[SYCL][ESIMD] Add test on simd load ctor (#769)

Author: vasilytric

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

@@ -0,0 +1,233 @@
+//==------- ctor_load_core.cpp  - DPC++ ESIMD on-device test ---------------==//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+// REQUIRES: gpu, level_zero
+// XREQUIRES: gpu
+// TODO gpu and level_zero in REQUIRES due to only this platforms supported yet.
+// The current "REQUIRES" should be replaced with "gpu" only as mentioned in
+// "XREQUIRES".
+// UNSUPPORTED: cuda, hip
+// RUN: %clangxx -fsycl %s -fsycl-device-code-split=per_kernel -o %t.out
+// RUN: %GPU_RUN_PLACEHOLDER %t.out
+//
+// Test for simd load constructor.
+// The test uses reference data and different alignment flags. Invokes simd
+// constructors in different contexts with provided reference data and alignment
+// flag.
+// It is expected for destination simd instance to store a bitwise same data as
+// the reference one.
+
+#include "common.hpp"
+
+using namespace sycl::ext::intel::experimental::esimd;
+using namespace esimd_test::api::functional;
+
+// Descriptor class for the case of calling constructor in initializer context.
+struct initializer {
+  static std::string get_description() { return "initializer"; }
+
+  template <typename DataT, int NumElems, typename AlignmentT>
+  static void call_simd_ctor(const DataT *ref_data, DataT *const out,
+                             AlignmentT alignment) {
+    simd<DataT, NumElems> simd_by_init =
+        simd<DataT, NumElems>(ref_data, alignment);
+    simd_by_init.copy_to(out);
+  }
+};
+
+// Descriptor class for the case of calling constructor in variable declaration
+// context.
+struct var_decl {
+  static std::string get_description() { return "variable declaration"; }
+
+  template <typename DataT, int NumElems, typename AlignmentT>
+  static void call_simd_ctor(const DataT *ref_data, DataT *const out,
+                             AlignmentT alignment) {
+    simd<DataT, NumElems> simd_by_var_decl(ref_data, alignment);
+    simd_by_var_decl.copy_to(out);
+  }
+};
+
+// Descriptor class for the case of calling constructor in rvalue in an
+// expression context.
+struct rval_in_expr {
+  static std::string get_description() { return "rvalue in an expression"; }
+
+  template <typename DataT, int NumElems, typename AlignmentT>
+  static void call_simd_ctor(const DataT *ref_data, DataT *const out,
+                             AlignmentT alignment) {
+    simd<DataT, NumElems> simd_by_rval;
+    simd_by_rval = simd<DataT, NumElems>(ref_data, alignment);
+    simd_by_rval.copy_to(out);
+  }
+};
+
+// Descriptor class for the case of calling constructor in const reference
+// context.
+class const_ref {
+public:
+  static std::string get_description() { return "const reference"; }
+
+  template <typename DataT, int NumElems, typename AlignmentT>
+  static void call_simd_ctor(const DataT *ref_data, DataT *const out,
+                             AlignmentT alignment) {
+    call_simd_by_const_ref<DataT, NumElems>(
+        simd<DataT, NumElems>(ref_data, alignment), out);
+  }
+
+private:
+  template <typename DataT, int NumElems>
+  static void
+  call_simd_by_const_ref(const simd<DataT, NumElems> &simd_by_const_ref,
+                         DataT *out) {
+    simd_by_const_ref.copy_to(out);
+  }
+};
+
+// Dummy kernel for submitting some code into device side.
+template <typename DataT, int NumElems, typename T, typename Alignment>
+struct Kernel_for_load_ctor;
+
+namespace alignment {
+
+struct element {
+  template <typename DataT, int> static size_t get_size() {
+    return alignof(DataT);
+  }
+  static constexpr auto get_value() { return element_aligned; }
+};
+
+struct vector {
+  template <typename DataT, int NumElems> static size_t get_size() {
+    // Referring to the simd class specialization on the host side is by design.
+    return alignof(simd<DataT, NumElems>);
+  }
+  static constexpr auto get_value() { return vector_aligned; }
+};
+
+struct overal {
+  template <typename, int> static size_t get_size() {
+    return alignof(std::max_align_t);
+  }
+  static constexpr auto get_value() {
+    return overaligned<alignof(std::max_align_t)>;
+  }
+};
+
+} // namespace alignment
+
+// The main test routine.
+// Using functor class to be able to iterate over the pre-defined data types.
+template <typename DataT, typename DimT, typename TestCaseT,
+          typename AlignmentT>
+class run_test {
+  static constexpr int NumElems = DimT::value;
+
+public:
+  bool operator()(sycl::queue &queue, const std::string &data_type) {
+    bool passed = true;
+    const std::vector<DataT> ref_data = generate_ref_data<DataT, NumElems>();
+
+    // If current number of elements is equal to one, then run test with each
+    // one value from reference data.
+    // If current number of elements is greater than one, then run tests with
+    // whole reference data.
+    if constexpr (NumElems == 1) {
+      for (size_t i = 0; i < ref_data.size(); ++i) {
+        passed = run_verification(queue, {ref_data[i]}, data_type);
+      }
+    } else {
+      passed = run_verification(queue, ref_data, data_type);
+    }
+    return passed;
+  }
+
+private:
+  bool run_verification(sycl::queue &queue, const std::vector<DataT> &ref_data,
+                        const std::string &data_type) {
+    assert(ref_data.size() == NumElems &&
+           "Reference data size is not equal to the simd vector length.");
+
+    bool passed = true;
+
+    const size_t alignment_value =
+        AlignmentT::template get_size<DataT, NumElems>();
+    const size_t container_extra_size = alignment_value / sizeof(DataT) + 1;
+    const size_t offset = 1;
+
+    shared_allocator<DataT> allocator(queue);
+    shared_vector<DataT> result(NumElems, allocator);
+    shared_vector<DataT> shared_ref_data(NumElems + container_extra_size +
+                                             offset,
+                                         shared_allocator<DataT>(queue));
+
+    const size_t object_size = NumElems * sizeof(DataT);
+    size_t buffer_size = object_size + container_extra_size * sizeof(DataT);
+
+    // When we allocate USM there is a high probability that this memory will
+    // have stronger alignment that required. We increment our pointer by fixed
+    // offset value to avoid bigger alignment of USM shared.
+    // The std::align can provide expected alignment on the small values of an
+    // alignment.
+    void *ref = shared_ref_data.data() + offset;
+    if (std::align(alignment_value, object_size, ref, buffer_size) == nullptr) {
+      return false;
+    }
+    DataT *const ref_aligned = static_cast<DataT *>(ref);
+
+    for (size_t i = 0; i < NumElems; ++i) {
+      ref_aligned[i] = ref_data[i];
+    }
+
+    queue.submit([&](sycl::handler &cgh) {
+      DataT *const out = result.data();
+
+      cgh.single_task<
+          Kernel_for_load_ctor<DataT, NumElems, TestCaseT, AlignmentT>>(
+          [=]() SYCL_ESIMD_KERNEL {
+            const auto alignment = AlignmentT::get_value();
+            TestCaseT::template call_simd_ctor<DataT, NumElems>(ref_aligned,
+                                                                out, alignment);
+          });
+    });
+    queue.wait_and_throw();
+
+    for (size_t i = 0; i < result.size(); ++i) {
+      if (!are_bitwise_equal(ref_data[i], result[i])) {
+        passed = false;
+
+        const auto description =
+            ctors::TestDescription<DataT, NumElems, TestCaseT>(
+                i, result[i], ref_data[i], data_type);
+        log::fail(description);
+      }
+    }
+
+    return passed;
+  }
+};
+
+int main(int, char **) {
+  sycl::queue queue(esimd_test::ESIMDSelector{},
+                    esimd_test::createExceptionHandler());
+
+  bool passed = true;
+  const auto types = get_tested_types<tested_types::all>();
+  const auto dims = get_all_dimensions();
+
+  const auto contexts = unnamed_type_pack<initializer, var_decl, rval_in_expr,
+                                          const_ref>::generate();
+  const auto alignments =
+      unnamed_type_pack<alignment::element, alignment::vector,
+                        alignment::overal>::generate();
+
+  passed &=
+      for_all_combinations<run_test>(types, dims, contexts, alignments, queue);
+
+  std::cout << (passed ? "=== Test passed\n" : "=== Test FAILED\n");
+  return passed ? 0 : 1;
+}