Mehdi Amini
2015-Aug-03 19:22 UTC
[LLVMdev] RFC: what is the purpose of test/CodeGen/Generic?
Hi all, I wonder about this directory and the 162 tests that are in it. Most tests have only this run line: ; RUN: llc < %s And when I look at the log of some tests like: test/CodeGen/Generic/2002-04-14-UnexpectedUnsignedType.ll The comment inside is: ; This caused the backend to assert out with: ; SparcInstrInfo.cpp:103: failed assertion `0 && "Unexpected unsigned type”’ I doubt we are effectively testing the Sparc backend unless you run the validation on a Sparc. Since “CodeGen” can’t really be generic (you’ll need a target), I’m not sure what is intended to be in it. I’d like to move most of the tests in their associated backend directory, but I’d like to hear what is legit to be here? Thanks, Mehdi
Some of the tests are used to test the command line options in llc and the behavior of the LLVMTargetMachine class, like stop-after.ll and run-pass.ll. I think that they are expected to work on every target and should stay here. Cheers, Alex 2015-08-03 12:22 GMT-07:00 Mehdi Amini <mehdi.amini at apple.com>:> Hi all, > > I wonder about this directory and the 162 tests that are in it. > > Most tests have only this run line: ; RUN: llc < %s > > And when I look at the log of some tests like: > test/CodeGen/Generic/2002-04-14-UnexpectedUnsignedType.ll > The comment inside is: > > ; This caused the backend to assert out with: > ; SparcInstrInfo.cpp:103: failed assertion `0 && "Unexpected unsigned > type”’ > > I doubt we are effectively testing the Sparc backend unless you run the > validation on a Sparc. > > Since “CodeGen” can’t really be generic (you’ll need a target), I’m not > sure what is intended to be in it. I’d like to move most of the tests in > their associated backend directory, but I’d like to hear what is legit to > be here? > > Thanks, > > Mehdi > > > > > > > _______________________________________________ > LLVM Developers mailing list > LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu > http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20150803/dc2bd4d0/attachment.html>
Eric Christopher
2015-Aug-03 19:59 UTC
[LLVMdev] RFC: what is the purpose of test/CodeGen/Generic?
Agreed :) -eric On Mon, Aug 3, 2015 at 12:51 PM Alex L <arphaman at gmail.com> wrote:> Some of the tests are used to test the command line options in llc and the > behavior of the LLVMTargetMachine class, like stop-after.ll and > run-pass.ll. I think that they are expected to work on every target and > should stay here. > > Cheers, > Alex > > 2015-08-03 12:22 GMT-07:00 Mehdi Amini <mehdi.amini at apple.com>: > >> Hi all, >> >> I wonder about this directory and the 162 tests that are in it. >> >> Most tests have only this run line: ; RUN: llc < %s >> >> And when I look at the log of some tests like: >> test/CodeGen/Generic/2002-04-14-UnexpectedUnsignedType.ll >> The comment inside is: >> >> ; This caused the backend to assert out with: >> ; SparcInstrInfo.cpp:103: failed assertion `0 && "Unexpected unsigned >> type”’ >> >> I doubt we are effectively testing the Sparc backend unless you run the >> validation on a Sparc. >> >> Since “CodeGen” can’t really be generic (you’ll need a target), I’m not >> sure what is intended to be in it. I’d like to move most of the tests in >> their associated backend directory, but I’d like to hear what is legit to >> be here? >> >> Thanks, >> >> Mehdi >> >> >> >> >> >> >> _______________________________________________ >> LLVM Developers mailing list >> LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu >> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev >> > > _______________________________________________ > LLVM Developers mailing list > LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu > http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20150803/93145552/attachment.html>
Mehdi Amini
2015-Aug-03 20:02 UTC
[LLVMdev] RFC: what is the purpose of test/CodeGen/Generic?
Ok for these, that is probably a small fraction of the 162 test in here :) In the meantime here are two patches to require “native” feature available for these tests (and others): http://reviews.llvm.org/D11730 <http://reviews.llvm.org/D11730> http://reviews.llvm.org/D11731 <http://reviews.llvm.org/D11731> (need reviewers) Thanks, — Mehdi> On Aug 3, 2015, at 12:48 PM, Alex L <arphaman at gmail.com> wrote: > > Some of the tests are used to test the command line options in llc and the behavior of the LLVMTargetMachine class, like stop-after.ll and run-pass.ll. I think that they are expected to work on every target and should stay here. > > Cheers, > Alex > > 2015-08-03 12:22 GMT-07:00 Mehdi Amini <mehdi.amini at apple.com <mailto:mehdi.amini at apple.com>>: > Hi all, > > I wonder about this directory and the 162 tests that are in it. > > Most tests have only this run line: ; RUN: llc < %s > > And when I look at the log of some tests like: test/CodeGen/Generic/2002-04-14-UnexpectedUnsignedType.ll > The comment inside is: > > ; This caused the backend to assert out with: > ; SparcInstrInfo.cpp:103: failed assertion `0 && "Unexpected unsigned type”’ > > I doubt we are effectively testing the Sparc backend unless you run the validation on a Sparc. > > Since “CodeGen” can’t really be generic (you’ll need a target), I’m not sure what is intended to be in it. I’d like to move most of the tests in their associated backend directory, but I’d like to hear what is legit to be here? > > Thanks, > > Mehdi > > > > > > > _______________________________________________ > LLVM Developers mailing list > LLVMdev at cs.uiuc.edu <mailto:LLVMdev at cs.uiuc.edu> http://llvm.cs.uiuc.edu <http://llvm.cs.uiuc.edu/> > http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev <http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev> >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20150803/f527ca43/attachment.html>