sarchlab/mgpusim

Issues

• Can the GPU simulator simulate different GPU interconnect topologies such as RING and ALL-TO-ALL?

  #89 opened 22 days ago by christindbose
  1

• Report Block Count in each simulation

  #87 opened a month ago by syifan
  0

• Cache access count does not match in metrics.csv and mem.trace

  #85 opened 2 months ago by syifan
  0

• BUG with command -use-unified-memory

  #81 opened 2 months ago by InnoBogus
  1

• Redesign RDMA

  #80 opened 2 months ago by DX990307
  0

• total_kernel_time = 0 when using unified-gpus

  #78 opened 2 months ago by InnoBogus
  1

• vgg16 benchmark

  #71 opened 3 months ago by syifan
  0

• DRAM Latency is not recorded in `metrics.csv`

  #69 opened 3 months ago by syifan
  0

• Failed Execution for Program Compiled with ROCm 1.7.1

  #64 opened 4 months ago by MyNewAcc1234
  2

• Support HIP, as well as OpenCL kernels compiled by new ROCm versions

  #34 opened 8 months ago by syifan
  3

• Rewrite the DNN in a way that all the memory are passed from outside

  #53 opened 5 months ago by syifan
  0

• Update instruction tracing so that all instructions show in one interface

  #56 opened 5 months ago by syifan
  0

• adding perfRegister

  #51 opened 6 months ago by DX990307
  0

• Perf analyzer cannot be used in mgpusim

  #49 opened 6 months ago by DX990307
  0

• MMU is not being traced for visualization

  #44 opened 6 months ago by syifan
  0

• Larger simulations are not deterministic

  #32 opened 9 months ago by syifan
  11

• Parallel simulation results is very different from serial simulation

  #45 opened 6 months ago by syifan
  1

• where is private segment memory allocation for kernel load ?

  #42 opened 6 months ago by bufans
  5

• We need a GMMU implementation that can walk the page table locally on each GPU

  #41 opened 7 months ago by syifan
  0

• Add a flag `--no-monitor` to disable AkitaRTM

  #39 opened 7 months ago by syifan
  0

• NVIDIA: Support block dispatching that considers SM capacity.

  #37 opened 8 months ago by syifan
  0

• Add units to metrics.csv report

  #4 opened a year ago by syifan
  0

• Report WF count per GPU

  #24 opened a year ago by syifan
  0

• Deterministic test is not working

  #17 opened 9 months ago by syifan
  0

• Report RDMA Trace

  #26 opened 9 months ago by DX990307
  1

• While VALU should take 4 cycles to execute, it should not be a pipeline.

  #31 opened 9 months ago by syifan
  0

• VALU instructions should take 4 cycles rather than 5 cycles.

  #29 opened 9 months ago by syifan
  0

• DRAM latency is reported as 0

  #19 opened 9 months ago by syifan
  0

• Show progress bar for memory copy

  #23 opened a year ago by syifan
  0

• Analyzer is not working

  #16 opened a year ago by syifan
  1

• Merge `gitlab.com/akita/dnn` into `github.com/sarchlab/mgpusim`

  #12 opened a year ago by syifan
  0

• RDMA itself should not be a performance bottleneck

  #8 opened a year ago by syifan
  0

• RMDA should have a builder

  #10 opened a year ago by syifan
  0

• ISADebugger: Tracer is only tracing a specific WF. The condition should remove.

  #11 opened a year ago by syifan
  0

• The design of the scratchpad is impacting the performance

  #7 opened a year ago by syifan
  0

• MultiNode Simulation

  #6 opened a year ago by amelfatima1231
  13