summaryrefslogtreecommitdiff
path: root/src/cuda-sim
AgeCommit message (Collapse)Author
2018-11-03merged with memory subsytem. Regression is passing but tensorcore kernel is ↵aamir
stuck in deadlock
2018-11-02bug in the last commitaamir
2018-11-02some minor changes asked in the code reviewaamir
2018-11-02added incount and outcount, addressed all of the pull request commentsaamir
2018-10-26made the changes compatible with old config filesMd Aamir Raihan
2018-10-24merged tensor-cores codeaamir
2018-10-05Merge pull request #15 from gjulianm/devTim Rogers
Naive fix of OpenCL parse errors
2018-10-05Merge pull request #76 from tgrogers/devTim Rogers
Changes to allow correct occupancy with newer cards, while still running PTXPlus
2018-10-05Merge pull request #66 from pigrew/cleanup-clang-warningsTim Rogers
Cleanup clang warnings
2018-09-22debug printsaamir
2018-09-16print for mem transactionaamir
2018-09-14corrected the bug in vpcore functional modelaamir
2018-09-10added support for generating correct no of mem txnaamir
2018-09-09generating correct number of mem transactionaamir
2018-09-09minor changes for generating mem transaction in timing model. NOTE NOT COMPLETEDaamir
2018-08-16add latencynegargoli
2018-08-16Timing model for VCOREnegargoli93
2018-08-16fix alignment bug in parser, modified ptxjitconfig, minor fixesJonathan
2018-08-14parsing changes for timing modelaamir
2018-08-13generalized v4 completed;Note I have only added row major support for matrix baamir
2018-08-12tested b4 b8 b16 modeaamir
2018-08-10added mma and st inst. Only different b mode needs to be addedaamir
2018-08-10fixes ptxjitplus and cuLaunchKernelJonathan
2018-08-10corrected a bug in bfi implementationaamir
2018-08-09added loadaamir
2018-08-08brev implementationJ
2018-08-07implemented prmt and started working on variable precision mul instaamir
2018-08-07working fix for deadlock due to operand collector, parser changes to support ↵J
culaunchkernel
2018-07-23brought back ptxinfoJonathan
2018-07-22added regression and debugged the failing testcaseaamir
2018-07-21merging the changes of cutlass on negar tensorcore branchaamir
2018-07-20load param offsets with refactored code and look in middle of malloc'd area ↵J
for pointer
2018-07-18added c++filt that works better, param offset dumping, protection for ↵Jonathan
failing system calls
2018-07-16fp32 to fp16 and fp16 to fp32 cvtJonathan
2018-07-16fix pipline for tensor_core and change confignegargoli93
2018-07-13Add more print statements to debug api callsJonathan
Fixed bug with 32bit rem Fixed bug with launching kernels in ptxjitplus Removed extraneous files Cleaned up more pointers Added documentation for ptxjitplus
2018-07-09fixed pointer bug in ptx jit and added support for executing ptxjitconfig ↵Jonathan
when running ptxjit
2018-07-09better launchkernels and more printouts for debuggingJonathan
2018-07-04dumps ptx kernels and scripts to launch all kernelsJonathan
2018-07-04dump and load block and grid size and launchJonathan
2018-07-03loading params works nowJonathan
2018-06-30Cleaning up the setup env message a bit and having the simulator properly ↵tgrogers
dump to stderr on error
2018-06-30changing the flow s.t. you cannot compile the simulator with a different ↵tgrogers
version of CUDA from the app.
2018-06-29Fixing the PTXPLUS + new CUDA execution case.tgrogers
2018-06-29dumps global using pointer, dump output from ptxjit to fileJonathan
2018-06-29Allowing modern GPU configrations to properly run PTXPLUS.tgrogers
There are several problems with this: 1) To get proper occupancy, based on register file usage, you must querry ptxas with the sm version that corresponds to the register usage in the config file. To enable this, a new config option has been added that determines what sm version you pass to ptxas to compute reg-usage. This configuration option is always required in the gpgpusim.config file 2) If you are running PTXPLUS with a modern card (i.e. volta/pascal), you need ptxas from CUDA 9.1. However since PTXPLUS only supports sm_13 - you need a version of CUDA where cudaobjectdump supports sm_1x. This ended at CUDA 5 - and PTXPLUS requires CUDA 4.2. Therefore, to run PTXPLUS on a modern card, you need CUDA 4.2 + modern CUDA installed. To fascilitate this, a new envronment varaible is added and the setup envrionment script prints an appropraite warning if you are using a newer CUDA. We have tried to make this as fail-proof as possible - and die appropraitely when something is wrong.
2018-06-28dumps pointers by accessing global memoryJonathan
2018-06-28Tests to find conditions that a value is a pointer and new mallocPtr_SizeJonathan
2018-06-27WIP dump paramsJonathan
2018-06-26exports and imports param data in new debug tool: WatchYourStepJonathan