Refactor and add CUDA JIT compilation to voigt_profile()
#113
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
smokestacklightnin
requested review from
wkerzendorf,
andrewfullard,
jaladh-singhal,
jvshields,
isaacgsmith,
epassaro and
atharva-2001
Codecov Report
@@ Coverage Diff @@
## main #113 +/- ##
==========================================
- Coverage 72.87% 69.37% -3.50%
==========================================
Files 21 21
Lines 693 738 +45
==========================================
+ Hits 505 512 +7
- Misses 188 226 +38
📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more |
smokestacklightnin
force-pushed
the
gpu/opacities/voigt-profile/add-cuda
branch
3 times, most recently
from
dde8a02
to
79f9446
Compare
smokestacklightnin
force-pushed
the
gpu/opacities/voigt-profile/add-cuda
branch
from
da68531
to
42dbca0
Compare
smokestacklightnin
force-pushed
the
gpu/opacities/voigt-profile/add-cuda
branch
from
62f61d5
to
02cdb29
Compare
|
*beep* *bop* All benchmarks: All benchmarks:
before after ratio
[9c2da67c] [02cdb293]
13.4±0.04s 13.2±0.01s 0.98 run_stardis.BenchmarkRunStardis.time_run_stardis
|
jvshields
approved these changes
Aug 18, 2023
jvshields
added a commit
that referenced
this pull request
Aug 21, 2023
* setting up radiation field structure * further work moving structures around * hook up radiation field to run_stardis() * fix bug with output * removing copied files * add back voigt test * further clean up stardis/base.py * remove transport files * update documentation * Add CUDA GPU JIT compilation to `faddeeva()` (#107) * Add cuda import * Add cuda boilerplate for faddeeva * Add square root of pi as constant to scope of entire file * Add faddeeva_cuda test * Vectorize faddeeva * Refactor faddeeva to be vectorized and work with cuda * Clean up faddeeva_cuda test * Rename variables to keep with convention * Refactor faddeeva_gpu and include associated tests * Add functionality for more datatypes for faddeeva_cuda Also add associated tests * Clean up faddeeva_gpu tests by testing numpy and cuda array types separately * Optimize faddeeva function to be branchless * Use cupy arrays instead of numba.cuda arrays Also call faddeeva_cuda with prespecified numbers of blocks and threads * Only import cupy if the machine has GPUs available * Typecast input to complex * Size should be of output array * Return cupy array by default * Change name of blackbody function. Hook up RadiationField.source_function to raytrace. * add units to stardis outputs (#122) * add units to stardis outputs * explicitly declare units to F_lambda * Benchmark internal STARDIS functions (#85) * Fix posting comment * Try addding token as env var * Put env var at correct place * fix benchmark comment bug * Benchmark raytrace function * Correct path for config soft link * Increase timeout time for raytrace benchmark * Add code to benchmark calc_alpha_line_at_nu * Add push trigger for all branches temporarily * Group all the benchmarking code under one class * Add tracing_lambdas attribute to benchmark class * Update benchmark setup * Fix breaking benchmark pipeline (#128) * Refactor and add CUDA JIT compilation to `voigt_profile()` (#113) * Add pi as a constant with file scope * Use real attribute of datatype rather than numpy call * Vectorize voigt_profile and add tests * Change indenting to improve readability * Typecast inputs to float and remove tests that include complex inputs * Add cuda versions of voigt_profile along with associated tests * Return cupy array by default * Size should be of output array * Fix typo in mathematical formula for `voigt_profile()` * Fix unit tests * Add CUDA JIT to `calc_n_effective()` (#119) * Add basic test for `calc_doppler_width()` * Refactor `calc_doppler_width()` and add test for vectorized implementation * Typecast to float * Add unwrapped cuda implementation of doppler_width Also typecast all global constants to float * Add wrapped cuda implementation of calc_doppler_width * Return cupy array by default * Add test for `calc_n_effective()` * Typecast inputs and refactor * Add cuda implementations and tests for `calc_n_effective()` * Return cupy array by default * Revert changes to `voigt_profile()` formula (#131) * setting up radiation field structure * further work moving structures around * hook up radiation field to run_stardis() * fix bug with output * add back voigt test * removing copied files * further clean up stardis/base.py * remove transport files * update documentation * Change name of blackbody function. Hook up RadiationField.source_function to raytrace. * fix imports * fix broadening merge * apply black * change benchmark paths --------- Co-authored-by: smokestacklightnin <[email protected]> Co-authored-by: light2802 <[email protected]>
jvshields
added a commit
to jvshields/stardis
that referenced
this pull request
Aug 22, 2023
* setting up radiation field structure * further work moving structures around * hook up radiation field to run_stardis() * fix bug with output * removing copied files * add back voigt test * further clean up stardis/base.py * remove transport files * update documentation * Add CUDA GPU JIT compilation to `faddeeva()` (tardis-sn#107) * Add cuda import * Add cuda boilerplate for faddeeva * Add square root of pi as constant to scope of entire file * Add faddeeva_cuda test * Vectorize faddeeva * Refactor faddeeva to be vectorized and work with cuda * Clean up faddeeva_cuda test * Rename variables to keep with convention * Refactor faddeeva_gpu and include associated tests * Add functionality for more datatypes for faddeeva_cuda Also add associated tests * Clean up faddeeva_gpu tests by testing numpy and cuda array types separately * Optimize faddeeva function to be branchless * Use cupy arrays instead of numba.cuda arrays Also call faddeeva_cuda with prespecified numbers of blocks and threads * Only import cupy if the machine has GPUs available * Typecast input to complex * Size should be of output array * Return cupy array by default * Change name of blackbody function. Hook up RadiationField.source_function to raytrace. * add units to stardis outputs (tardis-sn#122) * add units to stardis outputs * explicitly declare units to F_lambda * Benchmark internal STARDIS functions (tardis-sn#85) * Fix posting comment * Try addding token as env var * Put env var at correct place * fix benchmark comment bug * Benchmark raytrace function * Correct path for config soft link * Increase timeout time for raytrace benchmark * Add code to benchmark calc_alpha_line_at_nu * Add push trigger for all branches temporarily * Group all the benchmarking code under one class * Add tracing_lambdas attribute to benchmark class * Update benchmark setup * Fix breaking benchmark pipeline (tardis-sn#128) * Refactor and add CUDA JIT compilation to `voigt_profile()` (tardis-sn#113) * Add pi as a constant with file scope * Use real attribute of datatype rather than numpy call * Vectorize voigt_profile and add tests * Change indenting to improve readability * Typecast inputs to float and remove tests that include complex inputs * Add cuda versions of voigt_profile along with associated tests * Return cupy array by default * Size should be of output array * Fix typo in mathematical formula for `voigt_profile()` * Fix unit tests * Add CUDA JIT to `calc_n_effective()` (tardis-sn#119) * Add basic test for `calc_doppler_width()` * Refactor `calc_doppler_width()` and add test for vectorized implementation * Typecast to float * Add unwrapped cuda implementation of doppler_width Also typecast all global constants to float * Add wrapped cuda implementation of calc_doppler_width * Return cupy array by default * Add test for `calc_n_effective()` * Typecast inputs and refactor * Add cuda implementations and tests for `calc_n_effective()` * Return cupy array by default * Revert changes to `voigt_profile()` formula (tardis-sn#131) * setting up radiation field structure * further work moving structures around * hook up radiation field to run_stardis() * fix bug with output * add back voigt test * removing copied files * further clean up stardis/base.py * remove transport files * update documentation * Change name of blackbody function. Hook up RadiationField.source_function to raytrace. * fix imports * fix broadening merge * apply black * change benchmark paths --------- Co-authored-by: smokestacklightnin <[email protected]> Co-authored-by: light2802 <[email protected]>
andrewfullard
pushed a commit
that referenced
this pull request
Sep 11, 2023
* Restructure: Add radiation field (#123) * setting up radiation field structure * further work moving structures around * hook up radiation field to run_stardis() * fix bug with output * removing copied files * add back voigt test * further clean up stardis/base.py * remove transport files * update documentation * Add CUDA GPU JIT compilation to `faddeeva()` (#107) * Add cuda import * Add cuda boilerplate for faddeeva * Add square root of pi as constant to scope of entire file * Add faddeeva_cuda test * Vectorize faddeeva * Refactor faddeeva to be vectorized and work with cuda * Clean up faddeeva_cuda test * Rename variables to keep with convention * Refactor faddeeva_gpu and include associated tests * Add functionality for more datatypes for faddeeva_cuda Also add associated tests * Clean up faddeeva_gpu tests by testing numpy and cuda array types separately * Optimize faddeeva function to be branchless * Use cupy arrays instead of numba.cuda arrays Also call faddeeva_cuda with prespecified numbers of blocks and threads * Only import cupy if the machine has GPUs available * Typecast input to complex * Size should be of output array * Return cupy array by default * Change name of blackbody function. Hook up RadiationField.source_function to raytrace. * add units to stardis outputs (#122) * add units to stardis outputs * explicitly declare units to F_lambda * Benchmark internal STARDIS functions (#85) * Fix posting comment * Try addding token as env var * Put env var at correct place * fix benchmark comment bug * Benchmark raytrace function * Correct path for config soft link * Increase timeout time for raytrace benchmark * Add code to benchmark calc_alpha_line_at_nu * Add push trigger for all branches temporarily * Group all the benchmarking code under one class * Add tracing_lambdas attribute to benchmark class * Update benchmark setup * Fix breaking benchmark pipeline (#128) * Refactor and add CUDA JIT compilation to `voigt_profile()` (#113) * Add pi as a constant with file scope * Use real attribute of datatype rather than numpy call * Vectorize voigt_profile and add tests * Change indenting to improve readability * Typecast inputs to float and remove tests that include complex inputs * Add cuda versions of voigt_profile along with associated tests * Return cupy array by default * Size should be of output array * Fix typo in mathematical formula for `voigt_profile()` * Fix unit tests * Add CUDA JIT to `calc_n_effective()` (#119) * Add basic test for `calc_doppler_width()` * Refactor `calc_doppler_width()` and add test for vectorized implementation * Typecast to float * Add unwrapped cuda implementation of doppler_width Also typecast all global constants to float * Add wrapped cuda implementation of calc_doppler_width * Return cupy array by default * Add test for `calc_n_effective()` * Typecast inputs and refactor * Add cuda implementations and tests for `calc_n_effective()` * Return cupy array by default * Revert changes to `voigt_profile()` formula (#131) * setting up radiation field structure * further work moving structures around * hook up radiation field to run_stardis() * fix bug with output * add back voigt test * removing copied files * further clean up stardis/base.py * remove transport files * update documentation * Change name of blackbody function. Hook up RadiationField.source_function to raytrace. * fix imports * fix broadening merge * apply black * change benchmark paths --------- Co-authored-by: smokestacklightnin <[email protected]> Co-authored-by: light2802 <[email protected]> * fix voigt test that got overwritten by the rebase * add black to voigt.py * include fixto voight profile from pr 131 * fix benchmarks * further fix of benchmarks * remove blackbody import comment in radiation_field_solvers/base.py --------- Co-authored-by: smokestacklightnin <[email protected]> Co-authored-by: light2802 <[email protected]>
smokestacklightnin
added a commit
to smokestacklightnin/stardis
that referenced
this pull request
Sep 20, 2023
…#113) * Add pi as a constant with file scope * Use real attribute of datatype rather than numpy call * Vectorize voigt_profile and add tests * Change indenting to improve readability * Typecast inputs to float and remove tests that include complex inputs * Add cuda versions of voigt_profile along with associated tests * Return cupy array by default * Size should be of output array * Fix typo in mathematical formula for `voigt_profile()` * Fix unit tests
smokestacklightnin
added a commit
to smokestacklightnin/stardis
that referenced
this pull request
Sep 20, 2023
* setting up radiation field structure * further work moving structures around * hook up radiation field to run_stardis() * fix bug with output * removing copied files * add back voigt test * further clean up stardis/base.py * remove transport files * update documentation * Add CUDA GPU JIT compilation to `faddeeva()` (tardis-sn#107) * Add cuda import * Add cuda boilerplate for faddeeva * Add square root of pi as constant to scope of entire file * Add faddeeva_cuda test * Vectorize faddeeva * Refactor faddeeva to be vectorized and work with cuda * Clean up faddeeva_cuda test * Rename variables to keep with convention * Refactor faddeeva_gpu and include associated tests * Add functionality for more datatypes for faddeeva_cuda Also add associated tests * Clean up faddeeva_gpu tests by testing numpy and cuda array types separately * Optimize faddeeva function to be branchless * Use cupy arrays instead of numba.cuda arrays Also call faddeeva_cuda with prespecified numbers of blocks and threads * Only import cupy if the machine has GPUs available * Typecast input to complex * Size should be of output array * Return cupy array by default * Change name of blackbody function. Hook up RadiationField.source_function to raytrace. * add units to stardis outputs (tardis-sn#122) * add units to stardis outputs * explicitly declare units to F_lambda * Benchmark internal STARDIS functions (tardis-sn#85) * Fix posting comment * Try addding token as env var * Put env var at correct place * fix benchmark comment bug * Benchmark raytrace function * Correct path for config soft link * Increase timeout time for raytrace benchmark * Add code to benchmark calc_alpha_line_at_nu * Add push trigger for all branches temporarily * Group all the benchmarking code under one class * Add tracing_lambdas attribute to benchmark class * Update benchmark setup * Fix breaking benchmark pipeline (tardis-sn#128) * Refactor and add CUDA JIT compilation to `voigt_profile()` (tardis-sn#113) * Add pi as a constant with file scope * Use real attribute of datatype rather than numpy call * Vectorize voigt_profile and add tests * Change indenting to improve readability * Typecast inputs to float and remove tests that include complex inputs * Add cuda versions of voigt_profile along with associated tests * Return cupy array by default * Size should be of output array * Fix typo in mathematical formula for `voigt_profile()` * Fix unit tests * Add CUDA JIT to `calc_n_effective()` (tardis-sn#119) * Add basic test for `calc_doppler_width()` * Refactor `calc_doppler_width()` and add test for vectorized implementation * Typecast to float * Add unwrapped cuda implementation of doppler_width Also typecast all global constants to float * Add wrapped cuda implementation of calc_doppler_width * Return cupy array by default * Add test for `calc_n_effective()` * Typecast inputs and refactor * Add cuda implementations and tests for `calc_n_effective()` * Return cupy array by default * Revert changes to `voigt_profile()` formula (tardis-sn#131) * setting up radiation field structure * further work moving structures around * hook up radiation field to run_stardis() * fix bug with output * add back voigt test * removing copied files * further clean up stardis/base.py * remove transport files * update documentation * Change name of blackbody function. Hook up RadiationField.source_function to raytrace. * fix imports * fix broadening merge * apply black * change benchmark paths --------- Co-authored-by: smokestacklightnin <[email protected]> Co-authored-by: light2802 <[email protected]>
smokestacklightnin
added a commit
to smokestacklightnin/stardis
that referenced
this pull request
Sep 20, 2023
* Restructure: Add radiation field (tardis-sn#123) * setting up radiation field structure * further work moving structures around * hook up radiation field to run_stardis() * fix bug with output * removing copied files * add back voigt test * further clean up stardis/base.py * remove transport files * update documentation * Add CUDA GPU JIT compilation to `faddeeva()` (tardis-sn#107) * Add cuda import * Add cuda boilerplate for faddeeva * Add square root of pi as constant to scope of entire file * Add faddeeva_cuda test * Vectorize faddeeva * Refactor faddeeva to be vectorized and work with cuda * Clean up faddeeva_cuda test * Rename variables to keep with convention * Refactor faddeeva_gpu and include associated tests * Add functionality for more datatypes for faddeeva_cuda Also add associated tests * Clean up faddeeva_gpu tests by testing numpy and cuda array types separately * Optimize faddeeva function to be branchless * Use cupy arrays instead of numba.cuda arrays Also call faddeeva_cuda with prespecified numbers of blocks and threads * Only import cupy if the machine has GPUs available * Typecast input to complex * Size should be of output array * Return cupy array by default * Change name of blackbody function. Hook up RadiationField.source_function to raytrace. * add units to stardis outputs (tardis-sn#122) * add units to stardis outputs * explicitly declare units to F_lambda * Benchmark internal STARDIS functions (tardis-sn#85) * Fix posting comment * Try addding token as env var * Put env var at correct place * fix benchmark comment bug * Benchmark raytrace function * Correct path for config soft link * Increase timeout time for raytrace benchmark * Add code to benchmark calc_alpha_line_at_nu * Add push trigger for all branches temporarily * Group all the benchmarking code under one class * Add tracing_lambdas attribute to benchmark class * Update benchmark setup * Fix breaking benchmark pipeline (tardis-sn#128) * Refactor and add CUDA JIT compilation to `voigt_profile()` (tardis-sn#113) * Add pi as a constant with file scope * Use real attribute of datatype rather than numpy call * Vectorize voigt_profile and add tests * Change indenting to improve readability * Typecast inputs to float and remove tests that include complex inputs * Add cuda versions of voigt_profile along with associated tests * Return cupy array by default * Size should be of output array * Fix typo in mathematical formula for `voigt_profile()` * Fix unit tests * Add CUDA JIT to `calc_n_effective()` (tardis-sn#119) * Add basic test for `calc_doppler_width()` * Refactor `calc_doppler_width()` and add test for vectorized implementation * Typecast to float * Add unwrapped cuda implementation of doppler_width Also typecast all global constants to float * Add wrapped cuda implementation of calc_doppler_width * Return cupy array by default * Add test for `calc_n_effective()` * Typecast inputs and refactor * Add cuda implementations and tests for `calc_n_effective()` * Return cupy array by default * Revert changes to `voigt_profile()` formula (tardis-sn#131) * setting up radiation field structure * further work moving structures around * hook up radiation field to run_stardis() * fix bug with output * add back voigt test * removing copied files * further clean up stardis/base.py * remove transport files * update documentation * Change name of blackbody function. Hook up RadiationField.source_function to raytrace. * fix imports * fix broadening merge * apply black * change benchmark paths --------- Co-authored-by: smokestacklightnin <[email protected]> Co-authored-by: light2802 <[email protected]> * fix voigt test that got overwritten by the rebase * add black to voigt.py * include fixto voight profile from pr 131 * fix benchmarks * further fix of benchmarks * remove blackbody import comment in radiation_field_solvers/base.py --------- Co-authored-by: smokestacklightnin <[email protected]> Co-authored-by: light2802 <[email protected]>
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.
📝 Description
Type: 🚀
featureThis pull request refactors
voigt_profile()and adds CUDA JIT compilation to the same function. Basic unit tests are also included.This pull request should be merged (hopefully immediately) after pull request #107
🚦 Testing
How did you test these changes?
☑️ Checklist