GitHub-Proxy/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-06-04 18:54:00 +08:00
Code Issues Packages Projects Releases Wiki Activity

Added benchmarks for type casting of float16

Browse Source
This commit is contained in:
Benoit Steiner 2016-02-26 12:24:58 -08:00
parent 002824e32d
commit 93485d86bc
2 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
bench/tensors/tensor_benchmarks.h
View File

@ -48,12 +48,12 @@ template <typename Device, typename T> class BenchmarkSuite {
Eigen::array<TensorIndex, 2> sizes; Eigen::array<TensorIndex, 2> sizes;
sizes[0] = m_; sizes[0] = m_;
sizes[1] = k_; sizes[1] = k_;
const TensorMap<Tensor<T, 2, 0, TensorIndex>, Eigen::Aligned> A(a_, sizes); const TensorMap<Tensor<int, 2, 0, TensorIndex>, Eigen::Aligned> A((int*)a_, sizes);
TensorMap<Tensor<int, 2, 0, TensorIndex>, Eigen::Aligned> B((int*)b_, sizes); TensorMap<Tensor<T, 2, 0, TensorIndex>, Eigen::Aligned> B(b_, sizes);
StartBenchmarkTiming(); StartBenchmarkTiming();
for (int iter = 0; iter < num_iters; ++iter) { for (int iter = 0; iter < num_iters; ++iter) {
B.device(device_) = A.template cast<int>(); B.device(device_) = A.template cast<T>();
} }
// Record the number of values copied per second // Record the number of values copied per second
finalizeBenchmark(static_cast<int64_t>(m_) * k_ * num_iters); finalizeBenchmark(static_cast<int64_t>(m_) * k_ * num_iters);

2
bench/tensors/tensor_benchmarks_fp16_gpu.cu
View File

@ -19,7 +19,7 @@
BENCHMARK_RANGE(BM_##FUNC, 10, 5000); BENCHMARK_RANGE(BM_##FUNC, 10, 5000);
BM_FuncGPU(memcpy); BM_FuncGPU(memcpy);
//BM_FuncGPU(typeCasting); BM_FuncGPU(typeCasting);
//BM_FuncGPU(random); //BM_FuncGPU(random);
BM_FuncGPU(slicing); BM_FuncGPU(slicing);
BM_FuncGPU(rowChip); BM_FuncGPU(rowChip);