Migrate processor/feature detection from OpenJ9 to OMR Part 1 #4503

AidanHa · 2019-10-23T15:18:40Z

OpenJ9 port library has processor and feature detections that can be sunk down to OMR. This can be divided into 4 steps:
1: Creation of processor and feature detection in OMR (from OpenJ9).
2: Implementation of the new feature throughout out OMR.
3: Implementation of the new feature throughout OpenJ9.
4: Deletion of the processor detection in OpenJ9 port library.

This PR serves to implement step 1 of the list above. This PR includes the following:

Move and rename necessary port library macros from OpenJ9 to OMR's omrport.h (Commit 1)
Move and implement processor/feature detection for windows systems[WIP] (Commit 2)
Move and implement processor/feature detection for unix systems (x86, Power, AIX, Z) (Commit 3)
Move and test the respective port tests that helps verify that processor detection works (Commit 4)

Issue: #4339

AidanHa · 2019-10-29T14:25:30Z

These functions have been tested on Z, Power and X86 for Windows and Unix systems! I think we're ready for a full review!

Leonardo2718 · 2019-10-29T16:00:34Z

@genie-omr build all

AidanHa · 2019-11-01T13:36:16Z

@Leonardo2718 Could you run the PR builds again? There were some build issues on LinuxPPC and AIX, and I've pushed changes to fix these issues.

AidanHa · 2019-11-01T13:36:47Z

Also FYI, I've broken down the PR into several commits for clarity, but will merge them when the changes are approved!

rwy0717 · 2019-11-01T14:42:40Z

@genie-omr build all

rwy0717 · 2019-11-01T16:19:27Z

Aidan, for the new tests that you're adding, you should just use core gtest functionality, rather than using the port test helpers. Specifically, don't call reportTestEntry (or exit), and use ASSERT/EXPECT, rather than calling outputErrorMessage. As well, your tests should be in TEST blocks (I'm looking at testGetProcessorDescription).

rwy0717 · 2019-11-01T16:36:02Z

This looks like a pretty straightforward copy/paste/rename job, so if you want to do clean up in a later PR, I would be fine with that, it's up to you.

rwy0717 · 2019-11-01T20:00:48Z

The AIX failure is interesting, I don't think it's related to your change. EDIT: actually, it seems it is your problem.

@genie-omr build aix

The x86-64 build, however...

[2019-11-01T14:46:05.106Z] si.cpp:132:11: error: variable ‘rc’ set but not used [-Werror=unused-but-set-variable]
[2019-11-01T14:46:05.106Z]   uint32_t rc = 0;

The rc in question.

port/common/omrsysinfo.c

port/common/omrsysinfo_helpers.c

port/win32/omrsysinfo.c

port/common/omrsysinfo.c

port/zos390/omrportpg.h

rwy0717 · 2019-11-04T14:54:14Z

@genie-omr build aix

AidanHa · 2019-11-04T15:51:57Z

Aidan, for the new tests that you're adding, you should just use core gtest functionality, rather than using the port test helpers. Specifically, don't call reportTestEntry (or exit), and use ASSERT/EXPECT, rather than calling outputErrorMessage. As well, your tests should be in TEST blocks (I'm looking at testGetProcessorDescription).

I removed testGetProcessorDescription and added the tests in a TEST block. Any reason why we want to be using ASSERTS rather than outputErrorMessage? When writing the tests, I was following the format of the other tests on si.cpp

rwy0717 · 2019-11-04T16:25:18Z

Wherever possible, the port test helpers are being replaced with gtest. I opened an issue #4530. Any new tests should just be using gtest.

port/zos390/omrportpg.h

rwy0717 · 2019-11-04T16:36:20Z

port/unix/omrsysinfo.c

+		setFeature(desc, OMRPORT_PPC_FEATURE_POWER4);
+	}
+#endif /* !defined(J9OS_I5_V6R1) */
+#if !defined(J9OS_I5_V7R2) && !defined(J9OS_I5_V6R1)


These are never defined in OMR, and the code below fails to compile on aix 6.1.

port/common/omrsysinfo.c

fvtest/porttest/si.cpp

port/unix/omrsysinfo.c

fvtest/porttest/si.cpp

port/common/omrsysinfo_helpers.c

port/unix/omrsysinfo.c

port/common/omrsysinfo_helpers.h

AidanHa · 2019-11-11T12:54:56Z

Changes are fixed! @rwy0717 When you are done reviewing, I'll merge all of my changes into 1 single commit.

rwy0717 · 2019-11-22T17:13:01Z

@genie-omr build all

rwy0717 · 2019-11-28T19:12:58Z

@genie-omr build all

fjeremic · 2019-12-03T14:52:03Z

Let's make sure to squash the review addressing commits here before merging (after committer assigned has approved).

rwy0717 · 2019-12-07T00:15:46Z

@genie-omr build all

rwy0717 · 2019-12-07T01:50:39Z

@fjeremic when you say "Let's make sure to squash", are you asking @AidanHa to squash, or are you asking me to use the "squash and merge" button? Either way, I think this PR is ready to go.


        Add OMR macros for processor feature detection

Move required macros, enums and structs from OpenJ9 to OMR's port library Add function signatures for processor/feature detection to OMR Implement processor detection on Windows Add processor and feature detection for Unix Includes Linux, AIX PPC and ZOS systems Add test to verify processor/feature detection. Signed-off-by: Aidan Ha <qbha@edu.uwaterloo.ca>

AidanHa · 2019-12-07T02:51:47Z

I've gone ahead and squashed my commits!

rwy0717 · 2019-12-07T03:28:20Z

@genie-omr build all

AidanHa requested review from charliegracie, mstoodle, rwy0717 and youngar as code owners Oct 23, 2019

Leonardo2718 added the comp:port label Oct 23, 2019

AidanHa force-pushed the AidanHa:Issue4339 branch from d98afc1 to dc3700a Oct 29, 2019

AidanHa force-pushed the AidanHa:Issue4339 branch 2 times, most recently from 978156e to ffa758b Nov 1, 2019