A More Efficient If (VM Code) Than Shimon's

Shimon (or one of his underlings - maybe Yaron)

expression
if-goto IF_TRUE0
goto IF_FALSE0
label IF_TRUE0
if-statements
goto IF_END0
label IF_FALSE0
else-statements
label IF_END0

when all you really need is :

expression
if-goto IF_TRUE0
else-statements
goto IF_END0
label IF_TRUE0
if-statements
label IF_END0

So, why would they choose to implement the less optimal thing?

Nand2Tetris : if-goto (influences stack?)

Yes!!

You want to push something onto the stack to test it.

So, if you don't restore the stack-pointer with the if-goto, then you're leaky..

Nand2Tetris 2 : Project 07 Cheat Sheet

If you're like me wondering if documentation for project 7 could be improved - the answer is a resounding yes. Do C_ARITHMETIC commands include the logical ones? YES!! Book says so clearly on Pg 127. But, you can watch the videos twice and still be left with this q.

How about implementing the gt, lt, eq? You have to use jumps, but how do you do that - you're putting out code with a tonne of @DONE's.

And remember (?), the VM represents true as -1 or 0xFFFF and not 1 :) How should you know and what significance does this have? I don't know..

Tip : if using python, you should use raw strings (r prefix) for the match pattern so you can pass special tokens to the regex engine using backslashes that don't get interpreted by the parser itself. (Is it poss for me to confuse you more:) ?)

Shimon's More Efficient Push - Nice

In the video, where he does push constant 17, he shows

@17
D=A
@SP
A=M
M=D
@SP
M=M+1

Notice that needs 3 A instructions. But, if you pay attention, during the demo, with BasicTest.vm, for  push constant 10, his implementation is :

@10
D=A
@SP
AM=M+1
A=A-1
M=D

Saved an instruction! Nice. Ha - did you think Stuxnet was easy to build my man? Truly, since he scrolls down in his ASM file, he's given you the implementation of "pop argument 2" and so it should be okay to put that down here - but that would be mean :)

Not that they're hard to implement, but he's given you add and sub as well :) Both cute, and using 5 instructions each. Booleans are super easy. For the conditionals, you'll have to use the jumps to conditionally write the appropriate value to the stack.

Nand2Tetris Hack Divide By 2 (Right Shift)

// Divides R0 by 2 and stores the dividend in R1 and remainder in R2
// (R0, R1, R2 refer to RAM[0], RAM[1], and RAM[2], respectively.)

// how : counter goes from 2 to 32768 ( doubling each time )
// start with 1 and dump result of the AND in R2
// start off setting R1 to 0 and after that, based on result of the AND of
// counter with R0, you either OR R1 with counter or do nothing

@R1
M = 0
D = 1
@R0
D = M & D
@R2
M = D // remainder captured
@lagcount
M = 1
@2
D = A
@counter
M = D

(LOOP)
@R0
D = M & D // we're expecting D (pre) to already have the counter value based on knowing our code
@NO_ACTION
D, JEQ
@lagcount // else, we need to OR R1 with the lagging counter
D = M
@R1
M = M | D
(NO_ACTION)
@16384 // can only load a 15 bit number :)
D = A
D = D + A
@counter
D = M - D
@END
D, JEQ
// else, we need to double counter and double lagcount
@lagcount
D = M
M = M + D
@counter
D = M
MD = M + D

@LOOP
0, JMP

(END)
@END
0, JMP

Nand2Tetris Hack Language Min Function

Make a more efficient mult:)

// finds min of R0 and R1 and stores the result in R2.
// (R0, R1, R2 refer to RAM[0], RAM[1], and RAM[2], respectively.)

// how : start off assuming R0 is smaller. Then, subtract R1 from R0 and, if 
// result is > 0 => R1 is smaller, put R1 in R2.

@R0
D=M
@R2
M=D // until we find that R1 is smaller
@R1
D=M
@R0
M=M-D // we have R0 - R1
@R0 // if > 0, then we need to put R1 in R2
D=M
@R1SMALLER
D, JGT // if R1 is smaller than R0, this will be > 0 ..
@END
0, JMP // else, we're done

(R1SMALLER)
@R1
D=M
@R2
M=D

(END)
@END
0, JMP

nand2tetris Hilf Help Shimon!

I load a program into the CPU emulator and nothing happens.

Might need to drag the lower right corners south-east a little to see the status line - you'll probably see a message there. Why the hell they don't have a View > Parser Log through the Menu just beats me. But hey, it's free :)

I do

@LOOP
JMP

and it doesn't like it. Ya, coz you HAVE to supply the core part of a C (compute) instruction always - only the jump and destination are optional. So, in this case, just do

0, JMP