From 5e7877b3300998bb6e0b9e0329baba333d9077f7 Mon Sep 17 00:00:00 2001 From: Luke Kenneth Casson Leighton Date: Mon, 23 Sep 2019 09:28:06 +0100 Subject: [PATCH] add challenges section --- nlnet_2019_gcc.mdwn | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) diff --git a/nlnet_2019_gcc.mdwn b/nlnet_2019_gcc.mdwn index 73134991d..8d1d3ba00 100644 --- a/nlnet_2019_gcc.mdwn +++ b/nlnet_2019_gcc.mdwn @@ -78,6 +78,22 @@ Whatever other compiler projects exist, they are just not compatible at the asse This is just a standard part of processor innovation. We will also have to do the same thing for LLVM at some point. +# What are significant technical challenges you expect to solve during the project, if any? + +Compiler development is known, traditionally, to be extremely technically +challenging. There are not many people in the world who work on it. +Vectorisation support is even more challenging, and is a fast-moving +research topic. Fortunately there is convergent research in this area, +however with this processor's Vectorisation being literally unique, +and also in active development (requiring an iterative process), this is +going to be a huge challenge. Luckily, there is low-hanging fruit that +will allow significant performance increases for relatively little compiler +effort. + +Keeping the work upstream is made difficult because there is not yet any +active silicon. Part of the tasks will therefore be to ensure that the +code is kept up-to-date until such time as active silicon is available. + ## Describe the ecosystem of the project, and how you will engage with relevant actors and promote the outcomes? As mentioned in the 2018 submission, the Libre RISC-V -- 2.30.2