bytes_and_flops with CMake

Fix some warnings as well
ldh4 · Mar 4, 2024 · 2773390 · 2773390
1 parent e836198
commit 2773390
Show file tree

Hide file tree

Showing 10 changed files with 32 additions and 27 deletions.
diff --git a/benchmarks/CMakeLists.txt b/benchmarks/CMakeLists.txt
@@ -1,3 +1,4 @@
+KOKKOS_ADD_BENCHMARK_DIRECTORIES(bytes_and_flops)
 KOKKOS_ADD_BENCHMARK_DIRECTORIES(gups)
 KOKKOS_ADD_BENCHMARK_DIRECTORIES(launch_latency)
 KOKKOS_ADD_BENCHMARK_DIRECTORIES(stream)
diff --git a/benchmarks/bytes_and_flops/CMakeLists.txt b/benchmarks/bytes_and_flops/CMakeLists.txt
@@ -0,0 +1,4 @@
+KOKKOS_ADD_EXECUTABLE(
+  bytes_and_flops
+  SOURCES bench_double.cpp bench_float.cpp bench_int32_t.cpp bench_int64_t.cpp main.cpp
+)
diff --git a/benchmarks/bytes_and_flops/bench.hpp b/benchmarks/bytes_and_flops/bench.hpp
@@ -37,22 +37,22 @@ struct RunStride {
 };
 
 #define STRIDE 1
-#include <bench_stride.hpp>
+#include "bench_stride.hpp"
 #undef STRIDE
 #define STRIDE 2
-#include <bench_stride.hpp>
+#include "bench_stride.hpp"
 #undef STRIDE
 #define STRIDE 4
-#include <bench_stride.hpp>
+#include "bench_stride.hpp"
 #undef STRIDE
 #define STRIDE 8
-#include <bench_stride.hpp>
+#include "bench_stride.hpp"
 #undef STRIDE
 #define STRIDE 16
-#include <bench_stride.hpp>
+#include "bench_stride.hpp"
 #undef STRIDE
 #define STRIDE 32
-#include <bench_stride.hpp>
+#include "bench_stride.hpp"
 #undef STRIDE
 
 template <class Scalar>

diff --git a/benchmarks/bytes_and_flops/bench_double.cpp b/benchmarks/bytes_and_flops/bench_double.cpp
@@ -14,7 +14,7 @@
 //
 //@HEADER
 
-#include <bench.hpp>
+#include "bench.hpp"
 
 template void run_stride_unroll<double>(int N, int K, int R, int D, int U,
                                         int F, int T, int S, int B, int I);
diff --git a/benchmarks/bytes_and_flops/bench_float.cpp b/benchmarks/bytes_and_flops/bench_float.cpp
@@ -14,7 +14,7 @@
 //
 //@HEADER
 
-#include <bench.hpp>
+#include "bench.hpp"
 
 template void run_stride_unroll<float>(int N, int K, int R, int D, int U, int F,
                                        int T, int S, int B, int I);
diff --git a/benchmarks/bytes_and_flops/bench_int32_t.cpp b/benchmarks/bytes_and_flops/bench_int32_t.cpp
@@ -14,7 +14,7 @@
 //
 //@HEADER
 
-#include <bench.hpp>
+#include "bench.hpp"
 
 template void run_stride_unroll<int32_t>(int N, int K, int R, int D, int U,
                                          int F, int T, int S, int B, int I);
diff --git a/benchmarks/bytes_and_flops/bench_int64_t.cpp b/benchmarks/bytes_and_flops/bench_int64_t.cpp
@@ -14,7 +14,7 @@
 //
 //@HEADER
 
-#include <bench.hpp>
+#include "bench.hpp"
 
 template void run_stride_unroll<int64_t>(int N, int K, int R, int D, int U,
                                          int F, int T, int S, int B, int I);
diff --git a/benchmarks/bytes_and_flops/bench_stride.hpp b/benchmarks/bytes_and_flops/bench_stride.hpp
@@ -15,28 +15,28 @@
 //@HEADER
 
 #define UNROLL 1
-#include <bench_unroll_stride.hpp>
+#include "bench_unroll_stride.hpp"
 #undef UNROLL
 #define UNROLL 2
-#include <bench_unroll_stride.hpp>
+#include "bench_unroll_stride.hpp"
 #undef UNROLL
 #define UNROLL 3
-#include <bench_unroll_stride.hpp>
+#include "bench_unroll_stride.hpp"
 #undef UNROLL
 #define UNROLL 4
-#include <bench_unroll_stride.hpp>
+#include "bench_unroll_stride.hpp"
 #undef UNROLL
 #define UNROLL 5
-#include <bench_unroll_stride.hpp>
+#include "bench_unroll_stride.hpp"
 #undef UNROLL
 #define UNROLL 6
-#include <bench_unroll_stride.hpp>
+#include "bench_unroll_stride.hpp"
 #undef UNROLL
 #define UNROLL 7
-#include <bench_unroll_stride.hpp>
+#include "bench_unroll_stride.hpp"
 #undef UNROLL
 #define UNROLL 8
-#include <bench_unroll_stride.hpp>
+#include "bench_unroll_stride.hpp"
 #undef UNROLL
 
 template <class Scalar>

diff --git a/benchmarks/bytes_and_flops/bench_unroll_stride.hpp b/benchmarks/bytes_and_flops/bench_unroll_stride.hpp
@@ -26,7 +26,7 @@ struct Run<Scalar, UNROLL, STRIDE> {
     Kokkos::deep_copy(C, Scalar(3.5));
 
     Kokkos::Timer timer;
-    for (int i = 0; i < I; ++i) {
+    for (int iter = 0; iter < I; ++iter) {
       Kokkos::parallel_for(
           "BenchmarkKernel",
           Kokkos::TeamPolicy<>(N, T).set_scratch_size(0, Kokkos::PerTeam(S)),
@@ -87,25 +87,25 @@ struct Run<Scalar, UNROLL, STRIDE> {
                     C(n, i, 0) = a1;
 #endif
 #if (UNROLL == 2)
-                    C(n, i, 0) = a1 + a2;
+                    C(n, iter, 0) = a1 + a2;
 #endif
 #if (UNROLL == 3)
-                    C(n, i, 0) = a1 + a2 + a3;
+                    C(n, iter, 0) = a1 + a2 + a3;
 #endif
 #if (UNROLL == 4)
-                    C(n, i, 0) = a1 + a2 + a3 + a4;
+                    C(n, iter, 0) = a1 + a2 + a3 + a4;
 #endif
 #if (UNROLL == 5)
-                    C(n, i, 0) = a1 + a2 + a3 + a4 + a5;
+                    C(n, iter, 0) = a1 + a2 + a3 + a4 + a5;
 #endif
 #if (UNROLL == 6)
-                    C(n, i, 0) = a1 + a2 + a3 + a4 + a5 + a6;
+                    C(n, iter, 0) = a1 + a2 + a3 + a4 + a5 + a6;
 #endif
 #if (UNROLL == 7)
-                    C(n, i, 0) = a1 + a2 + a3 + a4 + a5 + a6 + a7;
+                    C(n, iter, 0) = a1 + a2 + a3 + a4 + a5 + a6 + a7;
 #endif
 #if (UNROLL == 8)
-                    C(n, i, 0) = a1 + a2 + a3 + a4 + a5 + a6 + a7 + a8;
+                    C(n, iter, 0) = a1 + a2 + a3 + a4 + a5 + a6 + a7 + a8;
 #endif
                   });
             }

diff --git a/benchmarks/bytes_and_flops/main.cpp b/benchmarks/bytes_and_flops/main.cpp
@@ -16,7 +16,7 @@
 
 #include <Kokkos_Core.hpp>
 #include <Kokkos_Timer.hpp>
-#include <bench.hpp>
+#include "bench.hpp"
 #include <cstdlib>
 
 extern template void run_stride_unroll<float>(int, int, int, int, int, int, int,