more fixes

Src/Base/AMReX_GpuLaunch.nolint.H

@@ -123,6 +123,16 @@
         ); \
     }
 
+#if defined(AMREX_USE_CUDA) && defined(_WIN32)
+#define AMREX_LAUNCH_HOST_DEVICE_LAMBDA_FLAG(where_to_run,box,tbox,block) \
+    if ((where_to_run == RunOn::Device) && (Gpu::inLaunchRegion())) \
+    { \
+        amrex::launch(box, tbox, [=] AMREX_GPU_DEVICE () { block }); \
+    } else { \
+        auto tbox = box; \
+        block; \
+    }
+#else
 #define AMREX_LAUNCH_HOST_DEVICE_LAMBDA_FLAG(where_to_run,box,tbox,block) \
     if ((where_to_run == RunOn::Device) && (Gpu::inLaunchRegion())) \
     { \
@@ -131,6 +141,7 @@
         auto tbox = box; \
         block; \
     }
+#endif
 
 #define AMREX_LAUNCH_HOST_DEVICE_LAMBDA_RANGE_FLAG(where_to_run,bx1,tbx1,block1) \
     if ((where_to_run == RunOn::Device) && (Gpu::inLaunchRegion())) \

Src/FFT/AMReX_FFT_R2C.H

@@ -672,7 +672,11 @@ void R2C<T,D,S>::post_forward_doit (F const& post_forward)
         amrex::Abort("xxxxx todo: post_forward");
 #if (AMREX_SPACEDIM > 1)
     } else if (m_r2c_sub) {
+#if defined(AMREX_USE_CUDA) && defined(_WIN32)
+        if           (Depth == 0) {
+#else
         if constexpr (Depth == 0) {
+#endif
             // We need to pass the originally ordered indices to post_forward.
 #if (AMREX_SPACEDIM == 2)
             // The original domain is (1,ny). The sub domain is (ny,1).

Src/FFT/AMReX_FFT_R2X.H

@@ -565,7 +565,11 @@ void R2X<T>::forwardThenBackward_doit (MF const& inmf, MF& outmf,
     BL_PROFILE("FFT::R2X::forwardbackward");
 
     if (m_r2x_sub) {
+#if defined(AMREX_USE_CUDA) && defined(_WIN32)
+        if           (Depth == 0)
+#else
         if constexpr (Depth == 0)
+#endif
         {
             bool inmf_safe = m_sub_helper.ghost_safe(inmf.nGrowVect());
             MF inmf_sub, inmf_tmp;

Src/LinearSolvers/MLMG/AMReX_MLABecLaplacian.H

@@ -673,20 +673,27 @@ MLABecLaplacianT<MF>::averageDownCoeffsSameAmrLevel (int amrlev, Vector<MF>& a,
                              auto const& bz = b[mglev][2].array(mfi));
                 auto const& osm = this->m_overset_mask[amrlev][mglev]->const_array(mfi);
 #if defined(AMREX_USE_CUDA) && defined(_WIN32)
-                amrex::launch(xbx, [=] AMREX_GPU_DEVICE (Box const& t_xbx)
-                {
-                    overset_rescale_bcoef_x(t_xbx, bx, osm, ncomp, osfac);
-                });
-                amrex::launch(ybx, [=] AMREX_GPU_DEVICE (Box const& t_ybx)
-                {
-                    overset_rescale_bcoef_y(t_ybx, by, osm, ncomp, osfac);
-                });
-                amrex::launch(zbx, [=] AMREX_GPU_DEVICE (Box const& t_zbx)
-                {
-                    overset_rescale_bcoef_z(t_zbx, bz, osm, ncomp, osfac);
-                });
+                AMREX_GPU_LAUNCH_HOST_DEVICE_LAMBDA_RANGE
+                    (xbx, t_xbx,
+                     {
+                         overset_rescale_bcoef_x(t_xbx, bx, osm, ncomp, osfac);
+                     });
+#if (AMREX_SPACEDIM >= 2)
+                AMREX_GPU_LAUNCH_HOST_DEVICE_LAMBDA_RANGE
+                    (ybx, t_ybx,
+                     {
+                         overset_rescale_bcoef_y(t_ybx, by, osm, ncomp, osfac);
+                     });
+#endif
+#if (AMREX_SPACEDIM == 3)
+                AMREX_GPU_LAUNCH_HOST_DEVICE_LAMBDA_RANGE
+                    (zbx, t_zbx,
+                     {
+                         overset_rescale_bcoef_z(t_zbx, bz, osm, ncomp, osfac);
+                     });
+#endif
 #else
-                AMREX_LAUNCH_HOST_DEVICE_LAMBDA_DIM
+                AMREX_GPU_LAUNCH_HOST_DEVICE_LAMBDA_RANGE_DIM
                     (xbx, t_xbx,
                      {
                          overset_rescale_bcoef_x(t_xbx, bx, osm, ncomp, osfac);
@@ -1179,15 +1186,15 @@ MLABecLaplacianT<MF>::FFlux (Box const& box, Real const* dxinv, RT bscalar,
         RT fac = bscalar*static_cast<RT>(dxinv[0]);
         Box blo = amrex::bdryLo(box, 0);
         int blen = box.length(0);
-        AMREX_LAUNCH_HOST_DEVICE_LAMBDA ( blo, tbox,
+        AMREX_GPU_LAUNCH_HOST_DEVICE_LAMBDA_RANGE ( blo, tbox,
         {
             mlabeclap_flux_xface(tbox, fxarr, solarr, bx, fac, blen, ncomp);
         });
 #if (AMREX_SPACEDIM >= 2)
         fac = bscalar*static_cast<RT>(dxinv[1]);
         blo = amrex::bdryLo(box, 1);
         blen = box.length(1);
-        AMREX_LAUNCH_HOST_DEVICE_LAMBDA ( blo, tbox,
+        AMREX_GPU_LAUNCH_HOST_DEVICE_LAMBDA_RANGE ( blo, tbox,
         {
             mlabeclap_flux_yface(tbox, fyarr, solarr, by, fac, blen, ncomp);
         });
@@ -1196,7 +1203,7 @@ MLABecLaplacianT<MF>::FFlux (Box const& box, Real const* dxinv, RT bscalar,
         fac = bscalar*static_cast<RT>(dxinv[2]);
         blo = amrex::bdryLo(box, 2);
         blen = box.length(2);
-        AMREX_LAUNCH_HOST_DEVICE_LAMBDA ( blo, tbox,
+        AMREX_GPU_LAUNCH_HOST_DEVICE_LAMBDA_RANGE ( blo, tbox,
         {
             mlabeclap_flux_zface(tbox, fzarr, solarr, bz, fac, blen, ncomp);
         });
@@ -1206,22 +1213,22 @@ MLABecLaplacianT<MF>::FFlux (Box const& box, Real const* dxinv, RT bscalar,
     {
         RT fac = bscalar*static_cast<RT>(dxinv[0]);
         Box bflux = amrex::surroundingNodes(box, 0);
-        AMREX_LAUNCH_HOST_DEVICE_LAMBDA ( bflux, tbox,
+        AMREX_GPU_LAUNCH_HOST_DEVICE_LAMBDA_RANGE ( bflux, tbox,
         {
             mlabeclap_flux_x(tbox, fxarr, solarr, bx, fac, ncomp);
         });
 #if (AMREX_SPACEDIM >= 2)
         fac = bscalar*static_cast<RT>(dxinv[1]);
         bflux = amrex::surroundingNodes(box, 1);
-        AMREX_LAUNCH_HOST_DEVICE_LAMBDA ( bflux, tbox,
+        AMREX_GPU_LAUNCH_HOST_DEVICE_LAMBDA_RANGE ( bflux, tbox,
         {
             mlabeclap_flux_y(tbox, fyarr, solarr, by, fac, ncomp);
         });
 #endif
 #if (AMREX_SPACEDIM == 3)
         fac = bscalar*static_cast<RT>(dxinv[2]);
         bflux = amrex::surroundingNodes(box, 2);
-        AMREX_LAUNCH_HOST_DEVICE_LAMBDA ( bflux, tbox,
+        AMREX_GPU_LAUNCH_HOST_DEVICE_LAMBDA_RANGE ( bflux, tbox,
         {
             mlabeclap_flux_z(tbox, fzarr, solarr, bz, fac, ncomp);
         });