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.

3 files changed, 48 insertions, 0 deletions

diff --git a/erts/emulator/beam/beam_emu.c b/erts/emulator/beam/beam_emu.c
index 2c10e7ae7c..4e6e7aa48a 100644
--- a/erts/emulator/beam/beam_emu.c
+++ b/erts/emulator/beam/beam_emu.c
@@ -545,6 +545,20 @@ void** beam_ops;
    HTOP += 2;                      \
   } while (0)
 
+#define Swap(R1, R2)				\
+  do {						\
+    Eterm V = R1;				\
+    R1 = R2;					\
+    R2 = V;					\
+  } while (0)
+
+#define SwapTemp(R1, R2, Tmp)			\
+  do {						\
+    Eterm V = R1;				\
+    R1 = R2;					\
+    R2 = Tmp = V;				\
+  } while (0)
+
 #define Move(Src, Dst, Store)      \
    do {                            \
        Eterm term = (Src);         \
diff --git a/erts/emulator/beam/beam_load.c b/erts/emulator/beam/beam_load.c
index 871c2b1f55..1e3690fcd6 100644
--- a/erts/emulator/beam/beam_load.c
+++ b/erts/emulator/beam/beam_load.c
@@ -2702,6 +2702,12 @@ same_label(LoaderState* stp, GenOpArg Target, GenOpArg Label)
 	Target.val == Label.val;
 }
 
+static int
+is_killed(LoaderState* stp, GenOpArg Reg, GenOpArg Live)
+{
+    return Reg.type == TAG_x && Live.type == TAG_u &&
+	Live.val <= Reg.val;
+}
 
 /*
  * Generate an instruction for element/2.
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