address_space(3))) void* __get_dynamicgroupbaseptr()
 {
   return hc::get_dynamic_group_segment_base_pointer();
 }
+#endif
 
 
 

src/hip_hcc.cpp

@@ -100,7 +100,7 @@ hsa_agent_t g_cpu_agent;
 ihipSignal_t::ihipSignal_t() :  _sig_id(0)
 {
     if (hsa_signal_create(0/*value*/, 0, NULL, &_hsa_signal) != HSA_STATUS_SUCCESS) {
-        throw ihipException(hipErrorOutOfResources);
+        throw ihipException(hipErrorRuntimeMemory);
     }
     //tprintf (DB_SIGNAL, "  allocated hsa_signal=%lu\n", (_hsa_signal.handle));
 }
@@ -110,7 +110,7 @@ ihipSignal_t::~ihipSignal_t()
 {
     tprintf (DB_SIGNAL, "  destroy hsa_signal #%lu (#%lu)\n", (_hsa_signal.handle), _sig_id);
     if (hsa_signal_destroy(_hsa_signal) != HSA_STATUS_SUCCESS) {
-       throw ihipException(hipErrorOutOfResources);
+       throw ihipException(hipErrorRuntimeOther);
     }
 };
 
@@ -1137,9 +1137,7 @@ const char *ihipErrorString(hipError_t hip_error)
     switch (hip_error) {
         case hipSuccess                         : return "hipSuccess";
         case hipErrorMemoryAllocation           : return "hipErrorMemoryAllocation";
-        case hipErrorMemoryFree                 : return "hipErrorMemoryFree";
-        case hipErrorUnknownSymbol              : return "hipErrorUnknownSymbol";
-        case hipErrorOutOfResources             : return "hipErrorOutOfResources";
+        case hipErrorLaunchOutOfResources       : return "hipErrorLaunchOutOfResources";
         case hipErrorInvalidValue               : return "hipErrorInvalidValue";
         case hipErrorInvalidResourceHandle      : return "hipErrorInvalidResourceHandle";
         case hipErrorInvalidDevice              : return "hipErrorInvalidDevice";

src/hip_memory.cpp

@@ -89,6 +89,8 @@ hipError_t hipHostGetDevicePointer(void **devicePointer, void *hostPointer, unsi
 
     hipError_t e = hipSuccess;
 
+    *devicePointer = NULL;
+
     // Flags must be 0:
     if (flags != 0) {
         e = hipErrorInvalidValue;
@@ -100,7 +102,6 @@ hipError_t hipHostGetDevicePointer(void **devicePointer, void *hostPointer, unsi
             *devicePointer = amPointerInfo._devicePointer;
         } else {
             e = hipErrorMemoryAllocation;
-            *devicePointer = NULL;
         }
     }
     return ihipLogStatus(e);
@@ -530,6 +531,9 @@ hipError_t hipFree(void* ptr)
                 hipStatus = hipSuccess;
             }
         }
+    } else {
+        // free NULL pointer succeeds and is common technique to initialize runtime
+        hipStatus = hipSuccess;  
     }
 
     return ihipLogStatus(hipStatus);
@@ -540,10 +544,11 @@ hipError_t hipHostFree(void* ptr)
 {
     HIP_INIT_API(ptr);
 
-    // TODO - ensure this pointer was created by hipMallocHost and not hipMalloc
-    std::call_once(hip_initialized, ihipInit);
+   // Synchronize to ensure all work has finished.
+    ihipGetTlsDefaultDevice()->locked_waitAllStreams(); // ignores non-blocking streams, this waits for all activity to finish.
 
-    hipError_t hipStatus = hipErrorInvalidDevicePointer;
+
+    hipError_t hipStatus = hipErrorInvalidValue;
     if (ptr) {
         hc::accelerator acc;
         hc::AmPointerInfo amPointerInfo(NULL, NULL, 0, acc, 0, 0);
@@ -554,6 +559,9 @@ hipError_t hipHostFree(void* ptr)
                 hipStatus = hipSuccess;
             }
         }
+    } else {
+        // free NULL pointer succeeds and is common technique to initialize runtime
+        hipStatus = hipSuccess;  
     }
 
     return ihipLogStatus(hipStatus);