Support Linux on z as a Swift platform #2541

bryanpkc · 2016-05-16T05:10:09Z

What's in this pull request?

This pull request adds support for Linux on IBM z Systems to Swift. It depends on pull requests in the other repos:

apple/swift-clang#13
apple/swift-llvm#8
apple/swift-lldb#27

The first commit in this pull request adds a new platform condition check named endian(), for which there are two possible values (little and big), to allow conditional compilation of endianness-specific Swift code. The second commit constitutes the bulk of infrastructure changes, including new test cases, to add support for the new target. The remaining commits deal with endianness issues in various parts of the compiler and Swift runtime. In particular, handling of enums is not yet completely fixed for big-endian systems, and we are researching a full solution (to be contributed in a future pull request). Nevertheless, the port gives us a usable compiler and REPL. Validation test results show a 0.2% failure rate:

  Expected Passes    : 7308
  Expected Failures  : 71
  Unsupported Tests  : 797
  Unexpected Failures: 19

No regression is observed on x86_64.

Before merging this pull request to apple/swift repository:

Test pull request on Swift continuous integration.

Triggering Swift CI

The swift-ci is triggered by writing a comment on this PR addressed to the GitHub user @swift-ci. Different tests will run depending on the specific comment that you use. The currently available comments are:

Smoke Testing

Platform	Comment
All supported platforms	@swift-ci Please smoke test
OS X platform	@swift-ci Please smoke test OS X platform
Linux platform	@swift-ci Please smoke test Linux platform

Validation Testing

Platform	Comment
All supported platforms	@swift-ci Please test
OS X platform	@swift-ci Please test OS X platform
Linux platform	@swift-ci Please test Linux platform

Note: Only members of the Apple organization can trigger swift-ci.

gribozavr · 2016-05-16T05:51:24Z

The first commit in this pull request adds a new platform condition check named endian(), for which there are two possible values (little and big), to allow conditional compilation of endianness-specific Swift code.

This change needs to go through the Swift evolution process. We can unblock the port by adding it as _endian for now.

gribozavr · 2016-05-16T05:56:21Z

stdlib/private/SwiftPrivateLibcExtras/SwiftPrivateLibcExtras.swift

@@ -81,6 +81,12 @@ public struct _stdlib_fd_set {
  }
 }

+#if arch(arm) || arch(i386)
+public typealias _stdlib_fd_set = __stdlib_fd_set<UInt32>
+#else


Please list 64-bit platforms explicitly. This ensures that this code breaks when we have a new port, and it is added to the appropriate list.

gribozavr · 2016-05-16T06:16:31Z

CC @rjmccall @slavapestov for IRGen and runtime changes.

The library parts and test changes LGTM.

gribozavr · 2016-05-16T06:17:14Z

@bryanpkc This is fantastic! I'm just curious -- what are the failing tests?

gribozavr · 2016-05-16T06:17:41Z

Let's run CI to verify that there are no regressions on existing platforms.

@swift-ci Please test

gribozavr · 2016-05-16T06:57:52Z

@bryanpkc Looks like there are test failures on OS X, please take a look.

bryanpkc · 2016-05-16T15:17:21Z

@gribozavr Thanks for your review. I have made changes according to your suggestions and fixed the test failure on OS X.

gribozavr · 2016-05-16T15:50:14Z

@swift-ci Please test

gribozavr · 2016-05-16T17:36:33Z

@bryanpkc OS X failed again.

lattner · 2016-05-17T04:49:00Z

@gribozavr thank you for the extensive review, I really appreciate it. @bryanpkc please do submit a proposal for adding the endian #if check. I expect it to be mostly a bike shedding exercise, but it is an important one to do and get past.

rintaro · 2016-05-17T05:00:54Z

@bryanpkc
FYI, http://thread.gmane.org/gmane.comp.lang.swift.evolution/12267

bryanpkc · 2016-05-17T13:40:33Z

@gribozavr I think I have fixed the OS X test case for real this time (sorry I cannot run that test myself). I have also simplified the StringUTF8 fix as you suggested. It also helps us pass a few more test cases.

@lattner, @rintaro: Thanks for the link to the draft proposal. It seems that the bitwidth() part of the proposal is more complicated and warrants more discussion, but I can submit the endian() part of it, if @sadun doesn't mind.

rintaro · 2016-05-17T14:04:45Z

stdlib/public/core/StringUTF8.swift

@@ -48,8 +48,10 @@ extension _StringCore {
        src: UnsafeMutablePointer(startASCII + i),
        size: numericCast(utf16Count))

-      return (i + utf16Count, result)
+      // Convert the _UTF8Chunk into host endianness.
+      return (i + utf16Count, _UTF8Chunk(littleEndian: result))


I think, we have another endianness problem in _StringCore._nthContiguous(_:).
Sorry, I cannot test myself.

@rintaro You are right about _nthContiguous() being little-endian only but that is only part of a larger problem. There does not seem to be any support for UTF-16BE in Swift, and supporting it may require considerable changes to the _StringCore code. I am pretty sure that conversion from UTF-8 to UTF-16 is currently broken in this port, but we hope to tackle that in a future pull request (soonish).

gribozavr · 2016-05-17T16:10:19Z

@bryanpkc Thank you!

@swift-ci Please test

gribozavr · 2016-05-19T05:36:18Z

@swift-ci Please test OS X platform

gribozavr · 2016-05-20T01:47:59Z

@swift-ci Please test OS X platform

gribozavr · 2016-05-20T05:21:18Z

@swift-ci Please test

gribozavr · 2016-05-20T07:46:07Z

The CI failed for unrelated reasons.

Ping @rjmccall @slavapestov for IRGen and runtime changes.

… and "big".

…is used to ensure correct operations of C++ code in the runtime. This patch also includes some (incomplete) changes to enum handling to make enums work in most common cases.

…hunk is in host endianness, for correct operations on both little- and big-endian systems.

…and must be loaded with a 16-bit load instruction for correctness on big-endian systems.

…ks correctly for 64-bit big-endian systems.

…phics module.

bryanpkc · 2016-05-25T11:31:31Z

I have rebased the commits on to the tip of master, and added a couple more fixes for test failures.

gribozavr · 2016-05-25T19:22:55Z

@slavapestov has reviewed the changes.

gribozavr · 2016-05-25T19:23:03Z

@swift-ci Please test and merge

bryanpkc · 2016-05-25T22:59:45Z

@gribozavr Thanks for testing. The TestNSJSONSerialization failure does not seem related to my commits.

gribozavr · 2016-05-25T23:06:07Z

Yes, it isn't.

bryanpkc · 2016-05-26T14:11:13Z

Thanks for merging.

Resolve conflicts with the `main` branch

gribozavr reviewed May 16, 2016
View reviewed changes

bryanpkc force-pushed the master-s390x branch from e85999e to 6b78d2c Compare May 16, 2016 15:00

modocache mentioned this pull request May 16, 2016

Add an option to specify the install prefix for CoreFoundation depend… swiftlang/swift-corelibs-xctest#106

Merged

bryanpkc force-pushed the master-s390x branch from 6b78d2c to 06daa6c Compare May 17, 2016 13:32

rintaro reviewed May 17, 2016
View reviewed changes

bryanpkc force-pushed the master-s390x branch from 11c2e5f to 6f19851 Compare May 20, 2016 14:06

bryanpkc and others added 11 commits May 24, 2016 19:39

Add new _endian() platform condition check; valid values are "little"…

bb3486d

… and "big".

Add support for Linux s390x. LLVM's Swift calling convention support …

85fde8b

…is used to ensure correct operations of C++ code in the runtime. This patch also includes some (incomplete) changes to enum handling to make enums work in most common cases.

When loading Unicode data into a 64-bit _UTF8Chunk, ensure that the c…

7127bff

…hunk is in host endianness, for correct operations on both little- and big-endian systems.

The instance alignment mask in the class metadata is a 16-bit field, …

9c62086

…and must be loaded with a 16-bit load instruction for correctness on big-endian systems.

Change the type of _stdlib_fd_set._data to [UInt] so that it also wor…

59e80c5

…ks correctly for 64-bit big-endian systems.

Fix test case failure on Linux s390x, which does not have the CoreGra…

3b39a03

…phics module.

Fixed Unimplemented error in UnsafePointer test case

b0a3727

Fix test/IRGen/condfail.sil and test/IRGen/c_layout.sil for s390x

e66f7fd

swift_allocError should use the Swift calling convention

71abe8b

Fix String.init test in FixedPoint.swift.gyb for big-endian systems.

9b881cd

Add support for s390x to UnsafeBitMap test

bcc43c0

bryanpkc force-pushed the master-s390x branch from 6f19851 to bcc43c0 Compare May 25, 2016 11:29

gribozavr merged commit 2daa140 into swiftlang:master May 25, 2016

garyliu1 mentioned this pull request Jun 24, 2016

Support Linux on z as a Swift platform: Part 2 #3193

Merged

1 task

garyliu1 mentioned this pull request Aug 23, 2016

Fix s390x Enum load/storeMultiPayloadValue #4461

Merged

1 task

MaxDesiatov pushed a commit that referenced this pull request Apr 19, 2021

Merge pull request #2541 from swiftwasm/maxd/main-merge

548c397

Resolve conflicts with the `main` branch

Support Linux on z as a Swift platform #2541

Support Linux on z as a Swift platform #2541

Uh oh!

Conversation

bryanpkc commented May 16, 2016 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What's in this pull request?

Uh oh!

gribozavr commented May 16, 2016

Uh oh!

gribozavr May 16, 2016

Choose a reason for hiding this comment

Uh oh!

gribozavr commented May 16, 2016

Uh oh!

gribozavr commented May 16, 2016

Uh oh!

gribozavr commented May 16, 2016

Uh oh!

gribozavr commented May 16, 2016

Uh oh!

bryanpkc commented May 16, 2016

Uh oh!

gribozavr commented May 16, 2016

Uh oh!

gribozavr commented May 16, 2016

Uh oh!

lattner commented May 17, 2016

Uh oh!

rintaro commented May 17, 2016

Uh oh!

bryanpkc commented May 17, 2016

Uh oh!

rintaro May 17, 2016 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

bryanpkc May 20, 2016

Choose a reason for hiding this comment

Uh oh!

gribozavr commented May 17, 2016

Uh oh!

gribozavr commented May 19, 2016

Uh oh!

gribozavr commented May 20, 2016

Uh oh!

gribozavr commented May 20, 2016

Uh oh!

gribozavr commented May 20, 2016

Uh oh!

bryanpkc commented May 25, 2016

Uh oh!

gribozavr commented May 25, 2016

Uh oh!

gribozavr commented May 25, 2016

Uh oh!

bryanpkc commented May 25, 2016

Uh oh!

gribozavr commented May 25, 2016

Uh oh!

bryanpkc commented May 26, 2016

Uh oh!

Uh oh!

bryanpkc commented May 16, 2016 •

edited

Loading

rintaro May 17, 2016 •

edited

Loading