There are two kinds of tests in these directories. The tests in conformance exercise almost all of the possible operations (it doesn't fully exercise the Riemann class, but that class has other problems, such as not being dimension-independent). The tests in speed compare how fast the FTensor's are compared to ordinary arrays.

You may have to edit the Makefile's in each of the directories to set the compiler (CXX) and its flags (CXXFLAGS). You can set optimization options only by changing (CXXOPTIMIZE). Alternately, you can specify them on the command line, as in

make CXX=<compiler> CXXOPTIMIZE="<flags>" test_compiler

The default compiler is g++.

In the speed directory, there is a makefile.vc specifically for Nmake and Microsoft Visual Studio. You will need to open a Visual Studio command line and build it with

NMAKE /f makefile.vc

conformance

Once the Makefile is set to your satisfaction, running the tests in conformance should be as simple as typing

make
test_all > test_results

The file "test_results" will contain thousands of lines which should say

PASS: (unique name of test)

If a test fails, it will print out

FAIL: (unique name of test)

None of the tests should fail.

speed

Running the tests in speed should just be

one_over_script

in Linux and

    one_over_ps.ps1

in Windows Powershell.

This will compile all of the programs and then run them. It will time each one, with the output looking like

fast 1

real    0m9.866s
user    0m9.720s
sys     0m0.000s
Tensor 1

real    0m19.891s
user    0m19.740s
sys     0m0.040s
.
.
.

and then followed by a series of lines that say PASS: (unique name) or FAIL:(unique name).

This is timing two different implementations of an equation. One uses ordinary arrays, and the other uses FTensor's. As you can see, the FTensor implementation can be much slower than the ordinary array. It depends on the compiler. The PASS and FAIL messages are calculated by comparing the final results of the two computations. All of the tests should pass.

The conformance tests check almost all of the possible operations to make sure that they are coded correctly and the compiler didn't do anything funny. If you're just interested in testing your compiler, a quick and dirty method is to make just the "little" and "littlefast" executables in the speed directory. If it compiles, they should output the same thing:

7e+08 9e+08 1.1e+09

A more complete test is to make "test_compiler" in the conformance directory. This will make a subset of the complete conformance test, enough to check every construct that is used in the code. It should compile much faster, especially with optimizations.