nasmjf

1 Now having seen a correct response when nothing matches 2 idiotic input string "foo", let's take a look at input 3 that does match. I'll start with a defined word. How about 4 "FIND", since we know that exists. 5 6 I'm not feeling particularly clever tonight. So I'm just 7 going to step through everthing and cut down the log to 8 the good stuff. 9 10 The "FIND" below is me typing "FIND" when the interpreter 11 requests input. 12 13 (gdb) run 14 code_INTERPRET () at nasmjf.asm:209 15 _WORD.skip_non_words () at nasmjf.asm:308 16 _KEY () at nasmjf.asm:349 17 _KEY.get_more_input () at nasmjf.asm:359 18 ... 19 364 int 0x80 ; syscall! 20 (gdb) 21 FIND 22 23 Now let's see if FIND can find itself. :-) 24 25 code_INTERPRET () at nasmjf.asm:212 26 _FIND () at nasmjf.asm:485 27 485 push esi ; _FIND! Save esi, we'll use this reg for string comparison 28 488 mov edx,[var_LATEST] ; LATEST points to name header of the latest word in the diction 29 30 As before, we alternate between .test_word and .prev_word 31 for every word in the dictionary, starting with the latest 32 and using the stored pointers as a linked list until we 33 find a match or the beginning of the list. 34 35 _FIND.test_word () at nasmjf.asm:490 36 _FIND.prev_word () at nasmjf.asm:517 37 ... 38 490 test edx,edx ; NULL pointer? (end of the linked list) 39 491 je .not_found 40 496 xor eax,eax 41 497 mov al, [edx+4] ; al = flags+length field 42 498 and al,(F_HIDDEN|F_LENMASK) ; al = name length 43 499 cmp cl,al ; Length is the same? 44 500 jne .prev_word ; nope, try prev 45 46 Length of the word matched, now we'll check the actual 47 name string. 48 49 The key to understanding the comparison is knowing that 50 cmpsb implicitly uses the eci and edi registers as pointers 51 to the data to compare. The repe mnemonic stands for 52 "repeat while equal" and is a modifier for the cmpsb 53 instruction. 54 55 I've always been in the RISC camp (versus CISC), because 56 I love systems that compose from "simple" pieces. But I 57 have to admit that these "string" operations in x86 58 do make a lot of sense. After dwelling on this a bit 59 last night, I think I had a dream where I desiged some 60 new hardware (like Ben Eater's 8-bit breadboard CPU or 61 was it an FPGA?) and I was coming up with a new 62 instruction set architecture (ISA) that was strictly 63 "complex" instructions like these string operations, 64 but I think it was also inspired by the array languages 65 like APL and J. Anyway, I've abandoned my simplistic 66 noob stance on RISC: there are lots of different kinds 67 of simple and RISC trades one kind for another. 68 69 Let's see a repe cmpsb in action... 70 71 503 push ecx ; Save the length 72 504 push edi ; Save the address (repe cmpsb will move this pointer) 73 _FIND.test_word () at nasmjf.asm:505 74 505 lea esi,[edx+5] ; Dictionary string we are checking against. 75 506 repe cmpsb ; Compare the strings. 76 507 pop edi 77 _FIND.test_word () at nasmjf.asm:508 78 508 pop ecx 79 _FIND.test_word () at nasmjf.asm:509 80 509 jne .prev_word ; nope, try prev 81 512 pop esi 82 _FIND.test_word () at nasmjf.asm:513 83 513 mov eax,edx 84 514 ret ; FOUND! 85 86 Yay! We've got a match on the word. 87 88 Now back in INTERPRET, we have to do another comparison to 89 check the return value. Since we did, we can now act upon 90 the matched word. 91 92 code_INTERPRET () at nasmjf.asm:215 93 215 test eax,eax ; Found? 94 216 jz .try_literal 95 219 mov edi,eax ; edi = dictionary entry YES WE HAVE MATCHED A WORD!!! 96 220 mov al,[edi+4] ; Get name+flags. 97 221 push ax ; Just save it for now. 98 code_INTERPRET () at nasmjf.asm:222 99 222 call _TCFA ; Convert dictionary entry (in %edi) to codeword pointer. 100 101 So TCFA is the internal label for the Forth word 102 ">CFA" which I read as "To CFA" and Jones guesses 103 probably means "Code Field Address". It's job is 104 to take the given pointer to a word and return a 105 pointer to the word's code. Neat. 106 107 _TCFA () at nasmjf.asm:386 108 386 xor eax,eax 109 387 add edi,4 ; Skip link pointer. 110 388 mov al,[edi] ; Load flags+len into %al. 111 389 inc edi ; Skip flags+len byte. 112 390 and al,F_LENMASK ; Just the length, not the flags. 113 391 add edi,eax ; Skip the name. 114 392 add edi,3 ; The codeword is 4-byte aligned. 115 393 and edi,-3 116 394 ret 117 118 Then we return to INTERPRET again now that the edi 119 register contains the address of the matched word's 120 code (in this case, the code for FIND). 121 122 We can be in immediate mode and/or executing state. 123 124 In this case, we are NOT in immediate mode. 125 126 code_INTERPRET () at nasmjf.asm:223 127 223 pop ax 128 code_INTERPRET () at nasmjf.asm:224 129 224 and al,F_IMMED ; is IMMED flag set? 130 225 mov eax,edi 131 226 jnz .execute ; If IMMED, jump straight to executing. 132 133 We ARE in executing state. 134 135 227 jmp .check_state 136 code_INTERPRET.check_state () at nasmjf.asm:238 137 238 mov edx,[var_STATE] 138 239 test edx,edx 139 240 jz .execute ; Jump if executing. 140 code_INTERPRET.execute () at nasmjf.asm:253 141 253 mov ecx,[interpret_is_lit] ; Literal? 142 254 test ecx,ecx ; Literal? 143 255 jnz .do_literal 144 145 To execute the matched word, we simply jump 146 to the code address... 147 148 259 jmp [eax] 149 150 ...and now we're executing FIND, just as expected. 151 152 code_FIND () at nasmjf.asm:478 153 478 pop ecx ; length of word 154 479 pop edi ; buffer with word 155 480 call _FIND 156 _FIND () at nasmjf.asm:485 157 485 push esi ; _FIND! Save esi, we'll use this reg for string comparison 158 ... 159 160 I'm not even sure what FIND is looking for now 161 since I didn't bother examining any memory during 162 the rest of the run. I was just happy to see the 163 interpreter finding and executing the requested 164 word! 165 166 Eventually it got through the linked list and didn't 167 match anything. 168 169 _FIND.test_word () at nasmjf.asm:490 170 490 test edx,edx ; NULL pointer? (end of the linked list) 171 491 je .not_found 172 _FIND.not_found () at nasmjf.asm:521 173 521 pop esi 174 _FIND.not_found () at nasmjf.asm:522 175 176 And then Forth exited normally having run out 177 of code (the interpreter does not yet loop, so 178 it always exits after the first bit of input. 179 180 190 mov eax, 1 ; exit syscall 181 191 int 80h ; call kernel 182 [Inferior 1 (process 2531) exited normally] 183 184 Next will be handling numeric literals.