From 327ae489df8129efe40363b3a74fc65c2b9777af Mon Sep 17 00:00:00 2001 From: Philippe Tillet Date: Thu, 6 Feb 2020 13:59:42 -0500 Subject: [PATCH] [GENERAL] Updated README.md --- README.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index b34943354..2f10338f5 100644 --- a/README.md +++ b/README.md @@ -1,6 +1,6 @@ # Triton -This is the development repository of Triton, a language and compiler for writing highly efficient custom Deep-Learning primitives. The aim of Triton is to provide an open-source environment to write custom ops at higher productivity than CUDA, but also with much higher flexibility than [TVM](https://github.com/apache/incubator-tvm). +This is the development repository of Triton, a language and compiler for writing highly efficient custom Deep-Learning primitives. The aim of Triton is to provide an open-source environment to write fast code at higher productivity than CUDA, but also with much higher flexibility than [TVM](https://github.com/apache/incubator-tvm) and without having to manually specify compute schedules. The main scope of Triton at the moment are: - **Triton-C**: An imperative, single-threaded language for writing highly efficient compute-kernels at a relatively high abstraction level using numpy-like extensions of the C language. @@ -16,7 +16,7 @@ The formal foundations of this project are described in the following MAPL2019 p ## Installation -Triton is a fairly self-contained package and uses its own parser (forked from [wgtcc](https://github.com/wgtdkp/wgtcc)) and LLVM code-generator. However, at the moment it relies on LLVM-8.0+ for PTX code generation. +Triton is a fairly self-contained package and uses its own parser (forked from [wgtcc](https://github.com/wgtdkp/wgtcc)) and LLVM code-generator. However, at the moment it relies on LLVM-8.0+ for PTX code generation. The whole compiler stack (~30k lines of C++ code) should take around 15 secs to compile. ``` sudo apt-get install llvm-8-dev