Merge pull request intel#893 from bb-sycl/xmain

Auto pulldown and update tc files for xmain branch on 20220423

Author: myler

.github/CODEOWNERS

@@ -1,9 +1,8 @@
-* @vladimirlaz @romanovvlad @bader
+* @pvchupin
 
 # Use runtime team as the umbrella for most of the tests
 /SYCL/ @intel/llvm-reviewers-runtime
 
-
 # SYCL sub-directory matchers are grouped by code owner first, followed by
 # alphabetical order within the group. Please, keep this ordering.
 

SYCL/Basic/buffer/buffer_create.cpp

@@ -15,20 +15,17 @@ int main() {
   buffer<::cl_int, 1> Buffer(Size);
   Queue.submit([&](handler &cgh) {
     accessor Accessor{Buffer, cgh, read_write};
-    if (NumOfDevices > 1)
-      // Currently the Level Zero plugin uses host allocations for multi-device
-      // contexts because such allocations are accessible by all devices.
-      std::cerr << "Multi GPU should use zeMemAllocHost\n";
-    else if (D.get_info<info::device::host_unified_memory>())
+    if (D.get_info<info::device::host_unified_memory>())
       std::cerr << "Integrated GPU should use zeMemAllocHost\n";
     else
       std::cerr << "Discrete GPU should use zeMemAllocDevice\n";
-    cgh.parallel_for<class CreateBuffer>(range<1>(Size), [=](id<1> ID) {});
+    cgh.parallel_for<class CreateBuffer>(range<1>(Size),
+                                         [=](id<1> ID) { Accessor[ID] = 0; });
   });
   Queue.wait();
 
   return 0;
 }
 
-// CHECK: {{Integrated|Multi|Discrete}} GPU should use [[API:zeMemAllocHost|zeMemAllocHost|zeMemAllocDevice]]
+// CHECK: {{Integrated|Discrete}} GPU should use [[API:zeMemAllocHost|zeMemAllocDevice]]
 // CHECK: ZE ---> [[API]](

SYCL/Basic/buffer/buffer_migrate.cpp

@@ -0,0 +1,47 @@
+// REQUIRES: gpu
+// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
+// RUN: %GPU_RUN_PLACEHOLDER %t.out
+// RUN: env NEOReadDebugKeys=1 CreateMultipleRootDevices=2 %GPU_RUN_PLACEHOLDER %t.out
+// RUN: env NEOReadDebugKeys=1 CreateMultipleRootDevices=3 %GPU_RUN_PLACEHOLDER %t.out
+//
+// Test for buffer use in a context with multiple devices (all found
+// root-devices)
+//
+
+#include <CL/sycl.hpp>
+using namespace cl::sycl;
+
+int main() {
+
+  int Data = 0;
+  int Result = 0;
+  buffer<int, 1> Buffer(&Data, range<1>(1));
+
+  const auto &Devices =
+      platform(gpu_selector{}).get_devices(info::device_type::gpu);
+  std::cout << Devices.size() << " devices found" << std::endl;
+  context C(Devices);
+
+  int Index = 0;
+  for (auto D : Devices) {
+    std::cout << "Using on device " << Index << ": "
+              << D.get_info<info::device::name>() << std::endl;
+    Result |= (1 << Index);
+
+    queue Q(C, D);
+    Q.submit([&](handler &cgh) {
+      accessor Accessor{Buffer, cgh, read_write};
+      cgh.parallel_for<class MigrateBuffer>(
+          range<1>(1), [=](id<1> ID) { Accessor[ID] |= (1 << Index); });
+    });
+    Q.wait();
+    ++Index;
+  }
+
+  auto HostAcc = Buffer.get_host_access();
+  auto Passed = (HostAcc[0] == Result);
+  std::cout << "Checking result on host: " << (Passed ? "passed" : "FAILED")
+            << std::endl;
+  std::cout << HostAcc[0] << " ?= " << Result << std::endl;
+  return !Passed;
+}

SYCL/Basic/context-with-multiple-devices.cpp

@@ -1,3 +1,6 @@
+// UNSUPPORTED: windows || linux
+//   temporarily disabled
+
 // REQUIRES: accelerator, opencl-aot
 
 // RUN: %clangxx -fsycl -fintelfpga -fsycl-unnamed-lambda %s -o %t2.out

SYCL/Basic/interop/check_carrying_real_kernel_IDs.cpp

@@ -1,3 +1,6 @@
+// UNSUPPORTED: windows || linux
+//   temporarily disabled
+
 // REQUIRES: opencl, opencl_icd
 // RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out %opencl_lib
 // RUN: %ACC_RUN_PLACEHOLDER %t.out

SYCL/DiscardEvents/invalid_event_exceptions.cpp

@@ -1,3 +1,6 @@
+// UNSUPPORTED: windows || linux
+//   temporarily disabled
+
 // RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
 //
 // RUN: %HOST_RUN_PLACEHOLDER %t.out

SYCL/ESIMD/api/functional/ctors/ctor_load.hpp

@@ -46,12 +46,7 @@ struct vector {
   static constexpr auto get_value() { return esimd::vector_aligned; }
 };
 
-template <unsigned int size = 16 /*oword alignment*/> struct overal {
-  // Use 16 instead of std::max_align_t because of the fact that long double is
-  // not a native type in Intel GPUs. So 16 is not driven by any type, but
-  // rather the "oword alignment" requirement for all block loads. In that
-  // sense, std::max_align_t would give wrong idea.
-
+template <unsigned int size> struct overal {
   static std::string to_string() {
     return "overaligned<" + std::to_string(size) + ">";
   }

SYCL/ESIMD/api/functional/ctors/ctor_load_acc_core.cpp

@@ -47,12 +47,13 @@ int main(int, char **) {
   const auto offset_generators = ctors::coverage::get_all_offset_generators();
 
   // Run over all types with element and vector alignment
-  // Default tag for overaligned is configured in a way appropriate for any
-  // type from the type coverage, so run over it as well
+  // Using overaligned tag with alignment requirements appropriate for any
+  // type from the type coverage
   {
+    constexpr auto oword = 16;
     const auto alignments =
         named_type_pack<ctors::alignment::element, ctors::alignment::vector,
-                        ctors::alignment::overal<>>::generate();
+                        ctors::alignment::overal<oword>>::generate();
 
     // Run for all combinations possible
     passed &= for_all_combinations<ctors::run_test>(

SYCL/ESIMD/api/functional/ctors/ctor_load_acc_fp_extra.cpp

@@ -47,12 +47,13 @@ int main(int, char **) {
   const auto offset_generators = ctors::coverage::get_all_offset_generators();
 
   // Run over all types with element and vector alignment
-  // Default tag for overaligned is configured in a way appropriate for any
-  // type from the type coverage, so run over it as well
+  // Using overaligned tag with alignment requirements appropriate for any
+  // type from the type coverage
   {
+    constexpr auto oword = 16;
     const auto alignments =
         named_type_pack<ctors::alignment::element, ctors::alignment::vector,
-                        ctors::alignment::overal<>>::generate();
+                        ctors::alignment::overal<oword>>::generate();
 
     // Run for all combinations possible
     passed &= for_all_combinations<ctors::run_test>(

SYCL/ESIMD/api/functional/ctors/ctor_load_usm_core.cpp

@@ -30,6 +30,7 @@ int main(int, char **) {
                     esimd_test::createExceptionHandler());
 
   bool passed = true;
+  constexpr auto oword = 16;
 
   const auto types = get_tested_types<tested_types::core>();
   const auto sizes = get_all_sizes();
@@ -39,7 +40,7 @@ int main(int, char **) {
                         ctors::rval_in_expr, ctors::const_ref>::generate();
   const auto alignments =
       named_type_pack<ctors::alignment::element, ctors::alignment::vector,
-                      ctors::alignment::overal<>>::generate();
+                      ctors::alignment::overal<oword>>::generate();
 
   passed &= for_all_combinations<ctors::run_test>(types, sizes, contexts,
                                                   alignments, queue);

SYCL/ESIMD/api/functional/ctors/ctor_load_usm_fp_extra.cpp

@@ -30,6 +30,8 @@ int main(int, char **) {
                     esimd_test::createExceptionHandler());
 
   bool passed = true;
+  constexpr auto oword = 16;
+
   const auto types = get_tested_types<tested_types::fp_extra>();
   const auto dims = get_all_dimensions();
 
@@ -38,7 +40,7 @@ int main(int, char **) {
                         ctors::rval_in_expr, ctors::const_ref>::generate();
   const auto alignments =
       named_type_pack<ctors::alignment::element, ctors::alignment::vector,
-                      ctors::alignment::overal<>>::generate();
+                      ctors::alignment::overal<oword>>::generate();
 
   passed &= for_all_combinations<ctors::run_test>(types, dims, contexts,
                                                   alignments, queue);

SYCL/ESIMD/api/functional/functions/common.hpp

@@ -19,8 +19,7 @@ namespace esimd_test::api::functional::functions {
 
 namespace esimd = sycl::ext::intel::esimd;
 
-template <int NumElems, int NumSelectedElems, int Stride, int Offset,
-          typename ContextT>
+template <int NumElems, typename ContextT>
 class TestDescription : public ITestDescription {
 public:
   TestDescription(const std::string &data_type) : m_data_type(data_type) {}
@@ -30,10 +29,6 @@ class TestDescription : public ITestDescription {
 
     test_description += m_data_type + ", " + std::to_string(NumElems) + ">";
     test_description += ", with context: " + ContextT::get_description();
-    test_description +=
-        ", with size selected elems: " + std::to_string(NumSelectedElems);
-    test_description += ", with stride: " + std::to_string(Stride);
-    test_description += ", with offset: " + std::to_string(Offset);
 
     return test_description;
   }

SYCL/ESIMD/api/functional/functions/functions_1d_select.hpp

@@ -109,8 +109,7 @@ class run_test {
   static constexpr int NumSelectedElems = NumSelectedElemsT::value;
   static constexpr int Stride = StrideT::value;
   static constexpr int Offset = OffsetT::value;
-  using TestDescriptionT =
-      TestDescription<NumElems, NumSelectedElems, Stride, Offset, TestCaseT>;
+  using TestDescriptionT = TestDescription<NumElems, TestCaseT>;
 
 public:
   bool operator()(sycl::queue &queue, const std::string &data_type) {
@@ -198,7 +197,9 @@ class run_test {
   bool fail_test(size_t i, DataT expected, DataT retrieved,
                  const std::string &data_type) {
     log::fail(TestDescriptionT(data_type), "Unexpected value at index ", i,
-              ", retrieved: ", retrieved, ", expected: ", expected);
+              ", retrieved: ", retrieved, ", expected: ", expected,
+              ", with size selected elems: ", NumSelectedElems,
+              ", with stride: ", Stride, ", with offset: ", Offset);
 
     return false;
   }