diff options
Diffstat (limited to 'erts/emulator/beam/ops.tab')
| -rw-r--r-- | erts/emulator/beam/ops.tab | 580 |
1 files changed, 369 insertions, 211 deletions
diff --git a/erts/emulator/beam/ops.tab b/erts/emulator/beam/ops.tab index ce1df74f03..9b168889dd 100644 --- a/erts/emulator/beam/ops.tab +++ b/erts/emulator/beam/ops.tab @@ -1,19 +1,19 @@ # # %CopyrightBegin% -# -# Copyright Ericsson AB 1997-2009. All Rights Reserved. -# +# +# Copyright Ericsson AB 1997-2012. All Rights Reserved. +# # The contents of this file are subject to the Erlang Public License, # Version 1.1, (the "License"); you may not use this file except in # compliance with the License. You should have received a copy of the # Erlang Public License along with this software. If not, it can be # retrieved online at http://www.erlang.org/. -# +# # Software distributed under the License is distributed on an "AS IS" # basis, WITHOUT WARRANTY OF ANY KIND, either express or implied. See # the License for the specific language governing rights and limitations # under the License. -# +# # %CopyrightEnd% # @@ -25,30 +25,12 @@ # instruction transformations; thus, they never occur in BEAM files. # -# Special instruction used to generate an error message when -# trying to load a module compiled by the V1 compiler (R5 & R6). -# (Specially treated in beam_load.c.) +# The too_old_compiler/0 instruction is specially handled in beam_load.c +# to produce a user-friendly message informing the user that the module +# needs to be re-compiled with a modern compiler. too_old_compiler/0 -too_old_compiler - -# -# Obsolete instruction usage follow. (Nowdays we use f with -# a zero label instead of p.) -# - -is_list p S => too_old_compiler -is_nonempty_list p R => too_old_compiler -is_nil p R => too_old_compiler - -is_tuple p S => too_old_compiler -test_arity p S Arity => too_old_compiler - -is_integer p R => too_old_compiler -is_float p R => too_old_compiler -is_atom p R => too_old_compiler - -is_eq_exact p S1 S2 => too_old_compiler +too_old_compiler | never() => # In R9C and earlier, the loader used to insert special instructions inside # the module_info/0,1 functions. (In R10B and later, the compiler inserts @@ -60,12 +42,17 @@ func_info M=a a==am_module_info A=u==0 | label L | move n r => too_old_compiler func_info M=a a==am_module_info A=u==1 | label L | move n r => too_old_compiler # The undocumented and unsupported guard BIF is_constant/1 was removed -# in R13. The is_constant/2 operation is marked as obosolete in genop.tab, +# in R13. The is_constant/2 operation is marked as obsolete in genop.tab, # so the loader will automatically generate a too_old_compiler message # it is used, but we need to handle the is_constant/1 BIF specially here. bif1 Fail u$func:erlang:is_constant/1 Src Dst => too_old_compiler +# Since the constant pool was introduced in R12B, empty tuples ({}) +# are literals. Therefore we no longer need to allow put_tuple/2 +# with a tuple size of zero. + +put_tuple u==0 d => too_old_compiler # # All the other instructions. @@ -79,14 +66,46 @@ i_trace_breakpoint i_mtrace_breakpoint i_debug_breakpoint i_count_breakpoint +i_time_breakpoint +i_return_time_trace i_return_to_trace i_yield -i_global_cons -i_global_tuple -i_global_copy return +# +# To ensure that a "move Src x(0)" instruction can be combined +# with the following call instruction, we need to make sure that +# there is no line/1 instruction between the move and the call. +# + +move S r | line Loc | call_ext Ar Func => \ + line Loc | move S r | call_ext Ar Func +move S r | line Loc | call_ext_last Ar Func=u$is_bif D => \ + line Loc | move S r | call_ext_last Ar Func D +move S r | line Loc | call_ext_only Ar Func=u$is_bif => \ + line Loc | move S r | call_ext_only Ar Func +move S r | line Loc | call Ar Func => \ + line Loc | move S r | call Ar Func + +# +# A tail-recursive call to an external function (non-BIF) will +# never be saved on the stack, so there is no reason to keep +# the line instruction. (The compiler did not remove the line +# instruction because it cannot tell the difference between +# BIFs and ordinary Erlang functions.) +# + +line Loc | call_ext_last Ar Func=u$is_not_bif D => \ + call_ext_last Ar Func D +line Loc | call_ext_only Ar Func=u$is_not_bif => \ + call_ext_only Ar Func + +line Loc | func_info M F A => func_info M F A | line Loc + +line I + + %macro: allocate Allocate -pack %macro: allocate_zero AllocateZero -pack %macro: allocate_heap AllocateHeap -pack @@ -94,16 +113,16 @@ return %macro: test_heap TestHeap -pack allocate t t -allocate_heap I I I +allocate_heap t I t deallocate I init y allocate_zero t t -allocate_heap_zero I I I +allocate_heap_zero t I t trim N Remaining => i_trim N i_trim I -test_heap I I +test_heap I t allocate_heap S u==0 R => allocate S R allocate_heap_zero S u==0 R => allocate_zero S R @@ -115,15 +134,9 @@ init Y1 | init Y2 => init2 Y1 Y2 %macro: init2 Init2 -pack %macro: init3 Init3 -pack -# -# Warning: The put_string instruction is specially treated in the loader. -# Don't change the instruction format unless you change the loader too. -# -put_string I I d - # Selecting values -select_val S=q Fail=f Size=u Rest=* => const_select_val(S, Fail, Size, Rest) +select_val S=aiq Fail=f Size=u Rest=* => const_select_val(S, Fail, Size, Rest) select_val S=s Fail=f Size=u Rest=* | use_jump_tab(Size, Rest) => \ gen_jump_tab(S, Fail, Size, Rest) @@ -131,34 +144,59 @@ select_val S=s Fail=f Size=u Rest=* | use_jump_tab(Size, Rest) => \ is_integer Fail=f S | select_val S=s Fail=f Size=u Rest=* | use_jump_tab(Size, Rest) => \ gen_jump_tab(S, Fail, Size, Rest) +is_integer TypeFail=f S | select_val S=s Fail=f Size=u Rest=* | \ + mixed_types(Size, Rest) => \ + gen_split_values(S, TypeFail, Fail, Size, Rest) + select_val S=s Fail=f Size=u Rest=* | mixed_types(Size, Rest) => \ - gen_split_values(S, Fail, Size, Rest) + gen_split_values(S, Fail, Fail, Size, Rest) -is_integer Fail=f S | select_val S=s Fail=f Size=u Rest=* | \ +is_integer Fail=f S | select_val S=d Fail=f Size=u Rest=* | \ fixed_size_values(Size, Rest) => gen_select_val(S, Fail, Size, Rest) -is_atom Fail=f S | select_val S=s Fail=f Size=u Rest=* | \ +is_atom Fail=f S | select_val S=d Fail=f Size=u Rest=* | \ fixed_size_values(Size, Rest) => gen_select_val(S, Fail, Size, Rest) -select_val S=s Fail=f Size=u Rest=* | fixed_size_values(Size, Rest) => \ - gen_select_val(S, Fail, Size, Rest) +select_val S=s Fail=f Size=u Rest=* | floats_or_bignums(Size, Rest) => \ + gen_select_literals(S, Fail, Size, Rest) -select_val S=s Fail=f Size=u Rest=* | all_values_are_big(Size, Rest) => \ - gen_select_big(S, Fail, Size, Rest) +select_val S=d Fail=f Size=u Rest=* | fixed_size_values(Size, Rest) => \ + gen_select_val(S, Fail, Size, Rest) -is_tuple Fail=f S | select_tuple_arity S=s Fail=f Size=u Rest=* => \ +is_tuple Fail=f S | select_tuple_arity S=d Fail=f Size=u Rest=* => \ gen_select_tuple_arity(S, Fail, Size, Rest) -select_tuple_arity S=s Fail=f Size=u Rest=* => \ +select_tuple_arity S=d Fail=f Size=u Rest=* => \ gen_select_tuple_arity(S, Fail, Size, Rest) -i_select_val s f I -i_select_tuple_arity s f I -i_select_big s f -i_select_float s f I +i_select_val r f I +i_select_val x f I +i_select_val y f I + +i_select_val2 r f c f c f +i_select_val2 x f c f c f +i_select_val2 y f c f c f + +i_select_tuple_arity2 r f A f A f +i_select_tuple_arity2 x f A f A f +i_select_tuple_arity2 y f A f A f + +i_select_tuple_arity r f I +i_select_tuple_arity x f I +i_select_tuple_arity y f I -i_jump_on_val_zero s f I -i_jump_on_val s f I I +i_jump_on_val_zero r f I +i_jump_on_val_zero x f I +i_jump_on_val_zero y f I + +i_jump_on_val r f I I +i_jump_on_val x f I I +i_jump_on_val y f I I + +jump Target | label Lbl | same_label(Target, Lbl) => label Lbl + +is_ne_exact L1 S1 S2 | jump Fail | label L2 | same_label(L1, L2) => \ + is_eq_exact Fail S1 S2 | label L2 %macro: get_list GetList -pack get_list x x x @@ -233,11 +271,17 @@ is_number Fail Literal=q => move Literal x | is_number Fail x jump f -case_end Literal=q => move Literal x | case_end x -badmatch Literal=q => move Literal x | badmatch x +case_end Literal=cq => move Literal x | case_end x +badmatch Literal=cq => move Literal x | badmatch x + +case_end r +case_end x +case_end y + +badmatch r +badmatch x +badmatch y -case_end s -badmatch s if_end raise s s @@ -245,14 +289,33 @@ raise s s badarg j system_limit j -move R R => +move C=cxy r | jump Lbl => move_jump Lbl C + +%macro: move_jump MoveJump -nonext +move_jump f n +move_jump f c +move_jump f x +move_jump f y move X1=x Y1=y | move X2=x Y2=y => move2 X1 Y1 X2 Y2 move Y1=y X1=x | move Y2=y X2=x => move2 Y1 X1 Y2 X2 +move X1=x X2=x | move X3=x X4=x => move2 X1 X2 X3 X4 + +move C=aiq X=x==1 => move_x1 C +move C=aiq X=x==2 => move_x2 C + +move_x1 c +move_x2 c %macro: move2 Move2 -pack move2 x y x y move2 y x y x +move2 x x x x + +# The compiler almost never generates a "move Literal y(Y)" instruction, +# so let's cheat if we encounter one. +move S=n D=y => init D +move S=c D=y => move S x | move x D %macro:move Move -pack -gen_dest move x x @@ -264,15 +327,10 @@ move r x move r y move c r move c x -move c y move n x move n r move y y -%cold -move s d -%hot - # Receive operations. loop_rec Fail Src | smp_mark_target_label(Fail) => i_loop_rec Fail Src @@ -305,58 +363,78 @@ i_wait_error_locked send # -# Comparisions. +# Optimized comparisons with one immediate/literal operand. +# + +is_eq_exact Lbl R=rxy C=ian => i_is_eq_exact_immed Lbl R C +is_eq_exact Lbl R=rxy C=q => i_is_eq_exact_literal R Lbl C + +is_ne_exact Lbl R=rxy C=ian => i_is_ne_exact_immed Lbl R C +is_ne_exact Lbl R=rxy C=q => i_is_ne_exact_literal R Lbl C + +%macro: i_is_eq_exact_immed EqualImmed -fail_action +i_is_eq_exact_immed f r c +i_is_eq_exact_immed f x c +i_is_eq_exact_immed f y c + +i_is_eq_exact_literal r f c +i_is_eq_exact_literal x f c +i_is_eq_exact_literal y f c + +%macro: i_is_ne_exact_immed NotEqualImmed -fail_action +i_is_ne_exact_immed f r c +i_is_ne_exact_immed f x c +i_is_ne_exact_immed f y c + +i_is_ne_exact_literal r f c +i_is_ne_exact_literal x f c +i_is_ne_exact_literal y f c + +# +# All other comparisons. # -is_eq_exact Lbl=f R=rxy C=ian => i_is_eq_immed Lbl R C -is_eq Lbl=f R=rxy C=an => i_is_eq_immed Lbl R C +is_eq_exact Lbl S1 S2 => i_fetch S1 S2 | i_is_eq_exact Lbl +is_ne_exact Lbl S1 S2 => i_fetch S1 S2 | i_is_ne_exact Lbl is_ge Lbl S1 S2 => i_fetch S1 S2 | i_is_ge Lbl is_lt Lbl S1 S2 => i_fetch S1 S2 | i_is_lt Lbl is_eq Lbl S1 S2 => i_fetch S1 S2 | i_is_eq Lbl is_ne Lbl S1 S2 => i_fetch S1 S2 | i_is_ne Lbl -is_eq_exact Lbl=f S1 S2 => i_fetch S1 S2 | i_is_eq_exact Lbl -is_ne_exact Lbl S1 S2 => i_fetch S1 S2 | i_is_ne_exact Lbl - +i_is_eq_exact f +i_is_ne_exact f i_is_lt f i_is_ge f i_is_eq f i_is_ne f -i_is_eq_exact f -i_is_ne_exact f - -%macro: i_is_eq_immed EqualImmed -fail_action -i_is_eq_immed f r c -i_is_eq_immed f x c -i_is_eq_immed f y c # # Putting things. # -put_tuple u==0 Dst => i_put_tuple_only u Dst -put_tuple Arity Dst | put V => i_put_tuple Arity V Dst +put_tuple Arity Dst => i_put_tuple Dst u -i_put_tuple_only A d +i_put_tuple Dst Arity Puts=* | put S1 | put S2 | \ + put S3 | put S4 | put S5 => \ + tuple_append_put5(Arity, Dst, Puts, S1, S2, S3, S4, S5) -%macro: i_put_tuple PutTuple -pack -i_put_tuple A x x -i_put_tuple A y x -i_put_tuple A r x -i_put_tuple A n x -i_put_tuple A c x -i_put_tuple A x y -i_put_tuple A x r -i_put_tuple A y r -i_put_tuple A n r -i_put_tuple A c r +i_put_tuple Dst Arity Puts=* | put S => \ + tuple_append_put(Arity, Dst, Puts, S) -%cold -i_put_tuple A r y -i_put_tuple A y y -i_put_tuple A c y -%hot +i_put_tuple/2 + +%macro:i_put_tuple PutTuple -pack -goto:do_put_tuple +i_put_tuple r I +i_put_tuple x I +i_put_tuple y I + +# +# The instruction "put_list Const [] Dst" will not be generated by +# the current BEAM compiler. But until R15A, play it safe by handling +# that instruction with the following transformation. +# +put_list Const=c n Dst => move Const x | put_list x n Dst %macro:put_list PutList -pack -gen_dest @@ -364,10 +442,8 @@ put_list x n x put_list y n x put_list x x x put_list y x x -put_list c n x put_list x x r put_list y r r -put_list c n r put_list y y x put_list x y x @@ -378,6 +454,13 @@ put_list y y r put_list y r x put_list r n x +put_list x r x +put_list x y r +put_list y x r +put_list y x x + +put_list x r r + # put_list SrcReg Constant Dst put_list r c r put_list r c x @@ -405,17 +488,9 @@ put_list c y x put_list c y y %cold -put_list x r r put_list s s d %hot -%macro: put Put -put x -put r -put y -put c -put n - %macro: i_fetch FetchArgs -pack i_fetch c c i_fetch c r @@ -466,19 +541,20 @@ move_return n r move S r | deallocate D | return => move_deallocate_return S r D -%macro: move_deallocate_return MoveDeallocateReturn -nonext -move_deallocate_return x r P -move_deallocate_return y r P -move_deallocate_return c r P -move_deallocate_return n r P +%macro: move_deallocate_return MoveDeallocateReturn -pack -nonext +move_deallocate_return x r Q +move_deallocate_return y r Q +move_deallocate_return c r Q +move_deallocate_return n r Q deallocate D | return => deallocate_return D %macro: deallocate_return DeallocateReturn -nonext -deallocate_return P +deallocate_return Q test_heap Need u==1 | put_list Y=y r r => test_heap_1_put_list Need Y +%macro: test_heap_1_put_list TestHeapPutList -pack test_heap_1_put_list I y # Test tuple & arity (head) @@ -519,8 +595,6 @@ get_tuple_element Reg P Dst => i_get_tuple_element Reg P Dst | original_reg Reg original_reg Reg Pos => -get_tuple_element Reg P Dst => i_get_tuple_element Reg P Dst - original_reg/2 extract_next_element D1=xy | original_reg Reg P1 | get_tuple_element Reg P2 D2=xy | \ @@ -578,14 +652,14 @@ is_list f y is_nonempty_list Fail=f S=rx | allocate Need Rs => is_nonempty_list_allocate Fail S Need Rs -%macro:is_nonempty_list_allocate IsNonemptyListAllocate -fail_action -is_nonempty_list_allocate f x I I -is_nonempty_list_allocate f r I I +%macro:is_nonempty_list_allocate IsNonemptyListAllocate -fail_action -pack +is_nonempty_list_allocate f x I t +is_nonempty_list_allocate f r I t is_nonempty_list F=f r | test_heap I1 I2 => is_non_empty_list_test_heap F r I1 I2 -%macro: is_non_empty_list_test_heap IsNonemptyListTestHeap -fail_action -is_non_empty_list_test_heap f r I I +%macro: is_non_empty_list_test_heap IsNonemptyListTestHeap -fail_action -pack +is_non_empty_list_test_heap f r I t %macro: is_nonempty_list IsNonemptyList -fail_action is_nonempty_list f x @@ -771,11 +845,11 @@ call_ext_only u==3 u$func:erlang:apply/3 => i_apply_only # thus there is no need to generate any return instruction. # -call_ext_last u==1 Bif=u$bif:erlang:exit/1 D => call_bif1 Bif -call_ext_last u==1 Bif=u$bif:erlang:throw/1 D => call_bif1 Bif +call_ext_last u==1 Bif=u$bif:erlang:exit/1 D => call_bif Bif +call_ext_last u==1 Bif=u$bif:erlang:throw/1 D => call_bif Bif -call_ext_only u==1 Bif=u$bif:erlang:exit/1 => call_bif1 Bif -call_ext_only u==1 Bif=u$bif:erlang:throw/1 => call_bif1 Bif +call_ext_only u==1 Bif=u$bif:erlang:exit/1 => call_bif Bif +call_ext_only u==1 Bif=u$bif:erlang:throw/1 => call_bif Bif # # The error/1 and error/2 BIFs never execute the instruction following them; @@ -785,13 +859,13 @@ call_ext_only u==1 Bif=u$bif:erlang:throw/1 => call_bif1 Bif # the continuation pointer on the stack. # -call_ext_last u==1 Bif=u$bif:erlang:error/1 D => call_bif1 Bif -call_ext_last u==2 Bif=u$bif:erlang:error/2 D => call_bif2 Bif +call_ext_last u==1 Bif=u$bif:erlang:error/1 D => call_bif Bif +call_ext_last u==2 Bif=u$bif:erlang:error/2 D => call_bif Bif call_ext_only Ar=u==1 Bif=u$bif:erlang:error/1 => \ - allocate u Ar | call_bif1 Bif + allocate u Ar | call_bif Bif call_ext_only Ar=u==2 Bif=u$bif:erlang:error/2 => \ - allocate u Ar | call_bif2 Bif + allocate u Ar | call_bif Bif # # The yield/0 BIF is an instruction @@ -809,21 +883,93 @@ call_ext_last u==3 u$func:erlang:hibernate/3 D => i_hibernate call_ext_only u==3 u$func:erlang:hibernate/3 => i_hibernate # -# Hybrid memory architecture need special cons and tuple instructions -# that allocate on the message area. These looks like BIFs in the BEAM code. -# - -call_ext u==2 u$func:hybrid:cons/2 => i_global_cons -call_ext_last u==2 u$func:hybrid:cons/2 D => i_global_cons | deallocate_return D -call_ext_only Ar=u==2 u$func:hybrid:cons/2 => i_global_cons | return - -call_ext u==1 u$func:hybrid:tuple/1 => i_global_tuple -call_ext_last u==1 u$func:hybrid:tuple/1 D => i_global_tuple | deallocate_return D -call_ext_only Ar=u==1 u$func:hybrid:tuple/1 => i_global_tuple | return - -call_ext u==1 u$func:hybrid:copy/1 => i_global_copy -call_ext_last u==1 u$func:hybrid:copy/1 D => i_global_copy | deallocate_return D -call_ext_only u==1 Ar=u$func:hybrid:copy/1 => i_global_copy | return +# If VM probes are not enabled, we want to short-circult calls to +# the dt tag BIFs to make them as cheap as possible. +# + +%unless USE_VM_PROBES + +call_ext Arity u$func:erlang:dt_get_tag/0 => \ + move a=am_undefined r +call_ext_last Arity u$func:erlang:dt_get_tag/0 D => \ + move a=am_undefined r | deallocate D | return +call_ext_only Arity u$func:erlang:dt_get_tag/0 => \ + move a=am_undefined r | return + +move Any r | call_ext Arity u$func:erlang:dt_put_tag/1 => \ + move a=am_undefined r +move Any r | call_ext_last Arity u$func:erlang:dt_put_tag/1 D => \ + move a=am_undefined r | deallocate D | return +move Any r | call_ext_only Arity u$func:erlang:dt_put_tag/1 => \ + move a=am_undefined r | return +call_ext Arity u$func:erlang:dt_put_tag/1 => \ + move a=am_undefined r +call_ext_last Arity u$func:erlang:dt_put_tag/1 D => \ + move a=am_undefined r | deallocate D | return +call_ext_only Arity u$func:erlang:dt_put_tag/1 => \ + move a=am_undefined r | return + +call_ext Arity u$func:erlang:dt_get_tag_data/0 => \ + move a=am_undefined r +call_ext_last Arity u$func:erlang:dt_get_tag_data/0 D => \ + move a=am_undefined r | deallocate D | return +call_ext_only Arity u$func:erlang:dt_get_tag_data/0 => \ + move a=am_undefined r | return + +move Any r | call_ext Arity u$func:erlang:dt_spread_tag/1 => \ + move a=am_true r +move Any r | call_ext_last Arity u$func:erlang:dt_spread_tag/1 D => \ + move a=am_true r | deallocate D | return +move Any r | call_ext_only Arity u$func:erlang:dt_spread_tag/1 => \ + move a=am_true r | return +call_ext Arity u$func:erlang:dt_spread_tag/1 => \ + move a=am_true r +call_ext_last Arity u$func:erlang:dt_spread_tag/1 D => \ + move a=am_true r | deallocate D | return +call_ext_only Arity u$func:erlang:dt_spread_tag/1 => \ + move a=am_true r | return + +move Any r | call_ext Arity u$func:erlang:dt_restore_tag/1 => \ + move a=am_true r +move Any r | call_ext_last Arity u$func:erlang:dt_restore_tag/1 D => \ + move a=am_true r | deallocate D | return +move Any r | call_ext_only Arity u$func:erlang:dt_restore_tag/1 => \ + move a=am_true r | return +call_ext Arity u$func:erlang:dt_restore_tag/1 => \ + move a=am_true r +call_ext_last Arity u$func:erlang:dt_restore_tag/1 D => \ + move a=am_true r | deallocate D | return +call_ext_only Arity u$func:erlang:dt_restore_tag/1 => \ + move a=am_true r | return + +move Any r | call_ext Arity u$func:erlang:dt_prepend_vm_tag_data/1 => \ + move Any r +move Any r | call_ext_last Arity u$func:erlang:dt_prepend_vm_tag_data/1 D => \ + move Any r | deallocate D | return +move Any r | call_ext_only Arity u$func:erlang:dt_prepend_vm_tag_data/1 => \ + move Any r | return +call_ext Arity u$func:erlang:dt_prepend_vm_tag_data/1 => +call_ext_last Arity u$func:erlang:dt_prepend_vm_tag_data/1 D => \ + deallocate D | return +call_ext_only Arity u$func:erlang:dt_prepend_vm_tag_data/1 => \ + return + +move Any r | call_ext Arity u$func:erlang:dt_append_vm_tag_data/1 => \ + move Any r +move Any r | call_ext_last Arity u$func:erlang:dt_append_vm_tag_data/1 D => \ + move Any r | deallocate D | return +move Any r | call_ext_only Arity u$func:erlang:dt_append_vm_tag_data/1 => \ + move Any r | return +call_ext Arity u$func:erlang:dt_append_vm_tag_data/1 => +call_ext_last Arity u$func:erlang:dt_append_vm_tag_data/1 D => \ + deallocate D | return +call_ext_only Arity u$func:erlang:dt_append_vm_tag_data/1 => \ + return + +# Can happen after one of the transformations above. +move Discarded r | move Something r => move Something r + +%endif # # The general case for BIFs that have no special instructions. @@ -832,24 +978,12 @@ call_ext_only u==1 Ar=u$func:hybrid:copy/1 => i_global_copy | return # To make trapping and stack backtraces work correctly, we make sure that # the continuation pointer is always stored on the stack. -call_ext u==0 Bif=u$is_bif => call_bif0 Bif -call_ext u==1 Bif=u$is_bif => call_bif1 Bif -call_ext u==2 Bif=u$is_bif => call_bif2 Bif -call_ext u==3 Bif=$is_bif => call_bif3 Bif +call_ext u Bif=u$is_bif => call_bif Bif -call_ext_last u==0 Bif=u$is_bif D => call_bif0 Bif | deallocate_return D -call_ext_last u==1 Bif=u$is_bif D => call_bif1 Bif | deallocate_return D -call_ext_last u==2 Bif=u$is_bif D => call_bif2 Bif | deallocate_return D -call_ext_last u==3 Bif=u$is_bif D => call_bif3 Bif | deallocate_return D +call_ext_last u Bif=u$is_bif D => call_bif Bif | deallocate_return D -call_ext_only Ar=u==0 Bif=u$is_bif => \ - allocate u Ar | call_bif0 Bif | deallocate_return u -call_ext_only Ar=u==1 Bif=u$is_bif => \ - allocate u Ar | call_bif1 Bif | deallocate_return u -call_ext_only Ar=u==2 Bif=u$is_bif => \ - allocate u Ar | call_bif2 Bif | deallocate_return u -call_ext_only Ar=u==3 Bif=u$is_bif => \ - allocate u Ar | call_bif3 Bif | deallocate_return u +call_ext_only Ar=u Bif=u$is_bif => \ + allocate u Ar | call_bif Bif | deallocate_return u # # Any remaining calls are calls to Erlang functions, not BIFs. @@ -862,9 +996,9 @@ move S=c r | call_ext Ar=u Func=u$is_not_bif => i_move_call_ext S r Func move S=c r | call_ext_last Ar=u Func=u$is_not_bif D => i_move_call_ext_last Func D S r move S=c r | call_ext_only Ar=u Func=u$is_not_bif => i_move_call_ext_only Func S r -call_ext Ar=u Func => i_call_ext Func -call_ext_last Ar=u Func D => i_call_ext_last Func D -call_ext_only Ar=u Func => i_call_ext_only Func +call_ext Ar Func => i_call_ext Func +call_ext_last Ar Func D => i_call_ext_last Func D +call_ext_only Ar Func => i_call_ext_only Func i_apply i_apply_last P @@ -876,10 +1010,7 @@ i_apply_fun_only i_hibernate -call_bif0 e -call_bif1 e -call_bif2 e -call_bif3 e +call_bif e # # Calls to non-building and guard BIFs. @@ -898,7 +1029,7 @@ bif1 p Bif S1 Dst => bif1_body Bif S1 Dst bif1_body Bif Literal=q Dst => move Literal x | bif1_body Bif x Dst bif2 p Bif S1 S2 Dst => i_fetch S1 S2 | i_bif2_body Bif Dst -bif2 Fail=f Bif S1 S2 Dst => i_fetch S1 S2 | i_bif2 Fail Bif Dst +bif2 Fail Bif S1 S2 Dst => i_fetch S1 S2 | i_bif2 Fail Bif Dst i_get s d @@ -914,8 +1045,13 @@ node x node y %hot -i_fast_element j I s d -i_element j s s d +i_fast_element r j I d +i_fast_element x j I d +i_fast_element y j I d + +i_element r j s d +i_element x j s d +i_element y j s d bif1 f b s d bif1_body b s d @@ -942,11 +1078,11 @@ move S r | call_last Ar P=f D => move_call_last S r P D i_move_call_last f P c r -%macro:move_call_last MoveCallLast -arg_f -nonext +%macro:move_call_last MoveCallLast -arg_f -nonext -pack move_call_last/4 -move_call_last x r f P -move_call_last y r f P +move_call_last x r f Q +move_call_last y r f Q move S=c r | call_only Ar P=f => i_move_call_only P S r move S=x r | call_only Ar P=f => move_call_only S r P @@ -976,8 +1112,8 @@ i_move_call_ext_only e c r # Fun calls. -call_fun Arity=u | deallocate D | return => i_call_fun_last Arity D -call_fun Arity=u => i_call_fun Arity +call_fun Arity | deallocate D | return => i_call_fun_last Arity D +call_fun Arity => i_call_fun Arity i_call_fun I i_call_fun_last I P @@ -995,7 +1131,7 @@ is_function f y is_function f r is_function Fail=f c => jump Fail -func_info M=a F=a A=u | label L => gen_func_info(M, F, A, L) +func_info M F A => i_func_info u M F A # ================================================================ # New bit syntax matching (R11B). @@ -1165,7 +1301,7 @@ i_bs_init_heap I I I d i_bs_init_heap_bin_heap I I I d -bs_init_bits Fail Sz Words Regs Flags Dst | binary_too_big_bits(Sz) => system_limit Fail +bs_init_bits Fail Sz=o Words Regs Flags Dst => system_limit Fail bs_init_bits Fail Sz=u Words=u==0 Regs Flags Dst => i_bs_init_bits Sz Regs Dst bs_init_bits Fail Sz=u Words Regs Flags Dst => i_bs_init_bits_heap Sz Words Regs Dst @@ -1184,12 +1320,6 @@ i_bs_init_bits_fail_heap I j I d i_bs_init_bits I I d i_bs_init_bits_heap I I I d -bs_bits_to_bytes Fail Src Dst => i_bs_bits_to_bytes Src Fail Dst - -i_bs_bits_to_bytes r j d -i_bs_bits_to_bytes x j d -i_bs_bits_to_bytes y j d - bs_add Fail S1=i==0 S2 Unit=u==1 D => move S2 D bs_add Fail S1 S2 Unit D => i_fetch S1 S2 | i_bs_add Fail Unit D @@ -1239,13 +1369,13 @@ i_bs_utf16_size s d bs_put_utf8 Fail=j Flags=u Literal=q => \ move Literal x | bs_put_utf8 Fail Flags x -bs_put_utf8 Fail=j u Src=s => i_bs_put_utf8 Fail Src +bs_put_utf8 Fail u Src=s => i_bs_put_utf8 Fail Src i_bs_put_utf8 j s bs_put_utf16 Fail=j Flags=u Literal=q => \ move Literal x | bs_put_utf16 Fail Flags x -bs_put_utf16 Fail=j Flags=u Src=s => i_bs_put_utf16 Fail Flags Src +bs_put_utf16 Fail Flags=u Src=s => i_bs_put_utf16 Fail Flags Src i_bs_put_utf16 j I s @@ -1311,10 +1441,12 @@ fmul p FR1 FR2 FR3 => i_fmul FR1 FR2 FR3 fdiv p FR1 FR2 FR3 => i_fdiv FR1 FR2 FR3 fnegate p FR1 FR2 => i_fnegate FR1 FR2 -fconv Int=iq Dst=l => move Int x | fconv x Dst +fconv Arg=iqan Dst=l => move Arg x | fconv x Dst fmove q l fmove d l +fmove l d + fconv d l i_fadd l l l @@ -1330,12 +1462,6 @@ fcheckerror p => i_fcheckerror i_fcheckerror fclearerror -fmove FR=l Dst=d | new_float_allocation() => fmove_new FR Dst - -# The new instruction for moving a float out of a floating point register. -# (No allocation.) -fmove_new l d - # # New apply instructions in R10B. # @@ -1344,7 +1470,21 @@ apply I apply_last I P # -# New GCing arithmetic instructions. +# Optimize addition and subtraction of small literals using +# the i_increment/4 instruction (in bodies, not in guards). +# + +gc_bif2 p Live u$bif:erlang:splus/2 Int=i Reg=d Dst => \ + gen_increment(Reg, Int, Live, Dst) +gc_bif2 p Live u$bif:erlang:splus/2 Reg=d Int=i Dst => \ + gen_increment(Reg, Int, Live, Dst) + +gc_bif2 p Live u$bif:erlang:sminus/2 Reg=d Int=i Dst | \ + negation_is_small(Int) => \ + gen_increment_from_minus(Reg, Int, Live, Dst) + +# +# GCing arithmetic instructions. # gc_bif2 Fail I u$bif:erlang:splus/2 S1 S2 Dst=d => i_fetch S1 S2 | i_plus Fail I Dst @@ -1367,6 +1507,10 @@ gc_bif1 Fail I u$bif:erlang:bnot/1 Src Dst=d => i_int_bnot Fail Src I Dst gc_bif1 Fail I u$bif:erlang:sminus/1 Src Dst=d => i_fetch i Src | i_minus Fail I Dst gc_bif1 Fail I u$bif:erlang:splus/1 Src Dst=d => i_fetch i Src | i_plus Fail I Dst +i_increment r I I d +i_increment x I I d +i_increment y I I d + i_plus j I d i_minus j I d i_times j I d @@ -1396,35 +1540,49 @@ bif1 Fail u$bif:erlang:trunc/1 s d => too_old_compiler # # Guard BIFs. # -gc_bif1 Fail I Bif=u$bif:erlang:length/1 Src Dst=d => \ - gen_guard_bif(Fail, I, Bif, Src, Dst) +gc_bif1 Fail I Bif Src Dst => \ + gen_guard_bif1(Fail, I, Bif, Src, Dst) -gc_bif1 Fail I Bif=u$bif:erlang:size/1 Src Dst=d => \ - gen_guard_bif(Fail, I, Bif, Src, Dst) +gc_bif2 Fail I Bif S1 S2 Dst => \ + gen_guard_bif2(Fail, I, Bif, S1, S2, Dst) -gc_bif1 Fail I Bif=u$bif:erlang:bit_size/1 Src Dst=d => \ - gen_guard_bif(Fail, I, Bif, Src, Dst) +gc_bif3 Fail I Bif S1 S2 S3 Dst => \ + gen_guard_bif3(Fail, I, Bif, S1, S2, S3, Dst) -gc_bif1 Fail I Bif=u$bif:erlang:byte_size/1 Src Dst=d => \ - gen_guard_bif(Fail, I, Bif, Src, Dst) +i_gc_bif1 Fail Bif V=q Live D => move V x | i_gc_bif1 Fail Bif x Live D -gc_bif1 Fail I Bif=u$bif:erlang:abs/1 Src Dst=d => \ - gen_guard_bif(Fail, I, Bif, Src, Dst) +i_gc_bif1 j I s I d -gc_bif1 Fail I Bif=u$bif:erlang:float/1 Src Dst=d => \ - gen_guard_bif(Fail, I, Bif, Src, Dst) +ii_gc_bif2/6 -gc_bif1 Fail I Bif=u$bif:erlang:round/1 Src Dst=d => \ - gen_guard_bif(Fail, I, Bif, Src, Dst) +ii_gc_bif2 Fail Bif S1 S2 Live D => i_fetch S1 S2 | i_gc_bif2 Fail Bif Live D -gc_bif1 Fail I Bif=u$bif:erlang:trunc/1 Src Dst=d => \ - gen_guard_bif(Fail, I, Bif, Src, Dst) +i_gc_bif2 j I I d -i_gc_bif1 Fail Bif V=q Live D => move V x | i_gc_bif1 Fail Bif x Live D +ii_gc_bif3/7 -i_gc_bif1 j I s I d +ii_gc_bif3 Fail Bif S1 S2 S3 Live D => move S1 x | i_fetch S2 S3 | i_gc_bif3 Fail Bif x Live D + +i_gc_bif3 j I s I d + +# +# The following instruction is specially handled in beam_load.c +# to produce a user-friendly message if an unsupported guard BIF is +# encountered. +# +unsupported_guard_bif/3 +unsupported_guard_bif A B C | never() => # # R13B03 # on_load + +# +# R14A. +# +recv_mark f + +recv_set Fail | label Lbl | loop_rec Lf Reg => \ + i_recv_set | label Lbl | loop_rec Lf Reg +i_recv_set |