Fh kokkos b spline by hsiehhRPI · Pull Request #3 · SCOREC/simLandIceMeshGen

hsiehhRPI · 2026-05-05T15:34:03Z

This branch included Kokkos version of the BSpline inplementation. The main file is BSplineKokkos2D.h which contains Kokkos version of the BSpline that utilize 2D view for control points. 1st and 2nd derivative evaluation are implemented.

…simLandIceMeshGen into fh_Kokkos_BSpline

cwsmith · 2026-05-11T15:34:53Z

@hsiehhRPI Was clang-format ran on this? I saw some long lines (>80 chars) that I would have expected it to break into multiple lines.

cwsmith

Please carefully review the comments on the rank 1 implementation/tests/etc and ensure that they are addressed in the rank 2 version if the same code exists there.

…sieh <hsiehh@checkers.scorec.rpi.edu>

…: Hsieh <hsiehh@checkers.scorec.rpi.edu>

…tter: Hsieh <hsiehh@checkers.scorec.rpi.edu>

…hh@checkers.scorec.rpi.edu>

…itter: Hsieh <hsiehh@checkers.scorec.rpi.edu>

…eh <hsiehh@checkers.scorec.rpi.edu>

…r: Hsieh <hsiehh@checkers.scorec.rpi.edu>

… test for GPU# Committer: Hsieh <hsiehh@checkers.scorec.rpi.edu>

…ieh <hsiehh@checkers.scorec.rpi.edu>

… Hsieh <hsiehh@checkers.scorec.rpi.edu>

…itter: Hsieh <hsiehh@checkers.scorec.rpi.edu>

…h cpu and gpu# Committer: Hsieh <hsiehh@checkers.scorec.rpi.edu>

…mespace# Committer: Hsieh <hsiehh@checkers.scorec.rpi.edu>

…views in derivative call functions Committer: Hsieh <hsiehh@checkers.scorec.rpi.edu>

…tion to simplify the input to the device functions # Committer: Hsieh <hsiehh@checkers.scorec.rpi.edu>

…ckers.scorec.rpi.edu>

…s # Committer: Hsieh <hsiehh@checkers.scorec.rpi.edu>

…s.scorec.rpi.edu>

hsiehhRPI · 2026-07-26T01:07:50Z

Hello Dr. Smith, I have ran clang-format and have a commit with only clang formatting and no logical changes. The line lengths should now be at most 80 characters.

cwsmith · 2026-08-07T13:18:00Z

+      const int order, const Kokkos::View<double *, MemSpace> knots,
+      const Kokkos::View<double *[2], MemSpace> ctrlPts_1stD) const {
+    // DeBoor's algorithm for BSpline 1st deriv calculation
+    const int MAX_DEGREE = 3;


this could be useful outside this function - where else is this hardcoded?

This was in both Kokkos De Boor's Device functions. I will move them outside.

When I move it outside of the function, should it be declared as a static constexpr inside the class or as a member variable?

cwsmith · 2026-08-07T13:19:22Z

+    const int MAX_DEGREE = 3;
+    int lKnot = order;
+    lKnot--;
+    int resultOrder = lKnot;


int resultOrder = order -1; would be much easier to read

cwsmith · 2026-08-07T13:21:23Z

    int leftPt = 0;
-
-    while (knots(lKnot+1) < x) {
+    // double x = xVals(offset);


please remove code that is commented out

cwsmith · 2026-08-07T13:22:44Z

    }
-
-    int order_t = lKnot-1;
+    const int MAX_DEGREE = 3;


move this definition to a place both functions can access it - defining constants in multiple locations is likely to cause problems later (i.e., someone changes one but forgets/doesn't know to change the other)

cwsmith · 2026-08-07T13:30:23Z

 #include <numeric>
+#include <string>

 double EPSILON = 1e-12;


Please make this const and move it to where it is used within the function.

This has been resolved.

cwsmith · 2026-08-07T13:31:01Z

 #include <numeric>
+#include <string>

 double EPSILON = 1e-12;


see comment on other test

This has been resolved.

cwsmith · 2026-08-07T13:36:50Z

+// For checking if the content of the splines are correct
 //

 double EPSILON = 1e-12;


same comment as other tests

This has been resolved.

cwsmith · 2026-08-07T13:40:43Z

    auto intView = Kokkos::create_mirror_view(kokkosBSP.getOrder());
    Kokkos::deep_copy(intView, kokkosBSP.getOrder());


use the kokkos function create_mirror_view_and_copy if you just want a host copy of the 'order' device array

This has been resolved.

cwsmith · 2026-08-07T13:42:05Z

    auto double2DView = Kokkos::create_mirror_view(kokkosBSP.getCtrlPts());
    Kokkos::deep_copy(double2DView, kokkosBSP.getCtrlPts());


same as above

This has been resolved.

cwsmith · 2026-08-07T13:42:14Z

    auto doubleView = Kokkos::create_mirror_view(kokkosBSP.getKnots());
    Kokkos::deep_copy(doubleView, kokkosBSP.getKnots());


same as above

This has been resolved.

cwsmith · 2026-08-07T13:43:16Z

same as above

This has been resolved.

Hsieh and others added 30 commits March 8, 2026 12:37

Attempt to create constructor for KokkosBSpline

9251c86

Attempt derivative coeff function, started eval

420575e

Attempt eval function, need compilation, testing

0f46eb0

Create init test for BSplineKokkos

06ae6f3

Add BSplineKokkos executable to CMakeLists.txt

7e97698

Fixed some syntax error and templating issues

c07c382

Fix compiling error

a6cfa5a

Link Kokkos::kokkos to testSplineKokkosBase

789f6bc

Fixed vector initialization error

333c669

Moving constructor from .cpp to .h

fbe5082

Moved functions use execution space to .h file

9822f38

Modified test case to suit the 2D structure

f87d67a

Fixed 1st, 2nd derivative coef function

004e5ee

Added constructor from a vector of kokkos splines

1b0c522

fixed 1 spline initialization and test

603aee7

Multispline constructor, offset problem in progress

09b8b83

Fixed ctrlpts, knots offset error

8030c59

First attempt on eval1stDerivative

79b1947

Still debugging derivative coef calculation error

defbba2

Fixing 1st deriv coe err 2nd deriv coe in progress

8e4b58c

Fixing2nd deriv coe, 1&2 deriv coe need to verify

d5707a8

1st attempt at eval2ndDeriv

935a31d

fixed indent & remove repeat knots, ongoing debug

dfafacb

fixing indentation to match existing code

f788662

fixing memory access error for spline constructors

bc562b9

Fixing print function

2b8d27c

fixed multi spline constructor offset index

e861781

Coeff function adapt with mirror views

865a5a3

eval1stD debug in progress

c794c9c

Add serial spline & init SplineKokkos w/ same data

2289339

Fix conflict & Merge branch 'fh_Kokkos_BSpline' of github.com:SCOREC/…

f09d1dd

…simLandIceMeshGen into fh_Kokkos_BSpline

cwsmith requested changes May 11, 2026

View reviewed changes

cwsmith reviewed May 11, 2026

View reviewed changes

Hsieh added 21 commits June 5, 2026 03:31

modified/deleted files according to pr review

53c10f6

Fixed 2 DeBoors naming and lKnot copying

012cbef

Update parallel_for name "eval DeBoor for _ deriv"

7c91403

pr fixes, deleted an unused SplineKokkos2D constructor # Committer: H…

7df648b

…sieh <hsiehh@checkers.scorec.rpi.edu>

changed order_t naming and set array size to MAX_ORDER + 1# Committer…

bf15e6d

…: Hsieh <hsiehh@checkers.scorec.rpi.edu>

Change both deriv device function's array size to max degree+1# Commi…

f60c944

…tter: Hsieh <hsiehh@checkers.scorec.rpi.edu>

Changed arr size from order + 1 to degree + 1# Committer: Hsieh <hsie…

738bf2f

…hh@checkers.scorec.rpi.edu>

Created prelim CSR, still working on implementing the functions# Comm…

b9b915a

…itter: Hsieh <hsiehh@checkers.scorec.rpi.edu>

Test file adjusted to the new batch derivative format# Committer: Hsi…

e5b0f23

…eh <hsiehh@checkers.scorec.rpi.edu>

Batch 1st derivative work in progress, does not pass serial# Committe…

98787ea

…r: Hsieh <hsiehh@checkers.scorec.rpi.edu>

Batch 1st derivative passes all test running with NATIVE(cpu ) yet to…

a13c022

… test for GPU# Committer: Hsieh <hsiehh@checkers.scorec.rpi.edu>

Batch 1st deriv CPU pass all test, GPU runs but errors# Committer: Hs…

8216795

…ieh <hsiehh@checkers.scorec.rpi.edu>

Batch 1st deriv passes all test cases for serial and gpu # Committer:…

ad5c6a2

… Hsieh <hsiehh@checkers.scorec.rpi.edu>

Test updated to take in batch 2nd derivative splines using csr.# Comm…

2a47b27

…itter: Hsieh <hsiehh@checkers.scorec.rpi.edu>

Updated 2nd derivative for batch evaluation, passes all tests for bot…

6571dd7

…h cpu and gpu# Committer: Hsieh <hsiehh@checkers.scorec.rpi.edu>

Adjusted syntax after CSR struct is wrapped in the KokkosBspline2D na…

ecd16ea

…mespace# Committer: Hsieh <hsiehh@checkers.scorec.rpi.edu>

Wrapped csr struct in BSplineKokkos2D namespace, working on using sub…

2146b41

…views in derivative call functions Committer: Hsieh <hsiehh@checkers.scorec.rpi.edu>

Passing parametric coordinate and order as subview to the device func…

ec1a0a4

…tion to simplify the input to the device functions # Committer: Hsieh <hsiehh@checkers.scorec.rpi.edu>

Deleted commented out print statements # Committer: Hsieh <hsiehh@che…

5a794ae

…ckers.scorec.rpi.edu>

Fixed some indentation error and delete commented out print statement…

c68f0c0

…s # Committer: Hsieh <hsiehh@checkers.scorec.rpi.edu>

Ran clang formatting on these files# Committer: Hsieh <hsiehh@checker…

19e52f9

…s.scorec.rpi.edu>

cwsmith reviewed Aug 7, 2026

View reviewed changes

cwsmith requested changes Aug 7, 2026

View reviewed changes

hsiehhRPI added 2 commits August 7, 2026 23:40

const EPSILON + create_mirror_view_and_copy

d365b09

FORMAT FIX ONLY: ran format fix

bb7fcc4

		auto intView = Kokkos::create_mirror_view(kokkosBSP.getOrder());
		Kokkos::deep_copy(intView, kokkosBSP.getOrder());

		auto double2DView = Kokkos::create_mirror_view(kokkosBSP.getCtrlPts());
		Kokkos::deep_copy(double2DView, kokkosBSP.getCtrlPts());

		auto doubleView = Kokkos::create_mirror_view(kokkosBSP.getKnots());
		Kokkos::deep_copy(doubleView, kokkosBSP.getKnots());

Uh oh!

Conversation

hsiehhRPI commented May 5, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

cwsmith commented May 11, 2026

Uh oh!

cwsmith left a comment

Choose a reason for hiding this comment

Uh oh!

hsiehhRPI commented Jul 26, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

hsiehhRPI Aug 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

hsiehhRPI Aug 8, 2026 •

edited

Loading