The Plasma processor is a public domain 32-bit soft processor designed by Steve Rhoads which implements most of the MIPS I (TM) instruction set. It contains a DDR SDRAM controller and a UART core. Because it has the same instruction set as a MIPS processor, it uses the same GNU toolchain. Additionally, a small RTOS with a TCP/IP stack is available in the opencores.org package, but it is not used here. The Plasma project overview is available from the opencores.org website here.

The design provided here simply instantiates the CPU module, links it to the LCD display. A small application is provided that initializes the LCD display and writes a few characters to it.

The plasma will synthetise as is for the Spartan 3E Starter Kit board.
To use the LCD, it is necessary to connect the LCD pins to the processor's GPIO out registers (this design only does LCD writes). This is done by simply adding the following code in the architecture of the top level module:

	LCD_D  <= gpio0_out(14 downto 11);
	LCD_RS <= gpio0_out(10);
	LCD_RW <= gpio0_out(9);
	LCD_E  <= gpio0_out(8);

Provided as a tool in the Plasma package is a serial communication program, ethermip, which dynamically loads a program executable by the Plasma CPU in memory and then listens on the serial interface, and additionally is also able to support Ethernet traffic over UART. I found it problematic to use as it will not run without WinPCAP or compile with GCC, so I wrote another version (without the Ethernet capability) which can be found here.

The sample program lcd_leds initializes the LCD, writes a few characters to it, and also flashes a few LEDS on the boards. It can be compiled with the following addition in the plasma provided makefile:

	lcd_leds:
	$(AS_MIPS) -o boot.o boot.asm
	$(GCC_MIPS) lcd_leds.c
	$(GCC_MIPS) no_os.c
	$(LD_MIPS) -Ttext 0x10000000 -eentry -Map test.map -s -N -o test.axf \
		boot.o lcd_leds.o no_os.o
	-$(DUMP_MIPS) --disassemble test.axf > test.lst
	convert_bin.exe
	$(CP) code.txt $(VHDL_DIR)
      

C and assembler programs are compiled differently depending on how the program will run. If the program is intended to be put in RAM during synthesis, then the entry point specified will be 0. If the program is intended to be downloaded to DDR SDRAM (with the plasma CPU running the bootldr application) then the entry point must be specified as 0x100000000, which is the memory entry point.

A program is put in RAM during synthesis as described in this comment that can be found in the plasma file ram_image.vhd:

--    Compile the MIPS C and assembly code into "test.axf".
--    Run convert.exe to change "test.axf" to "code.txt" which
--    will contain the hex values of the opcodes.
--    Next run "ram_image ram_xilinx.vhd code.txt ram_image.vhd",
--    to create the "ram_image.vhd" file that will have the opcodes
--    correctly placed inside the INIT_00 => strings.
--    Then include ram_image.vhd in the simulation/synthesis.
---------------------------------------------------------------------

In the makefile, look at the -Ttext option for the various test programs that can be built in order to know how they are supposed to run. "count" for example is compiled so as to be put in ram_image.vhd, whereas "count2" is intented to be loaded via the serial port when the plasma processor is synthetized with the ram_image.vhd application "bootldr". spartan_plasma.zip

---

Author: <lsilvestri@psilv.net>
2008/03/19 10:54:00 AM