From 8cb9bfab870c1f55ea9c69233a832e92c8de189d Mon Sep 17 00:00:00 2001
From: Benoit Steiner <benoit.steiner.goog@gmail.com>
Date: Tue, 23 Feb 2016 05:28:02 +0000
Subject: Extended the tensor benchmark suite to support types other than
 floats

---
 bench/tensors/tensor_benchmarks_cpu.cc | 42 +++++++++++++++++-----------------
 1 file changed, 21 insertions(+), 21 deletions(-)

(limited to 'bench/tensors/tensor_benchmarks_cpu.cc')

diff --git a/bench/tensors/tensor_benchmarks_cpu.cc b/bench/tensors/tensor_benchmarks_cpu.cc
index 6754e1a32..8947f4b7f 100644
--- a/bench/tensors/tensor_benchmarks_cpu.cc
+++ b/bench/tensors/tensor_benchmarks_cpu.cc
@@ -9,13 +9,13 @@ Eigen::ThreadPool pool(threads);                \
 Eigen::ThreadPoolDevice device(&pool, threads);
 
 // Simple functions
-#define BM_FuncCPU(FUNC, THREADS)                                \
-  static void BM_##FUNC##_##THREADS##T(int iters, int N) {       \
-    StopBenchmarkTiming();                                       \
-    CREATE_THREAD_POOL(THREADS);                                 \
-    BenchmarkSuite<Eigen::ThreadPoolDevice> suite(device, N);    \
-    suite.FUNC(iters);                                           \
-  }                                                              \
+#define BM_FuncCPU(FUNC, THREADS)                                    \
+  static void BM_##FUNC##_##THREADS##T(int iters, int N) {           \
+    StopBenchmarkTiming();                                           \
+    CREATE_THREAD_POOL(THREADS);                                     \
+    BenchmarkSuite<Eigen::ThreadPoolDevice, float> suite(device, N); \
+    suite.FUNC(iters);                                               \
+  }                                                                  \
   BENCHMARK_RANGE(BM_##FUNC##_##THREADS##T, 10, 5000);
 
 BM_FuncCPU(memcpy, 4);
@@ -80,19 +80,19 @@ BM_FuncCPU(colReduction, 12);
 
 
 // Contractions
-#define BM_FuncWithInputDimsCPU(FUNC, D1, D2, D3, THREADS)                     \
-  static void BM_##FUNC##_##D1##x##D2##x##D3##_##THREADS##T(int iters, int N) {\
-    StopBenchmarkTiming();                                                     \
-    if (THREADS == 1) {                                                        \
-      Eigen::DefaultDevice device;                                             \
-      BenchmarkSuite<Eigen::DefaultDevice> suite(device, D1, D2, D3);          \
-      suite.FUNC(iters);                                                       \
-    } else {                                                                   \
-      CREATE_THREAD_POOL(THREADS);                                             \
-      BenchmarkSuite<Eigen::ThreadPoolDevice> suite(device, D1, D2, D3);       \
-      suite.FUNC(iters);                                                       \
-    }                                                                          \
-  }                                                                            \
+#define BM_FuncWithInputDimsCPU(FUNC, D1, D2, D3, THREADS)                      \
+  static void BM_##FUNC##_##D1##x##D2##x##D3##_##THREADS##T(int iters, int N) { \
+    StopBenchmarkTiming();                                                      \
+    if (THREADS == 1) {                                                         \
+      Eigen::DefaultDevice device;                                              \
+      BenchmarkSuite<Eigen::DefaultDevice, float> suite(device, D1, D2, D3);    \
+      suite.FUNC(iters);                                                        \
+    } else {                                                                    \
+      CREATE_THREAD_POOL(THREADS);                                              \
+      BenchmarkSuite<Eigen::ThreadPoolDevice, float> suite(device, D1, D2, D3); \
+      suite.FUNC(iters);                                                        \
+    }                                                                           \
+  }                                                                             \
   BENCHMARK_RANGE(BM_##FUNC##_##D1##x##D2##x##D3##_##THREADS##T, 10, 5000);
 
 
@@ -138,7 +138,7 @@ BM_FuncWithInputDimsCPU(contraction, N, N, 1, 16);
   static void BM_##FUNC##_##DIM1##x##DIM2##_##THREADS##T(int iters, int N) {   \
     StopBenchmarkTiming();                                                     \
     CREATE_THREAD_POOL(THREADS);                                               \
-    BenchmarkSuite<Eigen::ThreadPoolDevice> suite(device, N);                  \
+    BenchmarkSuite<Eigen::ThreadPoolDevice, float> suite(device, N);	       \
     suite.FUNC(iters, DIM1, DIM2);                                             \
   }                                                                            \
   BENCHMARK_RANGE(BM_##FUNC##_##DIM1##x##DIM2##_##THREADS##T, 128, 5000);
-- 
cgit v1.2.3