vladism / mark_ii Goto Github PK

CPU need access to larger data space. This space can be provided by external SDRAM. We need to write SDRAM driver for IS42S16160B that is located on DE0 nano board.

SDRAM datasheet: IS42S16160B.pdf

There are few bugs in IO board design.

• PS2 connector have wrong connection
• PS2 footprint have pretty stupid dimensions (PS2 con from ebay doesn't fit)

There is probably problem with gitignore. It is look like git is ignoring all changes of files in fpga/src directory.

There is bug in linker somewhere. In one file I have

.IMPORT main
CALL    main

in second file

.EXPORT main
main: 

linker is not able to deal with this properly.

MARK-II need I2S bus driver to be able produce sound on new board.

Consider support for cursor in VGA driver. VRAM is now consist of 15 bits wide words, Use bit 16 to implement "cursor enable". It should be implemented as blinking underscore.

There is something terrible wrong with labels in macros. Trying to compile this code

#macro jump
    OR R0 R0 R0
label:
    OR R0 R0 R0
    BZ R0 label
#endmacro

OR R1 R1 R1
$jump

And horrible things will happen:

$ m2-assembler test.asm 
Traceback (most recent call last):
  File "/opt/MARK_II_tools/bin/m2-assembler", line 773, in <module>
    main()
  File "/opt/MARK_II_tools/bin/m2-assembler", line 759, in main
    a.pass0()
  File "/opt/MARK_II_tools/bin/m2-assembler", line 56, in pass0
    self.t.parse(self.mainFileName)
  File "/opt/MARK_II_tools/src/assembler/tokenizer.py", line 127, in parse
    self.load_file(p)
  File "/opt/MARK_II_tools/src/assembler/tokenizer.py", line 286, in load_file
    macro_item.invoke(self.parser_buffer, given_args, line_counter, file_name)
  File "/opt/MARK_II_tools/src/assembler/tokenizer.py", line 96, in invoke
    found_arg, value_arg = self.find_arg_value(given_args, line_token)
  File "/opt/MARK_II_tools/src/assembler/tokenizer.py", line 66, in find_arg_value
    return found, given_args[counter]
IndexError: list index out of range

There is something terribly wrong in tokenizer.py.

Please check how assembler deal with negative values as arguments. Negative values should be probably emited by compiler directly in hexadecimal form.

Implement in-line assembler into mark vbcc backend. This should be done easy. Also implement interrupt handlers.

It is necessary to add support for transmitting using PS/2. CPU need to talk with keyboard bidirectionally.

Probably bug in vbcc backend:

char strA[2][32] = {{"Hello world A!"}, {"Hello World B!"}};
char *strB[2];

int main(){	
	strB[0] = &strA[0][0];
	strB[1] = &strA[1][0];		
	return 0;
}

strB is array of pointers to char, vbcc compile it into:

.EXPORT 	 main 
main: 
	PUSH 	 R13
	OR   	 R0 SP R13
	PUSH 	 R1
	PUSH 	 R2
	PUSH 	 R3
	PUSH 	 R4
	MVIA 	 R1 strA
	ST  	 R1 strB
	MVIA 	 R1 strA
	MVIA 	 R2 strB
	MVIA 	 R3 1
	ADD 	 R2 R3 R2
	STI 	 R1 R2
	OR  	 R0 R0 R5
L_1:
	POP 	 R4
	POP 	 R3
	POP 	 R2
	POP 	 R1
	OR  	 R0 R13 SP
	POP 	 R13
	RET
.EXPORT 	 strA
strA:
	.DAT 	 72
	.DAT 	 101
	.DAT 	 108
	.DAT 	 108
	.DAT 	 111
	.DAT 	 32
	.DAT 	 119
	.DAT 	 111
	.DAT 	 114
	.DAT 	 108
	.DAT 	 100
	.DAT 	 32
	.DAT 	 65
	.DAT 	 33
	.DAT 	 0
	.DS 	17
	.DAT 	 72
	.DAT 	 101
	.DAT 	 108
	.DAT 	 108
	.DAT 	 111
	.DAT 	 32
	.DAT 	 87
	.DAT 	 111
	.DAT 	 114
	.DAT 	 108
	.DAT 	 100
	.DAT 	 32
	.DAT 	 66
	.DAT 	 33
	.DAT 	 0
	.DS 	17
.EXPORT 	 strB
strB:
	.DS 	2

in main function, after function head, vbcc try to load address of strA[0][0] into strB[0], and that is correct, but it will load same strA[0][0] address into strB[1], in strB[1] should be strA[1][0] instead

I tested it with gcc and code worked as expected.

Signal int_accept from instruction decoder is asynchronous, but should be synchronous as others signals from ID.

It would be nice to have simple random number generator, for example something like free running linear feedback shift register. It may be necessary to have one for generating random seeds for pseudo random numbers by c library.

Simple overview of Standard peripheral library should be great. Also add some examples of using it.

There is a lot of I2C devices on the new board. MARK-II need driver for this bus.

According discussion at reddit there is priority encoder in comparison unit. But this priority encoder is unwanted. Suggested solution is use of case statement.

It should be possible read from VRAM too. Implement it!

CPU core should be improved, but should be compatible with old one.

What new core should be able:

• FPU unit (single precision)
• Improved barrel shifter
• Improved comparator
• Faster execution generally

There is a lot of place to speed up CPU, ideal should be 50MHz or 100MHz. Just now CPU is running at 50MHz on Cyclone IV, and at 25MHz on low cost MAX10. Things that can be done to speed up:

• Pipeline CPU
• Add cache memory
• Replace FIFO in SDRAM driver with better domain crossing
• Don't close row in SDRAM
• Load next word from SDRAM into temp register when reading instruction

MARK-II need some SPI driver that will allow communication with NOR Flash IC on the board. If it can work with SD card too, it would be nice.