# Emulation/CPU=correct,Execution/GPU=incorrect emulation **URL:** \ **Category:** CUDA Programming and Performance\ **Created:** [August 26, 2008, 5:39pm UTC](https://forums.developer.nvidia.com/t/emulation-cpu-correct-execution-gpu-incorrect-emulation/5190 "2008-08-26T17:39:44Z")\ **Posts on this page:** 7\ **Page:** 2
**Author:** ![chrismc](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@chrismc](https://forums.developer.nvidia.com/u/chrismc)\ **Post date:** [September 1, 2008, 10:31am UTC](https://forums.developer.nvidia.com/t/emulation-cpu-correct-execution-gpu-incorrect-emulation/5190/21 "2008-09-01T10:31:27Z")
> [@](#): > > Not necessarily. At the moment I have only one thread perform one function while the rest wait (I haven’t yet thought how that function can be parallelised). I though the lack of syncthreads would be shown in the emulation anyway. > > \_\_syncthreads synchronises threads in a block, not in a warp, is that correct? BUt threads are executed in warps ie groups of 32? > > So when working with 400 threads/particles in 1 block and I place a \_\_syncthreads call after a function call in a kernel, not one thread should pass that \_\_syncthreads until all other 399 have reached that particular \_\_syncthreads? So with a maximum of 128 threads executing concurrently on a C870, ie 4 warps, with 400 threads not all threads are being computed concurrently. > > I’ll try multiples of warps on both machines. I expect something interesting to happen with 5 warps. > > [snapback]432808[/snapback] results from both the parallel machine and the C870 agree upto and including 6 warps, but from 7 warps onwards the results from the C870 show unexpected behaviour while the parallel machine results are as expected. Plus there is an unexpected performance gain at warps\>=7, in that the execution time drops dramatically from seconds to fractions of a millisecond. Any suggestions? ---
**Author:** ![E.D\_Riedijk](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@E.D\_Riedijk](https://forums.developer.nvidia.com/u/E.D_Riedijk)\ **Post date:** [September 1, 2008, 12:49pm UTC](https://forums.developer.nvidia.com/t/emulation-cpu-correct-execution-gpu-incorrect-emulation/5190/22 "2008-09-01T12:49:00Z")
check for errors after your kernel launch, you probably have to many registers for 7 warps per block. I think you are experiencing a too many resources requested error. ---
**Author:** ![chrismc](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@chrismc](https://forums.developer.nvidia.com/u/chrismc)\ **Post date:** [September 1, 2008, 12:58pm UTC](https://forums.developer.nvidia.com/t/emulation-cpu-correct-execution-gpu-incorrect-emulation/5190/23 "2008-09-01T12:58:05Z")
> [@](#): > > check for errors after your kernel launch, you probably have to many registers for 7 warps per block. I think you are experiencing a too many resources requested error. > > [snapback]433952[/snapback] I’m beginning to suspect it could be something like that. How do I check for the error you mention? From the example projects I have in my code //////////////////////////////////// // Launch the device computation SPH\<\<\\>\>(//variables); CUT\_CHECK\_ERROR(“kernel error”); //////////////////////////////////// Do I need to write something else? ---
**Author:** ![theMarix](https://sea2.discourse-cdn.com/nvidia/user_avatar/forums.developer.nvidia.com/themarix/32/12058_2.png) [@theMarix](https://forums.developer.nvidia.com/u/theMarix)\ **Post date:** [September 1, 2008, 2:37pm UTC](https://forums.developer.nvidia.com/t/emulation-cpu-correct-execution-gpu-incorrect-emulation/5190/24 "2008-09-01T14:37:19Z")
Try to cudaThreadSynchronize before checking for the error value. (Actually cudaThreadSynchronize will report the error if there is one.) ---
**Author:** ![cbuchner1](https://sea2.discourse-cdn.com/nvidia/user_avatar/forums.developer.nvidia.com/cbuchner1/32/14168_2.png) [@cbuchner1](https://forums.developer.nvidia.com/u/cbuchner1)\ **Post date:** [September 1, 2008, 9:57pm UTC](https://forums.developer.nvidia.com/t/emulation-cpu-correct-execution-gpu-incorrect-emulation/5190/25 "2008-09-01T21:57:07Z")
I just tripped over the fact that \_\_sinf only works within [-pi, pi] whereas in Emu mode it works correctly at all times. Similarly with \_\_cosf, I suppose. This screwed over my Hough Transform. ;) ---
**Author:** ![alex\_dubinsky](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@alex\_dubinsky](https://forums.developer.nvidia.com/u/alex_dubinsky)\ **Post date:** [September 1, 2008, 10:14pm UTC](https://forums.developer.nvidia.com/t/emulation-cpu-correct-execution-gpu-incorrect-emulation/5190/26 "2008-09-01T22:14:37Z")
> [@](#): > > //////////////////////////////////// > > // Launch the device computation > > SPH\<\<\\>\>(//variables); > > CUT\_CHECK\_ERROR(“kernel error”); > > //////////////////////////////////// > > Do I need to write something else? > > [snapback]433956[/snapback] Don’t forget to run in Debug mode. Check the custom build step of the sample in case anything doesn’t work. ---
**Author:** ![Sylvain\_Collange](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@Sylvain\_Collange](https://forums.developer.nvidia.com/u/Sylvain_Collange)\ **Post date:** [September 2, 2008, 5:01pm UTC](https://forums.developer.nvidia.com/t/emulation-cpu-correct-execution-gpu-incorrect-emulation/5190/27 "2008-09-02T17:01:35Z")
> [@](#): > > I just tripped over the fact that \_\_sinf only works within [-pi, pi]  whereas in Emu mode it works correctly at all times. Similarly with \_\_cosf, I suppose. This screwed over my Hough Transform. ;) > > [snapback]434167[/snapback] More precisely, the **absolute** error of \_\_sinf and \_\_cosf is only guaranteed within [-pi, pi]. Outside this range, it will be less accurate, but still usable for most purposes (well, maybe not yours!) You can think of \_\_sinf(x) as behaving much like sinf(fmodf(x, (float)TWO\_PI)). Only for very large inputs it will give a completely wrong answer in absolute terms. But even inside [-pi, pi], CPU implementations can be much more accurate than \_\_sinf and \_\_cosf, if accurate means a smaller **relative** error (think \_\_sinf(x) when x is very close to 0). [Previous page](https://forums.developer.nvidia.com/t/emulation-cpu-correct-execution-gpu-incorrect-emulation/5190.md?page=1)