Don't create adjustments from a type to itself

[dotdash]

Currently, we're generating adjustments, for example, to get from &[u8]
to &[u8], which is unneeded and kicks us out of trans_into()
into trans() which means an additional stack slot and copy in the
unoptimized code.

[rust-highfive]

r? @nikomatsakis

(rust_highfive has picked a reviewer for you, use r? to override)

[dotdash]

I just noticed that this fails in a number of cases, probably because of region mismatches in various &[u8] types. Is there a way to compare types while ignoring regions?

[nikomatsakis]

The problem is that the adjustments are needed for the type system to be happy. I think we can trivially detect and optimize this sort of thing once MIR lands, however, which is probably my preferred strategy. Alternatively, it'd probably be fairly easy to detect and do a kind of peephole optimization for this case in trans.

[dotdash]

Oh, right, I completely forgot about that. I guess the equality check as is
in this PR could be added though, right? That at least gets rid of a few
unnecessary copies.
Am 16.09.2015 18:42 schrieb "Niko Matsakis" notifications@github.com:

The problem is that the adjustments are needed for the type system to be
happy. I think we can trivially detect and optimize this sort of thing once
MIR lands, however, which is probably my preferred strategy. Alternatively,
it'd probably be fairly easy to detect and do a kind of peephole
optimization for this case in trans.

—
Reply to this email directly or view it on GitHub
#28428 (comment).

[nikomatsakis]

@dotdash yeah, the equality check seems harmless. This by the way is an example of a case where it'd be nice to get some kind of unit test to help us ensure that micro-optimizations of generated code don't get lost.

[dotdash]

Added a test. Some time soon, I'll remember to add those right away...

[nikomatsakis]

@bors r+

@dotdash nice.

[bors]

📌 Commit bc201fe has been approved by nikomatsakis

[bors]

⌛ Testing commit bc201fe with merge d8d8830...

[bors]

💔 Test failed - auto-mac-32-opt

[dotdash]

@bors r=nikomatsakis 6def06c

sigh I skipped the size in the memcpy() call because I knew it's target dependent, but forgot that the same is true for the length field in the slice. Fixed.

[bors]

⌛ Testing commit 6def06c with merge 2be0d0a...

[bors]

☀️ Test successful - auto-linux-32-nopt-t, auto-linux-32-opt, auto-linux-64-nopt-t, auto-linux-64-opt, auto-linux-64-x-android-t, auto-mac-32-opt, auto-mac-64-nopt-t, auto-mac-64-opt, auto-win-gnu-32-nopt-t, auto-win-gnu-32-opt, auto-win-gnu-64-nopt-t, auto-win-gnu-64-opt, auto-win-msvc-32-opt, auto-win-msvc-64-opt