Fix profiling calculation for CPU profiling path.

-There was a precision problem with timestamp calculation, all math was using
integers which are not very precise in overflow scenarios
-Change the logic to use doubles and cast back to uint64_t at the end.

Change-Id: Ia08d504a90a43df7330f398af966535ed944650d

runtime/os_interface/windows/os_time.cpp

@@ -85,11 +85,9 @@ bool OSTimeWin::getCpuGpuTime(TimeStampData *pGpuCpuTime) {
 
 bool OSTimeWin::getCpuTime(uint64_t *timeStamp) {
     uint64_t time;
-    uint64_t frequency;
-    QueryPerformanceCounter((LARGE_INTEGER *)&time);
-    QueryPerformanceFrequency((LARGE_INTEGER *)&frequency);
+    this->QueryPerfomanceCounterFnc((LARGE_INTEGER *)&time);
 
-    *timeStamp = time * NSEC_PER_SEC / frequency;
+    *timeStamp = static_cast<uint64_t>((static_cast<double>(time) * NSEC_PER_SEC / frequency.QuadPart));
     return true;
 };
 
@@ -126,7 +124,7 @@ double OSTimeWin::getDynamicDeviceTimerResolution(HardwareInfo const &hwInfo) co
 
 uint64_t OSTimeWin::getCpuRawTimestamp() {
     LARGE_INTEGER cpuRawTimestamp = {};
-    QueryPerformanceCounter(&cpuRawTimestamp);
+    this->QueryPerfomanceCounterFnc(&cpuRawTimestamp);
     return cpuRawTimestamp.QuadPart;
 }
 } // namespace OCLRT

runtime/os_interface/windows/os_time.h

@@ -42,6 +42,7 @@ class OSTimeWin : public OSTime {
     Wddm *wddm;
     LARGE_INTEGER frequency;
     OSTimeWin() {}
+    decltype(&QueryPerformanceCounter) QueryPerfomanceCounterFnc = QueryPerformanceCounter;
 };
 
 typedef enum GTDI_ESCAPE_FUNCTION_ENUM {