add kerevalmeth=0 vs 1 acc check to CI via test/finufft_check.sh; doc in CHANGELOG

Author: ahbarnett

CHANGELOG

@@ -7,6 +7,7 @@ V 2.3.0beta (6/21/24)
   kernel evaluation, templating by ns with AVX-width-dependent decisions.
   Up to 80% faster, dep on compiler. (Marco Barbone with help from Libin Lu).
     NOTE: introduces new dependency (XSIMD), added to cMake and makefile.
+* new test/finufft3dkernel_test checks kerevalmeth=0,1 same to tol (M Barbone).
 * new perftest/compare_spreads.jl compares two spreadinterp libs (A Barnett).
 * new benchmarker perftest/spreadtestndall sweeps all kernel widths (M Barbone).
 * cufinufft now supports modeord(type 1,2 only): 0 CMCL-style increasing mode
@@ -26,15 +27,15 @@ V 2.3.0beta (6/21/24)
 * improved GPU python docs: migration guide; usage from cupy, numba, torch,
   pycuda. PyPI pkg still at 2.2.0beta.
 * Added a clang-format pre-commit hook to ensure consistent code style.
-  Created a .clang-format file to define the style similar to the existing style.
+  Created a .clang-format file to define a style similar to the existing style.
   Applied clang-format to all cmake, C, C++, and CUDA code. Ignored the blame
   using .git-blame-ignore-revs. Added a contributing.md for developers.
-* cuFINUFFT interface update: number of nonuniform points M is now a 64-bit integer
-as opposed to 32-bit. While this does modify the ABI, most code will just need to
-recompile against the new library as compilers will silently upcast any 32-bit
-integers to 64-bit when calling cufinufft(f)_setpts. Note that internally, 32-bit
-integers are still used, so calling cufinufft with more than 2e9 points will fail.
-This restriction may be lifted in the future.
+* cuFINUFFT interface update: number of nonuniform points M is now a 64-bit int
+  as opposed to 32-bit. While this does modify the ABI, most code will just
+  need to recompile against the new library as compilers will silently upcast
+  any 32-bit integers to 64-bit when calling cufinufft(f)_setpts. Note that
+  internally, 32-bit integers are still used, so calling cufinufft with more
+  than 2e9 points will fail. This restriction may be lifted in the future.
 
 V 2.2.0 (12/12/23)
 
@@ -52,7 +53,7 @@ V 2.2.0 (12/12/23)
 * CMake build structure (thanks: Wenda Zhou, Marco Barbone, Libin Lu)
   - Note: the plan is to continue to support GNU makefile and make.inc.* but
     to transition to CMake as the main build system.
-  - CI workflow using CMake on 3 OSes, 2 compilers each, PR #382 (Libin Lu)	
+  - CI workflow using CMake on 3 OSes, 2 compilers each, PR #382 (Libin Lu)
 * Docs: new tutorial content on iterative inverse NUFFTs; troubleshooting.
 * GitHub-facing badges
 * include/finufft/finufft_eitherprec.h moved up directory to be public (bea316c)

test/README

@@ -9,7 +9,7 @@ OpenMP static scheduling for rand-# generation means that the test data should
 be reproducible (non-stochastic). Reordering of thread ops in FINUFFT itself
 leads to machine-rounding sized variations only.
 
-These CPU test executables have suffix "f" fior single precision, else double.
+These CPU test executables have suffix "f" for single precision, else double.
 The source codes do not have the suffix:
 
 basicpassfail{f} : basic double and single-prec smoke tests of the math.
@@ -22,13 +22,8 @@ finufft{1,2,3}dmany_test{f}: accuracy/speed tests for vectorized transforms,
                    in a given dimension. Types 1, 2, and 3 are tested.
                    (exit code 0 is a pass).
                    Call with no arguments for argument documentation.
-finufft3dkernel_test{f} : test of the kernel evaluation methods.
-                   Only 3D NUFFT are tested.
-                   Types 1, 2, and 3 are tested.
-                   It requires Nmodes1 Nmodes2 Nmodes3 Nsrc
-                   sizes of the 3d grid and the number of sources.
-                   Optional arguments are the tolerance, debug flags,
-                   sort flag and upsampfac.
+finufft3dkernel_test{f} : test kerevalmeth=0,1 give same answer within tol.
+                   Types 1, 2, and 3 are tested, in d=3 only.
                    (exit code 0 is a pass).
 dumbinputs{f} :    test of edge cases, invalid inputs, and plan interface.
                    No arguments needed (exit code 0 is a pass).

test/check_finufft.sh

@@ -14,7 +14,7 @@
 
 # Barnett 3/14/17. numdiff-free option 3/16/17. simpler, dual-prec 7/3/20,
 # execs now have exit codes, removed any numdiff dep 8/18/20
-# removed diff 6/16/23.
+# removed diff 6/16/23. Added kerevalmeth=0 vs 1 test 7/8/24.
 
 # precision-specific settings
 if [[ $1 == "SINGLE" ]]; then
@@ -93,6 +93,12 @@ T=finufft3dmany_test$PRECSUF
 E=${PIPESTATUS[0]}
 if [[ $E -eq 0 ]]; then echo passed; elif [[ $E -eq $SIGSEGV ]]; then echo crashed; ((CRASHES++)); else echo failed; ((FAILS++)); fi
 
+((N++))
+T=finufft3dkernel_test$PRECSUF
+./$T$FEX 20 50 30 1e3 $FINUFFT_REQ_TOL 2>$DIR/$T.err.out | tee $DIR/$T.out
+E=${PIPESTATUS[0]}
+if [[ $E -eq 0 ]]; then echo passed; elif [[ $E -eq $SIGSEGV ]]; then echo crashed; ((CRASHES++)); else echo failed; ((FAILS++)); fi
+
 ((N++))
 T=dumbinputs$PRECSUF
 ./$T$FEX 2>$DIR/$T.err.out | tee $DIR/$T.out

test/finufft3dkernel_test.cpp

@@ -5,7 +5,7 @@ using namespace std;
 using namespace finufft::utils;
 
 const char *help[] = {
-    "Tester for FINUFFT in 3d, all 3 types, either precision.",
+    "Test spread_kerevalmeth=0 & 1 match, for 3 types of 3D transf, either prec.",
     "Usage: finufft3dkernel_test Nmodes1 Nmodes2 Nmodes3 Nsrc",
     "\t[tol] error tolerance (default 1e-6)",
     "\t[debug] (default 0) 0: silent, 1: text, 2: as 1 but also spreader",