X-Git-Url: http://git.megacz.com/?a=blobdiff_plain;f=ghc%2Frts%2FPrimOps.cmm;h=6a2b36f3b90245887ee36c247be90ed26298730e;hb=04db0e9fa47ce4dfbcb73ec1752d94195f3b394e;hp=e50b17f2c8b1651e673a27abf3bb7da4c7481ca8;hpb=96757f6a4ec72dc609468d3da442db38a73df23e;p=ghc-hetmet.git diff --git a/ghc/rts/PrimOps.cmm b/ghc/rts/PrimOps.cmm index e50b17f..6a2b36f 100644 --- a/ghc/rts/PrimOps.cmm +++ b/ghc/rts/PrimOps.cmm @@ -49,7 +49,7 @@ newByteArrayzh_fast n = R1; payload_words = ROUNDUP_BYTES_TO_WDS(n); words = BYTES_TO_WDS(SIZEOF_StgArrWords) + payload_words; - "ptr" p = foreign "C" allocate(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; @@ -73,7 +73,7 @@ newPinnedByteArrayzh_fast words = words + 1; } - "ptr" p = foreign "C" allocatePinned(words); + "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 @@ -97,10 +97,10 @@ newArrayzh_fast MAYBE_GC(R2_PTR,newArrayzh_fast); words = BYTES_TO_WDS(SIZEOF_StgMutArrPtrs) + n; - "ptr" arr = foreign "C" allocate(words); + "ptr" arr = foreign "C" allocateLocal(MyCapability() "ptr",words) [R2]; TICK_ALLOC_PRIM(SIZEOF_StgMutArrPtrs, WDS(n), 0); - SET_HDR(arr, stg_MUT_ARR_PTRS_info, W_[CCCS]); + SET_HDR(arr, stg_MUT_ARR_PTRS_DIRTY_info, W_[CCCS]); StgMutArrPtrs_ptrs(arr) = n; // Initialise all elements of the the array with the value in R2 @@ -118,8 +118,6 @@ newArrayzh_fast unsafeThawArrayzh_fast { - SET_INFO(R1,stg_MUT_ARR_PTRS_info); - // SUBTLETY TO DO WITH THE OLD GEN MUTABLE LIST // // A MUT_ARR_PTRS lives on the mutable list, but a MUT_ARR_PTRS_FROZEN @@ -127,16 +125,26 @@ unsafeThawArrayzh_fast // it on the mutable list for the GC to remove (removing something from // the mutable list is not easy, because the mut_list is only singly-linked). // + // So that we can tell whether a MUT_ARR_PTRS_FROZEN is on the mutable list, + // when we freeze it we set the info ptr to be MUT_ARR_PTRS_FROZEN0 + // to indicate that it is still on the mutable list. + // // So, when we thaw a MUT_ARR_PTRS_FROZEN, we must cope with two cases: // either it is on a mut_list, or it isn't. We adopt the convention that - // the mut_link field is NULL if it isn't on a mut_list, and the GC - // maintains this invariant. + // the closure type is MUT_ARR_PTRS_FROZEN0 if it is on the mutable list, + // and MUT_ARR_PTRS_FROZEN otherwise. In fact it wouldn't matter if + // we put it on the mutable list more than once, but it would get scavenged + // multiple times during GC, which would be unnecessarily slow. // - if (StgMutClosure_mut_link(R1) == NULL) { - foreign "C" recordMutable(R1 "ptr"); + if (StgHeader_info(R1) != stg_MUT_ARR_PTRS_FROZEN0_info) { + SET_INFO(R1,stg_MUT_ARR_PTRS_DIRTY_info); + foreign "C" recordMutableLock(R1 "ptr") [R1]; + // must be done after SET_INFO, because it ASSERTs closure_MUTABLE() + RET_P(R1); + } else { + SET_INFO(R1,stg_MUT_ARR_PTRS_DIRTY_info); + RET_P(R1); } - - RET_P(R1); } /* ----------------------------------------------------------------------------- @@ -151,7 +159,7 @@ newMutVarzh_fast ALLOC_PRIM( SIZEOF_StgMutVar, R1_PTR, newMutVarzh_fast); mv = Hp - SIZEOF_StgMutVar + WDS(1); - SET_HDR(mv,stg_MUT_VAR_info,W_[CCCS]); + SET_HDR(mv,stg_MUT_VAR_DIRTY_info,W_[CCCS]); StgMutVar_var(mv) = R1; RET_P(mv); @@ -179,18 +187,18 @@ atomicModifyMutVarzh_fast */ #if MIN_UPD_SIZE > 1 -#define THUNK_1_SIZE (SIZEOF_StgHeader + WDS(MIN_UPD_SIZE)) +#define THUNK_1_SIZE (SIZEOF_StgThunkHeader + WDS(MIN_UPD_SIZE)) #define TICK_ALLOC_THUNK_1() TICK_ALLOC_UP_THK(WDS(1),WDS(MIN_UPD_SIZE-1)) #else -#define THUNK_1_SIZE (SIZEOF_StgHeader + WDS(1)) +#define THUNK_1_SIZE (SIZEOF_StgThunkHeader + WDS(1)) #define TICK_ALLOC_THUNK_1() TICK_ALLOC_UP_THK(WDS(1),0) #endif #if MIN_UPD_SIZE > 2 -#define THUNK_2_SIZE (SIZEOF_StgHeader + WDS(MIN_UPD_SIZE)) +#define THUNK_2_SIZE (SIZEOF_StgThunkHeader + WDS(MIN_UPD_SIZE)) #define TICK_ALLOC_THUNK_2() TICK_ALLOC_UP_THK(WDS(2),WDS(MIN_UPD_SIZE-2)) #else -#define THUNK_2_SIZE (SIZEOF_StgHeader + WDS(2)) +#define THUNK_2_SIZE (SIZEOF_StgThunkHeader + WDS(2)) #define TICK_ALLOC_THUNK_2() TICK_ALLOC_UP_THK(WDS(2),0) #endif @@ -198,6 +206,10 @@ atomicModifyMutVarzh_fast HP_CHK_GEN_TICKY(SIZE, R1_PTR & R2_PTR, atomicModifyMutVarzh_fast); +#if defined(THREADED_RTS) + foreign "C" ACQUIRE_LOCK(atomic_modify_mutvar_mutex "ptr") [R1,R2]; +#endif + x = StgMutVar_var(R1); TICK_ALLOC_THUNK_2(); @@ -205,46 +217,31 @@ atomicModifyMutVarzh_fast z = Hp - THUNK_2_SIZE + WDS(1); SET_HDR(z, stg_ap_2_upd_info, W_[CCCS]); LDV_RECORD_CREATE(z); - StgClosure_payload(z,0) = R2; - StgClosure_payload(z,1) = x; + StgThunk_payload(z,0) = R2; + StgThunk_payload(z,1) = x; TICK_ALLOC_THUNK_1(); CCCS_ALLOC(THUNK_1_SIZE); y = z - THUNK_1_SIZE; SET_HDR(y, stg_sel_0_upd_info, W_[CCCS]); LDV_RECORD_CREATE(y); - StgClosure_payload(y,0) = z; + StgThunk_payload(y,0) = z; StgMutVar_var(R1) = y; + foreign "C" dirty_MUT_VAR(BaseReg "ptr", R1 "ptr") [R1]; TICK_ALLOC_THUNK_1(); CCCS_ALLOC(THUNK_1_SIZE); r = y - THUNK_1_SIZE; SET_HDR(r, stg_sel_1_upd_info, W_[CCCS]); LDV_RECORD_CREATE(r); - StgClosure_payload(r,0) = z; - - RET_P(r); -} - -/* ----------------------------------------------------------------------------- - Foreign Object Primitives - -------------------------------------------------------------------------- */ - -mkForeignObjzh_fast -{ - /* R1 = ptr to foreign object, - */ - W_ result; - - ALLOC_PRIM( SIZEOF_StgForeignObj, NO_PTRS, mkForeignObjzh_fast); + StgThunk_payload(r,0) = z; - result = Hp - SIZEOF_StgForeignObj + WDS(1); - SET_HDR(result,stg_FOREIGN_info,W_[CCCS]); - StgForeignObj_data(result) = R1; +#if defined(THREADED_RTS) + foreign "C" RELEASE_LOCK(atomic_modify_mutvar_mutex "ptr") []; +#endif - /* returns (# s#, ForeignObj# #) */ - RET_P(result); + RET_P(r); } /* ----------------------------------------------------------------------------- @@ -277,7 +274,7 @@ mkWeakzh_fast StgWeak_link(w) = W_[weak_ptr_list]; W_[weak_ptr_list] = w; - IF_DEBUG(weak, foreign "C" fprintf(stderr,stg_weak_msg,w)); + IF_DEBUG(weak, foreign "C" debugBelch(stg_weak_msg,w) []); RET_P(w); } @@ -515,6 +512,7 @@ word64ToIntegerzh_fast /* ToDo: this is shockingly inefficient */ +#ifndef THREADED_RTS section "bss" { mp_tmp1: bits8 [SIZEOF_MP_INT]; @@ -526,101 +524,120 @@ section "bss" { } section "bss" { - result1: + mp_result1: bits8 [SIZEOF_MP_INT]; } section "bss" { - result2: + mp_result2: bits8 [SIZEOF_MP_INT]; } +#endif -#define GMP_TAKE2_RET1(name,mp_fun) \ -name \ -{ \ - CInt s1, s2; \ - W_ d1, d2; \ - \ - /* 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" mpz_init(result1); \ - \ - /* Perform the operation */ \ - foreign "C" mp_fun(result1,mp_tmp1,mp_tmp2); \ - \ - RET_NP(TO_W_(MP_INT__mp_size(result1)), \ - MP_INT__mp_d(result1) - SIZEOF_StgArrWords); \ +#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" mpz_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; \ - \ - /* 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" mpz_init(result1); \ - \ - /* Perform the operation */ \ - foreign "C" mp_fun(result1,mp_tmp1); \ - \ - RET_NP(TO_W_(MP_INT__mp_size(result1)), \ - MP_INT__mp_d(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" mpz_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; \ - \ - /* 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" mpz_init(result1); \ - foreign "C" mpz_init(result2); \ - \ - /* Perform the operation */ \ - foreign "C" mp_fun(result1,result2,mp_tmp1,mp_tmp2); \ - \ - RET_NPNP(TO_W_(MP_INT__mp_size(result1)), \ - MP_INT__mp_d(result1) - SIZEOF_StgArrWords, \ - TO_W_(MP_INT__mp_size(result2)), \ - MP_INT__mp_d(result2) - 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" mpz_init(mp_result1 "ptr") []; \ + foreign "C" mpz_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, mpz_add) @@ -638,17 +655,20 @@ GMP_TAKE1_RET1(complementIntegerzh_fast, mpz_com) GMP_TAKE2_RET2(quotRemIntegerzh_fast, mpz_tdiv_qr) GMP_TAKE2_RET2(divModIntegerzh_fast, mpz_fdiv_qr) +#ifndef THREADED_RTS section "bss" { - aa: W_; // NB. aa is really an mp_limb_t + 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_[aa] = R1; - r = foreign "C" mpn_gcd_1(aa, 1, R2); + W_[mp_tmp_w] = R1; + r = foreign "C" mpn_gcd_1(mp_tmp_w "ptr", 1, R2) []; R1 = r; /* Result parked in R1, return via info-pointer at TOS */ @@ -659,7 +679,7 @@ gcdIntzh_fast gcdIntegerIntzh_fast { /* R1 = s1; R2 = d1; R3 = the int */ - R1 = foreign "C" mpn_gcd_1( BYTE_ARR_CTS(R2) "ptr", R1, R3); + R1 = foreign "C" mpn_gcd_1( BYTE_ARR_CTS(R2) "ptr", R1, R3) []; /* Result parked in R1, return via info-pointer at TOS */ jump %ENTRY_CODE(Sp(0)); @@ -740,7 +760,7 @@ cmpIntegerzh_fast up = BYTE_ARR_CTS(R2); vp = BYTE_ARR_CTS(R4); - cmp = foreign "C" mpn_cmp(up "ptr", vp "ptr", size); + cmp = foreign "C" mpn_cmp(up "ptr", vp "ptr", size) []; if (cmp == 0 :: CInt) { R1 = 0; @@ -794,14 +814,12 @@ integer2Wordzh_fast jump %ENTRY_CODE(Sp(0)); } -section "bss" { - exponent: W_; -} - decodeFloatzh_fast { W_ p; F_ arg; + FETCH_MP_TEMP(mp_tmp1); + FETCH_MP_TEMP(mp_tmp_w); /* arguments: F1 = Float# */ arg = F1; @@ -816,10 +834,10 @@ decodeFloatzh_fast MP_INT__mp_d(mp_tmp1) = BYTE_ARR_CTS(p); /* Perform the operation */ - foreign "C" __decodeFloat(mp_tmp1,exponent,arg); + foreign "C" __decodeFloat(mp_tmp1 "ptr",mp_tmp_w "ptr" ,arg) []; /* returns: (Int# (expn), Int#, ByteArray#) */ - RET_NNP(W_[exponent], TO_W_(MP_INT__mp_size(mp_tmp1)), p); + RET_NNP(W_[mp_tmp_w], TO_W_(MP_INT__mp_size(mp_tmp1)), p); } #define DOUBLE_MANTISSA_SIZE SIZEOF_DOUBLE @@ -829,6 +847,8 @@ decodeDoublezh_fast { D_ arg; W_ p; + FETCH_MP_TEMP(mp_tmp1); + FETCH_MP_TEMP(mp_tmp_w); /* arguments: D1 = Double# */ arg = D1; @@ -843,10 +863,10 @@ decodeDoublezh_fast MP_INT__mp_d(mp_tmp1) = BYTE_ARR_CTS(p); /* Perform the operation */ - foreign "C" __decodeDouble(mp_tmp1,exponent,arg); + foreign "C" __decodeDouble(mp_tmp1 "ptr", mp_tmp_w "ptr",arg) []; /* returns: (Int# (expn), Int#, ByteArray#) */ - RET_NNP(W_[exponent], TO_W_(MP_INT__mp_size(mp_tmp1)), p); + RET_NNP(W_[mp_tmp_w], TO_W_(MP_INT__mp_size(mp_tmp1)), p); } /* ----------------------------------------------------------------------------- @@ -860,9 +880,10 @@ forkzh_fast MAYBE_GC(R1_PTR, forkzh_fast); // create it right now, return ThreadID in R1 - "ptr" R1 = foreign "C" createIOThread( RtsFlags_GcFlags_initialStkSize(RtsFlags), - R1 "ptr"); - foreign "C" scheduleThread(R1 "ptr"); + "ptr" R1 = foreign "C" createIOThread( MyCapability() "ptr", + RtsFlags_GcFlags_initialStkSize(RtsFlags), + R1 "ptr") [R1]; + foreign "C" scheduleThread(MyCapability() "ptr", R1 "ptr") [R1]; // switch at the earliest opportunity CInt[context_switch] = 1 :: CInt; @@ -887,7 +908,7 @@ labelThreadzh_fast R1 = ThreadId# R2 = Addr# */ #ifdef DEBUG - foreign "C" labelThread(R1 "ptr", R2 "ptr"); + foreign "C" labelThread(R1 "ptr", R2 "ptr") []; #endif jump %ENTRY_CODE(Sp(0)); } @@ -896,7 +917,7 @@ isCurrentThreadBoundzh_fast { /* no args */ W_ r; - r = foreign "C" isThreadBound(CurrentTSO); + r = foreign "C" isThreadBound(CurrentTSO) []; RET_N(r); } @@ -915,47 +936,17 @@ isCurrentThreadBoundzh_fast // Catch retry frame ------------------------------------------------------------ +#define CATCH_RETRY_FRAME_ERROR(label) \ + label { foreign "C" barf("catch_retry_frame incorrectly entered!"); } -#define CATCH_RETRY_FRAME_ENTRY_TEMPLATE(label,ret) \ - label \ - { \ - W_ r, frame, trec, outer; \ - IF_NOT_REG_R1(W_ rval; rval = Sp(0); Sp_adj(1); ) \ - \ - frame = Sp; \ - trec = StgTSO_trec(CurrentTSO); \ - "ptr" outer = foreign "C" stmGetEnclosingTRec(trec "ptr"); \ - r = foreign "C" stmCommitTransaction(trec "ptr"); \ - if (r) { \ - /* Succeeded (either first branch or second branch) */ \ - StgTSO_trec(CurrentTSO) = outer; \ - Sp = Sp + SIZEOF_StgCatchRetryFrame; \ - IF_NOT_REG_R1(Sp_adj(-1); Sp(0) = rval;) \ - jump ret; \ - } else { \ - /* Did not commit: retry */ \ - W_ new_trec; \ - "ptr" new_trec = foreign "C" stmStartTransaction(outer "ptr"); \ - StgTSO_trec(CurrentTSO) = new_trec; \ - if (StgCatchRetryFrame_running_alt_code(frame)) { \ - R1 = StgCatchRetryFrame_alt_code(frame); \ - } else { \ - R1 = StgCatchRetryFrame_first_code(frame); \ - StgCatchRetryFrame_first_code_trec(frame) = new_trec; \ - } \ - Sp_adj(-1); \ - jump RET_LBL(stg_ap_v); \ - } \ - } - -CATCH_RETRY_FRAME_ENTRY_TEMPLATE(stg_catch_retry_frame_0_ret,%RET_VEC(Sp(SP_OFF),0)) -CATCH_RETRY_FRAME_ENTRY_TEMPLATE(stg_catch_retry_frame_1_ret,%RET_VEC(Sp(SP_OFF),1)) -CATCH_RETRY_FRAME_ENTRY_TEMPLATE(stg_catch_retry_frame_2_ret,%RET_VEC(Sp(SP_OFF),2)) -CATCH_RETRY_FRAME_ENTRY_TEMPLATE(stg_catch_retry_frame_3_ret,%RET_VEC(Sp(SP_OFF),3)) -CATCH_RETRY_FRAME_ENTRY_TEMPLATE(stg_catch_retry_frame_4_ret,%RET_VEC(Sp(SP_OFF),4)) -CATCH_RETRY_FRAME_ENTRY_TEMPLATE(stg_catch_retry_frame_5_ret,%RET_VEC(Sp(SP_OFF),5)) -CATCH_RETRY_FRAME_ENTRY_TEMPLATE(stg_catch_retry_frame_6_ret,%RET_VEC(Sp(SP_OFF),6)) -CATCH_RETRY_FRAME_ENTRY_TEMPLATE(stg_catch_retry_frame_7_ret,%RET_VEC(Sp(SP_OFF),7)) +CATCH_RETRY_FRAME_ERROR(stg_catch_retry_frame_0_ret) +CATCH_RETRY_FRAME_ERROR(stg_catch_retry_frame_1_ret) +CATCH_RETRY_FRAME_ERROR(stg_catch_retry_frame_2_ret) +CATCH_RETRY_FRAME_ERROR(stg_catch_retry_frame_3_ret) +CATCH_RETRY_FRAME_ERROR(stg_catch_retry_frame_4_ret) +CATCH_RETRY_FRAME_ERROR(stg_catch_retry_frame_5_ret) +CATCH_RETRY_FRAME_ERROR(stg_catch_retry_frame_6_ret) +CATCH_RETRY_FRAME_ERROR(stg_catch_retry_frame_7_ret) #if MAX_VECTORED_RTN > 8 #error MAX_VECTORED_RTN has changed: please modify stg_catch_retry_frame too. @@ -980,77 +971,64 @@ INFO_TABLE_RET(stg_catch_retry_frame, stg_catch_retry_frame_5_ret, stg_catch_retry_frame_6_ret, stg_catch_retry_frame_7_ret) -CATCH_RETRY_FRAME_ENTRY_TEMPLATE(,%ENTRY_CODE(Sp(SP_OFF))) - +{ + W_ r, frame, trec, outer; + IF_NOT_REG_R1(W_ rval; rval = Sp(0); Sp_adj(1); ) + + frame = Sp; + trec = StgTSO_trec(CurrentTSO); + "ptr" outer = foreign "C" stmGetEnclosingTRec(trec "ptr") []; + r = foreign "C" stmCommitNestedTransaction(MyCapability() "ptr", trec "ptr") []; + if (r) { + /* Succeeded (either first branch or second branch) */ + StgTSO_trec(CurrentTSO) = outer; + Sp = Sp + SIZEOF_StgCatchRetryFrame; + IF_NOT_REG_R1(Sp_adj(-1); Sp(0) = rval;) + jump %ENTRY_CODE(Sp(SP_OFF)); + } else { + /* Did not commit: retry */ + W_ new_trec; + "ptr" new_trec = foreign "C" stmStartTransaction(MyCapability() "ptr", outer "ptr") []; + StgTSO_trec(CurrentTSO) = new_trec; + if (StgCatchRetryFrame_running_alt_code(frame)) { + R1 = StgCatchRetryFrame_alt_code(frame); + } else { + R1 = StgCatchRetryFrame_first_code(frame); + StgCatchRetryFrame_first_code_trec(frame) = new_trec; + } + jump stg_ap_v_fast; + } +} // Atomically frame ------------------------------------------------------------- -#define ATOMICALLY_FRAME_ENTRY_TEMPLATE(label,ret) \ - label \ - { \ - W_ frame, trec, valid; \ - IF_NOT_REG_R1(W_ rval; rval = Sp(0); Sp_adj(1); ) \ - \ - frame = Sp; \ - trec = StgTSO_trec(CurrentTSO); \ - if (StgAtomicallyFrame_waiting(frame)) { \ - /* The TSO is currently waiting: should we stop waiting? */ \ - valid = foreign "C" stmReWait(CurrentTSO "ptr"); \ - if (valid) { \ - /* Previous attempt is still valid: no point trying again yet */ \ - IF_NOT_REG_R1(Sp_adj(-1); Sp(0) = rval;) \ - jump stg_block_noregs; \ - } else { \ - /* Previous attempt is no longer valid: try again */ \ - "ptr" trec = foreign "C" stmStartTransaction(NO_TREC "ptr"); \ - StgTSO_trec(CurrentTSO) = trec; \ - StgAtomicallyFrame_waiting(frame) = 0 :: CInt; /* false; */ \ - R1 = StgAtomicallyFrame_code(frame); \ - Sp_adj(-1); \ - jump RET_LBL(stg_ap_v); \ - } \ - } else { \ - /* The TSO is not currently waiting: try to commit the transaction */ \ - valid = foreign "C" stmCommitTransaction(trec "ptr"); \ - if (valid) { \ - /* Transaction was valid: commit succeeded */ \ - StgTSO_trec(CurrentTSO) = NO_TREC; \ - Sp = Sp + SIZEOF_StgAtomicallyFrame; \ - IF_NOT_REG_R1(Sp_adj(-1); Sp(0) = rval;) \ - jump ret; \ - } else { \ - /* Transaction was not valid: try again */ \ - "ptr" trec = foreign "C" stmStartTransaction(NO_TREC "ptr"); \ - StgTSO_trec(CurrentTSO) = trec; \ - R1 = StgAtomicallyFrame_code(frame); \ - Sp_adj(-1); \ - jump RET_LBL(stg_ap_v); \ - } \ - } \ - } -ATOMICALLY_FRAME_ENTRY_TEMPLATE(stg_atomically_frame_0_ret,%RET_VEC(Sp(SP_OFF),0)) -ATOMICALLY_FRAME_ENTRY_TEMPLATE(stg_atomically_frame_1_ret,%RET_VEC(Sp(SP_OFF),1)) -ATOMICALLY_FRAME_ENTRY_TEMPLATE(stg_atomically_frame_2_ret,%RET_VEC(Sp(SP_OFF),2)) -ATOMICALLY_FRAME_ENTRY_TEMPLATE(stg_atomically_frame_3_ret,%RET_VEC(Sp(SP_OFF),3)) -ATOMICALLY_FRAME_ENTRY_TEMPLATE(stg_atomically_frame_4_ret,%RET_VEC(Sp(SP_OFF),4)) -ATOMICALLY_FRAME_ENTRY_TEMPLATE(stg_atomically_frame_5_ret,%RET_VEC(Sp(SP_OFF),5)) -ATOMICALLY_FRAME_ENTRY_TEMPLATE(stg_atomically_frame_6_ret,%RET_VEC(Sp(SP_OFF),6)) -ATOMICALLY_FRAME_ENTRY_TEMPLATE(stg_atomically_frame_7_ret,%RET_VEC(Sp(SP_OFF),7)) +#define ATOMICALLY_FRAME_ERROR(label) \ + label { foreign "C" barf("atomically_frame incorrectly entered!"); } + +ATOMICALLY_FRAME_ERROR(stg_atomically_frame_0_ret) +ATOMICALLY_FRAME_ERROR(stg_atomically_frame_1_ret) +ATOMICALLY_FRAME_ERROR(stg_atomically_frame_2_ret) +ATOMICALLY_FRAME_ERROR(stg_atomically_frame_3_ret) +ATOMICALLY_FRAME_ERROR(stg_atomically_frame_4_ret) +ATOMICALLY_FRAME_ERROR(stg_atomically_frame_5_ret) +ATOMICALLY_FRAME_ERROR(stg_atomically_frame_6_ret) +ATOMICALLY_FRAME_ERROR(stg_atomically_frame_7_ret) #if MAX_VECTORED_RTN > 8 #error MAX_VECTORED_RTN has changed: please modify stg_atomically_frame too. #endif #if defined(PROFILING) -#define ATOMICALLY_FRAME_BITMAP 7 -#define ATOMICALLY_FRAME_WORDS 4 +#define ATOMICALLY_FRAME_BITMAP 3 +#define ATOMICALLY_FRAME_WORDS 3 #else -#define ATOMICALLY_FRAME_BITMAP 1 -#define ATOMICALLY_FRAME_WORDS 2 +#define ATOMICALLY_FRAME_BITMAP 0 +#define ATOMICALLY_FRAME_WORDS 1 #endif + INFO_TABLE_RET(stg_atomically_frame, ATOMICALLY_FRAME_WORDS, ATOMICALLY_FRAME_BITMAP, ATOMICALLY_FRAME, @@ -1062,8 +1040,64 @@ INFO_TABLE_RET(stg_atomically_frame, stg_atomically_frame_5_ret, stg_atomically_frame_6_ret, stg_atomically_frame_7_ret) -ATOMICALLY_FRAME_ENTRY_TEMPLATE(,%ENTRY_CODE(Sp(SP_OFF))) +{ + W_ frame, trec, valid; + IF_NOT_REG_R1(W_ rval; rval = Sp(0); Sp_adj(1); ) + + frame = Sp; + trec = StgTSO_trec(CurrentTSO); + /* The TSO is not currently waiting: try to commit the transaction */ + valid = foreign "C" stmCommitTransaction(MyCapability() "ptr", trec "ptr") []; + if (valid) { + /* Transaction was valid: commit succeeded */ + StgTSO_trec(CurrentTSO) = NO_TREC; + Sp = Sp + SIZEOF_StgAtomicallyFrame; + IF_NOT_REG_R1(Sp_adj(-1); Sp(0) = rval;) + jump %ENTRY_CODE(Sp(SP_OFF)); + } else { + /* Transaction was not valid: try again */ + "ptr" trec = foreign "C" stmStartTransaction(MyCapability() "ptr", NO_TREC "ptr") []; + StgTSO_trec(CurrentTSO) = trec; + R1 = StgAtomicallyFrame_code(frame); + jump stg_ap_v_fast; + } +} + +INFO_TABLE_RET(stg_atomically_waiting_frame, + ATOMICALLY_FRAME_WORDS, ATOMICALLY_FRAME_BITMAP, + ATOMICALLY_FRAME, + stg_atomically_frame_0_ret, + stg_atomically_frame_1_ret, + stg_atomically_frame_2_ret, + stg_atomically_frame_3_ret, + stg_atomically_frame_4_ret, + stg_atomically_frame_5_ret, + stg_atomically_frame_6_ret, + stg_atomically_frame_7_ret) +{ + W_ frame, trec, valid; + IF_NOT_REG_R1(W_ rval; rval = Sp(0); Sp_adj(1); ) + + frame = Sp; + + /* The TSO is currently waiting: should we stop waiting? */ + valid = foreign "C" stmReWait(MyCapability() "ptr", CurrentTSO "ptr") []; + if (valid) { + /* Previous attempt is still valid: no point trying again yet */ + IF_NOT_REG_R1(Sp_adj(-2); + Sp(1) = stg_NO_FINALIZER_closure; + Sp(0) = stg_ut_1_0_unreg_info;) + jump stg_block_noregs; + } else { + /* Previous attempt is no longer valid: try again */ + "ptr" trec = foreign "C" stmStartTransaction(MyCapability() "ptr", NO_TREC "ptr") []; + StgTSO_trec(CurrentTSO) = trec; + StgHeader_info(frame) = stg_atomically_frame_info; + R1 = StgAtomicallyFrame_code(frame); + jump stg_ap_v_fast; + } +} // STM catch frame -------------------------------------------------------------- @@ -1130,25 +1164,33 @@ atomicallyzh_fast W_ old_trec; W_ new_trec; + // stmStartTransaction may allocate + MAYBE_GC (R1_PTR, atomicallyzh_fast); + /* Args: R1 = m :: STM a */ STK_CHK_GEN(SIZEOF_StgAtomicallyFrame + WDS(1), R1_PTR, atomicallyzh_fast); + old_trec = StgTSO_trec(CurrentTSO); + + /* Nested transactions are not allowed; raise an exception */ + if (old_trec != NO_TREC) { + R1 = GHCziIOBase_NestedAtomically_closure; + jump raisezh_fast; + } + /* Set up the atomically frame */ Sp = Sp - SIZEOF_StgAtomicallyFrame; frame = Sp; - SET_HDR(frame,stg_atomically_frame_info,CCCS); - StgAtomicallyFrame_waiting(frame) = 0 :: CInt; // False + SET_HDR(frame,stg_atomically_frame_info, W_[CCCS]); StgAtomicallyFrame_code(frame) = R1; /* Start the memory transcation */ - old_trec = StgTSO_trec(CurrentTSO); - "ptr" new_trec = foreign "C" stmStartTransaction(old_trec "ptr"); + "ptr" new_trec = foreign "C" stmStartTransaction(MyCapability() "ptr", old_trec "ptr") [R1]; StgTSO_trec(CurrentTSO) = new_trec; /* Apply R1 to the realworld token */ - Sp_adj(-1); - jump RET_LBL(stg_ap_v); + jump stg_ap_v_fast; } @@ -1164,12 +1206,11 @@ catchSTMzh_fast Sp = Sp - SIZEOF_StgCatchSTMFrame; frame = Sp; - SET_HDR(frame, stg_catch_stm_frame_info, CCCS); + SET_HDR(frame, stg_catch_stm_frame_info, W_[CCCS]); StgCatchSTMFrame_handler(frame) = R2; /* Apply R1 to the realworld token */ - Sp_adj(-1); - jump RET_LBL(stg_ap_v); + jump stg_ap_v_fast; } @@ -1179,28 +1220,30 @@ catchRetryzh_fast W_ new_trec; W_ trec; + // stmStartTransaction may allocate + MAYBE_GC (R1_PTR & R2_PTR, catchRetryzh_fast); + /* Args: R1 :: STM a */ /* Args: R2 :: STM a */ STK_CHK_GEN(SIZEOF_StgCatchRetryFrame + WDS(1), R1_PTR & R2_PTR, catchRetryzh_fast); /* Start a nested transaction within which to run the first code */ trec = StgTSO_trec(CurrentTSO); - "ptr" new_trec = foreign "C" stmStartTransaction(trec "ptr"); + "ptr" new_trec = foreign "C" stmStartTransaction(MyCapability() "ptr", trec "ptr") [R1,R2]; StgTSO_trec(CurrentTSO) = new_trec; /* Set up the catch-retry frame */ Sp = Sp - SIZEOF_StgCatchRetryFrame; frame = Sp; - SET_HDR(frame, stg_catch_retry_frame_info, CCCS); + SET_HDR(frame, stg_catch_retry_frame_info, W_[CCCS]); StgCatchRetryFrame_running_alt_code(frame) = 0 :: CInt; // false; StgCatchRetryFrame_first_code(frame) = R1; StgCatchRetryFrame_alt_code(frame) = R2; StgCatchRetryFrame_first_code_trec(frame) = new_trec; /* Apply R1 to the realworld token */ - Sp_adj(-1); - jump RET_LBL(stg_ap_v); + jump stg_ap_v_fast; } @@ -1217,9 +1260,9 @@ retryzh_fast // Find the enclosing ATOMICALLY_FRAME or CATCH_RETRY_FRAME retry_pop_stack: trec = StgTSO_trec(CurrentTSO); - "ptr" outer = foreign "C" stmGetEnclosingTRec(trec "ptr"); + "ptr" outer = foreign "C" stmGetEnclosingTRec(trec "ptr") []; StgTSO_sp(CurrentTSO) = Sp; - frame_type = foreign "C" findRetryFrameHelper(CurrentTSO "ptr"); + frame_type = foreign "C" findRetryFrameHelper(CurrentTSO "ptr") []; Sp = StgTSO_sp(CurrentTSO); frame = Sp; @@ -1228,19 +1271,20 @@ retry_pop_stack: ASSERT(outer != NO_TREC); if (!StgCatchRetryFrame_running_alt_code(frame)) { // Retry in the first code: try the alternative - "ptr" trec = foreign "C" stmStartTransaction(outer "ptr"); + "ptr" trec = foreign "C" stmStartTransaction(MyCapability() "ptr", outer "ptr") []; StgTSO_trec(CurrentTSO) = trec; StgCatchRetryFrame_running_alt_code(frame) = 1 :: CInt; // true; R1 = StgCatchRetryFrame_alt_code(frame); - Sp_adj(-1); - jump RET_LBL(stg_ap_v); + jump stg_ap_v_fast; } else { // Retry in the alternative code: propagate W_ other_trec; other_trec = StgCatchRetryFrame_first_code_trec(frame); - r = foreign "C" stmMergeForWaiting(trec "ptr", other_trec "ptr"); + r = foreign "C" stmCommitNestedTransaction(MyCapability() "ptr", other_trec "ptr") []; if (r) { - r = foreign "C" stmCommitTransaction(trec "ptr"); + r = foreign "C" stmCommitNestedTransaction(MyCapability() "ptr", trec "ptr") []; + } else { + foreign "C" stmAbortTransaction(MyCapability() "ptr", trec "ptr") []; } if (r) { // Merge between siblings succeeded: commit it back to enclosing transaction @@ -1250,13 +1294,12 @@ retry_pop_stack: goto retry_pop_stack; } else { // Merge failed: we musn't propagate the retry. Try both paths again. - "ptr" trec = foreign "C" stmStartTransaction(outer "ptr"); + "ptr" trec = foreign "C" stmStartTransaction(MyCapability() "ptr", outer "ptr") []; StgCatchRetryFrame_first_code_trec(frame) = trec; StgCatchRetryFrame_running_alt_code(frame) = 0 :: CInt; // false; StgTSO_trec(CurrentTSO) = trec; R1 = StgCatchRetryFrame_first_code(frame); - Sp_adj(-1); - jump RET_LBL(stg_ap_v); + jump stg_ap_v_fast; } } } @@ -1264,20 +1307,24 @@ retry_pop_stack: // We've reached the ATOMICALLY_FRAME: attempt to wait ASSERT(frame_type == ATOMICALLY_FRAME); ASSERT(outer == NO_TREC); - r = foreign "C" stmWait(CurrentTSO "ptr", trec "ptr"); + r = foreign "C" stmWait(MyCapability() "ptr", CurrentTSO "ptr", trec "ptr") []; if (r) { // Transaction was valid: stmWait put us on the TVars' queues, we now block - StgAtomicallyFrame_waiting(frame) = 1 :: CInt; // true + StgHeader_info(frame) = stg_atomically_waiting_frame_info; Sp = frame; - jump stg_block_noregs; + // Fix up the stack in the unregisterised case: the return convention is different. + IF_NOT_REG_R1(Sp_adj(-2); + Sp(1) = stg_NO_FINALIZER_closure; + Sp(0) = stg_ut_1_0_unreg_info;) + R3 = trec; // passing to stmWaitUnblock() + jump stg_block_stmwait; } else { // Transaction was not valid: retry immediately - "ptr" trec = foreign "C" stmStartTransaction(outer "ptr"); + "ptr" trec = foreign "C" stmStartTransaction(MyCapability() "ptr", outer "ptr") []; StgTSO_trec(CurrentTSO) = trec; R1 = StgAtomicallyFrame_code(frame); Sp = frame; - Sp_adj(-1); - jump RET_LBL(stg_ap_v); + jump stg_ap_v_fast; } } @@ -1285,15 +1332,13 @@ retry_pop_stack: newTVarzh_fast { W_ tv; + W_ new_value; /* Args: R1 = initialisation value */ - ALLOC_PRIM( SIZEOF_StgTVar, R1_PTR, newTVarzh_fast); - tv = Hp - SIZEOF_StgTVar + WDS(1); - SET_HDR(tv,stg_TVAR_info,W_[CCCS]); - StgTVar_current_value(tv) = R1; - StgTVar_first_wait_queue_entry(tv) = stg_END_STM_WAIT_QUEUE_closure; - + MAYBE_GC (R1_PTR, newTVarzh_fast); + new_value = R1; + "ptr" tv = foreign "C" stmNewTVar(MyCapability() "ptr", new_value "ptr") []; RET_P(tv); } @@ -1309,7 +1354,7 @@ readTVarzh_fast MAYBE_GC (R1_PTR, readTVarzh_fast); // Call to stmReadTVar may allocate trec = StgTSO_trec(CurrentTSO); tvar = R1; - "ptr" result = foreign "C" stmReadTVar(trec "ptr", tvar "ptr"); + "ptr" result = foreign "C" stmReadTVar(MyCapability() "ptr", trec "ptr", tvar "ptr") []; RET_P(result); } @@ -1328,7 +1373,7 @@ writeTVarzh_fast trec = StgTSO_trec(CurrentTSO); tvar = R1; new_value = R2; - foreign "C" stmWriteTVar(trec "ptr", tvar "ptr", new_value "ptr"); + foreign "C" stmWriteTVar(MyCapability() "ptr", trec "ptr", tvar "ptr", new_value "ptr") []; jump %ENTRY_CODE(Sp(0)); } @@ -1408,7 +1453,6 @@ newMVarzh_fast StgTSO_sp(tso) = StgTSO_sp(tso) + WDS(3); \ lval = W_[StgTSO_sp(tso) - WDS(1)]; - takeMVarzh_fast { W_ mvar, val, info, tso; @@ -1416,7 +1460,11 @@ takeMVarzh_fast /* args: R1 = MVar closure */ mvar = R1; +#if defined(THREADED_RTS) + "ptr" info = foreign "C" lockClosure(mvar "ptr") []; +#else info = GET_INFO(mvar); +#endif /* If the MVar is empty, put ourselves on its blocking queue, * and wait until we're woken up. @@ -1451,26 +1499,34 @@ takeMVarzh_fast #if defined(GRAN) || defined(PAR) /* ToDo: check 2nd arg (mvar) is right */ - "ptr" tso = foreign "C" unblockOne(StgMVar_head(mvar),mvar); + "ptr" tso = foreign "C" unblockOne(StgMVar_head(mvar),mvar) []; StgMVar_head(mvar) = tso; #else - "ptr" tso = foreign "C" unblockOne(StgMVar_head(mvar) "ptr"); + "ptr" tso = foreign "C" unblockOne(MyCapability() "ptr", + StgMVar_head(mvar) "ptr") []; StgMVar_head(mvar) = tso; #endif + if (StgMVar_head(mvar) == stg_END_TSO_QUEUE_closure) { StgMVar_tail(mvar) = stg_END_TSO_QUEUE_closure; } + +#if defined(THREADED_RTS) + foreign "C" unlockClosure(mvar "ptr", stg_FULL_MVAR_info) []; +#endif RET_P(val); } else { /* No further putMVars, MVar is now empty */ - - /* do this last... we might have locked the MVar in the SMP case, - * and writing the info pointer will unlock it. - */ - SET_INFO(mvar,stg_EMPTY_MVAR_info); StgMVar_value(mvar) = stg_END_TSO_QUEUE_closure; + +#if defined(THREADED_RTS) + foreign "C" unlockClosure(mvar "ptr", stg_EMPTY_MVAR_info) []; +#else + SET_INFO(mvar,stg_EMPTY_MVAR_info); +#endif + RET_P(val); } } @@ -1484,9 +1540,16 @@ tryTakeMVarzh_fast mvar = R1; +#if defined(THREADED_RTS) + "ptr" info = foreign "C" lockClosure(mvar "ptr") []; +#else info = GET_INFO(mvar); +#endif if (info == stg_EMPTY_MVAR_info) { +#if defined(THREADED_RTS) + foreign "C" unlockClosure(mvar "ptr", stg_EMPTY_MVAR_info) []; +#endif /* HACK: we need a pointer to pass back, * so we abuse NO_FINALIZER_closure */ @@ -1497,6 +1560,7 @@ tryTakeMVarzh_fast val = StgMVar_value(mvar); if (StgMVar_head(mvar) != stg_END_TSO_QUEUE_closure) { + /* There are putMVar(s) waiting... * wake up the first thread on the queue */ @@ -1508,26 +1572,30 @@ tryTakeMVarzh_fast #if defined(GRAN) || defined(PAR) /* ToDo: check 2nd arg (mvar) is right */ - "ptr" tso = foreign "C" unblockOne(StgMVar_head(mvar) "ptr", mvar "ptr"); + "ptr" tso = foreign "C" unblockOne(StgMVar_head(mvar) "ptr", mvar "ptr") []; StgMVar_head(mvar) = tso; #else - "ptr" tso = foreign "C" unblockOne(StgMVar_head(mvar) "ptr"); + "ptr" tso = foreign "C" unblockOne(MyCapability() "ptr", + StgMVar_head(mvar) "ptr") []; StgMVar_head(mvar) = tso; #endif if (StgMVar_head(mvar) == stg_END_TSO_QUEUE_closure) { StgMVar_tail(mvar) = stg_END_TSO_QUEUE_closure; } +#if defined(THREADED_RTS) + foreign "C" unlockClosure(mvar "ptr", stg_FULL_MVAR_info) []; +#endif } else { /* No further putMVars, MVar is now empty */ StgMVar_value(mvar) = stg_END_TSO_QUEUE_closure; - - /* do this last... we might have locked the MVar in the SMP case, - * and writing the info pointer will unlock it. - */ +#if defined(THREADED_RTS) + foreign "C" unlockClosure(mvar "ptr", stg_EMPTY_MVAR_info) []; +#else SET_INFO(mvar,stg_EMPTY_MVAR_info); +#endif } RET_NP(1, val); @@ -1541,7 +1609,11 @@ putMVarzh_fast /* args: R1 = MVar, R2 = value */ mvar = R1; +#if defined(THREADED_RTS) + "ptr" info = foreign "C" lockClosure(mvar "ptr") [R2]; +#else info = GET_INFO(mvar); +#endif if (info == stg_FULL_MVAR_info) { if (StgMVar_head(mvar) == stg_END_TSO_QUEUE_closure) { @@ -1558,6 +1630,7 @@ putMVarzh_fast } if (StgMVar_head(mvar) != stg_END_TSO_QUEUE_closure) { + /* There are takeMVar(s) waiting: wake up the first one */ ASSERT(StgTSO_why_blocked(StgMVar_head(mvar)) == BlockedOnMVar::I16); @@ -1568,10 +1641,10 @@ putMVarzh_fast #if defined(GRAN) || defined(PAR) /* ToDo: check 2nd arg (mvar) is right */ - "ptr" tso = foreign "C" unblockOne(StgMVar_head(mvar) "ptr",mvar "ptr"); + "ptr" tso = foreign "C" unblockOne(MyCapability() "ptr", StgMVar_head(mvar) "ptr",mvar "ptr") []; StgMVar_head(mvar) = tso; #else - "ptr" tso = foreign "C" unblockOne(StgMVar_head(mvar) "ptr"); + "ptr" tso = foreign "C" unblockOne(MyCapability() "ptr", StgMVar_head(mvar) "ptr") []; StgMVar_head(mvar) = tso; #endif @@ -1579,14 +1652,21 @@ putMVarzh_fast StgMVar_tail(mvar) = stg_END_TSO_QUEUE_closure; } +#if defined(THREADED_RTS) + foreign "C" unlockClosure(mvar "ptr", stg_EMPTY_MVAR_info) []; +#endif jump %ENTRY_CODE(Sp(0)); } else { /* No further takes, the MVar is now full. */ StgMVar_value(mvar) = R2; - /* unlocks the MVar in the SMP case */ + +#if defined(THREADED_RTS) + foreign "C" unlockClosure(mvar "ptr", stg_FULL_MVAR_info) []; +#else SET_INFO(mvar,stg_FULL_MVAR_info); +#endif jump %ENTRY_CODE(Sp(0)); } @@ -1601,13 +1681,21 @@ tryPutMVarzh_fast /* args: R1 = MVar, R2 = value */ mvar = R1; +#if defined(THREADED_RTS) + "ptr" info = foreign "C" lockClosure(mvar "ptr") [R2]; +#else info = GET_INFO(mvar); +#endif if (info == stg_FULL_MVAR_info) { +#if defined(THREADED_RTS) + foreign "C" unlockClosure(mvar "ptr", stg_FULL_MVAR_info) []; +#endif RET_N(0); } if (StgMVar_head(mvar) != stg_END_TSO_QUEUE_closure) { + /* There are takeMVar(s) waiting: wake up the first one */ ASSERT(StgTSO_why_blocked(StgMVar_head(mvar)) == BlockedOnMVar::I16); @@ -1618,10 +1706,10 @@ tryPutMVarzh_fast #if defined(GRAN) || defined(PAR) /* ToDo: check 2nd arg (mvar) is right */ - "ptr" tso = foreign "C" unblockOne(StgMVar_head(mvar) "ptr",mvar "ptr"); + "ptr" tso = foreign "C" unblockOne(MyCapability() "ptr", StgMVar_head(mvar) "ptr",mvar "ptr") []; StgMVar_head(mvar) = tso; #else - "ptr" tso = foreign "C" unblockOne(StgMVar_head(mvar) "ptr"); + "ptr" tso = foreign "C" unblockOne(MyCapability() "ptr", StgMVar_head(mvar) "ptr") []; StgMVar_head(mvar) = tso; #endif @@ -1629,17 +1717,23 @@ tryPutMVarzh_fast StgMVar_tail(mvar) = stg_END_TSO_QUEUE_closure; } - jump %ENTRY_CODE(Sp(0)); +#if defined(THREADED_RTS) + foreign "C" unlockClosure(mvar "ptr", stg_EMPTY_MVAR_info) []; +#endif } else { /* No further takes, the MVar is now full. */ StgMVar_value(mvar) = R2; - /* unlocks the MVar in the SMP case */ + +#if defined(THREADED_RTS) + foreign "C" unlockClosure(mvar "ptr", stg_FULL_MVAR_info) []; +#else SET_INFO(mvar,stg_FULL_MVAR_info); - jump %ENTRY_CODE(Sp(0)); +#endif } + RET_N(1); /* ToDo: yield afterward for better communication performance? */ } @@ -1654,18 +1748,18 @@ makeStableNamezh_fast ALLOC_PRIM( SIZEOF_StgStableName, R1_PTR, makeStableNamezh_fast ); - index = foreign "C" lookupStableName(R1 "ptr"); + index = foreign "C" lookupStableName(R1 "ptr") []; /* Is there already a StableName for this heap object? - * stable_ptr_table is an array of snEntry structs. + * stable_ptr_table is a pointer to an array of snEntry structs. */ - if ( snEntry_sn_obj(stable_ptr_table + index*SIZEOF_snEntry) == NULL ) { + if ( snEntry_sn_obj(W_[stable_ptr_table] + index*SIZEOF_snEntry) == NULL ) { sn_obj = Hp - SIZEOF_StgStableName + WDS(1); SET_HDR(sn_obj, stg_STABLE_NAME_info, W_[CCCS]); StgStableName_sn(sn_obj) = index; - snEntry_sn_obj(stable_ptr_table + index*SIZEOF_snEntry) = sn_obj; + snEntry_sn_obj(W_[stable_ptr_table] + index*SIZEOF_snEntry) = sn_obj; } else { - sn_obj = snEntry_sn_obj(stable_ptr_table + index*SIZEOF_snEntry); + sn_obj = snEntry_sn_obj(W_[stable_ptr_table] + index*SIZEOF_snEntry); } RET_P(sn_obj); @@ -1677,7 +1771,7 @@ makeStablePtrzh_fast /* Args: R1 = a */ W_ sp; MAYBE_GC(R1_PTR, makeStablePtrzh_fast); - "ptr" sp = foreign "C" getStablePtr(R1 "ptr"); + "ptr" sp = foreign "C" getStablePtr(R1 "ptr") []; RET_N(sp); } @@ -1686,7 +1780,7 @@ deRefStablePtrzh_fast /* Args: R1 = the stable ptr */ W_ r, sp; sp = R1; - r = snEntry_addr(stable_ptr_table + sp*SIZEOF_snEntry); + r = snEntry_addr(W_[stable_ptr_table] + sp*SIZEOF_snEntry); RET_P(r); } @@ -1781,7 +1875,7 @@ waitReadzh_fast /* args: R1 */ #ifdef THREADED_RTS foreign "C" barf("waitRead# on threaded RTS"); -#endif +#else ASSERT(StgTSO_why_blocked(CurrentTSO) == NotBlocked::I16); StgTSO_why_blocked(CurrentTSO) = BlockedOnRead::I16; @@ -1790,6 +1884,7 @@ waitReadzh_fast // threaded RTS anyway. APPEND_TO_BLOCKED_QUEUE(CurrentTSO); jump stg_block_noregs; +#endif } waitWritezh_fast @@ -1797,7 +1892,7 @@ waitWritezh_fast /* args: R1 */ #ifdef THREADED_RTS foreign "C" barf("waitWrite# on threaded RTS"); -#endif +#else ASSERT(StgTSO_why_blocked(CurrentTSO) == NotBlocked::I16); StgTSO_why_blocked(CurrentTSO) = BlockedOnWrite::I16; @@ -1806,13 +1901,14 @@ waitWritezh_fast // threaded RTS anyway. APPEND_TO_BLOCKED_QUEUE(CurrentTSO); jump stg_block_noregs; +#endif } STRING(stg_delayzh_malloc_str, "delayzh_fast") delayzh_fast { -#ifdef mingw32_TARGET_OS +#ifdef mingw32_HOST_OS W_ ares; CInt reqID; #else @@ -1821,13 +1917,13 @@ delayzh_fast #ifdef THREADED_RTS foreign "C" barf("delay# on threaded RTS"); -#endif +#else /* args: R1 (microsecond delay amount) */ ASSERT(StgTSO_why_blocked(CurrentTSO) == NotBlocked::I16); StgTSO_why_blocked(CurrentTSO) = BlockedOnDelay::I16; -#ifdef mingw32_TARGET_OS +#ifdef mingw32_HOST_OS /* could probably allocate this on the heap instead */ "ptr" ares = foreign "C" stgMallocBytes(SIZEOF_StgAsyncIOResult, @@ -1871,10 +1967,11 @@ while: } jump stg_block_noregs; #endif +#endif /* !THREADED_RTS */ } -#ifdef mingw32_TARGET_OS +#ifdef mingw32_HOST_OS STRING(stg_asyncReadzh_malloc_str, "asyncReadzh_fast") asyncReadzh_fast { @@ -1883,7 +1980,7 @@ asyncReadzh_fast #ifdef THREADED_RTS foreign "C" barf("asyncRead# on threaded RTS"); -#endif +#else /* args: R1 = fd, R2 = isSock, R3 = len, R4 = buf */ ASSERT(StgTSO_why_blocked(CurrentTSO) == NotBlocked::I16); @@ -1891,14 +1988,16 @@ asyncReadzh_fast /* could probably allocate this on the heap instead */ "ptr" ares = foreign "C" stgMallocBytes(SIZEOF_StgAsyncIOResult, - stg_asyncReadzh_malloc_str); - reqID = foreign "C" addIORequest(R1, 0/*FALSE*/,R2,R3,R4 "ptr"); + stg_asyncReadzh_malloc_str) + [R1,R2,R3,R4]; + reqID = foreign "C" addIORequest(R1, 0/*FALSE*/,R2,R3,R4 "ptr") []; StgAsyncIOResult_reqID(ares) = reqID; StgAsyncIOResult_len(ares) = 0; StgAsyncIOResult_errCode(ares) = 0; StgTSO_block_info(CurrentTSO) = ares; APPEND_TO_BLOCKED_QUEUE(CurrentTSO); jump stg_block_async; +#endif } STRING(stg_asyncWritezh_malloc_str, "asyncWritezh_fast") @@ -1909,15 +2008,16 @@ asyncWritezh_fast #ifdef THREADED_RTS foreign "C" barf("asyncWrite# on threaded RTS"); -#endif +#else /* args: R1 = fd, R2 = isSock, R3 = len, R4 = buf */ ASSERT(StgTSO_why_blocked(CurrentTSO) == NotBlocked::I16); StgTSO_why_blocked(CurrentTSO) = BlockedOnWrite::I16; "ptr" ares = foreign "C" stgMallocBytes(SIZEOF_StgAsyncIOResult, - stg_asyncWritezh_malloc_str); - reqID = foreign "C" addIORequest(R1, 1/*TRUE*/,R2,R3,R4 "ptr"); + stg_asyncWritezh_malloc_str) + [R1,R2,R3,R4]; + reqID = foreign "C" addIORequest(R1, 1/*TRUE*/,R2,R3,R4 "ptr") []; StgAsyncIOResult_reqID(ares) = reqID; StgAsyncIOResult_len(ares) = 0; @@ -1925,6 +2025,7 @@ asyncWritezh_fast StgTSO_block_info(CurrentTSO) = ares; APPEND_TO_BLOCKED_QUEUE(CurrentTSO); jump stg_block_async; +#endif } STRING(stg_asyncDoProczh_malloc_str, "asyncDoProczh_fast") @@ -1933,20 +2034,26 @@ asyncDoProczh_fast W_ ares; CInt reqID; +#ifdef THREADED_RTS + foreign "C" barf("asyncDoProc# on threaded RTS"); +#else + /* args: R1 = proc, R2 = param */ ASSERT(StgTSO_why_blocked(CurrentTSO) == NotBlocked::I16); StgTSO_why_blocked(CurrentTSO) = BlockedOnDoProc::I16; /* could probably allocate this on the heap instead */ "ptr" ares = foreign "C" stgMallocBytes(SIZEOF_StgAsyncIOResult, - stg_asyncDoProczh_malloc_str); - reqID = foreign "C" addDoProcRequest(R1 "ptr",R2 "ptr"); + stg_asyncDoProczh_malloc_str) + [R1,R2]; + reqID = foreign "C" addDoProcRequest(R1 "ptr",R2 "ptr") []; StgAsyncIOResult_reqID(ares) = reqID; StgAsyncIOResult_len(ares) = 0; StgAsyncIOResult_errCode(ares) = 0; StgTSO_block_info(CurrentTSO) = ares; APPEND_TO_BLOCKED_QUEUE(CurrentTSO); jump stg_block_async; +#endif } #endif