rodinia/gaussian sycl and ndpx implementation#307
Conversation
adarshyoga
left a comment
There was a problem hiding this comment.
Overall looks in good shape!
dpbench/benchmarks/rodinia/gaussian/gaussian_sycl_native_ext/gaussian_sycl/_gaussian_kernel.hpp
Outdated
Show resolved
Hide resolved
dpbench/benchmarks/rodinia/gaussian/gaussian_sycl_native_ext/gaussian_sycl/_gaussian_kernel.hpp
Outdated
Show resolved
Hide resolved
dpbench/benchmarks/rodinia/gaussian/gaussian_sycl_native_ext/gaussian_sycl/_gaussian_sycl.cpp
Show resolved
Hide resolved
640d447 to
77fafac
Compare
These numbers look scary. 500X slowdown in numba_dpex over sycl. |
77fafac to
2ad66eb
Compare
|
@roxx30198 please split commits so one commit is responsible for infrastructure changes and another one is for the gaussian implementation And let's set the input the way dpex implementation does not exceed 1s for S input. @adarshyoga does input sizes for M and L sounds reasonable for you? |
0074445 to
1820e49
Compare
We would need to experiment with this workload to arrive at an appropriate M and L. I think we can merge this PR with the current values of M and L. We can change them as we do deeper analysis into this workload and Rodinia, in general. |
4179cf5 to
5bec653
Compare
5bec653 to
19ff9ca
Compare
19ff9ca to
0f52f5f
Compare
6855db3 to
263776f
Compare
001d060 to
2d7bb7c
Compare
0ba2a06 to
8da8d99
Compare
Added the sycl and numba-dpex implementation for gaussian elimination benchmark for rodinia.
Runtime for 100x100 matrix with fp64 on cpu:
sycl:
max execution times: 12.190916ms (12190916 ns)
min execution times: 6.433963ms (6433963 ns)
median execution times: 7.595746ms (7595746 ns)
repeats: 10
numba_dpex_k:
max execution times: 3.994273296s (3994273296 ns)
min execution times: 2.814416425s (2814416425 ns)
median execution times: 3.585554875s (3585554875 ns)
repeats: 10