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
ARROW-11037: [Rust] Optimized creation of string array from iterator. #9016
Conversation
Codecov Report
@@ Coverage Diff @@
## master #9016 +/- ##
==========================================
- Coverage 82.60% 82.60% -0.01%
==========================================
Files 204 204
Lines 50175 50176 +1
==========================================
- Hits 41447 41446 -1
- Misses 8728 8730 +2
Continue to review full report at Codecov.
|
I think the previous time I tried this in a kernel, I got worse perf results than copying them from the |
That is a good point. I assumed that such a memcopy would always be worse for sufficiently large arrays, but I have not validated this. |
I am moving this to draft until the benches are in place. @Dandandan is right that this is not necessarily an improvement. I am investigating this.\ |
let mut values = Vec::new(); | ||
let mut offsets = | ||
MutableBuffer::new((v.len() + 1) * std::mem::size_of::<OffsetSize>()); | ||
let mut values = MutableBuffer::new(0); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Might be interesting to test here whether it is cheaper to calculate offsets + total size first for values
buffer based on final length_so_far
so that it doesn't require extra allocations.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
So something like
for s in &v {
length_so_far = length_so_far + OffsetSize::from_usize(s.len()).unwrap();
offsets.extend_from_slice(length_so_far.to_byte_slice());
}
let mut values = MutableBuffer::new(length_so_far);
for s in &v {
values.extend_from_slice(s.as_bytes());
}
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Probably the offset buffer is also faster after converting it to a slice using typed_data_mut::<OffsetSize>()
and iterating it with iter_mut
, this way it can skip the capacity checks, and also current conversion to a byte slice may add some overhead?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Both great ideas. If it is ok for you, I will leave them to future work, as I am focusing my work on the MutableBuffer
atm.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I.e. the general direction that I am pushing towards is to stop using Vec
and replace it by MutableBuffer
to avoid extra allocations. Once that is in place, we can replace these by iter_mut
.
also current conversion to a byte slice may add some overhead?
I believe so, as to_byte_slice
returns a slice of unknown size, both for &T
and &[T]
. Either the compiler optimizes it, or there is an extra cost. I benched a 10% diff in building buffers when introduced a method that was not doing bound checks on these. IMO we should be using to_le_bytes
, and have a method MutableBuffer::push<ToByteSlice>
. I think we need the crate byteorder
because AFAIK std's to_le_bytes
does not have a trait (which we need).
I have re-opened this PR as #9032 shows that creating a buffer from a mutable buffer is 2x faster. |
The full set of Rust CI tests did not run on this PR :( Can you please rebase this PR against apache/master to pick up the changes in #9056 so that they do? I apologize for the inconvenience. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looks good to me. Thanks @jorgecarleitao -- I didn't test this PR out or run the benchmarks but I think your results are good for me;
I did read the code and I understand the changes. 👍
This PR refactors `MutableBuffer::extend_from_slice` to remove the need to use `to_byte_slice` on every call, thereby removing its level of indirection, that does not allow the compiler to optimize out some code. This is the second performance improvement originally presented in #8796 and, together with #9027 , brings the performance of "MutableBuffer" to the same level as `Vec<u8>`, in particular to building buffers on the fly. Basically, when converting to a byte slice `&[u8]`, the compiler loses the type size information, and thus needs to perform extra checks and can't just optimize out the code. This PR adopts the same API as `Vec<T>::extend_from_slice`, but since our buffers are in `u8` (i.e. a la `Vec<u8>`), I made the signature ``` pub fn extend_from_slice<T: ToByteSlice>(&mut self, items: &[T]) pub fn push<T: ToByteSlice>(&mut self, item: &T) ``` i.e. it consumes something that can be converted to a byte slice, but internally makes the conversion to bytes (as `to_byte_slice` was doing). Credits for the root cause analysis that lead to this PR go to @Dandandan, [originally fielded here](#9016 (comment)). > [...] current conversion to a byte slice may add some overhead? - @Dandandan Benches (against master, so, both this PR and #9044 ): ``` Switched to branch 'perf_buffer' Your branch and 'origin/perf_buffer' have diverged, and have 6 and 1 different commits each, respectively. (use "git pull" to merge the remote branch into yours) Compiling arrow v3.0.0-SNAPSHOT (/Users/jorgecarleitao/projects/arrow/rust/arrow) Finished bench [optimized] target(s) in 1m 00s Running /Users/jorgecarleitao/projects/arrow/rust/target/release/deps/buffer_create-915da5f1abaf0471 Gnuplot not found, using plotters backend mutable time: [463.11 us 463.57 us 464.07 us] change: [-19.508% -18.571% -17.526%] (p = 0.00 < 0.05) Performance has improved. Found 10 outliers among 100 measurements (10.00%) 1 (1.00%) high mild 9 (9.00%) high severe mutable prepared time: [527.84 us 528.46 us 529.14 us] change: [-13.356% -12.522% -11.790%] (p = 0.00 < 0.05) Performance has improved. Found 12 outliers among 100 measurements (12.00%) 5 (5.00%) high mild 7 (7.00%) high severe Benchmarking from_slice: Warming up for 3.0000 s Warning: Unable to complete 100 samples in 5.0s. You may wish to increase target time to 6.1s, enable flat sampling, or reduce sample count to 60. from_slice time: [1.1968 ms 1.1979 ms 1.1991 ms] change: [-6.8697% -6.2029% -5.5812%] (p = 0.00 < 0.05) Performance has improved. Found 10 outliers among 100 measurements (10.00%) 3 (3.00%) high mild 7 (7.00%) high severe from_slice prepared time: [917.49 us 918.89 us 920.60 us] change: [-6.5111% -5.9102% -5.3038%] (p = 0.00 < 0.05) Performance has improved. Found 10 outliers among 100 measurements (10.00%) 4 (4.00%) high mild 6 (6.00%) high severe ``` Closes #9076 from jorgecarleitao/perf_buffer Authored-by: Jorge C. Leitao <jorgecarleitao@gmail.com> Signed-off-by: Jorge C. Leitao <jorgecarleitao@gmail.com>
This PR refactors `MutableBuffer::extend_from_slice` to remove the need to use `to_byte_slice` on every call, thereby removing its level of indirection, that does not allow the compiler to optimize out some code. This is the second performance improvement originally presented in #8796 and, together with #9027 , brings the performance of "MutableBuffer" to the same level as `Vec<u8>`, in particular to building buffers on the fly. Basically, when converting to a byte slice `&[u8]`, the compiler loses the type size information, and thus needs to perform extra checks and can't just optimize out the code. This PR adopts the same API as `Vec<T>::extend_from_slice`, but since our buffers are in `u8` (i.e. a la `Vec<u8>`), I made the signature ``` pub fn extend_from_slice<T: ToByteSlice>(&mut self, items: &[T]) pub fn push<T: ToByteSlice>(&mut self, item: &T) ``` i.e. it consumes something that can be converted to a byte slice, but internally makes the conversion to bytes (as `to_byte_slice` was doing). Credits for the root cause analysis that lead to this PR go to @Dandandan, [originally fielded here](#9016 (comment)). > [...] current conversion to a byte slice may add some overhead? - @Dandandan Benches (against master, so, both this PR and #9044 ): ``` Switched to branch 'perf_buffer' Your branch and 'origin/perf_buffer' have diverged, and have 6 and 1 different commits each, respectively. (use "git pull" to merge the remote branch into yours) Compiling arrow v3.0.0-SNAPSHOT (/Users/jorgecarleitao/projects/arrow/rust/arrow) Finished bench [optimized] target(s) in 1m 00s Running /Users/jorgecarleitao/projects/arrow/rust/target/release/deps/buffer_create-915da5f1abaf0471 Gnuplot not found, using plotters backend mutable time: [463.11 us 463.57 us 464.07 us] change: [-19.508% -18.571% -17.526%] (p = 0.00 < 0.05) Performance has improved. Found 10 outliers among 100 measurements (10.00%) 1 (1.00%) high mild 9 (9.00%) high severe mutable prepared time: [527.84 us 528.46 us 529.14 us] change: [-13.356% -12.522% -11.790%] (p = 0.00 < 0.05) Performance has improved. Found 12 outliers among 100 measurements (12.00%) 5 (5.00%) high mild 7 (7.00%) high severe Benchmarking from_slice: Warming up for 3.0000 s Warning: Unable to complete 100 samples in 5.0s. You may wish to increase target time to 6.1s, enable flat sampling, or reduce sample count to 60. from_slice time: [1.1968 ms 1.1979 ms 1.1991 ms] change: [-6.8697% -6.2029% -5.5812%] (p = 0.00 < 0.05) Performance has improved. Found 10 outliers among 100 measurements (10.00%) 3 (3.00%) high mild 7 (7.00%) high severe from_slice prepared time: [917.49 us 918.89 us 920.60 us] change: [-6.5111% -5.9102% -5.3038%] (p = 0.00 < 0.05) Performance has improved. Found 10 outliers among 100 measurements (10.00%) 4 (4.00%) high mild 6 (6.00%) high severe ``` Closes #9076 from jorgecarleitao/perf_buffer Authored-by: Jorge C. Leitao <jorgecarleitao@gmail.com> Signed-off-by: Jorge C. Leitao <jorgecarleitao@gmail.com>
This PR refactors `MutableBuffer::extend_from_slice` to remove the need to use `to_byte_slice` on every call, thereby removing its level of indirection, that does not allow the compiler to optimize out some code. This is the second performance improvement originally presented in #8796 and, together with #9027 , brings the performance of "MutableBuffer" to the same level as `Vec<u8>`, in particular to building buffers on the fly. Basically, when converting to a byte slice `&[u8]`, the compiler loses the type size information, and thus needs to perform extra checks and can't just optimize out the code. This PR adopts the same API as `Vec<T>::extend_from_slice`, but since our buffers are in `u8` (i.e. a la `Vec<u8>`), I made the signature ``` pub fn extend_from_slice<T: ToByteSlice>(&mut self, items: &[T]) pub fn push<T: ToByteSlice>(&mut self, item: &T) ``` i.e. it consumes something that can be converted to a byte slice, but internally makes the conversion to bytes (as `to_byte_slice` was doing). Credits for the root cause analysis that lead to this PR go to @Dandandan, [originally fielded here](apache/arrow#9016 (comment)). > [...] current conversion to a byte slice may add some overhead? - @Dandandan Benches (against master, so, both this PR and #9044 ): ``` Switched to branch 'perf_buffer' Your branch and 'origin/perf_buffer' have diverged, and have 6 and 1 different commits each, respectively. (use "git pull" to merge the remote branch into yours) Compiling arrow v3.0.0-SNAPSHOT (/Users/jorgecarleitao/projects/arrow/rust/arrow) Finished bench [optimized] target(s) in 1m 00s Running /Users/jorgecarleitao/projects/arrow/rust/target/release/deps/buffer_create-915da5f1abaf0471 Gnuplot not found, using plotters backend mutable time: [463.11 us 463.57 us 464.07 us] change: [-19.508% -18.571% -17.526%] (p = 0.00 < 0.05) Performance has improved. Found 10 outliers among 100 measurements (10.00%) 1 (1.00%) high mild 9 (9.00%) high severe mutable prepared time: [527.84 us 528.46 us 529.14 us] change: [-13.356% -12.522% -11.790%] (p = 0.00 < 0.05) Performance has improved. Found 12 outliers among 100 measurements (12.00%) 5 (5.00%) high mild 7 (7.00%) high severe Benchmarking from_slice: Warming up for 3.0000 s Warning: Unable to complete 100 samples in 5.0s. You may wish to increase target time to 6.1s, enable flat sampling, or reduce sample count to 60. from_slice time: [1.1968 ms 1.1979 ms 1.1991 ms] change: [-6.8697% -6.2029% -5.5812%] (p = 0.00 < 0.05) Performance has improved. Found 10 outliers among 100 measurements (10.00%) 3 (3.00%) high mild 7 (7.00%) high severe from_slice prepared time: [917.49 us 918.89 us 920.60 us] change: [-6.5111% -5.9102% -5.3038%] (p = 0.00 < 0.05) Performance has improved. Found 10 outliers among 100 measurements (10.00%) 4 (4.00%) high mild 6 (6.00%) high severe ``` Closes #9076 from jorgecarleitao/perf_buffer Authored-by: Jorge C. Leitao <jorgecarleitao@gmail.com> Signed-off-by: Jorge C. Leitao <jorgecarleitao@gmail.com>
Avoids a memcopy from 2 `Vec<T>` to `Buffer`, by building the buffers on the fly. Closes apache#9016 from jorgecarleitao/optimize_string Authored-by: Jorge C. Leitao <jorgecarleitao@gmail.com> Signed-off-by: Andrew Lamb <andrew@nerdnetworks.org>
This PR refactors `MutableBuffer::extend_from_slice` to remove the need to use `to_byte_slice` on every call, thereby removing its level of indirection, that does not allow the compiler to optimize out some code. This is the second performance improvement originally presented in apache#8796 and, together with apache#9027 , brings the performance of "MutableBuffer" to the same level as `Vec<u8>`, in particular to building buffers on the fly. Basically, when converting to a byte slice `&[u8]`, the compiler loses the type size information, and thus needs to perform extra checks and can't just optimize out the code. This PR adopts the same API as `Vec<T>::extend_from_slice`, but since our buffers are in `u8` (i.e. a la `Vec<u8>`), I made the signature ``` pub fn extend_from_slice<T: ToByteSlice>(&mut self, items: &[T]) pub fn push<T: ToByteSlice>(&mut self, item: &T) ``` i.e. it consumes something that can be converted to a byte slice, but internally makes the conversion to bytes (as `to_byte_slice` was doing). Credits for the root cause analysis that lead to this PR go to @Dandandan, [originally fielded here](apache#9016 (comment)). > [...] current conversion to a byte slice may add some overhead? - @Dandandan Benches (against master, so, both this PR and apache#9044 ): ``` Switched to branch 'perf_buffer' Your branch and 'origin/perf_buffer' have diverged, and have 6 and 1 different commits each, respectively. (use "git pull" to merge the remote branch into yours) Compiling arrow v3.0.0-SNAPSHOT (/Users/jorgecarleitao/projects/arrow/rust/arrow) Finished bench [optimized] target(s) in 1m 00s Running /Users/jorgecarleitao/projects/arrow/rust/target/release/deps/buffer_create-915da5f1abaf0471 Gnuplot not found, using plotters backend mutable time: [463.11 us 463.57 us 464.07 us] change: [-19.508% -18.571% -17.526%] (p = 0.00 < 0.05) Performance has improved. Found 10 outliers among 100 measurements (10.00%) 1 (1.00%) high mild 9 (9.00%) high severe mutable prepared time: [527.84 us 528.46 us 529.14 us] change: [-13.356% -12.522% -11.790%] (p = 0.00 < 0.05) Performance has improved. Found 12 outliers among 100 measurements (12.00%) 5 (5.00%) high mild 7 (7.00%) high severe Benchmarking from_slice: Warming up for 3.0000 s Warning: Unable to complete 100 samples in 5.0s. You may wish to increase target time to 6.1s, enable flat sampling, or reduce sample count to 60. from_slice time: [1.1968 ms 1.1979 ms 1.1991 ms] change: [-6.8697% -6.2029% -5.5812%] (p = 0.00 < 0.05) Performance has improved. Found 10 outliers among 100 measurements (10.00%) 3 (3.00%) high mild 7 (7.00%) high severe from_slice prepared time: [917.49 us 918.89 us 920.60 us] change: [-6.5111% -5.9102% -5.3038%] (p = 0.00 < 0.05) Performance has improved. Found 10 outliers among 100 measurements (10.00%) 4 (4.00%) high mild 6 (6.00%) high severe ``` Closes apache#9076 from jorgecarleitao/perf_buffer Authored-by: Jorge C. Leitao <jorgecarleitao@gmail.com> Signed-off-by: Jorge C. Leitao <jorgecarleitao@gmail.com>
This PR refactors `MutableBuffer::extend_from_slice` to remove the need to use `to_byte_slice` on every call, thereby removing its level of indirection, that does not allow the compiler to optimize out some code. This is the second performance improvement originally presented in apache#8796 and, together with apache#9027 , brings the performance of "MutableBuffer" to the same level as `Vec<u8>`, in particular to building buffers on the fly. Basically, when converting to a byte slice `&[u8]`, the compiler loses the type size information, and thus needs to perform extra checks and can't just optimize out the code. This PR adopts the same API as `Vec<T>::extend_from_slice`, but since our buffers are in `u8` (i.e. a la `Vec<u8>`), I made the signature ``` pub fn extend_from_slice<T: ToByteSlice>(&mut self, items: &[T]) pub fn push<T: ToByteSlice>(&mut self, item: &T) ``` i.e. it consumes something that can be converted to a byte slice, but internally makes the conversion to bytes (as `to_byte_slice` was doing). Credits for the root cause analysis that lead to this PR go to @Dandandan, [originally fielded here](apache#9016 (comment)). > [...] current conversion to a byte slice may add some overhead? - @Dandandan Benches (against master, so, both this PR and apache#9044 ): ``` Switched to branch 'perf_buffer' Your branch and 'origin/perf_buffer' have diverged, and have 6 and 1 different commits each, respectively. (use "git pull" to merge the remote branch into yours) Compiling arrow v3.0.0-SNAPSHOT (/Users/jorgecarleitao/projects/arrow/rust/arrow) Finished bench [optimized] target(s) in 1m 00s Running /Users/jorgecarleitao/projects/arrow/rust/target/release/deps/buffer_create-915da5f1abaf0471 Gnuplot not found, using plotters backend mutable time: [463.11 us 463.57 us 464.07 us] change: [-19.508% -18.571% -17.526%] (p = 0.00 < 0.05) Performance has improved. Found 10 outliers among 100 measurements (10.00%) 1 (1.00%) high mild 9 (9.00%) high severe mutable prepared time: [527.84 us 528.46 us 529.14 us] change: [-13.356% -12.522% -11.790%] (p = 0.00 < 0.05) Performance has improved. Found 12 outliers among 100 measurements (12.00%) 5 (5.00%) high mild 7 (7.00%) high severe Benchmarking from_slice: Warming up for 3.0000 s Warning: Unable to complete 100 samples in 5.0s. You may wish to increase target time to 6.1s, enable flat sampling, or reduce sample count to 60. from_slice time: [1.1968 ms 1.1979 ms 1.1991 ms] change: [-6.8697% -6.2029% -5.5812%] (p = 0.00 < 0.05) Performance has improved. Found 10 outliers among 100 measurements (10.00%) 3 (3.00%) high mild 7 (7.00%) high severe from_slice prepared time: [917.49 us 918.89 us 920.60 us] change: [-6.5111% -5.9102% -5.3038%] (p = 0.00 < 0.05) Performance has improved. Found 10 outliers among 100 measurements (10.00%) 4 (4.00%) high mild 6 (6.00%) high severe ``` Closes apache#9076 from jorgecarleitao/perf_buffer Authored-by: Jorge C. Leitao <jorgecarleitao@gmail.com> Signed-off-by: Jorge C. Leitao <jorgecarleitao@gmail.com>
Avoids a memcopy from 2
Vec<T>
toBuffer
, by building the buffers on the fly.