Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[vm/ffi] Optimize memcopy in nested struct assignment #43967

Closed
dcharkes opened this issue Oct 28, 2020 · 1 comment
Closed

[vm/ffi] Optimize memcopy in nested struct assignment #43967

dcharkes opened this issue Oct 28, 2020 · 1 comment
Labels
area-vm Use area-vm for VM related issues, including code coverage, and the AOT and JIT backends. library-ffi type-enhancement A request for a change that isn't a bug type-performance Issue relates to performance or code size

Comments

@dcharkes
Copy link
Contributor

dcharkes commented Oct 28, 2020

In https://dart-review.googlesource.com/c/sdk/+/169221 we implement a memcopy function that enables copying:

  • Pointer -> Pointer
  • Pointer -> TypedData
  • TypedData -> Pointer
  • TypedData -> TypedData

We should recognize the memcopy method and implement it with MemoryCopyInstr in the VM for performance.

edit: adding ideas from #48139:

dart:ffi uses the _memCopy helper when working with structs.

We should make sure to

  • Provide fast recognized implementation for this method (probably bottoming in MemoryCopyInstr.
  • Small constant size copies and unroll the copying loop in place.
@dcharkes dcharkes added area-vm Use area-vm for VM related issues, including code coverage, and the AOT and JIT backends. library-ffi labels Oct 28, 2020
@mkustermann mkustermann added the type-enhancement A request for a change that isn't a bug label Dec 16, 2020
@dcharkes dcharkes added the type-performance Issue relates to performance or code size label Dec 19, 2022
copybara-service bot pushed a commit that referenced this issue Dec 29, 2022
`_memCopy` inside `dart:ffi` is currently doing a per-byte copy in
Dart. This is rather slow, we should optimize this with the
`MemoryCopyInstr` in the VM.

This CL adds benchmarks to report the number of bytes copied per
second.

Adds only benchmarks with copies of 32^(0..3), as non-power-of-two
benchmarks did not seem to behave differently.

Since legacy mode is no longer benchmarked, the dart2 version of this
benchmark is omitted.

Benchmarks set up according to
https://dart-review.googlesource.com/c/sdk/+/200188

Bug: #43967
Change-Id: I3d9be8de725820fd3365a7dc85d15174bddc1ae6
Cq-Include-Trybots: luci.dart.try:benchmark-linux-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/277522
Reviewed-by: Jonas Termansen <sortie@google.com>
Commit-Queue: Daco Harkes <dacoharkes@google.com>
copybara-service bot pushed a commit that referenced this issue Jan 2, 2023
Replace the `_memCopy` implementation in Dart with `MemoryCopyInstr`.

Speeds up copying from and to `Pointer`s by up to a 100x for large
structs.

TEST=tests/ffi/function_callbacks_structs_by_value_test.dart

BENCHMARK=benchmarks/FfiStructCopy/dart/FfiStructCopy.dart

Before (x64 JIT):

FfiStructCopy.Copy1Bytes(RunTimeRaw): 86.26189953740506 ns.
FfiStructCopy.Copy16Bytes(RunTimeRaw): 14.006969563521945 ns.
FfiStructCopy.Copy1024Bytes(RunTimeRaw): 8.567228629242734 ns.
FfiStructCopy.Copy65536Bytes(RunTimeRaw): 8.469047080663412 ns.

After (x64 JIT):

FfiStructCopy.Copy1Bytes(RunTimeRaw): 56.79513144264321 ns.
FfiStructCopy.Copy16Bytes(RunTimeRaw): 3.562479879234367 ns.
FfiStructCopy.Copy1024Bytes(RunTimeRaw): 0.10771875669750132 ns.
FfiStructCopy.Copy65536Bytes(RunTimeRaw): 0.02758346614218262 ns.

Bug: #43967
Change-Id: I5d51c39a13b8c9522ee88d81bba7365caef70a27
Cq-Include-Trybots: luci.dart.try:vm-ffi-android-debug-arm-try,vm-ffi-android-debug-arm64c-try,vm-precomp-ffi-qemu-linux-release-arm-try,vm-precomp-ffi-qemu-linux-release-riscv64-try,vm-kernel-reload-rollback-linux-debug-x64-try,vm-kernel-reload-linux-debug-x64-try,vm-kernel-nnbd-mac-debug-arm64-try,vm-kernel-nnbd-mac-debug-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-debug-x64c-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/277523
Commit-Queue: Daco Harkes <dacoharkes@google.com>
Reviewed-by: Aske Simon Christensen <askesc@google.com>
@dcharkes
Copy link
Contributor Author

Small constant size copies and unroll the copying loop in place.

Tracking in:

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
area-vm Use area-vm for VM related issues, including code coverage, and the AOT and JIT backends. library-ffi type-enhancement A request for a change that isn't a bug type-performance Issue relates to performance or code size
Projects
None yet
Development

No branches or pull requests

2 participants