ptx

In CUDA 9, nVIDIA seems to have this new notion of "cooperative groups"; and for some reason not entirely clear to me, __ballot() is now (= CUDA 9) deprecated in favor of __ballot_sync() . Is that an alias or have the semantics changed? ... similar question for other builtins which now have __sync() added to their names. No the semantics are not the same. The function calls themselves are different, one is not an alias for another, new functionality has been exposed, and the implementation behavior is now different between Volta architecture and previous architectures. First of all, to set the

:) While I was trying to manage my kernel resources I decided to look into PTX but there are a couple of things that I do not understand. Here is a very simple kernel I wrote: __global__ void foo(float* out, float* in, uint32_t n) { uint32_t idx = blockIdx.x * blockDim.x + threadIdx.x; uint32_t one = 5; out[idx] = in[idx]+one; } Then I compiled it using: nvcc --ptxas-options=-v -keep main.cu and I got this output on the console: ptxas info : 0 bytes gmem ptxas info : Compiling entry function '_Z3fooPfS_j' for 'sm_10' ptxas info : Used 2 registers, 36 bytes smem And the resulting ptx is the

When reading through CUDA 5.0 Programming Guide I stumbled on a feature called "Funnel shift" which is present in 3.5 compute-capable device, but not 3.0. It contains an annotation "see reference manual", but when I search for the "funnel shift" term in the manual, I don't find anything. I tried googling for it, but only found a mention on http://www.cudahandbook.com , in the chapter 8: 8.2.3 Funnel Shift (SM 3.5) GK110 added a 64-bit “funnel shift” instruction that may be accessed with the following intrinsics: __funnelshift_lc(): returns most significant 32 bits of a left funnel shift. _

In CUDA, each thread knows its block index in the grid and thread index within the block. But two important values do not seem to be explicitly available to it: Its index as a lane within its warp (its "lane id") The index of the warp of which it is a lane within the block (its "warp id") Assuming the grid is 1-dimensional(a.k.a. linear, i.e. blockDim.y and blockDim.z are 1), one can obviously obtain these as follows: enum : unsigned { warp_size = 32 }; auto lane_id = threadIdx.x % warp_size; auto warp_id = threadIdx.x / warp_size; and if you don't trust the compiler to optimize that, you

Is there any way by which I can know the number of free/active SMs? Or atleast to read the voltage/power or temperature values of each SM by which I can know whether its working or not? (in real time while some job is getting executed on the gpu device). %smid helped me in knowing the Id of each SM. Something similar would be helpful. Thanks and Regards, Rakesh The CUDA Profiling Tools Interface ( CUPTI ) contains an Events API that enables run time sampling of GPU PM counters. The CUPTI SDK ships as part of the CUDA Toolkit. Documentation on sampling can be found in the section CUPTI Events

Can I use a PTX function contained in a PTX file as an external device function to link it to another .cu file which should call that function? This is another question from CUDA - link kernels together where the function itself is not contained in a .cu file but I rather have a PTX function to be linked somehow. JackOLantern You can load the file containing PTX code in your own code from the filesystem by cuModuleLoad and cuModuleGetFunction as follows: CUmodule module; CUfunction function; const char* module_file = "my_ptx_file.ptx"; const char* kernel_name = "my_kernel_name"; err =