Fix new compiler warning failures (#1199)

Author: cliffburdick

CMakeLists.txt

@@ -206,6 +206,10 @@ if (NOT ${IS_NVCPP} GREATER -1)
             -Wlogical-op
             -Wnull-dereference)
     endif()
+else()
+    # nvc++ diagnoses libcudacxx's if-consteval usage under all-warnings.
+    set(WARN_FLAGS ${WARN_FLAGS}
+        $<$<COMPILE_LANGUAGE:CUDA>:-Xcompiler=--diag_suppress,if_consteval_in_nonconstexpr_function>)
 endif()
 
 if (CMAKE_CUDA_COMPILER_ID STREQUAL "Clang")

include/matx/core/pybind.h

@@ -423,7 +423,7 @@ class MATX_PYBIND_VISIBILITY MatXPybind {
     // copy a smaller tensor from a larger NumPy array, so keep that path bounded.
     for (int d = 0; d < RANK; d++) {
       const auto np_size = static_cast<index_t>(info.shape[d]);
-      const bool valid_size = exact_shape ? ten.Size(d) == np_size : ten.Size(d) <= np_size;
+      [[maybe_unused]] const bool valid_size = exact_shape ? ten.Size(d) == np_size : ten.Size(d) <= np_size;
       MATX_ASSERT_STR(valid_size, matxInvalidSize,
                       "Numpy array dimension size is incompatible with tensor size");
     }

include/matx/kernels/sar_bp.cuh

@@ -651,11 +651,12 @@ __global__ void SarBp(OutImageType output, const InitialImageType initial_image,
     };
 
     const auto get_reference_phase_direct = [&phase_correction_partial](strict_compute_t range_delta) -> loose_complex_compute_t {
+        const strict_compute_t phase = static_cast<strict_compute_t>(phase_correction_partial * range_delta);
         strict_compute_t sinx, cosx;
         if constexpr (cuda::std::is_same_v<strict_compute_t, double>) {
-            ::sincos(phase_correction_partial * range_delta, &sinx, &cosx);
+            ::sincos(phase, &sinx, &cosx);
         } else {
-            ::sincosf(phase_correction_partial * range_delta, &sinx, &cosx);
+            ::sincosf(phase, &sinx, &cosx);
         }
         return loose_complex_compute_t{
             static_cast<loose_compute_t>(cosx), static_cast<loose_compute_t>(sinx)