Detailed documentation on what GPGPU-Sim models are, how to configure them, and a guide to the source code can be found here: http://gpgpu-sim.org/manual/. Instructions for building Doxygen source code documentation are included below. Detailed documentation on GPUWattch, including how to configure it and a guide to the source code, can be found here: http://gpgpu-sim.org/gpuwattch/.
GPGPU-Sim dependencies: gcc, g++, make, makedepend, xutils, bison, flex, zlib, CUDA Toolkit
GPGPU-Sim documentation dependencies: doxygen, graphvi
AerialVision dependencies: python-pmw, python-ply, python-numpy, libpng12-dev, python-matplotlib
CUDA SDK dependencies: install libxi-dev libxmu-dev libglut3-dev
$source setup_environment <build_type>
- For debugging, build_type ->
debug - For normal release -> you dont need to specify build_type, blank as empty.
To build the simulator user$ make, clean executables of the simulator user$ make clean, build the Doxygen generated documentation user$ make docs, clean the docs user$ make cleandocs. This doxygen-generated documentation will help you to understand classes, templates, functions, etc.
1- ApproxTrackerL1D flushes memory requests that will probably cause a miss on the L1 data cache for load operations within the same SM. It also writes zero to the operands of the corresponding memory request.
2- ApproxTrackerL2 flushes memory requests that will probably cause a miss on one of the L2 caches for load operations. It also writes zero to the operands of the corresponding memory request.
| Config Type | Descriptions |
|---|---|
| -l1dApproximator 1 | This flag enables the ApproxTrackerL1D. |
| -l1dSlicingAccessInterval 2000 | This configuration option helps determine slicing interval size to capture temporal memory problems through the runtime. |
| -l2Approximator 0 | enables ApproxTrackerL2. |
| -l2SlicingAccessInterval 2000 | Slicing interval size for ApproxTrackerL2. |
| -errorPercentage 100 | You can set the error percentage with this flag. For example, you can think of 100 as 1%, 50 as 2%, and 200 as 0.5%. |
| -kernelSpecificApprox 1 | You may simulate the approximator for a kernel within multiple of them. Here, you need to specify the kernel id, which starts from 0. |
| -kernelIdApprox 0 | kernel id |
Note: GPPRMon metric collection parts are still valid for this simulator version.
Inside the ApproxTrackerExperiments folder, we share the simulation results for Naive, ApproxTrackerL1D, and ApproxTrackerL2 with both 2% (50) and 1% (100) flushes of missed memory requests.