Skip to content
This repository has been archived by the owner on Sep 12, 2018. It is now read-only.

Latest commit

 

History

History
37 lines (34 loc) · 2.2 KB

INSTALL.md

File metadata and controls

37 lines (34 loc) · 2.2 KB
Raw

TH++: installation

TH++ requires a Linux x86_64 system with a recent version of gcc (4.8+) that supports C++11. We confirmed this installation procedure on Ubuntu 13.10 and 14.04 LTS, but other recent versions of Linux should work as well with some changes.

tl;dr - Install Torch. Then run THPP_NOFB=1 build.sh. For full functionality, also install folly and fbthrift, and instead run build.sh.

  1. Install Torch. The Torch home page has simple scripts to automate installing Torch on Ubuntu. NOTE that, even though you might already have Torch installed, you should reinstall, as older versions do not install LuaJIT with Lua 5.2 compatibility. To check, run luajit -e ';;' -- if you get an error ("unexpected symbol near ';'"), then you need to reinstall.
  2. (optional) Install folly. The folly README lists the packages (all from the standard Ubuntu distribution) that you need installed on your system before compiling folly.
  3. (optional) Install fbthrift. fbthrift depends on folly, and fbthrift's README lists additional required packages (again, from the standard Ubuntu distribution).
  4. Compile and build TH++. This is a standard cmake project; see cd thpp; ./build.sh, or use cmake directly. If you didn't install folly or fbthrift, you will have to exclude parts that use them from the build. This can be achieved with NO_FOLLY and NO_THRIFT cmake options.
  5. Just like most cmake projects, TH++ builds in a separate build directory; if anything goes wrong during the build and you want to start over, just delete the build directory and run build.sh again. THPP_NOFB=1 build.sh installs without folly and thrift. Some functions will be broken, but most of the usability is there.
  6. Confirm installation; if you used the default installation options, you should have /usr/local/include/thpp/Tensor.h and /usr/local/lib/libthpp.so.