boostorg/multiprecision
1
0
Fork 0
mirror of https://github.com/boostorg/multiprecision.git synced 2026-07-21 13:23:49 +00:00
Code Issues Packages Projects Releases Wiki Activity

performance: Fix GCC compiles [CI SKIP]

Browse Source
This commit is contained in:
jzmaddock
2019-12-10 18:46:11 +00:00
parent 1c9bef59e1
commit 6d6f776b86
9 changed files with 1393 additions and 41 deletions
Download Patch File Download Diff File Expand all files Collapse all files
performance/Jamfile.v2
+2
View File
@@ -33,6 +33,7 @@ project : requirements
lib gmp ;
lib mpfr ;
lib quadmath ;
if $(tommath_path)
{
@@ -71,6 +72,7 @@ exe sf_performance : sf_performance.cpp sf_performance_basic.cpp sf_performance_
: release
[ check-target-builds ../config//has_gmp : <define>TEST_MPF <define>TEST_MPZ <source>gmp : ]
[ check-target-builds ../config//has_mpfr : <define>TEST_MPFR <source>mpfr : ]
[ check-target-builds ../config//has_float128 : <source>quadmath : ]
<define>TEST_CPP_DEC_FLOAT
<define>TEST_CPP_BIN_FLOAT
<define>TEST_FLOAT
performance/delaunay_test.log
+74
View File
@@ -0,0 +1,74 @@
loading data...
calculating...
Running calculations for: int64_t, int64_t
Number of calculations = 100000000
Number of flips = 25765000
Total execution time = 4.87739
Time per calculation = 4.87739e-08
Running calculations for: arithmetic_backend<int64_t>, arithmetic_backend<int64_t>
Number of calculations = 100000000
Number of flips = 25765000
Total execution time = 42.5469
Time per calculation = 4.25469e-07
Running calculations for: int64_t, arithmetic_backend<int64_t>
Number of calculations = 100000000
Number of flips = 25765000
Total execution time = 9.24092
Time per calculation = 9.24092e-08
Running calculations for: multiprecision::int64_t, multiprecision::int64_t
Number of calculations = 100000000
Number of flips = 24627000
Total execution time = 66.3086
Time per calculation = 6.63086e-07
Running calculations for: int64_t, int128_t
Number of calculations = 100000000
Number of flips = 23892000
Total execution time = 10.3215
Time per calculation = 1.03215e-07
Running calculations for: int64_t, boost::multiprecision::int128_t
Number of calculations = 100000000
Number of flips = 23892000
Total execution time = 12.848
Time per calculation = 1.2848e-07
Running calculations for: int64_t, int128_t (ET)
Number of calculations = 100000000
Number of flips = 23892000
Total execution time = 12.7641
Time per calculation = 1.27641e-07
Running calculations for: multiprecision::int64_t, multiprecision::int128_t
Number of calculations = 100000000
Number of flips = 23892000
Total execution time = 76.4744
Time per calculation = 7.64744e-07
Running calculations for: int64_t, cpp_int
Number of calculations = 100000000
Number of flips = 23892000
Total execution time = 17.5545
Time per calculation = 1.75545e-07
Running calculations for: int64_t, cpp_int (no ET's)
Number of calculations = 100000000
Number of flips = 23892000
Total execution time = 17.5139
Time per calculation = 1.75139e-07
Running calculations for: int64_t, cpp_int(128-bit cache)
Number of calculations = 100000000
Number of flips = 23892000
Total execution time = 17.4647
Time per calculation = 1.74647e-07
Running calculations for: int64_t, cpp_int (128-bit Cache no ET's)
Number of calculations = 100000000
Number of flips = 23892000
Total execution time = 18.6492
Time per calculation = 1.86492e-07
performance/miller_rabin_performance.hpp
+2
View File
@@ -81,6 +81,8 @@ boost::chrono::duration<double> test_miller_rabin(const char* name)
return t;
}
boost::chrono::duration<double> test_miller_rabin_gmp();
void test01();
void test02();
void test03();
performance/miller_rabin_performance.log
+40
View File
@@ -0,0 +1,40 @@
Time for cpp_int (no Expression templates) = 0.0632349
Number of primes found = 9
Time for cpp_int = 0.0597808
Number of primes found = 9
Time for cpp_int (128-bit cache) = 0.061515
Number of primes found = 9
Time for cpp_int (256-bit cache) = 0.061159
Number of primes found = 9
Time for cpp_int (512-bit cache) = 0.0521137
Number of primes found = 9
Time for cpp_int (1024-bit cache) = 0.051393
Number of primes found = 9
Time for uint1024_t = 0.047259
Number of primes found = 9
Time for checked_uint1024_t = 0.0477963
Number of primes found = 9
Time for mpz_int (no Expression templates) = 0.00748244
Number of primes found = 9
Time for mpz_int = 0.00692059
Number of primes found = 9
Time for mpz_int (native Miller Rabin Test) = 0.000384121 seconds
Time for tom_int (no Expression templates) = 0.0288748
Number of primes found = 9
Time for tom_int = 0.0288394
Number of primes found = 9
[table
[[Integer Type][Relative Performance (Actual time in parenthesis)]]
[[checked_uint1024_t][6.90638(0.0477963s)]]
[[cpp_int][8.63811(0.0597808s)]]
[[cpp_int (1024-bit cache)][7.4261(0.051393s)]]
[[cpp_int (128-bit cache)][8.88868(0.061515s)]]
[[cpp_int (256-bit cache)][8.83724(0.061159s)]]
[[cpp_int (512-bit cache)][7.53024(0.0521137s)]]
[[cpp_int (no Expression templates)][9.1372(0.0632349s)]]
[[mpz_int][1(0.00692059s)]]
[[mpz_int (no Expression templates)][1.08118(0.00748244s)]]
[[tom_int][4.16719(0.0288394s)]]
[[tom_int (no Expression templates)][4.1723(0.0288748s)]]
[[uint1024_t][6.82875(0.047259s)]]
]
performance/performance_test.cpp
+1 -1
View File
@@ -121,7 +121,7 @@ void quickbook_platform_details()
#endif
std::cout << "[[Compiler][" << BOOST_COMPILER << "]]\n";
#ifdef TEST_MPZ
std::cout << "[[GMP][" << GMP_VERSION << "]]\n";
std::cout << "[[GMP][" << gmp_version << "]]\n";
#endif
#ifdef TEST_MPFR
std::cout << "[[MPFR][" << MPFR_VERSION << "]]\n";
performance/performance_test.log
+1040
View File
File diff suppressed because it is too large Load Diff
performance/sf_performance.cpp
+1 -1
View File
@@ -5,7 +5,7 @@
#include "sf_performance.hpp"
boost::atomic<unsigned> allocation_count = 0;
boost::atomic<unsigned> allocation_count(0);
std::map<std::string, std::map<std::string, std::pair<double, unsigned> > > result_table;
void* (*alloc_func_ptr)(size_t);
performance/sf_performance.log
+224
View File
@@ -0,0 +1,224 @@
Allocation Counts for Horner Evaluation:
Allocation count for type mpfr_float_50 = 3
Allocation count for type mpfr_float_50 - no expression templates = 1
Allocation Counts for boost::math::tools::evaluate_polynomial:
Allocation count for type mpfr_float_50 = 13
Allocation count for type mpfr_float_50 - no expression templates = 9
Time for double = 0.000688113 seconds
Total allocations for double = 14
Time for real_concept = 0.00149196 seconds
Total allocations for real_concept = 0
Time for arithmetic_backend<double> = 0.00133409 seconds
Total allocations for arithmetic_backend<double> = 0
Time for arithmetic_backend<double> - no expression templates = 0.000637403 seconds
Total allocations for arithmetic_backend<double> - no expression templates = 0
Time for double = 0.0366824 seconds
Total allocations for double = 0
Time for real_concept = 0.0867356 seconds
Total allocations for real_concept = 0
Time for arithmetic_backend<double> = 0.0956691 seconds
Total allocations for arithmetic_backend<double> = 0
Time for arithmetic_backend<double> - no expression templates = 0.040428 seconds
Total allocations for arithmetic_backend<double> - no expression templates = 0
Testing Bessel Functions at 50 digits.....
Time for mpfr_float_50 = 0.204043 seconds
Total allocations for mpfr_float_50 = 813605
Time for mpfr_float_50 (3 theads) = 0.238272 seconds
Total allocations for mpfr_float_50 = 3255602
Time for mpfr_float_50 (no expression templates = 0.216221 seconds
Total allocations for mpfr_float_50 (no expression templates = 1001796
Time for mpfr_float_50 (no expression templates (3 theads) = 0.267326 seconds
Total allocations for mpfr_float_50 (no expression templates = 4008534
Time for static_mpfr_float_50 = 0.146508 seconds
Total allocations for static_mpfr_float_50 = 18685
Time for static_mpfr_float_50 (3 theads) = 0.180008 seconds
Total allocations for static_mpfr_float_50 = 75946
Time for static_mpfr_float_50 (no expression templates) = 0.143153 seconds
Total allocations for static_mpfr_float_50 (no expression templates) = 46826
Time for static_mpfr_float_50 (no expression templates) (3 theads) = 0.161564 seconds
Total allocations for static_mpfr_float_50 (no expression templates) = 189191
Time for mpf_float_50 = 0.167874 seconds
Total allocations for mpf_float_50 = 870726
Time for mpf_float_50 (3 theads) = 0.205629 seconds
Total allocations for mpf_float_50 = 3482852
Time for mpf_float_50 (no expression templates = 0.171495 seconds
Total allocations for mpf_float_50 (no expression templates = 1071229
Time for mpf_float_50 (no expression templates (3 theads) = 0.22507 seconds
Total allocations for mpf_float_50 (no expression templates = 4284874
Time for cpp_dec_float_50 = 0.377193 seconds
Total allocations for cpp_dec_float_50 = 3
Time for cpp_dec_float_50 (3 theads) = 0.410275 seconds
Total allocations for cpp_dec_float_50 = 36
Time for cpp_bin_float_50 = 0.315701 seconds
Total allocations for cpp_bin_float_50 = 4
Time for cpp_bin_float_50 (3 theads) = 0.338848 seconds
Total allocations for cpp_bin_float_50 = 43
Testing Bessel Functions at 100 digits.....
Time for mpfr_float_100 = 0.504992 seconds
Total allocations for mpfr_float_100 = 1666200
Time for mpfr_float_100 (no expression templates = 0.503061 seconds
Total allocations for mpfr_float_100 (no expression templates = 2120706
Time for static_mpfr_float_100 = 0.375447 seconds
Total allocations for static_mpfr_float_100 = 19391
Time for mpf_float_100 = 0.406251 seconds
Total allocations for mpf_float_100 = 1817791
Time for mpf_float_100 (no expression templates = 0.418247 seconds
Total allocations for mpf_float_100 (no expression templates = 2221056
Time for cpp_dec_float_100 = 1.13054 seconds
Total allocations for cpp_dec_float_100 = 3
Time for cpp_bin_float_100 = 1.19595 seconds
Total allocations for cpp_bin_float_100 = 6
Testing Polynomial Evaluation at 50 digits.....
Time for mpfr_float_50 = 0.00172618 seconds
Total allocations for mpfr_float_50 = 12976
Time for mpfr_float_50 (no expression templates = 0.00140102 seconds
Total allocations for mpfr_float_50 (no expression templates = 8984
Time for static_mpfr_float_50 = 0.000774017 seconds
Total allocations for static_mpfr_float_50 = 0
Time for mpf_float_50 = 0.00122821 seconds
Total allocations for mpf_float_50 = 12976
Time for mpf_float_50 (no expression templates = 0.000938302 seconds
Total allocations for mpf_float_50 (no expression templates = 8984
Time for cpp_dec_float_50 = 0.000761543 seconds
Total allocations for cpp_dec_float_50 = 0
Time for cpp_bin_float_50 = 0.00100391 seconds
Total allocations for cpp_bin_float_50 = 0
Testing Polynomial Evaluation at 100 digits.....
Time for mpfr_float_100 = 0.00198759 seconds
Total allocations for mpfr_float_100 = 12976
Time for mpfr_float_100 (no expression templates = 0.00197391 seconds
Total allocations for mpfr_float_100 (no expression templates = 8984
Time for static_mpfr_float_100 = 0.00115947 seconds
Total allocations for static_mpfr_float_100 = 0
Time for mpf_float_100 = 0.00130242 seconds
Total allocations for mpf_float_100 = 12976
Time for mpf_float_100 (no expression templates = 0.000967883 seconds
Total allocations for mpf_float_100 (no expression templates = 8984
Time for cpp_dec_float_100 = 0.00110386 seconds
Total allocations for cpp_dec_float_100 = 0
Time for cpp_bin_float_100 = 0.00144396 seconds
Total allocations for cpp_bin_float_100 = 0
Testing Non-Central T at 50 digits.....
Time for mpfr_float_50 = 45.8621 seconds
Total allocations for mpfr_float_50 = 196986977
Time for mpfr_float_50 (no expression templates = 41.5062 seconds
Total allocations for mpfr_float_50 (no expression templates = 194554719
Time for static_mpfr_float_50 = 31.1014 seconds
Total allocations for static_mpfr_float_50 = 1798554
Time for mpf_float_50 = 28.2162 seconds
Total allocations for mpf_float_50 = 162346817
Time for mpf_float_50 (no expression templates = 24.8611 seconds
Total allocations for mpf_float_50 (no expression templates = 154896018
Time for cpp_dec_float_50 = 61.5326 seconds
Total allocations for cpp_dec_float_50 = 4
Time for cpp_bin_float_50 = 59.0179 seconds
Total allocations for cpp_bin_float_50 = 1
Time for cpp_bin_float_50 (3 theads) = 79.9276 seconds
Total allocations for cpp_bin_float_50 = 28
Testing Non-Central T at 100 digits.....
Time for mpfr_float_100 = 92.8415 seconds
Total allocations for mpfr_float_100 = 333396307
Time for mpfr_float_100 (no expression templates = 84.5783 seconds
Total allocations for mpfr_float_100 (no expression templates = 325676494
Time for static_mpfr_float_100 = 69.8614 seconds
Total allocations for static_mpfr_float_100 = 15326962
Time for mpf_float_100 = 60.903 seconds
Total allocations for mpf_float_100 = 280052298
Time for mpf_float_100 (no expression templates = 52.3144 seconds
Total allocations for mpf_float_100 (no expression templates = 265239868
Time for cpp_dec_float_100 = 184.481 seconds
Total allocations for cpp_dec_float_100 = 5
Time for cpp_bin_float_100 = 180.534 seconds
Total allocations for cpp_bin_float_100 = 2
[table Bessel Functions (100 digit precision)
[[Type][Time][# Allocations]]
[[cpp_bin_float_100][3.18541 (1.19595s)][6]]
[[cpp_dec_float_100][3.01119 (1.13054s)][3]]
[[mpf_float_100][1.08204 (0.406251s)][1817791]]
[[mpf_float_100 (no expression templates][1.114 (0.418247s)][2221056]]
[[mpfr_float_100][1.34504 (0.504992s)][1666200]]
[[mpfr_float_100 (no expression templates][1.3399 (0.503061s)][2120706]]
[[static_mpfr_float_100][1 (0.375447s)][19391]]
]
[table Bessel Functions (16 digit precision)
[[Type][Time][# Allocations]]
[[arithmetic_backend<double>][2.09301 (0.00133409s)][0]]
[[arithmetic_backend<double> - no expression templates][1 (0.000637403s)][0]]
[[double][1.07956 (0.000688113s)][14]]
[[real_concept][2.34069 (0.00149196s)][0]]
]
[table Bessel Functions (50 digit precision)
[[Type][Time][# Allocations]]
[[cpp_bin_float_50][2.20534 (0.315701s)][4]]
[[cpp_bin_float_50 (3 concurrent threads)][2.36703 (0.338848s)][44]]
[[cpp_dec_float_50][2.6349 (0.377193s)][3]]
[[cpp_dec_float_50 (3 concurrent threads)][2.86599 (0.410275s)][37]]
[[mpf_float_50][1.17269 (0.167874s)][870726]]
[[mpf_float_50 (3 concurrent threads)][1.43643 (0.205629s)][3482853]]
[[mpf_float_50 (no expression templates][1.19799 (0.171495s)][1071229]]
[[mpf_float_50 (no expression templates (3 concurrent threads)][1.57224 (0.22507s)][4284875]]
[[mpfr_float_50][1.42535 (0.204043s)][813605]]
[[mpfr_float_50 (3 concurrent threads)][1.66446 (0.238272s)][3255603]]
[[mpfr_float_50 (no expression templates][1.51042 (0.216221s)][1001796]]
[[mpfr_float_50 (no expression templates (3 concurrent threads)][1.86742 (0.267326s)][4008535]]
[[static_mpfr_float_50][1.02344 (0.146508s)][18685]]
[[static_mpfr_float_50 (3 concurrent threads)][1.25745 (0.180008s)][75947]]
[[static_mpfr_float_50 (no expression templates)][1 (0.143153s)][46826]]
[[static_mpfr_float_50 (no expression templates) (3 concurrent threads)][1.12861 (0.161564s)][189192]]
]
[table Non-central T (16 digit precision)
[[Type][Time][# Allocations]]
[[arithmetic_backend<double>][2.60804 (0.0956691s)][0]]
[[arithmetic_backend<double> - no expression templates][1.10211 (0.040428s)][0]]
[[double][1 (0.0366824s)][0]]
[[real_concept][2.3645 (0.0867356s)][0]]
]
[table Non-central T Distribution (100 digit precision)
[[Type][Time][# Allocations]]
[[cpp_bin_float_100][3.45094 (180.534s)][2]]
[[cpp_dec_float_100][3.52639 (184.481s)][5]]
[[mpf_float_100][1.16417 (60.903s)][280052298]]
[[mpf_float_100 (no expression templates][1 (52.3144s)][265239868]]
[[mpfr_float_100][1.77469 (92.8415s)][333396307]]
[[mpfr_float_100 (no expression templates][1.61673 (84.5783s)][325676494]]
[[static_mpfr_float_100][1.33541 (69.8614s)][15326962]]
]
[table Non-central T Distribution (50 digit precision)
[[Type][Time][# Allocations]]
[[cpp_bin_float_50][2.3739 (59.0179s)][1]]
[[cpp_bin_float_50 (3 concurrent threads)][3.21496 (79.9276s)][29]]
[[cpp_dec_float_50][2.47505 (61.5326s)][4]]
[[mpf_float_50][1.13495 (28.2162s)][162346817]]
[[mpf_float_50 (no expression templates][1 (24.8611s)][154896018]]
[[mpfr_float_50][1.84473 (45.8621s)][196986977]]
[[mpfr_float_50 (no expression templates][1.66952 (41.5062s)][194554719]]
[[static_mpfr_float_50][1.25101 (31.1014s)][1798554]]
]
[table Polynomial Evaluation (100 digit precision)
[[Type][Time][# Allocations]]
[[cpp_bin_float_100][1.49187 (0.00144396s)][0]]
[[cpp_dec_float_100][1.14049 (0.00110386s)][0]]
[[mpf_float_100][1.34564 (0.00130242s)][12976]]
[[mpf_float_100 (no expression templates][1 (0.000967883s)][8984]]
[[mpfr_float_100][2.05354 (0.00198759s)][12976]]
[[static_mpfr_float_100][1.19794 (0.00115947s)][0]]
]
[table Polynomial Evaluation (50 digit precision)
[[Type][Time][# Allocations]]
[[cpp_bin_float_50][1.31826 (0.00100391s)][0]]
[[cpp_dec_float_50][1 (0.000761543s)][0]]
[[mpf_float_50][1.61279 (0.00122821s)][12976]]
[[mpf_float_50 (no expression templates][1.23211 (0.000938302s)][8984]]
[[mpfr_float_100 (no expression templates][2.59199 (0.00197391s)][8984]]
[[mpfr_float_50][2.26669 (0.00172618s)][12976]]
[[mpfr_float_50 (no expression templates][1.83971 (0.00140102s)][8984]]
[[static_mpfr_float_50][1.01638 (0.000774017s)][0]]
]
performance/sf_performance_basic.cpp
+9 -39
View File
@@ -7,43 +7,13 @@
void basic_tests()
{
std::cout << "Allocation Counts for Horner Evaluation:\n";
#ifdef TEST_MPFR
basic_allocation_test("mpfr_float_50", mpfr_float_50(2));
basic_allocation_test("mpfr_float_50 - no expression templates", number<mpfr_float_backend<50>, et_off>(2));
#endif
#ifdef TEST_MPFR_CLASS
basic_allocation_test("mpfr_class", mpfr_class(2));
#endif
#ifdef TEST_MPREAL
basic_allocation_test("mpfr::mpreal", mpfr::mpreal(2));
#endif
std::cout << "Allocation Counts for boost::math::tools::evaluate_polynomial:\n";
#ifdef TEST_MPFR
poly_allocation_test("mpfr_float_50", mpfr_float_50(2));
poly_allocation_test("mpfr_float_50 - no expression templates", number<mpfr_float_backend<50>, et_off>(2));
#endif
#ifdef TEST_MPFR_CLASS
poly_allocation_test("mpfr_class", mpfr_class(2));
#endif
#ifdef TEST_MPREAL
poly_allocation_test("mpfr::mpreal", mpfr::mpreal(2));
#endif
//
// Comparison for builtin floats:
//
#ifdef TEST_FLOAT
time_proc("Bessel Functions - double", test_bessel<double>);
time_proc("Bessel Functions - real_concept", test_bessel<boost::math::concepts::real_concept>);
time_proc("Bessel Functions - arithmetic_backend<double>", test_bessel<number<arithmetic_backend<double>, et_on> >);
time_proc("Bessel Functions - arithmetic_backend<double> - no expression templates", test_bessel<number<arithmetic_backend<double>, et_off> >);
time_proc("Non-central T - double", test_nct<double>);
time_proc("Non-central T - real_concept", test_nct<boost::math::concepts::real_concept>);
time_proc("Non-central T - arithmetic_backend<double>", test_nct<number<arithmetic_backend<double>, et_on> >);
time_proc("Non-central T - arithmetic_backend<double> - no expression templates", test_nct<number<arithmetic_backend<double>, et_off> >);
#endif
basic_tests_1();
basic_tests_2();
basic_tests_3();
basic_tests_4();
basic_tests_5();
basic_tests_6();
basic_tests_7();
basic_tests_8();
basic_tests_9();
}