1 /* -----------------------------------------------------------------------------
3 * (c) The GHC Team, 1998-2011
5 * Out-of-line primitive operations
7 * This file contains the implementations of all the primitive
8 * operations ("primops") which are not expanded inline. See
9 * ghc/compiler/prelude/primops.txt.pp for a list of all the primops;
10 * this file contains code for most of those with the attribute
13 * Entry convention: the entry convention for a primop is that all the
14 * args are in Stg registers (R1, R2, etc.). This is to make writing
15 * the primops easier. (see compiler/codeGen/CgCallConv.hs).
17 * Return convention: results from a primop are generally returned
18 * using the ordinary unboxed tuple return convention. The C-- parser
19 * implements the RET_xxxx() macros to perform unboxed-tuple returns
20 * based on the prevailing return convention.
22 * This file is written in a subset of C--, extended with various
23 * features specific to GHC. It is compiled by GHC directly. For the
24 * syntax of .cmm files, see the parser in ghc/compiler/cmm/CmmParse.y.
26 * ---------------------------------------------------------------------------*/
31 import pthread_mutex_lock;
32 import pthread_mutex_unlock;
34 import base_ControlziExceptionziBase_nestedAtomically_closure;
35 import EnterCriticalSection;
36 import LeaveCriticalSection;
37 import ghczmprim_GHCziTypes_False_closure;
38 #if !defined(mingw32_HOST_OS)
42 /*-----------------------------------------------------------------------------
45 Basically just new*Array - the others are all inline macros.
47 The size arg is always passed in R1, and the result returned in R1.
49 The slow entry point is for returning from a heap check, the saved
50 size argument must be re-loaded from the stack.
51 -------------------------------------------------------------------------- */
53 /* for objects that are *less* than the size of a word, make sure we
54 * round up to the nearest word for the size of the array.
59 W_ words, payload_words, n, p;
60 MAYBE_GC(NO_PTRS,stg_newByteArrayzh);
62 payload_words = ROUNDUP_BYTES_TO_WDS(n);
63 words = BYTES_TO_WDS(SIZEOF_StgArrWords) + payload_words;
64 ("ptr" p) = foreign "C" allocate(MyCapability() "ptr",words) [];
65 TICK_ALLOC_PRIM(SIZEOF_StgArrWords,WDS(payload_words),0);
66 SET_HDR(p, stg_ARR_WORDS_info, W_[CCCS]);
67 StgArrWords_bytes(p) = n;
72 #define BA_MASK (BA_ALIGN-1)
74 stg_newPinnedByteArrayzh
76 W_ words, n, bytes, payload_words, p;
78 MAYBE_GC(NO_PTRS,stg_newPinnedByteArrayzh);
81 /* payload_words is what we will tell the profiler we had to allocate */
82 payload_words = ROUNDUP_BYTES_TO_WDS(bytes);
83 /* When we actually allocate memory, we need to allow space for the
85 bytes = bytes + SIZEOF_StgArrWords;
86 /* And we want to align to BA_ALIGN bytes, so we need to allow space
87 to shift up to BA_ALIGN - 1 bytes: */
88 bytes = bytes + BA_ALIGN - 1;
89 /* Now we convert to a number of words: */
90 words = ROUNDUP_BYTES_TO_WDS(bytes);
92 ("ptr" p) = foreign "C" allocatePinned(MyCapability() "ptr", words) [];
93 TICK_ALLOC_PRIM(SIZEOF_StgArrWords,WDS(payload_words),0);
95 /* Now we need to move p forward so that the payload is aligned
97 p = p + ((-p - SIZEOF_StgArrWords) & BA_MASK);
99 SET_HDR(p, stg_ARR_WORDS_info, W_[CCCS]);
100 StgArrWords_bytes(p) = n;
104 stg_newAlignedPinnedByteArrayzh
106 W_ words, n, bytes, payload_words, p, alignment;
108 MAYBE_GC(NO_PTRS,stg_newAlignedPinnedByteArrayzh);
112 /* we always supply at least word-aligned memory, so there's no
113 need to allow extra space for alignment if the requirement is less
114 than a word. This also prevents mischief with alignment == 0. */
115 if (alignment <= SIZEOF_W) { alignment = 1; }
119 /* payload_words is what we will tell the profiler we had to allocate */
120 payload_words = ROUNDUP_BYTES_TO_WDS(bytes);
122 /* When we actually allocate memory, we need to allow space for the
124 bytes = bytes + SIZEOF_StgArrWords;
125 /* And we want to align to <alignment> bytes, so we need to allow space
126 to shift up to <alignment - 1> bytes: */
127 bytes = bytes + alignment - 1;
128 /* Now we convert to a number of words: */
129 words = ROUNDUP_BYTES_TO_WDS(bytes);
131 ("ptr" p) = foreign "C" allocatePinned(MyCapability() "ptr", words) [];
132 TICK_ALLOC_PRIM(SIZEOF_StgArrWords,WDS(payload_words),0);
134 /* Now we need to move p forward so that the payload is aligned
135 to <alignment> bytes. Note that we are assuming that
136 <alignment> is a power of 2, which is technically not guaranteed */
137 p = p + ((-p - SIZEOF_StgArrWords) & (alignment - 1));
139 SET_HDR(p, stg_ARR_WORDS_info, W_[CCCS]);
140 StgArrWords_bytes(p) = n;
146 W_ words, n, init, arr, p, size;
147 /* Args: R1 = words, R2 = initialisation value */
150 MAYBE_GC(R2_PTR,stg_newArrayzh);
152 // the mark area contains one byte for each 2^MUT_ARR_PTRS_CARD_BITS words
153 // in the array, making sure we round up, and then rounding up to a whole
155 size = n + mutArrPtrsCardWords(n);
156 words = BYTES_TO_WDS(SIZEOF_StgMutArrPtrs) + size;
157 ("ptr" arr) = foreign "C" allocate(MyCapability() "ptr",words) [R2];
158 TICK_ALLOC_PRIM(SIZEOF_StgMutArrPtrs, WDS(n), 0);
160 SET_HDR(arr, stg_MUT_ARR_PTRS_DIRTY_info, W_[CCCS]);
161 StgMutArrPtrs_ptrs(arr) = n;
162 StgMutArrPtrs_size(arr) = size;
164 // Initialise all elements of the the array with the value in R2
166 p = arr + SIZEOF_StgMutArrPtrs;
168 if (p < arr + WDS(words)) {
173 // Initialise the mark bits with 0
175 if (p < arr + WDS(size)) {
184 stg_unsafeThawArrayzh
186 // SUBTLETY TO DO WITH THE OLD GEN MUTABLE LIST
188 // A MUT_ARR_PTRS lives on the mutable list, but a MUT_ARR_PTRS_FROZEN
189 // normally doesn't. However, when we freeze a MUT_ARR_PTRS, we leave
190 // it on the mutable list for the GC to remove (removing something from
191 // the mutable list is not easy).
193 // So that we can tell whether a MUT_ARR_PTRS_FROZEN is on the mutable list,
194 // when we freeze it we set the info ptr to be MUT_ARR_PTRS_FROZEN0
195 // to indicate that it is still on the mutable list.
197 // So, when we thaw a MUT_ARR_PTRS_FROZEN, we must cope with two cases:
198 // either it is on a mut_list, or it isn't. We adopt the convention that
199 // the closure type is MUT_ARR_PTRS_FROZEN0 if it is on the mutable list,
200 // and MUT_ARR_PTRS_FROZEN otherwise. In fact it wouldn't matter if
201 // we put it on the mutable list more than once, but it would get scavenged
202 // multiple times during GC, which would be unnecessarily slow.
204 if (StgHeader_info(R1) != stg_MUT_ARR_PTRS_FROZEN0_info) {
205 SET_INFO(R1,stg_MUT_ARR_PTRS_DIRTY_info);
206 recordMutable(R1, R1);
207 // must be done after SET_INFO, because it ASSERTs closure_MUTABLE()
210 SET_INFO(R1,stg_MUT_ARR_PTRS_DIRTY_info);
216 /* -----------------------------------------------------------------------------
218 -------------------------------------------------------------------------- */
223 /* Args: R1 = initialisation value */
225 ALLOC_PRIM( SIZEOF_StgMutVar, R1_PTR, stg_newMutVarzh);
227 mv = Hp - SIZEOF_StgMutVar + WDS(1);
228 SET_HDR(mv,stg_MUT_VAR_DIRTY_info,W_[CCCS]);
229 StgMutVar_var(mv) = R1;
235 /* MutVar# s a -> a -> a -> State# s -> (# State#, Int#, a #) */
243 (h) = foreign "C" cas(mv + SIZEOF_StgHeader + OFFSET_StgMutVar_var,
253 stg_atomicModifyMutVarzh
255 W_ mv, f, z, x, y, r, h;
256 /* Args: R1 :: MutVar#, R2 :: a -> (a,b) */
258 /* If x is the current contents of the MutVar#, then
259 We want to make the new contents point to
263 and the return value is
267 obviously we can share (f x).
269 z = [stg_ap_2 f x] (max (HS + 2) MIN_UPD_SIZE)
270 y = [stg_sel_0 z] (max (HS + 1) MIN_UPD_SIZE)
271 r = [stg_sel_1 z] (max (HS + 1) MIN_UPD_SIZE)
275 #define THUNK_1_SIZE (SIZEOF_StgThunkHeader + WDS(MIN_UPD_SIZE))
276 #define TICK_ALLOC_THUNK_1() TICK_ALLOC_UP_THK(WDS(1),WDS(MIN_UPD_SIZE-1))
278 #define THUNK_1_SIZE (SIZEOF_StgThunkHeader + WDS(1))
279 #define TICK_ALLOC_THUNK_1() TICK_ALLOC_UP_THK(WDS(1),0)
283 #define THUNK_2_SIZE (SIZEOF_StgThunkHeader + WDS(MIN_UPD_SIZE))
284 #define TICK_ALLOC_THUNK_2() TICK_ALLOC_UP_THK(WDS(2),WDS(MIN_UPD_SIZE-2))
286 #define THUNK_2_SIZE (SIZEOF_StgThunkHeader + WDS(2))
287 #define TICK_ALLOC_THUNK_2() TICK_ALLOC_UP_THK(WDS(2),0)
290 #define SIZE (THUNK_2_SIZE + THUNK_1_SIZE + THUNK_1_SIZE)
292 HP_CHK_GEN_TICKY(SIZE, R1_PTR & R2_PTR, stg_atomicModifyMutVarzh);
297 TICK_ALLOC_THUNK_2();
298 CCCS_ALLOC(THUNK_2_SIZE);
299 z = Hp - THUNK_2_SIZE + WDS(1);
300 SET_HDR(z, stg_ap_2_upd_info, W_[CCCS]);
301 LDV_RECORD_CREATE(z);
302 StgThunk_payload(z,0) = f;
304 TICK_ALLOC_THUNK_1();
305 CCCS_ALLOC(THUNK_1_SIZE);
306 y = z - THUNK_1_SIZE;
307 SET_HDR(y, stg_sel_0_upd_info, W_[CCCS]);
308 LDV_RECORD_CREATE(y);
309 StgThunk_payload(y,0) = z;
311 TICK_ALLOC_THUNK_1();
312 CCCS_ALLOC(THUNK_1_SIZE);
313 r = y - THUNK_1_SIZE;
314 SET_HDR(r, stg_sel_1_upd_info, W_[CCCS]);
315 LDV_RECORD_CREATE(r);
316 StgThunk_payload(r,0) = z;
319 x = StgMutVar_var(mv);
320 StgThunk_payload(z,1) = x;
322 (h) = foreign "C" cas(mv + SIZEOF_StgHeader + OFFSET_StgMutVar_var, x, y) [];
323 if (h != x) { goto retry; }
325 StgMutVar_var(mv) = y;
328 if (GET_INFO(mv) == stg_MUT_VAR_CLEAN_info) {
329 foreign "C" dirty_MUT_VAR(BaseReg "ptr", mv "ptr") [];
335 /* -----------------------------------------------------------------------------
336 Weak Pointer Primitives
337 -------------------------------------------------------------------------- */
339 STRING(stg_weak_msg,"New weak pointer at %p\n")
345 R3 = finalizer (or NULL)
350 R3 = stg_NO_FINALIZER_closure;
353 ALLOC_PRIM( SIZEOF_StgWeak, R1_PTR & R2_PTR & R3_PTR, stg_mkWeakzh );
355 w = Hp - SIZEOF_StgWeak + WDS(1);
356 SET_HDR(w, stg_WEAK_info, W_[CCCS]);
358 // We don't care about cfinalizer here.
359 // Should StgWeak_cfinalizer(w) be stg_NO_FINALIZER_closure or
363 StgWeak_value(w) = R2;
364 StgWeak_finalizer(w) = R3;
365 StgWeak_cfinalizer(w) = stg_NO_FINALIZER_closure;
367 ACQUIRE_LOCK(sm_mutex);
368 StgWeak_link(w) = W_[weak_ptr_list];
369 W_[weak_ptr_list] = w;
370 RELEASE_LOCK(sm_mutex);
372 IF_DEBUG(weak, foreign "C" debugBelch(stg_weak_msg,w) []);
377 stg_mkWeakForeignEnvzh
383 R5 = has environment (0 or 1)
386 W_ w, payload_words, words, p;
388 W_ key, val, fptr, ptr, flag, eptr;
397 ALLOC_PRIM( SIZEOF_StgWeak, R1_PTR & R2_PTR, stg_mkWeakForeignEnvzh );
399 w = Hp - SIZEOF_StgWeak + WDS(1);
400 SET_HDR(w, stg_WEAK_info, W_[CCCS]);
403 words = BYTES_TO_WDS(SIZEOF_StgArrWords) + payload_words;
404 ("ptr" p) = foreign "C" allocate(MyCapability() "ptr", words) [];
406 TICK_ALLOC_PRIM(SIZEOF_StgArrWords,WDS(payload_words),0);
407 SET_HDR(p, stg_ARR_WORDS_info, W_[CCCS]);
409 StgArrWords_bytes(p) = WDS(payload_words);
410 StgArrWords_payload(p,0) = fptr;
411 StgArrWords_payload(p,1) = ptr;
412 StgArrWords_payload(p,2) = eptr;
413 StgArrWords_payload(p,3) = flag;
415 // We don't care about the value here.
416 // Should StgWeak_value(w) be stg_NO_FINALIZER_closure or something else?
418 StgWeak_key(w) = key;
419 StgWeak_value(w) = val;
420 StgWeak_finalizer(w) = stg_NO_FINALIZER_closure;
421 StgWeak_cfinalizer(w) = p;
423 ACQUIRE_LOCK(sm_mutex);
424 StgWeak_link(w) = W_[weak_ptr_list];
425 W_[weak_ptr_list] = w;
426 RELEASE_LOCK(sm_mutex);
428 IF_DEBUG(weak, foreign "C" debugBelch(stg_weak_msg,w) []);
442 if (GET_INFO(w) == stg_DEAD_WEAK_info) {
443 RET_NP(0,stg_NO_FINALIZER_closure);
449 // A weak pointer is inherently used, so we do not need to call
450 // LDV_recordDead_FILL_SLOP_DYNAMIC():
451 // LDV_recordDead_FILL_SLOP_DYNAMIC((StgClosure *)w);
452 // or, LDV_recordDead():
453 // LDV_recordDead((StgClosure *)w, sizeofW(StgWeak) - sizeofW(StgProfHeader));
454 // Furthermore, when PROFILING is turned on, dead weak pointers are exactly as
455 // large as weak pointers, so there is no need to fill the slop, either.
456 // See stg_DEAD_WEAK_info in StgMiscClosures.hc.
460 // Todo: maybe use SET_HDR() and remove LDV_recordCreate()?
462 SET_INFO(w,stg_DEAD_WEAK_info);
463 LDV_RECORD_CREATE(w);
465 f = StgWeak_finalizer(w);
466 arr = StgWeak_cfinalizer(w);
468 StgDeadWeak_link(w) = StgWeak_link(w);
470 if (arr != stg_NO_FINALIZER_closure) {
471 foreign "C" runCFinalizer(StgArrWords_payload(arr,0),
472 StgArrWords_payload(arr,1),
473 StgArrWords_payload(arr,2),
474 StgArrWords_payload(arr,3)) [];
477 /* return the finalizer */
478 if (f == stg_NO_FINALIZER_closure) {
479 RET_NP(0,stg_NO_FINALIZER_closure);
491 if (GET_INFO(w) == stg_WEAK_info) {
493 val = StgWeak_value(w);
501 /* -----------------------------------------------------------------------------
502 Floating point operations.
503 -------------------------------------------------------------------------- */
505 stg_decodeFloatzuIntzh
512 STK_CHK_GEN( WDS(2), NO_PTRS, stg_decodeFloatzuIntzh );
514 mp_tmp1 = Sp - WDS(1);
515 mp_tmp_w = Sp - WDS(2);
517 /* arguments: F1 = Float# */
520 /* Perform the operation */
521 foreign "C" __decodeFloat_Int(mp_tmp1 "ptr", mp_tmp_w "ptr", arg) [];
523 /* returns: (Int# (mantissa), Int# (exponent)) */
524 RET_NN(W_[mp_tmp1], W_[mp_tmp_w]);
527 stg_decodeDoublezu2Intzh
536 STK_CHK_GEN( WDS(4), NO_PTRS, stg_decodeDoublezu2Intzh );
538 mp_tmp1 = Sp - WDS(1);
539 mp_tmp2 = Sp - WDS(2);
540 mp_result1 = Sp - WDS(3);
541 mp_result2 = Sp - WDS(4);
543 /* arguments: D1 = Double# */
546 /* Perform the operation */
547 foreign "C" __decodeDouble_2Int(mp_tmp1 "ptr", mp_tmp2 "ptr",
548 mp_result1 "ptr", mp_result2 "ptr",
552 (Int# (mant sign), Word# (mant high), Word# (mant low), Int# (expn)) */
553 RET_NNNN(W_[mp_tmp1], W_[mp_tmp2], W_[mp_result1], W_[mp_result2]);
556 /* -----------------------------------------------------------------------------
557 * Concurrency primitives
558 * -------------------------------------------------------------------------- */
562 /* args: R1 = closure to spark */
564 MAYBE_GC(R1_PTR, stg_forkzh);
570 ("ptr" threadid) = foreign "C" createIOThread( MyCapability() "ptr",
571 RtsFlags_GcFlags_initialStkSize(RtsFlags),
574 /* start blocked if the current thread is blocked */
575 StgTSO_flags(threadid) = %lobits16(
576 TO_W_(StgTSO_flags(threadid)) |
577 TO_W_(StgTSO_flags(CurrentTSO)) & (TSO_BLOCKEX | TSO_INTERRUPTIBLE));
579 foreign "C" scheduleThread(MyCapability() "ptr", threadid "ptr") [];
581 // context switch soon, but not immediately: we don't want every
582 // forkIO to force a context-switch.
583 Capability_context_switch(MyCapability()) = 1 :: CInt;
590 /* args: R1 = cpu, R2 = closure to spark */
592 MAYBE_GC(R2_PTR, stg_forkOnzh);
600 ("ptr" threadid) = foreign "C" createIOThread( MyCapability() "ptr",
601 RtsFlags_GcFlags_initialStkSize(RtsFlags),
604 /* start blocked if the current thread is blocked */
605 StgTSO_flags(threadid) = %lobits16(
606 TO_W_(StgTSO_flags(threadid)) |
607 TO_W_(StgTSO_flags(CurrentTSO)) & (TSO_BLOCKEX | TSO_INTERRUPTIBLE));
609 foreign "C" scheduleThreadOn(MyCapability() "ptr", cpu, threadid "ptr") [];
611 // context switch soon, but not immediately: we don't want every
612 // forkIO to force a context-switch.
613 Capability_context_switch(MyCapability()) = 1 :: CInt;
620 jump stg_yield_noregs;
635 foreign "C" labelThread(R1 "ptr", R2 "ptr") [];
637 jump %ENTRY_CODE(Sp(0));
640 stg_isCurrentThreadBoundzh
644 (r) = foreign "C" isThreadBound(CurrentTSO) [];
650 /* args: R1 :: ThreadId# */
658 what_next = TO_W_(StgTSO_what_next(tso));
659 why_blocked = TO_W_(StgTSO_why_blocked(tso));
660 // Note: these two reads are not atomic, so they might end up
661 // being inconsistent. It doesn't matter, since we
662 // only return one or the other. If we wanted to return the
663 // contents of block_info too, then we'd have to do some synchronisation.
665 if (what_next == ThreadComplete) {
666 ret = 16; // NB. magic, matches up with GHC.Conc.threadStatus
668 if (what_next == ThreadKilled) {
675 cap = TO_W_(Capability_no(StgTSO_cap(tso)));
677 if ((TO_W_(StgTSO_flags(tso)) & TSO_LOCKED) != 0) {
683 RET_NNN(ret,cap,locked);
686 /* -----------------------------------------------------------------------------
688 * -------------------------------------------------------------------------- */
692 // Catch retry frame ------------------------------------------------------------
694 INFO_TABLE_RET(stg_catch_retry_frame, CATCH_RETRY_FRAME,
695 #if defined(PROFILING)
696 W_ unused1, W_ unused2,
698 W_ unused3, P_ unused4, P_ unused5)
700 W_ r, frame, trec, outer;
703 trec = StgTSO_trec(CurrentTSO);
704 outer = StgTRecHeader_enclosing_trec(trec);
705 (r) = foreign "C" stmCommitNestedTransaction(MyCapability() "ptr", trec "ptr") [];
707 /* Succeeded (either first branch or second branch) */
708 StgTSO_trec(CurrentTSO) = outer;
709 Sp = Sp + SIZEOF_StgCatchRetryFrame;
710 jump %ENTRY_CODE(Sp(SP_OFF));
712 /* Did not commit: re-execute */
714 ("ptr" new_trec) = foreign "C" stmStartTransaction(MyCapability() "ptr", outer "ptr") [];
715 StgTSO_trec(CurrentTSO) = new_trec;
716 if (StgCatchRetryFrame_running_alt_code(frame) != 0::I32) {
717 R1 = StgCatchRetryFrame_alt_code(frame);
719 R1 = StgCatchRetryFrame_first_code(frame);
726 // Atomically frame ------------------------------------------------------------
728 INFO_TABLE_RET(stg_atomically_frame, ATOMICALLY_FRAME,
729 #if defined(PROFILING)
730 W_ unused1, W_ unused2,
732 P_ code, P_ next_invariant_to_check, P_ result)
734 W_ frame, trec, valid, next_invariant, q, outer;
737 trec = StgTSO_trec(CurrentTSO);
739 outer = StgTRecHeader_enclosing_trec(trec);
741 if (outer == NO_TREC) {
742 /* First time back at the atomically frame -- pick up invariants */
743 ("ptr" q) = foreign "C" stmGetInvariantsToCheck(MyCapability() "ptr", trec "ptr") [];
744 StgAtomicallyFrame_next_invariant_to_check(frame) = q;
745 StgAtomicallyFrame_result(frame) = result;
748 /* Second/subsequent time back at the atomically frame -- abort the
749 * tx that's checking the invariant and move on to the next one */
750 StgTSO_trec(CurrentTSO) = outer;
751 q = StgAtomicallyFrame_next_invariant_to_check(frame);
752 StgInvariantCheckQueue_my_execution(q) = trec;
753 foreign "C" stmAbortTransaction(MyCapability() "ptr", trec "ptr") [];
754 /* Don't free trec -- it's linked from q and will be stashed in the
755 * invariant if we eventually commit. */
756 q = StgInvariantCheckQueue_next_queue_entry(q);
757 StgAtomicallyFrame_next_invariant_to_check(frame) = q;
761 q = StgAtomicallyFrame_next_invariant_to_check(frame);
763 if (q != END_INVARIANT_CHECK_QUEUE) {
764 /* We can't commit yet: another invariant to check */
765 ("ptr" trec) = foreign "C" stmStartTransaction(MyCapability() "ptr", trec "ptr") [];
766 StgTSO_trec(CurrentTSO) = trec;
768 next_invariant = StgInvariantCheckQueue_invariant(q);
769 R1 = StgAtomicInvariant_code(next_invariant);
774 /* We've got no more invariants to check, try to commit */
775 (valid) = foreign "C" stmCommitTransaction(MyCapability() "ptr", trec "ptr") [];
777 /* Transaction was valid: commit succeeded */
778 StgTSO_trec(CurrentTSO) = NO_TREC;
779 R1 = StgAtomicallyFrame_result(frame);
780 Sp = Sp + SIZEOF_StgAtomicallyFrame;
781 jump %ENTRY_CODE(Sp(SP_OFF));
783 /* Transaction was not valid: try again */
784 ("ptr" trec) = foreign "C" stmStartTransaction(MyCapability() "ptr", NO_TREC "ptr") [];
785 StgTSO_trec(CurrentTSO) = trec;
786 StgAtomicallyFrame_next_invariant_to_check(frame) = END_INVARIANT_CHECK_QUEUE;
787 R1 = StgAtomicallyFrame_code(frame);
793 INFO_TABLE_RET(stg_atomically_waiting_frame, ATOMICALLY_FRAME,
794 #if defined(PROFILING)
795 W_ unused1, W_ unused2,
797 P_ code, P_ next_invariant_to_check, P_ result)
799 W_ frame, trec, valid;
803 /* The TSO is currently waiting: should we stop waiting? */
804 (valid) = foreign "C" stmReWait(MyCapability() "ptr", CurrentTSO "ptr") [];
806 /* Previous attempt is still valid: no point trying again yet */
807 jump stg_block_noregs;
809 /* Previous attempt is no longer valid: try again */
810 ("ptr" trec) = foreign "C" stmStartTransaction(MyCapability() "ptr", NO_TREC "ptr") [];
811 StgTSO_trec(CurrentTSO) = trec;
812 StgHeader_info(frame) = stg_atomically_frame_info;
813 R1 = StgAtomicallyFrame_code(frame);
818 // STM catch frame --------------------------------------------------------------
822 /* Catch frames are very similar to update frames, but when entering
823 * one we just pop the frame off the stack and perform the correct
824 * kind of return to the activation record underneath us on the stack.
827 INFO_TABLE_RET(stg_catch_stm_frame, CATCH_STM_FRAME,
828 #if defined(PROFILING)
829 W_ unused1, W_ unused2,
831 P_ unused3, P_ unused4)
833 W_ r, frame, trec, outer;
835 trec = StgTSO_trec(CurrentTSO);
836 outer = StgTRecHeader_enclosing_trec(trec);
837 (r) = foreign "C" stmCommitNestedTransaction(MyCapability() "ptr", trec "ptr") [];
839 /* Commit succeeded */
840 StgTSO_trec(CurrentTSO) = outer;
841 Sp = Sp + SIZEOF_StgCatchSTMFrame;
846 ("ptr" new_trec) = foreign "C" stmStartTransaction(MyCapability() "ptr", outer "ptr") [];
847 StgTSO_trec(CurrentTSO) = new_trec;
848 R1 = StgCatchSTMFrame_code(frame);
854 // Primop definition ------------------------------------------------------------
862 // stmStartTransaction may allocate
863 MAYBE_GC (R1_PTR, stg_atomicallyzh);
865 /* Args: R1 = m :: STM a */
866 STK_CHK_GEN(SIZEOF_StgAtomicallyFrame + WDS(1), R1_PTR, stg_atomicallyzh);
868 old_trec = StgTSO_trec(CurrentTSO);
870 /* Nested transactions are not allowed; raise an exception */
871 if (old_trec != NO_TREC) {
872 R1 = base_ControlziExceptionziBase_nestedAtomically_closure;
876 /* Set up the atomically frame */
877 Sp = Sp - SIZEOF_StgAtomicallyFrame;
880 SET_HDR(frame,stg_atomically_frame_info, W_[CCCS]);
881 StgAtomicallyFrame_code(frame) = R1;
882 StgAtomicallyFrame_result(frame) = NO_TREC;
883 StgAtomicallyFrame_next_invariant_to_check(frame) = END_INVARIANT_CHECK_QUEUE;
885 /* Start the memory transcation */
886 ("ptr" new_trec) = foreign "C" stmStartTransaction(MyCapability() "ptr", old_trec "ptr") [R1];
887 StgTSO_trec(CurrentTSO) = new_trec;
889 /* Apply R1 to the realworld token */
898 /* Args: R1 :: STM a */
899 /* Args: R2 :: Exception -> STM a */
900 STK_CHK_GEN(SIZEOF_StgCatchSTMFrame + WDS(1), R1_PTR & R2_PTR, stg_catchSTMzh);
902 /* Set up the catch frame */
903 Sp = Sp - SIZEOF_StgCatchSTMFrame;
906 SET_HDR(frame, stg_catch_stm_frame_info, W_[CCCS]);
907 StgCatchSTMFrame_handler(frame) = R2;
908 StgCatchSTMFrame_code(frame) = R1;
910 /* Start a nested transaction to run the body of the try block in */
913 cur_trec = StgTSO_trec(CurrentTSO);
914 ("ptr" new_trec) = foreign "C" stmStartTransaction(MyCapability() "ptr", cur_trec "ptr");
915 StgTSO_trec(CurrentTSO) = new_trec;
917 /* Apply R1 to the realworld token */
928 // stmStartTransaction may allocate
929 MAYBE_GC (R1_PTR & R2_PTR, stg_catchRetryzh);
931 /* Args: R1 :: STM a */
932 /* Args: R2 :: STM a */
933 STK_CHK_GEN(SIZEOF_StgCatchRetryFrame + WDS(1), R1_PTR & R2_PTR, stg_catchRetryzh);
935 /* Start a nested transaction within which to run the first code */
936 trec = StgTSO_trec(CurrentTSO);
937 ("ptr" new_trec) = foreign "C" stmStartTransaction(MyCapability() "ptr", trec "ptr") [R1,R2];
938 StgTSO_trec(CurrentTSO) = new_trec;
940 /* Set up the catch-retry frame */
941 Sp = Sp - SIZEOF_StgCatchRetryFrame;
944 SET_HDR(frame, stg_catch_retry_frame_info, W_[CCCS]);
945 StgCatchRetryFrame_running_alt_code(frame) = 0 :: CInt; // false;
946 StgCatchRetryFrame_first_code(frame) = R1;
947 StgCatchRetryFrame_alt_code(frame) = R2;
949 /* Apply R1 to the realworld token */
962 MAYBE_GC (NO_PTRS, stg_retryzh); // STM operations may allocate
964 // Find the enclosing ATOMICALLY_FRAME or CATCH_RETRY_FRAME
967 (frame_type) = foreign "C" findRetryFrameHelper(MyCapability(), CurrentTSO "ptr") [];
970 trec = StgTSO_trec(CurrentTSO);
971 outer = StgTRecHeader_enclosing_trec(trec);
973 if (frame_type == CATCH_RETRY_FRAME) {
974 // The retry reaches a CATCH_RETRY_FRAME before the atomic frame
975 ASSERT(outer != NO_TREC);
976 // Abort the transaction attempting the current branch
977 foreign "C" stmAbortTransaction(MyCapability() "ptr", trec "ptr") [];
978 foreign "C" stmFreeAbortedTRec(MyCapability() "ptr", trec "ptr") [];
979 if (!StgCatchRetryFrame_running_alt_code(frame) != 0::I32) {
980 // Retry in the first branch: try the alternative
981 ("ptr" trec) = foreign "C" stmStartTransaction(MyCapability() "ptr", outer "ptr") [];
982 StgTSO_trec(CurrentTSO) = trec;
983 StgCatchRetryFrame_running_alt_code(frame) = 1 :: CInt; // true;
984 R1 = StgCatchRetryFrame_alt_code(frame);
987 // Retry in the alternative code: propagate the retry
988 StgTSO_trec(CurrentTSO) = outer;
989 Sp = Sp + SIZEOF_StgCatchRetryFrame;
990 goto retry_pop_stack;
994 // We've reached the ATOMICALLY_FRAME: attempt to wait
995 ASSERT(frame_type == ATOMICALLY_FRAME);
996 if (outer != NO_TREC) {
997 // We called retry while checking invariants, so abort the current
998 // invariant check (merging its TVar accesses into the parents read
999 // set so we'll wait on them)
1000 foreign "C" stmAbortTransaction(MyCapability() "ptr", trec "ptr") [];
1001 foreign "C" stmFreeAbortedTRec(MyCapability() "ptr", trec "ptr") [];
1003 StgTSO_trec(CurrentTSO) = trec;
1004 outer = StgTRecHeader_enclosing_trec(trec);
1006 ASSERT(outer == NO_TREC);
1008 (r) = foreign "C" stmWait(MyCapability() "ptr", CurrentTSO "ptr", trec "ptr") [];
1010 // Transaction was valid: stmWait put us on the TVars' queues, we now block
1011 StgHeader_info(frame) = stg_atomically_waiting_frame_info;
1013 // Fix up the stack in the unregisterised case: the return convention is different.
1014 R3 = trec; // passing to stmWaitUnblock()
1015 jump stg_block_stmwait;
1017 // Transaction was not valid: retry immediately
1018 ("ptr" trec) = foreign "C" stmStartTransaction(MyCapability() "ptr", outer "ptr") [];
1019 StgTSO_trec(CurrentTSO) = trec;
1020 R1 = StgAtomicallyFrame_code(frame);
1031 /* Args: R1 = invariant closure */
1032 MAYBE_GC (R1_PTR, stg_checkzh);
1034 trec = StgTSO_trec(CurrentTSO);
1036 foreign "C" stmAddInvariantToCheck(MyCapability() "ptr",
1040 jump %ENTRY_CODE(Sp(0));
1049 /* Args: R1 = initialisation value */
1051 MAYBE_GC (R1_PTR, stg_newTVarzh);
1053 ("ptr" tv) = foreign "C" stmNewTVar(MyCapability() "ptr", new_value "ptr") [];
1064 /* Args: R1 = TVar closure */
1066 MAYBE_GC (R1_PTR, stg_readTVarzh); // Call to stmReadTVar may allocate
1067 trec = StgTSO_trec(CurrentTSO);
1069 ("ptr" result) = foreign "C" stmReadTVar(MyCapability() "ptr", trec "ptr", tvar "ptr") [];
1079 result = StgTVar_current_value(R1);
1080 if (%INFO_PTR(result) == stg_TREC_HEADER_info) {
1092 /* Args: R1 = TVar closure */
1093 /* R2 = New value */
1095 MAYBE_GC (R1_PTR & R2_PTR, stg_writeTVarzh); // Call to stmWriteTVar may allocate
1096 trec = StgTSO_trec(CurrentTSO);
1099 foreign "C" stmWriteTVar(MyCapability() "ptr", trec "ptr", tvar "ptr", new_value "ptr") [];
1101 jump %ENTRY_CODE(Sp(0));
1105 /* -----------------------------------------------------------------------------
1108 * take & putMVar work as follows. Firstly, an important invariant:
1110 * If the MVar is full, then the blocking queue contains only
1111 * threads blocked on putMVar, and if the MVar is empty then the
1112 * blocking queue contains only threads blocked on takeMVar.
1115 * MVar empty : then add ourselves to the blocking queue
1116 * MVar full : remove the value from the MVar, and
1117 * blocking queue empty : return
1118 * blocking queue non-empty : perform the first blocked putMVar
1119 * from the queue, and wake up the
1120 * thread (MVar is now full again)
1122 * putMVar is just the dual of the above algorithm.
1124 * How do we "perform a putMVar"? Well, we have to fiddle around with
1125 * the stack of the thread waiting to do the putMVar. See
1126 * stg_block_putmvar and stg_block_takemvar in HeapStackCheck.c for
1127 * the stack layout, and the PerformPut and PerformTake macros below.
1129 * It is important that a blocked take or put is woken up with the
1130 * take/put already performed, because otherwise there would be a
1131 * small window of vulnerability where the thread could receive an
1132 * exception and never perform its take or put, and we'd end up with a
1135 * -------------------------------------------------------------------------- */
1139 /* args: R1 = MVar closure */
1141 if (StgMVar_value(R1) == stg_END_TSO_QUEUE_closure) {
1153 ALLOC_PRIM ( SIZEOF_StgMVar, NO_PTRS, stg_newMVarzh );
1155 mvar = Hp - SIZEOF_StgMVar + WDS(1);
1156 SET_HDR(mvar,stg_MVAR_DIRTY_info,W_[CCCS]);
1157 // MVARs start dirty: generation 0 has no mutable list
1158 StgMVar_head(mvar) = stg_END_TSO_QUEUE_closure;
1159 StgMVar_tail(mvar) = stg_END_TSO_QUEUE_closure;
1160 StgMVar_value(mvar) = stg_END_TSO_QUEUE_closure;
1165 #define PerformTake(stack, value) \
1167 sp = StgStack_sp(stack); \
1168 W_[sp + WDS(1)] = value; \
1169 W_[sp + WDS(0)] = stg_gc_unpt_r1_info;
1171 #define PerformPut(stack,lval) \
1173 sp = StgStack_sp(stack) + WDS(3); \
1174 StgStack_sp(stack) = sp; \
1175 lval = W_[sp - WDS(1)];
1179 W_ mvar, val, info, tso, q;
1181 /* args: R1 = MVar closure */
1184 #if defined(THREADED_RTS)
1185 ("ptr" info) = foreign "C" lockClosure(mvar "ptr") [];
1187 info = GET_INFO(mvar);
1190 if (info == stg_MVAR_CLEAN_info) {
1191 foreign "C" dirty_MVAR(BaseReg "ptr", mvar "ptr") [];
1194 /* If the MVar is empty, put ourselves on its blocking queue,
1195 * and wait until we're woken up.
1197 if (StgMVar_value(mvar) == stg_END_TSO_QUEUE_closure) {
1199 // Note [mvar-heap-check] We want to do the heap check in the
1200 // branch here, to avoid the conditional in the common case.
1201 // However, we've already locked the MVar above, so we better
1202 // be careful to unlock it again if the the heap check fails.
1203 // Unfortunately we don't have an easy way to inject any code
1204 // into the heap check generated by the code generator, so we
1205 // have to do it in stg_gc_gen (see HeapStackCheck.cmm).
1206 HP_CHK_GEN_TICKY(SIZEOF_StgMVarTSOQueue, R1_PTR, stg_takeMVarzh);
1208 q = Hp - SIZEOF_StgMVarTSOQueue + WDS(1);
1210 SET_HDR(q, stg_MVAR_TSO_QUEUE_info, CCS_SYSTEM);
1211 StgMVarTSOQueue_link(q) = END_TSO_QUEUE;
1212 StgMVarTSOQueue_tso(q) = CurrentTSO;
1214 if (StgMVar_head(mvar) == stg_END_TSO_QUEUE_closure) {
1215 StgMVar_head(mvar) = q;
1217 StgMVarTSOQueue_link(StgMVar_tail(mvar)) = q;
1218 foreign "C" recordClosureMutated(MyCapability() "ptr",
1219 StgMVar_tail(mvar)) [];
1221 StgTSO__link(CurrentTSO) = q;
1222 StgTSO_block_info(CurrentTSO) = mvar;
1223 StgTSO_why_blocked(CurrentTSO) = BlockedOnMVar::I16;
1224 StgMVar_tail(mvar) = q;
1227 jump stg_block_takemvar;
1230 /* we got the value... */
1231 val = StgMVar_value(mvar);
1233 q = StgMVar_head(mvar);
1235 if (q == stg_END_TSO_QUEUE_closure) {
1236 /* No further putMVars, MVar is now empty */
1237 StgMVar_value(mvar) = stg_END_TSO_QUEUE_closure;
1238 unlockClosure(mvar, stg_MVAR_DIRTY_info);
1241 if (StgHeader_info(q) == stg_IND_info ||
1242 StgHeader_info(q) == stg_MSG_NULL_info) {
1243 q = StgInd_indirectee(q);
1247 // There are putMVar(s) waiting... wake up the first thread on the queue
1249 tso = StgMVarTSOQueue_tso(q);
1250 StgMVar_head(mvar) = StgMVarTSOQueue_link(q);
1251 if (StgMVar_head(mvar) == stg_END_TSO_QUEUE_closure) {
1252 StgMVar_tail(mvar) = stg_END_TSO_QUEUE_closure;
1255 ASSERT(StgTSO_why_blocked(tso) == BlockedOnMVar::I16);
1256 ASSERT(StgTSO_block_info(tso) == mvar);
1258 // actually perform the putMVar for the thread that we just woke up
1260 stack = StgTSO_stackobj(tso);
1261 PerformPut(stack, StgMVar_value(mvar));
1263 // indicate that the MVar operation has now completed.
1264 StgTSO__link(tso) = stg_END_TSO_QUEUE_closure;
1266 // no need to mark the TSO dirty, we have only written END_TSO_QUEUE.
1268 foreign "C" tryWakeupThread(MyCapability() "ptr", tso) [];
1270 unlockClosure(mvar, stg_MVAR_DIRTY_info);
1277 W_ mvar, val, info, tso, q;
1279 /* args: R1 = MVar closure */
1282 #if defined(THREADED_RTS)
1283 ("ptr" info) = foreign "C" lockClosure(mvar "ptr") [];
1285 info = GET_INFO(mvar);
1288 /* If the MVar is empty, put ourselves on its blocking queue,
1289 * and wait until we're woken up.
1291 if (StgMVar_value(mvar) == stg_END_TSO_QUEUE_closure) {
1292 #if defined(THREADED_RTS)
1293 unlockClosure(mvar, info);
1295 /* HACK: we need a pointer to pass back,
1296 * so we abuse NO_FINALIZER_closure
1298 RET_NP(0, stg_NO_FINALIZER_closure);
1301 if (info == stg_MVAR_CLEAN_info) {
1302 foreign "C" dirty_MVAR(BaseReg "ptr", mvar "ptr") [];
1305 /* we got the value... */
1306 val = StgMVar_value(mvar);
1308 q = StgMVar_head(mvar);
1310 if (q == stg_END_TSO_QUEUE_closure) {
1311 /* No further putMVars, MVar is now empty */
1312 StgMVar_value(mvar) = stg_END_TSO_QUEUE_closure;
1313 unlockClosure(mvar, stg_MVAR_DIRTY_info);
1316 if (StgHeader_info(q) == stg_IND_info ||
1317 StgHeader_info(q) == stg_MSG_NULL_info) {
1318 q = StgInd_indirectee(q);
1322 // There are putMVar(s) waiting... wake up the first thread on the queue
1324 tso = StgMVarTSOQueue_tso(q);
1325 StgMVar_head(mvar) = StgMVarTSOQueue_link(q);
1326 if (StgMVar_head(mvar) == stg_END_TSO_QUEUE_closure) {
1327 StgMVar_tail(mvar) = stg_END_TSO_QUEUE_closure;
1330 ASSERT(StgTSO_why_blocked(tso) == BlockedOnMVar::I16);
1331 ASSERT(StgTSO_block_info(tso) == mvar);
1333 // actually perform the putMVar for the thread that we just woke up
1335 stack = StgTSO_stackobj(tso);
1336 PerformPut(stack, StgMVar_value(mvar));
1338 // indicate that the MVar operation has now completed.
1339 StgTSO__link(tso) = stg_END_TSO_QUEUE_closure;
1341 // no need to mark the TSO dirty, we have only written END_TSO_QUEUE.
1343 foreign "C" tryWakeupThread(MyCapability() "ptr", tso) [];
1345 unlockClosure(mvar, stg_MVAR_DIRTY_info);
1352 W_ mvar, val, info, tso, q;
1354 /* args: R1 = MVar, R2 = value */
1358 #if defined(THREADED_RTS)
1359 ("ptr" info) = foreign "C" lockClosure(mvar "ptr") [];
1361 info = GET_INFO(mvar);
1364 if (info == stg_MVAR_CLEAN_info) {
1365 foreign "C" dirty_MVAR(BaseReg "ptr", mvar "ptr");
1368 if (StgMVar_value(mvar) != stg_END_TSO_QUEUE_closure) {
1370 // see Note [mvar-heap-check] above
1371 HP_CHK_GEN_TICKY(SIZEOF_StgMVarTSOQueue, R1_PTR & R2_PTR, stg_putMVarzh);
1373 q = Hp - SIZEOF_StgMVarTSOQueue + WDS(1);
1375 SET_HDR(q, stg_MVAR_TSO_QUEUE_info, CCS_SYSTEM);
1376 StgMVarTSOQueue_link(q) = END_TSO_QUEUE;
1377 StgMVarTSOQueue_tso(q) = CurrentTSO;
1379 if (StgMVar_head(mvar) == stg_END_TSO_QUEUE_closure) {
1380 StgMVar_head(mvar) = q;
1382 StgMVarTSOQueue_link(StgMVar_tail(mvar)) = q;
1383 foreign "C" recordClosureMutated(MyCapability() "ptr",
1384 StgMVar_tail(mvar)) [];
1386 StgTSO__link(CurrentTSO) = q;
1387 StgTSO_block_info(CurrentTSO) = mvar;
1388 StgTSO_why_blocked(CurrentTSO) = BlockedOnMVar::I16;
1389 StgMVar_tail(mvar) = q;
1393 jump stg_block_putmvar;
1396 q = StgMVar_head(mvar);
1398 if (q == stg_END_TSO_QUEUE_closure) {
1399 /* No further takes, the MVar is now full. */
1400 StgMVar_value(mvar) = val;
1401 unlockClosure(mvar, stg_MVAR_DIRTY_info);
1402 jump %ENTRY_CODE(Sp(0));
1404 if (StgHeader_info(q) == stg_IND_info ||
1405 StgHeader_info(q) == stg_MSG_NULL_info) {
1406 q = StgInd_indirectee(q);
1410 // There are takeMVar(s) waiting: wake up the first one
1412 tso = StgMVarTSOQueue_tso(q);
1413 StgMVar_head(mvar) = StgMVarTSOQueue_link(q);
1414 if (StgMVar_head(mvar) == stg_END_TSO_QUEUE_closure) {
1415 StgMVar_tail(mvar) = stg_END_TSO_QUEUE_closure;
1418 ASSERT(StgTSO_why_blocked(tso) == BlockedOnMVar::I16);
1419 ASSERT(StgTSO_block_info(tso) == mvar);
1421 // actually perform the takeMVar
1423 stack = StgTSO_stackobj(tso);
1424 PerformTake(stack, val);
1426 // indicate that the MVar operation has now completed.
1427 StgTSO__link(tso) = stg_END_TSO_QUEUE_closure;
1429 if (TO_W_(StgStack_dirty(stack)) == 0) {
1430 foreign "C" dirty_STACK(MyCapability() "ptr", stack "ptr") [];
1433 foreign "C" tryWakeupThread(MyCapability() "ptr", tso) [];
1435 unlockClosure(mvar, stg_MVAR_DIRTY_info);
1436 jump %ENTRY_CODE(Sp(0));
1442 W_ mvar, val, info, tso, q;
1444 /* args: R1 = MVar, R2 = value */
1448 #if defined(THREADED_RTS)
1449 ("ptr" info) = foreign "C" lockClosure(mvar "ptr") [];
1451 info = GET_INFO(mvar);
1454 if (info == stg_MVAR_CLEAN_info) {
1455 foreign "C" dirty_MVAR(BaseReg "ptr", mvar "ptr");
1458 if (StgMVar_value(mvar) != stg_END_TSO_QUEUE_closure) {
1459 #if defined(THREADED_RTS)
1460 unlockClosure(mvar, info);
1465 q = StgMVar_head(mvar);
1467 if (q == stg_END_TSO_QUEUE_closure) {
1468 /* No further takes, the MVar is now full. */
1469 StgMVar_value(mvar) = val;
1470 unlockClosure(mvar, stg_MVAR_DIRTY_info);
1473 if (StgHeader_info(q) == stg_IND_info ||
1474 StgHeader_info(q) == stg_MSG_NULL_info) {
1475 q = StgInd_indirectee(q);
1479 // There are takeMVar(s) waiting: wake up the first one
1481 tso = StgMVarTSOQueue_tso(q);
1482 StgMVar_head(mvar) = StgMVarTSOQueue_link(q);
1483 if (StgMVar_head(mvar) == stg_END_TSO_QUEUE_closure) {
1484 StgMVar_tail(mvar) = stg_END_TSO_QUEUE_closure;
1487 ASSERT(StgTSO_why_blocked(tso) == BlockedOnMVar::I16);
1488 ASSERT(StgTSO_block_info(tso) == mvar);
1490 // actually perform the takeMVar
1492 stack = StgTSO_stackobj(tso);
1493 PerformTake(stack, val);
1495 // indicate that the MVar operation has now completed.
1496 StgTSO__link(tso) = stg_END_TSO_QUEUE_closure;
1498 if (TO_W_(StgStack_dirty(stack)) == 0) {
1499 foreign "C" dirty_STACK(MyCapability() "ptr", stack "ptr") [];
1502 foreign "C" tryWakeupThread(MyCapability() "ptr", tso) [];
1504 unlockClosure(mvar, stg_MVAR_DIRTY_info);
1509 /* -----------------------------------------------------------------------------
1510 Stable pointer primitives
1511 ------------------------------------------------------------------------- */
1513 stg_makeStableNamezh
1517 ALLOC_PRIM( SIZEOF_StgStableName, R1_PTR, stg_makeStableNamezh );
1519 (index) = foreign "C" lookupStableName(R1 "ptr") [];
1521 /* Is there already a StableName for this heap object?
1522 * stable_ptr_table is a pointer to an array of snEntry structs.
1524 if ( snEntry_sn_obj(W_[stable_ptr_table] + index*SIZEOF_snEntry) == NULL ) {
1525 sn_obj = Hp - SIZEOF_StgStableName + WDS(1);
1526 SET_HDR(sn_obj, stg_STABLE_NAME_info, W_[CCCS]);
1527 StgStableName_sn(sn_obj) = index;
1528 snEntry_sn_obj(W_[stable_ptr_table] + index*SIZEOF_snEntry) = sn_obj;
1530 sn_obj = snEntry_sn_obj(W_[stable_ptr_table] + index*SIZEOF_snEntry);
1541 MAYBE_GC(R1_PTR, stg_makeStablePtrzh);
1542 ("ptr" sp) = foreign "C" getStablePtr(R1 "ptr") [];
1546 stg_deRefStablePtrzh
1548 /* Args: R1 = the stable ptr */
1551 r = snEntry_addr(W_[stable_ptr_table] + sp*SIZEOF_snEntry);
1555 /* -----------------------------------------------------------------------------
1556 Bytecode object primitives
1557 ------------------------------------------------------------------------- */
1567 W_ bco, bitmap_arr, bytes, words;
1571 words = BYTES_TO_WDS(SIZEOF_StgBCO) + BYTE_ARR_WDS(bitmap_arr);
1574 ALLOC_PRIM( bytes, R1_PTR&R2_PTR&R3_PTR&R5_PTR, stg_newBCOzh );
1576 bco = Hp - bytes + WDS(1);
1577 SET_HDR(bco, stg_BCO_info, W_[CCCS]);
1579 StgBCO_instrs(bco) = R1;
1580 StgBCO_literals(bco) = R2;
1581 StgBCO_ptrs(bco) = R3;
1582 StgBCO_arity(bco) = HALF_W_(R4);
1583 StgBCO_size(bco) = HALF_W_(words);
1585 // Copy the arity/bitmap info into the BCO
1589 if (i < BYTE_ARR_WDS(bitmap_arr)) {
1590 StgBCO_bitmap(bco,i) = StgArrWords_payload(bitmap_arr,i);
1601 // R1 = the BCO# for the AP
1605 // This function is *only* used to wrap zero-arity BCOs in an
1606 // updatable wrapper (see ByteCodeLink.lhs). An AP thunk is always
1607 // saturated and always points directly to a FUN or BCO.
1608 ASSERT(%INFO_TYPE(%GET_STD_INFO(R1)) == HALF_W_(BCO) &&
1609 StgBCO_arity(R1) == HALF_W_(0));
1611 HP_CHK_GEN_TICKY(SIZEOF_StgAP, R1_PTR, stg_mkApUpd0zh);
1612 TICK_ALLOC_UP_THK(0, 0);
1613 CCCS_ALLOC(SIZEOF_StgAP);
1615 ap = Hp - SIZEOF_StgAP + WDS(1);
1616 SET_HDR(ap, stg_AP_info, W_[CCCS]);
1618 StgAP_n_args(ap) = HALF_W_(0);
1626 /* args: R1 = closure to analyze */
1627 // TODO: Consider the absence of ptrs or nonptrs as a special case ?
1629 W_ info, ptrs, nptrs, p, ptrs_arr, nptrs_arr;
1630 info = %GET_STD_INFO(UNTAG(R1));
1632 // Some closures have non-standard layout, so we omit those here.
1634 type = TO_W_(%INFO_TYPE(info));
1635 switch [0 .. N_CLOSURE_TYPES] type {
1636 case THUNK_SELECTOR : {
1641 case THUNK, THUNK_1_0, THUNK_0_1, THUNK_2_0, THUNK_1_1,
1642 THUNK_0_2, THUNK_STATIC, AP, PAP, AP_STACK, BCO : {
1648 ptrs = TO_W_(%INFO_PTRS(info));
1649 nptrs = TO_W_(%INFO_NPTRS(info));
1654 W_ ptrs_arr_sz, ptrs_arr_cards, nptrs_arr_sz;
1655 nptrs_arr_sz = SIZEOF_StgArrWords + WDS(nptrs);
1656 ptrs_arr_cards = mutArrPtrsCardWords(ptrs);
1657 ptrs_arr_sz = SIZEOF_StgMutArrPtrs + WDS(ptrs) + WDS(ptrs_arr_cards);
1659 ALLOC_PRIM (ptrs_arr_sz + nptrs_arr_sz, R1_PTR, stg_unpackClosurezh);
1664 ptrs_arr = Hp - nptrs_arr_sz - ptrs_arr_sz + WDS(1);
1665 nptrs_arr = Hp - nptrs_arr_sz + WDS(1);
1667 SET_HDR(ptrs_arr, stg_MUT_ARR_PTRS_FROZEN_info, W_[CCCS]);
1668 StgMutArrPtrs_ptrs(ptrs_arr) = ptrs;
1669 StgMutArrPtrs_size(ptrs_arr) = ptrs + ptrs_arr_cards;
1674 W_[ptrs_arr + SIZEOF_StgMutArrPtrs + WDS(p)] = StgClosure_payload(clos,p);
1678 /* We can leave the card table uninitialised, since the array is
1679 allocated in the nursery. The GC will fill it in if/when the array
1682 SET_HDR(nptrs_arr, stg_ARR_WORDS_info, W_[CCCS]);
1683 StgArrWords_bytes(nptrs_arr) = WDS(nptrs);
1687 W_[BYTE_ARR_CTS(nptrs_arr) + WDS(p)] = StgClosure_payload(clos, p+ptrs);
1691 RET_NPP(info, ptrs_arr, nptrs_arr);
1694 /* -----------------------------------------------------------------------------
1695 Thread I/O blocking primitives
1696 -------------------------------------------------------------------------- */
1698 /* Add a thread to the end of the blocked queue. (C-- version of the C
1699 * macro in Schedule.h).
1701 #define APPEND_TO_BLOCKED_QUEUE(tso) \
1702 ASSERT(StgTSO__link(tso) == END_TSO_QUEUE); \
1703 if (W_[blocked_queue_hd] == END_TSO_QUEUE) { \
1704 W_[blocked_queue_hd] = tso; \
1706 foreign "C" setTSOLink(MyCapability() "ptr", W_[blocked_queue_tl] "ptr", tso) []; \
1708 W_[blocked_queue_tl] = tso;
1714 foreign "C" barf("waitRead# on threaded RTS") never returns;
1717 ASSERT(StgTSO_why_blocked(CurrentTSO) == NotBlocked::I16);
1718 StgTSO_why_blocked(CurrentTSO) = BlockedOnRead::I16;
1719 StgTSO_block_info(CurrentTSO) = R1;
1720 // No locking - we're not going to use this interface in the
1721 // threaded RTS anyway.
1722 APPEND_TO_BLOCKED_QUEUE(CurrentTSO);
1723 jump stg_block_noregs;
1731 foreign "C" barf("waitWrite# on threaded RTS") never returns;
1734 ASSERT(StgTSO_why_blocked(CurrentTSO) == NotBlocked::I16);
1735 StgTSO_why_blocked(CurrentTSO) = BlockedOnWrite::I16;
1736 StgTSO_block_info(CurrentTSO) = R1;
1737 // No locking - we're not going to use this interface in the
1738 // threaded RTS anyway.
1739 APPEND_TO_BLOCKED_QUEUE(CurrentTSO);
1740 jump stg_block_noregs;
1745 STRING(stg_delayzh_malloc_str, "stg_delayzh")
1748 #ifdef mingw32_HOST_OS
1756 foreign "C" barf("delay# on threaded RTS") never returns;
1759 /* args: R1 (microsecond delay amount) */
1760 ASSERT(StgTSO_why_blocked(CurrentTSO) == NotBlocked::I16);
1761 StgTSO_why_blocked(CurrentTSO) = BlockedOnDelay::I16;
1763 #ifdef mingw32_HOST_OS
1765 /* could probably allocate this on the heap instead */
1766 ("ptr" ares) = foreign "C" stgMallocBytes(SIZEOF_StgAsyncIOResult,
1767 stg_delayzh_malloc_str);
1768 (reqID) = foreign "C" addDelayRequest(R1);
1769 StgAsyncIOResult_reqID(ares) = reqID;
1770 StgAsyncIOResult_len(ares) = 0;
1771 StgAsyncIOResult_errCode(ares) = 0;
1772 StgTSO_block_info(CurrentTSO) = ares;
1774 /* Having all async-blocked threads reside on the blocked_queue
1775 * simplifies matters, so change the status to OnDoProc put the
1776 * delayed thread on the blocked_queue.
1778 StgTSO_why_blocked(CurrentTSO) = BlockedOnDoProc::I16;
1779 APPEND_TO_BLOCKED_QUEUE(CurrentTSO);
1780 jump stg_block_async_void;
1786 (time) = foreign "C" getourtimeofday() [R1];
1787 divisor = TO_W_(RtsFlags_MiscFlags_tickInterval(RtsFlags));
1791 divisor = divisor * 1000;
1792 target = ((R1 + divisor - 1) / divisor) /* divide rounding up */
1793 + time + 1; /* Add 1 as getourtimeofday rounds down */
1794 StgTSO_block_info(CurrentTSO) = target;
1796 /* Insert the new thread in the sleeping queue. */
1798 t = W_[sleeping_queue];
1800 if (t != END_TSO_QUEUE && StgTSO_block_info(t) < target) {
1802 t = StgTSO__link(t);
1806 StgTSO__link(CurrentTSO) = t;
1808 W_[sleeping_queue] = CurrentTSO;
1810 foreign "C" setTSOLink(MyCapability() "ptr", prev "ptr", CurrentTSO) [];
1812 jump stg_block_noregs;
1814 #endif /* !THREADED_RTS */
1818 #ifdef mingw32_HOST_OS
1819 STRING(stg_asyncReadzh_malloc_str, "stg_asyncReadzh")
1826 foreign "C" barf("asyncRead# on threaded RTS") never returns;
1829 /* args: R1 = fd, R2 = isSock, R3 = len, R4 = buf */
1830 ASSERT(StgTSO_why_blocked(CurrentTSO) == NotBlocked::I16);
1831 StgTSO_why_blocked(CurrentTSO) = BlockedOnRead::I16;
1833 /* could probably allocate this on the heap instead */
1834 ("ptr" ares) = foreign "C" stgMallocBytes(SIZEOF_StgAsyncIOResult,
1835 stg_asyncReadzh_malloc_str)
1837 (reqID) = foreign "C" addIORequest(R1, 0/*FALSE*/,R2,R3,R4 "ptr") [];
1838 StgAsyncIOResult_reqID(ares) = reqID;
1839 StgAsyncIOResult_len(ares) = 0;
1840 StgAsyncIOResult_errCode(ares) = 0;
1841 StgTSO_block_info(CurrentTSO) = ares;
1842 APPEND_TO_BLOCKED_QUEUE(CurrentTSO);
1843 jump stg_block_async;
1847 STRING(stg_asyncWritezh_malloc_str, "stg_asyncWritezh")
1854 foreign "C" barf("asyncWrite# on threaded RTS") never returns;
1857 /* args: R1 = fd, R2 = isSock, R3 = len, R4 = buf */
1858 ASSERT(StgTSO_why_blocked(CurrentTSO) == NotBlocked::I16);
1859 StgTSO_why_blocked(CurrentTSO) = BlockedOnWrite::I16;
1861 ("ptr" ares) = foreign "C" stgMallocBytes(SIZEOF_StgAsyncIOResult,
1862 stg_asyncWritezh_malloc_str)
1864 (reqID) = foreign "C" addIORequest(R1, 1/*TRUE*/,R2,R3,R4 "ptr") [];
1866 StgAsyncIOResult_reqID(ares) = reqID;
1867 StgAsyncIOResult_len(ares) = 0;
1868 StgAsyncIOResult_errCode(ares) = 0;
1869 StgTSO_block_info(CurrentTSO) = ares;
1870 APPEND_TO_BLOCKED_QUEUE(CurrentTSO);
1871 jump stg_block_async;
1875 STRING(stg_asyncDoProczh_malloc_str, "stg_asyncDoProczh")
1882 foreign "C" barf("asyncDoProc# on threaded RTS") never returns;
1885 /* args: R1 = proc, R2 = param */
1886 ASSERT(StgTSO_why_blocked(CurrentTSO) == NotBlocked::I16);
1887 StgTSO_why_blocked(CurrentTSO) = BlockedOnDoProc::I16;
1889 /* could probably allocate this on the heap instead */
1890 ("ptr" ares) = foreign "C" stgMallocBytes(SIZEOF_StgAsyncIOResult,
1891 stg_asyncDoProczh_malloc_str)
1893 (reqID) = foreign "C" addDoProcRequest(R1 "ptr",R2 "ptr") [];
1894 StgAsyncIOResult_reqID(ares) = reqID;
1895 StgAsyncIOResult_len(ares) = 0;
1896 StgAsyncIOResult_errCode(ares) = 0;
1897 StgTSO_block_info(CurrentTSO) = ares;
1898 APPEND_TO_BLOCKED_QUEUE(CurrentTSO);
1899 jump stg_block_async;
1904 /* -----------------------------------------------------------------------------
1907 * noDuplicate# tries to ensure that none of the thunks under
1908 * evaluation by the current thread are also under evaluation by
1909 * another thread. It relies on *both* threads doing noDuplicate#;
1910 * the second one will get blocked if they are duplicating some work.
1912 * The idea is that noDuplicate# is used within unsafePerformIO to
1913 * ensure that the IO operation is performed at most once.
1914 * noDuplicate# calls threadPaused which acquires an exclusive lock on
1915 * all the thunks currently under evaluation by the current thread.
1917 * Consider the following scenario. There is a thunk A, whose
1918 * evaluation requires evaluating thunk B, where thunk B is an
1919 * unsafePerformIO. Two threads, 1 and 2, bother enter A. Thread 2
1920 * is pre-empted before it enters B, and claims A by blackholing it
1921 * (in threadPaused). Thread 1 now enters B, and calls noDuplicate#.
1924 * +-----------+ +---------------+
1925 * | -------+-----> A <-------+------- |
1926 * | update | BLACKHOLE | marked_update |
1927 * +-----------+ +---------------+
1930 * | | +---------------+
1933 * | update | BLACKHOLE
1936 * At this point: A is a blackhole, owned by thread 2. noDuplicate#
1937 * calls threadPaused, which walks up the stack and
1938 * - claims B on behalf of thread 1
1939 * - then it reaches the update frame for A, which it sees is already
1940 * a BLACKHOLE and is therefore owned by another thread. Since
1941 * thread 1 is duplicating work, the computation up to the update
1942 * frame for A is suspended, including thunk B.
1943 * - thunk B, which is an unsafePerformIO, has now been reverted to
1944 * an AP_STACK which could be duplicated - BAD!
1945 * - The solution is as follows: before calling threadPaused, we
1946 * leave a frame on the stack (stg_noDuplicate_info) that will call
1947 * noDuplicate# again if the current computation is suspended and
1950 * See the test program in concurrent/prog003 for a way to demonstrate
1951 * this. It needs to be run with +RTS -N3 or greater, and the bug
1952 * only manifests occasionally (once very 10 runs or so).
1953 * -------------------------------------------------------------------------- */
1955 INFO_TABLE_RET(stg_noDuplicate, RET_SMALL)
1958 jump stg_noDuplicatezh;
1963 STK_CHK_GEN( WDS(1), NO_PTRS, stg_noDuplicatezh );
1964 // leave noDuplicate frame in case the current
1965 // computation is suspended and restarted (see above).
1967 Sp(0) = stg_noDuplicate_info;
1969 SAVE_THREAD_STATE();
1970 ASSERT(StgTSO_what_next(CurrentTSO) == ThreadRunGHC::I16);
1971 foreign "C" threadPaused (MyCapability() "ptr", CurrentTSO "ptr") [];
1973 if (StgTSO_what_next(CurrentTSO) == ThreadKilled::I16) {
1974 jump stg_threadFinished;
1976 LOAD_THREAD_STATE();
1977 ASSERT(StgTSO_what_next(CurrentTSO) == ThreadRunGHC::I16);
1978 // remove the stg_noDuplicate frame if it is still there.
1979 if (Sp(0) == stg_noDuplicate_info) {
1982 jump %ENTRY_CODE(Sp(0));
1986 /* -----------------------------------------------------------------------------
1988 -------------------------------------------------------------------------- */
1992 W_ ap_stack, offset, val, ok;
1994 /* args: R1 = AP_STACK, R2 = offset */
1998 if (%INFO_PTR(ap_stack) == stg_AP_STACK_info) {
2000 val = StgAP_STACK_payload(ap_stack,offset);
2008 // Write the cost center stack of the first argument on stderr; return
2009 // the second. Possibly only makes sense for already evaluated
2016 ccs = StgHeader_ccs(UNTAG(R1));
2017 foreign "C" fprintCCS_stderr(ccs "ptr") [R2];
2028 #ifndef THREADED_RTS
2029 RET_NP(0,ghczmprim_GHCziTypes_False_closure);
2031 (spark) = foreign "C" findSpark(MyCapability());
2035 RET_NP(0,ghczmprim_GHCziTypes_False_closure);
2044 (n) = foreign "C" dequeElements(Capability_sparks(MyCapability()));
2056 #if defined(TRACING) || defined(DEBUG)
2058 foreign "C" traceUserMsg(MyCapability() "ptr", msg "ptr") [];
2060 #elif defined(DTRACE)
2064 // We should go through the macro HASKELLEVENT_USER_MSG_ENABLED from
2065 // RtsProbes.h, but that header file includes unistd.h, which doesn't
2067 #if !defined(solaris2_TARGET_OS)
2068 (enabled) = foreign "C" __dtrace_isenabled$HaskellEvent$user__msg$v1() [];
2070 // Solaris' DTrace can't handle the
2071 // __dtrace_isenabled$HaskellEvent$user__msg$v1
2072 // call above. This call is just for testing whether the user__msg
2073 // probe is enabled, and is here for just performance optimization.
2074 // Since preparation for the probe is not that complex I disable usage of
2075 // this test above for Solaris and enable the probe usage manually
2076 // here. Please note that this does not mean that the probe will be
2077 // used during the runtime! You still need to enable it by consumption
2078 // in your dtrace script as you do with any other probe.
2082 foreign "C" dtraceUserMsgWrapper(MyCapability() "ptr", msg "ptr") [];
2086 jump %ENTRY_CODE(Sp(0));