ghc/includes/StgMacros.lh

   1 %
   2 % (c) The GRASP/AQUA Project, Glasgow University, 1993-1994
   3 %
   4 \section[StgMacros]{C macros used in GHC-generated \tr{.hc} files}
   5
   6 \begin{code}
   7 #ifndef STGMACROS_H
   8 #define STGMACROS_H
   9 \end{code}
  10
  11 %************************************************************************
  12 %*                                                                      *
  13 \subsection[StgMacros-abbrev]{Abbreviatory(?) and general macros}
  14 %*                                                                      *
  15 %************************************************************************
  16
  17 Mere abbreviations:
  18 \begin{code}
  19 /* for function declarations */
  20 #define STGFUN(f)  F_ f(STG_NO_ARGS)
  21 #define STATICFUN(f) static F_ f(STG_NO_ARGS)
  22
  23 /* for functions/data that are really external to this module */
  24 #define EXTFUN(f)       extern F_ f(STG_NO_ARGS)
  25 #define EXTDATA(d)      extern W_ d[]
  26 #define EXTDATA_RO(d)   extern const W_ d[] /* read-only */
  27
  28 /* for fwd decls to functions/data somewhere else in this module */
  29 /* (identical for the mo') */
  30 #define INTFUN(f)       static F_ f(STG_NO_ARGS)
  31 #define INTDATA(d)      extern W_ d[]
  32 #define INTDATA_RO(d)   extern const W_ d[] /* read-only */
  33
  34 /* short forms of most of the above */
  35
  36 #define FN_(f)          F_ f(STG_NO_ARGS)
  37 #define IFN_(f)         static F_ f(STG_NO_ARGS)
  38 #define EF_(f)          extern F_ f(STG_NO_ARGS)
  39 #define ED_(d)          extern W_ d[]
  40 #define ED_RO_(d)       extern const W_ d[] /* read-only */
  41 #define IF_(f)          static F_ f(STG_NO_ARGS)
  42
  43 /* GCC is uncooperative about the next one: */
  44 /* But, the "extern" prevents initialisation... ADR */
  45 #if defined(__GNUC__)
  46 #define ID_(d)          extern W_ d[]
  47 #define ID_RO_(d)       extern const W_ d[] /* read-only */
  48 #else
  49 #define ID_(d)          static W_ d[]
  50 #define ID_RO_(d)       static const W_ d[] /* read-only */
  51 #endif /* not GCC */
  52 \end{code}
  53
  54 General things; note: general-but-``machine-dependent'' macros are
  55 given in \tr{StgMachDeps.lh}.
  56 \begin{code}
  57 I_ STG_MAX PROTO((I_, I_)); /* GCC -Wall loves prototypes */
  58
  59 extern STG_INLINE
  60 I_
  61 STG_MAX(I_ a, I_ b) { return((a >= b) ? a : b); }
  62 /* NB: the naive #define macro version of STG_MAX
  63    can lead to exponential CPP explosion, if you
  64    have very-nested STG_MAXes.
  65 */
  66
  67 /*
  68 Macros to combine two short words into a single
  69 word and split such a word back into two.
  70
  71 Dependent on machine word size :-)
  72 */
  73
  74 #define COMBINE_WORDS(word,short1,short2)               \
  75         do {                                            \
  76             ((packed_shorts *)&(word))->wu.s1 = short1; \
  77             ((packed_shorts *)&(word))->wu.s2 = short2; \
  78         } while(0)
  79
  80 #define SPLIT_WORD(word,short1,short2)                  \
  81         do {                                            \
  82             short1 = ((packed_shorts *)&(word))->wu.s1; \
  83             short2 = ((packed_shorts *)&(word))->wu.s2; \
  84         } while(0)
  85
  86 \end{code}
  87
  88 %************************************************************************
  89 %*                                                                      *
  90 \subsection[StgMacros-gen-stg]{General STGish macros}
  91 %*                                                                      *
  92 %************************************************************************
  93
  94 Common sizes of vector-return tables.
  95
  96 Claim: don't need fwd decls for return pts in \tr{VECTBL*}, because
  97 the AbsC flattener ensures that things come out sufficiently
  98 ``backwards''.
  99
 100 \begin{code}
 101 #ifdef __STG_REV_TBLS__
 102 #define UNVECTBL(staticp,label,a)   /* nothing */
 103 #else
 104 #define UNVECTBL(staticp,label,a) \
 105 EXTFUN(a); \
 106 staticp const W_ label[] = { \
 107   (W_) a \
 108 };
 109 #endif
 110 \end{code}
 111
 112 \begin{code}
 113 #if defined(USE_SPLIT_MARKERS)
 114 #define __STG_SPLIT_MARKER(n) FN_(CAT2(__stg_split_marker,n)){ }
 115 #else
 116 #define __STG_SPLIT_MARKER(n) /* nothing */
 117 #endif
 118 \end{code}
 119
 120 %************************************************************************
 121 %*                                                                      *
 122 \subsection[StgMacros-exceptions]{Exception-checking macros}
 123 %*                                                                      *
 124 %************************************************************************
 125
 126 Argument-satisfaction check, stack(s) overflow check, heap overflow
 127 check.
 128
 129 The @SUBTRACT(upper, lower)@ macros return a positive result in words
 130 indicating the amount by which upper is above lower on the stack.
 131
 132 \begin{code}
 133 #define SUBTRACT_A_STK( upper, lower ) AREL( (lower) - (upper) )
 134 #define SUBTRACT_B_STK( upper, lower ) BREL( (lower) - (upper) )
 135 \end{code}
 136
 137 %************************************************************************
 138 %*                                                                      *
 139 \subsubsection[StgMacros-arg-satis]{Argument-satisfaction checks}
 140 %*                                                                      *
 141 %************************************************************************
 142
 143 @ARGS_CHK(n)@ sees of there are @n@ words of args on the A/B stack.
 144 If not, it jumps to @UpdatePAP@.
 145
 146 @ARGS_CHK@ args are pre-directionified.
 147 Notice that we do the comparisons in the form (x < a+n), for
 148 some constant n.  This generates more efficient code (with GCC at least)
 149 than (x-a < n).
 150
 151 \begin{code}
 152 #define ARGS_CHK_A(n)                                           \
 153         if (SuA /*SUBTRACT_A_STK( SpA, SuA )*/ < (SpA+(n))) {   \
 154                 JMP_( UpdatePAP );                              \
 155         }
 156
 157 #define ARGS_CHK_A_LOAD_NODE(n, closure_addr)                   \
 158         if (SuA /*SUBTRACT_A_STK( SpA, SuA )*/ < (SpA+(n))) {   \
 159                 Node = (P_) closure_addr;                       \
 160                 JMP_( UpdatePAP );                              \
 161         }
 162
 163 #define ARGS_CHK_B(n)                                           \
 164         if (SpB /*SUBTRACT_B_STK( SpB, SuB )*/ < (SuB-(n))) {   \
 165                 JMP_( UpdatePAP );                              \
 166         }
 167
 168
 169 #define ARGS_CHK_B_LOAD_NODE(n, closure_addr)                   \
 170         if (SpB /*SUBTRACT_B_STK( SpB, SuB )*/ < (SuB-(n))) {   \
 171                 Node = (P_) closure_addr;                       \
 172                 JMP_( UpdatePAP );                              \
 173         }
 174 \end{code}
 175
 176 %************************************************************************
 177 %*                                                                      *
 178 \subsubsection[StgMacros-stk-chks]{Stack-overflow check}
 179 %*                                                                      *
 180 %************************************************************************
 181
 182 @STK_CHK(a,b)@ [misc args omitted...] checks that we can allocate @a@
 183 words of A stack and @b@ words of B stack.  If not, it calls
 184 @StackOverflow@ (which dies).
 185
 186 (It will be different in the parallel case.)
 187
 188 NB: args @a@ and @b@ are pre-direction-ified!
 189 \begin{code}
 190 I_ SqueezeUpdateFrames PROTO((P_, P_, P_));
 191 int sanityChk_StkO (P_ stko); /* ToDo: move to a sane place */
 192
 193 #if ! defined(CONCURRENT)
 194
 195 extern void StackOverflow(STG_NO_ARGS) STG_NORETURN;
 196
 197 #if STACK_CHECK_BY_PAGE_FAULT
 198
 199 #define STACK_OVERFLOW(liveness,hda,hdb,spa,spb,rtype,reenter)  \
 200     /* use memory protection instead; still need ticky-ness */
 201
 202 #else
 203
 204 #define STACK_OVERFLOW(liveness,hda,hdb,spa,spb,rtype,reenter)  \
 205     ULTRASAFESTGCALL0(void,(void *),StackOverflow)
 206
 207 #endif /* not using page-faulting */
 208
 209 #else /* threaded */
 210
 211 I_ StackOverflow PROTO((W_, W_));
 212
 213 /*
 214  * On a uniprocessor, we do *NOT* context switch on a stack overflow
 215  * (though we may GC).  Therefore, we never have to reenter node.
 216  */
 217
 218 #define STACK_OVERFLOW(liveness,hda,hdb,spa,spb,rtype,reenter) \
 219     DO_STACKOVERFLOW((hda+hdb)<<2|((rtype)<<1)|(reenter),((spa)<<20)|((spb)<<8)|(liveness))
 220
 221 #define STACK_OVERFLOW_HEADROOM(args,y)     ((args) >> 2)
 222 #define STACK_OVERFLOW_PRIM_RETURN(args,y)  ((args) & 2)
 223 #define STACK_OVERFLOW_REENTER(args,y)      ((args) & 1)
 224
 225 #define STACK_OVERFLOW_AWORDS(x,args)       (((args) >> 20) & 0x0fff)
 226 #define STACK_OVERFLOW_BWORDS(x,args)       (((args) >> 8) & 0x0fff)
 227 #define STACK_OVERFLOW_LIVENESS(x,args)     ((args) & 0xff)
 228
 229 #endif  /* CONCURRENT */
 230
 231 #define STK_CHK(liveness_mask,a_headroom,b_headroom,spa,spb,ret_type,reenter)\
 232 do {                                                            \
 233   DO_ASTK_HWM(); /* ticky-ticky profiling */                    \
 234   DO_BSTK_HWM();                                                \
 235   if (STKS_OVERFLOW_OP(((a_headroom) + 1), ((b_headroom) + 1))) {       \
 236     STACK_OVERFLOW(liveness_mask,a_headroom,b_headroom,spa,spb,ret_type,reenter);\
 237   }                                                             \
 238 }while(0)
 239 \end{code}
 240
 241 %************************************************************************
 242 %*                                                                      *
 243 \subsubsection[StgMacros-heap-chks]{Heap-overflow checks}
 244 %*                                                                      *
 245 %************************************************************************
 246
 247 Please see the general discussion/commentary about ``what really
 248 happens in a GC,'' in \tr{SMinterface.lh}.
 249
 250 \begin{code}
 251 void PerformGC PROTO((W_));
 252 void RealPerformGC PROTO((W_ liveness, W_ reqsize, W_  always_reenter_node, rtsBool do_full_collection));
 253 void checkInCCallGC(STG_NO_ARGS);
 254
 255 #ifndef PAR
 256 void StgPerformGarbageCollection(STG_NO_ARGS);
 257 #endif
 258
 259 #ifndef CONCURRENT
 260
 261 #define OR_MSG_PENDING  /* never */
 262
 263 #define HEAP_OVERFLOW(liveness,n,reenter)       \
 264     do {                                        \
 265     DO_GC((((W_)n)<<8)|(liveness));             \
 266     } while (0)
 267
 268 #define REQSIZE_BITMASK ((1L << ((BITS_IN(W_) - 8 + 1))) - 1)
 269 #define HEAP_OVERFLOW_REQSIZE(args)     (((args) >> 8) & REQSIZE_BITMASK)
 270 #define HEAP_OVERFLOW_REENTER(args)     0
 271 #define HEAP_OVERFLOW_LIVENESS(args)    ((args) & 0xff)
 272
 273 #else /* CONCURRENT */
 274
 275 void ReallyPerformThreadGC PROTO((W_, rtsBool));
 276
 277 #define HEAP_OVERFLOW(liveness,n,reenter)       \
 278     do {                                        \
 279     DO_GC((((W_)(n))<<9)|((reenter)<<8)|(liveness)); \
 280     } while (0)
 281
 282 #define REQSIZE_BITMASK ((1L << ((BITS_IN(W_) - 9 + 1))) - 1)
 283 #define HEAP_OVERFLOW_REQSIZE(args)     (((args) >> 9) & REQSIZE_BITMASK)
 284 #define HEAP_OVERFLOW_REENTER(args)     (((args) >> 8) & 0x1)
 285 #define HEAP_OVERFLOW_LIVENESS(args)    ((args) & 0xff)
 286
 287 #ifndef PAR
 288
 289 #define OR_MSG_PENDING  /* never */
 290
 291 #else
 292
 293 extern int PacketsWaiting;              /*Probes for incoming messages*/
 294 extern int heapChkCounter;              /*Not currently used! We check for messages when*/
 295                                         /*a thread is resheduled PWT*/
 296 /* #define OR_MSG_PENDING       || (--heapChkCounter == 0 && PacketsWaiting())*/
 297 #define OR_MSG_PENDING  /* never */
 298
 299 #endif  /* PAR */
 300 #endif  /* CONCURRENT */
 301
 302 #if 0 /* alpha_TARGET_ARCH */
 303 #define CACHE_LINE  4   /* words */
 304 #define LINES_AHEAD 3
 305 #define PRE_FETCH(n)                                    \
 306 do {                                                    \
 307  StgInt j;                                              \
 308  j = ((STG_VOLATILE StgInt *) Hp)[LINES_AHEAD * CACHE_LINE];    \
 309 } while(0);
 310 #define EXTRA_HEAP_WORDS (CACHE_LINE * LINES_AHEAD)
 311 #else
 312 #define PRE_FETCH(reg)
 313 #define EXTRA_HEAP_WORDS 0
 314 #endif
 315
 316 #if defined(GRAN)
 317 #define HEAP_CHK(liveness_mask,n,reenter)                       \
 318         do {                                                    \
 319         /* TICKY_PARANOIA(__FILE__, __LINE__); */               \
 320         /* THREAD_CONTEXT_SWITCH(liveness_mask,reenter); */             \
 321         ALLOC_HEAP(n); /* ticky profiling */                    \
 322         GRAN_ALLOC_HEAP(n,liveness_mask); /* Granularity Simulation */ \
 323         if (((Hp = Hp + (n)) > HpLim)) {                        \
 324             /* Old:  STGCALL3_GC(PerformGC,liveness_mask,n,StgFalse); */\
 325             HEAP_OVERFLOW(liveness_mask,n,StgFalse); \
 326         }}while(0)
 327
 328 #else
 329
 330 #define HEAP_CHK(liveness_mask,n,reenter)               \
 331 do {                                                    \
 332   /* TICKY_PARANOIA(__FILE__, __LINE__); */             \
 333   PRE_FETCH(n);                                         \
 334   ALLOC_HEAP(n); /* ticky profiling */                  \
 335   if (((Hp = Hp + (n)) > HpLim) OR_INTERVAL_EXPIRED OR_CONTEXT_SWITCH OR_MSG_PENDING) { \
 336     HEAP_OVERFLOW(liveness_mask,n,reenter);             \
 337   }                                                     \
 338 } while(0)
 339
 340 #endif  /* GRAN */
 341
 342 #ifdef CONCURRENT
 343
 344 #define HEAP_CHK_AND_RESTORE_N(liveness_mask,n,reenter) \
 345 do {                                                    \
 346   /* TICKY_PARANOIA(__FILE__, __LINE__); */             \
 347   PRE_FETCH(n);                                         \
 348   ALLOC_HEAP(n); /* ticky profiling */                  \
 349   if (((Hp = Hp + (n)) > HpLim) OR_INTERVAL_EXPIRED OR_CONTEXT_SWITCH OR_MSG_PENDING) { \
 350     HEAP_OVERFLOW(liveness_mask,n,reenter);             \
 351     n = TSO_ARG1(CurrentTSO);                           \
 352   }} while(0)
 353
 354 #else
 355
 356 #define HEAP_CHK_AND_RESTORE_N(liveness_mask,n,reenter)     \
 357     HEAP_CHK(liveness_mask,n,reenter)
 358
 359 #endif
 360
 361 \end{code}
 362
 363
 364 %************************************************************************
 365 %*                                                                      *
 366 \subsection[StgMacros-prim-ops]{Primitive operations}
 367 %*                                                                      *
 368 %************************************************************************
 369
 370 One thing to be {\em very careful about} with these macros that assign
 371 to results is that the assignment must come {\em last}.  Some of the
 372 other arguments may be in terms of addressing modes that get clobbered
 373 by the assignment.  (Dirty imperative programming RULES!)
 374
 375 The order here is roughly that in \tr{compiler/prelude/PrimOps.lhs}.
 376
 377 %************************************************************************
 378 %*                                                                      *
 379 \subsubsection[StgMacros-compare-primops]{Primitive comparison ops on basic types}
 380 %*                                                                      *
 381 %************************************************************************
 382
 383 We cast the chars in case one of them is a literal (so C things work right
 384 even for 8-bit chars).
 385 \begin{code}
 386 #define gtCharZh(r,a,b) r=(I_)((a)> (b))
 387 #define geCharZh(r,a,b) r=(I_)((a)>=(b))
 388 #define eqCharZh(r,a,b) r=(I_)((a)==(b))
 389 #define neCharZh(r,a,b) r=(I_)((a)!=(b))
 390 #define ltCharZh(r,a,b) r=(I_)((a)< (b))
 391 #define leCharZh(r,a,b) r=(I_)((a)<=(b))
 392
 393 /* Int comparisons: >#, >=# etc */
 394 #define ZgZh(r,a,b)     r=(I_)((a) >(b))
 395 #define ZgZeZh(r,a,b)   r=(I_)((a)>=(b))
 396 #define ZeZeZh(r,a,b)   r=(I_)((a)==(b))
 397 #define ZdZeZh(r,a,b)   r=(I_)((a)!=(b))
 398 #define ZlZh(r,a,b)     r=(I_)((a) <(b))
 399 #define ZlZeZh(r,a,b)   r=(I_)((a)<=(b))
 400
 401 #define gtWordZh(r,a,b) r=(I_)((a) >(b))
 402 #define geWordZh(r,a,b) r=(I_)((a)>=(b))
 403 #define eqWordZh(r,a,b) r=(I_)((a)==(b))
 404 #define neWordZh(r,a,b) r=(I_)((a)!=(b))
 405 #define ltWordZh(r,a,b) r=(I_)((a) <(b))
 406 #define leWordZh(r,a,b) r=(I_)((a)<=(b))
 407
 408 #define gtAddrZh(r,a,b) r=(I_)((a) >(b))
 409 #define geAddrZh(r,a,b) r=(I_)((a)>=(b))
 410 #define eqAddrZh(r,a,b) r=(I_)((a)==(b))
 411 #define neAddrZh(r,a,b) r=(I_)((a)!=(b))
 412 #define ltAddrZh(r,a,b) r=(I_)((a) <(b))
 413 #define leAddrZh(r,a,b) r=(I_)((a)<=(b))
 414
 415 #define gtFloatZh(r,a,b)  r=(I_)((a)> (b))
 416 #define geFloatZh(r,a,b)  r=(I_)((a)>=(b))
 417 #define eqFloatZh(r,a,b)  r=(I_)((a)==(b))
 418 #define neFloatZh(r,a,b)  r=(I_)((a)!=(b))
 419 #define ltFloatZh(r,a,b)  r=(I_)((a)< (b))
 420 #define leFloatZh(r,a,b)  r=(I_)((a)<=(b))
 421
 422 /* Double comparisons: >##, >=#@ etc */
 423 #define ZgZhZh(r,a,b)   r=(I_)((a) >(b))
 424 #define ZgZeZhZh(r,a,b) r=(I_)((a)>=(b))
 425 #define ZeZeZhZh(r,a,b) r=(I_)((a)==(b))
 426 #define ZdZeZhZh(r,a,b) r=(I_)((a)!=(b))
 427 #define ZlZhZh(r,a,b)   r=(I_)((a) <(b))
 428 #define ZlZeZhZh(r,a,b) r=(I_)((a)<=(b))
 429 \end{code}
 430
 431 %************************************************************************
 432 %*                                                                      *
 433 \subsubsection[StgMacros-char-primops]{Primitive @Char#@ ops (and @LitString#@ish things, too)}
 434 %*                                                                      *
 435 %************************************************************************
 436
 437 We cast the chars in case one of them is a literal (so C things work right
 438 even for 8-bit chars).
 439 \begin{code}
 440 #define ordZh(r,a)      r=(I_)((W_) (a))
 441 #define chrZh(r,a)      r=(StgChar)((W_)(a))
 442 \end{code}
 443
 444 %************************************************************************
 445 %*                                                                      *
 446 \subsubsection[StgMacros-int-primops]{Primitive @Int#@ ops}
 447 %*                                                                      *
 448 %************************************************************************
 449
 450 \begin{code}
 451 I_ stg_div PROTO((I_ a, I_ b));
 452
 453 #define ZpZh(r,a,b)             r=(a)+(b)
 454 #define ZmZh(r,a,b)             r=(a)-(b)
 455 #define ZtZh(r,a,b)             r=(a)*(b)
 456 #define quotIntZh(r,a,b)        r=(a)/(b)
 457 /* ZdZh not used??? --SDM */
 458 #define ZdZh(r,a,b)             r=ULTRASAFESTGCALL2(I_,(void *, I_, I_),stg_div,(a),(b))
 459 #define remIntZh(r,a,b)         r=(a)%(b)
 460 #define negateIntZh(r,a)        r=-(a)
 461 /* Ever used ? -- SOF */
 462 #define absIntZh(a)             r=(( (a) >= 0 ) ? (a) : (-(a)))
 463 \end{code}
 464
 465 %************************************************************************
 466 %*                                                                      *
 467 \subsubsection[StgMacros-word-primops]{Primitive @Word#@ ops}
 468 %*                                                                      *
 469 %************************************************************************
 470
 471 \begin{code}
 472 #define quotWordZh(r,a,b)       r=((W_)a)/((W_)b)
 473 #define remWordZh(r,a,b)        r=((W_)a)%((W_)b)
 474
 475 #define andZh(r,a,b)    r=(a)&(b)
 476 #define orZh(r,a,b)     r=(a)|(b)
 477 #define xorZh(r,a,b)    r=(a)^(b)
 478 #define notZh(r,a)      r=~(a)
 479
 480 #define shiftLZh(r,a,b)   r=(a)<<(b)
 481 #define shiftRAZh(r,a,b)  r=(a)>>(b)
 482 #define shiftRLZh(r,a,b)  r=(a)>>(b)
 483 #define iShiftLZh(r,a,b)  r=(a)<<(b)
 484 #define iShiftRAZh(r,a,b) r=(a)>>(b)
 485 #define iShiftRLZh(r,a,b) r=(a)>>(b)
 486
 487 #define int2WordZh(r,a) r=(W_)(a)
 488 #define word2IntZh(r,a) r=(I_)(a)
 489 \end{code}
 490
 491 %************************************************************************
 492 %*                                                                      *
 493 \subsubsection[StgMacros-addr-primops]{Primitive @Addr#@ ops}
 494 %*                                                                      *
 495 %************************************************************************
 496
 497 \begin{code}
 498 #define int2AddrZh(r,a) r=(A_)(a)
 499 #define addr2IntZh(r,a) r=(I_)(a)
 500 \end{code}
 501
 502 %************************************************************************
 503 %*                                                                      *
 504 \subsubsection[StgMacros-float-primops]{Primitive @Float#@ ops}
 505 %*                                                                      *
 506 %************************************************************************
 507
 508 \begin{code}
 509 #define plusFloatZh(r,a,b)      r=(a)+(b)
 510 #define minusFloatZh(r,a,b)     r=(a)-(b)
 511 #define timesFloatZh(r,a,b)     r=(a)*(b)
 512 #define divideFloatZh(r,a,b)    r=(a)/(b)
 513 #define negateFloatZh(r,a)      r=-(a)
 514
 515 #define int2FloatZh(r,a)        r=(StgFloat)(a)
 516 #define float2IntZh(r,a)        r=(I_)(a)
 517
 518 #define expFloatZh(r,a)         r=(StgFloat) SAFESTGCALL1(StgDouble,(void *, StgDouble),exp,a)
 519 #define logFloatZh(r,a)         r=(StgFloat) SAFESTGCALL1(StgDouble,(void *, StgDouble),log,a)
 520 #define sqrtFloatZh(r,a)        r=(StgFloat) SAFESTGCALL1(StgDouble,(void *, StgDouble),sqrt,a)
 521 #define sinFloatZh(r,a)         r=(StgFloat) SAFESTGCALL1(StgDouble,(void *, StgDouble),sin,a)
 522 #define cosFloatZh(r,a)         r=(StgFloat) SAFESTGCALL1(StgDouble,(void *, StgDouble),cos,a)
 523 #define tanFloatZh(r,a)         r=(StgFloat) SAFESTGCALL1(StgDouble,(void *, StgDouble),tan,a)
 524 #define asinFloatZh(r,a)        r=(StgFloat) SAFESTGCALL1(StgDouble,(void *, StgDouble),asin,a)
 525 #define acosFloatZh(r,a)        r=(StgFloat) SAFESTGCALL1(StgDouble,(void *, StgDouble),acos,a)
 526 #define atanFloatZh(r,a)        r=(StgFloat) SAFESTGCALL1(StgDouble,(void *, StgDouble),atan,a)
 527 #define sinhFloatZh(r,a)        r=(StgFloat) SAFESTGCALL1(StgDouble,(void *, StgDouble),sinh,a)
 528 #define coshFloatZh(r,a)        r=(StgFloat) SAFESTGCALL1(StgDouble,(void *, StgDouble),cosh,a)
 529 #define tanhFloatZh(r,a)        r=(StgFloat) SAFESTGCALL1(StgDouble,(void *, StgDouble),tanh,a)
 530 #define powerFloatZh(r,a,b)     r=(StgFloat) SAFESTGCALL2(StgDouble,(void *, StgDouble,StgDouble),pow,a,b)
 531
 532 /* encoding/decoding given w/ Integer stuff */
 533 \end{code}
 534
 535 %************************************************************************
 536 %*                                                                      *
 537 \subsubsection[StgMacros-double-primops]{Primitive @Double#@ ops}
 538 %*                                                                      *
 539 %************************************************************************
 540
 541 \begin{code}
 542 #define ZpZhZh(r,a,b)           r=(a)+(b)
 543 #define ZmZhZh(r,a,b)           r=(a)-(b)
 544 #define ZtZhZh(r,a,b)           r=(a)*(b)
 545 #define ZdZhZh(r,a,b)           r=(a)/(b)
 546 #define negateDoubleZh(r,a)     r=-(a)
 547
 548 #define int2DoubleZh(r,a)       r=(StgDouble)(a)
 549 #define double2IntZh(r,a)       r=(I_)(a)
 550
 551 #define float2DoubleZh(r,a)     r=(StgDouble)(a)
 552 #define double2FloatZh(r,a)     r=(StgFloat)(a)
 553
 554 #define expDoubleZh(r,a)        r=(StgDouble) SAFESTGCALL1(StgDouble,(void *, StgDouble),exp,a)
 555 #define logDoubleZh(r,a)        r=(StgDouble) SAFESTGCALL1(StgDouble,(void *, StgDouble),log,a)
 556 #define sqrtDoubleZh(r,a)       r=(StgDouble) SAFESTGCALL1(StgDouble,(void *, StgDouble),sqrt,a)
 557 #define sinDoubleZh(r,a)        r=(StgDouble) SAFESTGCALL1(StgDouble,(void *, StgDouble),sin,a)
 558 #define cosDoubleZh(r,a)        r=(StgDouble) SAFESTGCALL1(StgDouble,(void *, StgDouble),cos,a)
 559 #define tanDoubleZh(r,a)        r=(StgDouble) SAFESTGCALL1(StgDouble,(void *, StgDouble),tan,a)
 560 #define asinDoubleZh(r,a)       r=(StgDouble) SAFESTGCALL1(StgDouble,(void *, StgDouble),asin,a)
 561 #define acosDoubleZh(r,a)       r=(StgDouble) SAFESTGCALL1(StgDouble,(void *, StgDouble),acos,a)
 562 #define atanDoubleZh(r,a)       r=(StgDouble) SAFESTGCALL1(StgDouble,(void *, StgDouble),atan,a)
 563 #define sinhDoubleZh(r,a)       r=(StgDouble) SAFESTGCALL1(StgDouble,(void *, StgDouble),sinh,a)
 564 #define coshDoubleZh(r,a)       r=(StgDouble) SAFESTGCALL1(StgDouble,(void *, StgDouble),cosh,a)
 565 #define tanhDoubleZh(r,a)       r=(StgDouble) SAFESTGCALL1(StgDouble,(void *, StgDouble),tanh,a)
 566 /* Power: **## */
 567 #define ZtZtZhZh(r,a,b) r=(StgDouble) SAFESTGCALL2(StgDouble,(void *, StgDouble,StgDouble),pow,a,b)
 568 \end{code}
 569
 570 %************************************************************************
 571 %*                                                                      *
 572 \subsubsection[StgMacros-integer-primops]{Primitive @Integer@-related ops (GMP stuff)}
 573 %*                                                                      *
 574 %************************************************************************
 575
 576 Dirty macros we use for the real business.
 577
 578 INVARIANT: When one of these macros is called, the only live data is
 579 tidily on the STG stacks or in the STG registers (the code generator
 580 ensures this).  If there are any pointer-arguments, they will be in
 581 the first \tr{Ret*} registers (e.g., \tr{da} arg of \tr{gmpTake1Return1}).
 582
 583 OK, here are the real macros:
 584 \begin{code}
 585 #define gmpTake1Return1(size_chk_macro, liveness, mpz_op, ar,sr,dr, aa,sa,da)   \
 586 { MP_INT arg;                                                                   \
 587   MP_INT result;                                                                \
 588   I_ space = size_chk_macro(sa);                                                \
 589                                                                                 \
 590   /* Check that there will be enough heap & make Hp visible to GMP allocator */ \
 591   GMP_HEAP_LOOKAHEAD(liveness,space);                                           \
 592                                                                                 \
 593   /* Now we can initialise (post possible GC) */                                \
 594   arg.alloc     = (aa);                                                         \
 595   arg.size      = (sa);                                                         \
 596   arg.d         = (unsigned long int *) (BYTE_ARR_CTS(da));                     \
 597                                                                                 \
 598   SAFESTGCALL1(void,(void *, MP_INT *),mpz_init,&result);                       \
 599                                                                                 \
 600   /* Perform the operation */                                                   \
 601   SAFESTGCALL2(void,(void *, MP_INT *, MP_INT *),mpz_op,&result,&arg);          \
 602                                                                                 \
 603   GMP_HEAP_HANDBACK();          /* restore Hp */                                \
 604   (ar) = result.alloc;                                                          \
 605   (sr) = result.size;                                                           \
 606   (dr) = (B_) (result.d - DATA_HS);                                             \
 607   /* pt to *beginning* of object (GMP has been monkeying around in the middle) */ \
 608 }
 609
 610
 611 #define gmpTake2Return1(size_chk_macro, liveness, mpz_op, ar,sr,dr, a1,s1,d1, a2,s2,d2)\
 612 { MP_INT arg1;                                                                  \
 613   MP_INT arg2;                                                                  \
 614   MP_INT result;                                                                \
 615   I_ space = size_chk_macro(s1,s2);                                             \
 616                                                                                 \
 617   /* Check that there will be enough heap & make Hp visible to GMP allocator */ \
 618   GMP_HEAP_LOOKAHEAD(liveness,space);                                           \
 619                                                                                 \
 620   /* Now we can initialise (post possible GC) */                                \
 621   arg1.alloc    = (a1);                                                         \
 622   arg1.size     = (s1);                                                         \
 623   arg1.d        = (unsigned long int *) (BYTE_ARR_CTS(d1));                     \
 624   arg2.alloc    = (a2);                                                         \
 625   arg2.size     = (s2);                                                         \
 626   arg2.d        = (unsigned long int *) (BYTE_ARR_CTS(d2));                     \
 627                                                                                 \
 628   SAFESTGCALL1(void,(void *, MP_INT *),mpz_init,&result);                       \
 629                                                                                 \
 630   /* Perform the operation */                                                   \
 631   SAFESTGCALL3(void,(void *, MP_INT *, MP_INT *, MP_INT *),mpz_op,&result,&arg1,&arg2); \
 632                                                                                 \
 633   GMP_HEAP_HANDBACK();          /* restore Hp */                                \
 634   (ar) = result.alloc;                                                          \
 635   (sr) = result.size;                                                           \
 636   (dr) = (B_) (result.d - DATA_HS);                                             \
 637   /* pt to *beginning* of object (GMP has been monkeying around in the middle) */ \
 638 }
 639
 640 #define gmpTake2Return2(size_chk_macro, liveness, mpz_op, ar1,sr1,dr1, ar2,sr2,dr2, a1,s1,d1, a2,s2,d2) \
 641 { MP_INT arg1;                                                                  \
 642   MP_INT arg2;                                                                  \
 643   MP_INT result1;                                                               \
 644   MP_INT result2;                                                               \
 645   I_ space = size_chk_macro(s1,s2);                                             \
 646                                                                                 \
 647   /* Check that there will be enough heap & make Hp visible to GMP allocator */ \
 648   GMP_HEAP_LOOKAHEAD(liveness,space);                                           \
 649                                                                                 \
 650   /* Now we can initialise (post possible GC) */                                \
 651   arg1.alloc    = (a1);                                                         \
 652   arg1.size     = (s1);                                                         \
 653   arg1.d        = (unsigned long int *) (BYTE_ARR_CTS(d1));                     \
 654   arg2.alloc    = (a2);                                                         \
 655   arg2.size     = (s2);                                                         \
 656   arg2.d        = (unsigned long int *) (BYTE_ARR_CTS(d2));                     \
 657                                                                                 \
 658   SAFESTGCALL1(void,(void *, MP_INT *),mpz_init,&result1);                      \
 659   SAFESTGCALL1(void,(void *, MP_INT *),mpz_init,&result2);                      \
 660                                                                                 \
 661   /* Perform the operation */                                                   \
 662   SAFESTGCALL4(void,(void *, MP_INT *, MP_INT *, MP_INT *, MP_INT *),mpz_op,&result1,&result2,&arg1,&arg2); \
 663                                                                                 \
 664   GMP_HEAP_HANDBACK();          /* restore Hp */                                \
 665   (ar1) = result1.alloc;                                                        \
 666   (sr1) = result1.size;                                                         \
 667   (dr1) = (B_) (result1.d - DATA_HS);                                           \
 668   (ar2) = result2.alloc;                                                        \
 669   (sr2) = result2.size;                                                         \
 670   (dr2) = (B_) (result2.d - DATA_HS);                                           \
 671 }
 672 \end{code}
 673
 674 Some handy size-munging macros: sometimes gratuitously {\em conservative}.
 675 The \tr{+16} is to allow for the initial allocation of \tr{MP_INT} results.
 676 The \tr{__abs} stuff is because negative-ness of GMP things is encoded
 677 in their ``size''...
 678 \begin{code}
 679 #define __abs(a)                (( (a) >= 0 ) ? (a) : (-(a)))
 680 #define GMP_SIZE_ONE()          (2 + DATA_HS + 16)
 681 #define GMP_SAME_SIZE(a)        (__abs(a) + DATA_HS + 16)
 682 #define GMP_MAX_SIZE(a,b)       ((__abs(a) > __abs(b) ? __abs(a) : __abs(b)) + 1 + DATA_HS + 16)
 683                                 /* NB: the +1 is for the carry (or whatever) */
 684 #define GMP_2MAX_SIZE(a,b)      (2 * GMP_MAX_SIZE(a,b))
 685 #define GMP_ADD_SIZES(a,b)      (__abs(a) + __abs(b) + 1 + DATA_HS + 16)
 686                                 /* the +1 may just be paranoia */
 687 \end{code}
 688
 689 For the Integer/GMP stuff, we have macros that {\em look ahead} for
 690 some space, but don't actually grab it.
 691
 692 If there are live pointers at the time of the lookahead, the caller
 693 must make sure they are in \tr{Ret1}, \tr{Ret2}, ..., so they can be
 694 handled normally.  We achieve this by having the code generator {\em
 695 always} pass args to may-invoke-GC primitives in registers, using the
 696 normal pointers-first policy.  This means that, if we do go to garbage
 697 collection, everything is already in the Right Place.
 698
 699 Saving and restoring Hp register so the MP allocator can see them. If we are
 700 performing liftime profiling need to save and restore HpLim as well so that
 701 it can be bumped if allocation occurs.
 702
 703 The second argument to @GMP_HEAP_LOOKAHEAD@ must be an lvalue so that
 704 it can be restored from @TSO_ARG1@ after a failed @HEAP_CHK@ in
 705 threaded land.
 706
 707 \begin{code}
 708 #define GMP_HEAP_LOOKAHEAD(liveness,n)                  \
 709         do {                                            \
 710         HEAP_CHK_AND_RESTORE_N(liveness,n,0);           \
 711         Hp = Hp - (n);                                  \
 712         UN_ALLOC_HEAP(n);       /* Undo ticky-ticky */  \
 713         SAVE_Hp = Hp;           /* Hand over the hp */  \
 714         DEBUG_SetGMPAllocBudget(n)                      \
 715         }while(0)
 716
 717 #define GMP_HEAP_HANDBACK()                             \
 718         Hp = SAVE_Hp;                                   \
 719         DEBUG_ResetGMPAllocBudget()
 720 \end{code}
 721
 722 \begin{code}
 723 void *stgAllocForGMP   PROTO((size_t size_in_bytes));
 724 void *stgReallocForGMP PROTO((void *ptr, size_t old_size, size_t new_size));
 725 void stgDeallocForGMP  PROTO((void *ptr, size_t size));
 726
 727 #ifdef ALLOC_DEBUG
 728 extern StgInt DEBUG_GMPAllocBudget;
 729 #define DEBUG_SetGMPAllocBudget(n)  DEBUG_GMPAllocBudget = (n);
 730 #define DEBUG_ResetGMPAllocBudget() DEBUG_GMPAllocBudget = 0;
 731 #else
 732 #define DEBUG_SetGMPAllocBudget(n)  /*nothing*/
 733 #define DEBUG_ResetGMPAllocBudget() /*nothing*/
 734 #endif
 735 \end{code}
 736
 737 The real business (defining Integer primops):
 738 \begin{code}
 739 #define negateIntegerZh(ar,sr,dr, liveness, aa,sa,da) \
 740         gmpTake1Return1(GMP_SAME_SIZE, liveness, mpz_neg, ar,sr,dr, aa,sa,da)
 741
 742 #define plusIntegerZh(ar,sr,dr, liveness, a1,s1,d1, a2,s2,d2) \
 743         gmpTake2Return1(GMP_MAX_SIZE, liveness, mpz_add, ar,sr,dr, a1,s1,d1, a2,s2,d2)
 744 #define minusIntegerZh(ar,sr,dr, liveness, a1,s1,d1, a2,s2,d2) \
 745         gmpTake2Return1(GMP_MAX_SIZE, liveness, mpz_sub, ar,sr,dr, a1,s1,d1, a2,s2,d2)
 746 #define timesIntegerZh(ar,sr,dr, liveness, a1,s1,d1, a2,s2,d2) \
 747         gmpTake2Return1(GMP_ADD_SIZES, liveness, mpz_mul, ar,sr,dr, a1,s1,d1, a2,s2,d2)
 748
 749 /* div, mod, quot, rem are defined w/ quotRem & divMod */
 750
 751 #define quotRemIntegerZh(ar1,sr1,dr1, ar2,sr2,dr2, liveness, a1,s1,d1, a2,s2,d2) \
 752         gmpTake2Return2(GMP_2MAX_SIZE, liveness, mpz_divmod, ar1,sr1,dr1, ar2,sr2,dr2, a1,s1,d1, a2,s2,d2)
 753 #define divModIntegerZh(ar1,sr1,dr1, ar2,sr2,dr2, liveness,  a1,s1,d1, a2,s2,d2) \
 754         gmpTake2Return2(GMP_2MAX_SIZE, liveness, mpz_mdivmod, ar1,sr1,dr1, ar2,sr2,dr2, a1,s1,d1, a2,s2,d2)
 755 \end{code}
 756
 757 Comparison ops (@<@, @>=@, etc.) are defined in terms of the cmp
 758 fellow (returns -ve, 0, or +ve).
 759 \begin{code}
 760 #define cmpIntegerZh(r, hp, a1,s1,d1, a2,s2,d2) /* calls mpz_cmp */             \
 761 { MP_INT arg1;                                                                  \
 762   MP_INT arg2;                                                                  \
 763   /* Does not allocate memory */                                                \
 764                                                                                 \
 765   arg1.alloc    = (a1);                                                         \
 766   arg1.size     = (s1);                                                         \
 767   arg1.d        = (unsigned long int *) (BYTE_ARR_CTS(d1));                     \
 768   arg2.alloc    = (a2);                                                         \
 769   arg2.size     = (s2);                                                         \
 770   arg2.d        = (unsigned long int *) (BYTE_ARR_CTS(d2));                     \
 771                                                                                 \
 772   (r) = SAFESTGCALL2(I_,(void *, MP_INT *, MP_INT *),mpz_cmp,&arg1,&arg2);      \
 773 }
 774 \end{code}
 775
 776 Coercions:
 777
 778 \begin{code}
 779 #define integer2IntZh(r, hp, aa,sa,da)                                          \
 780 { MP_INT arg;                                                                   \
 781   /* Does not allocate memory */                                                \
 782                                                                                 \
 783   arg.alloc     = (aa);                                                         \
 784   arg.size      = (sa);                                                         \
 785   arg.d         = (unsigned long int *) (BYTE_ARR_CTS(da));                     \
 786                                                                                 \
 787   (r) = SAFESTGCALL1(I_,(void *, MP_INT *),mpz_get_si,&arg);                    \
 788 }
 789
 790 /* Since we're forced to know a little bit about MP_INT layout to do this with
 791    pre-allocated heap, we just inline the whole of mpz_init_set_si here.
 792         ** DIRE WARNING.  if mpz_init_set_si changes, so does this! ***
 793 */
 794
 795 #define int2IntegerZh(ar,sr,dr, hp, i)                                          \
 796 { StgInt val; /* to snaffle arg to avoid aliasing */                            \
 797                                                                                 \
 798   val = (i);  /* snaffle... */                                                  \
 799                                                                                 \
 800   SET_DATA_HDR((hp),ArrayOfData_info,CCC,DATA_VHS+MIN_MP_INT_SIZE,0);           \
 801                                                                                 \
 802   if      ((val) < 0) { (sr) = -1; (hp)[DATA_HS] = -(val); }                    \
 803   else if ((val) > 0) { (sr) =  1; (hp)[DATA_HS] =  (val); }                    \
 804   else /* val==0 */   { (sr) =  0; }                                            \
 805   (ar) = 1;                                                                     \
 806   (dr) = (B_)(hp);              /* dr is an StgByteArray */                     \
 807 }
 808
 809 #define word2IntegerZh(ar,sr,dr, hp, i)                                         \
 810 { StgWord val; /* to snaffle arg to avoid aliasing */                           \
 811                                                                                 \
 812   val = (i);  /* snaffle... */                                                  \
 813                                                                                 \
 814   SET_DATA_HDR((hp),ArrayOfData_info,CCC,DATA_VHS+MIN_MP_INT_SIZE,0);           \
 815                                                                                 \
 816   if ((val) != 0)     { (sr) =  1; (hp)[DATA_HS] =  (val); }                    \
 817   else /* val==0 */   { (sr) =  0; }                                            \
 818   (ar) = 1;                                                                     \
 819   (dr) = (B_)(hp);              /* dr is an StgByteArray */                     \
 820 }
 821
 822 #define integer2WordZh(r, hp, aa,sa,da)                                         \
 823 { MP_INT arg;                                                                   \
 824   /* Does not allocate memory */                                                \
 825                                                                                 \
 826   arg.alloc     = (aa);                                                         \
 827   arg.size      = (sa);                                                         \
 828   arg.d         = (unsigned long int *) (BYTE_ARR_CTS(da));                     \
 829                                                                                 \
 830   (r) = SAFESTGCALL1(I_,(void *, MP_INT *),mpz_get_ui,&arg);                    \
 831 }
 832
 833 \end{code}
 834
 835 Then there are a few oddments to make life easier:
 836 \begin{code}
 837 /*
 838    DIRE WARNING.
 839    The "str" argument must be a literal C string.
 840
 841         addr2Integer( ..., "foo")   OK!
 842
 843         x = "foo";
 844         addr2Integer( ..., x)       NO! NO!
 845 */
 846
 847 #define addr2IntegerZh(ar,sr,dr, liveness, str)                                 \
 848 { MP_INT result;                                                                \
 849   /* taking the number of bytes/8 as the number of words of lookahead           \
 850      is plenty conservative */                                                  \
 851   I_ space = GMP_SAME_SIZE(sizeof(str) / 8 + 1);                                \
 852                                                                                 \
 853   GMP_HEAP_LOOKAHEAD(liveness, space);                                          \
 854                                                                                 \
 855   /* Perform the operation */                                                   \
 856   if (SAFESTGCALL3(I_,(void *, MP_INT *, char *, int), mpz_init_set_str,&result,(str),/*base*/10)) \
 857       abort();                                                                  \
 858                                                                                 \
 859   GMP_HEAP_HANDBACK();          /* restore Hp */                                \
 860   (ar) = result.alloc;                                                          \
 861   (sr) = result.size;                                                           \
 862   (dr) = (B_) (result.d - DATA_HS);                                             \
 863   /* pt to *beginning* of object (GMP has been monkeying around in the middle) */ \
 864 }
 865 \end{code}
 866
 867 Encoding and decoding float-ish things is pretty Integer-ish.  We use
 868 these pretty magical support functions, essentially stolen from Lennart:
 869 \begin{code}
 870 StgFloat  __encodeFloat  PROTO((MP_INT *, I_));
 871 void      __decodeFloat  PROTO((MP_INT * /*result1*/,
 872                                 I_ * /*result2*/,
 873                                 StgFloat));
 874
 875 StgDouble __encodeDouble PROTO((MP_INT *, I_));
 876 void      __decodeDouble PROTO((MP_INT * /*result1*/,
 877                                 I_ * /*result2*/,
 878                                 StgDouble));
 879 \end{code}
 880
 881 Some floating-point format info, made with the \tr{enquire} program
 882 (version~4.3) [comes with gcc].
 883 \begin{code}
 884 /* this should be done by CPU architecture, insofar as possible [WDP] */
 885
 886 #if sparc_TARGET_ARCH   \
 887  || alpha_TARGET_ARCH   \
 888  || hppa1_1_TARGET_ARCH \
 889  || i386_TARGET_ARCH    \
 890  || m68k_TARGET_ARCH    \
 891  || mipsel_TARGET_ARCH  \
 892  || mipseb_TARGET_ARCH  \
 893  || powerpc_TARGET_ARCH \
 894  || rs6000_TARGET_ARCH
 895
 896 /* yes, it is IEEE floating point */
 897 #include "ieee-flpt.h"
 898
 899 #if alpha_TARGET_ARCH   \
 900  || i386_TARGET_ARCH            \
 901  || mipsel_TARGET_ARCH
 902
 903 #undef BIGENDIAN /* little-endian weirdos... */
 904 #else
 905 #define BIGENDIAN 1
 906 #endif
 907
 908 #else /* unknown floating-point format */
 909
 910 ******* ERROR *********** Any ideas about floating-point format?
 911
 912 #endif /* unknown floating-point */
 913 \end{code}
 914
 915 \begin{code}
 916 #if alpha_TARGET_ARCH
 917 #define encodeFloatZh(r, hp, aa,sa,da, expon)   encodeDoubleZh(r, hp, aa,sa,da, expon)
 918 #else
 919 #define encodeFloatZh(r, hp, aa,sa,da, expon)   \
 920 { MP_INT arg;                                   \
 921   /* Does not allocate memory */                \
 922                                                 \
 923   arg.alloc     = aa;                           \
 924   arg.size      = sa;                           \
 925   arg.d         = (unsigned long int *) (BYTE_ARR_CTS(da)); \
 926                                                 \
 927   r = SAFESTGCALL2(StgFloat,(void *, MP_INT *, I_), __encodeFloat,&arg,(expon));        \
 928 }
 929 #endif /* ! alpha */
 930
 931 #define encodeDoubleZh(r, hp, aa,sa,da, expon)  \
 932 { MP_INT arg;                                   \
 933   /* Does not allocate memory */                \
 934                                                 \
 935   arg.alloc     = aa;                           \
 936   arg.size      = sa;                           \
 937   arg.d         = (unsigned long int *) (BYTE_ARR_CTS(da)); \
 938                                                 \
 939   r = SAFESTGCALL2(StgDouble,(void *, MP_INT *, I_), __encodeDouble,&arg,(expon));\
 940 }
 941
 942 #if alpha_TARGET_ARCH
 943 #define decodeFloatZh(exponr, ar,sr,dr, hp, f)  decodeDoubleZh(exponr, ar,sr,dr, hp, f)
 944 #else
 945 #define decodeFloatZh(exponr, ar,sr,dr, hp, f)                          \
 946 { MP_INT mantissa;                                                      \
 947   I_ exponent;                                                          \
 948   StgFloat arg = (f);                                                   \
 949                                                                         \
 950   /* Be prepared to tell Lennart-coded __decodeFloat    */              \
 951   /* where mantissa.d can be put (it does not care about the rest) */   \
 952   SET_DATA_HDR(hp,ArrayOfData_info,CCC,DATA_VHS+MIN_MP_INT_SIZE,0);     \
 953   mantissa.d = (hp) + DATA_HS;                                          \
 954                                                                         \
 955   /* Perform the operation */                                           \
 956   SAFESTGCALL3(void,(void *, MP_INT *, I_ *, StgFloat),__decodeFloat,&mantissa,&exponent,arg);          \
 957   exponr= exponent;                                                     \
 958   ar    = mantissa.alloc;                                               \
 959   sr    = mantissa.size;                                                \
 960   dr    = (B_)(hp);                                                     \
 961 }
 962 #endif /* !alpha */
 963
 964 #define decodeDoubleZh(exponr, ar,sr,dr, hp, f)                         \
 965 { MP_INT mantissa;                                                      \
 966   I_ exponent;                                                          \
 967   StgDouble arg = (f);                                                  \
 968                                                                         \
 969   /* Be prepared to tell Lennart-coded __decodeDouble   */              \
 970   /* where mantissa.d can be put (it does not care about the rest) */   \
 971   SET_DATA_HDR(hp,ArrayOfData_info,CCC,DATA_VHS+MIN_MP_INT_SIZE,0);     \
 972   mantissa.d = (hp) + DATA_HS;                                          \
 973                                                                         \
 974   /* Perform the operation */                                           \
 975   SAFESTGCALL3(void,(void *, MP_INT *, I_ *, StgDouble),__decodeDouble,&mantissa,&exponent,arg);                \
 976   exponr= exponent;                                                     \
 977   ar    = mantissa.alloc;                                               \
 978   sr    = mantissa.size;                                                \
 979   dr    = (B_)(hp);                                                     \
 980 }
 981 \end{code}
 982
 983 %************************************************************************
 984 %*                                                                      *
 985 \subsubsection[StgMacros-mv-floats]{Moving floats and doubles around (e.g., to/from stacks)}
 986 %*                                                                      *
 987 %************************************************************************
 988
 989 With GCC, we use magic non-standard inlining; for other compilers, we
 990 just use functions (see also \tr{runtime/prims/PrimArith.lc}).
 991
 992 (The @OMIT_...@ is only used in compiling some of the RTS, none of
 993 which uses these anyway.)
 994
 995 \begin{code}
 996 #if alpha_TARGET_ARCH   \
 997  || i386_TARGET_ARCH    \
 998  || m68k_TARGET_ARCH
 999
1000 #define ASSIGN_FLT(dst, src) *(StgFloat *)(dst) = (src);
1001 #define PK_FLT(src) (*(StgFloat *)(src))
1002
1003 #define ASSIGN_DBL(dst, src) *(StgDouble *)(dst) = (src);
1004 #define PK_DBL(src) (*(StgDouble *)(src))
1005
1006 #else   /* not m68k || alpha || i[34]86 */
1007
1008 /* Special handling for machines with troublesome alignment constraints */
1009
1010 #define FLOAT_ALIGNMENT_TROUBLES    TRUE
1011
1012 #if ! defined(__GNUC__) || ! defined(__STG_GCC_REGS__)
1013
1014 void        ASSIGN_DBL PROTO((W_ [], StgDouble));
1015 StgDouble   PK_DBL     PROTO((W_ []));
1016 void        ASSIGN_FLT PROTO((W_ [], StgFloat));
1017 StgFloat    PK_FLT     PROTO((W_ []));
1018
1019 #else /* yes, its __GNUC__ && we really want them */
1020
1021 #if sparc_TARGET_ARCH
1022
1023 #define ASSIGN_FLT(dst, src) *(StgFloat *)(dst) = (src);
1024 #define PK_FLT(src) (*(StgFloat *)(src))
1025
1026 #define ASSIGN_DBL(dst,src) \
1027       __asm__("st %2,%0\n\tst %R2,%1" : "=m" (((P_)(dst))[0]), \
1028         "=m" (((P_)(dst))[1]) : "f" (src));
1029
1030 #define PK_DBL(src) \
1031     ( { register double d; \
1032       __asm__("ld %1,%0\n\tld %2,%R0" : "=f" (d) : \
1033         "m" (((P_)(src))[0]), "m" (((P_)(src))[1])); d; \
1034     } )
1035
1036 #else /* ! sparc */
1037
1038 /* (not very) forward prototype declarations */
1039 void        ASSIGN_DBL PROTO((W_ [], StgDouble));
1040 StgDouble   PK_DBL     PROTO((W_ []));
1041 void        ASSIGN_FLT PROTO((W_ [], StgFloat));
1042 StgFloat    PK_FLT     PROTO((W_ []));
1043
1044 extern STG_INLINE
1045 void
1046 ASSIGN_DBL(W_ p_dest[], StgDouble src)
1047 {
1048     double_thing y;
1049     y.d = src;
1050     p_dest[0] = y.du.dhi;
1051     p_dest[1] = y.du.dlo;
1052 }
1053
1054 /* GCC also works with this version, but it generates
1055    the same code as the previous one, and is not ANSI
1056
1057 #define ASSIGN_DBL( p_dest, src ) \
1058         *p_dest = ((double_thing) src).du.dhi; \
1059         *(p_dest+1) = ((double_thing) src).du.dlo \
1060 */
1061
1062 extern STG_INLINE
1063 StgDouble
1064 PK_DBL(W_ p_src[])
1065 {
1066     double_thing y;
1067     y.du.dhi = p_src[0];
1068     y.du.dlo = p_src[1];
1069     return(y.d);
1070 }
1071
1072 extern STG_INLINE
1073 void
1074 ASSIGN_FLT(W_ p_dest[], StgFloat src)
1075 {
1076     float_thing y;
1077     y.f = src;
1078     *p_dest = y.fu;
1079 }
1080
1081 extern STG_INLINE
1082 StgFloat
1083 PK_FLT(W_ p_src[])
1084 {
1085     float_thing y;
1086     y.fu = *p_src;
1087     return(y.f);
1088 }
1089
1090 #endif /* ! sparc */
1091
1092 #endif /* __GNUC__ */
1093
1094 #endif /* not __m68k__ */
1095 \end{code}
1096
1097 %************************************************************************
1098 %*                                                                      *
1099 \subsubsection[StgMacros-array-primops]{Primitive arrays}
1100 %*                                                                      *
1101 %************************************************************************
1102
1103 We regularly use this macro to fish the ``contents'' part
1104 out of a DATA or TUPLE closure, which is what is used for
1105 non-ptr and ptr arrays (respectively).
1106
1107 BYTE_ARR_CTS returns a @C_ *@!
1108
1109 We {\em ASSUME} we can use the same macro for both!!
1110 \begin{code}
1111
1112 #ifdef DEBUG
1113 #define BYTE_ARR_CTS(a)                                 \
1114  ({ ASSERT(INFO_PTR(a) == (W_) ArrayOfData_info);       \
1115     ((C_ *) (((StgPtr) (a))+DATA_HS)); })
1116 #define PTRS_ARR_CTS(a)                                 \
1117  ({ ASSERT((INFO_PTR(a) == (W_) ArrayOfPtrs_info)       \
1118         || (INFO_PTR(a) == (W_) ImMutArrayOfPtrs_info));\
1119     ((a)+MUTUPLE_HS);} )
1120 #else
1121 #define BYTE_ARR_CTS(a)         ((char *) (((StgPtr) (a))+DATA_HS))
1122 #define PTRS_ARR_CTS(a)         ((a)+MUTUPLE_HS)
1123 #endif
1124
1125 /* sigh */
1126 extern I_ genSymZh(STG_NO_ARGS);
1127 extern I_ resetGenSymZh(STG_NO_ARGS);
1128 extern I_ incSeqWorldZh(STG_NO_ARGS);
1129
1130 extern I_ byteArrayHasNUL__ PROTO((const char *, I_));
1131 \end{code}
1132
1133 OK, the easy ops first: (all except \tr{newArr*}:
1134
1135 (OLD:) VERY IMPORTANT! The read/write/index primitive ops
1136 on @ByteArray#@s index the array using a {\em BYTE} offset, even
1137 if the thing begin gotten out is a multi-byte @Int#@, @Float#@ etc.
1138 This is because you might be trying to take apart a C struct, where
1139 the offset from the start of the struct isn't a multiple of the
1140 size of the thing you're getting.  Hence the @(char *)@ casts.
1141
1142 EVEN MORE IMPORTANT! The above is a lie.  The offsets for BlahArrays
1143 are in Blahs.  WDP 95/08
1144
1145 In the case of messing with @StgAddrs@ (@A_@), which are really \tr{void *},
1146 we cast to @P_@, because you can't index off an uncast \tr{void *}.
1147
1148 In the case of @Array#@ (which contain pointers), the offset is in units
1149 of one ptr (not bytes).
1150
1151 \begin{code}
1152 #define sameMutableArrayZh(r,a,b)       r=(I_)((a)==(b))
1153 #define sameMutableByteArrayZh(r,a,b)   r=(I_)((B_)(a)==(B_)(b))
1154
1155 #define readArrayZh(r,a,i)       r=((PP_) PTRS_ARR_CTS(a))[(i)]
1156
1157 #define readCharArrayZh(r,a,i)   indexCharOffAddrZh(r,BYTE_ARR_CTS(a),i)
1158 #define readIntArrayZh(r,a,i)    indexIntOffAddrZh(r,BYTE_ARR_CTS(a),i)
1159 #define readWordArrayZh(r,a,i)   indexWordOffAddrZh(r,BYTE_ARR_CTS(a),i)
1160 #define readAddrArrayZh(r,a,i)   indexAddrOffAddrZh(r,BYTE_ARR_CTS(a),i)
1161 #define readFloatArrayZh(r,a,i)  indexFloatOffAddrZh(r,BYTE_ARR_CTS(a),i)
1162 #define readDoubleArrayZh(r,a,i) indexDoubleOffAddrZh(r,BYTE_ARR_CTS(a),i)
1163
1164 /* result ("r") arg ignored in write macros! */
1165 #define writeArrayZh(a,i,v)     ((PP_) PTRS_ARR_CTS(a))[(i)]=(v)
1166
1167 #define writeCharArrayZh(a,i,v)   ((C_ *)(BYTE_ARR_CTS(a)))[i] = (v)
1168 #define writeIntArrayZh(a,i,v)    ((I_ *)(BYTE_ARR_CTS(a)))[i] = (v)
1169 #define writeWordArrayZh(a,i,v)   ((W_ *)(BYTE_ARR_CTS(a)))[i] = (v)
1170 #define writeAddrArrayZh(a,i,v)   ((PP_)(BYTE_ARR_CTS(a)))[i] = (v)
1171 #define writeFloatArrayZh(a,i,v)  \
1172         ASSIGN_FLT((P_) (((StgFloat *)(BYTE_ARR_CTS(a))) + i),v)
1173 #define writeDoubleArrayZh(a,i,v) \
1174         ASSIGN_DBL((P_) (((StgDouble *)(BYTE_ARR_CTS(a))) + i),v)
1175
1176 #define indexArrayZh(r,a,i)       r=((PP_) PTRS_ARR_CTS(a))[(i)]
1177
1178 #define indexCharArrayZh(r,a,i)   indexCharOffAddrZh(r,BYTE_ARR_CTS(a),i)
1179 #define indexIntArrayZh(r,a,i)    indexIntOffAddrZh(r,BYTE_ARR_CTS(a),i)
1180 #define indexWordArrayZh(r,a,i)   indexWordOffAddrZh(r,BYTE_ARR_CTS(a),i)
1181 #define indexAddrArrayZh(r,a,i)   indexAddrOffAddrZh(r,BYTE_ARR_CTS(a),i)
1182 #define indexFloatArrayZh(r,a,i)  indexFloatOffAddrZh(r,BYTE_ARR_CTS(a),i)
1183 #define indexDoubleArrayZh(r,a,i) indexDoubleOffAddrZh(r,BYTE_ARR_CTS(a),i)
1184
1185 #define indexCharOffForeignObjZh(r,fo,i)   indexCharOffAddrZh(r,ForeignObj_CLOSURE_DATA(fo),i)
1186 #define indexIntOffForeignObjZh(r,fo,i)    indexIntOffAddrZh(r,ForeignObj_CLOSURE_DATA(fo),i)
1187 #define indexWordOffForeignObjZh(r,fo,i)   indexWordOffAddrZh(r,ForeignObj_CLOSURE_DATA(fo),i)
1188 #define indexAddrOffForeignObjZh(r,fo,i)   indexAddrOffAddrZh(r,ForeignObj_CLOSURE_DATA(fo),i)
1189 #define indexFloatOffForeignObjZh(r,fo,i)  indexFloatOffAddrZh(r,ForeignObj_CLOSURE_DATA(fo),i)
1190 #define indexDoubleOffForeignObjZh(r,fo,i) indexDoubleOffAddrZh(r,ForeignObj_CLOSURE_DATA(fo),i)
1191
1192 #define indexCharOffAddrZh(r,a,i)   r= ((C_ *)(a))[i]
1193 #define indexIntOffAddrZh(r,a,i)    r= ((I_ *)(a))[i]
1194 #define indexWordOffAddrZh(r,a,i)   r= ((W_ *)(a))[i]
1195 #define indexAddrOffAddrZh(r,a,i)   r= ((PP_)(a))[i]
1196 #define indexFloatOffAddrZh(r,a,i)  r= PK_FLT((P_) (((StgFloat *)(a)) + i))
1197 #define indexDoubleOffAddrZh(r,a,i) r= PK_DBL((P_) (((StgDouble *)(a)) + i))
1198
1199 /* Freezing arrays-of-ptrs requires changing an info table, for the
1200    benefit of the generational collector.  It needs to scavenge mutable
1201    objects, even if they are in old space.  When they become immutable,
1202    they can be removed from this scavenge list.  */
1203 #define unsafeFreezeArrayZh(r,a)                                \
1204         do {                                            \
1205         P_ result;                                      \
1206         result=(P_) (a);                                \
1207         FREEZE_MUT_HDR(result,ImMutArrayOfPtrs_info);   \
1208         r = result;                                     \
1209         }while(0)
1210
1211 #define unsafeFreezeByteArrayZh(r,a)    r=(B_)(a)
1212
1213 #define sizeofByteArrayZh(r,a)        r=(W_)sizeof(W_)*(W_)(DATA_CLOSURE_SIZE(a)-DATA_VHS)
1214 #define sizeofMutableByteArrayZh(r,a) r=(W_)sizeof(W_)*(W_)(DATA_CLOSURE_SIZE(a)-DATA_VHS)
1215 \end{code}
1216
1217 Now the \tr{newArr*} ops:
1218
1219 \begin{code}
1220 /*
1221 --------------------
1222 Will: ToDo: we need to find suitable places to put this comment, and the
1223 "in-general" one which follows.
1224
1225 ************ Nota Bene.  The "n" in this macro is guaranteed to
1226 be a register, *not* (say) Node[1].  That means that it is guaranteed
1227 to survive GC, provided only that the register is kept unaltered.
1228 This is important, because "n" is used after the HEAP_CHK.
1229
1230 In general, *all* parameters to these primitive-op macros are always
1231 registers.  (Will: For exactly *which* primitive-op macros is this guaranteed?
1232 Exactly those which can trigger GC?)
1233 ------------------------
1234
1235 NOTE: the above may now be OLD (WDP 94/02/10)
1236 */
1237 \end{code}
1238
1239 For char arrays, the size is in {\em BYTES}.
1240
1241 \begin{code}
1242 #define newCharArrayZh(r,liveness,n)    newByteArray(r,liveness,(n) * sizeof(C_))
1243 #define newIntArrayZh(r,liveness,n)     newByteArray(r,liveness,(n) * sizeof(I_))
1244 #define newWordArrayZh(r,liveness,n)    newByteArray(r,liveness,(n) * sizeof(W_))
1245 #define newAddrArrayZh(r,liveness,n)    newByteArray(r,liveness,(n) * sizeof(P_))
1246 #define newFloatArrayZh(r,liveness,n)   newByteArray(r,liveness,(n) * sizeof(StgFloat))
1247 #define newDoubleArrayZh(r,liveness,n)  newByteArray(r,liveness,(n) * sizeof(StgDouble))
1248
1249 #define newByteArray(r,liveness,n)                              \
1250 {                                                               \
1251   P_ result;                                                    \
1252   I_ size;                                                      \
1253                                                                 \
1254   HEAP_CHK(liveness,DATA_HS+BYTES_TO_STGWORDS(n),0);            \
1255   size = BYTES_TO_STGWORDS(n);                                  \
1256   ALLOC_PRIM(DATA_HS,size,0,DATA_HS+size) /* ticky ticky */;    \
1257   CC_ALLOC(CCC,DATA_HS+size,ARR_K);                             \
1258                                                                 \
1259   result = Hp-(DATA_HS+size)+1;                                 \
1260   SET_DATA_HDR(result,ArrayOfData_info,CCC,DATA_VHS+size,0);    \
1261   r = (B_) result;                                              \
1262 }
1263 \end{code}
1264
1265 Arrays of pointers need to be initialised; uses \tr{TUPLES}!
1266 The initialisation value is guaranteed to be in a register,
1267 and will be indicated by the liveness mask, so it's ok to do
1268 a \tr{HEAP_CHK}, which may trigger GC.
1269
1270 \begin{code}
1271 /* The new array initialization routine for the NCG */
1272 void newArrZh_init PROTO((P_ result, I_ n, P_ init));
1273
1274 #define newArrayZh(r,liveness,n,init)                   \
1275 {                                                       \
1276   P_ p;                                                 \
1277   P_ result;                                            \
1278                                                         \
1279   HEAP_CHK(liveness, MUTUPLE_HS+(n),0);                 \
1280   ALLOC_PRIM(MUTUPLE_HS,(n),0,MUTUPLE_HS+(n)) /* ticky ticky */; \
1281   CC_ALLOC(CCC,MUTUPLE_HS+(n),ARR_K); /* cc prof */     \
1282                                                         \
1283   result = Hp + 1 - (MUTUPLE_HS+(n));                   \
1284   SET_MUTUPLE_HDR(result,ArrayOfPtrs_info,CCC,MUTUPLE_VHS+(n),0) \
1285   for (p = result+MUTUPLE_HS; p < (result+MUTUPLE_HS+(n)); p++) { \
1286         *p = (W_) (init);                               \
1287   }                                                     \
1288                                                         \
1289   r = result;                                           \
1290 }
1291 \end{code}
1292
1293 %************************************************************************
1294 %*                                                                      *
1295 \subsubsection[StgMacros-SynchVar-primops]{Synchronizing Variables PrimOps}
1296 %*                                                                      *
1297 %************************************************************************
1298
1299 \begin{code}
1300 ED_(PrelBase_Z91Z93_closure);
1301
1302 #define newSynchVarZh(r, hp)                            \
1303 {                                                       \
1304   ALLOC_PRIM(MUTUPLE_HS,3,0,MUTUPLE_HS+3) /* ticky ticky */; \
1305   CC_ALLOC(CCC,MUTUPLE_HS+3,ARR_K); /* cc prof */       \
1306   SET_SVAR_HDR(hp,EmptySVar_info,CCC);                  \
1307   SVAR_HEAD(hp) = SVAR_TAIL(hp) = SVAR_VALUE(hp) = PrelBase_Z91Z93_closure;     \
1308   r = hp;                                               \
1309 }
1310 \end{code}
1311
1312 \begin{code}
1313 #ifdef CONCURRENT
1314
1315 void Yield PROTO((W_));
1316
1317 #define takeMVarZh(r, liveness, node)                   \
1318 {                                                       \
1319   while (INFO_PTR(node) != (W_) FullSVar_info) {        \
1320     if (SVAR_HEAD(node) == PrelBase_Z91Z93_closure)             \
1321       SVAR_HEAD(node) = CurrentTSO;                     \
1322     else                                                \
1323       TSO_LINK(SVAR_TAIL(node)) = CurrentTSO;           \
1324     TSO_LINK(CurrentTSO) = (P_) PrelBase_Z91Z93_closure;                \
1325     SVAR_TAIL(node) = CurrentTSO;                       \
1326     DO_YIELD(liveness << 1);                            \
1327   }                                                     \
1328   SET_INFO_PTR(node, EmptySVar_info);                   \
1329   r = SVAR_VALUE(node);                                 \
1330   SVAR_VALUE(node) = PrelBase_Z91Z93_closure;                           \
1331 }
1332
1333 #else
1334
1335 #define takeMVarZh(r, liveness, node)                   \
1336 {                                                       \
1337   if (INFO_PTR(node) != (W_) FullSVar_info) {           \
1338     /* Don't wrap the calls; we're done with STG land */\
1339     fflush(stdout);                                     \
1340     fprintf(stderr, "takeMVar#: MVar is empty.\n");     \
1341     EXIT(EXIT_FAILURE);                                 \
1342   }                                                     \
1343   SET_INFO_PTR(node, EmptySVar_info);                   \
1344   r = SVAR_VALUE(node);                                 \
1345   SVAR_VALUE(node) = PrelBase_Z91Z93_closure;                           \
1346 }
1347
1348 #endif
1349 \end{code}
1350
1351 \begin{code}
1352 #ifdef CONCURRENT
1353
1354 #ifdef GRAN
1355
1356 /* Only difference to the !GRAN def: RunnableThreadsHd has been replaced by */
1357 /* ThreadQueueHd i.e. the tso is added at the end of the thread queue on */
1358 /* the CurrentProc. This means we have an implicit context switch after */
1359 /* putMVar even if unfair scheduling is used in GranSim (default)!  -- HWL */
1360
1361 #define putMVarZh(node, value)                          \
1362 {                                                       \
1363   P_ tso;                                               \
1364   if (INFO_PTR(node) == (W_) FullSVar_info) {           \
1365     /* Don't wrap the calls; we're done with STG land */\
1366     fflush(stdout);                                     \
1367     fprintf(stderr, "putMVar#: MVar already full.\n");  \
1368     EXIT(EXIT_FAILURE);                                 \
1369   }                                                     \
1370   SET_INFO_PTR(node, FullSVar_info);                    \
1371   SVAR_VALUE(node) = value;                             \
1372   tso = SVAR_HEAD(node);                                \
1373   if (tso != (P_) PrelBase_Z91Z93_closure) {                            \
1374     if (DO_QP_PROF)                                     \
1375       STGCALL3(void,(void *, char *, P_, P_),QP_Event2,do_qp_prof > 1 ? "RA" : "RG",tso,CurrentTSO);    \
1376     if (ThreadQueueHd == PrelBase_Z91Z93_closure)               \
1377       ThreadQueueHd = tso;                      \
1378     else                                                \
1379       TSO_LINK(ThreadQueueTl) = tso;            \
1380     ThreadQueueTl = tso;                                \
1381     SVAR_HEAD(node) = TSO_LINK(tso);                    \
1382     TSO_LINK(tso) = (P_) PrelBase_Z91Z93_closure;                       \
1383     if(SVAR_HEAD(node) == (P_) PrelBase_Z91Z93_closure)                 \
1384       SVAR_TAIL(node) = (P_) PrelBase_Z91Z93_closure;           \
1385   }                                                     \
1386 }
1387
1388 #else /* !GRAN */
1389
1390 #define putMVarZh(node, value)                          \
1391 {                                                       \
1392   P_ tso;                                               \
1393   if (INFO_PTR(node) == (W_) FullSVar_info) {           \
1394     /* Don't wrap the calls; we're done with STG land */\
1395     fflush(stdout);                                     \
1396     fprintf(stderr, "putMVar#: MVar already full.\n");  \
1397     EXIT(EXIT_FAILURE);                                 \
1398   }                                                     \
1399   SET_INFO_PTR(node, FullSVar_info);                    \
1400   SVAR_VALUE(node) = value;                             \
1401   tso = SVAR_HEAD(node);                                \
1402   if (tso != (P_) PrelBase_Z91Z93_closure) {                            \
1403     if (DO_QP_PROF)                                     \
1404       STGCALL3(void,(void *, char *, P_, P_),QP_Event2,do_qp_prof > 1 ? "RA" : "RG",tso,CurrentTSO);    \
1405     if (RunnableThreadsHd == PrelBase_Z91Z93_closure)                   \
1406       RunnableThreadsHd = tso;                          \
1407     else                                                \
1408       TSO_LINK(RunnableThreadsTl) = tso;                \
1409     RunnableThreadsTl = tso;                            \
1410     SVAR_HEAD(node) = TSO_LINK(tso);                    \
1411     TSO_LINK(tso) = (P_) PrelBase_Z91Z93_closure;                       \
1412     if(SVAR_HEAD(node) == (P_) PrelBase_Z91Z93_closure)                 \
1413       SVAR_TAIL(node) = (P_) PrelBase_Z91Z93_closure;           \
1414   }                                                     \
1415 }
1416
1417 #endif  /* GRAN */
1418
1419 #else
1420
1421 #define putMVarZh(node, value)                          \
1422 {                                                       \
1423   P_ tso;                                               \
1424   if (INFO_PTR(node) == (W_) FullSVar_info) {           \
1425     /* Don't wrap the calls; we're done with STG land */\
1426     fflush(stdout);                                     \
1427     fprintf(stderr, "putMVar#: MVar already full.\n");  \
1428     EXIT(EXIT_FAILURE);                                 \
1429   }                                                     \
1430   SET_INFO_PTR(node, FullSVar_info);                    \
1431   SVAR_VALUE(node) = value;                             \
1432 }
1433
1434 #endif
1435 \end{code}
1436
1437 \begin{code}
1438 #ifdef CONCURRENT
1439
1440 #define readIVarZh(r, liveness, node)                   \
1441 {                                                       \
1442   if (INFO_PTR(node) != (W_) ImMutArrayOfPtrs_info) {   \
1443     if (SVAR_HEAD(node) == PrelBase_Z91Z93_closure)             \
1444       SVAR_HEAD(node) = CurrentTSO;                     \
1445     else                                                \
1446       TSO_LINK(SVAR_TAIL(node)) = CurrentTSO;           \
1447     TSO_LINK(CurrentTSO) = (P_) PrelBase_Z91Z93_closure;                \
1448     SVAR_TAIL(node) = CurrentTSO;                       \
1449     DO_YIELD(liveness << 1);                            \
1450   }                                                     \
1451   r = SVAR_VALUE(node);                                 \
1452 }
1453
1454 #else
1455
1456 #define readIVarZh(r, liveness, node)                   \
1457 {                                                       \
1458   if (INFO_PTR(node) != (W_) ImMutArrayOfPtrs_info) {   \
1459     /* Don't wrap the calls; we're done with STG land */\
1460     fflush(stdout);                                     \
1461     fprintf(stderr, "readIVar#: IVar is empty.\n");     \
1462     EXIT(EXIT_FAILURE);                                 \
1463   }                                                     \
1464   r = SVAR_VALUE(node);                                 \
1465 }
1466
1467 #endif
1468 \end{code}
1469
1470 \begin{code}
1471 #ifdef CONCURRENT
1472
1473 #ifdef GRAN
1474
1475 /* Only difference to the !GRAN def: RunnableThreadsHd has been replaced by */
1476 /* ThreadQueueHd i.e. the tso is added at the end of the thread queue on */
1477 /* the CurrentProc. This means we have an implicit context switch after */
1478 /* writeIVar even if unfair scheduling is used in GranSim (default)!  -- HWL */
1479
1480 #define writeIVarZh(node, value)                        \
1481 {                                                       \
1482   P_ tso;                                               \
1483   if (INFO_PTR(node) == (W_) ImMutArrayOfPtrs_info) {   \
1484     /* Don't wrap the calls; we're done with STG land */\
1485     fflush(stdout);                                     \
1486     fprintf(stderr, "writeIVar#: IVar already full.\n");\
1487     EXIT(EXIT_FAILURE);                                 \
1488   }                                                     \
1489   tso = SVAR_HEAD(node);                                \
1490   if (tso != (P_) PrelBase_Z91Z93_closure) {                            \
1491     if (ThreadQueueHd == PrelBase_Z91Z93_closure)               \
1492       ThreadQueueHd = tso;                      \
1493     else                                                \
1494       TSO_LINK(ThreadQueueTl) = tso;            \
1495     while(TSO_LINK(tso) != PrelBase_Z91Z93_closure) {                   \
1496       if (DO_QP_PROF)                                   \
1497         STGCALL3(void,(void *, char *, P_, P_),QP_Event2,do_qp_prof > 1 ? "RA" : "RG",tso,CurrentTSO);  \
1498       tso = TSO_LINK(tso);                              \
1499     }                                                   \
1500     if (DO_QP_PROF)                                     \
1501       STGCALL3(void,(void *, char *, P_, P_),QP_Event2,do_qp_prof > 1 ? "RA" : "RG",tso,CurrentTSO);    \
1502     ThreadQueueTl = tso;                                \
1503   }                                                     \
1504   /* Don't use freeze, since it's conditional on GC */  \
1505   SET_INFO_PTR(node, ImMutArrayOfPtrs_info);            \
1506   MUTUPLE_CLOSURE_SIZE(node) = (MUTUPLE_VHS+1);         \
1507   SVAR_VALUE(node) = value;                             \
1508 }
1509
1510 #else /* !GRAN */
1511
1512 #define writeIVarZh(node, value)                        \
1513 {                                                       \
1514   P_ tso;                                               \
1515   if (INFO_PTR(node) == (W_) ImMutArrayOfPtrs_info) {   \
1516     /* Don't wrap the calls; we're done with STG land */\
1517     fflush(stdout);                                     \
1518     fprintf(stderr, "writeIVar#: IVar already full.\n");\
1519     EXIT(EXIT_FAILURE);                                 \
1520   }                                                     \
1521   tso = SVAR_HEAD(node);                                \
1522   if (tso != (P_) PrelBase_Z91Z93_closure) {                            \
1523     if (RunnableThreadsHd == PrelBase_Z91Z93_closure)                   \
1524       RunnableThreadsHd = tso;                          \
1525     else                                                \
1526       TSO_LINK(RunnableThreadsTl) = tso;                \
1527     while(TSO_LINK(tso) != PrelBase_Z91Z93_closure) {                   \
1528       if (DO_QP_PROF)                                   \
1529         STGCALL3(void,(void *, char *, P_, P_),QP_Event2,do_qp_prof > 1 ? "RA" : "RG",tso,CurrentTSO);  \
1530       tso = TSO_LINK(tso);                              \
1531     }                                                   \
1532     if (DO_QP_PROF)                                     \
1533       STGCALL3(void,(void *, char *, P_, P_),QP_Event2,do_qp_prof > 1 ? "RA" : "RG",tso,CurrentTSO);    \
1534     RunnableThreadsTl = tso;                            \
1535   }                                                     \
1536   /* Don't use freeze, since it's conditional on GC */  \
1537   SET_INFO_PTR(node, ImMutArrayOfPtrs_info);            \
1538   MUTUPLE_CLOSURE_SIZE(node) = (MUTUPLE_VHS+1);         \
1539   SVAR_VALUE(node) = value;                             \
1540 }
1541
1542 #endif  /* GRAN */
1543
1544 #else
1545
1546 #define writeIVarZh(node, value)                        \
1547 {                                                       \
1548   P_ tso;                                               \
1549   if (INFO_PTR(node) == (W_) ImMutArrayOfPtrs_info) {   \
1550     /* Don't wrap the calls; we're done with STG land */\
1551     fflush(stdout);                                     \
1552     fprintf(stderr, "writeIVar#: IVar already full.\n");\
1553     EXIT(EXIT_FAILURE);                                 \
1554   }                                                     \
1555   /* Don't use freeze, since it's conditional on GC */  \
1556   SET_INFO_PTR(node, ImMutArrayOfPtrs_info);            \
1557   MUTUPLE_CLOSURE_SIZE(node) = (MUTUPLE_VHS+1);         \
1558   SVAR_VALUE(node) = value;                             \
1559 }
1560
1561 #endif
1562 \end{code}
1563
1564 %************************************************************************
1565 %*                                                                      *
1566 \subsubsection[StgMacros-Wait-primops]{Delay/Wait PrimOps}
1567 %*                                                                      *
1568 %************************************************************************
1569
1570 \begin{code}
1571 #ifdef CONCURRENT
1572
1573 /* ToDo: for GRAN */
1574
1575 #define delayZh(liveness, us)                           \
1576   {                                                     \
1577     if (WaitingThreadsTl == PrelBase_Z91Z93_closure)            \
1578       WaitingThreadsHd = CurrentTSO;                    \
1579     else                                                \
1580       TSO_LINK(WaitingThreadsTl) = CurrentTSO;          \
1581     WaitingThreadsTl = CurrentTSO;                      \
1582     TSO_LINK(CurrentTSO) = PrelBase_Z91Z93_closure;                     \
1583     TSO_EVENT(CurrentTSO) = (W_) ((us) < 1 ? 1 : (us)); \
1584     DO_YIELD(liveness << 1);                            \
1585   }
1586
1587 #else
1588
1589 #define delayZh(liveness, us)                           \
1590   {                                                     \
1591     fflush(stdout);                                     \
1592     fprintf(stderr, "delay#: unthreaded build.\n");     \
1593     EXIT(EXIT_FAILURE);                                 \
1594   }
1595
1596 #endif
1597
1598 #ifdef CONCURRENT
1599
1600 /* ToDo: something for GRAN */
1601
1602 #define waitReadZh(liveness, fd)                        \
1603   {                                                     \
1604     if (WaitingThreadsTl == PrelBase_Z91Z93_closure)            \
1605       WaitingThreadsHd = CurrentTSO;                    \
1606     else                                                \
1607       TSO_LINK(WaitingThreadsTl) = CurrentTSO;          \
1608     WaitingThreadsTl = CurrentTSO;                      \
1609     TSO_LINK(CurrentTSO) = PrelBase_Z91Z93_closure;                     \
1610     TSO_EVENT(CurrentTSO) = (W_) (-(fd));               \
1611     DO_YIELD(liveness << 1);                            \
1612   }
1613
1614 #else
1615
1616 #define waitReadZh(liveness, fd)                        \
1617   {                                                     \
1618     fflush(stdout);                                     \
1619     fprintf(stderr, "waitRead#: unthreaded build.\n");  \
1620     EXIT(EXIT_FAILURE);                                 \
1621   }
1622
1623 #endif
1624
1625 #ifdef CONCURRENT
1626
1627 /* ToDo: something for GRAN */
1628
1629 #ifdef HAVE_SYS_TYPES_H
1630 #include <sys/types.h>
1631 #endif  HAVE_SYS_TYPES_H */
1632
1633 #define waitWriteZh(liveness, fd)                       \
1634   {                                                     \
1635     if (WaitingThreadsTl == PrelBase_Z91Z93_closure)            \
1636       WaitingThreadsHd = CurrentTSO;                    \
1637     else                                                \
1638       TSO_LINK(WaitingThreadsTl) = CurrentTSO;          \
1639     WaitingThreadsTl = CurrentTSO;                      \
1640     TSO_LINK(CurrentTSO) = PrelBase_Z91Z93_closure;                     \
1641     TSO_EVENT(CurrentTSO) = (W_) (-(fd+FD_SETSIZE));    \
1642     DO_YIELD(liveness << 1);                            \
1643   }
1644
1645 #else
1646
1647 #define waitWriteZh(liveness, fd)                       \
1648   {                                                     \
1649     fflush(stdout);                                     \
1650     fprintf(stderr, "waitWrite#: unthreaded build.\n"); \
1651     EXIT(EXIT_FAILURE);                                 \
1652   }
1653
1654 #endif
1655
1656 \end{code}
1657
1658 %************************************************************************
1659 %*                                                                      *
1660 \subsubsection[StgMacros-IO-primops]{Primitive I/O, error-handling primops}
1661 %*                                                                      *
1662 %************************************************************************
1663
1664 \begin{code}
1665 extern P_ TopClosure;
1666 EXTFUN(ErrorIO_innards);
1667 EXTFUN(__std_entry_error__);
1668
1669 #define errorIOZh(a)            \
1670     do { TopClosure=(a);        \
1671          (void) SAFESTGCALL1(I_,(void *, FILE *),fflush,stdout); \
1672          (void) SAFESTGCALL1(I_,(void *, FILE *),fflush,stderr); \
1673          JMP_(ErrorIO_innards); \
1674     } while(0)
1675
1676 #if !defined(CALLER_SAVES_SYSTEM)
1677 /* can use the macros */
1678 #define stg_getc(stream)        getc((FILE *) (stream))
1679 #define stg_putc(c,stream)      putc((c),((FILE *) (stream)))
1680 #else
1681 /* must not use the macros (they contain embedded calls to _filbuf/whatnot) */
1682 #define stg_getc(stream)        SAFESTGCALL1(I_,(void *, FILE *),fgetc,(FILE *) (stream))
1683 #define stg_putc(c,stream)      SAFESTGCALL2(I_,(void *, char, FILE *),fputc,(c),((FILE *) (stream)))
1684 #endif
1685
1686 int initialize_virtual_timer(int us);
1687 int install_segv_handler(STG_NO_ARGS);
1688 int install_vtalrm_handler(STG_NO_ARGS);
1689 void initUserSignals(STG_NO_ARGS);
1690 void blockUserSignals(STG_NO_ARGS);
1691 void unblockUserSignals(STG_NO_ARGS);
1692 IF_RTS(void blockVtAlrmSignal(STG_NO_ARGS);)
1693 IF_RTS(void unblockVtAlrmSignal(STG_NO_ARGS);)
1694 IF_RTS(void AwaitEvent(I_ delta);)
1695
1696 #if  defined(_POSIX_SOURCE) && !defined(nextstep3_TARGET_OS)
1697         /* For nextstep3_TARGET_OS comment see stgdefs.h. CaS */
1698 extern I_ sig_install PROTO((I_, I_, sigset_t *));
1699 #define stg_sig_ignore(s,m)     SAFESTGCALL3(I_,(void *, I_, I_),sig_install,s,STG_SIG_IGN,(sigset_t *)m)
1700 #define stg_sig_default(s,m)    SAFESTGCALL3(I_,(void *, I_, I_),sig_install,s,STG_SIG_DFL,(sigset_t *)m)
1701 #define stg_sig_catch(s,sp,m)   SAFESTGCALL3(I_,(void *, I_, I_),sig_install,s,sp,(sigset_t *)m)
1702 #else
1703 extern I_ sig_install PROTO((I_, I_));
1704 #define stg_sig_ignore(s,m)     SAFESTGCALL2(I_,(void *, I_, I_),sig_install,s,STG_SIG_IGN)
1705 #define stg_sig_default(s,m)    SAFESTGCALL2(I_,(void *, I_, I_),sig_install,s,STG_SIG_DFL)
1706 #define stg_sig_catch(s,sp,m)   SAFESTGCALL2(I_,(void *, I_, I_),sig_install,s,sp)
1707 #endif
1708
1709 #define STG_SIG_DFL     (-1)
1710 #define STG_SIG_IGN     (-2)
1711 #define STG_SIG_ERR     (-3)
1712
1713 StgInt getErrorHandler(STG_NO_ARGS);
1714 #ifndef PAR
1715 void   raiseError PROTO((StgStablePtr handler));
1716 StgInt catchError PROTO((StgStablePtr newErrorHandler));
1717 #endif
1718 void decrementErrorCount(STG_NO_ARGS);
1719
1720 #define stg_catchError(sp)        SAFESTGCALL1(I_,(void *, StgStablePtr),catchError,sp)
1721 #define stg_decrementErrorCount() SAFESTGCALL0(void,(void *),decrementErrorCount)
1722 \end{code}
1723
1724 %************************************************************************
1725 %*                                                                      *
1726 \subsubsection[StgMacros-stable-ptr]{Primitive ops for manipulating stable pointers}
1727 %*                                                                      *
1728 %************************************************************************
1729
1730
1731 The type of these should be:
1732
1733 \begin{verbatim}
1734 makeStablePointer#  :: a -> State# _RealWorld -> StateAndStablePtr# _RealWorld a
1735 deRefStablePointer# :: StablePtr# a -> State# _RealWorld -> StateAndPtr _RealWorld a
1736 \end{verbatim}
1737
1738 Since world-tokens are no longer explicitly passed around, the
1739 implementations have a few less arguments/results.
1740
1741 The simpler one is @deRefStablePointer#@ (which is only a primop
1742 because it is more polymorphic than is allowed of a ccall).
1743
1744 \begin{code}
1745 #ifdef PAR
1746
1747 #define deRefStablePtrZh(ri,sp)                                     \
1748 do {                                                                \
1749     fflush(stdout);                                                 \
1750     fprintf(stderr, "deRefStablePtr#: no stable pointer support.\n");\
1751     EXIT(EXIT_FAILURE);                                             \
1752 } while(0)
1753
1754 #else /* !PAR */
1755
1756 extern StgPtr _deRefStablePointer PROTO((StgInt, StgPtr));
1757
1758 #define deRefStablePtrZh(ri,sp) \
1759    ri = SAFESTGCALL2(I_,(void *, I_, P_),_deRefStablePointer,sp,StorageMgrInfo.StablePointerTable);
1760 \end{code}
1761
1762 Declarations for other stable pointer operations.
1763
1764 \begin{code}
1765 void    freeStablePointer       PROTO((I_ stablePtr));
1766
1767 void    enterStablePtr          PROTO((StgStablePtr, StgFunPtr));
1768 void    performIO               PROTO((StgStablePtr));
1769 I_      enterInt                PROTO((StgStablePtr));
1770 I_      enterFloat              PROTO((StgStablePtr));
1771 P_      deRefStablePointer      PROTO((StgStablePtr));
1772 IF_RTS(I_ catchSoftHeapOverflow PROTO((StgStablePtr, I_));)
1773 IF_RTS(I_ getSoftHeapOverflowHandler(STG_NO_ARGS);)
1774 IF_RTS(extern StgStablePtr softHeapOverflowHandler;)
1775 IF_RTS(void shutdownHaskell(STG_NO_ARGS);)
1776
1777 EXTFUN(stopPerformIODirectReturn);
1778 EXTFUN(startPerformIO);
1779 EXTFUN(stopEnterIntDirectReturn);
1780 EXTFUN(startEnterInt);
1781 EXTFUN(stopEnterFloatDirectReturn);
1782 EXTFUN(startEnterFloat);
1783
1784 void enterStablePtr PROTO((StgStablePtr stableIndex, StgFunPtr startCode));
1785
1786 #endif /* !PAR */
1787
1788 IF_RTS(extern I_ ErrorIO_call_count;)
1789 \end{code}
1790
1791 Somewhat harder is @makeStablePointer#@ --- it is usually simple but
1792 if we're unlucky, it will have to allocate a new table and copy the
1793 old bit over.  Since we might, very occasionally, have to call the
1794 garbage collector, this has to be a macro... sigh!
1795
1796 NB @newSP@ is required because it is entirely possible that
1797 @stablePtr@ and @unstablePtr@ are aliases and so we can't do the
1798 assignment to @stablePtr@ until we've finished with @unstablePtr@.
1799
1800 Another obscure piece of coding is the recalculation of the size of
1801 the table.  We do this just in case Jim's threads decide they want to
1802 context switch---in which case any stack-allocated variables may get
1803 trashed.  (If only there was a special heap check which didn't
1804 consider context switching...)
1805
1806 \begin{code}
1807 #ifndef PAR
1808
1809 /* Calculate SP Table size from number of pointers */
1810 #define SPTSizeFromNoPtrs( newNP ) (DYN_VHS + 1 + 2 * (newNP))
1811
1812 /* Calculate number of pointers in new table from number in old table:
1813    any strictly increasing expression will do here */
1814 #define CalcNewNoSPtrs( i ) ((i)*2 + 100)
1815
1816 void enlargeSPTable PROTO((P_, P_));
1817
1818 #define makeStablePtrZh(stablePtr,liveness,unstablePtr)             \
1819 do {                                                                \
1820   EXTDATA_RO(StablePointerTable_info);                              \
1821   EXTDATA(UnusedSP);                                                \
1822   StgStablePtr newSP;                                               \
1823                                                                     \
1824   if (SPT_EMPTY(StorageMgrInfo.StablePointerTable)) { /* free stack is empty */ \
1825     { /* Variables used before the heap check */                    \
1826       I_ OldNoPtrs = SPT_NoPTRS( StorageMgrInfo.StablePointerTable ); \
1827       I_ NewNoPtrs = CalcNewNoSPtrs( OldNoPtrs );                   \
1828       I_ NewSize = SPTSizeFromNoPtrs( NewNoPtrs );                  \
1829       HEAP_CHK(liveness, _FHS+NewSize, 0);                          \
1830     }                                                               \
1831     { /* Variables used after the heap check - same values */       \
1832       I_ OldNoPtrs = SPT_NoPTRS( StorageMgrInfo.StablePointerTable ); \
1833       I_ NewNoPtrs = CalcNewNoSPtrs( OldNoPtrs );                   \
1834       I_ NewSize = SPTSizeFromNoPtrs( NewNoPtrs );                  \
1835       P_ SPTable = Hp + 1 - (_FHS + NewSize);                       \
1836                                                                     \
1837       CC_ALLOC(CCC, _FHS+NewSize, SPT_K); /* cc prof */             \
1838       SET_DYN_HDR(SPTable,StablePointerTable_info,CCC,NewSize,NewNoPtrs);\
1839       SAFESTGCALL2(void, (void *, P_, P_), enlargeSPTable, SPTable, StorageMgrInfo.StablePointerTable);      \
1840       StorageMgrInfo.StablePointerTable = SPTable;                  \
1841     }                                                               \
1842   }                                                                 \
1843                                                                     \
1844   newSP = SPT_POP(StorageMgrInfo.StablePointerTable);               \
1845   SPT_SPTR(StorageMgrInfo.StablePointerTable, newSP) = unstablePtr; \
1846   CHECK_SPT_CLOSURE( StorageMgrInfo.StablePointerTable );           \
1847   stablePtr = newSP;                                                \
1848 } while (0)
1849
1850 #else
1851
1852 #define makeStablePtrZh(stablePtr,liveness,unstablePtr)             \
1853 do {                                                                \
1854     fflush(stdout);                                                 \
1855     fprintf(stderr, "makeStablePtr#: no stable pointer support.\n");\
1856     EXIT(EXIT_FAILURE);                                             \
1857 } while(0)
1858
1859 #endif /* !PAR */
1860 \end{code}
1861
1862 %************************************************************************
1863 %*                                                                      *
1864 \subsubsection[StgMacros-unsafePointerEquality]{Primitive `op' for breaking referential transparency}
1865 %*                                                                      *
1866 %************************************************************************
1867
1868 The type of this is @reallyUnsafePtrEquality :: a -> a -> Int#@ so we
1869 can expect three parameters: the two arguments and a "register" to put
1870 the result into.
1871
1872 Message to Will: This primop breaks referential transparency so badly
1873 you might want to leave it out.  On the other hand, if you hide it
1874 away in an appropriate monad, it's perfectly safe. [ADR]
1875
1876 Note that this primop is non-deterministic: different results can be
1877 obtained depending on just what the garbage collector (and code
1878 optimiser??) has done.  However, we can guarantee that if two objects
1879 are pointer-equal, they have the same denotation --- the converse most
1880 certainly doesn't hold.
1881
1882 ToDo ADR: The degree of non-determinism could be greatly reduced by
1883 following indirections.
1884
1885 \begin{code}
1886 #define reallyUnsafePtrEqualityZh(r,a,b) r=((StgPtr)(a) == (StgPtr)(b))
1887 \end{code}
1888
1889 %************************************************************************
1890 %*                                                                      *
1891 \subsubsection[StgMacros-parallel-primop]{Primitive `op' for sparking (etc)}
1892 %*                                                                      *
1893 %************************************************************************
1894
1895 Assuming local sparking in some form, we can now inline the spark request.
1896
1897 We build a doubly-linked list in the heap, so that we can handle FIFO
1898 or LIFO scheduling as we please.
1899
1900 Anything with tag >= 0 is in WHNF, so we discard it.
1901
1902 \begin{code}
1903 #ifdef CONCURRENT
1904
1905 ED_(PrelBase_Z91Z93_closure);
1906 ED_(True_closure);
1907
1908 #if defined(GRAN)
1909 #define parZh(r,node)                           \
1910         PARZh(r,node,1,0,0,0,0,0)
1911
1912 #define parAtZh(r,node,where,identifier,gran_info,size_info,par_info,rest) \
1913         parATZh(r,node,where,identifier,gran_info,size_info,par_info,rest,1)
1914
1915 #define parAtAbsZh(r,node,proc,identifier,gran_info,size_info,par_info,rest) \
1916         parATZh(r,node,proc,identifier,gran_info,size_info,par_info,rest,2)
1917
1918 #define parAtRelZh(r,node,proc,identifier,gran_info,size_info,par_info,rest) \
1919         parATZh(r,node,proc,identifier,gran_info,size_info,par_info,rest,3)
1920
1921 #define parAtForNowZh(r,node,where,identifier,gran_info,size_info,par_info,rest)        \
1922         parATZh(r,node,where,identifier,gran_info,size_info,par_info,rest,0)
1923
1924 #define parATZh(r,node,where,identifier,gran_info,size_info,par_info,rest,local)        \
1925 {                                                       \
1926   sparkq result;                                                \
1927   if (SHOULD_SPARK(node)) {                             \
1928     SaveAllStgRegs();                                   \
1929     { sparkq result;                                            \
1930       result = NewSpark((P_)node,identifier,gran_info,size_info,par_info,local);        \
1931       if (local==2) {         /* special case for parAtAbs */   \
1932         GranSimSparkAtAbs(result,(I_)where,identifier);\
1933       } else if (local==3) {  /* special case for parAtRel */   \
1934         GranSimSparkAtAbs(result,(I_)(CurrentProc+where),identifier);   \
1935       } else {       \
1936         GranSimSparkAt(result,where,identifier);        \
1937       }        \
1938       context_switch = 1;                               \
1939     }                                                   \
1940     RestoreAllStgRegs();                                \
1941   } else if (do_qp_prof) {                              \
1942     I_ tid = threadId++;                                \
1943     SAFESTGCALL2(void,(I_, P_),QP_Event0,tid,node);     \
1944   }                                                     \
1945   r = 1; /* return code for successful spark -- HWL */  \
1946 }
1947
1948 #define parLocalZh(r,node,identifier,gran_info,size_info,par_info,rest) \
1949         PARZh(r,node,rest,identifier,gran_info,size_info,par_info,1)
1950
1951 #define parGlobalZh(r,node,identifier,gran_info,size_info,par_info,rest) \
1952         PARZh(r,node,rest,identifier,gran_info,size_info,par_info,0)
1953
1954 #if 1
1955
1956 #define PARZh(r,node,rest,identifier,gran_info,size_info,par_info,local) \
1957 {                                                       \
1958   if (SHOULD_SPARK(node)) {                             \
1959     SaveAllStgRegs();                                   \
1960     { sparkq result;                                            \
1961       result = NewSpark((P_)node,identifier,gran_info,size_info,par_info,local);\
1962       add_to_spark_queue(result);                               \
1963       GranSimSpark(local,(P_)node);                                     \
1964       context_switch = 1;                               \
1965     }                                                   \
1966     RestoreAllStgRegs();                                \
1967   } else if (do_qp_prof) {                              \
1968     I_ tid = threadId++;                                \
1969     SAFESTGCALL2(void,(I_, P_),QP_Event0,tid,node);     \
1970   }                                                     \
1971   r = 1; /* return code for successful spark -- HWL */  \
1972 }
1973
1974 #else
1975
1976 #define PARZh(r,node,rest,identifier,gran_info,size_info,par_info,local) \
1977 {                                                       \
1978   sparkq result;                                                \
1979   if (SHOULD_SPARK(node)) {                             \
1980     result = NewSpark((P_)node,identifier,gran_info,size_info,par_info,local);\
1981     ADD_TO_SPARK_QUEUE(result);                         \
1982     SAFESTGCALL2(void,(W_),GranSimSpark,local,(P_)node);        \
1983     /* context_switch = 1;  not needed any more -- HWL */       \
1984   } else if (do_qp_prof) {                              \
1985     I_ tid = threadId++;                                \
1986     SAFESTGCALL2(void,(I_, P_),QP_Event0,tid,node);     \
1987   }                                                     \
1988   r = 1; /* return code for successful spark -- HWL */  \
1989 }
1990
1991 #endif
1992
1993 #define copyableZh(r,node)                              \
1994   /* copyable not yet implemented!! */
1995
1996 #define noFollowZh(r,node)                              \
1997   /* noFollow not yet implemented!! */
1998
1999 #else  /* !GRAN */
2000
2001 extern I_ required_thread_count;
2002
2003 #ifdef PAR
2004 #define COUNT_SPARK     TSO_GLOBALSPARKS(CurrentTSO)++; sparksCreated++
2005 #else
2006 #define COUNT_SPARK
2007 #endif
2008
2009 /*
2010    Note that we must bump the required thread count NOW, rather
2011    than when the thread is actually created.
2012  */
2013
2014 #define forkZh(r,liveness,node)                         \
2015 {                                                       \
2016   while (PendingSparksTl[REQUIRED_POOL] == PendingSparksLim[REQUIRED_POOL]) \
2017     DO_YIELD((liveness << 1) | 1);                      \
2018   COUNT_SPARK;                                          \
2019   if (SHOULD_SPARK(node)) {                             \
2020     *PendingSparksTl[REQUIRED_POOL]++ = (P_)(node);     \
2021   } else if (DO_QP_PROF) {                              \
2022     I_ tid = threadId++;                                \
2023     SAFESTGCALL2(void,(I_, P_),QP_Event0,tid,node);     \
2024   }                                                     \
2025   required_thread_count++;                              \
2026   context_switch = 1;                                   \
2027   r = 1; /* Should not be necessary */                  \
2028 }
2029
2030 #define parZh(r,node)                                   \
2031 {                                                       \
2032   COUNT_SPARK;                                          \
2033   if (SHOULD_SPARK(node) &&                             \
2034    PendingSparksTl[ADVISORY_POOL] < PendingSparksLim[ADVISORY_POOL]) {  \
2035     *PendingSparksTl[ADVISORY_POOL]++ = (P_)(node);     \
2036   } else {                                              \
2037     sparksIgnored++;                                    \
2038     if (DO_QP_PROF) {                                   \
2039       I_ tid = threadId++;                              \
2040       SAFESTGCALL2(void,(I_, P_),QP_Event0,tid,node);   \
2041     }                                                   \
2042   }                                                     \
2043   r = 1; /* Should not be necessary */                  \
2044 }
2045
2046 #endif  /* GRAN */
2047
2048 #endif  /* CONCURRENT */
2049 \end{code}
2050
2051 The following seq# code should only be used in unoptimized code.
2052 Be warned: it's a potential bug-farm.
2053
2054 First we push two words on the B stack: the current value of RetReg
2055 (which may or may not be live), and a continuation snatched largely out
2056 of thin air (it's a point within this code block).  Then we set RetReg
2057 to the special polymorphic return code for seq, load up Node with the
2058 closure to be evaluated, and we're off.  When the eval returns to the
2059 polymorphic seq return point, the two words are popped off the B stack,
2060 RetReg is restored, and we jump to the continuation, completing the
2061 primop and going on our merry way.
2062
2063 \begin{code}
2064
2065 ED_RO_(vtbl_seq);
2066
2067 #define seqZh(r,liveness,node)              \
2068   ({                                        \
2069     __label__ cont;                         \
2070     /* STK_CHK(liveness,0,2,0,0,0,0); */    \
2071     /* SpB -= BREL(2); */                   \
2072     SpB[BREL(0)] = (W_) RetReg;             \
2073     SpB[BREL(1)] = (W_) &&cont;             \
2074     RetReg = (StgRetAddr) vtbl_seq;         \
2075     Node = node;                            \
2076     ENT_VIA_NODE();                         \
2077     InfoPtr = (D_)(INFO_PTR(Node));         \
2078     JMP_(ENTRY_CODE(InfoPtr));              \
2079     cont:                                   \
2080     r = 1; /* Should be unnecessary */      \
2081   })
2082
2083 \end{code}
2084
2085 %************************************************************************
2086 %*                                                                      *
2087 \subsubsection[StgMacros-foreign-objects]{Foreign Objects}
2088 %*                                                                      *
2089 %************************************************************************
2090
2091 [Based on previous MallocPtr comments -- SOF]
2092
2093 This macro is used to construct a ForeignObj on the heap.
2094
2095 What this does is plug the pointer (which will be in a local
2096 variable) together with its finalising/free routine, into a fresh heap
2097 object and then sets a result (which will be a register) to point
2098 to the fresh heap object.
2099
2100 To accommodate per-object finalisation, augment the macro with a
2101 finalisation routine argument. Nothing spectacular, just plug the
2102 pointer to the routine into the ForeignObj -- SOF 4/96
2103
2104 Question: what's this "SET_ACTIVITY" stuff - should I be doing this
2105 too?  (It's if you want to use the SPAT profiling tools to
2106 characterize program behavior by ``activity'' -- tail-calling,
2107 heap-checking, etc. -- see Ticky.lh.  It is quite specialized.
2108 WDP 95/1)
2109
2110 (Swapped first two arguments to make it come into line with what appears
2111 to be `standard' format, return register then liveness mask. -- SOF 4/96)
2112
2113 \begin{code}
2114 #ifndef PAR
2115
2116 StgInt eqForeignObj PROTO((StgForeignObj p1, StgForeignObj p2));
2117
2118 #define makeForeignObjZh(r, liveness, mptr, finalise)    \
2119 do {                                                     \
2120   P_ result;                                             \
2121                                                          \
2122   HEAP_CHK((W_)liveness, _FHS + ForeignObj_SIZE,0);              \
2123   CC_ALLOC(CCC,_FHS + ForeignObj_SIZE,ForeignObj_K); /* cc prof */   \
2124                                                                    \
2125   result = Hp + 1 - (_FHS + ForeignObj_SIZE);                      \
2126   SET_ForeignObj_HDR(result,ForeignObj_info,CCC,_FHS + ForeignObj_SIZE,0); \
2127   ForeignObj_CLOSURE_DATA(result)      = (P_)mptr;                 \
2128   ForeignObj_CLOSURE_FINALISER(result) = (P_)finalise;             \
2129   ForeignObj_CLOSURE_LINK(result) = StorageMgrInfo.ForeignObjList; \
2130   StorageMgrInfo.ForeignObjList = result;                          \
2131                                                         \
2132                                                         \
2133  /*fprintf(stderr,"DEBUG: ForeignObj(0x%x) = <0x%x, 0x%x, 0x%x, 0x%x>\n",       \
2134       result,                                           \
2135       result[0],result[1],                              \
2136       result[2],result[3]);*/                           \
2137                                                         \
2138   CHECK_ForeignObj_CLOSURE( result );                   \
2139   VALIDATE_ForeignObjList( StorageMgrInfo.ForeignObjList ); \
2140                                                         \
2141   (r) = (P_) result;                                    \
2142 } while (0)
2143
2144 #define writeForeignObjZh(res,datum)    ((PP_) ForeignObj_CLOSURE_DATA(res)) = ((P_)datum)
2145
2146 #else
2147 #define makeForeignObjZh(r, liveness, mptr, finalise)               \
2148 do {                                                                \
2149     fflush(stdout);                                                 \
2150     fprintf(stderr, "makeForeignObj#: no foreign object support.\n");\
2151     EXIT(EXIT_FAILURE);                                             \
2152 } while(0)
2153
2154 #define writeForeignObjZh(res,datum)    \
2155 do {                                                                \
2156     fflush(stdout);                                                 \
2157     fprintf(stderr, "writeForeignObj#: no foreign object support.\n");\
2158     EXIT(EXIT_FAILURE);                                             \
2159 } while(0)
2160
2161 #endif /* !PAR */
2162 \end{code}
2163
2164
2165 End-of-file's multi-slurp protection:
2166 \begin{code}
2167 #endif /* ! STGMACROS_H */
2168 \end{code}