OilfieldToolsNet/opm-simulators
4
0
Fork 0
mirror of https://github.com/OPM/opm-simulators.git synced 2025-01-14 04:41:56 -06:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #3644 from blattms/use-cl2.hpp

Browse Source 
[enhancement][opencl} Use cl2.hpp instead of cl.hpp (missing on bullseye)
This commit is contained in:
Atgeirr Flø Rasmussen 2021-11-01 10:10:30 +01:00 committed by GitHub
commit fa3b619ec9
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
9 changed files with 46 additions and 43 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
CMakeLists.txt
View File

@ -200,14 +200,14 @@ if(ENABLE_FPGA)
endif()
if(OpenCL_FOUND)
# the current OpenCL implementation relies on cl.hpp, not cl2.hpp
# the current OpenCL implementation relies on cl2.hpp, not cl.hpp
# make sure it is available, otherwise disable OpenCL
find_file(CL_HPP CL/cl.hpp HINTS ${OpenCL_INCLUDE_DIRS})
if(CL_HPP)
find_file(CL2_HPP CL/cl2.hpp HINTS ${OpenCL_INCLUDE_DIRS})
if(CL2_HPP)
set(HAVE_OPENCL 1)
include_directories(${OpenCL_INCLUDE_DIRS})
else()
message(WARNING " OpenCL was found, but this version of opm-simulators relies on CL/cl.hpp, which implements OpenCL 1.0, 1.1 and 1.2.\n Deactivating OpenCL")
message(WARNING " OpenCL was found, but this version of opm-simulators relies on CL/cl2.hpp, which implements OpenCL 1.0, 1.1 and 1.2.\n Deactivating OpenCL")
set(OpenCL_FOUND OFF)
set(OPENCL_FOUND OFF)
endif()

16
opm/simulators/linalg/bda/BILU0.cpp
View File

@ -654,10 +654,10 @@ void BILU0<block_size>::setKernelParameters(const unsigned int work_group_size_,
}
template <unsigned int block_size>
void BILU0<block_size>::setKernels(
cl::make_kernel<cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, const unsigned int, const unsigned int, cl::LocalSpaceArg> *ILU_apply1_,
cl::make_kernel<cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, const unsigned int, const unsigned int, cl::LocalSpaceArg> *ILU_apply2_,
cl::make_kernel<cl::Buffer&, const double, const unsigned int> *scale_,
cl::make_kernel<const unsigned int, const unsigned int, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, const int, cl::LocalSpaceArg> *ilu_decomp_
cl::KernelFunctor<cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, const unsigned int, const unsigned int, cl::LocalSpaceArg> *ILU_apply1_,
cl::KernelFunctor<cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, const unsigned int, const unsigned int, cl::LocalSpaceArg> *ILU_apply2_,
cl::KernelFunctor<cl::Buffer&, const double, const unsigned int> *scale_,
cl::KernelFunctor<const unsigned int, const unsigned int, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, const int, cl::LocalSpaceArg> *ilu_decomp_
){
this->ILU_apply1 = ILU_apply1_;
this->ILU_apply2 = ILU_apply2_;
@ -677,10 +677,10 @@ template void BILU0<n>::setOpenCLContext(cl::Context*);
template void BILU0<n>::setOpenCLQueue(cl::CommandQueue*); \
template void BILU0<n>::setKernelParameters(unsigned int, unsigned int, unsigned int); \
template void BILU0<n>::setKernels( \
cl::make_kernel<cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, const unsigned int, const unsigned int, cl::LocalSpaceArg> *, \
cl::make_kernel<cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, const unsigned int, const unsigned int, cl::LocalSpaceArg> *, \
cl::make_kernel<cl::Buffer&, const double, const unsigned int> *, \
cl::make_kernel<const unsigned int, const unsigned int, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, const int, cl::LocalSpaceArg> * \
cl::KernelFunctor<cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, const unsigned int, const unsigned int, cl::LocalSpaceArg> *, \
cl::KernelFunctor<cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, const unsigned int, const unsigned int, cl::LocalSpaceArg> *, \
cl::KernelFunctor<cl::Buffer&, const double, const unsigned int> *, \
cl::KernelFunctor<const unsigned int, const unsigned int, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, const int, cl::LocalSpaceArg> * \
);
INSTANTIATE_BDA_FUNCTIONS(1);

16
opm/simulators/linalg/bda/BILU0.hpp
View File

@ -85,10 +85,10 @@ namespace bda
ilu_apply1_kernel_type *ILU_apply1;
ilu_apply2_kernel_type *ILU_apply2;
cl::make_kernel<cl::Buffer&, const double, const unsigned int> *scale;
cl::make_kernel<const unsigned int, const unsigned int, cl::Buffer&, cl::Buffer&, cl::Buffer&,
cl::Buffer&, cl::Buffer&,
const int, cl::LocalSpaceArg> *ilu_decomp;
cl::KernelFunctor<cl::Buffer&, const double, const unsigned int> *scale;
cl::KernelFunctor<const unsigned int, const unsigned int, cl::Buffer&, cl::Buffer&, cl::Buffer&,
cl::Buffer&, cl::Buffer&,
const int, cl::LocalSpaceArg> *ilu_decomp;
GPU_storage s;
cl::Context *context;
@ -122,10 +122,10 @@ namespace bda
void setOpenCLQueue(cl::CommandQueue *queue);
void setKernelParameters(const unsigned int work_group_size, const unsigned int total_work_items, const unsigned int lmem_per_work_group);
void setKernels(
cl::make_kernel<cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, const unsigned int, const unsigned int, cl::LocalSpaceArg> *ILU_apply1,
cl::make_kernel<cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, const unsigned int, const unsigned int, cl::LocalSpaceArg> *ILU_apply2,
cl::make_kernel<cl::Buffer&, const double, const unsigned int> *scale,
cl::make_kernel<const unsigned int, const unsigned int, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, const int, cl::LocalSpaceArg> *ilu_decomp
cl::KernelFunctor<cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, const unsigned int, const unsigned int, cl::LocalSpaceArg> *ILU_apply1,
cl::KernelFunctor<cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, const unsigned int, const unsigned int, cl::LocalSpaceArg> *ILU_apply2,
cl::KernelFunctor<cl::Buffer&, const double, const unsigned int> *scale,
cl::KernelFunctor<const unsigned int, const unsigned int, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, const int, cl::LocalSpaceArg> *ilu_decomp
);
int* getToOrder()

4
opm/simulators/linalg/bda/ChowPatelIlu.cpp
View File

@ -488,7 +488,7 @@ void ChowPatelIlu::decomposition(
try {
// just put everything in the capture list
std::call_once(initialize_flag, [&](){
cl::Program::Sources source(1, std::make_pair(chow_patel_ilu_sweep_s, strlen(chow_patel_ilu_sweep_s))); // what does this '1' mean? cl::Program::Sources is of type 'std::vector<std::pair<const char*, long unsigned int> >'
cl::Program::Sources source(1, chow_patel_ilu_sweep_s); // what does this '1' mean? cl::Program::Sources is of type 'std::vector<std::pair<const char*, long unsigned int> >'
cl::Program program = cl::Program(*context, source, &err);
if (err != CL_SUCCESS) {
OPM_THROW(std::logic_error, "ChowPatelIlu OpenCL could not create Program");
@ -497,7 +497,7 @@ void ChowPatelIlu::decomposition(
std::vector<cl::Device> devices = context->getInfo<CL_CONTEXT_DEVICES>();
program.build(devices);
chow_patel_ilu_sweep_k.reset(new cl::make_kernel<cl::Buffer&, cl::Buffer&, cl::Buffer&,
chow_patel_ilu_sweep_k.reset(new cl::KernelFunctor<cl::Buffer&, cl::Buffer&, cl::Buffer&,
cl::Buffer&, cl::Buffer&, cl::Buffer&,
cl::Buffer&, cl::Buffer&, cl::Buffer&,
cl::Buffer&, cl::Buffer&,

2
opm/simulators/linalg/bda/ChowPatelIlu.hpp
View File

@ -47,7 +47,7 @@ namespace bda
cl_int err;
std::once_flag initialize_flag;
std::unique_ptr<cl::make_kernel<cl::Buffer&, cl::Buffer&, cl::Buffer&,
std::unique_ptr<cl::KernelFunctor<cl::Buffer&, cl::Buffer&, cl::Buffer&,
cl::Buffer&, cl::Buffer&, cl::Buffer&,
cl::Buffer&, cl::Buffer&, cl::Buffer&,
cl::Buffer&, cl::Buffer&,

7
opm/simulators/linalg/bda/opencl.hpp
View File

@ -20,9 +20,12 @@
/// This file includes the relevant OpenCL header(s)
/// All bda files using OpenCL declarations should include this header
#define __CL_ENABLE_EXCEPTIONS
#define CL_HPP_ENABLE_EXCEPTIONS
#define CL_TARGET_OPENCL_VERSION 120 // indicate OpenCL 1.2 is used
#include <CL/cl.hpp> // supports up to OpenCL 1.2
#define CL_HPP_TARGET_OPENCL_VERSION 120 // indicate OpenCL 1.2 is used
#define CL_HPP_MINIMUM_VERSION 120
#define CL_HPP_MINIMUM_OPENCL_VERSION 120
#include <CL/cl2.hpp> // supports up to OpenCL 1.2
#include <string>

12
opm/simulators/linalg/bda/openclKernels.hpp
View File

@ -27,21 +27,21 @@
namespace bda
{
using spmv_kernel_type = cl::make_kernel<cl::Buffer&, cl::Buffer&, cl::Buffer&, const unsigned int,
using spmv_kernel_type = cl::KernelFunctor<cl::Buffer&, cl::Buffer&, cl::Buffer&, const unsigned int,
cl::Buffer&, cl::Buffer&, const unsigned int, cl::LocalSpaceArg>;
using ilu_apply1_kernel_type = cl::make_kernel<cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&,
using ilu_apply1_kernel_type = cl::KernelFunctor<cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&,
cl::Buffer&, cl::Buffer&, const unsigned int, const unsigned int, cl::LocalSpaceArg>;
using ilu_apply2_kernel_type = cl::make_kernel<cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&,
using ilu_apply2_kernel_type = cl::KernelFunctor<cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&,
cl::Buffer&, cl::Buffer&, const unsigned int, const unsigned int, cl::LocalSpaceArg>;
using stdwell_apply_kernel_type = cl::make_kernel<cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&,
using stdwell_apply_kernel_type = cl::KernelFunctor<cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&,
cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&,
const unsigned int, const unsigned int, cl::Buffer&,
cl::LocalSpaceArg, cl::LocalSpaceArg, cl::LocalSpaceArg>;
using stdwell_apply_no_reorder_kernel_type = cl::make_kernel<cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&,
using stdwell_apply_no_reorder_kernel_type = cl::KernelFunctor<cl::Buffer&, cl::Buffer&, cl::Buffer&, cl::Buffer&,
cl::Buffer&, cl::Buffer&, cl::Buffer&,
const unsigned int, const unsigned int, cl::Buffer&,
cl::LocalSpaceArg, cl::LocalSpaceArg, cl::LocalSpaceArg>;
using ilu_decomp_kernel_type = cl::make_kernel<const unsigned int, const unsigned int, cl::Buffer&, cl::Buffer&,
using ilu_decomp_kernel_type = cl::KernelFunctor<const unsigned int, const unsigned int, cl::Buffer&, cl::Buffer&,
cl::Buffer&, cl::Buffer&, cl::Buffer&, const int, cl::LocalSpaceArg>;
/// Generate string with axpy kernel

14
opm/simulators/linalg/bda/openclSolverBackend.cpp
View File

@ -526,7 +526,7 @@ void openclSolverBackend<block_size>::initialize(int N_, int nnz_, int dim, doub
} // end initialize()
void add_kernel_string(cl::Program::Sources &sources, std::string &source) {
sources.emplace_back(std::make_pair(source.c_str(), source.size()));
sources.emplace_back(source);
}
template <unsigned int block_size>
@ -565,14 +565,14 @@ void openclSolverBackend<block_size>::get_opencl_kernels() {
program.build(devices);
// queue.enqueueNDRangeKernel() is a blocking/synchronous call, at least for NVIDIA
// cl::make_kernel<> myKernel(); myKernel(args, arg1, arg2); is also blocking
// cl::KernelFunctor<> myKernel(); myKernel(args, arg1, arg2); is also blocking
// actually creating the kernels
dot_k.reset(new cl::make_kernel<cl::Buffer&, cl::Buffer&, cl::Buffer&, const unsigned int, cl::LocalSpaceArg>(cl::Kernel(program, "dot_1")));
norm_k.reset(new cl::make_kernel<cl::Buffer&, cl::Buffer&, const unsigned int, cl::LocalSpaceArg>(cl::Kernel(program, "norm")));
axpy_k.reset(new cl::make_kernel<cl::Buffer&, const double, cl::Buffer&, const unsigned int>(cl::Kernel(program, "axpy")));
scale_k.reset(new cl::make_kernel<cl::Buffer&, const double, const unsigned int>(cl::Kernel(program, "scale")));
custom_k.reset(new cl::make_kernel<cl::Buffer&, cl::Buffer&, cl::Buffer&, const double, const double, const unsigned int>(cl::Kernel(program, "custom")));
dot_k.reset(new cl::KernelFunctor<cl::Buffer&, cl::Buffer&, cl::Buffer&, const unsigned int, cl::LocalSpaceArg>(cl::Kernel(program, "dot_1")));
norm_k.reset(new cl::KernelFunctor<cl::Buffer&, cl::Buffer&, const unsigned int, cl::LocalSpaceArg>(cl::Kernel(program, "norm")));
axpy_k.reset(new cl::KernelFunctor<cl::Buffer&, const double, cl::Buffer&, const unsigned int>(cl::Kernel(program, "axpy")));
scale_k.reset(new cl::KernelFunctor<cl::Buffer&, const double, const unsigned int>(cl::Kernel(program, "scale")));
custom_k.reset(new cl::KernelFunctor<cl::Buffer&, cl::Buffer&, cl::Buffer&, const double, const double, const unsigned int>(cl::Kernel(program, "custom")));
spmv_blocked_k.reset(new spmv_kernel_type(cl::Kernel(program, "spmv_blocked")));
ILU_apply1_k.reset(new ilu_apply1_kernel_type(cl::Kernel(program, "ILU_apply1")));
ILU_apply2_k.reset(new ilu_apply2_kernel_type(cl::Kernel(program, "ILU_apply2")));

10
opm/simulators/linalg/bda/openclSolverBackend.hpp
View File

@ -65,11 +65,11 @@ private:
// shared pointers are also passed to other objects
std::vector<cl::Device> devices;
std::unique_ptr<cl::make_kernel<cl::Buffer&, cl::Buffer&, cl::Buffer&, const unsigned int, cl::LocalSpaceArg> > dot_k;
std::unique_ptr<cl::make_kernel<cl::Buffer&, cl::Buffer&, const unsigned int, cl::LocalSpaceArg> > norm_k;
std::unique_ptr<cl::make_kernel<cl::Buffer&, const double, cl::Buffer&, const unsigned int> > axpy_k;
std::unique_ptr<cl::make_kernel<cl::Buffer&, const double, const unsigned int> > scale_k;
std::unique_ptr<cl::make_kernel<cl::Buffer&, cl::Buffer&, cl::Buffer&, const double, const double, const unsigned int> > custom_k;
std::unique_ptr<cl::KernelFunctor<cl::Buffer&, cl::Buffer&, cl::Buffer&, const unsigned int, cl::LocalSpaceArg> > dot_k;
std::unique_ptr<cl::KernelFunctor<cl::Buffer&, cl::Buffer&, const unsigned int, cl::LocalSpaceArg> > norm_k;
std::unique_ptr<cl::KernelFunctor<cl::Buffer&, const double, cl::Buffer&, const unsigned int> > axpy_k;
std::unique_ptr<cl::KernelFunctor<cl::Buffer&, const double, const unsigned int> > scale_k;
std::unique_ptr<cl::KernelFunctor<cl::Buffer&, cl::Buffer&, cl::Buffer&, const double, const double, const unsigned int> > custom_k;
std::unique_ptr<spmv_kernel_type> spmv_blocked_k;
std::shared_ptr<ilu_apply1_kernel_type> ILU_apply1_k;
std::shared_ptr<ilu_apply2_kernel_type> ILU_apply2_k;