cmd/compile: some SSA cleanup

Do some easy TODOs.
Move a bunch of other TODOs into bugs.

Change-Id: Iaba9dad6221a2af11b3cbcc512875f4a85842873
Reviewed-on: https://go-review.googlesource.com/20114
Run-TryBot: Todd Neal <todd@tneal.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Todd Neal <todd@tneal.org>

src/cmd/compile/internal/gc/ssa.go

@@ -1935,7 +1935,7 @@ func (s *state) expr(n *Node) *ssa.Value {
 		for !data.Type.IsPtr() {
 			switch {
 			case data.Type.IsArray():
-				data = s.newValue2(ssa.OpArrayIndex, data.Type.Elem(), data, s.constInt(Types[TINT], 0))
+				data = s.newValue1I(ssa.OpArrayIndex, data.Type.Elem(), 0, data)
 			case data.Type.IsStruct():
 				for i := data.Type.NumFields() - 1; i >= 0; i-- {
 					f := data.Type.FieldType(i)

src/cmd/compile/internal/ssa/TODO

 This is a list of things that need to be worked on.  It will hopefully
 be complete soon.
 
-Coverage
---------
-
 Correctness
 -----------
 - Debugging info (check & fix as much as we can)
@@ -14,24 +11,12 @@ Optimizations (better compiled code)
 - More strength reduction: multiply -> shift/add combos (Worth doing?)
 - Add a value range propagation pass (for bounds elim & bitwidth reduction)
 - Make dead store pass inter-block
-- redundant CMP in sequences like this:
-  SUBQ $8, AX
-  CMP AX, $0
-  JEQ ...
 - If there are a lot of MOVQ $0, ..., then load
   0 into a register and use the register as the source instead.
 - Allow arrays of length 1 (or longer, with all constant indexes?) to be SSAable.
-- Figure out how to make PARAMOUT variables ssa-able.
-  They need to get spilled automatically at end-of-function somehow.
 - If strings are being passed around without being interpreted (ptr
   and len fields being accessed) pass them in xmm registers?
   Same for interfaces?
-- OpArrayIndex should take its index in AuxInt, not a full value.
-- remove FLAGS from REP instruction clobbers
-- (x86) Combine loads into other ops
-  Note that this is challenging for ops that generate flags
-  because flagalloc wants to move those instructions around for
-  flag regeneration.
 - Non-constant rotate detection.
 - Do 0 <= x && x < n with one unsigned compare
 - nil-check removal in indexed load/store case:
@@ -44,17 +29,13 @@ Optimizations (better compiled code)
 
 Optimizations (better compiler)
 -------------------------------
-- Smaller Value.Type (int32 or ptr)?  Get rid of types altogether?
 - OpStore uses 3 args.  Increase the size of Value.argstorage to 3?
 - Use a constant cache for OpConstNil, OpConstInterface, OpConstSlice, maybe OpConstString
 - Handle signed division overflow and sign extension earlier
-- Implement 64 bit const division with high multiply, maybe in the frontend?
-- Add bit widths to complex ops
 
 Regalloc
 --------
 - Make less arch-dependent
-- Allow return values to be ssa-able
 - Handle 2-address instructions
 - Make liveness analysis non-quadratic
 

src/cmd/compile/internal/ssa/gen/AMD64Ops.go

@@ -438,7 +438,7 @@ func init() {
 			argLength: 4,
 			reg: regInfo{
 				inputs:   []regMask{buildReg("DI"), buildReg("CX"), buildReg("AX")},
-				clobbers: buildReg("DI CX FLAGS"),
+				clobbers: buildReg("DI CX"),
 			},
 		},
 

src/cmd/compile/internal/ssa/gen/generic.rules

@@ -400,7 +400,7 @@
 
 // indexing operations
 // Note: bounds check has already been done
-(ArrayIndex (Load ptr mem) idx) && b == v.Args[0].Block -> (Load (PtrIndex <v.Type.PtrTo()> ptr idx) mem)
+(ArrayIndex <t> [0] (Load ptr mem)) -> @v.Args[0].Block (Load <t> ptr mem)
 (PtrIndex <t> ptr idx) && config.PtrSize == 4 -> (AddPtr ptr (Mul32 <config.fe.TypeInt()> idx (Const32 <config.fe.TypeInt()> [t.Elem().Size()])))
 (PtrIndex <t> ptr idx) && config.PtrSize == 8 -> (AddPtr ptr (Mul64 <config.fe.TypeInt()> idx (Const64 <config.fe.TypeInt()> [t.Elem().Size()])))
 

src/cmd/compile/internal/ssa/gen/genericOps.go

@@ -335,9 +335,9 @@ var genericOps = []opData{
 	{name: "GetClosurePtr"},      // get closure pointer from dedicated register
 
 	// Indexing operations
-	{name: "ArrayIndex", argLength: 2},           // arg0=array, arg1=index.  Returns a[i]
-	{name: "PtrIndex", argLength: 2},             // arg0=ptr, arg1=index. Computes ptr+sizeof(*v.type)*index, where index is extended to ptrwidth type
-	{name: "OffPtr", argLength: 1, aux: "Int64"}, // arg0 + auxint (arg0 and result are pointers)
+	{name: "ArrayIndex", aux: "Int64", argLength: 1}, // arg0=array, auxint=index.  Returns a[i]
+	{name: "PtrIndex", argLength: 2},                 // arg0=ptr, arg1=index. Computes ptr+sizeof(*v.type)*index, where index is extended to ptrwidth type
+	{name: "OffPtr", argLength: 1, aux: "Int64"},     // arg0 + auxint (arg0 and result are pointers)
 
 	// Slices
 	{name: "SliceMake", argLength: 3},                // arg0=ptr, arg1=len, arg2=cap

src/cmd/compile/internal/ssa/opGen.go

@@ -3689,7 +3689,7 @@ var opcodeTable = [...]opInfo{
 				{1, 2},   // .CX
 				{2, 1},   // .AX
 			},
-			clobbers: 8589934722, // .CX .DI .FLAGS
+			clobbers: 130, // .CX .DI
 		},
 	},
 	{
@@ -5110,7 +5110,8 @@ var opcodeTable = [...]opInfo{
 	},
 	{
 		name:    "ArrayIndex",
-		argLen:  2,
+		auxType: auxInt64,
+		argLen:  1,
 		generic: true,
 	},
 	{

src/cmd/compile/internal/ssa/rewritegeneric.go

@@ -1142,25 +1142,25 @@ func rewriteValuegeneric_OpArg(v *Value, config *Config) bool {
 func rewriteValuegeneric_OpArrayIndex(v *Value, config *Config) bool {
 	b := v.Block
 	_ = b
-	// match: (ArrayIndex (Load ptr mem) idx)
-	// cond: b == v.Args[0].Block
-	// result: (Load (PtrIndex <v.Type.PtrTo()> ptr idx) mem)
+	// match: (ArrayIndex <t> [0] (Load ptr mem))
+	// cond:
+	// result: @v.Args[0].Block (Load <t> ptr mem)
 	for {
+		t := v.Type
+		if v.AuxInt != 0 {
+			break
+		}
 		if v.Args[0].Op != OpLoad {
 			break
 		}
 		ptr := v.Args[0].Args[0]
 		mem := v.Args[0].Args[1]
-		idx := v.Args[1]
-		if !(b == v.Args[0].Block) {
-			break
-		}
-		v.reset(OpLoad)
-		v0 := b.NewValue0(v.Line, OpPtrIndex, v.Type.PtrTo())
-		v0.AddArg(ptr)
-		v0.AddArg(idx)
+		b = v.Args[0].Block
+		v0 := b.NewValue0(v.Line, OpLoad, t)
+		v.reset(OpCopy)
 		v.AddArg(v0)
-		v.AddArg(mem)
+		v0.AddArg(ptr)
+		v0.AddArg(mem)
 		return true
 	}
 	return false