Simplify filterParticles Kernel (#3510)

## Summary On Summit, generation of this kernel shows compiler issues with `nvcc` 11.3.109: it compiles without warnings but leads to a segmentation fault at runtime. The fix for the compiler bug is to implement the trivial lambda that is passed to `copyParticles` w/o capture: `[]`. We also add explicit capture to a few other lambdas, to simplify compiler intake complexity. This simplifies the kernel generation, which also solves the issue seen with WarpX [for this line](https://github.com/ECP-WarpX/WarpX/blob/43d2ac7b54546c87d2cb540df7a2b3cb57592b84/Source/Diagnostics/WarpXOpenPMD.cpp#L585). Co-authored-by: AlexanderSinn <[email protected]>
AMReX-Codes · Aug 26, 2023 · d4761e8 · d4761e8
1 parent a45eade
commit d4761e8
Show file tree

Hide file tree

Showing 2 changed files with 5 additions and 5 deletions.
diff --git a/Src/Particle/AMReX_ParticleContainerI.H b/Src/Particle/AMReX_ParticleContainerI.H
@@ -991,7 +991,7 @@ ParticleContainer_impl<ParticleType, NArrayReal, NArrayInt, Allocator, CellAssig
 copyParticles (const PCType& other, bool local)
 {
     using PData = ConstParticleTileData<typename ParticleType::StorageParticleType, NArrayReal, NArrayInt>;
-    copyParticles(other, [=] AMREX_GPU_HOST_DEVICE (const PData& /*data*/, int /*i*/) { return 1; }, local);
+    copyParticles(other, [] AMREX_GPU_HOST_DEVICE (const PData& /*data*/, int /*i*/) { return 1; }, local);
 }
 
 template <typename ParticleType, int NArrayReal, int NArrayInt,
@@ -1002,7 +1002,7 @@ ParticleContainer_impl<ParticleType, NArrayReal, NArrayInt, Allocator, CellAssig
 addParticles (const PCType& other, bool local)
 {
     using PData = ConstParticleTileData<typename ParticleType::StorageParticleType, NArrayReal, NArrayInt>;
-    addParticles(other, [=] AMREX_GPU_HOST_DEVICE (const PData& /*data*/, int /*i*/) { return 1; }, local);
+    addParticles(other, [] AMREX_GPU_HOST_DEVICE (const PData& /*data*/, int /*i*/) { return 1; }, local);
 }
 
 template <typename ParticleType, int NArrayReal, int NArrayInt,

diff --git a/Src/Particle/AMReX_ParticleTransformation.H b/Src/Particle/AMReX_ParticleTransformation.H
@@ -404,7 +404,7 @@ Index filterParticles (DstTile& dst, const SrcTile& src, Pred&& p,
     const auto src_data = src.getConstParticleTileData();
 
     amrex::ParallelForRNG(n,
-    [=] AMREX_GPU_DEVICE (int i, amrex::RandomEngine const& engine) noexcept
+    [p, p_mask, src_data, src_start] AMREX_GPU_DEVICE (int i, amrex::RandomEngine const& engine) noexcept
     {
         amrex::ignore_unused(p, p_mask, src_data, src_start, engine);
         if constexpr (IsCallable<Pred,decltype(src_data),Index,RandomEngine>::value) {
@@ -577,7 +577,7 @@ int filterAndTransformParticles (DstTile1& dst1, DstTile2& dst2, const SrcTile&
     const auto src_data = src.getConstParticleTileData();
 
     amrex::ParallelForRNG(np,
-    [=] AMREX_GPU_DEVICE (int i, amrex::RandomEngine const& engine) noexcept
+    [p, p_mask, src_data] AMREX_GPU_DEVICE (int i, amrex::RandomEngine const& engine) noexcept
     {
         amrex::ignore_unused(p, p_mask, src_data, engine);
         if constexpr (IsCallable<Pred,decltype(src_data),int,RandomEngine>::value) {
@@ -620,7 +620,7 @@ Index filterAndTransformParticles (DstTile& dst, const SrcTile& src, Pred&& p, F
     const auto src_data = src.getConstParticleTileData();
 
     amrex::ParallelForRNG(np,
-    [=] AMREX_GPU_DEVICE (int i, amrex::RandomEngine const& engine) noexcept
+    [p, p_mask, src_data, src_start] AMREX_GPU_DEVICE (int i, amrex::RandomEngine const& engine) noexcept
     {
         amrex::ignore_unused(p, p_mask, src_data, src_start, engine);
         if constexpr (IsCallable<Pred,decltype(src_data),Index,RandomEngine>::value) {