hipe_x86: LSRA for SSE2

There is little point offering LSRA for x86 if we're still going to call
hipe_graph_coloring_regalloc for the floats. In particular, all
allocators except LSRA allocates an N^2 interference matrix, making them
unusable for really large functions.

5 files changed, 116 insertions, 43 deletions

diff --git a/lib/hipe/amd64/Makefile b/lib/hipe/amd64/Makefile
index 8dc2af2679..ea3559b7e6 100644
--- a/lib/hipe/amd64/Makefile
+++ b/lib/hipe/amd64/Makefile
@@ -57,10 +57,10 @@ MODULES=hipe_amd64_assemble \
 	hipe_amd64_ra_naive \
 	hipe_amd64_ra_postconditions \
 	hipe_amd64_ra_sse2_postconditions \
-	hipe_amd64_ra_x87_ls \
 	hipe_amd64_registers \
 	hipe_amd64_spill_restore \
 	hipe_amd64_x87 \
+	hipe_amd64_sse2 \
 	hipe_rtl_to_amd64
 
 ERL_FILES=$(MODULES:%=%.erl)
@@ -125,10 +125,10 @@ $(EBIN)/hipe_amd64_ra_ls.beam: ../main/hipe.hrl ../x86/hipe_x86_ra_ls.erl
 $(EBIN)/hipe_amd64_ra_naive.beam: ../main/hipe.hrl ../x86/hipe_x86.hrl ../x86/hipe_x86_ra_naive.erl
 $(EBIN)/hipe_amd64_ra_postconditions.beam: ../main/hipe.hrl ../x86/hipe_x86.hrl ../x86/hipe_x86_ra_postconditions.erl
 $(EBIN)/hipe_amd64_ra_sse2_postconditions.beam: ../main/hipe.hrl
-$(EBIN)/hipe_amd64_ra_x87_ls.beam: ../main/hipe.hrl ../x86/hipe_x86_ra_x87_ls.erl
 $(EBIN)/hipe_amd64_registers.beam: ../rtl/hipe_literals.hrl
 $(EBIN)/hipe_amd64_spill_restore.beam: ../main/hipe.hrl ../x86/hipe_x86.hrl ../flow/cfg.hrl ../x86/hipe_x86_spill_restore.erl
 $(EBIN)/hipe_amd64_x87.beam: ../x86/hipe_x86_x87.erl
+$(EBIN)/hipe_amd64_sse2.beam: ../main/hipe.hrl ../x86/hipe_x86.hrl
 $(EBIN)/hipe_rtl_to_amd64.beam: ../x86/hipe_rtl_to_x86.erl ../rtl/hipe_rtl.hrl
 
 $(TARGET_FILES): ../x86/hipe_x86.hrl ../misc/hipe_consttab.hrl
diff --git a/lib/hipe/amd64/hipe_amd64_ra_sse2_postconditions.erl b/lib/hipe/amd64/hipe_amd64_ra_sse2_postconditions.erl
index b1f7bd7572..8483d2d0d5 100644
--- a/lib/hipe/amd64/hipe_amd64_ra_sse2_postconditions.erl
+++ b/lib/hipe/amd64/hipe_amd64_ra_sse2_postconditions.erl
@@ -21,7 +21,7 @@
 
 -module(hipe_amd64_ra_sse2_postconditions).
 
--export([check_and_rewrite/2]).
+-export([check_and_rewrite/2, check_and_rewrite/3]).
 
 -include("../x86/hipe_x86.hrl").
 -define(HIPE_INSTRUMENT_COMPILER, true).
@@ -30,39 +30,43 @@
 
 
 check_and_rewrite(AMD64Defun, Coloring) ->
+  check_and_rewrite(AMD64Defun, Coloring, 'normal').
+
+check_and_rewrite(AMD64Defun, Coloring, Strategy) ->
   %%io:format("Converting\n"),
   TempMap = hipe_temp_map:cols2tuple(Coloring,hipe_amd64_specific_sse2),
   %%io:format("Rewriting\n"),
   #defun{code=Code0} = AMD64Defun,
-  {Code1, DidSpill} = do_insns(Code0, TempMap, [], false),
+  {Code1, DidSpill} = do_insns(Code0, TempMap, Strategy, [], false),
   {AMD64Defun#defun{code=Code1, var_range={0, hipe_gensym:get_var(x86)}}, 
    DidSpill}.
 
-do_insns([I|Insns], TempMap, Accum, DidSpill0) ->
-  {NewIs, DidSpill1} = do_insn(I, TempMap),
-  do_insns(Insns, TempMap, lists:reverse(NewIs, Accum), DidSpill0 or DidSpill1);
-do_insns([], _TempMap, Accum, DidSpill) ->
+do_insns([I|Insns], TempMap, Strategy, Accum, DidSpill0) ->
+  {NewIs, DidSpill1} = do_insn(I, TempMap, Strategy),
+  do_insns(Insns, TempMap, Strategy, lists:reverse(NewIs, Accum),
+	   DidSpill0 or DidSpill1);
+do_insns([], _TempMap, _Strategy, Accum, DidSpill) ->
   {lists:reverse(Accum), DidSpill}.
 
-do_insn(I, TempMap) ->	% Insn -> {Insn list, DidSpill}
+do_insn(I, TempMap, Strategy) ->	% Insn -> {Insn list, DidSpill}
   case I of
     #fmove{} ->
-      do_fmove(I, TempMap);
+      do_fmove(I, TempMap, Strategy);
     #fp_unop{} ->
-      do_fp_unop(I, TempMap);
+      do_fp_unop(I, TempMap, Strategy);
     #fp_binop{} ->
-      do_fp_binop(I, TempMap);
+      do_fp_binop(I, TempMap, Strategy);
     _ ->
       %% All non sse2 ops
       {[I], false}
   end.
 
 %%% Fix an fp_binop.
-do_fp_binop(I, TempMap) ->
+do_fp_binop(I, TempMap, Strategy) ->
   #fp_binop{src=Src,dst=Dst} = I,
   case is_mem_opnd(Dst, TempMap) of
     true ->
-      Tmp = clone(Dst),
+      Tmp = clone(Dst, Strategy),
       {[#fmove{src=Dst, dst=Tmp},
 	I#fp_binop{src=Src,dst=Tmp},
 	#fmove{src=Tmp,dst=Dst}],
@@ -71,11 +75,11 @@ do_fp_binop(I, TempMap) ->
       {[I], false}
   end.
 
-do_fp_unop(I, TempMap) ->
+do_fp_unop(I, TempMap, Strategy) ->
   #fp_unop{arg=Arg} = I,
   case is_mem_opnd(Arg, TempMap) of
     true ->
-      Tmp = clone(Arg),
+      Tmp = clone(Arg, Strategy),
       {[#fmove{src=Arg, dst=Tmp},
 	I#fp_unop{arg=Tmp},
 	#fmove{src=Tmp,dst=Arg}],
@@ -85,7 +89,7 @@ do_fp_unop(I, TempMap) ->
   end.
 
 %%% Fix an fmove op.
-do_fmove(I, TempMap) ->
+do_fmove(I, TempMap, Strategy) ->
   #fmove{src=Src,dst=Dst} = I,
   case
     (is_mem_opnd(Src, TempMap) andalso is_mem_opnd(Dst, TempMap))
@@ -93,7 +97,7 @@ do_fmove(I, TempMap) ->
     orelse ((not is_float_temp(Src)) andalso is_mem_opnd(Dst, TempMap))
   of
     true ->
-      Tmp = spill_temp(double),
+      Tmp = spill_temp(double, Strategy),
       {[#fmove{src=Src, dst=Tmp},I#fmove{src=Tmp,dst=Dst}],
        true};
     false ->
@@ -177,15 +181,21 @@ is_mem_opnd(Opnd, TempMap) ->
 
 %%% Make Reg a clone of Dst (attach Dst's type to Reg).
 
-clone(Dst) ->
+clone(Dst, Strategy) ->
   Type =
     case Dst of
       #x86_mem{} -> hipe_x86:mem_type(Dst);
       #x86_temp{} -> hipe_x86:temp_type(Dst)
     end,
-  spill_temp(Type).
-
-spill_temp(Type) ->
+  spill_temp(Type, Strategy).
+
+spill_temp(Type, 'normal') ->
+  hipe_x86:mk_new_temp(Type);
+spill_temp(double, 'linearscan') ->
+  hipe_x86:mk_temp(hipe_amd64_specific_sse2:temp0(), double);
+spill_temp(Type, 'linearscan') when Type =:= tagged; Type =/= untagged ->
+  %% We can make a new temp here since we have yet to allocate registers for
+  %% these types
   hipe_x86:mk_new_temp(Type).
 
 %%% Make a certain reg into a clone of Dst
diff --git a/lib/hipe/amd64/hipe_amd64_ra_x87_ls.erl b/lib/hipe/amd64/hipe_amd64_ra_x87_ls.erl
deleted file mode 100644
index 6da3f44cd3..0000000000
--- a/lib/hipe/amd64/hipe_amd64_ra_x87_ls.erl
+++ /dev/null
@@ -1,21 +0,0 @@
-%%
-%% %CopyrightBegin%
-%% 
-%% Copyright Ericsson AB 2004-2016. All Rights Reserved.
-%% 
-%% Licensed under the Apache License, Version 2.0 (the "License");
-%% you may not use this file except in compliance with the License.
-%% You may obtain a copy of the License at
-%%
-%%     http://www.apache.org/licenses/LICENSE-2.0
-%%
-%% Unless required by applicable law or agreed to in writing, software
-%% distributed under the License is distributed on an "AS IS" BASIS,
-%% WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-%% See the License for the specific language governing permissions and
-%% limitations under the License.
-%% 
-%% %CopyrightEnd%
-%%
-
--include("../x86/hipe_x86_ra_x87_ls.erl").
diff --git a/lib/hipe/amd64/hipe_amd64_registers.erl b/lib/hipe/amd64/hipe_amd64_registers.erl
index 780c2cc547..ada5311453 100644
--- a/lib/hipe/amd64/hipe_amd64_registers.erl
+++ b/lib/hipe/amd64/hipe_amd64_registers.erl
@@ -52,6 +52,7 @@
  	 tailcall_clobbered/0,
  	 temp0/0,
 	 temp1/0,
+	 sse2_temp0/0,
 	 %% fixed/0,
 	 wordsize/0
 	]).
@@ -107,6 +108,8 @@ heap_limit_offset() -> ?P_HP_LIMIT.
 -define(TEMP0, ?R14).
 -define(TEMP1, ?R13).
 
+-define(SSE2_TEMP0, 00).
+
 -define(PROC_POINTER, ?RBP).
 
 reg_name(R) ->
@@ -204,6 +207,8 @@ allocatable() ->
 allocatable_sse2() ->
   [00,01,02,03,04,05,06,07,08,09,10,11,12,13,14,15]. %% xmm0 - xmm15
 
+sse2_temp0() -> ?SSE2_TEMP0.
+
 allocatable_x87() ->
   [0,1,2,3,4,5,6].
 
diff --git a/lib/hipe/amd64/hipe_amd64_sse2.erl b/lib/hipe/amd64/hipe_amd64_sse2.erl
new file mode 100644
index 0000000000..df78941be5
--- /dev/null
+++ b/lib/hipe/amd64/hipe_amd64_sse2.erl
@@ -0,0 +1,79 @@
+%% -*- erlang-indent-level: 2 -*-
+%%
+%% %CopyrightBegin%
+%% 
+%% Copyright Ericsson AB 2016. All Rights Reserved.
+%% 
+%% Licensed under the Apache License, Version 2.0 (the "License");
+%% you may not use this file except in compliance with the License.
+%% You may obtain a copy of the License at
+%%
+%%     http://www.apache.org/licenses/LICENSE-2.0
+%%
+%% Unless required by applicable law or agreed to in writing, software
+%% distributed under the License is distributed on an "AS IS" BASIS,
+%% WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+%% See the License for the specific language governing permissions and
+%% limitations under the License.
+%% 
+%% %CopyrightEnd%
+%%
+%% Fix {mem, mem} floating point operations that result from linear scan
+%% allocated floats.
+
+-module(hipe_amd64_sse2).
+
+-export([map/1]).
+
+-include("../x86/hipe_x86.hrl").
+-include("../main/hipe.hrl").
+
+%%----------------------------------------------------------------------
+
+map(Defun = #defun{code=Code0}) ->
+  Code1 = do_insns(Code0, []),
+  Defun#defun{code=Code1}.
+
+do_insns([I|Insns], Accum) ->
+  NewIs = do_insn(I),
+  do_insns(Insns, lists:reverse(NewIs, Accum));
+do_insns([], Accum) ->
+  lists:reverse(Accum).
+
+do_insn(I) ->
+  case I of
+    #fp_binop{} -> do_fp_binop(I);
+    #fmove{}    -> do_fmove(I);
+    _           -> [I]
+  end.
+
+do_fp_binop(I = #fp_binop{src=Src0,dst=Dst}) ->
+  {FixSrc, Src} = fix_binary(Src0, Dst),
+  FixSrc ++ [I#fp_binop{src=Src}].
+
+do_fmove(I = #fmove{src=Src0,dst=Dst}) ->
+  {FixSrc, Src} = fix_binary(Src0, Dst),
+  FixSrc ++ [I#fmove{src=Src}].
+
+fix_binary(Src0, Dst) ->
+  case is_mem_opnd(Src0) of
+    false -> {[], Src0};
+    true ->
+      case is_mem_opnd(Dst) of
+	false -> {[], Src0};
+	true ->
+	  Src1 = spill_temp(),
+	  {[hipe_x86:mk_fmove(Src0, Src1)], Src1}
+      end
+  end.
+
+is_mem_opnd(#x86_fpreg{reg=Reg}) ->
+  not hipe_amd64_registers:is_precoloured_sse2(Reg);
+is_mem_opnd(#x86_temp{type=double, reg=Reg}) ->
+  not hipe_amd64_registers:is_precoloured_sse2(Reg);
+is_mem_opnd(#x86_temp{type=_, reg=Reg}) ->
+  not hipe_amd64_registers:is_precoloured(Reg);
+is_mem_opnd(#x86_mem{}) -> true.
+
+spill_temp() ->
+  hipe_x86:mk_temp(hipe_amd64_registers:sse2_temp0(), double).