[PATCH] D21810: Don't instantiate a full host toolchain in ASTMatchersTest.
Justin Lebar via cfe-commits
cfe-commits at lists.llvm.org
Tue Jun 28 12:00:04 PDT 2016
jlebar created this revision.
jlebar added a reviewer: chandlerc.
jlebar added a subscriber: cfe-commits.
Herald added a subscriber: klimek.
This test was stat()'ing large swaths of /usr/lib hundreds of times, as
every invocation of matchesConditionally*() created a new Linux
toolchain.
In addition to being slow, perf indicated this was causing substantial
contention in the kernel.
Something is...interesting in the kernel, as without this patch I
sometimes see ~11m spent in the kernel, and sometimes ~5m. This
corresponds to bimodal ninja check-clang times of ~30s and ~20s.
It's not clear to me exactly what causes the bimodality. In any case,
this change makes this test run in 2.5s, down from 17s, and it seems to
cause us to get the 20s ninja check-clang time unconditionally.
http://reviews.llvm.org/D21810
Files:
unittests/ASTMatchers/ASTMatchersTest.h
Index: unittests/ASTMatchers/ASTMatchersTest.h
===================================================================
--- unittests/ASTMatchers/ASTMatchersTest.h
+++ unittests/ASTMatchers/ASTMatchersTest.h
@@ -73,12 +73,16 @@
return testing::AssertionFailure() << "Could not add dynamic matcher";
std::unique_ptr<FrontendActionFactory> Factory(
newFrontendActionFactory(&Finder));
- // Some tests use typeof, which is a gnu extension.
- std::vector<std::string> Args;
- Args.push_back(CompileArg);
- // Some tests need rtti/exceptions on
- Args.push_back("-frtti");
- Args.push_back("-fexceptions");
+ // Some tests need rtti/exceptions on. Use an unknown-unknown triple so we
+ // don't instantiate the full system toolchain. On Linux, instantiting the
+ // toolchain involves stat'ing large portions of /usr/lib, and this slows down
+ // not only this test, but all other tests, via contention in the kernel.
+ //
+ // FIXME: This is a hack to work around the fact that there's no way to do the
+ // equivalent of runToolOnCodeWithArgs without instantiating a full Driver.
+ // We should consider having a function, at least for tests, that invokes cc1.
+ std::vector<std::string> Args = {CompileArg, "-frtti", "-fexceptions",
+ "-target", "i386-unknown-unknown"};
if (!runToolOnCodeWithArgs(
Factory->create(), Code, Args, Filename, "clang-tool",
std::make_shared<PCHContainerOperations>(), VirtualMappedFiles)) {
@@ -180,13 +184,12 @@
return testing::AssertionFailure() << "Could not add dynamic matcher";
std::unique_ptr<FrontendActionFactory> Factory(
newFrontendActionFactory(&Finder));
- // Some tests use typeof, which is a gnu extension.
- std::vector<std::string> Args;
- Args.push_back("-xcuda");
- Args.push_back("-fno-ms-extensions");
- Args.push_back("--cuda-host-only");
- Args.push_back("-nocudainc");
- Args.push_back(CompileArg);
+ // Some tests use typeof, which is a gnu extension. Using an explicit
+ // unknown-unknown triple is good for a large speedup, because it lets us
+ // avoid constructing a full system triple.
+ std::vector<std::string> Args = {
+ "-xcuda", "-fno-ms-extensions", "--cuda-host-only", "-nocudainc",
+ "-target", "nvptx64-unknown-unknown", CompileArg};
if (!runToolOnCodeWithArgs(Factory->create(),
CudaHeader + Code, Args)) {
return testing::AssertionFailure() << "Parsing error in \"" << Code << "\"";
@@ -230,8 +233,11 @@
Finder.addMatcher(AMatcher, &VerifyVerifiedResult);
std::unique_ptr<FrontendActionFactory> Factory(
newFrontendActionFactory(&Finder));
- // Some tests use typeof, which is a gnu extension.
- std::vector<std::string> Args(1, "-std=gnu++98");
+ // Some tests use typeof, which is a gnu extension. Using an explicit
+ // unknown-unknown triple is good for a large speedup, because it lets us
+ // avoid constructing a full system triple.
+ std::vector<std::string> Args = {"-std=gnu++98", "-target",
+ "i386-unknown-unknown"};
if (!runToolOnCodeWithArgs(Factory->create(), Code, Args)) {
return testing::AssertionFailure() << "Parsing error in \"" << Code << "\"";
}
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D21810.62123.patch
Type: text/x-patch
Size: 3271 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20160628/0833b02e/attachment.bin>
More information about the cfe-commits
mailing list