address review comments

makefile

@@ -175,10 +175,10 @@ usage:
 	@echo "Also see docs/install.rst and docs/README"
 
 # collect headers for implicit depends (we don't separate public from private)
-HEADERS = $(wildcard include/*.h include/finufft/*.h) $(DUCC_HEADERS)
+HEADERS = $(wildcard include/*.h include/finufft/*.h)
 
 # implicit rules for objects (note -o ensures writes to correct dir)
-%.o: %.cpp $(HEADERS) $(XSIMD_DIR)/include/xsimd/xsimd.hpp
+%.o: %.cpp $(HEADERS)
 	$(CXX) -c $(CXXFLAGS) $< -o $@
 %.o: %.c $(HEADERS)
 	$(CC) -c $(CFLAGS) $< -o $@
@@ -194,6 +194,9 @@ include/finufft/fft.h: $(DUCC_SETUP)
 SHEAD = $(wildcard src/*.h) $(XSIMD_DIR)/include/xsimd/xsimd.hpp
 src/spreadinterp.o: $(SHEAD)
 
+# we need xsimd functionality in finufft_core.h, which is included by many other
+# files, so make sure we install xsimd before we prcess any of those files.
+include/finufft/finufft_core.h: $(XSIMD_DIR)/include/xsimd/xsimd.hpp
 
 # lib -----------------------------------------------------------------------
 # build library with double/single prec both bundled in...

src/finufft_core.cpp

@@ -548,7 +548,7 @@ FINUFFT_PLAN_T<TF>::FINUFFT_PLAN_T(int type_, int dim_, const BIGINT *n_modes, i
   if (!opts_)      // use default opts
     finufft_default_opts_t(&opts);
   else             // or read from what's passed in
-    opts = *opts_; // keep a deep copy; changing *opts now has no effect
+    opts = *opts_; // keep a deep copy; changing *opts_ now has no effect
 
   if (opts.debug)  // do a hello world
     printf("[%s] new plan: FINUFFT version " FINUFFT_VER " .................\n",
@@ -907,9 +907,9 @@ int FINUFFT_PLAN_T<TF>::setpts(BIGINT nj, TF *xj, TF *yj, TF *zj, BIGINT nk, TF
       }
     } else
       for (BIGINT j = 0; j < nj; ++j)
-        prephase[j] = {1, 0}; // *** or keep flag so no mult in exec??
+        prephase[j] = {1.0, 0.0}; // *** or keep flag so no mult in exec??
 
-                              // rescale the target s_k etc to s'_k etc...
+                                  // rescale the target s_k etc to s'_k etc...
 #pragma omp parallel for num_threads(opts.nthreads) schedule(static)
     for (BIGINT k = 0; k < nk; ++k) {
       Sp[k] = t3P.h1 * t3P.gam1 * (s[k] - t3P.D1);   // so |s'_k| < pi/R