devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity

Added support for fast integer divisions by a constant

Browse Source 
Sped up tensor slicing by a factor of 3 by using these fast integer divisions.
This commit is contained in:
Benoit Steiner
2014-08-14 22:13:21 -07:00
parent 756292f8aa
commit 33c702c79f
5 changed files with 177 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
unsupported/test/CMakeLists.txt
View File

@@ -106,6 +106,7 @@ if(EIGEN_TEST_CXX11)
ei_add_test(cxx11_tensor_convolution "-std=c++0x")
ei_add_test(cxx11_tensor_expr "-std=c++0x")
# ei_add_test(cxx11_tensor_fixed_size "-std=c++0x")
ei_add_test(cxx11_tensor_intdiv "-std=c++0x")
ei_add_test(cxx11_tensor_lvalue "-std=c++0x")
ei_add_test(cxx11_tensor_map "-std=c++0x")
# ei_add_test(cxx11_tensor_morphing "-std=c++0x")

77
unsupported/test/cxx11_tensor_intdiv.cpp Normal file
View File

@@ -0,0 +1,77 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2014 Benoit Steiner <benoit.steiner.goog@gmail.com>
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#include "main.h"
#include <Eigen/CXX11/Tensor>
static void test_signed_32bit()
{
for (int32_t i = 1; i < 25000; ++i) {
const Eigen::internal::TensorIntDivisor<int32_t> div(i);
for (int32_t j = 0; j < 25000; ++j) {
const int32_t fast_div = j / div;
const int32_t slow_div = j / i;
VERIFY_IS_EQUAL(fast_div, slow_div);
}
}
}
static void test_unsigned_32bit()
{
for (uint32_t i = 1; i < 25000; ++i) {
const Eigen::internal::TensorIntDivisor<uint32_t> div(i);
for (uint32_t j = 0; j < 25000; ++j) {
const uint32_t fast_div = j / div;
const uint32_t slow_div = j / i;
VERIFY_IS_EQUAL(fast_div, slow_div);
}
}
}
static void test_signed_64bit()
{
for (int64_t i = 2; i < 25000; ++i) {
const Eigen::internal::TensorIntDivisor<int64_t> div(i);
for (int64_t j = 0; j < 25000; ++j) {
const int64_t fast_div = j / div;
const int64_t slow_div = j / i;
VERIFY_IS_EQUAL(fast_div, slow_div);
}
}
}
static void test_unsigned_64bit()
{
for (uint64_t i = 2; i < 25000; ++i) {
const Eigen::internal::TensorIntDivisor<uint64_t> div(i);
for (uint64_t j = 0; j < 25000; ++j) {
const uint64_t fast_div = j / div;
const uint64_t slow_div = j / i;
VERIFY_IS_EQUAL(fast_div, slow_div);
}
}
}
void test_cxx11_tensor_intdiv()
{
CALL_SUBTEST(test_signed_32bit());
CALL_SUBTEST(test_unsigned_32bit());
CALL_SUBTEST(test_signed_64bit());
CALL_SUBTEST(test_unsigned_64bit());
}