#include "Cmm.h"
#ifdef __PIC__
-#ifndef mingw32_HOST_OS
-import __gmpz_init;
-import __gmpz_add;
-import __gmpz_sub;
-import __gmpz_mul;
-import __gmpz_gcd;
-import __gmpn_gcd_1;
-import __gmpn_cmp;
-import __gmpz_tdiv_q;
-import __gmpz_tdiv_r;
-import __gmpz_tdiv_qr;
-import __gmpz_fdiv_qr;
-import __gmpz_divexact;
-import __gmpz_and;
-import __gmpz_xor;
-import __gmpz_ior;
-import __gmpz_com;
-#endif
import pthread_mutex_lock;
import pthread_mutex_unlock;
#endif
import base_ControlziExceptionziBase_nestedAtomically_closure;
import EnterCriticalSection;
import LeaveCriticalSection;
+import ghczmprim_GHCziBool_False_closure;
/*-----------------------------------------------------------------------------
Array Primitives
* round up to the nearest word for the size of the array.
*/
-newByteArrayzh_fast
+stg_newByteArrayzh
{
W_ words, payload_words, n, p;
- MAYBE_GC(NO_PTRS,newByteArrayzh_fast);
+ MAYBE_GC(NO_PTRS,stg_newByteArrayzh);
n = R1;
payload_words = ROUNDUP_BYTES_TO_WDS(n);
words = BYTES_TO_WDS(SIZEOF_StgArrWords) + payload_words;
RET_P(p);
}
-newPinnedByteArrayzh_fast
+#define BA_ALIGN 16
+#define BA_MASK (BA_ALIGN-1)
+
+stg_newPinnedByteArrayzh
{
- W_ words, payload_words, n, p;
+ W_ words, bytes, payload_words, p;
+
+ MAYBE_GC(NO_PTRS,stg_newPinnedByteArrayzh);
+ 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);
- MAYBE_GC(NO_PTRS,newPinnedByteArrayzh_fast);
- n = R1;
- payload_words = ROUNDUP_BYTES_TO_WDS(n);
+ ("ptr" p) = foreign "C" allocatePinned(words) [];
+ TICK_ALLOC_PRIM(SIZEOF_StgArrWords,WDS(payload_words),0);
- // 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;
- }
+ /* 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);
+}
+
+stg_newAlignedPinnedByteArrayzh
+{
+ W_ words, bytes, payload_words, p, alignment;
+
+ MAYBE_GC(NO_PTRS,stg_newAlignedPinnedByteArrayzh);
+ 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;
RET_P(p);
}
-newArrayzh_fast
+stg_newArrayzh
{
W_ words, n, init, arr, p;
/* Args: R1 = words, R2 = initialisation value */
n = R1;
- MAYBE_GC(R2_PTR,newArrayzh_fast);
+ MAYBE_GC(R2_PTR,stg_newArrayzh);
words = BYTES_TO_WDS(SIZEOF_StgMutArrPtrs) + n;
("ptr" arr) = foreign "C" allocateLocal(MyCapability() "ptr",words) [R2];
RET_P(arr);
}
-unsafeThawArrayzh_fast
+stg_unsafeThawArrayzh
{
// SUBTLETY TO DO WITH THE OLD GEN MUTABLE LIST
//
MutVar primitives
-------------------------------------------------------------------------- */
-newMutVarzh_fast
+stg_newMutVarzh
{
W_ mv;
/* Args: R1 = initialisation value */
- ALLOC_PRIM( SIZEOF_StgMutVar, R1_PTR, newMutVarzh_fast);
+ ALLOC_PRIM( SIZEOF_StgMutVar, R1_PTR, stg_newMutVarzh);
mv = Hp - SIZEOF_StgMutVar + WDS(1);
SET_HDR(mv,stg_MUT_VAR_DIRTY_info,W_[CCCS]);
RET_P(mv);
}
-atomicModifyMutVarzh_fast
+stg_atomicModifyMutVarzh
{
W_ mv, f, z, x, y, r, h;
/* Args: R1 :: MutVar#, R2 :: a -> (a,b) */
#define SIZE (THUNK_2_SIZE + THUNK_1_SIZE + THUNK_1_SIZE)
- HP_CHK_GEN_TICKY(SIZE, R1_PTR & R2_PTR, atomicModifyMutVarzh_fast);
+ HP_CHK_GEN_TICKY(SIZE, R1_PTR & R2_PTR, stg_atomicModifyMutVarzh);
mv = R1;
f = R2;
STRING(stg_weak_msg,"New weak pointer at %p\n")
-mkWeakzh_fast
+stg_mkWeakzh
{
/* R1 = key
R2 = value
R3 = stg_NO_FINALIZER_closure;
}
- ALLOC_PRIM( SIZEOF_StgWeak, R1_PTR & R2_PTR & R3_PTR, mkWeakzh_fast );
+ ALLOC_PRIM( SIZEOF_StgWeak, R1_PTR & R2_PTR & R3_PTR, stg_mkWeakzh );
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);
}
+stg_mkWeakForeignEnvzh
+{
+ /* 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, stg_mkWeakForeignEnvzh );
+
+ 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
+stg_finalizzeWeakzh
{
/* 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);
}
}
-deRefWeakzh_fast
+stg_deRefWeakzh
{
/* R1 = weak ptr */
W_ w, code, val;
}
/* -----------------------------------------------------------------------------
- Arbitrary-precision Integer operations.
-
- There are some assumptions in this code that mp_limb_t == W_. This is
- the case for all the platforms that GHC supports, currently.
+ Floating point operations.
-------------------------------------------------------------------------- */
-int2Integerzh_fast
-{
- /* arguments: R1 = Int# */
-
- W_ val, s, p; /* to avoid aliasing */
-
- val = R1;
- ALLOC_PRIM( SIZEOF_StgArrWords + WDS(1), NO_PTRS, int2Integerzh_fast );
-
- p = Hp - SIZEOF_StgArrWords;
- SET_HDR(p, stg_ARR_WORDS_info, W_[CCCS]);
- StgArrWords_words(p) = 1;
-
- /* mpz_set_si is inlined here, makes things simpler */
- if (%lt(val,0)) {
- s = -1;
- Hp(0) = -val;
- } else {
- if (%gt(val,0)) {
- s = 1;
- Hp(0) = val;
- } else {
- s = 0;
- }
- }
-
- /* returns (# size :: Int#,
- data :: ByteArray#
- #)
- */
- RET_NP(s,p);
-}
-
-word2Integerzh_fast
-{
- /* arguments: R1 = Word# */
-
- W_ val, s, p; /* to avoid aliasing */
-
- val = R1;
-
- ALLOC_PRIM( SIZEOF_StgArrWords + WDS(1), NO_PTRS, word2Integerzh_fast);
-
- p = Hp - SIZEOF_StgArrWords;
- SET_HDR(p, stg_ARR_WORDS_info, W_[CCCS]);
- StgArrWords_words(p) = 1;
-
- if (val != 0) {
- s = 1;
- W_[Hp] = val;
- } else {
- s = 0;
- }
-
- /* returns (# size :: Int#,
- data :: ByteArray# #)
- */
- RET_NP(s,p);
-}
-
-
-/*
- * 'long long' primops for converting to/from Integers.
- */
-
-#ifdef SUPPORT_LONG_LONGS
-
-int64ToIntegerzh_fast
-{
- /* arguments: L1 = Int64# */
-
- L_ val;
- W_ hi, lo, s, neg, words_needed, p;
-
- val = L1;
- neg = 0;
-
- hi = TO_W_(val >> 32);
- lo = TO_W_(val);
-
- if ( hi == 0 || (hi == 0xFFFFFFFF && lo != 0) ) {
- // minimum is one word
- words_needed = 1;
- } else {
- words_needed = 2;
- }
-
- ALLOC_PRIM( SIZEOF_StgArrWords + WDS(words_needed),
- NO_PTRS, int64ToIntegerzh_fast );
-
- p = Hp - SIZEOF_StgArrWords - WDS(words_needed) + WDS(1);
- SET_HDR(p, stg_ARR_WORDS_info, W_[CCCS]);
- StgArrWords_words(p) = words_needed;
-
- if ( %lt(hi,0) ) {
- neg = 1;
- lo = -lo;
- if(lo == 0) {
- hi = -hi;
- } else {
- hi = -hi - 1;
- }
- }
-
- if ( words_needed == 2 ) {
- s = 2;
- Hp(-1) = lo;
- Hp(0) = hi;
- } else {
- if ( lo != 0 ) {
- s = 1;
- Hp(0) = lo;
- } else /* val==0 */ {
- s = 0;
- }
- }
- if ( neg != 0 ) {
- s = -s;
- }
-
- /* returns (# size :: Int#,
- data :: ByteArray# #)
- */
- RET_NP(s,p);
-}
-word64ToIntegerzh_fast
-{
- /* arguments: L1 = Word64# */
-
- L_ val;
- W_ hi, lo, s, words_needed, p;
-
- val = L1;
- hi = TO_W_(val >> 32);
- lo = TO_W_(val);
-
- if ( hi != 0 ) {
- words_needed = 2;
- } else {
- words_needed = 1;
- }
-
- ALLOC_PRIM( SIZEOF_StgArrWords + WDS(words_needed),
- NO_PTRS, word64ToIntegerzh_fast );
-
- p = Hp - SIZEOF_StgArrWords - WDS(words_needed) + WDS(1);
- SET_HDR(p, stg_ARR_WORDS_info, W_[CCCS]);
- StgArrWords_words(p) = words_needed;
-
- if ( hi != 0 ) {
- s = 2;
- Hp(-1) = lo;
- Hp(0) = hi;
- } else {
- if ( lo != 0 ) {
- s = 1;
- Hp(0) = lo;
- } else /* val==0 */ {
- s = 0;
- }
- }
-
- /* returns (# size :: Int#,
- data :: ByteArray# #)
- */
- 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); \
- \
- /* call doYouWantToGC() */ \
- MAYBE_GC(R2_PTR & R4_PTR, name); \
- \
- s1 = W_TO_INT(R1); \
- d1 = R2; \
- s2 = W_TO_INT(R3); \
- d2 = R4; \
- \
- 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); \
- MP_INT__mp_alloc(mp_tmp2) = W_TO_INT(StgArrWords_words(d2)); \
- MP_INT__mp_size(mp_tmp2) = (s2); \
- MP_INT__mp_d(mp_tmp2) = BYTE_ARR_CTS(d2); \
- \
- foreign "C" __gmpz_init(mp_result1 "ptr") []; \
- \
- /* Perform the operation */ \
- foreign "C" mp_fun(mp_result1 "ptr",mp_tmp1 "ptr",mp_tmp2 "ptr") []; \
- \
- RET_NP(TO_W_(MP_INT__mp_size(mp_result1)), \
- MP_INT__mp_d(mp_result1) - SIZEOF_StgArrWords); \
-}
-
-#define GMP_TAKE1_RET1(name,mp_fun) \
-name \
-{ \
- CInt s1; \
- W_ d1; \
- FETCH_MP_TEMP(mp_tmp1); \
- FETCH_MP_TEMP(mp_result1) \
- \
- /* call doYouWantToGC() */ \
- MAYBE_GC(R2_PTR, name); \
- \
- d1 = R2; \
- s1 = W_TO_INT(R1); \
- \
- 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); \
- \
- foreign "C" __gmpz_init(mp_result1 "ptr") []; \
- \
- /* Perform the operation */ \
- foreign "C" mp_fun(mp_result1 "ptr",mp_tmp1 "ptr") []; \
- \
- RET_NP(TO_W_(MP_INT__mp_size(mp_result1)), \
- MP_INT__mp_d(mp_result1) - SIZEOF_StgArrWords); \
-}
-
-#define GMP_TAKE2_RET2(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) \
- \
- /* call doYouWantToGC() */ \
- MAYBE_GC(R2_PTR & R4_PTR, name); \
- \
- s1 = W_TO_INT(R1); \
- d1 = R2; \
- s2 = W_TO_INT(R3); \
- d2 = R4; \
- \
- 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); \
- MP_INT__mp_alloc(mp_tmp2) = W_TO_INT(StgArrWords_words(d2)); \
- MP_INT__mp_size(mp_tmp2) = (s2); \
- MP_INT__mp_d(mp_tmp2) = BYTE_ARR_CTS(d2); \
- \
- foreign "C" __gmpz_init(mp_result1 "ptr") []; \
- foreign "C" __gmpz_init(mp_result2 "ptr") []; \
- \
- /* Perform the operation */ \
- foreign "C" mp_fun(mp_result1 "ptr",mp_result2 "ptr",mp_tmp1 "ptr",mp_tmp2 "ptr") []; \
- \
- RET_NPNP(TO_W_(MP_INT__mp_size(mp_result1)), \
- MP_INT__mp_d(mp_result1) - SIZEOF_StgArrWords, \
- TO_W_(MP_INT__mp_size(mp_result2)), \
- MP_INT__mp_d(mp_result2) - SIZEOF_StgArrWords); \
-}
-
-GMP_TAKE2_RET1(plusIntegerzh_fast, __gmpz_add)
-GMP_TAKE2_RET1(minusIntegerzh_fast, __gmpz_sub)
-GMP_TAKE2_RET1(timesIntegerzh_fast, __gmpz_mul)
-GMP_TAKE2_RET1(gcdIntegerzh_fast, __gmpz_gcd)
-GMP_TAKE2_RET1(quotIntegerzh_fast, __gmpz_tdiv_q)
-GMP_TAKE2_RET1(remIntegerzh_fast, __gmpz_tdiv_r)
-GMP_TAKE2_RET1(divExactIntegerzh_fast, __gmpz_divexact)
-GMP_TAKE2_RET1(andIntegerzh_fast, __gmpz_and)
-GMP_TAKE2_RET1(orIntegerzh_fast, __gmpz_ior)
-GMP_TAKE2_RET1(xorIntegerzh_fast, __gmpz_xor)
-GMP_TAKE1_RET1(complementIntegerzh_fast, __gmpz_com)
-
-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] = R1;
- (r) = foreign "C" __gmpn_gcd_1(mp_tmp_w "ptr", 1, R2) [];
-
- R1 = r;
- /* Result parked in R1, return via info-pointer at TOS */
- jump %ENTRY_CODE(Sp(0));
-}
-
-
-gcdIntegerIntzh_fast
-{
- /* R1 = s1; R2 = d1; R3 = the int */
- W_ s1;
- (s1) = foreign "C" __gmpn_gcd_1( BYTE_ARR_CTS(R2) "ptr", R1, R3) [];
- R1 = s1;
-
- /* Result parked in R1, return via info-pointer at TOS */
- jump %ENTRY_CODE(Sp(0));
-}
-
-
-cmpIntegerIntzh_fast
-{
- /* R1 = s1; R2 = d1; R3 = the int */
- W_ usize, vsize, v_digit, u_digit;
-
- usize = R1;
- vsize = 0;
- v_digit = R3;
-
- // paraphrased from __gmpz_cmp_si() in the GMP sources
- if (%gt(v_digit,0)) {
- vsize = 1;
- } else {
- if (%lt(v_digit,0)) {
- vsize = -1;
- v_digit = -v_digit;
- }
- }
-
- if (usize != vsize) {
- R1 = usize - vsize;
- jump %ENTRY_CODE(Sp(0));
- }
-
- if (usize == 0) {
- R1 = 0;
- jump %ENTRY_CODE(Sp(0));
- }
-
- u_digit = W_[BYTE_ARR_CTS(R2)];
-
- if (u_digit == v_digit) {
- R1 = 0;
- jump %ENTRY_CODE(Sp(0));
- }
-
- if (%gtu(u_digit,v_digit)) { // NB. unsigned: these are mp_limb_t's
- R1 = usize;
- } else {
- R1 = -usize;
- }
-
- jump %ENTRY_CODE(Sp(0));
-}
-
-cmpIntegerzh_fast
-{
- /* R1 = s1; R2 = d1; R3 = s2; R4 = d2 */
- W_ usize, vsize, size, up, vp;
- CInt cmp;
-
- // paraphrased from __gmpz_cmp() in the GMP sources
- usize = R1;
- vsize = R3;
-
- if (usize != vsize) {
- R1 = usize - vsize;
- jump %ENTRY_CODE(Sp(0));
- }
-
- if (usize == 0) {
- R1 = 0;
- jump %ENTRY_CODE(Sp(0));
- }
-
- if (%lt(usize,0)) { // NB. not <, which is unsigned
- size = -usize;
- } else {
- size = usize;
- }
-
- up = BYTE_ARR_CTS(R2);
- vp = BYTE_ARR_CTS(R4);
-
- (cmp) = foreign "C" __gmpn_cmp(up "ptr", vp "ptr", size) [];
-
- if (cmp == 0 :: CInt) {
- R1 = 0;
- jump %ENTRY_CODE(Sp(0));
- }
-
- if (%lt(cmp,0 :: CInt) == %lt(usize,0)) {
- R1 = 1;
- } else {
- R1 = (-1);
- }
- /* Result parked in R1, return via info-pointer at TOS */
- jump %ENTRY_CODE(Sp(0));
-}
-
-integer2Intzh_fast
-{
- /* R1 = s; R2 = d */
- W_ r, s;
-
- s = R1;
- if (s == 0) {
- r = 0;
- } else {
- r = W_[R2 + SIZEOF_StgArrWords];
- if (%lt(s,0)) {
- r = -r;
- }
- }
- /* Result parked in R1, return via info-pointer at TOS */
- R1 = r;
- jump %ENTRY_CODE(Sp(0));
-}
-
-integer2Wordzh_fast
-{
- /* R1 = s; R2 = d */
- W_ r, s;
-
- s = R1;
- if (s == 0) {
- r = 0;
- } else {
- r = W_[R2 + SIZEOF_StgArrWords];
- if (%lt(s,0)) {
- r = -r;
- }
- }
- /* Result parked in R1, return via info-pointer at TOS */
- R1 = r;
- jump %ENTRY_CODE(Sp(0));
-}
-
-decodeFloatzh_fast
+stg_decodeFloatzuIntzh
{
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);
-}
+ W_ mp_tmp1;
+ W_ mp_tmp_w;
-decodeFloatzuIntzh_fast
-{
- W_ p;
- F_ arg;
- FETCH_MP_TEMP(mp_tmp1);
- FETCH_MP_TEMP(mp_tmp_w);
+ STK_CHK_GEN( WDS(2), NO_PTRS, stg_decodeFloatzuIntzh );
+
+ mp_tmp1 = Sp - WDS(1);
+ mp_tmp_w = Sp - WDS(2);
/* arguments: F1 = Float# */
arg = F1;
RET_NN(W_[mp_tmp1], W_[mp_tmp_w]);
}
-#define DOUBLE_MANTISSA_SIZE SIZEOF_DOUBLE
-#define ARR_SIZE (SIZEOF_StgArrWords + DOUBLE_MANTISSA_SIZE)
-
-decodeDoublezh_fast
+stg_decodeDoublezu2Intzh
{
D_ arg;
W_ p;
- FETCH_MP_TEMP(mp_tmp1);
- FETCH_MP_TEMP(mp_tmp_w);
+ W_ mp_tmp1;
+ W_ mp_tmp2;
+ W_ mp_result1;
+ W_ mp_result2;
- /* arguments: D1 = Double# */
- arg = D1;
+ STK_CHK_GEN( WDS(4), NO_PTRS, stg_decodeDoublezu2Intzh );
- ALLOC_PRIM( ARR_SIZE, NO_PTRS, decodeDoublezh_fast );
-
- /* Be prepared to tell Lennart-coded __decodeDouble
- where mantissa.d can be put (it does not care about the rest) */
- p = Hp - ARR_SIZE + WDS(1);
- SET_HDR(p, stg_ARR_WORDS_info, W_[CCCS]);
- StgArrWords_words(p) = BYTES_TO_WDS(DOUBLE_MANTISSA_SIZE);
- MP_INT__mp_d(mp_tmp1) = BYTE_ARR_CTS(p);
-
- /* Perform the operation */
- foreign "C" __decodeDouble(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);
-}
-
-decodeDoublezu2Intzh_fast
-{
- D_ arg;
- W_ p;
- FETCH_MP_TEMP(mp_tmp1);
- FETCH_MP_TEMP(mp_tmp2);
- FETCH_MP_TEMP(mp_result1);
- FETCH_MP_TEMP(mp_result2);
+ 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;
* Concurrency primitives
* -------------------------------------------------------------------------- */
-forkzh_fast
+stg_forkzh
{
/* args: R1 = closure to spark */
- MAYBE_GC(R1_PTR, forkzh_fast);
+ MAYBE_GC(R1_PTR, stg_forkzh);
W_ closure;
W_ threadid;
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);
}
-forkOnzh_fast
+stg_forkOnzh
{
/* args: R1 = cpu, R2 = closure to spark */
- MAYBE_GC(R2_PTR, forkOnzh_fast);
+ MAYBE_GC(R2_PTR, stg_forkOnzh);
W_ cpu;
W_ closure;
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);
}
-yieldzh_fast
+stg_yieldzh
{
jump stg_yield_noregs;
}
-myThreadIdzh_fast
+stg_myThreadIdzh
{
/* no args. */
RET_P(CurrentTSO);
}
-labelThreadzh_fast
+stg_labelThreadzh
{
/* args:
R1 = ThreadId#
jump %ENTRY_CODE(Sp(0));
}
-isCurrentThreadBoundzh_fast
+stg_isCurrentThreadBoundzh
{
/* no args */
W_ r;
RET_N(r);
}
-threadStatuszh_fast
+stg_threadStatuszh
{
/* args: R1 :: ThreadId# */
W_ tso;
#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;
frame = Sp;
trec = StgTSO_trec(CurrentTSO);
- ("ptr" outer) = foreign "C" stmGetEnclosingTRec(trec "ptr") [];
+ outer = StgTRecHeader_enclosing_trec(trec);
(r) = foreign "C" stmCommitNestedTransaction(MyCapability() "ptr", trec "ptr") [];
if (r != 0) {
/* Succeeded (either first branch or second branch) */
#if defined(PROFILING)
W_ unused1, W_ unused2,
#endif
- "ptr" W_ unused3, "ptr" W_ unused4)
+ P_ code, P_ next_invariant_to_check, P_ result)
{
W_ frame, trec, valid, next_invariant, q, outer;
- frame = Sp;
- trec = StgTSO_trec(CurrentTSO);
- ("ptr" outer) = foreign "C" stmGetEnclosingTRec(trec "ptr") [];
+ frame = Sp;
+ trec = StgTSO_trec(CurrentTSO);
+ result = R1;
+ outer = StgTRecHeader_enclosing_trec(trec);
if (outer == NO_TREC) {
/* First time back at the atomically frame -- pick up invariants */
("ptr" q) = foreign "C" stmGetInvariantsToCheck(MyCapability() "ptr", trec "ptr") [];
StgAtomicallyFrame_next_invariant_to_check(frame) = q;
+ StgAtomicallyFrame_result(frame) = result;
} else {
/* Second/subsequent time back at the atomically frame -- abort the
if (valid != 0) {
/* Transaction was valid: commit succeeded */
StgTSO_trec(CurrentTSO) = NO_TREC;
+ R1 = StgAtomicallyFrame_result(frame);
Sp = Sp + SIZEOF_StgAtomicallyFrame;
jump %ENTRY_CODE(Sp(SP_OFF));
} else {
#if defined(PROFILING)
W_ unused1, W_ unused2,
#endif
- "ptr" W_ unused3, "ptr" W_ unused4)
+ P_ code, P_ next_invariant_to_check, P_ result)
{
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;
trec = StgTSO_trec(CurrentTSO);
- ("ptr" outer) = foreign "C" stmGetEnclosingTRec(trec "ptr") [];
+ outer = StgTRecHeader_enclosing_trec(trec);
(r) = foreign "C" stmCommitNestedTransaction(MyCapability() "ptr", trec "ptr") [];
if (r != 0) {
/* Commit succeeded */
// Primop definition ------------------------------------------------------------
-atomicallyzh_fast
+stg_atomicallyzh
{
W_ frame;
W_ old_trec;
W_ new_trec;
// stmStartTransaction may allocate
- MAYBE_GC (R1_PTR, atomicallyzh_fast);
+ MAYBE_GC (R1_PTR, stg_atomicallyzh);
/* Args: R1 = m :: STM a */
- STK_CHK_GEN(SIZEOF_StgAtomicallyFrame + WDS(1), R1_PTR, atomicallyzh_fast);
+ STK_CHK_GEN(SIZEOF_StgAtomicallyFrame + WDS(1), R1_PTR, stg_atomicallyzh);
old_trec = StgTSO_trec(CurrentTSO);
/* Nested transactions are not allowed; raise an exception */
if (old_trec != NO_TREC) {
R1 = base_ControlziExceptionziBase_nestedAtomically_closure;
- jump raisezh_fast;
+ jump stg_raisezh;
}
/* Set up the atomically frame */
SET_HDR(frame,stg_atomically_frame_info, W_[CCCS]);
StgAtomicallyFrame_code(frame) = R1;
+ StgAtomicallyFrame_result(frame) = NO_TREC;
StgAtomicallyFrame_next_invariant_to_check(frame) = END_INVARIANT_CHECK_QUEUE;
/* Start the memory transcation */
}
-catchSTMzh_fast
+stg_catchSTMzh
{
W_ frame;
/* Args: R1 :: STM a */
/* Args: R2 :: Exception -> STM a */
- STK_CHK_GEN(SIZEOF_StgCatchSTMFrame + WDS(1), R1_PTR & R2_PTR, catchSTMzh_fast);
+ STK_CHK_GEN(SIZEOF_StgCatchSTMFrame + WDS(1), R1_PTR & R2_PTR, stg_catchSTMzh);
/* Set up the catch frame */
Sp = Sp - SIZEOF_StgCatchSTMFrame;
}
-catchRetryzh_fast
+stg_catchRetryzh
{
W_ frame;
W_ new_trec;
W_ trec;
// stmStartTransaction may allocate
- MAYBE_GC (R1_PTR & R2_PTR, catchRetryzh_fast);
+ MAYBE_GC (R1_PTR & R2_PTR, stg_catchRetryzh);
/* Args: R1 :: STM a */
/* Args: R2 :: STM a */
- STK_CHK_GEN(SIZEOF_StgCatchRetryFrame + WDS(1), R1_PTR & R2_PTR, catchRetryzh_fast);
+ STK_CHK_GEN(SIZEOF_StgCatchRetryFrame + WDS(1), R1_PTR & R2_PTR, stg_catchRetryzh);
/* Start a nested transaction within which to run the first code */
trec = StgTSO_trec(CurrentTSO);
}
-retryzh_fast
+stg_retryzh
{
W_ frame_type;
W_ frame;
W_ outer;
W_ r;
- MAYBE_GC (NO_PTRS, retryzh_fast); // STM operations may allocate
+ MAYBE_GC (NO_PTRS, stg_retryzh); // STM operations may allocate
// Find the enclosing ATOMICALLY_FRAME or CATCH_RETRY_FRAME
retry_pop_stack:
Sp = StgTSO_sp(CurrentTSO);
frame = Sp;
trec = StgTSO_trec(CurrentTSO);
- ("ptr" outer) = foreign "C" stmGetEnclosingTRec(trec "ptr") [];
+ outer = StgTRecHeader_enclosing_trec(trec);
if (frame_type == CATCH_RETRY_FRAME) {
// The retry reaches a CATCH_RETRY_FRAME before the atomic frame
foreign "C" stmFreeAbortedTRec(MyCapability() "ptr", trec "ptr") [];
trec = outer;
StgTSO_trec(CurrentTSO) = trec;
- ("ptr" outer) = foreign "C" stmGetEnclosingTRec(trec "ptr") [];
+ outer = StgTRecHeader_enclosing_trec(trec);
}
ASSERT(outer == NO_TREC);
}
-checkzh_fast
+stg_checkzh
{
W_ trec, closure;
/* Args: R1 = invariant closure */
- MAYBE_GC (R1_PTR, checkzh_fast);
+ MAYBE_GC (R1_PTR, stg_checkzh);
trec = StgTSO_trec(CurrentTSO);
closure = R1;
}
-newTVarzh_fast
+stg_newTVarzh
{
W_ tv;
W_ new_value;
/* Args: R1 = initialisation value */
- MAYBE_GC (R1_PTR, newTVarzh_fast);
+ MAYBE_GC (R1_PTR, stg_newTVarzh);
new_value = R1;
("ptr" tv) = foreign "C" stmNewTVar(MyCapability() "ptr", new_value "ptr") [];
RET_P(tv);
}
-readTVarzh_fast
+stg_readTVarzh
{
W_ trec;
W_ tvar;
/* Args: R1 = TVar closure */
- MAYBE_GC (R1_PTR, readTVarzh_fast); // Call to stmReadTVar may allocate
+ MAYBE_GC (R1_PTR, stg_readTVarzh); // Call to stmReadTVar may allocate
trec = StgTSO_trec(CurrentTSO);
tvar = R1;
("ptr" result) = foreign "C" stmReadTVar(MyCapability() "ptr", trec "ptr", tvar "ptr") [];
RET_P(result);
}
-readTVarIOzh_fast
+stg_readTVarIOzh
{
W_ result;
RET_P(result);
}
-writeTVarzh_fast
+stg_writeTVarzh
{
W_ trec;
W_ tvar;
/* Args: R1 = TVar closure */
/* R2 = New value */
- MAYBE_GC (R1_PTR & R2_PTR, writeTVarzh_fast); // Call to stmWriteTVar may allocate
+ MAYBE_GC (R1_PTR & R2_PTR, stg_writeTVarzh); // Call to stmWriteTVar may allocate
trec = StgTSO_trec(CurrentTSO);
tvar = R1;
new_value = R2;
*
* -------------------------------------------------------------------------- */
-isEmptyMVarzh_fast
+stg_isEmptyMVarzh
{
/* args: R1 = MVar closure */
}
}
-newMVarzh_fast
+stg_newMVarzh
{
/* args: none */
W_ mvar;
- ALLOC_PRIM ( SIZEOF_StgMVar, NO_PTRS, newMVarzh_fast );
+ ALLOC_PRIM ( SIZEOF_StgMVar, NO_PTRS, stg_newMVarzh );
mvar = Hp - SIZEOF_StgMVar + WDS(1);
SET_HDR(mvar,stg_MVAR_DIRTY_info,W_[CCCS]);
StgTSO_sp(tso) = StgTSO_sp(tso) + WDS(3); \
lval = W_[StgTSO_sp(tso) - WDS(1)];
-takeMVarzh_fast
+stg_takeMVarzh
{
W_ mvar, val, info, tso;
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;
tso = StgMVar_head(mvar);
PerformPut(tso,StgMVar_value(mvar));
- if (TO_W_(StgTSO_flags(tso)) & TSO_DIRTY == 0) {
+ if (TO_W_(StgTSO_dirty(tso)) == 0) {
foreign "C" dirty_TSO(MyCapability() "ptr", tso "ptr") [];
}
}
-tryTakeMVarzh_fast
+stg_tryTakeMVarzh
{
W_ mvar, val, info, tso;
/* actually perform the putMVar for the thread that we just woke up */
tso = StgMVar_head(mvar);
PerformPut(tso,StgMVar_value(mvar));
- if (TO_W_(StgTSO_flags(tso)) & TSO_DIRTY == 0) {
+ if (TO_W_(StgTSO_dirty(tso)) == 0) {
foreign "C" dirty_TSO(MyCapability() "ptr", tso "ptr") [];
}
}
-putMVarzh_fast
+stg_putMVarzh
{
W_ mvar, val, info, tso;
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;
/* actually perform the takeMVar */
tso = StgMVar_head(mvar);
PerformTake(tso, val);
- if (TO_W_(StgTSO_flags(tso)) & TSO_DIRTY == 0) {
+ if (TO_W_(StgTSO_dirty(tso)) == 0) {
foreign "C" dirty_TSO(MyCapability() "ptr", tso "ptr") [];
}
}
-tryPutMVarzh_fast
+stg_tryPutMVarzh
{
W_ mvar, info, tso;
/* actually perform the takeMVar */
tso = StgMVar_head(mvar);
PerformTake(tso, R2);
- if (TO_W_(StgTSO_flags(tso)) & TSO_DIRTY == 0) {
+ if (TO_W_(StgTSO_dirty(tso)) == 0) {
foreign "C" dirty_TSO(MyCapability() "ptr", tso "ptr") [];
}
Stable pointer primitives
------------------------------------------------------------------------- */
-makeStableNamezh_fast
+stg_makeStableNamezh
{
W_ index, sn_obj;
- ALLOC_PRIM( SIZEOF_StgStableName, R1_PTR, makeStableNamezh_fast );
+ ALLOC_PRIM( SIZEOF_StgStableName, R1_PTR, stg_makeStableNamezh );
(index) = foreign "C" lookupStableName(R1 "ptr") [];
}
-makeStablePtrzh_fast
+stg_makeStablePtrzh
{
/* Args: R1 = a */
W_ sp;
- MAYBE_GC(R1_PTR, makeStablePtrzh_fast);
+ MAYBE_GC(R1_PTR, stg_makeStablePtrzh);
("ptr" sp) = foreign "C" getStablePtr(R1 "ptr") [];
RET_N(sp);
}
-deRefStablePtrzh_fast
+stg_deRefStablePtrzh
{
/* Args: R1 = the stable ptr */
W_ r, sp;
Bytecode object primitives
------------------------------------------------------------------------- */
-newBCOzh_fast
+stg_newBCOzh
{
/* R1 = instrs
R2 = literals
words = BYTES_TO_WDS(SIZEOF_StgBCO) + StgArrWords_words(bitmap_arr);
bytes = WDS(words);
- ALLOC_PRIM( bytes, R1_PTR&R2_PTR&R3_PTR&R5_PTR, newBCOzh_fast );
+ ALLOC_PRIM( bytes, R1_PTR&R2_PTR&R3_PTR&R5_PTR, stg_newBCOzh );
bco = Hp - bytes + WDS(1);
SET_HDR(bco, stg_BCO_info, W_[CCCS]);
}
-mkApUpd0zh_fast
+stg_mkApUpd0zh
{
// R1 = the BCO# for the AP
//
ASSERT(%INFO_TYPE(%GET_STD_INFO(R1)) == HALF_W_(BCO) &&
StgBCO_arity(R1) == HALF_W_(0));
- HP_CHK_GEN_TICKY(SIZEOF_StgAP, R1_PTR, mkApUpd0zh_fast);
+ HP_CHK_GEN_TICKY(SIZEOF_StgAP, R1_PTR, stg_mkApUpd0zh);
TICK_ALLOC_UP_THK(0, 0);
CCCS_ALLOC(SIZEOF_StgAP);
RET_P(ap);
}
-unpackClosurezh_fast
+stg_unpackClosurezh
{
/* args: R1 = closure to analyze */
// TODO: Consider the absence of ptrs or nonptrs as a special case ?
nptrs_arr_sz = SIZEOF_StgArrWords + WDS(nptrs);
ptrs_arr_sz = SIZEOF_StgMutArrPtrs + WDS(ptrs);
- ALLOC_PRIM (ptrs_arr_sz + nptrs_arr_sz, R1_PTR, unpackClosurezh_fast);
+ ALLOC_PRIM (ptrs_arr_sz + nptrs_arr_sz, R1_PTR, stg_unpackClosurezh);
W_ clos;
clos = UNTAG(R1);
} \
W_[blocked_queue_tl] = tso;
-waitReadzh_fast
+stg_waitReadzh
{
/* args: R1 */
#ifdef THREADED_RTS
#endif
}
-waitWritezh_fast
+stg_waitWritezh
{
/* args: R1 */
#ifdef THREADED_RTS
}
-STRING(stg_delayzh_malloc_str, "delayzh_fast")
-delayzh_fast
+STRING(stg_delayzh_malloc_str, "stg_delayzh")
+stg_delayzh
{
#ifdef mingw32_HOST_OS
W_ ares;
#ifdef mingw32_HOST_OS
-STRING(stg_asyncReadzh_malloc_str, "asyncReadzh_fast")
-asyncReadzh_fast
+STRING(stg_asyncReadzh_malloc_str, "stg_asyncReadzh")
+stg_asyncReadzh
{
W_ ares;
CInt reqID;
#endif
}
-STRING(stg_asyncWritezh_malloc_str, "asyncWritezh_fast")
-asyncWritezh_fast
+STRING(stg_asyncWritezh_malloc_str, "stg_asyncWritezh")
+stg_asyncWritezh
{
W_ ares;
CInt reqID;
#endif
}
-STRING(stg_asyncDoProczh_malloc_str, "asyncDoProczh_fast")
-asyncDoProczh_fast
+STRING(stg_asyncDoProczh_malloc_str, "stg_asyncDoProczh")
+stg_asyncDoProczh
{
W_ ares;
CInt reqID;
// evaluation by the current thread are also under evaluation by
// another thread. It relies on *both* threads doing noDuplicate#;
// the second one will get blocked if they are duplicating some work.
-noDuplicatezh_fast
+stg_noDuplicatezh
{
SAVE_THREAD_STATE();
ASSERT(StgTSO_what_next(CurrentTSO) == ThreadRunGHC::I16);
}
}
-getApStackValzh_fast
+stg_getApStackValzh
{
W_ ap_stack, offset, val, ok;
RET_NP(ok,val);
}
-getSparkzh_fast
+/* -----------------------------------------------------------------------------
+ Misc. primitives
+ -------------------------------------------------------------------------- */
+
+// Write the cost center stack of the first argument on stderr; return
+// the second. Possibly only makes sense for already evaluated
+// things?
+stg_traceCcszh
+{
+ W_ ccs;
+
+#ifdef PROFILING
+ ccs = StgHeader_ccs(UNTAG(R1));
+ foreign "C" fprintCCS_stderr(ccs "ptr") [R2];
+#endif
+
+ R1 = R2;
+ ENTER();
+}
+
+stg_getSparkzh
{
W_ spark;
#ifndef THREADED_RTS
RET_NP(0,ghczmprim_GHCziBool_False_closure);
#else
- (spark) = foreign "C" tryStealSpark(MyCapability());
+ (spark) = foreign "C" findSpark(MyCapability());
if (spark != 0) {
RET_NP(1,spark);
} else {
- (spark) = foreign "C" stealWork (MyCapability());
- if (spark != 0) {
- RET_NP(1,spark);
- } else {
- RET_NP(0,ghczmprim_GHCziBool_False_closure);
-
- }
+ RET_NP(0,ghczmprim_GHCziBool_False_closure);
}
#endif
}
+
+stg_traceEventzh
+{
+ W_ msg;
+ msg = R1;
+#if defined(TRACING) || defined(DEBUG)
+ foreign "C" traceUserMsg(MyCapability() "ptr", msg "ptr") [];
+#endif
+ jump %ENTRY_CODE(Sp(0));
+}