nasmjf

1 Now we can start fleshing out this Forth implementation 2 with some basic words. I'll start with the stack manipulation 3 words as defined in assembly (for speed) in JonesFORTH: 4 5 ; drop top of stack 6 DEFCODE "DROP",4,,DROP 7 pop eax 8 NEXT 9 10 I'll test each one out as I go. Here's DROP: 11 12 65 EMIT 13 A 14 65 66 67 15 DROP EMIT 16 B 17 EMIT 18 A 19 20 Good, it dropped 67 ('c') from the top of the stack. 21 22 And swap: 23 24 ; swap top two elements 25 DEFCODE "SWAP",4,,SWAP 26 pop eax 27 pop ebx 28 push eax 29 push ebx 30 NEXT 31 32 Which is easy to test: 33 34 65 66 EMIT EMIT 35 BA 36 65 66 SWAP EMIT EMIT 37 AB 38 39 Finally, 40 41 ; duplicate element on top of stack 42 DEFCODE "DUP",3,,DUP 43 mov eax, [esp] 44 push eax 45 NEXT 46 47 65 DUP 66 DUP EMIT EMIT EMIT EMIT 48 BBAA 49 50 Awesome. 51 52 Next night: okay I've ported over OVER, ROT, -ROT, 53 2DROP, 2DUP, 2SWAP, and ?DUP. I'll just test them all 54 at once and won't bother with the source - they're 55 as simple as the above - so it's really easy to read 56 their assembly definitions in the source file. 57 58 : A 65 ; 59 A EMIT 60 A 61 : B 66 ; 62 : C 67 ; 63 : D 68 ; 64 : e EMIT ; 65 A B C D e e e e 66 DCBA 67 A B C OVER e e e e 68 BCBA 69 A B C ROT e e e 70 ACB 71 A B C -ROT e e e 72 BAC 73 A B C D DROP e e e 74 CBA 75 A B C D 2DROP e e 76 BA 77 A B C D 2SWAP e e e e 78 BADC 79 A B ?DUP e e e 80 BBA 81 A B 0 ?DUP e e e 82 83 Program received signal SIGSEGV, Segmentation fault. 84 _FIND.test_word () at nasmjf.asm:532 85 532 mov al, [edx+4] ; al = flags+length field 86 (gdb) 87 88 LOL, well, ?DUP is a little hard to test compared 89 to the others. Though a segfault for trying to print 90 an ASCII NUL seems weird, and is GDB telling me it 91 crashed in FIND? Hmm. I'll have to keep an eye on that. 92 93 The next series of words to define in assembly are 94 math operators, which is great. But I think I'm going 95 to need to be able to display numbers before that'll 96 be any fun to test. 97 98 ... 99 100 Several nights later: Done! Thanks to one of the utility 101 routines I wrote as part of the excellent asmtutor.com, 102 I was able to slap in a temporary stand-in for the DOT 103 word which Jones defines in pure Forth. 104 105 DOT (.) pops and prints the numeric value at the top of 106 the stack. At least, I think it's supposed to pop the 107 value...well, mine does, for now. 108 109 Starting program: /home/dave/nasmjf/nasmjf 110 Continuing. 111 1 . 112 1 113 11 . 114 11 115 111 . 116 111 117 1111 . 118 1111 119 11111 . 120 121 Program received signal SIGSEGV, Segmentation fault. 122 _FIND.test_word () at nasmjf.asm:532 123 532 mov al, [edx+4] ; al = flags+length field 124 125 Works great, but for some reason, accessing edx+4 causes a 126 segfault when we try to match a word with a 5-digit number? 127 I started stepping throuh it, but I wasn't feeling like 128 chasing the rabbit into that particular hole tonight. We've 129 got some new math-related words to test! 130 131 1+ and 1- increment and decrement. 132 4+ and 4- do the same, but by 4. 133 134 5 1+ . 135 6 136 5 1- . 137 4 138 5 4+ . 139 9 140 5 4- . 141 1 142 143 Then we have + and -. 144 145 5 4 + . 146 134521461 147 148 LOL, what? 149 150 5 4 + . 151 Program received signal SIGSEGV, Segmentation fault. 152 0x00000004 in ?? () 153 154 Well, clearly "+" has a bug. 155 156 5 4 - . 157 1 158 159 But "-" looks good. I'm curious about a negative 160 value: 161 162 4 5 - . 163 4294967295 164 165 I'll have to check thia answer. Is this -1? 166 167 Moving on, we have multiplication and division: 168 169 2 3 * . 170 6 171 10 5 /MOD . . 172 2 0 173 100 3 /MOD . . 174 33 1 175 176 Great! Note: I added space beween the quotient and remainder 177 for readability. The 'real' DOT in forth prints a space after 178 each digit. 179 180 Cool! So we've got math! 181 182 Oh, and we need a satisfying conclusion on the ?DUP operator 183 from our stack manipulation words from the beginning of this 184 particular log. 185 186 ?DUP duplicates the value on the top of the stack IF it's not 187 zero: 188 189 1 2 3 ?DUP . . . 190 332 191 1 2 0 ?DUP . . . 192 021 193 194 Yup, that works! 195 196 Looks like I'll find the bug in "+" next. :-) 197 198 Yup, fixed that with the next set of words. See next log.