Commit Graph

21 Commits

Author SHA1 Message Date
Chris Fallin
00f357c028 Cranelift: support 14-bit Type index with some bitpacking. (#4269)
* Cranelift: make `ir::Type` a `u16`.

* Cranelift: pack ValueData back into 64 bits.

After extending `Type` to a `u16`, `ValueData` became 12 bytes rather
than 8. This packs it back down to 8 bytes (64 bits) by stealing two
bits from the `Type` for the enum discriminant (leaving 14 bits for the
type itself).

Performance comparison (3-way between original (`ty-u8`), 16-bit `Type`
(`ty-u16`), and this PR (`ty-packed`)):

```
~/work/sightglass% target/release/sightglass-cli benchmark \
    -e ~/ty-u8.so -e ~/ty-u16.so -e ~/ty-packed.so \
    --iterations-per-process 10 --processes 2 \
    benchmarks-next/spidermonkey/benchmark.wasm

compilation
  benchmarks-next/spidermonkey/benchmark.wasm
    cycles
      [20654406874 21749213920.50 22958520306] /home/cfallin/ty-packed.so
      [22227738316 22584704883.90 22916433748] /home/cfallin/ty-u16.so
      [20659150490 21598675968.60 22588108428] /home/cfallin/ty-u8.so
    nanoseconds
      [5435333269 5723139427.25 6041072883] /home/cfallin/ty-packed.so
      [5848788229 5942729637.85 6030030341] /home/cfallin/ty-u16.so
      [5436002390 5683248226.10 5943626225] /home/cfallin/ty-u8.so
```

So, when compiling SpiderMonkey.wasm, making `Type` 16 bits regresses
performance by 4.5% (5.683s -> 5.723s), while this PR gets 14 bits for a 1.0%
cost (5.683s -> 5.723s). That's still not great, and we can likely do better,
but it's a start.

* Fix test failure: entities to/from u32 via `{from,to}_bits`, not `{from,to}_u32`.
2022-07-05 14:51:02 -07:00
bjorn3
4c75616a7c Remove unused constants from cranelift-codegen-shared (#3479) 2021-11-18 18:51:35 +01:00
bjorn3
91d4f36970 Move generate_table from cranelift-codegen-shared to cranelift-codegen-meta 2021-10-29 14:43:09 +02:00
bjorn3
2b89b13c57 Move condcodes from cranelift-codegen-shared to cranelift-codegen 2021-10-10 14:23:35 +02:00
bjorn3
9e34df33b9 Remove the old x86 backend 2021-09-29 16:13:46 +02:00
bjorn3
2fc964ea35 Add serde serialization support for the full clif ir 2021-02-18 11:27:02 +01:00
Andrew Brown
c9e8889d47 Update clippy annotation to use latest version (#2375) 2020-11-09 09:24:59 -06:00
Benjamin Bouvier
35d9ab19b7 Review fixes; 2020-07-24 19:29:12 +02:00
Benjamin Bouvier
987c616bf5 machinst x64: implement support for dynamic heaps and explicit bound checks; 2020-07-24 19:29:12 +02:00
Andrew Brown
71914c7668 Fix clippy warnings in EncodingBits 2020-01-10 08:38:40 -08:00
Sean Stangl
cf9e762f16 Add a DynRex recipe type for x86, decreasing the number of recipes (#1298)
This patch adds a third mode for templates: REX inference is requestable
at template instantiation time. This reduces the number of recipes
by removing rex()/nonrex() redundancy for many instructions.
2019-12-19 15:49:34 -07:00
Andrew Brown
d4df756acf Remove packed_struct dependency; closes #1271 and #1284 (#1282) 2019-12-12 17:01:31 -08:00
Sean Stangl
f8ae622003 Use a struct interface for creating and reading encoding bits on x86. #1156 (#1212) 2019-11-13 18:01:13 -07:00
Peter Huene
9f506692c2 Fix clippy warnings.
This commit fixes the current set of (stable) clippy warnings in the repo.
2019-10-24 17:20:12 -07:00
Benjamin Bouvier
f668869508 Share constants between codegen and the meta crate; 2019-10-10 16:45:48 +02:00
Benjamin Bouvier
097fa0c7b1 Clarify a comment in constant_hash::generate_table; 2019-10-10 16:45:48 +02:00
Benjamin Bouvier
d404368dea Share constant_hash code between the meta and codegen crates; 2019-10-10 16:45:48 +02:00
Ujjwal Sharma
c062f12d7c [codegen] legalize icmp for 64 and 128 bit operands
Add legalizations for icmp and icmp_imm for i64 and i128 operands for
the narrow legalization set, allowing 32-bit ISAs (like x86-32) to
compare 64-bit integers and all ISAs to compare 128-bit integers.

Fixes: https://github.com/bnjbvr/cranelift-x86/issues/2
2019-10-10 11:06:19 +02:00
Ujjwal Sharma
19444649e7 [codegen] add to_static_str method to IntCC
Add a method to_static_str to objects of type IntCC which consumes the
object to basically do the opposite of IntCC::new.

Refs: https://github.com/CraneStation/cranelift/pull/1081#discussion_r329042331
2019-10-10 11:06:19 +02:00
Ujjwal Sharma
43a891dfa2 [codegen] add intcc conditions for reading overflow flag
Add conditions to IntCC for checking the overflow flag (Overflow,
NotOverflow).
2019-09-25 11:42:58 +02:00
Andrew Brown
c648fa3d06 Move condcodes.rs to a new crate, cranelift-codegen-shared
This move allows the `IntCC`/`FloatCC` enums to be used in both meta (for predicate matching) and in codegen. To avoid breaking any code dependent on the previous location of condcodes.rs (`cranelift-codegen/src/condcodes.rs`), the module is re-exported under `cranelift_codegen::ir`.
2019-09-24 09:33:07 -07:00