Is it possible to measure the amount of GPU time/cycles a rendering sequence used that is vendor-independent? I am interested in measuring the performance of the GPU, not the CPU, and the question is not limited to any specific API. What's the most effective and efficient way of doing this?
I'm not sure about "vendor-independent" solution for each single case (actually with exception for CUDA, which is NVidia only, you get vendor independent timing)...
For CUDA - you can use cudaEvent_t type and cudaEventCreate, cudaEventRecord, cudaEventSynchronize and cudaEventElapsedTime (dont forget to cudaEventDestroy in the end) - google up the functions in their docs if you are using CUDA.
For OpenCL - you can use clGetEventProfilingInfo (this is vendor independent, but applies only on OpenCL code).
For OpenGL - read about GL_ARB_timer_query extension (promoted to core with OpenGL 3.3 ... technically taken this is also vendor independent)
For Direct3D - you can profile DirectX 11 with queries too - google for it, I think MJP (one of the guy from GameDev.net) had an article about them. Can't remember the link though. (And yeah, this is also vendor independent)