Add a test which calculate vectorDot product

vyast-softserveinc · Sep 24, 2024 · bfa53b3 · bfa53b3
1 parent 7f82ac6
commit bfa53b3
Show file tree

Hide file tree

Showing 3 changed files with 103 additions and 32 deletions.
diff --git a/examples/cpp/31_oklt_v3_moving_avg/CMakeLists.txt b/examples/cpp/31_oklt_v3_moving_avg/CMakeLists.txt
@@ -2,9 +2,10 @@ compile_cpp_example_with_modes(oklt_v3_moving_avg main.cpp)
 
 add_custom_target(cpp_example_oklt_v3_moving_avg_cpy ALL 
 	COMMAND ${CMAKE_COMMAND} -E copy ${CMAKE_CURRENT_SOURCE_DIR}/constants.h constants.h
-	COMMAND ${CMAKE_COMMAND} -E copy ${CMAKE_CURRENT_SOURCE_DIR}/movingAverage.okl movingAverage.okl)
+	COMMAND ${CMAKE_COMMAND} -E copy ${CMAKE_CURRENT_SOURCE_DIR}/movingAverage.okl movingAverage.okl
+	COMMAND ${CMAKE_COMMAND} -E copy ${CMAKE_CURRENT_SOURCE_DIR}/vectorDot.okl vectorDot.okl)
 add_dependencies(examples_cpp_oklt_v3_moving_avg cpp_example_oklt_v3_moving_avg_cpy)
 target_sources(examples_cpp_oklt_v3_moving_avg 
-	PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/movingAverage.okl
-        PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/constants.h
-	)
+  PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/movingAverage.okl
+  PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/vectorDot.okl
+  PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/constants.h)
diff --git a/examples/cpp/31_oklt_v3_moving_avg/main.cpp b/examples/cpp/31_oklt_v3_moving_avg/main.cpp
@@ -3,9 +3,25 @@
 #include <vector>
 #include "constants.h"
 
-std::vector<float> buildData(std::size_t size,
-                             float initialValue,
-                             float fluctuation)
+bool starts_with(const std::string &str, const std::string &substring) {
+    return str.rfind(substring, 0) == 0;
+}
+
+occa::json getDeviceOptions(int argc, const char **argv) {
+    for(int i  = 0; i < argc; ++i) {
+        std::string argument(argv[i]);
+        if((starts_with(argument,"-d") || starts_with(argument, "--device")) && i + 1 < argc)
+        {
+            std::string value(argv[i + 1]);
+            return occa::json::parse(value);
+        }
+    }
+    return occa::json::parse("{mode: 'Serial'}");
+}
+
+std::vector<float> buildMovingAverageData(std::size_t size,
+                                          float initialValue,
+                                          float fluctuation)
 {
     std::vector<float> buffer(size);
     float currentValue = initialValue;
@@ -34,36 +50,13 @@ std::vector<float> goldMovingAverage(const std::vector<float> &hostVector) {
     return result;
 }
 
-bool starts_with(const std::string &str, const std::string &substring) {
-    return str.rfind(substring, 0) == 0;
-}
-
-occa::json getDeviceOptions(int argc, const char **argv) {
-    for(int i  = 0; i < argc; ++i) {
-        std::string argument(argv[i]);
-        if((starts_with(argument,"-d") || starts_with(argument, "--device")) && i + 1 < argc)
-        {
-            std::string value(argv[i + 1]);
-            return occa::json::parse(value);
-        }
-    }
-    return occa::json::parse("{mode: 'Serial'}");
-}
-
-int main(int argc, const char **argv) {
-
-  occa::json deviceOpts = getDeviceOptions(argc, argv);
-  auto inputHostBuffer = buildData(THREADS_PER_BLOCK * WINDOW_SIZE + WINDOW_SIZE, 10.0f, 4.0f);
+int runMovingAverageTest(occa::device &device, occa::json &buildProps) {
+  auto inputHostBuffer = buildMovingAverageData(THREADS_PER_BLOCK * WINDOW_SIZE + WINDOW_SIZE, 10.0f, 4.0f);
   std::vector<float> outputHostBuffer(inputHostBuffer.size() - WINDOW_SIZE);
 
-  occa::device device(deviceOpts);
   occa::memory deviceInput = device.malloc<float>(inputHostBuffer.size());
   occa::memory deviceOutput = device.malloc<float>(outputHostBuffer.size());
 
-  occa::json buildProps({
-      {"transpiler-version", 3}
-  });
-
   occa::kernel movingAverageKernel = device.buildKernel("movingAverage.okl", "movingAverage32f", buildProps);
 
   deviceInput.copyFrom(inputHostBuffer.data(), inputHostBuffer.size());
@@ -86,8 +79,57 @@ int main(int argc, const char **argv) {
             return 1;
         }
   }
+
   std::cout << "Comparison with gold has passed" << std::endl;
   std::cout << "Moving average finished" << std::endl;
 
   return 0;
 }
+
+int runVectorDotTest(occa::device &device, occa::json &buildProps) {
+  const std::size_t size = 1e7;
+  auto vecA = std::vector<double>(size);
+  auto vecB = std::vector<double>(size);
+  auto vecT = std::vector<double>((size + THREADS_PER_BLOCK - 1) / THREADS_PER_BLOCK);
+
+  for (std::size_t i = 0; i < vecA.size(); i++)
+    vecA[i] = i + 1, vecB[i] = size - (i + 1);
+
+  occa::memory deviceVecA = device.malloc<double>(vecA.size());
+  deviceVecA.copyFrom(vecA.data(), vecA.size());
+
+  occa::memory deviceVecB = device.malloc<double>(vecB.size());
+  deviceVecB.copyFrom(vecB.data(), vecB.size());
+
+  occa::memory deviceVecT = device.malloc<double>(vecT.size(), 0);
+
+  occa::kernel vectorDotKernel = device.buildKernel("vectorDot.okl", "vectorDot",
+    buildProps);
+
+  vectorDotKernel(deviceVecT, deviceVecA.size(), deviceVecA, deviceVecB);
+
+  deviceVecT.copyTo(vecT.data(), deviceVecT.size());
+
+  double dot = 0;
+  for (std::size_t i = 0; i < vecT.size(); i++)
+    dot += vecT[i];
+
+  const double exact = (size * (size + 1.0) * (size - 1.0)) / 6;
+  return (std::fabs(dot - exact)/exact > 1e-8);
+}
+
+int main(int argc, const char **argv) {
+
+  occa::json deviceOpts = getDeviceOptions(argc, argv);
+  occa::device device(deviceOpts);
+
+  occa::json buildProps({
+      {"transpiler-version", 3}
+  });
+
+  int failure = 0;
+  failure |= runMovingAverageTest(device, buildProps);
+  failure |= runVectorDotTest(device, buildProps);
+
+  return failure;
+}
diff --git a/examples/cpp/31_oklt_v3_moving_avg/vectorDot.okl b/examples/cpp/31_oklt_v3_moving_avg/vectorDot.okl
@@ -0,0 +1,28 @@
+#include "constants.h"
+
+@kernel void vectorDot(double *temp, const unsigned int n, const double *a,
+    const double *b) {
+  for (unsigned int i = 0; i < (n + THREADS_PER_BLOCK - 1) / THREADS_PER_BLOCK; i++; @outer) {
+    @shared float s[THREADS_PER_BLOCK];
+
+    for (int j = 0; j < THREADS_PER_BLOCK; j++; @inner) {
+      int t = i * THREADS_PER_BLOCK + j;
+      if (t < n)
+        s[j] = a[t] * b[t];
+      else
+        s[j] = 0.0;
+    }
+
+    for (int k = (THREADS_PER_BLOCK + 1) / 2; k > 0; k /= 2) {
+      for (int j = 0; j < THREADS_PER_BLOCK; j++; @inner) {
+        if (j < k)
+          s[j] += s[j + k];
+      }
+    }
+
+    for (int j = 0; j < THREADS_PER_BLOCK; j++; @inner) {
+      if (j == 0)
+        temp[i] = s[0];
+    }
+  }
+}