3 files changed, 29 insertions, 13 deletions
diff --git a/src/cmd/compile/internal/ssa/gen/AMD64.rules b/src/cmd/compile/internal/ssa/gen/AMD64.rules
index 23b03d3080..dd71ccaf5e 100644
--- a/src/cmd/compile/internal/ssa/gen/AMD64.rules
+++ b/src/cmd/compile/internal/ssa/gen/AMD64.rules
@@ -2396,10 +2396,10 @@
 
 // Load args directly into the register class where it will be used.
 // We do this by just modifying the type of the Arg.
-(MOVQf2i <t> (Arg [off] {sym})) -> @b.Func.Entry (Arg <t> [off] {sym})
-(MOVLf2i <t> (Arg [off] {sym})) -> @b.Func.Entry (Arg <t> [off] {sym})
-(MOVQi2f <t> (Arg [off] {sym})) -> @b.Func.Entry (Arg <t> [off] {sym})
-(MOVLi2f <t> (Arg [off] {sym})) -> @b.Func.Entry (Arg <t> [off] {sym})
+(MOVQf2i <t> (Arg <u> [off] {sym})) && t.Size() == u.Size() -> @b.Func.Entry (Arg <t> [off] {sym})
+(MOVLf2i <t> (Arg <u> [off] {sym})) && t.Size() == u.Size() -> @b.Func.Entry (Arg <t> [off] {sym})
+(MOVQi2f <t> (Arg <u> [off] {sym})) && t.Size() == u.Size() -> @b.Func.Entry (Arg <t> [off] {sym})
+(MOVLi2f <t> (Arg <u> [off] {sym})) && t.Size() == u.Size() -> @b.Func.Entry (Arg <t> [off] {sym})
 
 // LEAQ is rematerializeable, so this helps to avoid register spill.
 // See issue 22947 for details
diff --git a/src/cmd/compile/internal/ssa/gen/AMD64Ops.go b/src/cmd/compile/internal/ssa/gen/AMD64Ops.go
index e54ba11da8..2f0364fb00 100644
--- a/src/cmd/compile/internal/ssa/gen/AMD64Ops.go
+++ b/src/cmd/compile/internal/ssa/gen/AMD64Ops.go
@@ -482,7 +482,7 @@ func init() {
 		{name: "MOVQi2f", argLength: 1, reg: gpfp, typ: "Float64"}, // move 64 bits from int to float reg
 		{name: "MOVQf2i", argLength: 1, reg: fpgp, typ: "UInt64"},  // move 64 bits from float to int reg
 		{name: "MOVLi2f", argLength: 1, reg: gpfp, typ: "Float32"}, // move 32 bits from int to float reg
-		{name: "MOVLf2i", argLength: 1, reg: fpgp, typ: "UInt32"},  // move 32 bits from float to int reg
+		{name: "MOVLf2i", argLength: 1, reg: fpgp, typ: "UInt32"},  // move 32 bits from float to int reg, zero extend
 
 		{name: "PXOR", argLength: 2, reg: fp21, asm: "PXOR", commutative: true, resultInArg0: true}, // exclusive or, applied to X regs for float negation.
 
diff --git a/src/cmd/compile/internal/ssa/rewriteAMD64.go b/src/cmd/compile/internal/ssa/rewriteAMD64.go
index 5ccd176e34..5c681b94b7 100644
--- a/src/cmd/compile/internal/ssa/rewriteAMD64.go
+++ b/src/cmd/compile/internal/ssa/rewriteAMD64.go
@@ -13143,8 +13143,8 @@ func rewriteValueAMD64_OpAMD64MOVLatomicload_0(v *Value) bool {
 func rewriteValueAMD64_OpAMD64MOVLf2i_0(v *Value) bool {
 	b := v.Block
 	_ = b
-	// match: (MOVLf2i <t> (Arg [off] {sym}))
-	// cond:
+	// match: (MOVLf2i <t> (Arg <u> [off] {sym}))
+	// cond: t.Size() == u.Size()
 	// result: @b.Func.Entry (Arg <t> [off] {sym})
 	for {
 		t := v.Type
@@ -13152,8 +13152,12 @@ func rewriteValueAMD64_OpAMD64MOVLf2i_0(v *Value) bool {
 		if v_0.Op != OpArg {
 			break
 		}
+		u := v_0.Type
 		off := v_0.AuxInt
 		sym := v_0.Aux
+		if !(t.Size() == u.Size()) {
+			break
+		}
 		b = b.Func.Entry
 		v0 := b.NewValue0(v.Pos, OpArg, t)
 		v.reset(OpCopy)
@@ -13167,8 +13171,8 @@ func rewriteValueAMD64_OpAMD64MOVLf2i_0(v *Value) bool {
 func rewriteValueAMD64_OpAMD64MOVLi2f_0(v *Value) bool {
 	b := v.Block
 	_ = b
-	// match: (MOVLi2f <t> (Arg [off] {sym}))
-	// cond:
+	// match: (MOVLi2f <t> (Arg <u> [off] {sym}))
+	// cond: t.Size() == u.Size()
 	// result: @b.Func.Entry (Arg <t> [off] {sym})
 	for {
 		t := v.Type
@@ -13176,8 +13180,12 @@ func rewriteValueAMD64_OpAMD64MOVLi2f_0(v *Value) bool {
 		if v_0.Op != OpArg {
 			break
 		}
+		u := v_0.Type
 		off := v_0.AuxInt
 		sym := v_0.Aux
+		if !(t.Size() == u.Size()) {
+			break
+		}
 		b = b.Func.Entry
 		v0 := b.NewValue0(v.Pos, OpArg, t)
 		v.reset(OpCopy)
@@ -15520,8 +15528,8 @@ func rewriteValueAMD64_OpAMD64MOVQatomicload_0(v *Value) bool {
 func rewriteValueAMD64_OpAMD64MOVQf2i_0(v *Value) bool {
 	b := v.Block
 	_ = b
-	// match: (MOVQf2i <t> (Arg [off] {sym}))
-	// cond:
+	// match: (MOVQf2i <t> (Arg <u> [off] {sym}))
+	// cond: t.Size() == u.Size()
 	// result: @b.Func.Entry (Arg <t> [off] {sym})
 	for {
 		t := v.Type
@@ -15529,8 +15537,12 @@ func rewriteValueAMD64_OpAMD64MOVQf2i_0(v *Value) bool {
 		if v_0.Op != OpArg {
 			break
 		}
+		u := v_0.Type
 		off := v_0.AuxInt
 		sym := v_0.Aux
+		if !(t.Size() == u.Size()) {
+			break
+		}
 		b = b.Func.Entry
 		v0 := b.NewValue0(v.Pos, OpArg, t)
 		v.reset(OpCopy)
@@ -15544,8 +15556,8 @@ func rewriteValueAMD64_OpAMD64MOVQf2i_0(v *Value) bool {
 func rewriteValueAMD64_OpAMD64MOVQi2f_0(v *Value) bool {
 	b := v.Block
 	_ = b
-	// match: (MOVQi2f <t> (Arg [off] {sym}))
-	// cond:
+	// match: (MOVQi2f <t> (Arg <u> [off] {sym}))
+	// cond: t.Size() == u.Size()
 	// result: @b.Func.Entry (Arg <t> [off] {sym})
 	for {
 		t := v.Type
@@ -15553,8 +15565,12 @@ func rewriteValueAMD64_OpAMD64MOVQi2f_0(v *Value) bool {
 		if v_0.Op != OpArg {
 			break
 		}
+		u := v_0.Type
 		off := v_0.AuxInt
 		sym := v_0.Aux
+		if !(t.Size() == u.Size()) {
+			break
+		}
 		b = b.Func.Entry
 		v0 := b.NewValue0(v.Pos, OpArg, t)
 		v.reset(OpCopy)