MSOL16 is a graphics acceleration chip, which can be integrated
in XPRO32Cache. Insert it by connecting the output signals of
MEMCON bigmem to MSOL16SOLVE20, MSOL16SOLVE12, MSOL16MULT20
and MSOL16MULT12 in sequence and the last ones' outputs to the
signals the bigmem signals pointed to or test it without connection.
The four chips jump their program counter to the value of the first input,
load the next 12 or 20 halfwords and store 3 or 4 words as
result. One result starts at one word. The next word is the
address, where the next input waits or zero, if the
thread is inactive. XPRO32Cache and MSOL16 work independently
and use the same addresses. The start and the length of
the MSOL16 memory can be changed with memstart and membits.
Every thread has got its own memory. The initial values of those
words can be changed in the modules S20MEMn, S12MEMn,
M20MEMn, and M12MEMn. If you want true results, you must
replace the DUMMY instances by SOLEON16Bit adders, subtracters,
multipliers and dividers. The program is too slow for
current personal computers then.