[Merged by Bors] - Add maxperf build profile #3608
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
michaelsproul
added
ready-for-review
michaelsproul
force-pushed
the
maxperf
branch
from
2fcd773 to
2b57262
Compare
|
Interestingly, the release builds on CI don't seem to take much longer than they did previously: I think this is because Github actions is already pretty close to single-threaded, so it wasn't gaining much from compiling the early crates in parallel (unlike the 5950X which gains a lot by parallelism). This has solidified my commitment to maxperf-by-default for binaries, and the faster |
|
Oops, accidental submission. I agree with option 1 for the time being |
paulhauner
added
ready-for-merge
|
Cool! I think I've got the release CI pretty much sorted, but would like to do a practice run before we do a release for real. Maybe just a "pre-release" that we release for beta testing, something like |
|
Updated the book and tweaked the defaults so that:
Running this on my fork here:
Will merge once a few of those builds succeed. |
|
bors r+ |
bors bot
pushed a commit
that referenced
this pull request
Sep 28, 2022
## Proposed Changes Add a new Cargo compilation profile called `maxperf` which enables more aggressive compiler optimisations at the expense of compilation time. Some rough initial benchmarks show that this can provide up to a 25% reduction to run time for CPU bound tasks like block processing: https://docs.google.com/spreadsheets/d/15jHuZe7lLHhZq9Nw8kc6EL0Qh_N_YAYqkW2NQ_Afmtk/edit The numbers in that spreadsheet compare the `consensus-context` branch from #3604 to the same branch compiled with the `maxperf` profile using: ``` PROFILE=maxperf make install-lcli ``` ## Additional Info The downsides of the maxperf profile are: - It increases compile times substantially, which will particularly impact low-spec hardware. Compiling `lcli` is about 3x slower. Compiling Lighthouse is about 5x slower on my 5950X: 17m 38s rather than 3m 28s. As a result I think we should not enable this everywhere by default. - **Option 1**: enable by default for our released binaries. This gives the majority of users the fastest version of `lighthouse` possible, at the expense of slowing down our release CI. Source builds will continue to use the default `release` profile unless users opt-in to `maxperf`. - **Option 2**: enable by default for source builds. This gives users building from source an edge, but makes them pay for it with compilation time. I think I would prefer Option 1. I'll try doing some benchmarking to see how long a maxperf build of Lighthouse would take on GitHub actions. Credit to Nicholas Nethercote for documenting these options in the Rust Performance Book: https://nnethercote.github.io/perf-book/build-configuration.html.
|
Build failed (retrying...): |
bors bot
pushed a commit
that referenced
this pull request
Sep 29, 2022
## Proposed Changes Add a new Cargo compilation profile called `maxperf` which enables more aggressive compiler optimisations at the expense of compilation time. Some rough initial benchmarks show that this can provide up to a 25% reduction to run time for CPU bound tasks like block processing: https://docs.google.com/spreadsheets/d/15jHuZe7lLHhZq9Nw8kc6EL0Qh_N_YAYqkW2NQ_Afmtk/edit The numbers in that spreadsheet compare the `consensus-context` branch from #3604 to the same branch compiled with the `maxperf` profile using: ``` PROFILE=maxperf make install-lcli ``` ## Additional Info The downsides of the maxperf profile are: - It increases compile times substantially, which will particularly impact low-spec hardware. Compiling `lcli` is about 3x slower. Compiling Lighthouse is about 5x slower on my 5950X: 17m 38s rather than 3m 28s. As a result I think we should not enable this everywhere by default. - **Option 1**: enable by default for our released binaries. This gives the majority of users the fastest version of `lighthouse` possible, at the expense of slowing down our release CI. Source builds will continue to use the default `release` profile unless users opt-in to `maxperf`. - **Option 2**: enable by default for source builds. This gives users building from source an edge, but makes them pay for it with compilation time. I think I would prefer Option 1. I'll try doing some benchmarking to see how long a maxperf build of Lighthouse would take on GitHub actions. Credit to Nicholas Nethercote for documenting these options in the Rust Performance Book: https://nnethercote.github.io/perf-book/build-configuration.html.
|
Build failed (retrying...): |
|
bors r- |
|
Canceled. |
|
bors r+ |
bors bot
pushed a commit
that referenced
this pull request
Sep 29, 2022
## Proposed Changes Add a new Cargo compilation profile called `maxperf` which enables more aggressive compiler optimisations at the expense of compilation time. Some rough initial benchmarks show that this can provide up to a 25% reduction to run time for CPU bound tasks like block processing: https://docs.google.com/spreadsheets/d/15jHuZe7lLHhZq9Nw8kc6EL0Qh_N_YAYqkW2NQ_Afmtk/edit The numbers in that spreadsheet compare the `consensus-context` branch from #3604 to the same branch compiled with the `maxperf` profile using: ``` PROFILE=maxperf make install-lcli ``` ## Additional Info The downsides of the maxperf profile are: - It increases compile times substantially, which will particularly impact low-spec hardware. Compiling `lcli` is about 3x slower. Compiling Lighthouse is about 5x slower on my 5950X: 17m 38s rather than 3m 28s. As a result I think we should not enable this everywhere by default. - **Option 1**: enable by default for our released binaries. This gives the majority of users the fastest version of `lighthouse` possible, at the expense of slowing down our release CI. Source builds will continue to use the default `release` profile unless users opt-in to `maxperf`. - **Option 2**: enable by default for source builds. This gives users building from source an edge, but makes them pay for it with compilation time. I think I would prefer Option 1. I'll try doing some benchmarking to see how long a maxperf build of Lighthouse would take on GitHub actions. Credit to Nicholas Nethercote for documenting these options in the Rust Performance Book: https://nnethercote.github.io/perf-book/build-configuration.html.
|
Build failed (retrying...): |
bors bot
pushed a commit
that referenced
this pull request
Sep 29, 2022
## Proposed Changes Add a new Cargo compilation profile called `maxperf` which enables more aggressive compiler optimisations at the expense of compilation time. Some rough initial benchmarks show that this can provide up to a 25% reduction to run time for CPU bound tasks like block processing: https://docs.google.com/spreadsheets/d/15jHuZe7lLHhZq9Nw8kc6EL0Qh_N_YAYqkW2NQ_Afmtk/edit The numbers in that spreadsheet compare the `consensus-context` branch from #3604 to the same branch compiled with the `maxperf` profile using: ``` PROFILE=maxperf make install-lcli ``` ## Additional Info The downsides of the maxperf profile are: - It increases compile times substantially, which will particularly impact low-spec hardware. Compiling `lcli` is about 3x slower. Compiling Lighthouse is about 5x slower on my 5950X: 17m 38s rather than 3m 28s. As a result I think we should not enable this everywhere by default. - **Option 1**: enable by default for our released binaries. This gives the majority of users the fastest version of `lighthouse` possible, at the expense of slowing down our release CI. Source builds will continue to use the default `release` profile unless users opt-in to `maxperf`. - **Option 2**: enable by default for source builds. This gives users building from source an edge, but makes them pay for it with compilation time. I think I would prefer Option 1. I'll try doing some benchmarking to see how long a maxperf build of Lighthouse would take on GitHub actions. Credit to Nicholas Nethercote for documenting these options in the Rust Performance Book: https://nnethercote.github.io/perf-book/build-configuration.html.
|
bors r- |
|
Canceled. |
|
bors r+ |
bors bot
pushed a commit
that referenced
this pull request
Sep 29, 2022
## Proposed Changes Add a new Cargo compilation profile called `maxperf` which enables more aggressive compiler optimisations at the expense of compilation time. Some rough initial benchmarks show that this can provide up to a 25% reduction to run time for CPU bound tasks like block processing: https://docs.google.com/spreadsheets/d/15jHuZe7lLHhZq9Nw8kc6EL0Qh_N_YAYqkW2NQ_Afmtk/edit The numbers in that spreadsheet compare the `consensus-context` branch from #3604 to the same branch compiled with the `maxperf` profile using: ``` PROFILE=maxperf make install-lcli ``` ## Additional Info The downsides of the maxperf profile are: - It increases compile times substantially, which will particularly impact low-spec hardware. Compiling `lcli` is about 3x slower. Compiling Lighthouse is about 5x slower on my 5950X: 17m 38s rather than 3m 28s. As a result I think we should not enable this everywhere by default. - **Option 1**: enable by default for our released binaries. This gives the majority of users the fastest version of `lighthouse` possible, at the expense of slowing down our release CI. Source builds will continue to use the default `release` profile unless users opt-in to `maxperf`. - **Option 2**: enable by default for source builds. This gives users building from source an edge, but makes them pay for it with compilation time. I think I would prefer Option 1. I'll try doing some benchmarking to see how long a maxperf build of Lighthouse would take on GitHub actions. Credit to Nicholas Nethercote for documenting these options in the Rust Performance Book: https://nnethercote.github.io/perf-book/build-configuration.html.
|
Pull request successfully merged into unstable. Build succeeded:
|
bors
bot
changed the title
Woodpile37
pushed a commit
to Woodpile37/lighthouse
that referenced
this pull request
Jan 6, 2024
## Proposed Changes Add a new Cargo compilation profile called `maxperf` which enables more aggressive compiler optimisations at the expense of compilation time. Some rough initial benchmarks show that this can provide up to a 25% reduction to run time for CPU bound tasks like block processing: https://docs.google.com/spreadsheets/d/15jHuZe7lLHhZq9Nw8kc6EL0Qh_N_YAYqkW2NQ_Afmtk/edit The numbers in that spreadsheet compare the `consensus-context` branch from sigp#3604 to the same branch compiled with the `maxperf` profile using: ``` PROFILE=maxperf make install-lcli ``` ## Additional Info The downsides of the maxperf profile are: - It increases compile times substantially, which will particularly impact low-spec hardware. Compiling `lcli` is about 3x slower. Compiling Lighthouse is about 5x slower on my 5950X: 17m 38s rather than 3m 28s. As a result I think we should not enable this everywhere by default. - **Option 1**: enable by default for our released binaries. This gives the majority of users the fastest version of `lighthouse` possible, at the expense of slowing down our release CI. Source builds will continue to use the default `release` profile unless users opt-in to `maxperf`. - **Option 2**: enable by default for source builds. This gives users building from source an edge, but makes them pay for it with compilation time. I think I would prefer Option 1. I'll try doing some benchmarking to see how long a maxperf build of Lighthouse would take on GitHub actions. Credit to Nicholas Nethercote for documenting these options in the Rust Performance Book: https://nnethercote.github.io/perf-book/build-configuration.html.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Proposed Changes
Add a new Cargo compilation profile called
maxperfwhich enables more aggressive compiler optimisations at the expense of compilation time.Some rough initial benchmarks show that this can provide up to a 25% reduction to run time for CPU bound tasks like block processing: https://docs.google.com/spreadsheets/d/15jHuZe7lLHhZq9Nw8kc6EL0Qh_N_YAYqkW2NQ_Afmtk/edit
The numbers in that spreadsheet compare the
consensus-contextbranch from #3604 to the same branch compiled with themaxperfprofile using:Additional Info
The downsides of the maxperf profile are:
lcliis about 3x slower. Compiling Lighthouse is about 5x slower on my 5950X: 17m 38s rather than 3m 28s.As a result I think we should not enable this everywhere by default.
lighthousepossible, at the expense of slowing down our release CI. Source builds will continue to use the defaultreleaseprofile unless users opt-in tomaxperf.I think I would prefer Option 1. I'll try doing some benchmarking to see how long a maxperf build of Lighthouse would take on GitHub actions.
Credit to Nicholas Nethercote for documenting these options in the Rust Performance Book: https://nnethercote.github.io/perf-book/build-configuration.html.