1 /*****************************************************************************/
5 /* Code segment structure */
9 /* (C) 2001 Ullrich von Bassewitz */
11 /* D-70597 Stuttgart */
12 /* EMail: uz@musoftware.de */
15 /* This software is provided 'as-is', without any expressed or implied */
16 /* warranty. In no event will the authors be held liable for any damages */
17 /* arising from the use of this software. */
19 /* Permission is granted to anyone to use this software for any purpose, */
20 /* including commercial applications, and to alter it and redistribute it */
21 /* freely, subject to the following restrictions: */
23 /* 1. The origin of this software must not be misrepresented; you must not */
24 /* claim that you wrote the original software. If you use this software */
25 /* in a product, an acknowledgment in the product documentation would be */
26 /* appreciated but is not required. */
27 /* 2. Altered source versions must be plainly marked as such, and must not */
28 /* be misrepresented as being the original software. */
29 /* 3. This notice may not be removed or altered from any source */
32 /*****************************************************************************/
57 /*****************************************************************************/
59 /*****************************************************************************/
63 /* Pointer to current code segment */
68 /*****************************************************************************/
69 /* Functions for parsing instructions */
70 /*****************************************************************************/
74 static CodeLabel* NewCodeSegLabel (CodeSeg* S, const char* Name, unsigned Hash)
75 /* Create a new label and insert it into the label hash table */
77 /* Not found - create a new one */
78 CodeLabel* L = NewCodeLabel (Name, Hash);
80 /* Enter the label into the hash table */
81 L->Next = S->LabelHash[L->Hash];
82 S->LabelHash[L->Hash] = L;
84 /* Return the new label */
90 static const char* SkipSpace (const char* S)
91 /* Skip white space and return an updated pointer */
93 while (IsSpace (*S)) {
101 static const char* ReadToken (const char* L, const char* Term,
102 char* Buf, unsigned BufSize)
103 /* Read the next token into Buf, return the updated line pointer. The
104 * token is terminated by one of the characters given in term.
107 /* Read/copy the token */
109 unsigned ParenCount = 0;
110 while (*L && (ParenCount > 0 || strchr (Term, *L) == 0)) {
116 } else if (*L == '(') {
122 /* Terminate the buffer contents */
125 /* Return the updated line pointer */
131 static CodeEntry* ParseInsn (CodeSeg* S, const char* L)
132 /* Parse an instruction nnd generate a code entry from it. If the line contains
133 * errors, output an error message and return NULL.
134 * For simplicity, we don't accept the broad range of input a "real" assembler
135 * does. The instruction and the argument are expected to be separated by
136 * white space, for example.
141 am_t AM = 0; /* Initialize to keep gcc silent */
148 L = ReadToken (L, " \t", Mnemo, sizeof (Mnemo));
150 /* Try to find the opcode description for the mnemonic */
151 OPC = FindOpcode (Mnemo);
153 /* If we didn't find the opcode, print an error and bail out */
155 Error ("ASM code error: %s is not a valid mnemonic", Mnemo);
159 /* Skip separator white space */
162 /* Get the addressing mode */
173 StrCopy (Expr, sizeof (Expr), L+1);
179 L = ReadToken (L+1, ",)", Expr, sizeof (Expr));
181 /* Check for errors */
183 Error ("ASM code error: syntax error");
187 /* Check the different indirect modes */
189 /* Expect zp x indirect */
191 if (toupper (*L) != 'X') {
192 Error ("ASM code error: `X' expected");
197 Error ("ASM code error: `)' expected");
202 Error ("ASM code error: syntax error");
206 } else if (*L == ')') {
207 /* zp indirect or zp indirect, y */
211 if (toupper (*L) != 'Y') {
212 Error ("ASM code error: `Y' expected");
217 Error ("ASM code error: syntax error");
221 } else if (*L == '\0') {
224 Error ("ASM code error: syntax error");
240 /* Absolute, maybe indexed */
241 L = ReadToken (L, ",", Expr, sizeof (Expr));
243 /* Assume absolute */
245 } else if (*L == ',') {
249 Error ("ASM code error: syntax error");
256 } else if (Reg == 'Y') {
259 Error ("ASM code error: syntax error");
263 Error ("ASM code error: syntax error");
272 /* If the instruction is a branch, check for the label and generate it
273 * if it does not exist. In case of a PC relative branch (*+x) we will
274 * not generate a label, because the target label will not be defined.
277 if ((OPC->Info & CI_MASK_BRA) == CI_BRA && Expr[0] != '*') {
281 /* Addressing mode must be alsobute or something is really wrong */
282 CHECK (AM == AM_ABS);
284 /* Addressing mode is a branch/jump */
287 /* Generate the hash over the label, then search for the label */
288 Hash = HashStr (Expr) % CS_LABEL_HASH_SIZE;
289 Label = FindCodeLabel (S, Expr, Hash);
291 /* If we don't have the label, it's a forward ref - create it */
293 /* Generate a new label */
294 Label = NewCodeSegLabel (S, Expr, Hash);
298 /* We do now have the addressing mode in AM. Allocate a new CodeEntry
299 * structure and initialize it.
301 E = NewCodeEntry (OPC, AM, Label);
302 if (Expr[0] != '\0') {
303 /* We have an additional expression */
304 E->Arg.Expr = xstrdup (Expr);
307 /* Return the new code entry */
313 /*****************************************************************************/
315 /*****************************************************************************/
319 CodeSeg* NewCodeSeg (const char* SegName, const char* FuncName)
320 /* Create a new code segment, initialize and return it */
324 /* Allocate memory */
325 CodeSeg* S = xmalloc (sizeof (CodeSeg));
327 /* Initialize the fields */
329 S->SegName = xstrdup (SegName);
330 S->FuncName = xstrdup (FuncName);
331 InitCollection (&S->Entries);
332 InitCollection (&S->Labels);
333 for (I = 0; I < sizeof(S->LabelHash) / sizeof(S->LabelHash[0]); ++I) {
337 /* Return the new struct */
343 void FreeCodeSeg (CodeSeg* S)
344 /* Free a code segment including all code entries */
352 /* Free the entries */
353 Count = CollCount (&S->Entries);
354 for (I = 0; I < Count; ++I) {
355 FreeCodeEntry (CollAt (&S->Entries, I));
358 /* Free the collections */
359 DoneCollection (&S->Entries);
360 DoneCollection (&S->Labels);
362 /* Free all labels */
363 for (I = 0; I < sizeof(S->LabelHash) / sizeof(S->LabelHash[0]); ++I) {
364 CodeLabel* L = S->LabelHash[I];
372 /* Free the struct */
378 void PushCodeSeg (CodeSeg* S)
379 /* Push the given code segment onto the stack */
387 CodeSeg* PopCodeSeg (void)
388 /* Remove the current code segment from the stack and return it */
390 /* Remember the current code segment */
393 /* Cannot pop on empty stack */
394 PRECONDITION (S != 0);
399 /* Return the popped code segment */
405 void AddCodeSegLine (CodeSeg* S, const char* Format, ...)
406 /* Add a line to the given code segment */
412 /* Format the line */
415 va_start (ap, Format);
416 xvsprintf (Buf, sizeof (Buf), Format, ap);
419 /* Skip whitespace */
422 /* Check which type of instruction we have */
423 E = 0; /* Assume no insn created */
427 /* Empty line, just ignore it */
431 /* Comment or hint, ignore it for now */
435 /* Control instruction */
436 ReadToken (L, " \t", Token, sizeof (Token));
437 Error ("ASM code error: Pseudo instruction `%s' not supported", Token);
441 E = ParseInsn (S, L);
445 /* If we have a code entry, transfer the labels and insert it */
448 /* Transfer the labels if we have any */
450 unsigned LabelCount = CollCount (&S->Labels);
451 for (I = 0; I < LabelCount; ++I) {
453 CodeLabel* L = CollAt (&S->Labels, I);
454 /* Mark it as defined */
456 /* Move it to the code entry */
457 CollAppend (&E->Labels, L);
460 /* Delete the transfered labels */
461 CollDeleteAll (&S->Labels);
463 /* Add the entry to the list of code entries in this segment */
464 CollAppend (&S->Entries, E);
471 void AddCodeLabel (CodeSeg* S, const char* Name)
472 /* Add a code label for the next instruction to follow */
474 /* Calculate the hash from the name */
475 unsigned Hash = HashStr (Name) % CS_LABEL_HASH_SIZE;
477 /* Try to find the code label if it does already exist */
478 CodeLabel* L = FindCodeLabel (S, Name, Hash);
480 /* Did we find it? */
482 /* We found it - be sure it does not already have an owner */
483 CHECK (L->Owner == 0);
485 /* Not found - create a new one */
486 L = NewCodeSegLabel (S, Name, Hash);
489 /* We do now have a valid label. Remember it for later */
490 CollAppend (&S->Labels, L);
495 void AddCodeSegHint (CodeSeg* S, unsigned Hint)
496 /* Add a hint for the preceeding instruction */
500 /* Get the number of entries in this segment */
501 unsigned EntryCount = CollCount (&S->Entries);
503 /* Must have at least one entry */
504 CHECK (EntryCount > 0);
506 /* Get the last entry */
507 E = CollAt (&S->Entries, EntryCount-1);
515 void DelCodeSegAfter (CodeSeg* S, unsigned Last)
516 /* Delete all entries including the given one */
518 /* Get the number of entries in this segment */
519 unsigned Count = CollCount (&S->Entries);
521 /* Remove all entries after the given one */
522 while (Last < Count) {
524 /* Get the next entry */
525 CodeEntry* E = CollAt (&S->Entries, Count-1);
527 /* We have to transfer all labels to the code segment label pool */
528 unsigned LabelCount = CollCount (&E->Labels);
529 while (LabelCount--) {
530 CodeLabel* L = CollAt (&E->Labels, LabelCount);
532 CollAppend (&S->Labels, L);
534 CollDeleteAll (&E->Labels);
536 /* Remove the code entry */
537 FreeCodeEntry (CollAt (&S->Entries, Count-1));
538 CollDelete (&S->Entries, Count-1);
545 void OutputCodeSeg (FILE* F, const CodeSeg* S)
546 /* Output the code segment data to a file */
550 /* Get the number of entries in this segment */
551 unsigned Count = CollCount (&S->Entries);
553 /* Output the segment directive */
554 fprintf (F, ".segment\t\"%s\"\n\n", S->SegName);
556 /* If this is a segment for a function, enter a function */
557 if (S->FuncName[0] != '\0') {
558 fprintf (F, ".proc\t_%s\n\n", S->FuncName);
561 /* Output all entries */
562 for (I = 0; I < Count; ++I) {
563 OutputCodeEntry (F, CollConstAt (&S->Entries, I));
566 /* If this is a segment for a function, leave the function */
567 if (S->FuncName[0] != '\0') {
568 fprintf (F, "\n.endproc\n\n");
574 CodeLabel* FindCodeLabel (CodeSeg* S, const char* Name, unsigned Hash)
575 /* Find the label with the given name. Return the label or NULL if not found */
577 /* Get the first hash chain entry */
578 CodeLabel* L = S->LabelHash[Hash];
580 /* Search the list */
582 if (strcmp (Name, L->Name) == 0) {
593 void MergeCodeLabels (CodeSeg* S)
594 /* Merge code labels. That means: For each instruction, remove all labels but
595 * one and adjust the code entries accordingly.
600 /* Walk over all code entries */
601 unsigned EntryCount = CollCount (&S->Entries);
602 for (I = 0; I < EntryCount; ++I) {
607 /* Get a pointer to the next entry */
608 CodeEntry* E = CollAt (&S->Entries, I);
610 /* If this entry has zero labels, continue with the next one */
611 unsigned LabelCount = CollCount (&E->Labels);
612 if (LabelCount == 0) {
616 /* We have at least one label. Use the first one as reference label. */
617 RefLab = CollAt (&E->Labels, 0);
619 /* Walk through the remaining labels and change references to these
620 * labels to a reference to the one and only label. Delete the labels
621 * that are no longer used. To increase performance, walk backwards
624 for (J = LabelCount-1; J >= 1; --J) {
628 /* Get the next label */
629 CodeLabel* L = CollAt (&E->Labels, J);
631 /* Walk through all instructions referencing this label */
632 unsigned RefCount = CollCount (&L->JumpFrom);
633 for (K = 0; K < RefCount; ++K) {
635 /* Get the next instrcuction that references this label */
636 CodeEntry* E = CollAt (&L->JumpFrom, K);
638 /* Change the reference */
639 CHECK (E->JumpTo == L);
641 CollAppend (&RefLab->JumpFrom, E);
645 /* Remove the label completely. */
647 CollDelete (&E->Labels, J);
650 /* The reference label is the only remaining label. Check if there
651 * are any references to this label, and delete it if this is not
654 if (CollCount (&RefLab->JumpFrom) == 0) {
655 /* Delete the label */
656 FreeCodeLabel (RefLab);
657 /* Remove it from the list */
658 CollDelete (&E->Labels, 0);
665 unsigned GetCodeSegEntries (const CodeSeg* S)
666 /* Return the number of entries for the given code segment */
668 return CollCount (&S->Entries);