cmd/compile: add generic rules to eliminate some unnecessary stores

Eliminates stores of values that have just been loaded from the same
location. Handles the common case where there are up to 3 intermediate
stores to non-overlapping struct fields.

For example the loads and stores of x.a, x.b and x.d in the following
function are now removed:

type T struct {
	a, b, c, d int
}

func f(x *T) {
	y := *x
	y.c += 8
	*x = y
}

Before this CL (s390x):

TEXT    "".f(SB)
	MOVD    "".x(R15), R5
	MOVD    (R5), R1
	MOVD    8(R5), R2
	MOVD    16(R5), R0
	MOVD    24(R5), R4
	ADD     $8, R0, R3
	STMG    R1, R4, (R5)
	RET

After this CL (s390x):

TEXT	"".f(SB)
	MOVD	"".x(R15), R1
	MOVD	16(R1), R0
	ADD	$8, R0, R0
	MOVD	R0, 16(R1)
	RET

In total these rules are triggered ~5091 times during all.bash,
which is broken down as:

Intermediate stores | Triggered
--------------------+----------
0                   | 1434
1                   | 2508
2                   | 888
3                   | 261
--------------------+----------

Change-Id: Ia4721ae40146aceec1fdd3e65b0e9283770bfba5
Reviewed-on: https://go-review.googlesource.com/38793
Run-TryBot: Michael Munday <munday@ca.ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>

3 files changed, 246 insertions, 9 deletions

diff --git a/src/cmd/compile/internal/ssa/gen/generic.rules b/src/cmd/compile/internal/ssa/gen/generic.rules
index 322de4de11..b8d7381420 100644
--- a/src/cmd/compile/internal/ssa/gen/generic.rules
+++ b/src/cmd/compile/internal/ssa/gen/generic.rules
@@ -691,6 +691,33 @@
 // Load of store of same address, with compatibly typed value and same size
 (Load <t1> p1 (Store {t2} p2 x _)) && isSamePtr(p1,p2) && t1.Compare(x.Type) == types.CMPeq && t1.Size() == t2.(*types.Type).Size() -> x
 
+// Eliminate stores of values that have just been loaded from the same location.
+// We also handle the common case where there are some intermediate stores to non-overlapping struct fields.
+(Store {t1} p1 (Load <t2> p2 mem) mem) &&
+	isSamePtr(p1, p2) &&
+	t2.Size() == t1.(*types.Type).Size() -> mem
+(Store {t1} (OffPtr [o1] p1) (Load <t2> (OffPtr [o1] p2) oldmem) mem:(Store {t3} (OffPtr [o3] p3) _ oldmem)) &&
+	isSamePtr(p1, p2) &&
+	isSamePtr(p1, p3) &&
+	t2.Size() == t1.(*types.Type).Size() &&
+	!overlap(o1, t2.Size(), o3, t3.(*types.Type).Size()) -> mem
+(Store {t1} (OffPtr [o1] p1) (Load <t2> (OffPtr [o1] p2) oldmem) mem:(Store {t3} (OffPtr [o3] p3) _ (Store {t4} (OffPtr [o4] p4) _ oldmem))) &&
+	isSamePtr(p1, p2) &&
+	isSamePtr(p1, p3) &&
+	isSamePtr(p1, p4) &&
+	t2.Size() == t1.(*types.Type).Size() &&
+	!overlap(o1, t2.Size(), o3, t3.(*types.Type).Size()) &&
+	!overlap(o1, t2.Size(), o4, t4.(*types.Type).Size()) -> mem
+(Store {t1} (OffPtr [o1] p1) (Load <t2> (OffPtr [o1] p2) oldmem) mem:(Store {t3} (OffPtr [o3] p3) _ (Store {t4} (OffPtr [o4] p4) _ (Store {t5} (OffPtr [o5] p5) _ oldmem)))) &&
+	isSamePtr(p1, p2) &&
+	isSamePtr(p1, p3) &&
+	isSamePtr(p1, p4) &&
+	isSamePtr(p1, p5) &&
+	t2.Size() == t1.(*types.Type).Size() &&
+	!overlap(o1, t2.Size(), o3, t3.(*types.Type).Size()) &&
+	!overlap(o1, t2.Size(), o4, t4.(*types.Type).Size()) &&
+	!overlap(o1, t2.Size(), o5, t5.(*types.Type).Size()) -> mem
+
 // Collapse OffPtr
 (OffPtr (OffPtr p [b]) [a]) -> (OffPtr p [a+b])
 (OffPtr p [0]) && v.Type.Compare(p.Type) == types.CMPeq -> p
diff --git a/src/cmd/compile/internal/ssa/rewrite.go b/src/cmd/compile/internal/ssa/rewrite.go
index f69ffc8c5c..06595586c1 100644
--- a/src/cmd/compile/internal/ssa/rewrite.go
+++ b/src/cmd/compile/internal/ssa/rewrite.go
@@ -628,3 +628,15 @@ func isARMImmRot(v uint32) bool {
 
 	return false
 }
+
+// overlap reports whether the ranges given by the given offset and
+// size pairs overlap.
+func overlap(offset1, size1, offset2, size2 int64) bool {
+	if offset1 >= offset2 && offset2+size2 > offset1 {
+		return true
+	}
+	if offset2 >= offset1 && offset1+size1 > offset2 {
+		return true
+	}
+	return false
+}
diff --git a/src/cmd/compile/internal/ssa/rewritegeneric.go b/src/cmd/compile/internal/ssa/rewritegeneric.go
index f41d3fa151..9f67d00404 100644
--- a/src/cmd/compile/internal/ssa/rewritegeneric.go
+++ b/src/cmd/compile/internal/ssa/rewritegeneric.go
@@ -20447,10 +20447,206 @@ func rewriteValuegeneric_OpSqrt_0(v *Value) bool {
 func rewriteValuegeneric_OpStore_0(v *Value) bool {
 	b := v.Block
 	_ = b
-	config := b.Func.Config
-	_ = config
 	fe := b.Func.fe
 	_ = fe
+	// match: (Store {t1} p1 (Load <t2> p2 mem) mem)
+	// cond: isSamePtr(p1, p2) && 	t2.Size() == t1.(*types.Type).Size()
+	// result: mem
+	for {
+		t1 := v.Aux
+		p1 := v.Args[0]
+		v_1 := v.Args[1]
+		if v_1.Op != OpLoad {
+			break
+		}
+		t2 := v_1.Type
+		p2 := v_1.Args[0]
+		mem := v_1.Args[1]
+		if mem != v.Args[2] {
+			break
+		}
+		if !(isSamePtr(p1, p2) && t2.Size() == t1.(*types.Type).Size()) {
+			break
+		}
+		v.reset(OpCopy)
+		v.Type = mem.Type
+		v.AddArg(mem)
+		return true
+	}
+	// match: (Store {t1} (OffPtr [o1] p1) (Load <t2> (OffPtr [o1] p2) oldmem) mem:(Store {t3} (OffPtr [o3] p3) _ oldmem))
+	// cond: isSamePtr(p1, p2) && 	isSamePtr(p1, p3) && 	t2.Size() == t1.(*types.Type).Size() && 	!overlap(o1, t2.Size(), o3, t3.(*types.Type).Size())
+	// result: mem
+	for {
+		t1 := v.Aux
+		v_0 := v.Args[0]
+		if v_0.Op != OpOffPtr {
+			break
+		}
+		o1 := v_0.AuxInt
+		p1 := v_0.Args[0]
+		v_1 := v.Args[1]
+		if v_1.Op != OpLoad {
+			break
+		}
+		t2 := v_1.Type
+		v_1_0 := v_1.Args[0]
+		if v_1_0.Op != OpOffPtr {
+			break
+		}
+		if v_1_0.AuxInt != o1 {
+			break
+		}
+		p2 := v_1_0.Args[0]
+		oldmem := v_1.Args[1]
+		mem := v.Args[2]
+		if mem.Op != OpStore {
+			break
+		}
+		t3 := mem.Aux
+		mem_0 := mem.Args[0]
+		if mem_0.Op != OpOffPtr {
+			break
+		}
+		o3 := mem_0.AuxInt
+		p3 := mem_0.Args[0]
+		if oldmem != mem.Args[2] {
+			break
+		}
+		if !(isSamePtr(p1, p2) && isSamePtr(p1, p3) && t2.Size() == t1.(*types.Type).Size() && !overlap(o1, t2.Size(), o3, t3.(*types.Type).Size())) {
+			break
+		}
+		v.reset(OpCopy)
+		v.Type = mem.Type
+		v.AddArg(mem)
+		return true
+	}
+	// match: (Store {t1} (OffPtr [o1] p1) (Load <t2> (OffPtr [o1] p2) oldmem) mem:(Store {t3} (OffPtr [o3] p3) _ (Store {t4} (OffPtr [o4] p4) _ oldmem)))
+	// cond: isSamePtr(p1, p2) && 	isSamePtr(p1, p3) && 	isSamePtr(p1, p4) && 	t2.Size() == t1.(*types.Type).Size() && 	!overlap(o1, t2.Size(), o3, t3.(*types.Type).Size()) && 	!overlap(o1, t2.Size(), o4, t4.(*types.Type).Size())
+	// result: mem
+	for {
+		t1 := v.Aux
+		v_0 := v.Args[0]
+		if v_0.Op != OpOffPtr {
+			break
+		}
+		o1 := v_0.AuxInt
+		p1 := v_0.Args[0]
+		v_1 := v.Args[1]
+		if v_1.Op != OpLoad {
+			break
+		}
+		t2 := v_1.Type
+		v_1_0 := v_1.Args[0]
+		if v_1_0.Op != OpOffPtr {
+			break
+		}
+		if v_1_0.AuxInt != o1 {
+			break
+		}
+		p2 := v_1_0.Args[0]
+		oldmem := v_1.Args[1]
+		mem := v.Args[2]
+		if mem.Op != OpStore {
+			break
+		}
+		t3 := mem.Aux
+		mem_0 := mem.Args[0]
+		if mem_0.Op != OpOffPtr {
+			break
+		}
+		o3 := mem_0.AuxInt
+		p3 := mem_0.Args[0]
+		mem_2 := mem.Args[2]
+		if mem_2.Op != OpStore {
+			break
+		}
+		t4 := mem_2.Aux
+		mem_2_0 := mem_2.Args[0]
+		if mem_2_0.Op != OpOffPtr {
+			break
+		}
+		o4 := mem_2_0.AuxInt
+		p4 := mem_2_0.Args[0]
+		if oldmem != mem_2.Args[2] {
+			break
+		}
+		if !(isSamePtr(p1, p2) && isSamePtr(p1, p3) && isSamePtr(p1, p4) && t2.Size() == t1.(*types.Type).Size() && !overlap(o1, t2.Size(), o3, t3.(*types.Type).Size()) && !overlap(o1, t2.Size(), o4, t4.(*types.Type).Size())) {
+			break
+		}
+		v.reset(OpCopy)
+		v.Type = mem.Type
+		v.AddArg(mem)
+		return true
+	}
+	// match: (Store {t1} (OffPtr [o1] p1) (Load <t2> (OffPtr [o1] p2) oldmem) mem:(Store {t3} (OffPtr [o3] p3) _ (Store {t4} (OffPtr [o4] p4) _ (Store {t5} (OffPtr [o5] p5) _ oldmem))))
+	// cond: isSamePtr(p1, p2) && 	isSamePtr(p1, p3) && 	isSamePtr(p1, p4) && 	isSamePtr(p1, p5) && 	t2.Size() == t1.(*types.Type).Size() && 	!overlap(o1, t2.Size(), o3, t3.(*types.Type).Size()) && 	!overlap(o1, t2.Size(), o4, t4.(*types.Type).Size()) && 	!overlap(o1, t2.Size(), o5, t5.(*types.Type).Size())
+	// result: mem
+	for {
+		t1 := v.Aux
+		v_0 := v.Args[0]
+		if v_0.Op != OpOffPtr {
+			break
+		}
+		o1 := v_0.AuxInt
+		p1 := v_0.Args[0]
+		v_1 := v.Args[1]
+		if v_1.Op != OpLoad {
+			break
+		}
+		t2 := v_1.Type
+		v_1_0 := v_1.Args[0]
+		if v_1_0.Op != OpOffPtr {
+			break
+		}
+		if v_1_0.AuxInt != o1 {
+			break
+		}
+		p2 := v_1_0.Args[0]
+		oldmem := v_1.Args[1]
+		mem := v.Args[2]
+		if mem.Op != OpStore {
+			break
+		}
+		t3 := mem.Aux
+		mem_0 := mem.Args[0]
+		if mem_0.Op != OpOffPtr {
+			break
+		}
+		o3 := mem_0.AuxInt
+		p3 := mem_0.Args[0]
+		mem_2 := mem.Args[2]
+		if mem_2.Op != OpStore {
+			break
+		}
+		t4 := mem_2.Aux
+		mem_2_0 := mem_2.Args[0]
+		if mem_2_0.Op != OpOffPtr {
+			break
+		}
+		o4 := mem_2_0.AuxInt
+		p4 := mem_2_0.Args[0]
+		mem_2_2 := mem_2.Args[2]
+		if mem_2_2.Op != OpStore {
+			break
+		}
+		t5 := mem_2_2.Aux
+		mem_2_2_0 := mem_2_2.Args[0]
+		if mem_2_2_0.Op != OpOffPtr {
+			break
+		}
+		o5 := mem_2_2_0.AuxInt
+		p5 := mem_2_2_0.Args[0]
+		if oldmem != mem_2_2.Args[2] {
+			break
+		}
+		if !(isSamePtr(p1, p2) && isSamePtr(p1, p3) && isSamePtr(p1, p4) && isSamePtr(p1, p5) && t2.Size() == t1.(*types.Type).Size() && !overlap(o1, t2.Size(), o3, t3.(*types.Type).Size()) && !overlap(o1, t2.Size(), o4, t4.(*types.Type).Size()) && !overlap(o1, t2.Size(), o5, t5.(*types.Type).Size())) {
+			break
+		}
+		v.reset(OpCopy)
+		v.Type = mem.Type
+		v.AddArg(mem)
+		return true
+	}
 	// match: (Store _ (StructMake0) mem)
 	// cond:
 	// result: mem
@@ -20633,6 +20829,15 @@ func rewriteValuegeneric_OpStore_0(v *Value) bool {
 		v.AddArg(mem)
 		return true
 	}
+	return false
+}
+func rewriteValuegeneric_OpStore_10(v *Value) bool {
+	b := v.Block
+	_ = b
+	config := b.Func.Config
+	_ = config
+	fe := b.Func.fe
+	_ = fe
 	// match: (Store {t} dst (Load src mem) (VarDef {x} mem))
 	// cond: !fe.CanSSA(t.(*types.Type))
 	// result: (Move {t} [t.(*types.Type).Size()] dst src (VarDef {x} mem))
@@ -20729,13 +20934,6 @@ func rewriteValuegeneric_OpStore_0(v *Value) bool {
 		v.AddArg(mem)
 		return true
 	}
-	return false
-}
-func rewriteValuegeneric_OpStore_10(v *Value) bool {
-	b := v.Block
-	_ = b
-	config := b.Func.Config
-	_ = config
 	// match: (Store (OffPtr (Load (OffPtr [c] (SP)) mem)) x mem)
 	// cond: isConstZero(x) 	&& mem.Op == OpStaticCall 	&& isSameSym(mem.Aux, "runtime.newobject") 	&& c == config.ctxt.FixedFrameSize() + config.RegSize
 	// result: mem