tree c057a9b7f941cf5bc2d6e2f8c0073c0cf5d5acc1
parent 547221f5dc50e62543733cd65bd71bd97785afb6
author David Benjamin <davidben@google.com> 1702407502 -0500
committer Boringssl LUCI CQ <boringssl-scoped@luci-project-accounts.iam.gserviceaccount.com> 1702414667 +0000

Disable 32-bit Arm assembly optimizations on iOS

The last iOS version that supported 32-bit was iOS 10, which I don't
believe any of our consumers support anymore. (Chromium does not,
neither does google/oss-policies-info.) Our iOS CI coverage comes from
Chromium, so this means we don't actually have any test coverage for
32-bit iOS, only compile coverage.

In addition to lacking any test coverage, 32-bit Arm assembly is more
platform-dependent than one might expect, between different limitiations
on patterns for PC-relative loads and lots of assembler quirks around
what kinds of label expressions (which show up in PC-relative loads a
lot) are allowed.

Finally, since iOS in that era did not do runtime detection of features
and relied on compiling a binary multiple times, the 32-bit assembly
would never enable AES acceleration anyway, so it's not as impactful as
on other platforms.

Between all that, it's no longer worth enabling this. Disable it in
target.h which, with the all the recent build simplifications, should be
sufficient to disable this code.

Update-Note: iOS on 32-bit Arm now disables assembly. This is unlikely
to impact anyone. As far as I can tell, 32-bit Arm for iOS thoroughly
does not exist anymore.

Change-Id: If31208b42047377ad1b4fb0af6fee17334f18330
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/64748
Auto-Submit: David Benjamin <davidben@google.com>
Reviewed-by: Bob Beck <bbe@google.com>
Commit-Queue: Bob Beck <bbe@google.com>