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.

Sign up for GitHub

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

Jump to bottom

Refactor for new memory syncronization API #15

Open
wants to merge 2 commits into
base: master
Choose a base branch
from
Open

Refactor for new memory syncronization API #15

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
8292e43
refactor/tensor: convert to the new memory management API
alexandermorozov Apr 22, 2016
84c6666
refactor/tests: fix and refactor tests and benches
alexandermorozov Apr 23, 2016
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
239 changes: 29 additions & 210 deletions benches/rblas_overhead.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,244 +16,63 @@ fn backend() -> Backend<Native> {
Backend::<Native>::default().unwrap()
}

#[bench]
fn bench_1000_dot_100_rblas(b: &mut Bencher) {
fn bench_dot_rblas(b: &mut Bencher, n: usize) {
let mut rng = thread_rng();
let slice_a = rng.gen_iter::<f32>().take(100).collect::<Vec<f32>>();
let slice_b = rng.gen_iter::<f32>().take(100).collect::<Vec<f32>>();
let slice_a = rng.gen_iter::<f32>().take(n).collect::<Vec<f32>>();
let slice_b = rng.gen_iter::<f32>().take(n).collect::<Vec<f32>>();

b.iter(|| {
for _ in 0..1000 {
let res = rblas::Dot::dot(&slice_a, &slice_b);
test::black_box(res);
}
let res = rblas::Dot::dot(&slice_a, &slice_b);
test::black_box(res);
});
}

#[bench]
fn bench_1000_dot_100_collenchyma(b: &mut Bencher) {
fn bench_dot_collenchyma(b: &mut Bencher, n: usize) {
let mut rng = thread_rng();
let slice_a = rng.gen_iter::<f32>().take(100).collect::<Vec<f32>>();
let slice_b = rng.gen_iter::<f32>().take(100).collect::<Vec<f32>>();
let slice_a = rng.gen_iter::<f32>().take(n).collect::<Vec<f32>>();
let slice_b = rng.gen_iter::<f32>().take(n).collect::<Vec<f32>>();

let backend = backend();
let shared_a = &mut SharedTensor::<f32>::new(backend.device(), &100).unwrap();
let shared_b = &mut SharedTensor::<f32>::new(backend.device(), &100).unwrap();
let shared_res = &mut SharedTensor::<f32>::new(backend.device(), &()).unwrap();
shared_a.get_mut(backend.device()).unwrap().as_mut_native().unwrap().as_mut_slice().clone_from_slice(&slice_a);
shared_b.get_mut(backend.device()).unwrap().as_mut_native().unwrap().as_mut_slice().clone_from_slice(&slice_b);
let shared_a = &mut SharedTensor::<f32>::new(&[n]);
let shared_b = &mut SharedTensor::<f32>::new(&[n]);
let shared_res = &mut SharedTensor::<f32>::new(&[1]);
shared_a.write_only(backend.device()).unwrap().as_mut_native().unwrap()
.as_mut_slice().clone_from_slice(&slice_a);
shared_b.write_only(backend.device()).unwrap().as_mut_native().unwrap()
.as_mut_slice().clone_from_slice(&slice_b);
let _ = backend.dot(shared_a, shared_b, shared_res);
bench_1000_dot_100_collenchyma_profile(b, &backend, shared_a, shared_b, shared_res);
}

#[inline(never)]
fn bench_1000_dot_100_collenchyma_profile(
b: &mut Bencher,
backend: &Backend<Native>,
shared_a: &mut SharedTensor<f32>,
shared_b: &mut SharedTensor<f32>,
shared_res: &mut SharedTensor<f32>
) {
b.iter(|| {
for _ in 0..1000 {
let _ = backend.dot(shared_a, shared_b, shared_res);
}
});
b.iter(|| backend.dot(shared_a, shared_b, shared_res).unwrap());
}

#[bench]
fn bench_1000_dot_100_collenchyma_plain(b: &mut Bencher) {
let mut rng = thread_rng();
let slice_a = rng.gen_iter::<f32>().take(100).collect::<Vec<f32>>();
let slice_b = rng.gen_iter::<f32>().take(100).collect::<Vec<f32>>();

let backend = backend();
let shared_a = &mut SharedTensor::<f32>::new(backend.device(), &100).unwrap();
let shared_b = &mut SharedTensor::<f32>::new(backend.device(), &100).unwrap();
let shared_res = &mut SharedTensor::<f32>::new(backend.device(), &()).unwrap();
shared_a.get_mut(backend.device()).unwrap().as_mut_native().unwrap().as_mut_slice().clone_from_slice(&slice_a);
shared_b.get_mut(backend.device()).unwrap().as_mut_native().unwrap().as_mut_slice().clone_from_slice(&slice_b);
let _ = backend.dot(shared_a, shared_b, shared_res);
bench_1000_dot_100_collenchyma_plain_profile(b, &backend, shared_a, shared_b, shared_res);
}

#[inline(never)]
fn bench_1000_dot_100_collenchyma_plain_profile(
b: &mut Bencher,
backend: &Backend<Native>,
shared_a: &mut SharedTensor<f32>,
shared_b: &mut SharedTensor<f32>,
shared_res: &mut SharedTensor<f32>
) {
b.iter(|| {
for _ in 0..1000 {
let _ = backend.dot_plain(shared_a, shared_b, shared_res);
}
});
}

#[bench]
fn bench_100_dot_1000_rblas(b: &mut Bencher) {
let mut rng = thread_rng();
let slice_a = rng.gen_iter::<f32>().take(1000).collect::<Vec<f32>>();
let slice_b = rng.gen_iter::<f32>().take(1000).collect::<Vec<f32>>();

b.iter(|| {
for _ in 0..100 {
let res = rblas::Dot::dot(&slice_a, &slice_b);
test::black_box(res);
}
});
}
fn bench_dot_100_rblas(b: &mut Bencher) { bench_dot_rblas(b, 100); }

#[bench]
fn bench_100_dot_1000_collenchyma(b: &mut Bencher) {
let mut rng = thread_rng();
let slice_a = rng.gen_iter::<f32>().take(1000).collect::<Vec<f32>>();
let slice_b = rng.gen_iter::<f32>().take(1000).collect::<Vec<f32>>();

let backend = backend();
let shared_a = &mut SharedTensor::<f32>::new(backend.device(), &1000).unwrap();
let shared_b = &mut SharedTensor::<f32>::new(backend.device(), &1000).unwrap();
let shared_res = &mut SharedTensor::<f32>::new(backend.device(), &()).unwrap();
shared_a.get_mut(backend.device()).unwrap().as_mut_native().unwrap().as_mut_slice().clone_from_slice(&slice_a);
shared_b.get_mut(backend.device()).unwrap().as_mut_native().unwrap().as_mut_slice().clone_from_slice(&slice_b);
let _ = backend.dot(shared_a, shared_b, shared_res);
bench_100_dot_1000_collenchyma_profile(b, &backend, shared_a, shared_b, shared_res);
}

#[inline(never)]
fn bench_100_dot_1000_collenchyma_profile(
b: &mut Bencher,
backend: &Backend<Native>,
shared_a: &mut SharedTensor<f32>,
shared_b: &mut SharedTensor<f32>,
shared_res: &mut SharedTensor<f32>
) {
b.iter(|| {
for _ in 0..100 {
let _ = backend.dot(shared_a, shared_b, shared_res);
}
});
}
fn bench_dot_100_collenchyma(b: &mut Bencher) { bench_dot_collenchyma(b, 100); }

#[bench]
fn bench_50_dot_2000_collenchyma(b: &mut Bencher) {
let mut rng = thread_rng();
let slice_a = rng.gen_iter::<f32>().take(2000).collect::<Vec<f32>>();
let slice_b = rng.gen_iter::<f32>().take(2000).collect::<Vec<f32>>();

let backend = backend();
let shared_a = &mut SharedTensor::<f32>::new(backend.device(), &2000).unwrap();
let shared_b = &mut SharedTensor::<f32>::new(backend.device(), &2000).unwrap();
let shared_res = &mut SharedTensor::<f32>::new(backend.device(), &()).unwrap();
shared_a.get_mut(backend.device()).unwrap().as_mut_native().unwrap().as_mut_slice().clone_from_slice(&slice_a);
shared_b.get_mut(backend.device()).unwrap().as_mut_native().unwrap().as_mut_slice().clone_from_slice(&slice_b);
let _ = backend.dot(shared_a, shared_b, shared_res);
bench_50_dot_2000_collenchyma_profile(b, &backend, shared_a, shared_b, shared_res);
}

#[inline(never)]
fn bench_50_dot_2000_collenchyma_profile(
b: &mut Bencher,
backend: &Backend<Native>,
shared_a: &mut SharedTensor<f32>,
shared_b: &mut SharedTensor<f32>,
shared_res: &mut SharedTensor<f32>
) {
b.iter(|| {
for _ in 0..50 {
let _ = backend.dot(shared_a, shared_b, shared_res);
}
});
}
fn bench_dot_1000_rblas(b: &mut Bencher) { bench_dot_rblas(b, 1000); }

#[bench]
fn bench_10_dot_10000_rblas(b: &mut Bencher) {
let mut rng = thread_rng();
let slice_a = rng.gen_iter::<f32>().take(10000).collect::<Vec<f32>>();
let slice_b = rng.gen_iter::<f32>().take(10000).collect::<Vec<f32>>();

b.iter(|| {
for _ in 0..10 {
let res = rblas::Dot::dot(&slice_a, &slice_b);
test::black_box(res);
}
});
}
fn bench_dot_1000_collenchyma(b: &mut Bencher) { bench_dot_collenchyma(b, 1000); }

#[bench]
fn bench_10_dot_10000_collenchyma(b: &mut Bencher) {
let mut rng = thread_rng();
let slice_a = rng.gen_iter::<f32>().take(10000).collect::<Vec<f32>>();
let slice_b = rng.gen_iter::<f32>().take(10000).collect::<Vec<f32>>();

let backend = backend();
let shared_a = &mut SharedTensor::<f32>::new(backend.device(), &10000).unwrap();
let shared_b = &mut SharedTensor::<f32>::new(backend.device(), &10000).unwrap();
let shared_res = &mut SharedTensor::<f32>::new(backend.device(), &()).unwrap();
shared_a.get_mut(backend.device()).unwrap().as_mut_native().unwrap().as_mut_slice().clone_from_slice(&slice_a);
shared_b.get_mut(backend.device()).unwrap().as_mut_native().unwrap().as_mut_slice().clone_from_slice(&slice_b);
let _ = backend.dot(shared_a, shared_b, shared_res);
bench_10_dot_10000_collenchyma_profile(b, &backend, shared_a, shared_b, shared_res);
}

#[inline(never)]
fn bench_10_dot_10000_collenchyma_profile(
b: &mut Bencher,
backend: &Backend<Native>,
shared_a: &mut SharedTensor<f32>,
shared_b: &mut SharedTensor<f32>,
shared_res: &mut SharedTensor<f32>
) {
b.iter(|| {
for _ in 0..10 {
let _ = backend.dot(shared_a, shared_b, shared_res);
}
});
}
fn bench_dot_2000_rblas(b: &mut Bencher) { bench_dot_rblas(b, 2000); }

#[bench]
fn bench_5_dot_20000_rblas(b: &mut Bencher) {
let mut rng = thread_rng();
let slice_a = rng.gen_iter::<f32>().take(20000).collect::<Vec<f32>>();
let slice_b = rng.gen_iter::<f32>().take(20000).collect::<Vec<f32>>();
fn bench_dot_2000_collenchyma(b: &mut Bencher) { bench_dot_collenchyma(b, 2000); }

b.iter(|| {
for _ in 0..5 {
let res = rblas::Dot::dot(&slice_a, &slice_b);
test::black_box(res);
}
});
}
#[bench]
fn bench_dot_10000_rblas(b: &mut Bencher) { bench_dot_rblas(b, 10000); }

#[bench]
fn bench_5_dot_20000_collenchyma(b: &mut Bencher) {
let mut rng = thread_rng();
let slice_a = rng.gen_iter::<f32>().take(20000).collect::<Vec<f32>>();
let slice_b = rng.gen_iter::<f32>().take(20000).collect::<Vec<f32>>();
fn bench_dot_10000_collenchyma(b: &mut Bencher) { bench_dot_collenchyma(b, 10000); }

let backend = backend();
let shared_a = &mut SharedTensor::<f32>::new(backend.device(), &20000).unwrap();
let shared_b = &mut SharedTensor::<f32>::new(backend.device(), &20000).unwrap();
let shared_res = &mut SharedTensor::<f32>::new(backend.device(), &()).unwrap();
shared_a.get_mut(backend.device()).unwrap().as_mut_native().unwrap().as_mut_slice().clone_from_slice(&slice_a);
shared_b.get_mut(backend.device()).unwrap().as_mut_native().unwrap().as_mut_slice().clone_from_slice(&slice_b);
let _ = backend.dot(shared_a, shared_b, shared_res);
bench_5_dot_20000_collenchyma_profile(b, &backend, shared_a, shared_b, shared_res);
}
#[bench]
fn bench_dot_20000_rblas(b: &mut Bencher) { bench_dot_rblas(b, 20000); }

#[inline(never)]
fn bench_5_dot_20000_collenchyma_profile(
b: &mut Bencher,
backend: &Backend<Native>,
shared_a: &mut SharedTensor<f32>,
shared_b: &mut SharedTensor<f32>,
shared_res: &mut SharedTensor<f32>
) {
b.iter(|| {
for _ in 0..5 {
let _ = backend.dot(shared_a, shared_b, shared_res);
}
});
}
#[bench]
fn bench_dot_20000_collenchyma(b: &mut Bencher) { bench_dot_collenchyma(b, 20000); }
Loading