RET_P(p);
}
+#define BA_ALIGN 16
+#define BA_MASK (BA_ALIGN-1)
+
newPinnedByteArrayzh_fast
{
- W_ words, payload_words, n, p;
+ W_ words, bytes, payload_words, p;
MAYBE_GC(NO_PTRS,newPinnedByteArrayzh_fast);
- n = R1;
- payload_words = ROUNDUP_BYTES_TO_WDS(n);
+ bytes = R1;
+ /* payload_words is what we will tell the profiler we had to allocate */
+ payload_words = ROUNDUP_BYTES_TO_WDS(bytes);
+ /* When we actually allocate memory, we need to allow space for the
+ header: */
+ bytes = bytes + SIZEOF_StgArrWords;
+ /* And we want to align to BA_ALIGN bytes, so we need to allow space
+ to shift up to BA_ALIGN - 1 bytes: */
+ bytes = bytes + BA_ALIGN - 1;
+ /* Now we convert to a number of words: */
+ words = ROUNDUP_BYTES_TO_WDS(bytes);
- // We want an 8-byte aligned array. allocatePinned() gives us
- // 8-byte aligned memory by default, but we want to align the
- // *goods* inside the ArrWords object, so we have to check the
- // size of the ArrWords header and adjust our size accordingly.
- words = BYTES_TO_WDS(SIZEOF_StgArrWords) + payload_words;
- if ((SIZEOF_StgArrWords & 7) != 0) {
- words = words + 1;
- }
+ ("ptr" p) = foreign "C" allocatePinned(words) [];
+ TICK_ALLOC_PRIM(SIZEOF_StgArrWords,WDS(payload_words),0);
+
+ /* Now we need to move p forward so that the payload is aligned
+ to BA_ALIGN bytes: */
+ p = p + ((-p - SIZEOF_StgArrWords) & BA_MASK);
+
+ SET_HDR(p, stg_ARR_WORDS_info, W_[CCCS]);
+ StgArrWords_words(p) = payload_words;
+ RET_P(p);
+}
+
+newAlignedPinnedByteArrayzh_fast
+{
+ W_ words, bytes, payload_words, p, alignment;
+
+ MAYBE_GC(NO_PTRS,newAlignedPinnedByteArrayzh_fast);
+ bytes = R1;
+ alignment = R2;
+
+ /* payload_words is what we will tell the profiler we had to allocate */
+ payload_words = ROUNDUP_BYTES_TO_WDS(bytes);
+
+ /* When we actually allocate memory, we need to allow space for the
+ header: */
+ bytes = bytes + SIZEOF_StgArrWords;
+ /* And we want to align to <alignment> bytes, so we need to allow space
+ to shift up to <alignment - 1> bytes: */
+ bytes = bytes + alignment - 1;
+ /* Now we convert to a number of words: */
+ words = ROUNDUP_BYTES_TO_WDS(bytes);
("ptr" p) = foreign "C" allocatePinned(words) [];
TICK_ALLOC_PRIM(SIZEOF_StgArrWords,WDS(payload_words),0);
- // Again, if the ArrWords header isn't a multiple of 8 bytes, we
- // have to push the object forward one word so that the goods
- // fall on an 8-byte boundary.
- if ((SIZEOF_StgArrWords & 7) != 0) {
- p = p + WDS(1);
- }
+ /* Now we need to move p forward so that the payload is aligned
+ to <alignment> bytes. Note that we are assuming that
+ <alignment> is a power of 2, which is technically not guaranteed */
+ p = p + ((-p - SIZEOF_StgArrWords) & (alignment - 1));
SET_HDR(p, stg_ARR_WORDS_info, W_[CCCS]);
StgArrWords_words(p) = payload_words;
w = Hp - SIZEOF_StgWeak + WDS(1);
SET_HDR(w, stg_WEAK_info, W_[CCCS]);
- StgWeak_key(w) = R1;
- StgWeak_value(w) = R2;
- StgWeak_finalizer(w) = R3;
+ // We don't care about cfinalizer here.
+ // Should StgWeak_cfinalizer(w) be stg_NO_FINALIZER_closure or
+ // something else?
+
+ StgWeak_key(w) = R1;
+ StgWeak_value(w) = R2;
+ StgWeak_finalizer(w) = R3;
+ StgWeak_cfinalizer(w) = stg_NO_FINALIZER_closure;
StgWeak_link(w) = W_[weak_ptr_list];
W_[weak_ptr_list] = w;
RET_P(w);
}
+mkWeakForeignEnvzh_fast
+{
+ /* R1 = key
+ R2 = value
+ R3 = finalizer
+ R4 = pointer
+ R5 = has environment (0 or 1)
+ R6 = environment
+ */
+ W_ w, payload_words, words, p;
+
+ W_ key, val, fptr, ptr, flag, eptr;
+
+ key = R1;
+ val = R2;
+ fptr = R3;
+ ptr = R4;
+ flag = R5;
+ eptr = R6;
+
+ ALLOC_PRIM( SIZEOF_StgWeak, R1_PTR & R2_PTR, mkWeakForeignEnvzh_fast );
+
+ w = Hp - SIZEOF_StgWeak + WDS(1);
+ SET_HDR(w, stg_WEAK_info, W_[CCCS]);
+
+ payload_words = 4;
+ words = BYTES_TO_WDS(SIZEOF_StgArrWords) + payload_words;
+ ("ptr" p) = foreign "C" allocateLocal(MyCapability() "ptr", words) [];
+
+ TICK_ALLOC_PRIM(SIZEOF_StgArrWords,WDS(payload_words),0);
+ SET_HDR(p, stg_ARR_WORDS_info, W_[CCCS]);
+
+ StgArrWords_words(p) = payload_words;
+ StgArrWords_payload(p,0) = fptr;
+ StgArrWords_payload(p,1) = ptr;
+ StgArrWords_payload(p,2) = eptr;
+ StgArrWords_payload(p,3) = flag;
+
+ // We don't care about the value here.
+ // Should StgWeak_value(w) be stg_NO_FINALIZER_closure or something else?
+
+ StgWeak_key(w) = key;
+ StgWeak_value(w) = val;
+ StgWeak_finalizer(w) = stg_NO_FINALIZER_closure;
+ StgWeak_cfinalizer(w) = p;
+
+ StgWeak_link(w) = W_[weak_ptr_list];
+ W_[weak_ptr_list] = w;
+
+ IF_DEBUG(weak, foreign "C" debugBelch(stg_weak_msg,w) []);
+
+ RET_P(w);
+}
finalizzeWeakzh_fast
{
/* R1 = weak ptr
*/
- W_ w, f;
+ W_ w, f, arr;
w = R1;
SET_INFO(w,stg_DEAD_WEAK_info);
LDV_RECORD_CREATE(w);
- f = StgWeak_finalizer(w);
+ f = StgWeak_finalizer(w);
+ arr = StgWeak_cfinalizer(w);
+
StgDeadWeak_link(w) = StgWeak_link(w);
+ if (arr != stg_NO_FINALIZER_closure) {
+ foreign "C" runCFinalizer(StgArrWords_payload(arr,0),
+ StgArrWords_payload(arr,1),
+ StgArrWords_payload(arr,2),
+ StgArrWords_payload(arr,3)) [];
+ }
+
/* return the finalizer */
if (f == stg_NO_FINALIZER_closure) {
RET_NP(0,stg_NO_FINALIZER_closure);
RET_NP(s,p);
}
-
-
#endif /* SUPPORT_LONG_LONGS */
-/* ToDo: this is shockingly inefficient */
-
-#ifndef THREADED_RTS
-section "bss" {
- mp_tmp1:
- bits8 [SIZEOF_MP_INT];
-}
-
-section "bss" {
- mp_tmp2:
- bits8 [SIZEOF_MP_INT];
-}
-
-section "bss" {
- mp_result1:
- bits8 [SIZEOF_MP_INT];
-}
-
-section "bss" {
- mp_result2:
- bits8 [SIZEOF_MP_INT];
-}
-#endif
-
-#ifdef THREADED_RTS
-#define FETCH_MP_TEMP(X) \
-W_ X; \
-X = BaseReg + (OFFSET_StgRegTable_r ## X);
-#else
-#define FETCH_MP_TEMP(X) /* Nothing */
-#endif
-
#define GMP_TAKE2_RET1(name,mp_fun) \
name \
{ \
CInt s1, s2; \
W_ d1, d2; \
- FETCH_MP_TEMP(mp_tmp1); \
- FETCH_MP_TEMP(mp_tmp2); \
- FETCH_MP_TEMP(mp_result1) \
- FETCH_MP_TEMP(mp_result2); \
+ W_ mp_tmp1; \
+ W_ mp_tmp2; \
+ W_ mp_result1; \
+ W_ mp_result2; \
\
/* call doYouWantToGC() */ \
MAYBE_GC(R2_PTR & R4_PTR, name); \
\
+ STK_CHK_GEN( 4 * SIZEOF_MP_INT, R2_PTR & R4_PTR, name ); \
+ \
s1 = W_TO_INT(R1); \
d1 = R2; \
s2 = W_TO_INT(R3); \
d2 = R4; \
\
+ mp_tmp1 = Sp - 1 * SIZEOF_MP_INT; \
+ mp_tmp2 = Sp - 2 * SIZEOF_MP_INT; \
+ mp_result1 = Sp - 3 * SIZEOF_MP_INT; \
+ mp_result2 = Sp - 4 * SIZEOF_MP_INT; \
MP_INT__mp_alloc(mp_tmp1) = W_TO_INT(StgArrWords_words(d1)); \
MP_INT__mp_size(mp_tmp1) = (s1); \
MP_INT__mp_d(mp_tmp1) = BYTE_ARR_CTS(d1); \
{ \
CInt s1; \
W_ d1; \
- FETCH_MP_TEMP(mp_tmp1); \
- FETCH_MP_TEMP(mp_result1) \
+ W_ mp_tmp1; \
+ W_ mp_result1; \
\
/* call doYouWantToGC() */ \
MAYBE_GC(R2_PTR, name); \
\
+ STK_CHK_GEN( 2 * SIZEOF_MP_INT, R2_PTR, name ); \
+ \
d1 = R2; \
s1 = W_TO_INT(R1); \
\
+ mp_tmp1 = Sp - 1 * SIZEOF_MP_INT; \
+ mp_result1 = Sp - 2 * SIZEOF_MP_INT; \
MP_INT__mp_alloc(mp_tmp1) = W_TO_INT(StgArrWords_words(d1)); \
MP_INT__mp_size(mp_tmp1) = (s1); \
MP_INT__mp_d(mp_tmp1) = BYTE_ARR_CTS(d1); \
{ \
CInt s1, s2; \
W_ d1, d2; \
- FETCH_MP_TEMP(mp_tmp1); \
- FETCH_MP_TEMP(mp_tmp2); \
- FETCH_MP_TEMP(mp_result1) \
- FETCH_MP_TEMP(mp_result2) \
+ W_ mp_tmp1; \
+ W_ mp_tmp2; \
+ W_ mp_result1; \
+ W_ mp_result2; \
\
/* call doYouWantToGC() */ \
MAYBE_GC(R2_PTR & R4_PTR, name); \
\
+ STK_CHK_GEN( 4 * SIZEOF_MP_INT, R2_PTR & R4_PTR, name ); \
+ \
s1 = W_TO_INT(R1); \
d1 = R2; \
s2 = W_TO_INT(R3); \
d2 = R4; \
\
+ mp_tmp1 = Sp - 1 * SIZEOF_MP_INT; \
+ mp_tmp2 = Sp - 2 * SIZEOF_MP_INT; \
+ mp_result1 = Sp - 3 * SIZEOF_MP_INT; \
+ mp_result2 = Sp - 4 * SIZEOF_MP_INT; \
MP_INT__mp_alloc(mp_tmp1) = W_TO_INT(StgArrWords_words(d1)); \
MP_INT__mp_size(mp_tmp1) = (s1); \
MP_INT__mp_d(mp_tmp1) = BYTE_ARR_CTS(d1); \
GMP_TAKE2_RET2(quotRemIntegerzh_fast, __gmpz_tdiv_qr)
GMP_TAKE2_RET2(divModIntegerzh_fast, __gmpz_fdiv_qr)
-#ifndef THREADED_RTS
-section "bss" {
- mp_tmp_w: W_; // NB. mp_tmp_w is really an here mp_limb_t
-}
-#endif
-
gcdIntzh_fast
{
/* R1 = the first Int#; R2 = the second Int# */
W_ r;
- FETCH_MP_TEMP(mp_tmp_w);
+ W_ mp_tmp_w;
+
+ STK_CHK_GEN( 1 * SIZEOF_MP_INT, NO_PTRS, gcdIntzh_fast );
+
+ mp_tmp_w = Sp - 1 * SIZEOF_MP_INT;
W_[mp_tmp_w] = R1;
(r) = foreign "C" __gmpn_gcd_1(mp_tmp_w "ptr", 1, R2) [];
jump %ENTRY_CODE(Sp(0));
}
-decodeFloatzh_fast
-{
- W_ p;
- F_ arg;
- FETCH_MP_TEMP(mp_tmp1);
- FETCH_MP_TEMP(mp_tmp_w);
-
- /* arguments: F1 = Float# */
- arg = F1;
-
- ALLOC_PRIM( SIZEOF_StgArrWords + WDS(1), NO_PTRS, decodeFloatzh_fast );
-
- /* Be prepared to tell Lennart-coded __decodeFloat
- where mantissa._mp_d can be put (it does not care about the rest) */
- p = Hp - SIZEOF_StgArrWords;
- SET_HDR(p,stg_ARR_WORDS_info,W_[CCCS]);
- StgArrWords_words(p) = 1;
- MP_INT__mp_d(mp_tmp1) = BYTE_ARR_CTS(p);
-
- /* Perform the operation */
- foreign "C" __decodeFloat(mp_tmp1 "ptr",mp_tmp_w "ptr" ,arg) [];
-
- /* returns: (Int# (expn), Int#, ByteArray#) */
- RET_NNP(W_[mp_tmp_w], TO_W_(MP_INT__mp_size(mp_tmp1)), p);
-}
-
decodeFloatzuIntzh_fast
{
W_ p;
F_ arg;
- FETCH_MP_TEMP(mp_tmp1);
- FETCH_MP_TEMP(mp_tmp_w);
+ W_ mp_tmp1;
+ W_ mp_tmp_w;
+
+ STK_CHK_GEN( WDS(2), NO_PTRS, decodeFloatzuIntzh_fast );
+
+ mp_tmp1 = Sp - WDS(1);
+ mp_tmp_w = Sp - WDS(2);
/* arguments: F1 = Float# */
arg = F1;
{
D_ arg;
W_ p;
- FETCH_MP_TEMP(mp_tmp1);
- FETCH_MP_TEMP(mp_tmp_w);
+ W_ mp_tmp1;
+ W_ mp_tmp_w;
+
+ STK_CHK_GEN( 2 * SIZEOF_MP_INT, NO_PTRS, decodeDoublezh_fast );
+
+ mp_tmp1 = Sp - 1 * SIZEOF_MP_INT;
+ mp_tmp_w = Sp - 2 * SIZEOF_MP_INT;
/* arguments: D1 = Double# */
arg = D1;
{
D_ arg;
W_ p;
- FETCH_MP_TEMP(mp_tmp1);
- FETCH_MP_TEMP(mp_tmp2);
- FETCH_MP_TEMP(mp_result1);
- FETCH_MP_TEMP(mp_result2);
+ W_ mp_tmp1;
+ W_ mp_tmp2;
+ W_ mp_result1;
+ W_ mp_result2;
+
+ STK_CHK_GEN( WDS(4), NO_PTRS, decodeDoublezu2Intzh_fast );
+
+ mp_tmp1 = Sp - WDS(1);
+ mp_tmp2 = Sp - WDS(2);
+ mp_result1 = Sp - WDS(3);
+ mp_result2 = Sp - WDS(4);
/* arguments: D1 = Double# */
arg = D1;
foreign "C" scheduleThread(MyCapability() "ptr", threadid "ptr") [];
- // switch at the earliest opportunity
+ // context switch soon, but not immediately: we don't want every
+ // forkIO to force a context-switch.
Capability_context_switch(MyCapability()) = 1 :: CInt;
RET_P(threadid);
foreign "C" scheduleThreadOn(MyCapability() "ptr", cpu, threadid "ptr") [];
- // switch at the earliest opportunity
+ // context switch soon, but not immediately: we don't want every
+ // forkIO to force a context-switch.
Capability_context_switch(MyCapability()) = 1 :: CInt;
RET_P(threadid);
#if defined(PROFILING)
W_ unused1, W_ unused2,
#endif
- W_ unused3, "ptr" W_ unused4, "ptr" W_ unused5)
+ W_ unused3, P_ unused4, P_ unused5)
{
W_ r, frame, trec, outer;
#if defined(PROFILING)
W_ unused1, W_ unused2,
#endif
- "ptr" W_ unused3, "ptr" W_ unused4)
+ P_ unused3, P_ unused4)
{
W_ frame, trec, valid, next_invariant, q, outer;
#if defined(PROFILING)
W_ unused1, W_ unused2,
#endif
- "ptr" W_ unused3, "ptr" W_ unused4)
+ P_ unused3, P_ unused4)
{
W_ frame, trec, valid;
#if defined(PROFILING)
W_ unused1, W_ unused2,
#endif
- "ptr" W_ unused3, "ptr" W_ unused4)
+ P_ unused3, P_ unused4)
{
W_ r, frame, trec, outer;
frame = Sp;
CurrentTSO) [];
}
StgTSO__link(CurrentTSO) = stg_END_TSO_QUEUE_closure;
- StgTSO_why_blocked(CurrentTSO) = BlockedOnMVar::I16;
StgTSO_block_info(CurrentTSO) = mvar;
+ // write barrier for throwTo(), which looks at block_info
+ // if why_blocked==BlockedOnMVar.
+ prim %write_barrier() [];
+ StgTSO_why_blocked(CurrentTSO) = BlockedOnMVar::I16;
StgMVar_tail(mvar) = CurrentTSO;
R1 = mvar;
CurrentTSO) [];
}
StgTSO__link(CurrentTSO) = stg_END_TSO_QUEUE_closure;
- StgTSO_why_blocked(CurrentTSO) = BlockedOnMVar::I16;
StgTSO_block_info(CurrentTSO) = mvar;
+ // write barrier for throwTo(), which looks at block_info
+ // if why_blocked==BlockedOnMVar.
+ prim %write_barrier() [];
+ StgTSO_why_blocked(CurrentTSO) = BlockedOnMVar::I16;
StgMVar_tail(mvar) = CurrentTSO;
R1 = mvar;
RET_NP(ok,val);
}
+/* -----------------------------------------------------------------------------
+ Misc. primitives
+ -------------------------------------------------------------------------- */
+
+// Write the cost center stack of the first argument on stderr; return
+// the second. Possibly only makes sense for already evaluated
+// things?
+traceCcszh_fast
+{
+ W_ ccs;
+
+#ifdef PROFILING
+ ccs = StgHeader_ccs(UNTAG(R1));
+ foreign "C" fprintCCS_stderr(ccs "ptr") [R2];
+#endif
+
+ R1 = R2;
+ ENTER();
+}
+
getSparkzh_fast
{
W_ spark;
projects / ghc-hetmet.git / blobdiff