Math functions for Swift #23140

stephentyrone · 2019-03-06T23:02:51Z

This PR implements SE-0246.

We still need to make a decision about how to handle obsoleting functions provided by the platform module. Because most programs pull in Foundation, they transitively pull in the existing names, which means that most functions calls in concrete contexts will get the tgmath functions, not the stdlib implementations. That's OK, but there are a few functions for which the behavior differs for edge cases (most notably pow), and we'd also like to remove these to cut down the overload sets.

rdar://problem/17625344

stephentyrone · 2019-03-06T23:03:01Z

@swift-ci please test

swift-ci · 2019-03-06T23:39:25Z

Build failed
Swift Test OS X Platform
Git Sha - f17dffb9b557546d631a6f5ec81e9765152c7d5d

swift-ci · 2019-03-06T23:40:22Z

Build failed
Swift Test Linux Platform
Git Sha - f17dffb9b557546d631a6f5ec81e9765152c7d5d

stephentyrone · 2019-03-15T12:13:05Z

@swift-ci please test

swift-ci · 2019-03-15T12:15:02Z

Build failed
Swift Test OS X Platform
Git Sha - f17dffb9b557546d631a6f5ec81e9765152c7d5d

swift-ci · 2019-03-15T12:15:12Z

Build failed
Swift Test Linux Platform
Git Sha - f17dffb9b557546d631a6f5ec81e9765152c7d5d

tkremenek · 2019-04-02T15:58:59Z

@swift-ci clean test

tkremenek · 2019-04-02T15:59:16Z

@stephentyrone @airspeedswift what is preventing this from moving forward?

swift-ci · 2019-04-02T16:35:09Z

Build failed
Swift Test Linux Platform
Git Sha - 7cb09f768aa2acbc890364c21abe60ae44d3b9b6

stephentyrone · 2019-04-02T17:24:11Z

@tkremenek Just shaking out fallout from dropping the Math module, specifically what that means for the existing Platform functions. I expect to push this forward today.

More precisely: when we had the Math module, it was possible for Platform to reexport it and preserve source compatibility while obsoleting the existing functions. That isn't possible any longer unless we have it reexport all of the stdlib, which seems like a bad idea. So I'm working on making the minimally invasive change to preserve the behavior of the platform module as much as possible while obsoleting a few specific functions as necessary to resolve ambiguities. This has turned out to be a bit fussier than I originally anticipated.

swift-ci · 2019-04-02T17:33:55Z

Build failed
Swift Test OS X Platform
Git Sha - 7cb09f768aa2acbc890364c21abe60ae44d3b9b6

stephentyrone · 2019-04-02T19:19:26Z

@swift-ci please test

swift-ci · 2019-04-02T19:21:39Z

Build failed
Swift Test Linux Platform
Git Sha - 7cb09f768aa2acbc890364c21abe60ae44d3b9b6

swift-ci · 2019-04-02T19:23:25Z

Build failed
Swift Test OS X Platform
Git Sha - 7cb09f768aa2acbc890364c21abe60ae44d3b9b6

stephentyrone · 2019-04-02T21:47:00Z

@swift-ci please test

swift-ci · 2019-04-02T21:49:08Z

Build failed
Swift Test Linux Platform
Git Sha - d54908b1b7cb6ba7b61cc209c8537530a0044fcb

swift-ci · 2019-04-02T22:08:03Z

Build failed
Swift Test OS X Platform
Git Sha - d54908b1b7cb6ba7b61cc209c8537530a0044fcb

Fundamental decision to be made; should implementation hooks present like `Float.exp(_ x: Float) -> Float` or like `Float.Maths.exp(_ x: Float) -> Float`? Having the intermediate namespace to group them as in the second is definitely nicer, but it requires a little bit of extra machinery, and much more importantly, there doesn't seem to be any way to make access to the static `Maths` associatedtype transparent.

This reverts commit d9dde28.

This commit implements SE-0246, by adding conformance to Real to the Float, CGFloat, Double, and Float80 types, implemented either in terms of the system's C math library, existing standard library functionality, or LLVM intrinsics. It includes basic test coverage for these new functions, and deprecates and obsoletes *some* existing functionality in the Platform overlay. We still need to make a decision about how to handle the remaining "tgmath" functions, because obsoleting them is technically a source-breaking change (if users have unqualified names like "exp(1)", it's fine, but it would break users who have used qualified names like "Darwin.exp(1)".)

swift-ci · 2019-04-02T22:53:03Z

Build failed
Swift Test Linux Platform
Git Sha - 44a5821b678e7c06a8bd367e4f8db95ed8f3bdb9

swift-ci · 2019-04-02T23:14:49Z

Build failed
Swift Test OS X Platform
Git Sha - 44a5821b678e7c06a8bd367e4f8db95ed8f3bdb9

stephentyrone · 2019-04-03T01:16:43Z

@swift-ci clean test

swift-ci · 2019-04-03T02:03:15Z

Build failed
Swift Test Linux Platform
Git Sha - 10fd131

swift-ci · 2019-04-03T03:07:58Z

Build failed
Swift Test OS X Platform
Git Sha - 10fd131

stephentyrone · 2019-04-03T12:17:40Z

@swift-ci please smoke test

stephentyrone · 2019-04-03T16:38:33Z

@swift-ci please test.

swift-ci · 2019-04-03T16:40:42Z

Build failed
Swift Test Linux Platform
Git Sha - 10fd131

swift-ci · 2019-04-03T16:42:41Z

Build failed
Swift Test OS X Platform
Git Sha - 10fd131

stephentyrone · 2019-04-03T19:16:32Z

@swift-ci please test.

swift-ci · 2019-04-03T19:17:10Z

Build failed
Swift Test OS X Platform
Git Sha - 45a76fe

swift-ci · 2019-04-03T19:18:42Z

Build failed
Swift Test Linux Platform
Git Sha - 45a76fe

compnerd · 2019-04-03T22:39:49Z

This broke the windows build! https://dev.azure.com/compnerd/windows-swift/_build/results?buildId=976

stephentyrone · 2019-04-03T22:44:21Z

@compnerd windows doesn't have hypot either?!

compnerd · 2019-04-03T22:46:45Z

@stephentyrone - clearly MSVCRT hates math :-(

Swift.obj : error LNK2019: unresolved external symbol hypotf referenced in function $sSf5hypotyS2f_SftFZ

I think it may be inlined or something, since they claim it is in the library:
https://docs.microsoft.com/en-us/cpp/c-runtime-library/reference/hypot-hypotf-hypotl-hypot-hypotf-hypotl?view=vs-2019

stephentyrone · 2019-04-03T22:49:24Z

@compnerd Can you replace _swift_stdlib_hypotf with a bogus stub on a branch and see if anything else breaks?

You might also be able to call _hypotf instead. Worst-case scenario, I can open-code an implementation for Windows, but let's see if we can use the MSVCRT first.

compnerd · 2019-04-03T23:09:37Z

Yeah, guess we need to switch it over to _hypotf:

    _Check_return_ __inline float __CRTDECL hypotf(_In_ float _X, _In_ float _Y)
    {
        return _hypotf(_X, _Y);
    }

I'll put up a patch momentarily.

xwu · 2019-04-04T02:54:43Z

stdlib/public/core/MathFunctions.swift.gyb

+/// ```
+///
+/// All of these are made available as free functions by importing the Math
+/// module:


This is outdated, I thought?

Yeah, I still need to update some comments.

This reverts commit 777750d, reversing changes made to 0c8920e.

* Revert "Merge pull request #23791 from compnerd/you-know-nothing-clang" This reverts commit 5150981, reversing changes made to 8fc305c. * Revert "Merge pull request #23780 from compnerd/math-is-terrible" This reverts commit 2d7fedd, reversing changes made to 0205150. * Revert "Merge pull request #23140 from stephentyrone/mafs" This reverts commit 777750d, reversing changes made to 0c8920e.

Done here swiftlang/swift#23140

xwu changed the title ~~[DNM] Elemenatary functions for Swift (aka Mathable)~~ [DNM] Elementary functions for Swift (aka Mathable) Mar 8, 2019

stephentyrone force-pushed the mafs branch from f17dffb to 5f6affe Compare March 15, 2019 12:11

stephentyrone force-pushed the mafs branch from 5f6affe to 7cb09f7 Compare March 21, 2019 17:24

stephentyrone force-pushed the mafs branch from 7cb09f7 to d54908b Compare April 2, 2019 19:19

stephentyrone marked this pull request as ready for review April 2, 2019 19:19

stephentyrone requested a review from airspeedswift April 2, 2019 19:19

stephentyrone changed the title ~~[DNM] Elementary functions for Swift (aka Mathable)~~ Math functions for Swift Apr 2, 2019

stephentyrone added 6 commits April 2, 2019 18:48

WOW\! 100% less gratuitous type machinery at runtime\!

b8d1481

First non-intrinsic functions.

2f7435d

Revert "build: improve the builds with Visual Studio"

cad116a

This reverts commit d9dde28.

Test fixes.

10fd131

stephentyrone added 2 commits April 3, 2019 08:10

update stability-stdlib-abi.swift.expected to match new baseline

42abd7f

Update tgmath.swift.gyb to work with Glibc and MSVCRT.

0c7c1f0

stephentyrone added 2 commits April 3, 2019 09:32

More test fixup for Linux.

2bc6040

fixup IRGen test for exp on linux

45a76fe

Obscene rant about linters.

bcc7e2c

tkremenek merged commit 777750d into swiftlang:master Apr 3, 2019

stephentyrone deleted the mafs branch April 3, 2019 21:54

xwu reviewed Apr 4, 2019

View reviewed changes

This was referenced Apr 4, 2019

Revert "Math functions for Swift" #23789

Closed

[SE-0252][TypeChecker] Keypath Dynamic Member Lookup #23436

Merged

stephentyrone added a commit to stephentyrone/swift that referenced this pull request Apr 4, 2019

Revert "Merge pull request swiftlang#23140 from stephentyrone/mafs"

d69391b

This reverts commit 777750d, reversing changes made to 0c8920e.

mackoj added a commit to mackoj/swift-evolution that referenced this pull request Jan 10, 2020

Update 0246-mathable.md

0b6072c

Done here swiftlang/swift#23140

Math functions for Swift #23140

Math functions for Swift #23140

Conversation

stephentyrone commented Mar 6, 2019 • edited Loading

stephentyrone commented Mar 6, 2019

swift-ci commented Mar 6, 2019

swift-ci commented Mar 6, 2019

stephentyrone commented Mar 15, 2019

swift-ci commented Mar 15, 2019

swift-ci commented Mar 15, 2019

tkremenek commented Apr 2, 2019

tkremenek commented Apr 2, 2019

swift-ci commented Apr 2, 2019

stephentyrone commented Apr 2, 2019 • edited Loading

swift-ci commented Apr 2, 2019

stephentyrone commented Apr 2, 2019

swift-ci commented Apr 2, 2019

swift-ci commented Apr 2, 2019

stephentyrone commented Apr 2, 2019

swift-ci commented Apr 2, 2019

swift-ci commented Apr 2, 2019

swift-ci commented Apr 2, 2019

swift-ci commented Apr 2, 2019

stephentyrone commented Apr 3, 2019

swift-ci commented Apr 3, 2019

swift-ci commented Apr 3, 2019

stephentyrone commented Apr 3, 2019

stephentyrone commented Apr 3, 2019

swift-ci commented Apr 3, 2019

swift-ci commented Apr 3, 2019

stephentyrone commented Apr 3, 2019

swift-ci commented Apr 3, 2019

swift-ci commented Apr 3, 2019

compnerd commented Apr 3, 2019

stephentyrone commented Apr 3, 2019

compnerd commented Apr 3, 2019

stephentyrone commented Apr 3, 2019 • edited Loading

compnerd commented Apr 3, 2019

xwu Apr 4, 2019

Choose a reason for hiding this comment

stephentyrone Apr 4, 2019

Choose a reason for hiding this comment

stephentyrone commented Mar 6, 2019 •

edited

Loading

stephentyrone commented Apr 2, 2019 •

edited

Loading

stephentyrone commented Apr 3, 2019 •

edited

Loading