vexexec executes a cvex shader on the command line. It can be used to debug simple VEX code without needing to start Houdini or another heavyweight context, for example by using the printf function.

The -j option specifies the number of CPU threads to run.

The -p option specifies the VEX array size on which the shader will execute (by default, the shader will be executed as a scalar program - a default of 1). Larger values of -p can be used to test how the program behaves or performs with a larger array size, which will be more indicative of how shaders are executed in Houdini and mantra. For example, the VOP SOP will have an array size equal to the number of points in the geometry.

The -t option turns on timing of the VEX execution.

The -v option specifies a comma separated list of additional inputs to bind.

The -V option specifies verbosity when running VEX. The argument to the -V option can consist of a combination of:

• 0:

  Set verbose level to 0.

• 1:

  Set verbose level to 1.

• 2:

  Set verbose level to 2.

• c:

  Enable coloring of log messages (default).

• C:

  Disable coloring of log messages.

• t:

  Enable time stamps on log messages (default).

• T:

  Disable time stamps on log messages.

• p:

  Enable VEX profiling.

• P:

  Enable VEX profiling and NAN detection.

For example:

• -V 2PC:

  Set verbosity to 2, perform VEX profiling and NAN detection and disable log message coloring.

There are a few predefined shader parameter bindings supported by vexexec that can be used to manage the context of the shader. You can add new bindings with the -v command-line option.

These values are queried by cvex after the shader executes:

cvex simple(int proc = 0; export float result = 0)
{
    result = proc * proc;
    printf("proc = %g, proc^2 = %g\n", proc, result);
}

Save to simple.vfl and execute the following commands:

vcc simple.vfl
vexexec -p3 simple.vex

Output:

proc = 0, proc^2 = 0

proc = 1, proc^2 = 1

proc = 2, proc^2 = 4