]> granicus.if.org Git - clang/commit
[CUDA] Allow external variables in separate compilation
authorJonas Hahnfeld <hahnjo@hahnjo.de>
Wed, 14 Feb 2018 16:04:03 +0000 (16:04 +0000)
committerJonas Hahnfeld <hahnjo@hahnjo.de>
Wed, 14 Feb 2018 16:04:03 +0000 (16:04 +0000)
commit751ae2ab8b6f9d5f64524ebf9052b6fbbdfdbf37
tree117dcf421998a853ec6853a82b8469a1950222bf
parent1532324f1979660b1fc429eff544e1a9c1b79332
[CUDA] Allow external variables in separate compilation

According to the CUDA Programming Guide this is prohibited in
whole program compilation mode. This makes sense because external
references cannot be satisfied in that mode anyway. However,
such variables are allowed in separate compilation mode which
is a valid use case.

Differential Revision: https://reviews.llvm.org/D42923

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@325136 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Sema/SemaDeclAttr.cpp
test/SemaCUDA/extern-shared.cu