Change compiler from dpcpp to icpx -fsycl Part 1 (#1180)

* CombinationalLogic/mandelbrot

* CombinationalLogic/sepia-filter

* DenseLinearAlgebra/complex_mult

* DenseLinearAlgebra/guided_jacobi

* DenseLinearAlgebra/matrix_mul

* DenseLinearAlgebra/Jacobi_Iterative

* DenseLinearAlgebra/simple-add

* GraphAlgorithms/all-pairs

* GraphTraversal/bitonic-sort

* GraphTraversal/hidden-markov-models

* GraphTraversal/odd-even-merge-sort

* MapReduce/MonteCarloPi

* OpenCLInterop

* ParallelPaterns/loop-unroll

* ParallelPatterns/PrefixSum

* SparseLinearAlgebra/merge-spmv

* SpectralMethods/DiscreteCosineTransform

* StructuredGrids/1d_HeatTransfer

* StructuredGrids/guided_HSOpticalFlow

* StructuredGrids/guided_iso3dfd

* StructuredGrids/iso2dfd

* StructuredGrids/iso3dfd

* StructuredGrids/particle-diffusion

* ParallelPatterns/dpc_reduce

* ParallelPatterns/histogram

* Remove exception Handlers and rename DPC++ extension to Intel

Co-authored-by: Kanclerz <[email protected]>

Author: KanclerzPiotr

DirectProgramming/DPC++/CombinationalLogic/mandelbrot/CMakeLists.txt

@@ -1,6 +1,6 @@
 cmake_minimum_required (VERSION 3.4)
 
-set(CMAKE_CXX_COMPILER dpcpp)
+set(CMAKE_CXX_COMPILER icpx)
 
 # Set default build type to RelWithDebInfo if not specified
 if (NOT CMAKE_BUILD_TYPE)

DirectProgramming/DPC++/CombinationalLogic/mandelbrot/src/CMakeLists.txt

@@ -1,4 +1,4 @@
-set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -g -std=c++17")
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -g -std=c++17 -fsycl")
 set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS}")
 
 add_executable(mandelbrot main.cpp)

DirectProgramming/DPC++/CombinationalLogic/mandelbrot/src/main.cpp

@@ -73,7 +73,7 @@ int main(int argc, char *argv[]) {
   try {
     // Create a queue on the default device. Set SYCL_DEVICE_TYPE environment
     // variable to (CPU|GPU|FPGA|HOST) to change the device.
-    queue q(default_selector{}, dpc_common::exception_handler);
+    queue q(default_selector_v);
 
     // Display the device info.
     ShowDevice(q);

DirectProgramming/DPC++/CombinationalLogic/sepia-filter/CMakeLists.txt

@@ -1,4 +1,4 @@
-set(CMAKE_CXX_COMPILER "dpcpp")
+set(CMAKE_CXX_COMPILER "icpx")
 # Set default build type to RelWithDebInfo if not specified
 if (NOT CMAKE_BUILD_TYPE)
     message (STATUS "Default CMAKE_BUILD_TYPE not set using Release with Debug Info")

DirectProgramming/DPC++/CombinationalLogic/sepia-filter/src/CMakeLists.txt

@@ -1,4 +1,4 @@
-set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++17")
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++17 -fsycl")
 set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS}")
 add_executable (sepia sepia_dpcpp.cpp)
 target_link_libraries(sepia OpenCL sycl)

DirectProgramming/DPC++/CombinationalLogic/sepia-filter/src/device_selector.hpp

@@ -18,12 +18,12 @@
 // which '()' returned the highest number, is selected. If a negative number
 // was returned for all devices, then the selection process will cause an
 // exception.
-class MyDeviceSelector : public cl::sycl::device_selector {
+class MyDeviceSelector {
  public:
   MyDeviceSelector() {}
 
   // This is the function which gives a "rating" to devices.
-  virtual int operator()(const cl::sycl::device &device) const override {
+  virtual int operator()(const cl::sycl::device &device) const {
     // The template parameter to device.get_info can be a variety of properties
     // defined by the SYCL spec's cl::sycl::info:: enum. Properties may have
     // different types. Here we query name which is a string.
@@ -37,7 +37,6 @@ class MyDeviceSelector : public cl::sycl::device_selector {
     if (device.is_gpu()) return 500;
     if (device.is_accelerator()) return 400;
     if (device.is_cpu()) return 300;
-    if (device.is_host()) return 100;
     return -1;
   }
 };

DirectProgramming/DPC++/CombinationalLogic/sepia-filter/src/sepia_dpcpp.cpp

@@ -6,7 +6,7 @@
 #include <chrono>
 #include <cmath>
 #include <iostream>
-#include "CL/sycl.hpp"
+#include <sycl/sycl.hpp>
 #include "device_selector.hpp"
 
 // dpc_common.hpp can be found in the dev-utilities include folder.
@@ -26,7 +26,7 @@ using namespace sycl;
 // Few useful acronyms.
 constexpr auto sycl_read = access::mode::read;
 constexpr auto sycl_write = access::mode::write;
-constexpr auto sycl_global_buffer = access::target::global_buffer;
+constexpr auto sycl_device = access::target::device;
 
 static void ReportTime(const string &msg, event e) {
   cl_ulong time_start =
@@ -72,8 +72,8 @@ class SepiaFunctor {
  public:
   // Constructor captures needed data into fields
   SepiaFunctor(
-      accessor<uint8_t, 1, sycl_read, sycl_global_buffer> &image_acc_,
-      accessor<uint8_t, 1, sycl_write, sycl_global_buffer> &image_exp_acc_)
+      accessor<uint8_t, 1, sycl_read, sycl_device> &image_acc_,
+      accessor<uint8_t, 1, sycl_write, sycl_device> &image_exp_acc_)
       : image_acc(image_acc_), image_exp_acc(image_exp_acc_) {}
 
   // The '()' operator is the actual kernel
@@ -83,8 +83,8 @@ class SepiaFunctor {
 
  private:
   // Captured values:
-  accessor<uint8_t, 1, sycl_read, sycl_global_buffer> image_acc;
-  accessor<uint8_t, 1, sycl_write, sycl_global_buffer> image_exp_acc;
+  accessor<uint8_t, 1, sycl_read, sycl_device> image_acc;
+  accessor<uint8_t, 1, sycl_write, sycl_device> image_exp_acc;
 };
 
 int main(int argc, char **argv) {

DirectProgramming/DPC++/DenseLinearAlgebra/complex_mult/Makefile

@@ -1,5 +1,5 @@
-CXX = dpcpp
-CXXFLAGS = -std=c++17 -g -o
+CXX = icpx
+CXXFLAGS = -std=c++17 -g  -fsycl -o 
 LDFLAGS =
 EXE_NAME = complex_mult.exe
 SOURCES = src/complex_mult.cpp

DirectProgramming/DPC++/DenseLinearAlgebra/complex_mult/src/complex_mult.cpp

@@ -4,7 +4,7 @@
 //
 // SPDX-License-Identifier: MIT
 // =============================================================
-#include <CL/sycl.hpp>
+#include <sycl/sycl.hpp>
 #include <iomanip>
 #include <vector>
 // dpc_common.hpp can be found in the dev-utilities include folder.
@@ -18,10 +18,10 @@ using namespace std;
 // Number of complex numbers passing to the SYCL code
 static const int num_elements = 10000;
 
-class CustomDeviceSelector : public device_selector {
+class CustomDeviceSelector {
  public:
   CustomDeviceSelector(std::string vendorName) : vendorName_(vendorName){};
-  int operator()(const device &dev) const override {
+  int operator()(const device &dev) const {
     int device_rating = 0;
     // In the below code we are querying for the custom device specific to a
     // Vendor and if it is a GPU device we are giving the highest rating. The
@@ -124,7 +124,7 @@ int main() {
     // std::string vendor_name = "Nvidia";
     // queue constructor passed exception handler
     CustomDeviceSelector selector(vendor_name);
-    queue q(selector, dpc_common::exception_handler);
+    queue q(selector);
     // Call the DpcppParallel with the required inputs and outputs
     DpcppParallel(q, input_vect1, input_vect2, out_vect_parallel);
   } catch (...) {

DirectProgramming/DPC++/DenseLinearAlgebra/guided_jacobi_iterative_gpu_optimization/CMakeLists.txt

@@ -1,6 +1,6 @@
 cmake_minimum_required (VERSION 3.5)
 
-set(CMAKE_CXX_COMPILER dpcpp)
+set(CMAKE_CXX_COMPILER icpx)
 
 # Set default build type to RelWithDebInfo if not specified
 if (NOT CMAKE_BUILD_TYPE)

DirectProgramming/DPC++/DenseLinearAlgebra/guided_jacobi_iterative_gpu_optimization/src/1_guided_jacobi_iterative_solver_cpu.cpp

@@ -6,7 +6,7 @@
 // =============================================================
 #include <bits/stdc++.h>
 
-#include <CL/sycl.hpp>
+#include <sycl/sycl.hpp>
 #include <algorithm>
 #include <chrono>
 #include <cstdlib>
@@ -26,7 +26,6 @@ static const int kMinRand = -1000;
 static const int kMaxRaand = 1000;
 static const int kMaxSweepps = 100;
 static const std::uint32_t kSeed = 666;
-cpu_selector selector;
 std::ofstream outfile;
 
 // Function responsible for generating a float type

DirectProgramming/DPC++/DenseLinearAlgebra/guided_jacobi_iterative_gpu_optimization/src/2_guided_jacobi_iterative_solver_gpu.cpp

@@ -6,7 +6,7 @@
 // =============================================================
 #include <bits/stdc++.h>
 
-#include <CL/sycl.hpp>
+#include <sycl/sycl.hpp>
 #include <algorithm>
 #include <chrono>
 #include <cstdlib>
@@ -26,7 +26,6 @@ static const int kMinRand = -1000;
 static const int kMaxRand = 1000;
 static const int kMaxSweeps = 100;
 static const std::uint32_t kSeed = 666;
-gpu_selector selector;
 std::ofstream outfile;
 
 // Function responsible for generating a float type
@@ -38,7 +37,7 @@ std::ofstream outfile;
 // drasticly.
 void GenerateMatrix(std::vector<float> &input_matrix,
                     std::vector<Real> &input_results) {
-  queue q(selector);
+  queue q(gpu_selector_v);
 
   buffer bufin_mat(input_matrix);
   buffer bufin_res(input_results);
@@ -126,7 +125,7 @@ int main(int argc, char *argv[]) {
   std::vector<float> input_matrix(kSize * kSize);
   std::vector<Real> input_results(kSize);
 
-  queue q(selector);
+  queue q(gpu_selector_v);
 
   std::cout << "Device : " << q.get_device().get_info<info::device::name>()
             << std::endl;

DirectProgramming/DPC++/DenseLinearAlgebra/guided_jacobi_iterative_gpu_optimization/src/3_guided_jacobi_iterative_solver_multi_gpu.cpp

@@ -6,7 +6,7 @@
 // =============================================================
 #include <bits/stdc++.h>
 
-#include <CL/sycl.hpp>
+#include <sycl/sycl.hpp>
 #include <algorithm>
 #include <chrono>
 #include <cstdlib>
@@ -26,7 +26,6 @@ static const int kMinRand = -1000;
 static const int kMaxRand = 1000;
 static const int kMaxSweeps = 100;
 static const std::uint32_t kSeed = 666;
-gpu_selector selector;
 std::ofstream outfile;
 
 // Function responsible for generating a float type

DirectProgramming/DPC++/DenseLinearAlgebra/guided_jacobi_iterative_gpu_optimization/src/CMakeLists.txt

@@ -1,7 +1,7 @@
-set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -g -std=c++17")
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -g -fsycl -std=c++17")
 set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS}")
 
-add_executable(1_guided_jacobi_iterative_solver_cpu 1_guided_jacobi_iterative_solver_cpu)
+add_executable(1_guided_jacobi_iterative_solver_cpu 1_guided_jacobi_iterative_solver_cpu.cpp)
 add_executable(2_guided_jacobi_iterative_solver_gpu 2_guided_jacobi_iterative_solver_gpu.cpp)
 add_executable(3_guided_jacobi_iterative_solver_multi_gpu 3_guided_jacobi_iterative_solver_multi_gpu.cpp)
 

DirectProgramming/DPC++/DenseLinearAlgebra/jacobi_iterative/CMakeLists.txt

@@ -1,6 +1,6 @@
 cmake_minimum_required (VERSION 3.4.0)
 
-set(CMAKE_CXX_COMPILER "dpcpp")
+set(CMAKE_CXX_COMPILER "icpx")
 
 project (jacobi_iterative)
 # Set default build type to RelWithDebInfo if not specified

DirectProgramming/DPC++/DenseLinearAlgebra/jacobi_iterative/sycl_dpct_migrated/CMakeLists.txt

@@ -1,4 +1,4 @@
-set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsycl")
 set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS}")
 include_directories(${CMAKE_SOURCE_DIR}/sycl_dpct_migrated/Common/)
 

DirectProgramming/DPC++/DenseLinearAlgebra/jacobi_iterative/sycl_migrated/CMakeLists.txt

@@ -1,4 +1,4 @@
-set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsycl")
 set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS}")
 
 add_executable (sycl_migrated src/main.cpp src/jacobi.cpp)

DirectProgramming/DPC++/DenseLinearAlgebra/jacobi_iterative/sycl_migrated/src/jacobi.cpp

@@ -33,7 +33,7 @@
  */
 
 
-#include <CL/sycl.hpp>
+#include <sycl/sycl.hpp>
 
 #include "jacobi.h"
 
@@ -194,9 +194,9 @@ double JacobiMethodGpu(const float *A, const double *b,
 
     if ((k & 1) == 0) {
       q.submit([&](handler &cgh) {
-        accessor<double, 1, access_mode::read_write, access::target::local>
+        local_accessor<double, 1>
             x_shared_acc_ct1(range<1>(N_ROWS), cgh);
-        accessor<double, 1, access_mode::read_write, access::target::local>
+        local_accessor<double, 1>
             b_shared_acc_ct1(range<1>(ROWS_PER_CTA + 1), cgh);
 
         cgh.parallel_for(nd_range<3>(nblocks * nthreads, nthreads),
@@ -210,9 +210,9 @@ double JacobiMethodGpu(const float *A, const double *b,
       });
     } else {
       q.submit([&](handler &cgh) {
-        accessor<double, 1, access_mode::read_write, access::target::local>
+        local_accessor<double, 1>
             x_shared_acc_ct1(range<1>(N_ROWS), cgh);
-        accessor<double, 1, access_mode::read_write, access::target::local>
+        local_accessor<double, 1>
             b_shared_acc_ct1(range<1>(ROWS_PER_CTA + 1), cgh);
 
         cgh.parallel_for(nd_range<3>(nblocks * nthreads, nthreads),
@@ -237,7 +237,7 @@ double JacobiMethodGpu(const float *A, const double *b,
 
       if ((k & 1) == 0) {
         q.submit([&](handler &cgh) {
-          accessor<uint8_t, 1, access_mode::read_write, access::target::local>
+          local_accessor<uint8_t, 1>
               sycl_local_acc_ct1(range<1>(sharedMemSize), cgh);
 
           cgh.parallel_for(nd_range<3>(nblocks * nthreads, nthreads),
@@ -249,7 +249,7 @@ double JacobiMethodGpu(const float *A, const double *b,
         });
       } else {
         q.submit([&](handler &cgh) {
-          accessor<uint8_t, 1, access_mode::read_write, access::target::local>
+          local_accessor<uint8_t, 1>
               sycl_local_acc_ct1(range<1>(sharedMemSize), cgh);
 
           cgh.parallel_for(nd_range<3>(nblocks * nthreads, nthreads),

DirectProgramming/DPC++/DenseLinearAlgebra/jacobi_iterative/sycl_migrated/src/main.cpp

@@ -33,7 +33,7 @@
  */
 
 
-#include <CL/sycl.hpp>
+#include <sycl/sycl.hpp>
 #include <chrono>
 
 #include "jacobi.h"
@@ -63,7 +63,7 @@ int main(int argc, char **argv) {
   double *b = NULL;
   float *A = NULL;
 
-  queue q{default_selector(), property::queue::in_order()};
+  queue q{default_selector_v, property::queue::in_order()};
 
   b = (double *)malloc(N_ROWS * sizeof(double));
   memset(b, 0, N_ROWS * sizeof(double));

DirectProgramming/DPC++/DenseLinearAlgebra/jacobi_iterative/sycl_migrated_optimized/CMakeLists.txt

@@ -1,4 +1,4 @@
-set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O2")
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O2 -fsycl")
 set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS}")
 
 add_executable (sycl_migrated_optimized src/main.cpp src/jacobi.cpp)

DirectProgramming/DPC++/DenseLinearAlgebra/jacobi_iterative/sycl_migrated_optimized/src/jacobi.cpp

@@ -33,7 +33,7 @@
  */
 
 
-#include <CL/sycl.hpp>
+#include <sycl/sycl.hpp>
 
 #include "jacobi.h"
 
@@ -181,9 +181,9 @@ double JacobiMethodGpu(const float *A, const double *b,
 
     if ((k & 1) == 0) {
       q.submit([&](handler &cgh) {
-        accessor<double, 1, access_mode::read_write, access::target::local>
+        local_accessor<double, 1>
             x_shared_acc_ct1(range<1>(N_ROWS), cgh);
-        accessor<double, 1, access_mode::read_write, access::target::local>
+        local_accessor<double, 1>
             b_shared_acc_ct1(range<1>(ROWS_PER_CTA + 1), cgh);
 
         cgh.parallel_for(nd_range<3>(nblocks * nthreads, nthreads),
@@ -197,9 +197,9 @@ double JacobiMethodGpu(const float *A, const double *b,
       });
     } else {
       q.submit([&](handler &cgh) {
-        accessor<double, 1, access_mode::read_write, access::target::local>
+        local_accessor<double, 1>
             x_shared_acc_ct1(range<1>(N_ROWS), cgh);
-        accessor<double, 1, access_mode::read_write, access::target::local>
+        local_accessor<double, 1>
             b_shared_acc_ct1(range<1>(ROWS_PER_CTA + 1), cgh);
 
         cgh.parallel_for(nd_range<3>(nblocks * nthreads, nthreads),
@@ -224,7 +224,7 @@ double JacobiMethodGpu(const float *A, const double *b,
 
       if ((k & 1) == 0) {
         q.submit([&](handler &cgh) {
-          accessor<uint8_t, 1, access_mode::read_write, access::target::local>
+          local_accessor<uint8_t, 1>
               sycl_local_acc_ct1(range<1>(sharedMemSize), cgh);
 
           cgh.parallel_for(nd_range<3>(nblocks * nthreads, nthreads),
@@ -236,7 +236,7 @@ double JacobiMethodGpu(const float *A, const double *b,
         });
       } else {
         q.submit([&](handler &cgh) {
-          accessor<uint8_t, 1, access_mode::read_write, access::target::local>
+          local_accessor<uint8_t, 1>
               sycl_local_acc_ct1(range<1>(sharedMemSize), cgh);
 
           cgh.parallel_for(nd_range<3>(nblocks * nthreads, nthreads),

DirectProgramming/DPC++/DenseLinearAlgebra/jacobi_iterative/sycl_migrated_optimized/src/main.cpp

@@ -33,7 +33,7 @@
  */
 
 
-#include <CL/sycl.hpp>
+#include <sycl/sycl.hpp>
 #include <chrono>
 
 #include "jacobi.h"
@@ -63,7 +63,7 @@ int main(int argc, char **argv) {
   double *b = NULL;
   float *A = NULL;
 
-  queue q{default_selector(), property::queue::in_order()};
+  queue q{default_selector_v, property::queue::in_order()};
 
   b = (double *)malloc(N_ROWS * sizeof(double));
   memset(b, 0, N_ROWS * sizeof(double));