Download QNICE processor files            


Donate money            


Project detail and discuss            


Get support            



Join this project:

To join this project, please contact the project administrators of of this project, as shown on the project summary page.

Get the source code:

Source code for this project may be available as downloads or through one of the SCM repositories used by the project, as accessible from the project develop page.

QNICE - an Elegant 16 Bit Processor

QNICE is a rather simple, yet powerful 16 bit processor architecture which is a result of some research done in the late 1990s and the following years.

The goal was to create a processor architectur which is as orthogonal as possible without complicating a hardware implementation unnecessarily. Since the design borrows heavily from the NICE processor which featured a 32 bit architecture, the resulting architecture was called QNICE for "Quite NICE".

The architecture is very well suited for teaching the basics of computer architecture and hardware implementation. Furthermore programming the QNICE processor is really easy and thus the architecture already served as a training device for some lectures about programming as well as computer architecture.

QNICE at a Glance
  • 16 bit architecture with RISC like instruction set (18 instructions so far):

    • MOVE src, dst:     dst := src
    • ADD src, dst:     dst := dst + src
    • ADDC src, dst:     dst := dst + src + C
    • SUB src, dst:     dst := dst - src
    • SUBC src, dst:     dst := dst - src - C
    • SHL src, dst:     dst << src, fill with X bit, shift to C bit
    • SHR src, dst:     dst >> src, fill with C bit, shift to X bit
    • SWAP src, dst:     dst := ((src << 8) & 0xFF00 | ((src >> 8) & 0xFF
    • NOT src, dst:     dst := !src
    • AND src, dst:     dst := dst & src
    • OR src, dst:     dst := dst | src
    • XOR src, dst:     dst := dst ^ src
    • CMP src, dst:     dst := -src
    • HALT:     Halt the processor
    • ABRA src, [!]cond:     Absolute branch, conditional
    • ASUB src, [!]cond:     Absolute subroutine call, conditional
    • RBRA src, [!]cond:     Relative branch, conditional
    • RSUB src, [!]cond:     Relative subroutine call, conditional

  • Four addressing modes (register direct, register indirect, register indirect with post increment and register indirect with pre decrement): Rxx, @Rxx, @Rxx++ and @--Rxx.
  • Each instruction occupies one 16 bit word and has a fixed instruction format (excluding branches/jumps) thus decoding instructions is easy:

  • Branch and subroutine call instructions share a common instruction format, too:

  • 16 registers divided into to areas: R0 to R7 are in fact a window to a register bank containing 256 times 8 registers while R8 to R15 are fixed. This architecture makes subroutine calls and saving registers very easy (just increment/decrement the register bank pointer which is part of the status register). All in all QNICE features 256 * 8 + 8 = 2056 registers.
  • Special registers: R13 serves as a stack pointer (used by the xSUB instructions), R14 is the status register which also controls the register bank windows described above while R15 is the program counter.
  • Branch/subroutinecall instructions can be executed conditionally, controlled by the status bits of the status register.
  • Address space is 64 kWords of 16 bits each.
  • Memory mapped IO.


0F80  0000          MOVE 0x0000, R0
0F84  1000          MOVE 0x1000, R1
1100        LOOP    ADD R1, R0
3F84  0001          SUB 0x0001, R1
FF8B  0004          ABRA LOOP, !Z
E000                HALT

The QNICE assembler program shown on the left sums all values from 0x0000 to 0x1000.

This example is really simple - the main idea is to count backwards from 0x1000 to 0x0000 and using the Z (zero) flag of the status register R14 as a control condition for the central loop.

Due to the possibility of making all branches and subroutine calls conditional by prefixing them with the name of one of the eight status register status bits the loop effectively consists of only two instructions, a subtraction and an absolute branch (ABRA).

Prefixing the flag controlling the instruction with an exclamation mark inverts the flag (without modifying the status register's contents) prior to testing thus the instruction ABRA LOOP, !Z will perform an absolute branch to the location labeled LOOP if the zero flag is not set.

Subroutine Calls

The code example below is a bit more complex and shows a subroutine which performs a string comparison (like the C standard library function strcmp). This function expects two pointers to the strings to be compared in the registers R9 and R10 and will return the result of the comparison in R8.

Since the routine needs some registers for temporary storage of data is has to make sure that the contents of these registers are somehow saved at the entry of the subroutine and restored just before jumping back to the calling program.

In traditional architectures this would involve pushing the registers to a stack (either explicitly, register by register, or controlled by a bit mask). Since QNICE features a register bank for its registers R0 to R7 saving and restoring the contents of these eight registers just makes it necessary to increment the register bank pointer which is made up by the eight upper bits of the status register R14 thus giving the routine access to its "own" set of registers R0 to R7. Before returning to the calling program the register bank pointer will be decremented making sure that the original register contents will be accessible again.

Incrementing the register bank pointer is done by performing the instruction ADD 0x0100, R14, decrementing it by one is done at the label STR$_STRCMP_EXIT using the subtraction SUB 0x0100, R14.

;* STR$STRCMP compares two strings
;* R9: Pointer to the first string (S0),
;* R10: Pointer to the second string (S1),
;* R8: negative if (S0 < S1), zero if (S0 == S1), positive if (S0 > S1)
;* The contents of R8 and R9 are being preserved during the run of 
;* this function
STR$STRCMP       ADD 0x0100, R14          ; Get a new register page
                 MOVE R9, R0              ; Save arguments
                 MOVE R10, R1
STR$_STRCMP_LOOP MOVE @R0, R8             ; while (*s1 == *s2++)
                 MOVE @R1++, R2
                 SUB R8, R2
                 RBRA STR$_STRCMP_END, !Z
                 MOVE @R0++, R8           ; if (*s1++ == 0)
                 RBRA STR$_STRCMP_EXIT, Z ;   return 0;
                 RBRA STR$_STRCMP_LOOP, 1 ; end-of-while-loop
STR$_STRCMP_END  MOVE @--R1, R2           ; return (*s1 - (--*s2));
                 SUB R2, R8
STR$_STRCMP_EXIT SUB 0x0100, R14          ; Restore previous register page
                 MOVE @R13++, R15         ; and return to calling program part

A typical call to this routine would look like this:

RSUB STR$STRCMP, 1      ; Was a halt command issued?
Contents of the Distribution Kit

The QNICE distribution kit currently contains the following items:

  • A short documentation of the processor architecture in form of a slide show prepared for a talk at a German DECUS symposium.
  • The source code of a QNICE cross assembler written entirely in pure C (this has been proven to run on LINUX, Mac OS X, OpenVMS and Windows).
  • The source code of the QNICE emulator, written in pure C, too. This emulator not only emulates the processor itself but also a 2681 UART (serial input/output is read from stdin/written to stdout) as well as a simple IDE disk. Both devices are memory mapped in the last 1 k Block of memory. (Since the UART emulation required the use of the curses library, the portability is a bit questionable - on systems like OpenVMS the usage of the UART part can be inhibited by undefining a constant in qnice.c.)
  • Some introductory examples like the summation program shown above and the first couple of routines for a monitor program.

Current state of the project

The project is currently in the alpha state - the assembler and emulator are both fully operational. The next steps will involve the following tasks:

  • Write a complete monitor program based on qmon1.asm found in the examples directory.
  • Create a hardware implementation of QNICE - either using TTL technology or (preferred) an FPGA.

Currently our main goal is to interest people in the QNICE architecture in general and in the implementation of the processor in hardware. If you are interested in joining the project, please contact the project administrators as shown on the project summary page.