From f0923b143ecfdacfe4873fc1e96c8ec69726a4d9 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= <bjorn@erlang.org>
Date: Tue, 16 Jun 2015 06:39:25 +0200
Subject: Allow X and Y registers to be overloaded with any literal

Consider the try_case_end instruction:

  try_case_end s

The 's' operand type means that the operand can either be a
literal of one of the types atom, integer, or empty list, or
a register. That worked well before R12. In R12 additional
types of literals where introduced. Because of way the
overloading was done, an 's' operand cannot handle the
new types of literals. Therefore, code such as the following
is necessary in ops.tab to avoid giving an 's' operand a
literal:

  try_case_end Literal=q => move Literal x | try_case_end x

While this work, it is error-prone in that it is easy to
forget to add that kind of rule. It would also be complicated
in case we wanted to introduce a new kind of addition operator
such as:

  i_plus jssd

Since there are two 's' operands, two scratch registers and
two 'move' instructions would be needed.

Therefore, we'll need to find a smarter way to find tag
register operands. We will overload the pid and port tags
for X and Y register, respectively. That works because pids
and port are immediate values (fit in one word), and there
are no literals for pids and ports.
---
 erts/emulator/beam/ops.tab | 29 ++---------------------------
 1 file changed, 2 insertions(+), 27 deletions(-)

(limited to 'erts/emulator/beam/ops.tab')

diff --git a/erts/emulator/beam/ops.tab b/erts/emulator/beam/ops.tab
index 1d32e72247..4fa6a087e2 100644
--- a/erts/emulator/beam/ops.tab
+++ b/erts/emulator/beam/ops.tab
@@ -237,12 +237,10 @@ try Y F => catch Y F
 try_case Y => try_end Y
 try_end y
 
-try_case_end Literal=q => move Literal x | try_case_end x
 try_case_end s
 
 # Destructive set tuple element
 
-set_tuple_element Lit=q Tuple Pos => move Lit x | set_tuple_element x Tuple Pos
 set_tuple_element s d P
 
 # Get tuple element
@@ -273,8 +271,8 @@ is_number Fail Literal=q => move Literal x | is_number Fail x
 
 jump f
 
-case_end Literal=cq => move Literal x | case_end x
-badmatch Literal=cq => move Literal x | badmatch x
+case_end Literal=c => move Literal x | case_end x
+badmatch Literal=c => move Literal x | badmatch x
 
 case_end r
 case_end x
@@ -849,7 +847,6 @@ is_boolean f y
 
 is_function2 Fail=f acq Arity => jump Fail
 is_function2 Fail=f Fun a => jump Fail
-is_function2 Fail Fun Literal=q => move Literal x | is_function2 Fail Fun x
 
 is_function2 f s s
 %macro: is_function2 IsFunction2 -fail_action
@@ -1122,11 +1119,8 @@ bif1 Fail Bif=u$bif:erlang:get/1 Src=s Dst=d => i_get Src Dst
 
 bif2 Jump=j u$bif:erlang:element/2 S1=s S2=rxy Dst=d => gen_element(Jump, S1, S2, Dst)
 
-bif1 Fail Bif Literal=q Dst => move Literal x | bif1 Fail Bif x Dst
 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 Bif S1 S2 Dst => i_fetch S1 S2 | i_bif2 Fail Bif Dst
 
@@ -1439,8 +1433,6 @@ i_bs_private_append j I d
 # Storing integers into binaries.
 #
 
-bs_put_integer Fail=j Sz=s Unit=u Flags=u Literal=q => \
-	move Literal x | bs_put_integer Fail Sz Unit Flags x
 bs_put_integer Fail=j Sz=sq Unit=u Flags=u Src=s => \
 			gen_put_integer(Fail, Sz, Unit, Flags, Src)
 
@@ -1454,32 +1446,22 @@ i_new_bs_put_integer_imm j I I s
 # Utf8/utf16/utf32 support. (R12B-5)
 #
 
-bs_utf8_size Fail Literal=q Dst=d => \
-	move Literal x | bs_utf8_size Fail x Dst
 bs_utf8_size j Src=s Dst=d => i_bs_utf8_size Src Dst
 
 i_bs_utf8_size s d
 
-bs_utf16_size Fail Literal=q Dst=d => \
-	move Literal x | bs_utf16_size Fail x Dst
 bs_utf16_size j Src=s Dst=d => i_bs_utf16_size Src Dst
 
 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 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 Flags=u Src=s => i_bs_put_utf16 Fail Flags Src
 
 i_bs_put_utf16 j I s
 
-bs_put_utf32 Fail=j Flags=u Literal=q => \
-	move Literal x | bs_put_utf32 Fail Flags x
 bs_put_utf32 Fail=j Flags=u Src=s => \
    i_bs_validate_unicode Fail Src | bs_put_integer Fail i=32 u=1 Flags Src
 
@@ -1490,9 +1472,6 @@ i_bs_validate_unicode j s
 #
 bs_put_float Fail Sz=q Unit Flags Val => badarg Fail
 
-bs_put_float Fail=j Sz Unit=u Flags=u Literal=q => \
-	move Literal x | bs_put_float Fail Sz Unit Flags x
-
 bs_put_float Fail=j Sz=s Unit=u Flags=u Src=s => \
 			gen_put_float(Fail, Sz, Unit, Flags, Src)
 
@@ -1506,8 +1485,6 @@ i_new_bs_put_float_imm j I I s
 # Storing binaries into binaries.
 #
 
-bs_put_binary Fail Sz Unit Flags Literal=q => \
- 	move Literal x | bs_put_binary Fail Sz Unit Flags x
 bs_put_binary Fail=j Sz=s Unit=u Flags=u Src=s => \
 			gen_put_binary(Fail, Sz, Unit, Flags, Src)
 
@@ -1731,8 +1708,6 @@ gc_bif2 Fail I Bif S1 S2 Dst => \
 gc_bif3 Fail I Bif S1 S2 S3 Dst => \
 	gen_guard_bif3(Fail, I, Bif, S1, S2, S3, Dst)
 
-i_gc_bif1 Fail Bif V=q Live D => move V x | i_gc_bif1 Fail Bif x Live D
-
 i_gc_bif1 j I s I d
 
 ii_gc_bif2/6
-- 
cgit v1.2.3


From 135c9821e8806defa455a47208fb897fff1a3ff8 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= <bjorn@erlang.org>
Date: Tue, 23 Jun 2015 08:53:50 +0200
Subject: Make the 'r' operand type optional

The 'r' type is now mandatory. That means in order to handle
both of the following instructions:

  move x(0) y(7)
  move x(1) y(7)

we would need to define two specific operations in ops.tab:

  move r y
  move x y

We want to make 'r' operands optional. That is, if we have
only this specific instruction:

  move x y

it will match both of the following instructions:

  move x(0) y(7)
  move x(1) y(7)

Make 'r' optional allows us to save code space when we don't
want to make handling of x(0) a special case, but we can still
use 'r' to optimize commonly used instructions.
---
 erts/emulator/beam/ops.tab | 456 +++++++++++++++------------------------------
 1 file changed, 155 insertions(+), 301 deletions(-)

(limited to 'erts/emulator/beam/ops.tab')

diff --git a/erts/emulator/beam/ops.tab b/erts/emulator/beam/ops.tab
index 4fa6a087e2..335f8e2db5 100644
--- a/erts/emulator/beam/ops.tab
+++ b/erts/emulator/beam/ops.tab
@@ -39,8 +39,8 @@ too_old_compiler | never() =>
 # necessary.) Since the instructions don't work correctly in R12B, simply
 # refuse to load the module.
 
-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
+func_info M=a a==am_module_info A=u==0 | label L | move n x==0 => too_old_compiler
+func_info M=a a==am_module_info A=u==1 | label L | move n x==0 => too_old_compiler
 
 # The undocumented and unsupported guard BIF is_constant/1 was removed
 # in R13. The is_constant/2 operation is marked as obsolete in genop.tab,
@@ -77,14 +77,14 @@ return
 # 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
+move S X0=x==0 | line Loc | call_ext Ar Func => \
+     line Loc | move S X0 | call_ext Ar Func
+move S X0=x==0 | line Loc | call_ext_last Ar Func=u$is_bif D => \
+     line Loc | move S X0 | call_ext_last Ar Func D
+move S X0=x==0 | line Loc | call_ext_only Ar Func=u$is_bif => \
+     line Loc | move S X0 | call_ext_only Ar Func
+move S X0=x==0 | line Loc | call Ar Func => \
+     line Loc | move S X0 | call Ar Func
 
 #
 # A tail-recursive call to an external function (non-BIF) will
@@ -167,31 +167,24 @@ is_tuple Fail=f S | select_tuple_arity S=d 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_bins r f I
 i_select_val_bins x f I
 i_select_val_bins y f I
 
-i_select_val_lins r f I
 i_select_val_lins x f I
 i_select_val_lins y f I
 
-i_select_val2 r f c c f f
 i_select_val2 x f c c f f
 i_select_val2 y f c c f f
 
-i_select_tuple_arity r f I
 i_select_tuple_arity x f I
 i_select_tuple_arity y f I
 
-i_select_tuple_arity2 r f A A f f
 i_select_tuple_arity2 x f A A f f
 i_select_tuple_arity2 y f A A f f
 
-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
 
@@ -203,30 +196,13 @@ is_ne_exact L1 S1 S2 | jump Fail | label L2 | same_label(L1, L2) => \
 %macro: get_list GetList -pack
 get_list x x x
 get_list x x y
-get_list x x r
 get_list x y x
 get_list x y y
-get_list x y r
-get_list x r x
-get_list x r y
 
 get_list y x x
 get_list y x y
-get_list y x r
 get_list y y x
 get_list y y y
-get_list y y r
-get_list y r x
-get_list y r y
-
-get_list r x x
-get_list r x y
-get_list r x r
-get_list r y x
-get_list r y y
-get_list r y r
-get_list r r x
-get_list r r y
 
 # Old-style catch.
 catch y f
@@ -247,21 +223,15 @@ set_tuple_element s d P
 
 %macro: i_get_tuple_element GetTupleElement -pack
 i_get_tuple_element x P x
-i_get_tuple_element r P x
 i_get_tuple_element y P x
-i_get_tuple_element x P r
-i_get_tuple_element y P r
 
 %cold
-i_get_tuple_element r P r
 i_get_tuple_element x P y
-i_get_tuple_element r P y
 i_get_tuple_element y P y
 %hot
 
 %macro: is_number IsNumber -fail_action
 %cold
-is_number f r
 is_number f x
 is_number f y
 %hot
@@ -271,16 +241,12 @@ is_number Fail Literal=q => move Literal x | is_number Fail x
 
 jump f
 
-case_end Literal=c => move Literal x | case_end x
-badmatch Literal=c => move Literal x | badmatch x
+case_end NotInX=cy => move NotInX x | case_end x
+badmatch NotInX=cy => move NotInX x | badmatch x
 
-case_end r
 case_end x
-case_end y
 
-badmatch r
 badmatch x
-badmatch y
 
 if_end
 raise s s
@@ -289,7 +255,7 @@ raise s s
 badarg j
 system_limit j
 
-move C=cxy r | jump Lbl => move_jump Lbl C
+move C=cxy x==0 | jump Lbl => move_jump Lbl C
 
 %macro: move_jump MoveJump -nonext
 move_jump f n
@@ -307,8 +273,6 @@ move_window/6
 
 # x -> y
 
-move S1=r S2=y | move X1=x Y1=y => move2 S1 S2 X1 Y1
-
 move X1=x Y1=y | move X2=x Y2=y | move X3=x Y3=y | succ(Y1,Y2) | succ(Y2,Y3) => \
     move_window X1 X2 X3 Y1 Y3
 
@@ -329,12 +293,13 @@ 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 S1=x S2=r | move S3=x S4=x => move2 S1 S2 S3 S4
-move S1=x S2=r | move X1=x Y1=y => move2 S1 S2 X1 Y1
-move S1=y S2=r | move X1=x Y1=y => move2 S1 S2 X1 Y1
+move X1=x X2=x | move X3=x Y1=y => move2 X1 X2 X3 Y1
+
+move S1=x S2=x | move X1=x Y1=y => move2 S1 S2 X1 Y1
+move S1=y S2=x | move X1=x Y1=y => move2 S1 S2 X1 Y1
 
-move Y1=y X1=x | move S1=r D1=x => move2 Y1 X1 S1 D1
-move S1=r D1=x | move Y1=y X1=x => move2 S1 D1 Y1 X1
+move Y1=y X1=x | move S1=x D1=x => move2 Y1 X1 S1 D1
+move S1=x D1=x | move Y1=y X1=x => move2 S1 D1 Y1 X1
 
 move2 X1=x Y1=y X2=x Y2=y | move X3=x Y3=y => move3 X1 Y1 X2 Y2 X3 Y3
 move2 Y1=y X1=x Y2=y X2=x | move Y3=y X3=x => move3 Y1 X1 Y2 X2 Y3 X3
@@ -351,14 +316,13 @@ move2 x y x y
 move2 y x y x
 move2 x x x x
 
-move2 x r x x
+move2 x x x y
 
-move2 x r x y
-move2 r y x y
-move2 y r x y
+move2 x y x y
+move2 y x x y
 
-move2 r x y x
-move2 y x r x
+move2 x x y x
+move2 y x x x
 
 %macro: move3 Move3
 move3 x y x y x y
@@ -373,20 +337,14 @@ move S=c D=y => move S x | move x D
 %macro:move Move -pack -gen_dest
 move x x
 move x y
-move x r
 move y x
-move y r
-move r x
-move r y
-move c r
 move c x
 move n x
-move n r
 move y y
 
 # Receive operations.
 
-loop_rec Fail Src | smp_mark_target_label(Fail) => i_loop_rec Fail Src
+loop_rec Fail x==0 | smp_mark_target_label(Fail) => i_loop_rec Fail
 
 label L | wait_timeout Fail Src | smp_already_locked(L) => label L | i_wait_timeout_locked Fail Src
 wait_timeout Fail Src => i_wait_timeout Fail Src
@@ -401,7 +359,7 @@ label L | timeout | smp_already_locked(L) => label L | timeout_locked
 remove_message
 timeout
 timeout_locked
-i_loop_rec f r
+i_loop_rec f
 loop_rec_end f
 wait f
 wait_locked f
@@ -419,29 +377,25 @@ send
 # 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_eq_exact Lbl R=xy C=ian => i_is_eq_exact_immed Lbl R C
+is_eq_exact Lbl R=xy C=q => i_is_eq_exact_literal Lbl R 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
+is_ne_exact Lbl R=xy C=ian => i_is_ne_exact_immed Lbl R C
+is_ne_exact Lbl R=xy C=q => i_is_ne_exact_literal Lbl R 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
+i_is_eq_exact_literal f x c
+i_is_eq_exact_literal f y 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
+i_is_ne_exact_literal f x c
+i_is_ne_exact_literal f y c
 
 #
 # Common Compare Specializations
@@ -449,31 +403,21 @@ i_is_ne_exact_literal y f c
 # to keep the instruction set small-ish
 #
 
-is_eq_exact Lbl S1=xy S2=r => is_eq_exact Lbl S2 S1
-is_eq_exact Lbl S1=rx S2=xy => i_is_eq_exact_spec Lbl S1 S2
+is_eq_exact Lbl S1=y S2=x => is_eq_exact Lbl S2 S1
+is_eq_exact Lbl S1=x S2=xy => i_is_eq_exact_spec Lbl S1 S2
 %macro: i_is_eq_exact_spec EqualExact -fail_action
 
 i_is_eq_exact_spec f x x
 i_is_eq_exact_spec f x y
-i_is_eq_exact_spec f r x
-i_is_eq_exact_spec f r y
-%cold
-i_is_eq_exact_spec f r r
-%hot
 
-is_lt Lbl S1=rxc S2=rxc  => i_is_lt_spec Lbl S1 S2
+is_lt Lbl S1=xc S2=xc  => i_is_lt_spec Lbl S1 S2
 
 %macro: i_is_lt_spec IsLessThan -fail_action
 
 i_is_lt_spec f x x
-i_is_lt_spec f x r
 i_is_lt_spec f x c
-i_is_lt_spec f r x
-i_is_lt_spec f r c
 i_is_lt_spec f c x
-i_is_lt_spec f c r
 %cold
-i_is_lt_spec f r r
 i_is_lt_spec f c c
 %hot
 
@@ -523,7 +467,6 @@ i_put_tuple Dst Arity Puts=* | put S => \
 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
 
@@ -540,48 +483,25 @@ put_list x n x
 put_list y n x
 put_list x x x
 put_list y x x
-put_list x x r
-put_list y r r
 
 put_list y y x
 put_list x y x
-put_list r x x
-put_list r y x
-put_list r x r
-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 y x x
 
 # put_list SrcReg Constant Dst
-put_list r c r
-put_list r c x
-put_list r c y
 
-put_list x c r
 put_list x c x
 put_list x c y
 
-put_list y c r
 put_list y c x
 put_list y c y
 
 # put_list Constant SrcReg Dst
-put_list c r r
-put_list c r x
-put_list c r y
 
-put_list c x r
 put_list c x x
 put_list c x y
 
-put_list c y r
 put_list c y x
 put_list c y y
 
@@ -590,18 +510,12 @@ put_list s s d
 %hot
 
 %macro: i_fetch FetchArgs -pack
-i_fetch c r
 i_fetch c x
 i_fetch c y
-i_fetch r c
-i_fetch r x
-i_fetch r y
 i_fetch x c
-i_fetch x r
 i_fetch x x
 i_fetch x y
 i_fetch y c
-i_fetch y r
 i_fetch y x
 i_fetch y y
 
@@ -629,27 +543,27 @@ return_trace
 # Note: There is no 'move_return y r', since there never are any y registers
 # when we do move_return (if we have y registers, we must do move_deallocate_return).
 
-move S r | return => move_return S r
+move S x==0 | return => move_return S
 
 %macro: move_return MoveReturn -nonext
-move_return x r
-move_return c r
-move_return n r
+move_return x
+move_return c
+move_return n
 
-move S r | deallocate D | return => move_deallocate_return S r D
+move S x==0 | deallocate D | return => move_deallocate_return S D
 
 %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
+move_deallocate_return x Q
+move_deallocate_return y Q
+move_deallocate_return c Q
+move_deallocate_return n Q
 
 deallocate D | return => deallocate_return D
 
 %macro: deallocate_return DeallocateReturn -nonext
 deallocate_return Q
 
-test_heap Need u==1 | put_list Y=y r r => test_heap_1_put_list Need Y
+test_heap Need u==1 | put_list Y=y x==0 x==0 => test_heap_1_put_list Need Y
 
 %macro: test_heap_1_put_list TestHeapPutList -pack
 test_heap_1_put_list I y
@@ -658,18 +572,16 @@ test_heap_1_put_list I y
 
 is_tuple Fail Literal=q => move Literal x | is_tuple Fail x
 is_tuple Fail=f c => jump Fail
-is_tuple Fail=f S=rxy | test_arity Fail=f S=rxy Arity => is_tuple_of_arity Fail S Arity
+is_tuple Fail=f S=xy | test_arity Fail=f S=xy Arity => is_tuple_of_arity Fail S Arity
 
 %macro:is_tuple_of_arity IsTupleOfArity -fail_action
 
 is_tuple_of_arity f x A
 is_tuple_of_arity f y A
-is_tuple_of_arity f r A
 
 %macro: is_tuple IsTuple -fail_action
 is_tuple f x
 is_tuple f y
-is_tuple f r
 
 test_arity Fail Literal=q Arity => move Literal x | test_arity Fail x Arity
 test_arity Fail=f c Arity => jump Fail
@@ -677,7 +589,6 @@ test_arity Fail=f c Arity => jump Fail
 %macro: test_arity IsArity -fail_action
 test_arity f x A
 test_arity f y A
-test_arity f r A
 
 is_tuple_of_arity Fail=f Reg Arity | get_tuple_element Reg P=u==0 Dst=xy => \
   is_tuple_of_arity Fail Reg Arity | extract_next_element Dst | original_reg Reg P
@@ -726,46 +637,40 @@ is_integer Fail=f i =>
 is_integer Fail=f an => jump Fail
 is_integer Fail Literal=q => move Literal x | is_integer Fail x
 
-is_integer Fail=f S=rx | allocate Need Regs => is_integer_allocate Fail S Need Regs
+is_integer Fail=f S=x | allocate Need Regs => is_integer_allocate Fail S Need Regs
 
 %macro: is_integer_allocate IsIntegerAllocate -fail_action
 is_integer_allocate f x I I
-is_integer_allocate f r I I
 
 %macro: is_integer IsInteger -fail_action
 is_integer f x
 is_integer f y
-is_integer f r
 
 is_list Fail=f n =>
 is_list Fail Literal=q => move Literal x | is_list Fail x
 is_list Fail=f c => jump Fail
 %macro: is_list IsList -fail_action
-is_list f r
 is_list f x
 %cold
 is_list f y
 %hot
 
-is_nonempty_list Fail=f S=rx | allocate Need Rs => is_nonempty_list_allocate Fail S Need Rs
+is_nonempty_list Fail=f S=x | allocate Need Rs => is_nonempty_list_allocate Fail S Need Rs
 
 %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
+is_nonempty_list F=f x==0 | test_heap I1 I2 => is_non_empty_list_test_heap F I1 I2
 
 %macro: is_non_empty_list_test_heap IsNonemptyListTestHeap -fail_action -pack
-is_non_empty_list_test_heap f r I t
+is_non_empty_list_test_heap f I t
 
 %macro: is_nonempty_list IsNonemptyList -fail_action
 is_nonempty_list f x
 is_nonempty_list f y
-is_nonempty_list f r
 
 %macro: is_atom IsAtom -fail_action
 is_atom f x
-is_atom f r
 %cold
 is_atom f y
 %hot
@@ -773,7 +678,6 @@ is_atom Fail=f a =>
 is_atom Fail=f niq => jump Fail
 
 %macro: is_float IsFloat -fail_action
-is_float f r
 is_float f x
 %cold
 is_float f y
@@ -787,12 +691,10 @@ is_nil Fail=f qia => jump Fail
 %macro: is_nil IsNil -fail_action
 is_nil f x
 is_nil f y
-is_nil f r
 
 is_binary Fail Literal=q => move Literal x | is_binary Fail x
 is_binary Fail=f c => jump Fail
 %macro: is_binary IsBinary -fail_action
-is_binary f r
 is_binary f x
 %cold
 is_binary f y
@@ -804,7 +706,6 @@ is_bitstr Fail Term => is_bitstring Fail Term
 is_bitstring Fail Literal=q => move Literal x | is_bitstring Fail x
 is_bitstring Fail=f c => jump Fail
 %macro: is_bitstring IsBitstring -fail_action
-is_bitstring f r
 is_bitstring f x
 %cold
 is_bitstring f y
@@ -812,7 +713,6 @@ is_bitstring f y
 
 is_reference Fail=f cq => jump Fail
 %macro: is_reference IsRef -fail_action
-is_reference f r
 is_reference f x
 %cold
 is_reference f y
@@ -820,7 +720,6 @@ is_reference f y
 
 is_pid Fail=f cq => jump Fail
 %macro: is_pid IsPid -fail_action
-is_pid f r
 is_pid f x
 %cold
 is_pid f y
@@ -828,7 +727,6 @@ is_pid f y
 
 is_port Fail=f cq => jump Fail
 %macro: is_port IsPort -fail_action
-is_port f r
 is_port f x
 %cold
 is_port f y
@@ -840,7 +738,6 @@ is_boolean Fail=f ac => jump Fail
 
 %cold
 %macro: is_boolean IsBoolean -fail_action
-is_boolean f r
 is_boolean f x
 is_boolean f y
 %hot
@@ -986,76 +883,76 @@ call_ext_only u==3 u$func:erlang:hibernate/3 => i_hibernate
 %unless USE_VM_PROBES
 
 call_ext Arity u$func:erlang:dt_get_tag/0 => \
-    move a=am_undefined r
+    move a=am_undefined x=0
 call_ext_last Arity u$func:erlang:dt_get_tag/0 D => \
-    move a=am_undefined r | deallocate D | return
+    move a=am_undefined x=0 | 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
+    move a=am_undefined x=0 | return
+
+move Any x==0 | call_ext Arity u$func:erlang:dt_put_tag/1 => \
+    move a=am_undefined x=0
+move Any x==0 | call_ext_last Arity u$func:erlang:dt_put_tag/1 D => \
+    move a=am_undefined x=0 | deallocate D | return
+move Any x==0 | call_ext_only Arity u$func:erlang:dt_put_tag/1 => \
+    move a=am_undefined x=0 | return
 call_ext Arity u$func:erlang:dt_put_tag/1 => \
-    move a=am_undefined r
+    move a=am_undefined x=0
 call_ext_last Arity u$func:erlang:dt_put_tag/1 D => \
-    move a=am_undefined r | deallocate D | return
+    move a=am_undefined x=0 | deallocate D | return
 call_ext_only Arity u$func:erlang:dt_put_tag/1 => \
-    move a=am_undefined r | return
+    move a=am_undefined x=0 | return
 
 call_ext Arity u$func:erlang:dt_get_tag_data/0 => \
-    move a=am_undefined r
+    move a=am_undefined x=0
 call_ext_last Arity u$func:erlang:dt_get_tag_data/0 D => \
-    move a=am_undefined r | deallocate D | return
+    move a=am_undefined x=0 | 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
+    move a=am_undefined x=0 | return
+
+move Any x==0 | call_ext Arity u$func:erlang:dt_spread_tag/1 => \
+    move a=am_true x=0
+move Any x==0 | call_ext_last Arity u$func:erlang:dt_spread_tag/1 D => \
+    move a=am_true x=0 | deallocate D | return
+move Any x==0 | call_ext_only Arity u$func:erlang:dt_spread_tag/1 => \
+    move a=am_true x=0 | return
 call_ext Arity u$func:erlang:dt_spread_tag/1 => \
-    move a=am_true r
+    move a=am_true x=0
 call_ext_last Arity u$func:erlang:dt_spread_tag/1 D => \
-    move a=am_true r | deallocate D | return
+    move a=am_true x=0 | deallocate D | return
 call_ext_only Arity u$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$func:erlang:dt_restore_tag/1 => \
-    move a=am_true r | return
+    move a=am_true x=0 | return
+
+move Any x==0 | call_ext Arity u$func:erlang:dt_restore_tag/1 => \
+    move a=am_true x=0
+move Any x==0 | call_ext_last Arity u$func:erlang:dt_restore_tag/1 D => \
+    move a=am_true x=0 | deallocate D | return
+move Any x==0 | call_ext_only Arity u$func:erlang:dt_restore_tag/1 => \
+    move a=am_true x=0 | return
 call_ext Arity u$func:erlang:dt_restore_tag/1 => \
-    move a=am_true r
+    move a=am_true x=0
 call_ext_last Arity u$func:erlang:dt_restore_tag/1 D => \
-    move a=am_true r | deallocate D | return
+    move a=am_true x=0 | 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
+    move a=am_true x=0 | return
+
+move Any x==0 | call_ext Arity u$func:erlang:dt_prepend_vm_tag_data/1 => \
+    move Any x=0
+move Any x==0 | call_ext_last Arity u$func:erlang:dt_prepend_vm_tag_data/1 D => \
+    move Any x=0 | deallocate D | return
+move Any x==0 | call_ext_only Arity u$func:erlang:dt_prepend_vm_tag_data/1 => \
+    move Any x=0 | return
 call_ext Arity u$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$func:erlang:dt_append_vm_tag_data/1 => \
-    move Any r | return
+move Any x==0 | call_ext Arity u$func:erlang:dt_append_vm_tag_data/1 => \
+    move Any x=0
+move Any x==0 | call_ext_last Arity u$func:erlang:dt_append_vm_tag_data/1 D => \
+    move Any x=0 | deallocate D | return
+move Any x==0 | call_ext_only Arity u$func:erlang:dt_append_vm_tag_data/1 => \
+    move Any x=0 | 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
@@ -1063,7 +960,7 @@ 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
+move Discarded x==0 | move Something x==0 => move Something x=0
 
 %endif
 
@@ -1088,9 +985,9 @@ call_ext_only Ar=u Bif=u$is_bif => \
 # with call instructions.
 #
 
-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$is_not_bif => i_move_call_ext_only Func S r 
+move S=c x==0 | call_ext Ar=u Func=u$is_not_bif => i_move_call_ext S Func
+move S=c x==0 | call_ext_last Ar=u Func=u$is_not_bif D => i_move_call_ext_last Func D S
+move S=c x==0 | call_ext_only Ar=u Func=u$is_not_bif => i_move_call_ext_only Func S
 
 call_ext Ar Func        => i_call_ext Func
 call_ext_last Ar Func D => i_call_ext_last Func D
@@ -1117,7 +1014,7 @@ bif0 u$bif:erlang:node/0 Dst=d => node Dst
 
 bif1 Fail Bif=u$bif:erlang:get/1 Src=s Dst=d => i_get Src Dst
 
-bif2 Jump=j u$bif:erlang:element/2 S1=s S2=rxy Dst=d => gen_element(Jump, S1, S2, Dst)
+bif2 Jump=j u$bif:erlang:element/2 S1=s S2=xy Dst=d => gen_element(Jump, S1, S2, Dst)
 
 bif1 p Bif S1 Dst => bif1_body Bif S1 Dst
 
@@ -1127,24 +1024,20 @@ bif2 Fail Bif S1 S2 Dst => i_fetch S1 S2 | i_bif2 Fail Bif Dst
 i_get s d
 
 %macro: self Self
-self r
 self x
 self y
 
 %macro: node Node
-node r
 node x
 %cold
 node y
 %hot
 
-i_fast_element r j I d
-i_fast_element x j I d
-i_fast_element y j I d
+i_fast_element j x I d
+i_fast_element j y I d
 
-i_element r j s d
-i_element x j s d
-i_element y j s d
+i_element j x s d
+i_element j y s d
 
 bif1 f b s d
 bif1_body b s d
@@ -1155,37 +1048,37 @@ i_bif2_body b d
 # Internal calls.
 #
 
-move S=c r | call Ar P=f => i_move_call S r P
-move S=s r | call Ar P=f => move_call S r P
+move S=c x==0 | call Ar P=f => i_move_call S P
+move S=s x==0 | call Ar P=f => move_call S P
 
-i_move_call c r f
+i_move_call c f
 
 %macro:move_call MoveCall -arg_f -size -nonext
-move_call/3
+move_call/2
 
-move_call x r f
-move_call y r f
+move_call x f
+move_call y f
 
-move S=c r | call_last Ar P=f D => i_move_call_last P D S r
-move S r | call_last Ar P=f D => move_call_last S r P D
+move S=c x==0 | call_last Ar P=f D => i_move_call_last P D S
+move S x==0 | call_last Ar P=f D => move_call_last S P D
 
-i_move_call_last f P c r
+i_move_call_last f P c
 
 %macro:move_call_last MoveCallLast -arg_f -nonext -pack
 
-move_call_last/4
-move_call_last x r f Q
-move_call_last y r f Q
+move_call_last/3
+move_call_last x f Q
+move_call_last y 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
+move S=c x==0 | call_only Ar P=f => i_move_call_only P S
+move S=x x==0 | call_only Ar P=f => move_call_only S P
 
-i_move_call_only f c r
+i_move_call_only f c
 
 %macro:move_call_only MoveCallOnly -arg_f -nonext
-move_call_only/3
+move_call_only/2
 
-move_call_only x r f
+move_call_only x f
 
 call Ar Func        => i_call Func
 call_last Ar Func D => i_call_last Func D
@@ -1199,9 +1092,9 @@ i_call_ext e
 i_call_ext_last e P
 i_call_ext_only e
 
-i_move_call_ext c r e
-i_move_call_ext_last e P c r
-i_move_call_ext_only e c r
+i_move_call_ext c e
+i_move_call_ext_last e P c
+i_move_call_ext_only e c
 
 # Fun calls.
 
@@ -1221,7 +1114,6 @@ i_make_fun I t
 %macro: is_function IsFunction -fail_action
 is_function f x
 is_function f y
-is_function f r
 is_function Fail=f c => jump Fail
 
 func_info M F A => i_func_info u M F A
@@ -1233,126 +1125,102 @@ func_info M F A => i_func_info u M F A
 %cold
 bs_start_match2 Fail=f ica X Y D => jump Fail
 bs_start_match2 Fail Bin X Y D => i_bs_start_match2 Bin Fail X Y D
-i_bs_start_match2 r f I I d
 i_bs_start_match2 x f I I d
 i_bs_start_match2 y f I I d
 
 bs_save2 Reg Index => gen_bs_save(Reg, Index)
-i_bs_save2 r I
 i_bs_save2 x I
 
 bs_restore2 Reg Index => gen_bs_restore(Reg, Index)
-i_bs_restore2 r I
 i_bs_restore2 x I
 
 # Matching integers
 bs_match_string Fail Ms Bits Val => i_bs_match_string Ms Fail Bits Val
 
-i_bs_match_string r f I I
 i_bs_match_string x f I I
 
 # Fetching integers from binaries.
-bs_get_integer2 Fail=f Ms=rx Live=u Sz=sq Unit=u Flags=u Dst=d => \
+bs_get_integer2 Fail=f Ms=x Live=u Sz=sq Unit=u Flags=u Dst=d => \
 			gen_get_integer2(Fail, Ms, Live, Sz, Unit, Flags, Dst)
 
-i_bs_get_integer_small_imm r I f I d
 i_bs_get_integer_small_imm x I f I d
-i_bs_get_integer_imm r I I f I d
 i_bs_get_integer_imm x I I f I d
 i_bs_get_integer f I I d
-i_bs_get_integer_8 r f d
 i_bs_get_integer_8 x f d
-i_bs_get_integer_16 r f d
 i_bs_get_integer_16 x f d
-i_bs_get_integer_32 r f I d
 i_bs_get_integer_32 x f I d
 
 # Fetching binaries from binaries.
-bs_get_binary2 Fail=f Ms=rx Live=u Sz=sq Unit=u Flags=u Dst=d => \
+bs_get_binary2 Fail=f Ms=x Live=u Sz=sq Unit=u Flags=u Dst=d => \
 			gen_get_binary2(Fail, Ms, Live, Sz, Unit, Flags, Dst)
 
 %macro: i_bs_get_binary_imm2 BsGetBinaryImm_2 -fail_action -gen_dest
 %macro: i_bs_get_binary2 BsGetBinary_2 -fail_action -gen_dest
 %macro: i_bs_get_binary_all2 BsGetBinaryAll_2 -fail_action -gen_dest
 
-i_bs_get_binary_imm2 f r I I I d
 i_bs_get_binary_imm2 f x I I I d
-i_bs_get_binary2 f r I s I d
 i_bs_get_binary2 f x I s I d
-i_bs_get_binary_all2 f r I I d
 i_bs_get_binary_all2 f x I I d
-i_bs_get_binary_all_reuse r f I
 i_bs_get_binary_all_reuse x f I
 
 # Fetching float from binaries.
-bs_get_float2 Fail=f Ms=rx Live=u Sz=s Unit=u Flags=u Dst=d => \
+bs_get_float2 Fail=f Ms=x Live=u Sz=s Unit=u Flags=u Dst=d => \
 		gen_get_float2(Fail, Ms, Live, Sz, Unit, Flags, Dst)
 
-bs_get_float2 Fail=f Ms=rx Live=u Sz=q Unit=u Flags=u Dst=d => jump Fail
+bs_get_float2 Fail=f Ms=x Live=u Sz=q Unit=u Flags=u Dst=d => jump Fail
 
 %macro: i_bs_get_float2 BsGetFloat2 -fail_action -gen_dest
-i_bs_get_float2 f r I s I d
 i_bs_get_float2 f x I s I d
 
 # Miscellanous
 
-bs_skip_bits2 Fail=f Ms=rx Sz=s Unit=u Flags=u => \
-			gen_skip_bits2(Fail, Ms, Sz, Unit, Flags)
-bs_skip_bits2 Fail=f Ms=rx Sz=q Unit=u Flags=u => \
+bs_skip_bits2 Fail=f Ms=x Sz=sq Unit=u Flags=u => \
 			gen_skip_bits2(Fail, Ms, Sz, Unit, Flags)
 
 %macro: i_bs_skip_bits_imm2 BsSkipBitsImm2 -fail_action
-i_bs_skip_bits_imm2 f r I
 i_bs_skip_bits_imm2 f x I
 
 %macro: i_bs_skip_bits2 BsSkipBits2 -fail_action
-i_bs_skip_bits2 f r x I
-i_bs_skip_bits2 f r y I
 i_bs_skip_bits2 f x x I
-i_bs_skip_bits2 f x r I
 i_bs_skip_bits2 f x y I
 
 %macro: i_bs_skip_bits_all2 BsSkipBitsAll2 -fail_action
-i_bs_skip_bits_all2 f r I
 i_bs_skip_bits_all2 f x I 
 
-bs_test_tail2 Fail=f Ms=rx Bits=u==0 => bs_test_zero_tail2 Fail Ms
-bs_test_tail2 Fail=f Ms=rx Bits=u => bs_test_tail_imm2 Fail Ms Bits
-bs_test_zero_tail2 f r
+bs_test_tail2 Fail=f Ms=x Bits=u==0 => bs_test_zero_tail2 Fail Ms
+bs_test_tail2 Fail=f Ms=x Bits=u => bs_test_tail_imm2 Fail Ms Bits
 bs_test_zero_tail2 f x
-bs_test_tail_imm2 f r I
 bs_test_tail_imm2 f x I
 
 bs_test_unit F Ms Unit=u==8 => bs_test_unit8 F Ms
-bs_test_unit f r I
 bs_test_unit f x I
-bs_test_unit8 f r
 bs_test_unit8 f x
 
-bs_context_to_binary r
+# An y register operand for bs_context_to_binary is rare,
+# but can happen because of inlining.
+
+bs_context_to_binary Y=y => move Y x | bs_context_to_binary x
+
 bs_context_to_binary x
-bs_context_to_binary y
 
 #
 # Utf8/utf16/utf32 support. (R12B-5)
 #
-bs_get_utf8 Fail=f Ms=rx u u Dst=d => i_bs_get_utf8 Ms Fail Dst
-i_bs_get_utf8 r f d
+bs_get_utf8 Fail=f Ms=x u u Dst=d => i_bs_get_utf8 Ms Fail Dst
 i_bs_get_utf8 x f d
 
-bs_skip_utf8 Fail=f Ms=rx u u => i_bs_get_utf8 Ms Fail x
+bs_skip_utf8 Fail=f Ms=x u u => i_bs_get_utf8 Ms Fail x
 
-bs_get_utf16 Fail=f Ms=rx u Flags=u Dst=d => i_bs_get_utf16 Ms Fail Flags Dst
-bs_skip_utf16 Fail=f Ms=rx u Flags=u => i_bs_get_utf16 Ms Fail Flags x
+bs_get_utf16 Fail=f Ms=x u Flags=u Dst=d => i_bs_get_utf16 Ms Fail Flags Dst
+bs_skip_utf16 Fail=f Ms=x u Flags=u => i_bs_get_utf16 Ms Fail Flags x
 
-i_bs_get_utf16 r f I d
 i_bs_get_utf16 x f I d
 
-bs_get_utf32 Fail=f Ms=rx Live=u Flags=u Dst=d => \
+bs_get_utf32 Fail=f Ms=x Live=u Flags=u Dst=d => \
 	bs_get_integer2 Fail Ms Live i=32 u=1 Flags Dst | \
 	i_fetch Dst Ms | \
 	i_bs_validate_unicode_retract Fail
-bs_skip_utf32 Fail=f Ms=rx Live=u Flags=u => \
+bs_skip_utf32 Fail=f Ms=x Live=u Flags=u => \
 	bs_get_integer2 Fail Ms Live i=32 u=1 Flags x | \
 	i_fetch x Ms | \
 	i_bs_validate_unicode_retract Fail
@@ -1379,9 +1247,8 @@ bs_init2 Fail Sz=u Words Regs Flags Dst => \
 bs_init2 Fail Sz Words=u==0 Regs Flags Dst => \
   i_bs_init_fail Sz Fail Regs Dst
 bs_init2 Fail Sz Words Regs Flags Dst => \
-  i_fetch Sz r | i_bs_init_fail_heap Words Fail Regs Dst
+  i_fetch Sz x=0 | i_bs_init_fail_heap Words Fail Regs Dst
 
-i_bs_init_fail r j I d
 i_bs_init_fail x j I d
 i_bs_init_fail y j I d
 
@@ -1402,9 +1269,8 @@ bs_init_bits Fail Sz=u Words Regs Flags Dst =>  i_bs_init_bits_heap Sz Words Reg
 bs_init_bits Fail Sz Words=u==0 Regs Flags Dst => \
   i_bs_init_bits_fail Sz Fail Regs Dst
 bs_init_bits Fail Sz Words Regs Flags Dst => \
-  i_fetch Sz r | i_bs_init_bits_fail_heap Words Fail Regs Dst
+  i_fetch Sz x=0 | i_bs_init_bits_fail_heap Words Fail Regs Dst
 
-i_bs_init_bits_fail r j I d
 i_bs_init_bits_fail x j I d
 i_bs_init_bits_fail y j I d
 
@@ -1577,7 +1443,6 @@ is_map Fail Lit=q | literal_is_map(Lit) =>
 is_map Fail cq => jump Fail
 
 %macro: is_map IsMap -fail_action
-is_map f r
 is_map f x
 is_map f y
 
@@ -1588,35 +1453,25 @@ has_map_fields Fail Src Size Rest=* => \
 
 ## Transform get_map_elements(s) #{ K1 := V1, K2 := V2 }
 
-get_map_elements Fail Src=rxy Size=u==2 Rest=* => \
+get_map_elements Fail Src=xy Size=u==2 Rest=* => \
     gen_get_map_element(Fail, Src, Size, Rest)
 get_map_elements Fail Src Size Rest=* | map_key_sort(Size, Rest) => \
    gen_get_map_elements(Fail, Src, Size, Rest)
 
 i_get_map_elements f s I
 
-i_get_map_element Fail Src=rxy Key=ry Dst => \
+i_get_map_element Fail Src=xy Key=y Dst => \
     move Key x | i_get_map_element Fail Src x Dst
 
 %macro: i_get_map_element_hash GetMapElementHash -fail_action
-i_get_map_element_hash f r c I r
-i_get_map_element_hash f x c I r
-i_get_map_element_hash f y c I r
-i_get_map_element_hash f r c I x
 i_get_map_element_hash f x c I x
 i_get_map_element_hash f y c I x
-i_get_map_element_hash f r c I y
 i_get_map_element_hash f x c I y
 i_get_map_element_hash f y c I y
 
 %macro: i_get_map_element GetMapElement -fail_action
-i_get_map_element f r x r
-i_get_map_element f x x r
-i_get_map_element f y x r
-i_get_map_element f r x x
 i_get_map_element f x x x
 i_get_map_element f y x x
-i_get_map_element f r x y
 i_get_map_element f x x y
 i_get_map_element f y x y
 
@@ -1625,9 +1480,9 @@ i_get_map_element f y x y
 # the i_increment/4 instruction (in bodies, not in guards).
 #
 
-gc_bif2 p Live u$bif:erlang:splus/2 Int=i Reg=d Dst => \
+gc_bif2 p Live u$bif:erlang:splus/2 Int=i Reg=x Dst => \
 	gen_increment(Reg, Int, Live, Dst)
-gc_bif2 p Live u$bif:erlang:splus/2 Reg=d Int=i Dst => \
+gc_bif2 p Live u$bif:erlang:splus/2 Reg=x Int=i Dst => \
 	gen_increment(Reg, Int, Live, Dst)
 
 gc_bif2 p Live u$bif:erlang:sminus/2 Reg=d Int=i Dst | \
@@ -1662,7 +1517,6 @@ gc_bif1 Fail I u$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
 
-- 
cgit v1.2.3


From c76b297463feee133adad510128d312536150392 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= <bjorn@erlang.org>
Date: Tue, 16 Jun 2015 10:06:20 +0200
Subject: Eliminate the use of i_fetch in arithmetic instructions

The i_fetch instruction fetches two operands and places them
in the tmp_arg1 and tmp_arg2 variables. The next instruction
(such as i_plus) does not have to handle different types of
operands, but can get get them simply from the tmp_arg*
variables. Thus, i_fetch was introduced as a way to temper
a potentail combinatorial explosion.

Unfortunately, clang will generate terrible code because of
the tmp_arg1 and tmp_arg2 variables being live across multiple
instructions. Note that Clang has no way to predict the control
flow from one instruction to another. Clang must assume that
any instruction can jump to any other instruction. Somehow GCC
manages to cope with this situation much better.

Therefore, to improve the quality of the code generated by clang, we
must eliminate all uses of the tmp_arg1 and tmp_arg2 variables. This
commit eliminates the use of i_fetch in combination with the
arithmetic and logical instructions.

While we are touching the code for the bsr and bsl instructions,
also move the tmp_big[] array from top scope of process main into
the block that encloses the bsr and bsl instructions.
---
 erts/emulator/beam/ops.tab | 93 ++++++++++++++++++++++++++++++----------------
 1 file changed, 60 insertions(+), 33 deletions(-)

(limited to 'erts/emulator/beam/ops.tab')

diff --git a/erts/emulator/beam/ops.tab b/erts/emulator/beam/ops.tab
index 335f8e2db5..e4a757ec8b 100644
--- a/erts/emulator/beam/ops.tab
+++ b/erts/emulator/beam/ops.tab
@@ -1475,68 +1475,95 @@ i_get_map_element f y x x
 i_get_map_element f x x y
 i_get_map_element f y x y
 
+#
+# Convert the plus operations to a generic plus instruction.
+#
+gen_plus/5
+gen_minus/5
+
+gc_bif1 Fail Live u$bif:erlang:splus/1 Src Dst => \
+   gen_plus Fail Live Src i Dst
+gc_bif2 Fail Live u$bif:erlang:splus/2 S1 S2 Dst => \
+   gen_plus Fail Live S1 S2 Dst
+
+gc_bif1 Fail Live u$bif:erlang:sminus/1 Src Dst => \
+   gen_minus Fail Live i Src Dst
+gc_bif2 Fail Live u$bif:erlang:sminus/2 S1 S2 Dst => \
+   gen_minus Fail Live S1 S2 Dst
+
 #
 # Optimize addition and subtraction of small literals using
 # the i_increment/4 instruction (in bodies, not in guards).
 #
 
-gc_bif2 p Live u$bif:erlang:splus/2 Int=i Reg=x Dst => \
+gen_plus p Live Int=i Reg=d Dst => \
 	gen_increment(Reg, Int, Live, Dst)
-gc_bif2 p Live u$bif:erlang:splus/2 Reg=x Int=i Dst => \
+gen_plus p Live 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_minus p Live 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$bif:erlang:splus/2 S1=x S2=x Dst=d => i_plus Fail I S1 S2 Dst
-gc_bif2 Fail I u$bif:erlang:splus/2 S1 S2 Dst=d => i_fetch S1 S2 | i_plus Fail I Dst
-gc_bif2 Fail I u$bif:erlang:sminus/2 S1=x S2=x Dst=d => i_minus Fail I S1 S2 Dst
-gc_bif2 Fail I u$bif:erlang:sminus/2 S1 S2 Dst=d => i_fetch S1 S2 | i_minus Fail I Dst
-gc_bif2 Fail I u$bif:erlang:stimes/2 S1 S2 Dst=d => i_fetch S1 S2 | i_times Fail I Dst
-gc_bif2 Fail I u$bif:erlang:div/2 S1 S2 Dst=d => i_fetch S1 S2 | i_m_div Fail I Dst
+gen_plus Fail Live S1 S2 Dst => i_plus Fail Live S1 S2 Dst
 
-gc_bif2 Fail I u$bif:erlang:intdiv/2 S1 S2 Dst=d => i_fetch S1 S2 | i_int_div Fail I Dst
-gc_bif2 Fail I u$bif:erlang:rem/2 S1=x S2=x Dst=d => i_rem Fail I S1 S2 Dst
-gc_bif2 Fail I u$bif:erlang:rem/2 S1 S2 Dst=d => i_fetch S1 S2 | i_rem Fail I Dst
+gen_minus Fail Live S1 S2 Dst => i_minus Fail Live S1 S2 Dst
 
-gc_bif2 Fail I u$bif:erlang:bsl/2 S1 S2 Dst=d => i_fetch S1 S2 | i_bsl Fail I Dst
-gc_bif2 Fail I u$bif:erlang:bsr/2 S1 S2 Dst=d => i_fetch S1 S2 | i_bsr Fail I Dst
+gc_bif2 Fail Live u$bif:erlang:stimes/2 S1 S2 Dst => \
+  i_times Fail Live S1 S2 Dst
 
-gc_bif2 Fail I u$bif:erlang:band/2 S1=x S2=c Dst=d => i_band Fail I S1 S2 Dst
-gc_bif2 Fail I u$bif:erlang:band/2 S1 S2 Dst=d => i_fetch S1 S2 | i_band Fail I Dst
-gc_bif2 Fail I u$bif:erlang:bor/2 S1 S2 Dst=d => i_fetch S1 S2 | i_bor Fail I Dst
-gc_bif2 Fail I u$bif:erlang:bxor/2 S1 S2 Dst=d => i_fetch S1 S2 | i_bxor Fail I Dst
+gc_bif2 Fail Live u$bif:erlang:div/2 S1 S2 Dst => \
+  i_m_div Fail Live S1 S2 Dst
+gc_bif2 Fail Live u$bif:erlang:intdiv/2 S1 S2 Dst => \
+  i_int_div Fail Live S1 S2 Dst
 
-gc_bif1 Fail I u$bif:erlang:bnot/1 Src Dst=d => i_int_bnot Fail Src I Dst
+gc_bif2 Fail Live u$bif:erlang:rem/2 S1 S2 Dst => \
+  i_rem Fail Live S1 S2 Dst
+
+gc_bif2 Fail Live u$bif:erlang:bsl/2 S1 S2 Dst => \
+  i_bsl Fail Live S1 S2 Dst
+gc_bif2 Fail Live u$bif:erlang:bsr/2 S1 S2 Dst => \
+  i_bsr Fail Live S1 S2 Dst
+
+gc_bif2 Fail Live u$bif:erlang:band/2 S1 S2 Dst => \
+  i_band Fail Live S1 S2 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
+gc_bif2 Fail Live u$bif:erlang:bor/2 S1 S2 Dst => \
+  i_bor Fail Live S1 S2 Dst
+
+gc_bif2 Fail Live u$bif:erlang:bxor/2 S1 S2 Dst => \
+  i_bxor Fail Live S1 S2 Dst
+
+gc_bif1 Fail I u$bif:erlang:bnot/1 Src Dst=d => i_int_bnot Fail Src I Dst
 
 i_increment x I I d
 i_increment y I I d
 
 i_plus j I x x d
-i_plus j I d
+i_plus j I s s d
+
 i_minus j I x x d
-i_minus j I d
-i_times j I d
-i_m_div j I d
-i_int_div j I d
+i_minus j I s s d
+
+i_times j I s s d
+
+i_m_div j I s s d
+i_int_div j I s s d
+
 i_rem j I x x d
-i_rem j I d
+i_rem j I s s d
 
-i_bsl j I d
-i_bsr j I d
+i_bsl j I s s d
+i_bsr j I s s d
 
 i_band j I x c d
-i_band j I d
-i_bor j I d
-i_bxor j I d
+i_band j I s s d
+
+i_bor j I s s d
+i_bxor j I s s d
 
 i_int_bnot j s I d
 
-- 
cgit v1.2.3


From 3c583d590589b64ab543102f2542bda0337ce2f9 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= <bjorn@erlang.org>
Date: Wed, 17 Jun 2015 06:52:50 +0200
Subject: Eliminate the use of i_fetch for relational operators

---
 erts/emulator/beam/ops.tab | 67 +++++++++++++++-------------------------------
 1 file changed, 22 insertions(+), 45 deletions(-)

(limited to 'erts/emulator/beam/ops.tab')

diff --git a/erts/emulator/beam/ops.tab b/erts/emulator/beam/ops.tab
index e4a757ec8b..57fb1ea026 100644
--- a/erts/emulator/beam/ops.tab
+++ b/erts/emulator/beam/ops.tab
@@ -397,59 +397,36 @@ i_is_ne_exact_immed f y c
 i_is_ne_exact_literal f x c
 i_is_ne_exact_literal f y c
 
-#
-# Common Compare Specializations
-# We don't do all of them since we want
-# to keep the instruction set small-ish
-#
-
-is_eq_exact Lbl S1=y S2=x => is_eq_exact Lbl S2 S1
-is_eq_exact Lbl S1=x S2=xy => i_is_eq_exact_spec Lbl S1 S2
-%macro: i_is_eq_exact_spec EqualExact -fail_action
-
-i_is_eq_exact_spec f x x
-i_is_eq_exact_spec f x y
-
-is_lt Lbl S1=xc S2=xc  => i_is_lt_spec Lbl S1 S2
-
-%macro: i_is_lt_spec IsLessThan -fail_action
-
-i_is_lt_spec f x x
-i_is_lt_spec f x c
-i_is_lt_spec f c x
+is_eq_exact Lbl Y=y X=x => is_eq_exact Lbl X Y
+%macro: is_eq_exact EqualExact -fail_action
+is_eq_exact f x x
+is_eq_exact f x y
+is_eq_exact f s s
+
+%macro: is_lt IsLessThan -fail_action
+is_lt f x x
+is_lt f x c
+is_lt f c x
 %cold
-i_is_lt_spec f c c
+is_lt f s s
 %hot
 
-is_ge Lbl S1=xc S2=xc => i_is_ge_spec Lbl S1 S2
-
-%macro: i_is_ge_spec IsGreaterEqual -fail_action
-
-i_is_ge_spec f x x
-i_is_ge_spec f x c
-i_is_ge_spec f c x
+%macro: is_ge IsGreaterEqual -fail_action
+is_ge f x x
+is_ge f x c
+is_ge f c x
 %cold
-i_is_ge_spec f c c
+is_ge f s s
 %hot
 
-#
-# All other comparisons.
-#
-
-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
+%macro: is_ne_exact NotEqualExact -fail_action
+is_ne_exact f s s
 
-is_lt Lbl S1 S2 => i_fetch S1 S2 | i_is_lt Lbl
-is_ge Lbl S1 S2 => i_fetch S1 S2 | i_is_ge 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
+%macro: is_eq Equal -fail_action
+is_eq f s s
 
-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
+%macro: is_ne NotEqual -fail_action
+is_ne f s s
 
 #
 # Putting things.
-- 
cgit v1.2.3


From 30204739a047ab96d2b7d59ae461d4cbb2131509 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= <bjorn@erlang.org>
Date: Wed, 17 Jun 2015 10:34:39 +0200
Subject: Eliminate the use of i_fetch for BIF instructions

---
 erts/emulator/beam/ops.tab | 21 ++++++++++-----------
 1 file changed, 10 insertions(+), 11 deletions(-)

(limited to 'erts/emulator/beam/ops.tab')

diff --git a/erts/emulator/beam/ops.tab b/erts/emulator/beam/ops.tab
index 57fb1ea026..492d4d240a 100644
--- a/erts/emulator/beam/ops.tab
+++ b/erts/emulator/beam/ops.tab
@@ -995,8 +995,8 @@ bif2 Jump=j u$bif:erlang:element/2 S1=s S2=xy Dst=d => gen_element(Jump, S1, S2,
 
 bif1 p Bif S1 Dst => bif1_body Bif S1 Dst
 
-bif2 p Bif S1 S2 Dst => i_fetch S1 S2 | i_bif2_body Bif Dst
-bif2 Fail Bif S1 S2 Dst => i_fetch S1 S2 | i_bif2 Fail Bif Dst
+bif2 p Bif S1 S2 Dst => i_bif2_body Bif S1 S2 Dst
+bif2 Fail Bif S1 S2 Dst => i_bif2 Fail Bif S1 S2 Dst
 
 i_get s d
 
@@ -1018,8 +1018,8 @@ i_element j y s d
 
 bif1 f b s d
 bif1_body b s d
-i_bif2 f b d
-i_bif2_body b d
+i_bif2 f b s s d
+i_bif2_body b s s d
 
 #
 # Internal calls.
@@ -1568,17 +1568,16 @@ gc_bif3 Fail I Bif S1 S2 S3 Dst => \
 
 i_gc_bif1 j I s I d
 
-ii_gc_bif2/6
-
-ii_gc_bif2 Fail Bif S1 S2 Live D => i_fetch S1 S2 | i_gc_bif2 Fail Bif Live D
-
-i_gc_bif2 j I I d
+i_gc_bif2 j I I s s d
 
 ii_gc_bif3/7
 
-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
+# A specific instruction can only have 6 operands, so we must
+# pass one of the arguments in an x register.
+ii_gc_bif3 Fail Bif Live S1 S2 S3 Dst => \
+  move S1 x | i_gc_bif3 Fail Bif Live S2 S3 Dst
 
-i_gc_bif3 j I s I d
+i_gc_bif3 j I I s s d
 
 #
 # The following instruction is specially handled in beam_load.c
-- 
cgit v1.2.3


From bede3941be8629efa4d91755c085a91b1416d432 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= <bjorn@erlang.org>
Date: Wed, 17 Jun 2015 15:31:36 +0200
Subject: Eliminate use of i_fetch for bit syntax instructions

---
 erts/emulator/beam/ops.tab | 29 +++++++++++++----------------
 1 file changed, 13 insertions(+), 16 deletions(-)

(limited to 'erts/emulator/beam/ops.tab')

diff --git a/erts/emulator/beam/ops.tab b/erts/emulator/beam/ops.tab
index 492d4d240a..637cef0a22 100644
--- a/erts/emulator/beam/ops.tab
+++ b/erts/emulator/beam/ops.tab
@@ -1122,7 +1122,7 @@ bs_get_integer2 Fail=f Ms=x Live=u Sz=sq Unit=u Flags=u Dst=d => \
 
 i_bs_get_integer_small_imm x I f I d
 i_bs_get_integer_imm x I I f I d
-i_bs_get_integer f I I d
+i_bs_get_integer f I I s s d
 i_bs_get_integer_8 x f d
 i_bs_get_integer_16 x f d
 i_bs_get_integer_32 x f I d
@@ -1195,14 +1195,12 @@ i_bs_get_utf16 x f I d
 
 bs_get_utf32 Fail=f Ms=x Live=u Flags=u Dst=d => \
 	bs_get_integer2 Fail Ms Live i=32 u=1 Flags Dst | \
-	i_fetch Dst Ms | \
-	i_bs_validate_unicode_retract Fail
+	i_bs_validate_unicode_retract Fail Dst Ms
 bs_skip_utf32 Fail=f Ms=x Live=u Flags=u => \
 	bs_get_integer2 Fail Ms Live i=32 u=1 Flags x | \
-	i_fetch x Ms | \
-	i_bs_validate_unicode_retract Fail
+	i_bs_validate_unicode_retract Fail x Ms
 
-i_bs_validate_unicode_retract j
+i_bs_validate_unicode_retract j s s
 %hot
 
 #
@@ -1224,12 +1222,12 @@ bs_init2 Fail Sz=u Words Regs Flags Dst => \
 bs_init2 Fail Sz Words=u==0 Regs Flags Dst => \
   i_bs_init_fail Sz Fail Regs Dst
 bs_init2 Fail Sz Words Regs Flags Dst => \
-  i_fetch Sz x=0 | i_bs_init_fail_heap Words Fail Regs Dst
+  i_bs_init_fail_heap Sz Words Fail Regs Dst
 
 i_bs_init_fail x j I d
 i_bs_init_fail y j I d
 
-i_bs_init_fail_heap I j I d
+i_bs_init_fail_heap s I j I d
 
 i_bs_init I I d
 i_bs_init_heap_bin I I d
@@ -1246,31 +1244,30 @@ bs_init_bits Fail Sz=u Words Regs Flags Dst =>  i_bs_init_bits_heap Sz Words Reg
 bs_init_bits Fail Sz Words=u==0 Regs Flags Dst => \
   i_bs_init_bits_fail Sz Fail Regs Dst
 bs_init_bits Fail Sz Words Regs Flags Dst => \
-  i_fetch Sz x=0 | i_bs_init_bits_fail_heap Words Fail Regs Dst
+  i_bs_init_bits_fail_heap Sz Words Fail Regs Dst
 
 i_bs_init_bits_fail x j I d
 i_bs_init_bits_fail y j I d
 
-i_bs_init_bits_fail_heap I j I d
+i_bs_init_bits_fail_heap s I j I d
 
 i_bs_init_bits I I d
 i_bs_init_bits_heap I I I 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
 
-i_bs_add j I d
+bs_add j s s I d
 
 bs_append Fail Size Extra Live Unit Bin Flags Dst => \
-  i_fetch Size Bin | i_bs_append Fail Extra Live Unit Dst
+  move Bin x | i_bs_append Fail Extra Live Unit Size Dst
 
 bs_private_append Fail Size Unit Bin Flags Dst => \
-  i_fetch Size Bin | i_bs_private_append Fail Unit Dst
+  i_bs_private_append Fail Unit Size Bin Dst
 
 bs_init_writable
 
-i_bs_append j I I I d
-i_bs_private_append j I d
+i_bs_append j I I I s d
+i_bs_private_append j I s s d
 
 #
 # Storing integers into binaries.
-- 
cgit v1.2.3


From bf2b06d3edd2c855071529be17c0c35e778a88fa Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= <bjorn@erlang.org>
Date: Wed, 17 Jun 2015 16:33:19 +0200
Subject: Remove the i_fetch instruction

---
 erts/emulator/beam/ops.tab | 15 ---------------
 1 file changed, 15 deletions(-)

(limited to 'erts/emulator/beam/ops.tab')

diff --git a/erts/emulator/beam/ops.tab b/erts/emulator/beam/ops.tab
index 637cef0a22..631e10215c 100644
--- a/erts/emulator/beam/ops.tab
+++ b/erts/emulator/beam/ops.tab
@@ -486,21 +486,6 @@ put_list c y y
 put_list s s d
 %hot
 
-%macro: i_fetch FetchArgs -pack
-i_fetch c x
-i_fetch c y
-i_fetch x c
-i_fetch x x
-i_fetch x y
-i_fetch y c
-i_fetch y x
-i_fetch y y
-
-%cold
-i_fetch c c
-i_fetch s s
-%hot
-
 #
 # Some more only used by the emulator
 #
-- 
cgit v1.2.3


From dddd225934cfe79c052e9013a8a2c1d6a29c8ae0 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= <bjorn@erlang.org>
Date: Tue, 23 Jun 2015 15:24:07 +0200
Subject: Remove the last use of tmp_arg1

---
 erts/emulator/beam/ops.tab | 53 ++++++++++++----------------------------------
 1 file changed, 13 insertions(+), 40 deletions(-)

(limited to 'erts/emulator/beam/ops.tab')

diff --git a/erts/emulator/beam/ops.tab b/erts/emulator/beam/ops.tab
index 631e10215c..853ba82f83 100644
--- a/erts/emulator/beam/ops.tab
+++ b/erts/emulator/beam/ops.tab
@@ -230,6 +230,12 @@ i_get_tuple_element x P y
 i_get_tuple_element y P y
 %hot
 
+%macro: i_get_tuple_element2 GetTupleElement2 -pack
+i_get_tuple_element2 x P x
+
+%macro: i_get_tuple_element3 GetTupleElement3 -pack
+i_get_tuple_element3 x P x
+
 %macro: is_number IsNumber -fail_action
 %cold
 is_number f x
@@ -552,48 +558,15 @@ test_arity Fail=f c Arity => jump Fail
 test_arity f x A
 test_arity f y A
 
-is_tuple_of_arity Fail=f Reg Arity | get_tuple_element Reg P=u==0 Dst=xy => \
-  is_tuple_of_arity Fail Reg Arity | extract_next_element Dst | original_reg Reg P
-
-test_arity Fail Reg Arity | get_tuple_element Reg P=u==0 Dst=xy => \
-  test_arity Fail Reg Arity | extract_next_element Dst | original_reg Reg P
-
-original_reg Reg P1 | get_tuple_element Reg P2 Dst=xy | succ(P1, P2) => \
-  extract_next_element Dst | original_reg Reg P2
-
-get_tuple_element Reg P Dst => i_get_tuple_element Reg P Dst | original_reg Reg P
-
-original_reg Reg Pos =>
-
-original_reg/2
-
-extract_next_element D1=xy | original_reg Reg P1 | get_tuple_element Reg P2 D2=xy | \
-succ(P1, P2) | succ(D1, D2) => \
-  extract_next_element2 D1 | original_reg Reg P2
-
-extract_next_element2 D1=xy | original_reg Reg P1 | get_tuple_element Reg P2 D2=xy | \
-succ(P1, P2) | succ2(D1, D2) => \
-  extract_next_element3 D1 | original_reg Reg P2
-
-#extract_next_element3 D1=xy | original_reg Reg P1 | get_tuple_element Reg P2 D2=xy | \
-#succ(P1, P2) | succ3(D1, D2) => \
-#  extract_next_element4 D1 | original_reg Reg P2
-
-%macro: extract_next_element ExtractNextElement -pack
-extract_next_element x
-extract_next_element y
-
-%macro: extract_next_element2 ExtractNextElement2 -pack
-extract_next_element2 x
-extract_next_element2 y
+get_tuple_element Reg=x P1 D1=x | get_tuple_element Reg=x P2 D2=x | \
+   get_tuple_element Reg=x P3 D3=x | \
+   succ(P1, P2) | succ(P2, P3) | \
+   succ(D1, D2) | succ(D2, D3) => i_get_tuple_element3 Reg P1 D1
 
-%macro: extract_next_element3 ExtractNextElement3 -pack
-extract_next_element3 x
-extract_next_element3 y
+get_tuple_element Reg=x P1 D1=x | get_tuple_element Reg=x P2 D2=x | \
+   succ(P1, P2) | succ(D1, D2) => i_get_tuple_element2 Reg P1 D1
 
-#%macro: extract_next_element4 ExtractNextElement4 -pack
-#extract_next_element4 x
-#extract_next_element4 y
+get_tuple_element Reg P Dst => i_get_tuple_element Reg P Dst
 
 is_integer Fail=f i =>
 is_integer Fail=f an => jump Fail
-- 
cgit v1.2.3


From 1f73d45327bb13a615f2f0a8d9d4888ddacb95a5 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= <bjorn@erlang.org>
Date: Tue, 23 Jun 2015 09:38:23 +0200
Subject: Add back frequently used x(0) instructions

---
 erts/emulator/beam/ops.tab | 37 +++++++++++++++++++++++++++++++++----
 1 file changed, 33 insertions(+), 4 deletions(-)

(limited to 'erts/emulator/beam/ops.tab')

diff --git a/erts/emulator/beam/ops.tab b/erts/emulator/beam/ops.tab
index 853ba82f83..4ab08faae3 100644
--- a/erts/emulator/beam/ops.tab
+++ b/erts/emulator/beam/ops.tab
@@ -204,6 +204,14 @@ get_list y x y
 get_list y y x
 get_list y y y
 
+# The following get_list instructions using x(0) are frequently used.
+get_list r x x
+get_list r r y
+get_list x r x
+get_list r x y
+get_list r y r
+get_list r x r
+
 # Old-style catch.
 catch y f
 catch_end y
@@ -348,6 +356,14 @@ move c x
 move n x
 move y y
 
+# The following move instructions using x(0) are frequently used.
+
+move x r
+move r x
+move y r
+move c r
+move r y
+
 # Receive operations.
 
 loop_rec Fail x==0 | smp_mark_target_label(Fail) => i_loop_rec Fail
@@ -390,6 +406,7 @@ is_ne_exact Lbl R=xy C=ian => i_is_ne_exact_immed Lbl R C
 is_ne_exact Lbl R=xy C=q => i_is_ne_exact_literal Lbl R 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
 
@@ -478,15 +495,21 @@ put_list x c x
 put_list x c y
 
 put_list y c x
-put_list y c y
 
 # put_list Constant SrcReg Dst
 
 put_list c x x
-put_list c x y
-
 put_list c y x
-put_list c y y
+
+# The following put_list instructions using x(0) are frequently used.
+
+put_list y r r
+put_list x r r
+put_list r n r
+put_list r n x
+put_list r x x
+put_list r x r
+put_list x x r
 
 %cold
 put_list s s d
@@ -544,10 +567,12 @@ is_tuple Fail=f S=xy | test_arity Fail=f S=xy Arity => is_tuple_of_arity Fail S
 
 %macro:is_tuple_of_arity IsTupleOfArity -fail_action
 
+is_tuple_of_arity f r A
 is_tuple_of_arity f x A
 is_tuple_of_arity f y A
 
 %macro: is_tuple IsTuple -fail_action
+is_tuple f r
 is_tuple f x
 is_tuple f y
 
@@ -593,6 +618,7 @@ is_list f y
 is_nonempty_list Fail=f S=x | allocate Need Rs => is_nonempty_list_allocate Fail S Need Rs
 
 %macro:is_nonempty_list_allocate IsNonemptyListAllocate -fail_action -pack
+is_nonempty_list_allocate f r I t
 is_nonempty_list_allocate f x I t
 
 is_nonempty_list F=f x==0 | test_heap I1 I2 => is_non_empty_list_test_heap F I1 I2
@@ -1440,6 +1466,7 @@ gen_minus p Live Reg=d Int=i Dst | negation_is_small(Int) => \
 # GCing arithmetic instructions.
 #
 
+gen_plus Fail Live Y=y X=x Dst => i_plus Fail Live X Y Dst
 gen_plus Fail Live S1 S2 Dst => i_plus Fail Live S1 S2 Dst
 
 gen_minus Fail Live S1 S2 Dst => i_minus Fail Live S1 S2 Dst
@@ -1471,10 +1498,12 @@ gc_bif2 Fail Live u$bif:erlang:bxor/2 S1 S2 Dst => \
 
 gc_bif1 Fail I u$bif:erlang:bnot/1 Src Dst=d => i_int_bnot Fail Src I Dst
 
+i_increment r I I d
 i_increment x I I d
 i_increment y I I d
 
 i_plus j I x x d
+i_plus j I x y d
 i_plus j I s s d
 
 i_minus j I x x d
-- 
cgit v1.2.3


From 3dc2bee53b4d36f41821a6ab512cf01c958c11f9 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= <bjorn@erlang.org>
Date: Tue, 23 Jun 2015 17:31:49 +0200
Subject: Introduce specialized versions of move2

Currently, move2/2 does the two moves sequentially to ensure
that the instruction will always work correctly.

We can do better than that. If the two move instructions have
any registers in common, we can introduce simpler and slightly
more efficient instructions to handle those cases:

  move_shift/3
  move_dup/3

For the remaining cases when the the move instructions
have no common registers, the move2/4 instruction can perform
the moves in parallel which is probably slightly more efficient.
For clarity's sake, we will remain the instruction to move2_par/4.
---
 erts/emulator/beam/ops.tab | 71 +++++++++++++++++++++++++++++++++-------------
 1 file changed, 51 insertions(+), 20 deletions(-)

(limited to 'erts/emulator/beam/ops.tab')

diff --git a/erts/emulator/beam/ops.tab b/erts/emulator/beam/ops.tab
index 4ab08faae3..97525c3b72 100644
--- a/erts/emulator/beam/ops.tab
+++ b/erts/emulator/beam/ops.tab
@@ -303,21 +303,40 @@ move_window3 x x x y
 move_window4 x x x x y
 move_window5 x x x x x 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 Src=x D1=x | move Src=x D2=x => move_dup Src D1 D2
+move Src=x SD=x | move SD=x D=x   => move_dup Src SD D
+move Src=x D1=x | move Src=x D2=y => move_dup Src D1 D2
+move Src=y SD=x | move SD=x D=y   => move_dup Src SD D
+move Src=x SD=x | move SD=x D=y   => move_dup Src SD D
+move Src=y SD=x | move SD=x D=x   => move_dup Src SD D
 
-move X1=x X2=x | move X3=x Y1=y => move2 X1 X2 X3 Y1
+move SD=x D=x | move Src=xy SD=x => move_shift Src SD D
+move SD=y D=x | move Src=x  SD=y => move_shift Src SD D
+move SD=x D=y | move Src=x  SD=x => move_shift Src SD D
 
-move S1=x S2=x | move X1=x Y1=y => move2 S1 S2 X1 Y1
-move S1=y S2=x | move X1=x Y1=y => move2 S1 S2 X1 Y1
+# The transformations above guarantee that the source for
+# the second move is not the same as the destination for
+# the first move. That means that we can do the moves in
+# parallel (fetch both values, then store them) which could
+# be faster.
 
-move Y1=y X1=x | move S1=x D1=x => move2 Y1 X1 S1 D1
-move S1=x D1=x | move Y1=y X1=x => move2 S1 D1 Y1 X1
+move X1=x Y1=y | move X2=x Y2=y => move2_par X1 Y1 X2 Y2
+move Y1=y X1=x | move Y2=y X2=x => move2_par Y1 X1 Y2 X2
 
-move2 X1=x Y1=y X2=x Y2=y | move X3=x Y3=y => move3 X1 Y1 X2 Y2 X3 Y3
-move2 Y1=y X1=x Y2=y X2=x | move Y3=y X3=x => move3 Y1 X1 Y2 X2 Y3 X3
-move2 X1=x X2=x X3=x X4=x | move X5=x X6=x => move3 X1 X2 X3 X4 X5 X6
+move X1=x X2=x | move X3=x X4=x => move2_par X1 X2 X3 X4
+
+move X1=x X2=x | move X3=x Y1=y => move2_par X1 X2 X3 Y1
+
+move S1=x S2=x | move X1=x Y1=y => move2_par S1 S2 X1 Y1
+
+move S1=y S2=x | move X1=x Y1=y => move2_par S1 S2 X1 Y1
+
+move Y1=y X1=x | move S1=x D1=x => move2_par Y1 X1 S1 D1
+move S1=x D1=x | move Y1=y X1=x => move2_par S1 D1 Y1 X1
+
+move2_par X1=x Y1=y X2=x Y2=y | move X3=x Y3=y => move3 X1 Y1 X2 Y2 X3 Y3
+move2_par Y1=y X1=x Y2=y X2=x | move Y3=y X3=x => move3 Y1 X1 Y2 X2 Y3 X3
+move2_par X1=x X2=x X3=x X4=x | move X5=x X6=x => move3 X1 X2 X3 X4 X5 X6
 
 move C=aiq X=x==1 => move_x1 C
 move C=aiq X=x==2 => move_x2 C
@@ -325,18 +344,30 @@ 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
+%macro: move_shift MoveShift -pack
+move_shift x x x
+move_shift y x x
+move_shift x y x
+move_shift x x y
+
+%macro: move_dup MoveDup -pack
+move_dup x x x
+move_dup x x y
+move_dup y x x
+move_dup y x y
+
+%macro: move2_par Move2Par -pack
+
+move2_par x y x y
+move2_par y x y x
+move2_par x x x x
 
-move2 x x x y
+move2_par x x x y
 
-move2 x y x y
-move2 y x x y
+move2_par y x x y
 
-move2 x x y x
-move2 y x x x
+move2_par x x y x
+move2_par y x x x
 
 %macro: move3 Move3
 move3 x y x y x y
-- 
cgit v1.2.3


From 62473daf8169a04a07409f344d938bc51a4536c3 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= <bjorn@erlang.org>
Date: Thu, 25 Jun 2015 13:21:14 +0200
Subject: Introduce swap_temp/3 and swap/2

Sequences of three move instructionst that effectively swap the
contents of two registers are fairly common. We can replace them
with a swap_temp/3 instruction. The third operand is the temporary
register to be used for swapping, since the temporary register
may actually be used.

If swap_temp/3 instruction is followed by a call, the temporary
register will often (but not always) be killed by the call. If
it is killed, we can replace the swap_temp/3 instruction with a
slightly cheaper swap/2 instruction.
---
 erts/emulator/beam/ops.tab | 28 ++++++++++++++++++++++++++++
 1 file changed, 28 insertions(+)

(limited to 'erts/emulator/beam/ops.tab')

diff --git a/erts/emulator/beam/ops.tab b/erts/emulator/beam/ops.tab
index 97525c3b72..386de0a658 100644
--- a/erts/emulator/beam/ops.tab
+++ b/erts/emulator/beam/ops.tab
@@ -303,6 +303,34 @@ move_window3 x x x y
 move_window4 x x x x y
 move_window5 x x x x x y
 
+# Swap registers.
+move R1=x Tmp=x | move R2=xy R1 | move Tmp R2 => swap_temp R1 R2 Tmp
+
+swap_temp R1 R2 Tmp | line Loc | apply Live | is_killed(Tmp, Live) => \
+  swap R1 R2 | line Loc | apply Live
+
+swap_temp R1 R2 Tmp | line Loc | call Live Addr | is_killed(Tmp, Live) => \
+  swap R1 R2 | line Loc | call Live Addr
+swap_temp R1 R2 Tmp | call_only Live Addr | \
+  is_killed(Tmp, Live) => swap R1 R2 | call_only Live Addr
+swap_temp R1 R2 Tmp | call_last Live Addr D | \
+  is_killed(Tmp, Live) => swap R1 R2 | call_last Live Addr D
+
+swap_temp R1 R2 Tmp | line Loc | call_ext Live Addr | is_killed(Tmp, Live) => \
+  swap R1 R2 | line Loc | call_ext Live Addr
+swap_temp R1 R2 Tmp | line Loc | call_ext_only Live Addr | \
+  is_killed(Tmp, Live) => swap R1 R2 | line Loc | call_ext_only Live Addr
+swap_temp R1 R2 Tmp | line Loc | call_ext_last Live Addr D | \
+  is_killed(Tmp, Live) => swap R1 R2 | line Loc | call_ext_last Live Addr D
+
+%macro: swap_temp SwapTemp -pack
+swap_temp x x x
+swap_temp x y x
+
+%macro: swap Swap -pack
+swap x x
+swap x y
+
 move Src=x D1=x | move Src=x D2=x => move_dup Src D1 D2
 move Src=x SD=x | move SD=x D=x   => move_dup Src SD D
 move Src=x D1=x | move Src=x D2=y => move_dup Src D1 D2
-- 
cgit v1.2.3


From a52bd8a9ebece53c67024e3ad17a899c22cea2ec Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= <bjorn@erlang.org>
Date: Mon, 29 Jun 2015 15:56:44 +0200
Subject: Ensure that the move_call_ext_{last,only} instructions are used

Update transformations to ensure that the move_call_ext_last
and move_call_ext_last are used.
---
 erts/emulator/beam/ops.tab | 23 ++++++++---------------
 1 file changed, 8 insertions(+), 15 deletions(-)

(limited to 'erts/emulator/beam/ops.tab')

diff --git a/erts/emulator/beam/ops.tab b/erts/emulator/beam/ops.tab
index 386de0a658..18202a4200 100644
--- a/erts/emulator/beam/ops.tab
+++ b/erts/emulator/beam/ops.tab
@@ -76,17 +76,6 @@ return
 # 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 X0=x==0 | line Loc | call_ext Ar Func => \
-     line Loc | move S X0 | call_ext Ar Func
-move S X0=x==0 | line Loc | call_ext_last Ar Func=u$is_bif D => \
-     line Loc | move S X0 | call_ext_last Ar Func D
-move S X0=x==0 | line Loc | call_ext_only Ar Func=u$is_bif => \
-     line Loc | move S X0 | call_ext_only Ar Func
-move S X0=x==0 | line Loc | call Ar Func => \
-     line Loc | move S X0 | 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
@@ -94,10 +83,14 @@ move S X0=x==0 | line Loc | call Ar Func => \
 # 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
+move S X0=x==0 | line Loc | call_ext Ar Func => \
+     line Loc | move S X0 | call_ext Ar Func
+move S X0=x==0 | line Loc | call_ext_last Ar Func=u$is_not_bif D => \
+     move S X0 | call_ext_last Ar Func D
+move S X0=x==0 | line Loc | call_ext_only Ar Func=u$is_not_bif => \
+     move S X0 | call_ext_only Ar Func
+move S X0=x==0 | line Loc | call Ar Func => \
+     line Loc | move S X0 | call Ar Func
 
 line Loc | func_info M F A => func_info M F A | line Loc
 
-- 
cgit v1.2.3


From fad052472def54fff1268b21313b15bd666437c3 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= <bjorn@erlang.org>
Date: Sun, 21 Jun 2015 18:09:52 +0200
Subject: Teach beam_makeops to pack operands for move3 and move_window

It is currently only possible to pack up to 4 operands. However,
the move_window4 instrucion has 5 operands and move_window5 and
move3 instrucations have 6 operands.

Teach beam_makeops to pack instructions with 5 or 6 operands.
Also rewrite the move_window instructions in beam_emu.c to macros
to allow their operands to get packed.
---
 erts/emulator/beam/ops.tab | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

(limited to 'erts/emulator/beam/ops.tab')

diff --git a/erts/emulator/beam/ops.tab b/erts/emulator/beam/ops.tab
index 18202a4200..bce6bacd5c 100644
--- a/erts/emulator/beam/ops.tab
+++ b/erts/emulator/beam/ops.tab
@@ -292,6 +292,10 @@ move_window X1=x X2=x X3=x X4=x Y1=y Y4=y | move X5=x Y5=y | succ(Y4,Y5) => \
 move_window X1=x X2=x X3=x Y1=y Y3=y => move_window3 X1 X2 X3 Y1
 move_window X1=x X2=x X3=x X4=x Y1=y Y4=y => move_window4 X1 X2 X3 X4 Y1
 
+%macro: move_window3 MoveWindow3 -pack
+%macro: move_window4 MoveWindow4 -pack
+%macro: move_window5 MoveWindow5 -pack
+
 move_window3 x x x y
 move_window4 x x x x y
 move_window5 x x x x x y
@@ -390,7 +394,7 @@ move2_par y x x y
 move2_par x x y x
 move2_par y x x x
 
-%macro: move3 Move3
+%macro: move3 Move3 -pack
 move3 x y x y x y
 move3 y x y x y x
 move3 x x x x x x
-- 
cgit v1.2.3


From 6d6fce257cf666113e982c0887121739dff5bf1d Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= <bjorn@erlang.org>
Date: Sun, 1 Apr 2012 09:28:34 +0200
Subject: Eliminate prefetch for conditional instructions

Not pre-fetching in conditional instructions (instructions that use
-fail_action) seems to improve performance slightly.

The reason for that is that conditional instructions may jump to the
failure label, wasting the pre-fetched instruction. Another reason
is that that many conditional instructions do a function call, and
the register pressure is therefore high. Avoiding the pre-fetch
may reduce the register pressure and pontentially result in more
efficient code.
---
 erts/emulator/beam/ops.tab | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'erts/emulator/beam/ops.tab')

diff --git a/erts/emulator/beam/ops.tab b/erts/emulator/beam/ops.tab
index bce6bacd5c..3d52da77d3 100644
--- a/erts/emulator/beam/ops.tab
+++ b/erts/emulator/beam/ops.tab
@@ -477,7 +477,7 @@ i_is_ne_exact_literal f x c
 i_is_ne_exact_literal f y c
 
 is_eq_exact Lbl Y=y X=x => is_eq_exact Lbl X Y
-%macro: is_eq_exact EqualExact -fail_action
+%macro: is_eq_exact EqualExact -fail_action -pack
 is_eq_exact f x x
 is_eq_exact f x y
 is_eq_exact f s s
-- 
cgit v1.2.3


From e820b8534fd440aaf4991f3c57b8990ce836c8b0 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Gustavsson?= <bjorn@erlang.org>
Date: Sat, 4 Jul 2015 08:58:33 +0200
Subject: Speed up list matching

The combination is_non_empty_list followed by get_list is extremly
common (but not in estone_SUITE, which is why it has not been noticed
before). Therefore it is worthwile to introduce a combined
instruction.
---
 erts/emulator/beam/ops.tab | 7 +++++++
 1 file changed, 7 insertions(+)

(limited to 'erts/emulator/beam/ops.tab')

diff --git a/erts/emulator/beam/ops.tab b/erts/emulator/beam/ops.tab
index 3d52da77d3..46fefb88af 100644
--- a/erts/emulator/beam/ops.tab
+++ b/erts/emulator/beam/ops.tab
@@ -682,6 +682,13 @@ is_nonempty_list F=f x==0 | test_heap I1 I2 => is_non_empty_list_test_heap F I1
 %macro: is_non_empty_list_test_heap IsNonemptyListTestHeap -fail_action -pack
 is_non_empty_list_test_heap f I t
 
+is_nonempty_list Fail=f S=x | get_list S D1=x D2=x => \
+  is_nonempty_list_get_list Fail S D1 D2
+
+%macro: is_nonempty_list_get_list IsNonemptyListGetList -fail_action -pack
+is_nonempty_list_get_list f r x x
+is_nonempty_list_get_list f x x x
+
 %macro: is_nonempty_list IsNonemptyList -fail_action
 is_nonempty_list f x
 is_nonempty_list f y
-- 
cgit v1.2.3


From c97f3332aeddf039ee2207196229b9ff07047c72 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn-Egil=20Dahlberg?= <egil@erlang.org>
Date: Wed, 2 Dec 2015 13:17:28 +0100
Subject: erts: Add i_get_hash instruction

Calculate hashvalue in load-time for constant process dictionary gets.
---
 erts/emulator/beam/ops.tab | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

(limited to 'erts/emulator/beam/ops.tab')

diff --git a/erts/emulator/beam/ops.tab b/erts/emulator/beam/ops.tab
index 1d32e72247..86ae189c27 100644
--- a/erts/emulator/beam/ops.tab
+++ b/erts/emulator/beam/ops.tab
@@ -1118,7 +1118,7 @@ call_bif e
 bif0 u$bif:erlang:self/0 Dst=d => self Dst
 bif0 u$bif:erlang:node/0 Dst=d => node Dst
 
-bif1 Fail Bif=u$bif:erlang:get/1 Src=s Dst=d => i_get Src Dst
+bif1 Fail Bif=u$bif:erlang:get/1 Src=s Dst=d => gen_get(Src, Dst)
 
 bif2 Jump=j u$bif:erlang:element/2 S1=s S2=rxy Dst=d => gen_element(Jump, S1, S2, Dst)
 
@@ -1130,6 +1130,7 @@ 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 Bif S1 S2 Dst => i_fetch S1 S2 | i_bif2 Fail Bif Dst
 
+i_get_hash c I d
 i_get s d
 
 %macro: self Self
-- 
cgit v1.2.3


From eea5f896780e07f7ca76685061d01e7be5a7abaa Mon Sep 17 00:00:00 2001
From: Lukas Larsson <lukas@erlang-solutions.com>
Date: Thu, 11 Sep 2014 18:26:26 +0200
Subject: erts, kernel: Add os:perf_counter function

The perf_counter is a very very cheap and high resolution timer
that can be used to timestamp system events. It does not have
monoticity guarantees, but should on most OS's expose a monotonous
time.

A special instruction has been created for this counter to further
speed up fetching it.

OTP-12908
---
 erts/emulator/beam/ops.tab | 9 +++++++++
 1 file changed, 9 insertions(+)

(limited to 'erts/emulator/beam/ops.tab')

diff --git a/erts/emulator/beam/ops.tab b/erts/emulator/beam/ops.tab
index 081c4108a0..9e53b4bfcc 100644
--- a/erts/emulator/beam/ops.tab
+++ b/erts/emulator/beam/ops.tab
@@ -988,6 +988,13 @@ move Discarded x==0 | move Something x==0 => move Something x=0
 
 %endif
 
+call_ext u==0 u$func:os:perf_counter/0 => \
+    i_perf_counter
+call_ext_last u==0 u$func:os:perf_counter/0 D => \
+    i_perf_counter | deallocate_return D
+call_ext_only u==0 u$func:os:perf_counter/0 => \
+    i_perf_counter | return
+
 #
 # The general case for BIFs that have no special instructions.
 # A BIF used in the tail must be followed by a return instruction.
@@ -1027,6 +1034,8 @@ i_apply_fun_only
 
 i_hibernate
 
+i_perf_counter
+
 call_bif e
 
 #
-- 
cgit v1.2.3