Stream: git-wasmtime

Topic: wasmtime / PR #4080 Rework x64 addressing-mode lowering t...

view this post on Zulip Wasmtime GitHub notifications bot (Apr 29 2022 at 22:24):

cfallin edited PR #4080 from x64-amodes to main:

This draft PR modifies the x64 backend's addressing-mode lowering to use both scaled indexing, e.g. [rax+rbx*4], and to consolidate offsets, so x+y+8+12 can become [rax+rbx+20]`.

Unfortunately this doesn't yet appear to be quite enough to improve on this really annoying sequence that occurs frequently in SpiderMonkey.wasm's hot blocks:

     0xC064911:  movq %rsi,%rdx
     0xC064914:  addl $3, %edx
     0xC064917:  movzbq 0(%r9,%rdx),%rdx

the issue being that edx+3, as a 32-bit add, can wrap around and so this is not quite equivalent to 3(%r9, %rsi) (assuming we could prove rsi is zero-extended). @abrown or others, if you have ideas on a better lowering here, I'd be happy to hear them!

This PR builds on top of #4072, #4078, and #4079; only the last commit is new.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 17:55):

abrown submitted PR review.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 17:55):

abrown submitted PR review.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 17:55):

abrown created PR review comment:

I think the rules following this would benefit from a section similar to the one just removed. I don't necessarily mean pseudo-code, I mean a section describing the high-level transformations being made, why they're being made, what limitations are involved, etc. For example, I don't completely understand why we need amode_initial to kick the transformation off with an invalid_reg. And I don't really understand why the priorities are the way they are--what I mean is that the logic doesn't quite jump off the page when reading the rules. Perhaps some better documentation here would be helpful?

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 18:21):

fitzgen submitted PR review.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 18:21):

fitzgen submitted PR review.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 18:21):

fitzgen created PR review comment:

At this point, we might want to just make the ISLE definition of Amode the canonical one (like we do for Inst) and then import it from ISLE where it is used elsewhere, rather than have duplicate definitions.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 18:21):

fitzgen created PR review comment:

Don't we (didn't we?) have a is_invalid method or something like that on Reg? It seems like an abstraction violation to be talking about sentinels outside of the type definition. (e.g. what if there was more than one bit pattern for Reg that is invalid?)

If we don't have that method anymore, can we add it and use it here? It would also make this function one line shorter since we wouldn't need the use import.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 18:21):

fitzgen created PR review comment:

Nitpick: align the pattern arguments:

                   (iconst (simm32 c)))

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 18:21):

fitzgen created PR review comment:

Ditto here and more below.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 18:21):

fitzgen created PR review comment:

While you're messing with the DSL, it would be nice to be able to override variables rather than pretend we're erlang :-p

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 21:34):

cfallin updated PR #4080 from x64-amodes to main.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 21:35):

cfallin submitted PR review.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 21:35):

cfallin created PR review comment:

Done. We still need to import the trait (it's a provided trait method) but we don't depend on the "equals the one sentinel value" assumption anymore.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 21:35):

cfallin submitted PR review.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 21:35):

cfallin created PR review comment:

Done, this was me adding priorities and not re-indenting, sorry!

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 21:35):

cfallin submitted PR review.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 21:35):

cfallin created PR review comment:

Fixed.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 21:37):

cfallin submitted PR review.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 21:37):

cfallin created PR review comment:

I agree it'd be nice! Looks like we have a tracking issue (#3526) already so I'll resolve this comment for now...

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 21:37):

cfallin submitted PR review.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 21:37):

cfallin created PR review comment:

Done!

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 21:38):

cfallin submitted PR review.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 21:38):

cfallin created PR review comment:

I added a bunch of comments here; hopefully it's clearer! Let me know if anything is still unclear -- I'll wait for your r+ as well before merging.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 21:58):

abrown submitted PR review.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 21:58):

abrown created PR review comment:

;; load/store's built-in `Offset32` and `invalid_reg` as the

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 21:58):

abrown submitted PR review.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 21:58):

abrown created PR review comment:

;; - Case 2: When we have an `ImmReg` with a valid register already,

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 21:58):

abrown created PR review comment:

;; `amode_add`, for each operand of the `iadd`. This is what allows us

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 22:06):

cfallin updated PR #4080 from x64-amodes to main.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 22:07):

cfallin updated PR #4080 from x64-amodes to main.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 22:07):

cfallin updated PR #4080 from x64-amodes to main.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 22:51):

cfallin updated PR #4080 from x64-amodes to main.

view this post on Zulip Wasmtime GitHub notifications bot (May 02 2022 at 23:20):

cfallin merged PR #4080.

Last updated: Nov 22 2024 at 16:03 UTC