Interrupt tasks with assembly code – Echelon Neuron User Manual

Page 49

pushd [SCRATCH16] ; (r(2), piDiv(2))

pushd [SCRATCH16] ; (r(2), r(2), piDiv(2))

; push the “second half” of pi:

pushd #PIFACTOR ; (PiFac(2), r(2), r(2), PiDiv(2))

; do the math:

call _mul16 ; ((piFac*r)(2), r(2), piDiv(2))

brf muldiv ; ( -- carea)

For this implementation, the Neuron assembly file imports the two external

functions: _mul16 and muldiv. The _mul16 function has a preceding

underscore because it is intended for use only with assembly functions. Note that

neither the _mul16 nor the muldiv function names has a percent sign (%) prefix.

The _mul16 system function is not defined in, or designed for, the Neuron C

application name space. The muldiv function is designed for the Neuron C

application name space, but is imported using the system keyword in Neuron C,

which makes the symbol available without the percent sign (%) prefix.
The assembly file then defines two mnemonics for the pi value approximation,

and defines a mnemonic for one of the general-purpose 16-bit registers. This

value as a pointer, these registers can hold any 16-bit value, even though they

are typically used for pointer values.
The %carea function is then declared, using the percent sign (%) prefix, with the

APEXP directive to make this symbol available to the Neuron C application

name space.
The %carea function then sets up the stack for calling the _mul16 and muldiv

functions. Note that because the _mul16 function is located in the near code

area, it can be called with the two-byte CALL instruction, rather than the

slightly less efficient three-byte CALLF instruction.
By setting the stack up correctly, the carea function can branch into the muldiv

function (using the BRF instruction), rather than having to call the muldiv

function (using the CALLF and RET instructions). The carea function can use

the branch instruction because the stack layout for executing the muldiv

function is correct, and contains only the arguments for the muldiv function.

Also, the result type of the carea function is the same as that of the muldiv

function.
For more information about the _mul16 function, see _mul16 (Multiply, 16 Bit).

For more information about the muldiv() function, see the Neuron C Reference

Guide.

Interrupt Tasks with Assembly Code

For Series 5000 and 6000 devices, you can write an interrupt task to handle

hardware interrupts. Although you write the interrupt task in Neuron C, you

can include a Neuron assembly function in the Neuron C task using the

#pragma include_assembly_file compiler directive, or call a function written

in assembly from your interrupt task.
In general, interrupts run in a separate processor context on the Series 5000 and

6000 Chips, but for the lowest two clock rates (clock multiplier values ½ and 1 in

Neuron Assembly Language Reference