You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@tvm.apache.org by Luis Vega <no...@github.com> on 2019/04/11 23:28:55 UTC

[dmlc/tvm] [RFC] [VTA] [TSIM] Enabling Cycle-Accurate Hardware Simulation for VTA #3009 (#3010)

RFC in #3009

You can view, comment on, or merge this pull request online at:

  https://github.com/dmlc/tvm/pull/3010

-- Commit Summary --

  * merge files
  * move verilator to the right place
  * change name to tsim

-- File Changes --

    M cmake/modules/VTA.cmake (5)
    A vta/apps/tsim/CMakeLists.txt (22)
    A vta/apps/tsim/Makefile (26)
    A vta/apps/tsim/cmake/modules/driver.cmake (8)
    A vta/apps/tsim/cmake/modules/tsim.cmake (135)
    A vta/apps/tsim/hardware/chisel/Makefile (2)
    A vta/apps/tsim/hardware/chisel/build.sbt (50)
    A vta/apps/tsim/hardware/chisel/project/build.properties (1)
    A vta/apps/tsim/hardware/chisel/project/plugins.sbt (1)
    A vta/apps/tsim/hardware/chisel/src/main/scala/accel/Accel.scala (33)
    A vta/apps/tsim/hardware/chisel/src/main/scala/accel/Compute.scala (95)
    A vta/apps/tsim/hardware/chisel/src/main/scala/accel/RegFile.scala (86)
    A vta/apps/tsim/hardware/chisel/src/test/scala/dut/TestAccel.scala (51)
    A vta/apps/tsim/hardware/verilog/Accel.v (103)
    A vta/apps/tsim/hardware/verilog/Compute.v (139)
    A vta/apps/tsim/hardware/verilog/RegFile.v (129)
    A vta/apps/tsim/hardware/verilog/TestAccel.v (98)
    A vta/apps/tsim/python/tsim/__init__.py (0)
    A vta/apps/tsim/python/tsim/config.json (7)
    A vta/apps/tsim/python/tsim/config.py (44)
    A vta/apps/tsim/python/tsim/load.py (39)
    A vta/apps/tsim/src/test_driver.cc (73)
    A vta/apps/tsim/tests/python/test_tsim.py (21)
    A vta/hardware/chisel/Makefile (2)
    A vta/hardware/chisel/build.sbt (49)
    A vta/hardware/chisel/project/build.properties (1)
    A vta/hardware/chisel/project/plugins.sbt (1)
    A vta/hardware/chisel/src/main/resources/verilog/VTAHostDPI.v (101)
    A vta/hardware/chisel/src/main/resources/verilog/VTAMemDPI.v (87)
    A vta/hardware/chisel/src/main/scala/dpi/VTAHostDPI.scala (53)
    A vta/hardware/chisel/src/main/scala/dpi/VTAMemDPI.scala (54)
    A vta/include/vta/verilator/dpi_module.h (25)
    A vta/include/vta/verilator/tsim.h (88)
    A vta/src/verilator/dpi_module.cc (353)
    A vta/src/verilator/tsim.cc (113)

-- Patch Links --

https://github.com/dmlc/tvm/pull/3010.patch
https://github.com/dmlc/tvm/pull/3010.diff

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/dmlc/tvm/pull/3010

Re: [dmlc/tvm] [RFC] [VTA] [TSIM] Enabling Cycle-Accurate Hardware Simulation for VTA #3009 (#3010)

Posted by Tianqi Chen <no...@github.com>.
High-level structural comments, given that dpi part(verilog shell and c file) of the codebase is compiled with the verilator. We should consider a distinct folder structure. Here is one possible recommendation

```
/vta
/vta/dpi  
      - VTATSimMemHost.v
      - vta_tsim_dpi.h 
      - vta_tsim_dpi.cc
      - README.md
/vta/include/vta
     - dpi_module.h (the common dpi module defs that can be re-used)
/vta/src
     - dpi_module.cc (consider links to vta instead of tvm runtime, import vta will import the dpi)
```


-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/dmlc/tvm/pull/3010#issuecomment-483472352

Re: [dmlc/tvm] [RFC] [VTA] [TSIM] Enabling Cycle-Accurate Hardware Simulation for VTA #3009 (#3010)

Posted by SharpTVRepairDubai <no...@github.com>.
Great discussion, learned a lot. Thanks, 
@misc{poster:Ganjeloo:2018
We repair all brands of tv. Get your tv repaired from the comfort of home. 100% safe & reliable [Sharp TV Repair Dubai](https://f2help.com/sharp-tv-repair/) Service by our trusted and top-rated technicians.

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/dmlc/tvm/pull/3010#issuecomment-514084973

Re: [dmlc/tvm] [RFC] [VTA] [TSIM] Enabling Cycle-Accurate Hardware Simulation for VTA #3009 (#3010)

Posted by Liangfu Chen <no...@github.com>.
Thanks @vegaluisjose and all the reviewers ! This is a great start for us to develop, evaluate and integrate chisel3 based implement of VTA.

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/dmlc/tvm/pull/3010#issuecomment-490792540

Re: [dmlc/tvm] [RFC] [VTA] [TSIM] Enabling Cycle-Accurate Hardware Simulation for VTA #3009 (#3010)

Posted by Thierry Moreau <no...@github.com>.
I've approved the changes @vegaluisjose @tqchen. There seems to be an issue with the pr-merge CI; perhaps the branch should be re-based?

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/dmlc/tvm/pull/3010#issuecomment-490158867

Re: [dmlc/tvm] [RFC] [VTA] [TSIM] Enabling Cycle-Accurate Hardware Simulation for VTA #3009 (#3010)

Posted by Thierry Moreau <no...@github.com>.
In order not to confuse adopters of TSIM, who would like to use TSIM to simulate any hardware instead of VTA itself, can we have a guide that is not part of the `install.md` file aimed at VTA developers?

Once VTA can be simulated by TSIM we can merge it back.

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/dmlc/tvm/pull/3010#issuecomment-489871096

Re: [dmlc/tvm] [RFC] [VTA] [TSIM] Enabling Cycle-Accurate Hardware Simulation for VTA #3009 (#3010)

Posted by Thierry Moreau <no...@github.com>.
Merged #3010 into master.

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/dmlc/tvm/pull/3010#event-2327864904

Re: [dmlc/tvm] [RFC] [VTA] [TSIM] Enabling Cycle-Accurate Hardware Simulation for VTA #3009 (#3010)

Posted by Tianqi Chen <no...@github.com>.
@tmoreau89 the ci is now green

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/dmlc/tvm/pull/3010#issuecomment-490532689

Re: [dmlc/tvm] [RFC] [VTA] [TSIM] Enabling Cycle-Accurate Hardware Simulation for VTA #3009 (#3010)

Posted by Jared Roesch <no...@github.com>.
Do we need to commit the Verilog code? isn't generated

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/dmlc/tvm/pull/3010#issuecomment-482743446

Re: [dmlc/tvm] [RFC] [VTA] [TSIM] Enabling Cycle-Accurate Hardware Simulation for VTA #3009 (#3010)

Posted by Tianqi Chen <no...@github.com>.
@tmoreau89 can you review again and https://docs.tvm.ai/contribute/code_review.html#approve-and-request-changes-explicitly?

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/dmlc/tvm/pull/3010#issuecomment-489866068

Re: [dmlc/tvm] [RFC] [VTA] [TSIM] Enabling Cycle-Accurate Hardware Simulation for VTA #3009 (#3010)

Posted by Luis Vega <no...@github.com>.
Ok I have addressed most of the issues, except for `unsigned long long` complains cpplint is giving on arguments for DPI functions. The reason why I used this type is because this is the same type used by Verilator when it compiles 64-bit hardware types to C. 

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/dmlc/tvm/pull/3010#issuecomment-488847732

Re: [dmlc/tvm] [RFC] [VTA] [TSIM] Enabling Cycle-Accurate Hardware Simulation for VTA #3009 (#3010)

Posted by Tianqi Chen <no...@github.com>.
@jroesch @tmoreau89 please help to review this PR

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/dmlc/tvm/pull/3010#issuecomment-488727075