Hi Tanya, The new type-legalization mode (-promote-elements) which enables vector-select in LLVM (and a nice perf boost for several workloads), is currently disabled because of a _single_ bug in the ARM codegen which makes a few tests fail. If ARM is not a supported target, can I mark these tests as 'XFAIL' and enable vector-select support in LLVM ? Thanks, Nadav -----Original Message----- From: llvmdev-bounces at cs.uiuc.edu [mailto:llvmdev-bounces at cs.uiuc.edu] On Behalf Of Tanya Lattner Sent: Saturday, October 08, 2011 01:10 To: Seb Cc: llvmdev at cs.uiuc.edu Subject: Re: [LLVMdev] LLC ARM Backend maintainer On Oct 7, 2011, at 1:07 AM, Seb wrote:> Hi all, > > To answer Eli question, I wanted to know who is actively working on ARM because I submitted some bug report (#11029, #9905) and don't know if someone is working on them, if/when the will be fixed. Maybe I just need to better understand LLVM release process, I've seen a mail in this list about it.Bugs get fixed if there are people to fix them. There are numerous people working on ARM and patches are also accepted. ARM is not currently a target that we support for releases. So those bugs are not release blockers. -Tanya> > -- Seb > _______________________________________________ > 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 --------------------------------------------------------------------- Intel Israel (74) Limited This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.
Hi Nadav,> The new type-legalization mode (-promote-elements) which enables vector-select in LLVM (and a nice perf boost for several workloads), is currently disabled because of a _single_ bug in the ARM codegen which makes a few tests fail. If ARM is not a supported target, can I mark these tests as 'XFAIL' and enable vector-select support in LLVM ?Which testcase is it? -- With best regards, Anton Korobeynikov Faculty of Mathematics and Mechanics, Saint Petersburg State University
Anton, "CodeGen/ARM/vrev.ll" is one of the failing tests. You can make it fail if you run it with '-promote-elements'. Here is a short description from PR10902: """ The failure in CodeGen/ARM/vrev.ll is due to the lack of support for legalizing trunc-store/load on ARM. The code we have in LegalizeDAG assumes that the scalars making the saved vector are legal. So, the vector <2 x i16> is scalarized to two i16, which are illegal on ARM. """ Thanks, Nadav -----Original Message----- From: Anton Korobeynikov [mailto:anton at korobeynikov.info] Sent: Saturday, October 08, 2011 19:46 To: Rotem, Nadav Cc: Tanya Lattner; llvmdev at cs.uiuc.edu Subject: Re: [LLVMdev] LLC ARM Backend maintainer Hi Nadav,> The new type-legalization mode (-promote-elements) which enables vector-select in LLVM (and a nice perf boost for several workloads), is currently disabled because of a _single_ bug in the ARM codegen which makes a few tests fail. If ARM is not a supported target, can I mark these tests as 'XFAIL' and enable vector-select support in LLVM ?Which testcase is it? -- With best regards, Anton Korobeynikov Faculty of Mathematics and Mechanics, Saint Petersburg State University --------------------------------------------------------------------- Intel Israel (74) Limited This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.
No. Note the qualifying phrase "for releases" on Tanya's statement. If, during release testing, a regression is found on ARM compared to 2.9 results, it is not required by process to be considered a release blocker. That does not mean features can or should be enabled which knowingly break ARM. That's an entirely different situation. -Jim On Oct 8, 2011, at 9:59 AM, Rotem, Nadav wrote:> Hi Tanya, > > The new type-legalization mode (-promote-elements) which enables vector-select in LLVM (and a nice perf boost for several workloads), is currently disabled because of a _single_ bug in the ARM codegen which makes a few tests fail. If ARM is not a supported target, can I mark these tests as 'XFAIL' and enable vector-select support in LLVM ? > > Thanks, > Nadav > > > -----Original Message----- > From: llvmdev-bounces at cs.uiuc.edu [mailto:llvmdev-bounces at cs.uiuc.edu] On Behalf Of Tanya Lattner > Sent: Saturday, October 08, 2011 01:10 > To: Seb > Cc: llvmdev at cs.uiuc.edu > Subject: Re: [LLVMdev] LLC ARM Backend maintainer > > > On Oct 7, 2011, at 1:07 AM, Seb wrote: > >> Hi all, >> >> To answer Eli question, I wanted to know who is actively working on ARM because I submitted some bug report (#11029, #9905) and don't know if someone is working on them, if/when the will be fixed. Maybe I just need to better understand LLVM release process, I've seen a mail in this list about it. > > Bugs get fixed if there are people to fix them. There are numerous people working on ARM and patches are also accepted. > > ARM is not currently a target that we support for releases. So those bugs are not release blockers. > > -Tanya > >> >> -- Seb >> _______________________________________________ >> 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 > --------------------------------------------------------------------- > Intel Israel (74) Limited > > This e-mail and any attachments may contain confidential material for > the sole use of the intended recipient(s). Any review or distribution > by others is strictly prohibited. If you are not the intended > recipient, please contact the sender and delete all copies. > > > _______________________________________________ > LLVM Developers mailing list > LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu > http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
Exactly as Jim said :) I'm only talking about releases and release blockers. Its also a bit of a grey area when you are talking about just ARM codgen tests in the regression test suite because those should always be passing regardless of what target/os you are testing on (assuming you built the arm backend which the release team typically does). As it stands now we don't have ARM as a supported target just because there is no qualification plan for it (ie. no one running the test suite or some other benchmarks on an ARM device for example) and no real support to qualify it. If someone wants to initiate this process, it can be talked about post-3.0. -Tanya On Oct 10, 2011, at 8:47 AM, Jim Grosbach wrote:> No. Note the qualifying phrase "for releases" on Tanya's statement. If, during release testing, a regression is found on ARM compared to 2.9 results, it is not required by process to be considered a release blocker. That does not mean features can or should be enabled which knowingly break ARM. That's an entirely different situation. > > -Jim > > > On Oct 8, 2011, at 9:59 AM, Rotem, Nadav wrote: > >> Hi Tanya, >> >> The new type-legalization mode (-promote-elements) which enables vector-select in LLVM (and a nice perf boost for several workloads), is currently disabled because of a _single_ bug in the ARM codegen which makes a few tests fail. If ARM is not a supported target, can I mark these tests as 'XFAIL' and enable vector-select support in LLVM ? >> >> Thanks, >> Nadav >> >> >> -----Original Message----- >> From: llvmdev-bounces at cs.uiuc.edu [mailto:llvmdev-bounces at cs.uiuc.edu] On Behalf Of Tanya Lattner >> Sent: Saturday, October 08, 2011 01:10 >> To: Seb >> Cc: llvmdev at cs.uiuc.edu >> Subject: Re: [LLVMdev] LLC ARM Backend maintainer >> >> >> On Oct 7, 2011, at 1:07 AM, Seb wrote: >> >>> Hi all, >>> >>> To answer Eli question, I wanted to know who is actively working on ARM because I submitted some bug report (#11029, #9905) and don't know if someone is working on them, if/when the will be fixed. Maybe I just need to better understand LLVM release process, I've seen a mail in this list about it. >> >> Bugs get fixed if there are people to fix them. There are numerous people working on ARM and patches are also accepted. >> >> ARM is not currently a target that we support for releases. So those bugs are not release blockers. >> >> -Tanya >> >>> >>> -- Seb >>> _______________________________________________ >>> 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 >> --------------------------------------------------------------------- >> Intel Israel (74) Limited >> >> This e-mail and any attachments may contain confidential material for >> the sole use of the intended recipient(s). Any review or distribution >> by others is strictly prohibited. If you are not the intended >> recipient, please contact the sender and delete all copies. >> >> >> _______________________________________________ >> LLVM Developers mailing list >> LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu >> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev >