SWDEV-396533 - Add test for intrinsic shfl API

catch/unit/cooperativeGrps/simple_shfl_functions.cc

@@ -0,0 +1,126 @@
+/*
+Copyright (c) 2023 Advanced Micro Devices, Inc. All rights reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.
+*/
+
+// Test Description:
+/* This test implements simple tests to verfy intrinsic shuffle functions.*/
+
+#include <hip_test_common.hh>
+#include <stdio.h>
+#include <vector>
+
+/* Test basic shfl functionality.
+ *
+ */
+void __global__ shfl_test(int *presult, int delta, int width) {
+  int id = threadIdx.x + blockIdx.x * blockDim.x;
+  int val = __shfl(id, id + delta, width);
+  presult[id] = val;
+}
+
+void __global__ shfl_up_test(int *presult, int delta, int width) {
+  int id = threadIdx.x + blockIdx.x * blockDim.x;
+  int val = __shfl_up(id, delta, width);
+  presult[id] = val;
+}
+
+void __global__ shfl_down_test(int *presult, int delta, int width) {
+  int id = threadIdx.x + blockIdx.x * blockDim.x;
+  int val = __shfl_down(id, delta, width);
+  presult[id] = val;
+}
+
+TEST_CASE("Unit_shfl_function") {
+
+  hipDeviceProp_t devProp;
+  HIP_CHECK(hipGetDeviceProperties(&devProp, 0))
+  unsigned int wave_size = devProp.warpSize;
+  size_t pow2 = static_cast<int>(log2(wave_size));
+  std::vector<size_t> width;
+  std::vector<size_t> delta;
+  int blockSize = 1;
+  int threadsPerBlock = wave_size;
+  int totalThreads = blockSize * threadsPerBlock;
+  int* dResults = NULL;
+  int* hPtr = NULL;
+  bool testpassed = true;
+
+  for (int i = 0; i < pow2; i++) {
+    width.push_back(pow(2, i));
+    delta.push_back(pow(2, i) + 1);
+  }
+
+  HIPCHECK(hipHostMalloc(&hPtr, sizeof(int) * totalThreads));
+  HIPCHECK(hipMalloc(&dResults, sizeof(int) * totalThreads));
+  SECTION("test shfl function") {
+    for (auto w : width) {
+      for (auto d : delta) {
+        hipLaunchKernelGGL(shfl_test, blockSize, threadsPerBlock,
+                          threadsPerBlock * sizeof(int), 0, dResults, d, w);
+        hipDeviceSynchronize();
+        HIPCHECK(hipMemcpy(hPtr, dResults, totalThreads * sizeof(int), hipMemcpyDeviceToHost));
+
+        for (size_t tid = 0; tid < wave_size; ++tid) {
+          auto src = ((tid + d) % w) + (tid / w) * w;
+          if (hPtr[tid] != src) {
+            testpassed = false;
+          }
+        }
+      }
+    }
+    REQUIRE(testpassed == true);
+  }
+  SECTION("test shfl up function") {
+    for (auto w : width) {
+      for (auto d : delta) {
+        hipLaunchKernelGGL(shfl_up_test, blockSize, threadsPerBlock,
+                          threadsPerBlock * sizeof(int), 0, dResults, d, w);
+        HIPCHECK(hipDeviceSynchronize());
+        HIPCHECK(hipMemcpy(hPtr, dResults, totalThreads * sizeof(int), hipMemcpyDeviceToHost));
+        for (size_t tid = d; tid < wave_size; ++tid) {
+          auto src = ((tid - d) < ((tid / w)*w)) ? tid : (tid - d);
+          if (hPtr[tid] != src) {
+            testpassed = false;
+          }
+        }
+      }
+    }
+    REQUIRE(testpassed == true);
+  }
+  SECTION("test shfl down function") {
+    for (auto w : width) {
+      for (auto d : delta) {
+        hipLaunchKernelGGL(shfl_down_test, blockSize, threadsPerBlock,
+                          threadsPerBlock * sizeof(int), 0, dResults, d, w);
+        HIPCHECK(hipDeviceSynchronize());
+        HIPCHECK(hipMemcpy(hPtr, dResults, totalThreads * sizeof(int), hipMemcpyDeviceToHost));
+        for (size_t tid = 0; tid < wave_size; ++tid) {
+          auto src = (((tid % w) + d) >= w) ? tid : tid + d;
+          if (hPtr[tid] != src) {
+            testpassed = false;
+          }
+        }
+      }
+    }
+    REQUIRE(testpassed == true);
+  }
+  HIPCHECK(hipFree(dResults));
+}