Rocm36 (#1263)

## Summary Update make system for hip-clang. Remove some hip workarounds. ## Additional background ## Checklist The proposed changes: - [ ] fix a bug or incorrect behavior in AMReX - [ ] add new capabilities to AMReX - [ ] changes answers in the test suite to more than roundoff level - [ ] are likely to significantly affect the results of downstream AMReX users - [ ] are described in the proposed changes to the AMReX documentation, if appropriate
AMReX-Codes · Aug 14, 2020 · 192fa35 · 192fa35
1 parent 6c36b5f
commit 192fa35
Show file tree

Hide file tree

Showing 10 changed files with 28 additions and 87 deletions.
diff --git a/Src/Base/AMReX_BaseFab.H b/Src/Base/AMReX_BaseFab.H
@@ -2456,7 +2456,7 @@ BaseFab<T>::maxabs (const Box& subbox, int comp) const noexcept
 #endif
     {
         T r = 0;
-        amrex::Loop(subbox, [=,&r] (int i, int j, int k) AMREX_NOEXCEPT
+        amrex::Loop(subbox, [=,&r] (int i, int j, int k) noexcept
         {
             r = amrex::max(r, amrex::Math::abs(a(i,j,k)));
         });
@@ -3929,7 +3929,7 @@ BaseFab<T>::sum (const Box& bx, DestComp dcomp, NumComps ncomp) const noexcept
     } else
 #endif
     {
-        amrex::LoopOnCpu(bx, ncomp.n, [=,&r] (int i, int j, int k, int n) AMREX_NOEXCEPT
+        amrex::LoopOnCpu(bx, ncomp.n, [=,&r] (int i, int j, int k, int n) noexcept
         {
             r += a(i,j,k,n+dcomp.i);
         });
@@ -3969,7 +3969,7 @@ BaseFab<T>::dot (const BaseFab<T>& src, const Box& bx, SrcComp scomp, DestComp d
     } else
 #endif
     {
-        amrex::LoopOnCpu(bx, ncomp.n, [=,&r] (int i, int j, int k, int n) AMREX_NOEXCEPT
+        amrex::LoopOnCpu(bx, ncomp.n, [=,&r] (int i, int j, int k, int n) noexcept
         {
             r += d(i,j,k,n+dcomp.i) * s(i,j,k,n+scomp.i);
         });
@@ -4015,7 +4015,7 @@ BaseFab<T>::dot (const Box& bx, DestComp dcomp, NumComps ncomp) const noexcept
     } else
 #endif
     {
-        amrex::LoopOnCpu(bx, ncomp.n, [=,&r] (int i, int j, int k, int n) AMREX_NOEXCEPT
+        amrex::LoopOnCpu(bx, ncomp.n, [=,&r] (int i, int j, int k, int n) noexcept
         {
             r += a(i,j,k,n+dcomp.i)*a(i,j,k,n+dcomp.i);
         });
@@ -4059,7 +4059,7 @@ BaseFab<T>::dotmask (const BaseFab<T>& src, const Box& bx, const BaseFab<int>& m
     } else
 #endif
     {
-        amrex::LoopOnCpu(bx, ncomp.n, [=,&r] (int i, int j, int k, int n) AMREX_NOEXCEPT
+        amrex::LoopOnCpu(bx, ncomp.n, [=,&r] (int i, int j, int k, int n) noexcept
         {
             int mi = static_cast<int>(static_cast<bool>(m(i,j,k)));
             r += d(i,j,k,n+dcomp.i)*s(i,j,k,n+scomp.i)*mi;

diff --git a/Src/Base/AMReX_BaseFabUtility.H b/Src/Base/AMReX_BaseFabUtility.H
@@ -9,11 +9,11 @@ template <class Tto, class Tfrom>
 AMREX_GPU_HOST_DEVICE
 void
 cast (BaseFab<Tto>& tofab, BaseFab<Tfrom> const& fromfab,
-      Box const& bx, SrcComp scomp, DestComp dcomp, NumComps ncomp) AMREX_NOEXCEPT 
+      Box const& bx, SrcComp scomp, DestComp dcomp, NumComps ncomp) noexcept
 {
     auto const& tdata = tofab.array();
     auto const& fdata = fromfab.const_array();
-    amrex::LoopConcurrent(bx, ncomp.n, [=] (int i, int j, int k, int n) AMREX_NOEXCEPT
+    amrex::LoopConcurrent(bx, ncomp.n, [=] (int i, int j, int k, int n) noexcept
     {
         tdata(i,j,k,n+dcomp.i) = static_cast<Tto>(fdata(i,j,k,n+scomp.i));
     });

diff --git a/Src/Base/AMReX_Extension.H b/Src/Base/AMReX_Extension.H
@@ -3,15 +3,6 @@
 
 #if !defined(BL_LANG_FORT)
 
-// HIP FIX HERE - noexcept
-
-#ifdef AMREX_HIP_PLATFORM_HCC
-#define AMREX_NOEXCEPT
-#else
-#define AMREX_NOEXCEPT noexcept
-#endif
-
-
 // restrict
 
 #ifdef __cplusplus

diff --git a/Src/Base/AMReX_GpuAsyncArray.H b/Src/Base/AMReX_GpuAsyncArray.H
@@ -9,15 +9,10 @@
 #include <AMReX_TypeTraits.H>
 #include <AMReX_GpuDevice.H>
 
-// HIP FIX HERE - MOVE TO ONE SPOT
-#ifdef __HIP_PLATFORM_HCC__
-#define HIPRT_CB 
-#endif
-
 #if defined(AMREX_USE_CUDA) || defined(AMREX_USE_HIP)
 extern "C" {
 #if defined(AMREX_USE_HIP)
-    void HIPRT_CB  amrex_asyncarray_delete ( hipStream_t stream,  hipError_t error, void* p);
+    void amrex_asyncarray_delete ( hipStream_t stream,  hipError_t error, void* p);
 #elif defined(__CUDACC__) && (__CUDACC_VER_MAJOR__ >= 10)
     void CUDART_CB amrex_asyncarray_delete (void* p);
 #elif defined(AMREX_USE_CUDA)

diff --git a/Src/Base/AMReX_GpuAsyncArray.cpp b/Src/Base/AMReX_GpuAsyncArray.cpp
@@ -5,7 +5,7 @@
 #if !defined(AMREX_USE_DPCPP)
 extern "C" {
 #if defined(AMREX_USE_HIP)
-    void HIPRT_CB  amrex_asyncarray_delete ( hipStream_t /*stream*/,  hipError_t /*error*/, void* p)
+    void amrex_asyncarray_delete ( hipStream_t /*stream*/,  hipError_t /*error*/, void* p)
 #elif defined(__CUDACC__) && (__CUDACC_VER_MAJOR__ >= 10)
     void CUDART_CB amrex_asyncarray_delete (void* p)
 #elif defined(AMREX_USE_CUDA)

diff --git a/Src/Base/AMReX_GpuElixir.cpp b/Src/Base/AMReX_GpuElixir.cpp
@@ -6,10 +6,6 @@
 #include <memory>
 #include <AMReX_GpuDevice.H>
 
-#ifdef __HIP_PLATFORM_HCC__
-#define HIPRT_CB 
-#endif
-
 namespace amrex {
 namespace Gpu {
 
@@ -19,7 +15,7 @@ namespace {
 
 extern "C" {
 #if defined(AMREX_USE_HIP)
-    void HIPRT_CB  amrex_elixir_delete ( hipStream_t /*stream*/,  hipError_t /*error*/, void* p)
+    void amrex_elixir_delete ( hipStream_t /*stream*/,  hipError_t /*error*/, void* p)
 #elif defined(__CUDACC__) && (__CUDACC_VER_MAJOR__ >= 10)
     void CUDART_CB amrex_elixir_delete (void* p)
 #elif defined(AMREX_USE_CUDA)

diff --git a/Src/Base/AMReX_GpuError.H b/Src/Base/AMReX_GpuError.H
@@ -89,17 +89,6 @@ namespace Gpu {
                            + " " + hipGetErrorString(amrex_i_err)); \
         amrex::Abort(errStr); \
     }}
-
-#ifdef AMREX_HIP_PLATFORM_NVCC
-#define AMREX_CUDA_SAFE_CALL(call) { \
-    cudaError_t amrex_i_err = call;          \
-    if (cudaSuccess != amrex_i_err) { \
-        std::string errStr(std::string("CUDA error in file ") + __FILE__ \
-                           + " line " + std::to_string(__LINE__) \
-                           + " " + cudaGetErrorString(amrex_i_err)); \
-        amrex::Abort(errStr); \
-    }}
-#endif
 #endif
 
 #define AMREX_GPU_ERROR_CHECK() amrex::Gpu::ErrorCheck(__FILE__, __LINE__)

diff --git a/Src/Base/AMReX_GpuLaunch.H b/Src/Base/AMReX_GpuLaunch.H
@@ -248,12 +248,12 @@ namespace Gpu {
 #define AMREX_HOST_DEVICE_PARALLEL_FOR_4D_FLAG(where_to_run,box,nc,i,j,k,n,block) \
     if ((where_to_run == RunOn::Device) && (Gpu::inLaunchRegion())) \
     { \
-        amrex::ParallelFor(box, nc, [=] AMREX_GPU_DEVICE (int i, int j, int k, int n) AMREX_NOEXCEPT \
+        amrex::ParallelFor(box, nc, [=] AMREX_GPU_DEVICE (int i, int j, int k, int n) noexcept \
             block \
         ); \
     } \
     else { \
-        amrex::LoopConcurrentOnCpu(box, nc, [=] (int i, int j, int k, int n) AMREX_NOEXCEPT \
+        amrex::LoopConcurrentOnCpu(box, nc, [=] (int i, int j, int k, int n) noexcept \
             block \
         ); \
     }
@@ -288,12 +288,12 @@ namespace Gpu {
 #define AMREX_HOST_DEVICE_FOR_4D_FLAG(where_to_run,box,nc,i,j,k,n,block) \
     if ((where_to_run == RunOn::Device) && (Gpu::inLaunchRegion())) \
     { \
-        amrex::ParallelFor(box, nc, [=] AMREX_GPU_DEVICE (int i, int j, int k, int n) AMREX_NOEXCEPT \
+        amrex::ParallelFor(box, nc, [=] AMREX_GPU_DEVICE (int i, int j, int k, int n) noexcept \
             block \
         ); \
     } \
     else { \
-        amrex::LoopOnCpu(box, nc, [=] (int i, int j, int k, int n) AMREX_NOEXCEPT \
+        amrex::LoopOnCpu(box, nc, [=] (int i, int j, int k, int n) noexcept \
             block \
         ); \
     }

diff --git a/Tools/GNUMake/Make.defs b/Tools/GNUMake/Make.defs
@@ -48,15 +48,12 @@ ifeq ($(USE_HIP),TRUE)
   COMP := hip
   HIP_PLATFORM = $(shell hipconfig --compiler)
   ifeq ($(HIP_PLATFORM),nvcc)
-  $(error HIP_PLATFORM nvcc is not supported at this time. Use USE_CUDA to compile for NVIDIA platforms)
-#    If nvcc is supported, put this back.
-#    CUDA_ARCH ?= 60
-#    CUDA_MAX_THREADS ?= 256
-#    CUDA_MAXREGCOUNT ?= 255
+    $(error HIP_PLATFORM nvcc is not supported at this time. Use USE_CUDA to compile for NVIDIA platforms)
   else ifeq ($(HIP_PLATFORM),hcc)
-# do nothing special for now, but also don't abort
+    $(error HIP_PLATFORM=hcc is no longer supported)
+  else ifeq ($(HIP_PLATFORM),clang)
   else
-    $(error only HIP_PLATFORM=hcc is supported)
+    $(error Unknown HIP_PLATFORM=$(HIP_PLATFORM))
   endif
 endif
 
@@ -672,14 +669,7 @@ ifeq ($(USE_HIP),TRUE)
 
     endif
 
-    DEFINES += -DAMREX_USE_HIP
-
-    ifeq ($(HIP_PLATFORM),nvcc)
-        $(error HIP_PLATFORM nvcc is not supported at this time. Use USE_CUDA to compile for NVIDIA platforms.)
- #       DEFINES += -DAMREX_USE_CUDA -DAMREX_HIP_PLATFORM_NVCC
-    else ifeq ($(HIP_PLATFORM),hcc)
-        DEFINES += -DAMREX_HIP_PLATFORM_HCC
-    endif
+    DEFINES += -DAMREX_USE_HIP -DAMREX_HIP_PLATFORM=$(HIP_PLATFORM)
 
     GPUSuffix := .HIP
 

diff --git a/Tools/GNUMake/comps/hip.mak b/Tools/GNUMake/comps/hip.mak
@@ -50,21 +50,22 @@ endif  # BL_NO_FORT
 
 # =============================================================================================
 
-# This is designed only for dogora for now.
-ifeq ($(HIP_PLATFORM),hcc)
+ifeq ($(HIP_PLATFORM),clang)
 
   ifeq ($(DEBUG),TRUE)
-    # From llvm
-    CXXFLAGS += -g
-    CFLAGS   += -g 
+    CXXFLAGS += -g -O0 -Wall -Wextra -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -ftrapv
+    CFLAGS   += -g -O0 -Wall -Wextra -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -ftrapv
     FFLAGS   += -g -O0 -ggdb -fbounds-check -fbacktrace -Wuninitialized -Wunused -ffpe-trap=invalid,zero -finit-real=snan -finit-integer=2147483647 -ftrapv
     F90FLAGS += -g -O0 -ggdb -fbounds-check -fbacktrace -Wuninitialized -Wunused -ffpe-trap=invalid,zero -finit-real=snan -finit-integer=2147483647 -ftrapv
-
   else  # DEBUG=FALSE flags
+    CXXFLAGS += -g -O3
+    CFLAGS   += -g -O3
+    FFLAGS   += -g -O3
+    F90FLAGS += -g -O3
   endif
 
   # Generic HIP info
-  ROC_PATH=/opt/rocm
+  ROC_PATH=$(realpath $(dir $(HIP_PATH)))
   INCLUDE_LOCATIONS += $(HIP_PATH)/include
 
   # rocRand
@@ -76,33 +77,12 @@ ifeq ($(HIP_PLATFORM),hcc)
   INCLUDE_LOCATIONS += $(ROC_PATH)/rocprim/include
 
   # rocThrust - Header only
-  INCLUDE_LOCATIONS += $(ROC_PATH)/rocthrust/include
+  # INCLUDE_LOCATIONS += $(ROC_PATH)/rocthrust/include
 
 # =============================================================================================
 
-# This is Summit. Likely broken.
 else ifeq ($(HIP_PLATFORM),nvcc)
   $(error HIP_PLATFORM nvcc is not supported at this time. Use USE_CUDA to compile for NVIDIA platforms.)
-#
-#  CXXFLAGS_FROM_HOST := -ccbin=$(CXX) --std=c++14
-#  CFLAGS_FROM_HOST := -ccbin=$(CXX)
-#  HIPCC_FLAGS = -Wno-deprecated-gpu-targets -m64 -arch=compute_$(CUDA_ARCH) -code=sm_$(CUDA_ARCH) -maxrregcount=$(CUDA_MAXREGCOUNT)
-#
-#  ifeq ($(DEBUG),TRUE)
-#    HIPCC_FLAGS += -g -G
-#  else
-#    HIPCC_FLAGS += -lineinfo --ptxas-options=-O3,-v
-#  endif
-#
-#  ifneq ($(USE_CUDA_FAST_MATH),FALSE)
-#    HIPCC_FLAGS += --use_fast_math
-#  endif
-#
-#  CXXFLAGS = $(CXXFLAGS_FROM_HOST) $(HIPCC_FLAGS) -c -dc
-#  CFLAGS   =   $(CFLAGS_FROM_HOST) $(HIPCC_FLAGS) -dc
-#
-#  CXXFLAGS += --expt-relaxed-constexpr --expt-extended-lambda
-#
 endif
 
 # =============================================================================================