cmd/compile: add intrinsic for reading caller's pc

First step towards removing the mandatory argument for
getcallerpc, which solves certain problems for the runtime.
This might also slightly improve performance.

Intrinsic enabled on 386, amd64, amd64p32,
runtime asm implementation removed on those architectures.

Now-superfluous argument remains in getcallerpc signature
(for a future CL; non-386/amd64 asm funcs ignore it).

Added getcallerpc to the "not a real function" test
in dcl.go, that story is a little odd with respect to
unexported functions but that is not this CL.

Fixes #17327.

Change-Id: I5df1ad91f27ee9ac1f0dd88fa48f1329d6306c3e
Reviewed-on: https://go-review.googlesource.com/31851
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>

src/cmd/compile/internal/amd64/ssa.go

@@ -795,6 +795,15 @@ func ssaGenValue(s *gc.SSAGenState, v *ssa.Value) {
 		}
 	case ssa.OpAMD64CALLstatic, ssa.OpAMD64CALLclosure, ssa.OpAMD64CALLinter:
 		s.Call(v)
+
+	case ssa.OpAMD64LoweredGetCallerPC:
+		p := s.Prog(x86.AMOVQ)
+		p.From.Type = obj.TYPE_MEM
+		p.From.Offset = -8 // PC is stored 8 bytes below first parameter.
+		p.From.Name = obj.NAME_PARAM
+		p.To.Type = obj.TYPE_REG
+		p.To.Reg = v.Reg()
+
 	case ssa.OpAMD64NEGQ, ssa.OpAMD64NEGL,
 		ssa.OpAMD64BSWAPQ, ssa.OpAMD64BSWAPL,
 		ssa.OpAMD64NOTQ, ssa.OpAMD64NOTL:

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

@@ -1077,9 +1077,9 @@ func makefuncsym(s *types.Sym) {
 	if s.IsBlank() {
 		return
 	}
-	if compiling_runtime && (s.Name == "getg" || s.Name == "getclosureptr") {
+	if compiling_runtime && (s.Name == "getg" || s.Name == "getclosureptr" || s.Name == "getcallerpc") {
-		// runtime.getg() and getclosureptr are not real functions and so do not
+		// runtime.getg(), getclosureptr(), and getcallerpc() are
-		// get funcsyms.
+		// not real functions and so do not get funcsyms.
 		return
 	}
 	if _, existed := s.Pkg.LookupOK(funcsymname(s)); !existed {

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

@@ -2584,6 +2584,11 @@ func init() {
 		},
 		all...)
 
+	addF("runtime", "getcallerpc",
+		func(s *state, n *Node, args []*ssa.Value) *ssa.Value {
+			return s.newValue0(ssa.OpGetCallerPC, s.f.Config.Types.Uintptr)
+		}, sys.AMD64, sys.I386)
+
 	/******** runtime/internal/sys ********/
 	addF("runtime/internal/sys", "Ctz32",
 		func(s *state, n *Node, args []*ssa.Value) *ssa.Value {
@@ -2614,7 +2619,6 @@ func init() {
 			return s.newValue1(ssa.OpSelect0, types.Types[TUINT32], v)
 		},
 		sys.AMD64, sys.ARM64, sys.S390X, sys.MIPS, sys.PPC64)
-
 	addF("runtime/internal/atomic", "Load64",
 		func(s *state, n *Node, args []*ssa.Value) *ssa.Value {
 			v := s.newValue2(ssa.OpAtomicLoad64, types.NewTuple(types.Types[TUINT64], types.TypeMem), args[0], s.mem())

src/cmd/compile/internal/ssa/gen/386.rules

@@ -383,6 +383,7 @@
 (NilCheck ptr mem) -> (LoweredNilCheck ptr mem)
 (GetG mem) -> (LoweredGetG mem)
 (GetClosurePtr) -> (LoweredGetClosurePtr)
+(GetCallerPC) -> (LoweredGetCallerPC)
 (Addr {sym} base) -> (LEAL {sym} base)
 
 // block rewrites

src/cmd/compile/internal/ssa/gen/386Ops.go

@@ -440,6 +440,11 @@ func init() {
 		// and sorts it to the very beginning of the block to prevent other
 		// use of DX (the closure pointer)
 		{name: "LoweredGetClosurePtr", reg: regInfo{outputs: []regMask{buildReg("DX")}}},
+		// LoweredGetCallerPC evaluates to the PC to which its "caller" will return.
+		// I.e., if f calls g "calls" getcallerpc,
+		// the result should be the PC within f that g will return to.
+		// See runtime/stubs.go for a more detailed discussion.
+		{name: "LoweredGetCallerPC", reg: gp01},
 		//arg0=ptr,arg1=mem, returns void.  Faults if ptr is nil.
 		{name: "LoweredNilCheck", argLength: 2, reg: regInfo{inputs: []regMask{gpsp}}, clobberFlags: true, nilCheck: true, faultOnNilArg0: true},
 

src/cmd/compile/internal/ssa/gen/AMD64.rules

@@ -482,6 +482,7 @@
 (NilCheck ptr mem) -> (LoweredNilCheck ptr mem)
 (GetG mem) -> (LoweredGetG mem)
 (GetClosurePtr) -> (LoweredGetClosurePtr)
+(GetCallerPC) -> (LoweredGetCallerPC)
 (Addr {sym} base) && config.PtrSize == 8 -> (LEAQ {sym} base)
 (Addr {sym} base) && config.PtrSize == 4 -> (LEAL {sym} base)
 

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

@@ -540,6 +540,11 @@ func init() {
 		// and sorts it to the very beginning of the block to prevent other
 		// use of DX (the closure pointer)
 		{name: "LoweredGetClosurePtr", reg: regInfo{outputs: []regMask{buildReg("DX")}}},
+		// LoweredGetCallerPC evaluates to the PC to which its "caller" will return.
+		// I.e., if f calls g "calls" getcallerpc,
+		// the result should be the PC within f that g will return to.
+		// See runtime/stubs.go for a more detailed discussion.
+		{name: "LoweredGetCallerPC", reg: gp01},
 		//arg0=ptr,arg1=mem, returns void.  Faults if ptr is nil.
 		{name: "LoweredNilCheck", argLength: 2, reg: regInfo{inputs: []regMask{gpsp}}, clobberFlags: true, nilCheck: true, faultOnNilArg0: true},
 

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

@@ -378,6 +378,7 @@ var genericOps = []opData{
 	// Pseudo-ops
 	{name: "GetG", argLength: 1}, // runtime.getg() (read g pointer). arg0=mem
 	{name: "GetClosurePtr"},      // get closure pointer from dedicated register
+	{name: "GetCallerPC"},        // for getcallerpc intrinsic
 
 	// Indexing operations
 	{name: "PtrIndex", argLength: 2},             // arg0=ptr, arg1=index. Computes ptr+sizeof(*v.type)*index, where index is extended to ptrwidth type

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

@@ -396,6 +396,7 @@ const (
 	Op386InvertFlags
 	Op386LoweredGetG
 	Op386LoweredGetClosurePtr
+	Op386LoweredGetCallerPC
 	Op386LoweredNilCheck
 	Op386MOVLconvert
 	Op386FlagEQ
@@ -653,6 +654,7 @@ const (
 	OpAMD64InvertFlags
 	OpAMD64LoweredGetG
 	OpAMD64LoweredGetClosurePtr
+	OpAMD64LoweredGetCallerPC
 	OpAMD64LoweredNilCheck
 	OpAMD64MOVQconvert
 	OpAMD64MOVLconvert
@@ -1905,6 +1907,7 @@ const (
 	OpNilCheck
 	OpGetG
 	OpGetClosurePtr
+	OpGetCallerPC
 	OpPtrIndex
 	OpOffPtr
 	OpSliceMake
@@ -4286,6 +4289,15 @@ var opcodeTable = [...]opInfo{
 			},
 		},
 	},
+	{
+		name:   "LoweredGetCallerPC",
+		argLen: 0,
+		reg: regInfo{
+			outputs: []outputInfo{
+				{0, 239}, // AX CX DX BX BP SI DI
+			},
+		},
+	},
 	{
 		name:           "LoweredNilCheck",
 		argLen:         2,
@@ -7914,6 +7926,15 @@ var opcodeTable = [...]opInfo{
 			},
 		},
 	},
+	{
+		name:   "LoweredGetCallerPC",
+		argLen: 0,
+		reg: regInfo{
+			outputs: []outputInfo{
+				{0, 65519}, // AX CX DX BX BP SI DI R8 R9 R10 R11 R12 R13 R14 R15
+			},
+		},
+	},
 	{
 		name:           "LoweredNilCheck",
 		argLen:         2,
@@ -22834,6 +22855,11 @@ var opcodeTable = [...]opInfo{
 		argLen:  0,
 		generic: true,
 	},
+	{
+		name:    "GetCallerPC",
+		argLen:  0,
+		generic: true,
+	},
 	{
 		name:    "PtrIndex",
 		argLen:  2,

src/cmd/compile/internal/ssa/rewrite386.go

@@ -329,6 +329,8 @@ func rewriteValue386(v *Value) bool {
 		return rewriteValue386_OpGeq8_0(v)
 	case OpGeq8U:
 		return rewriteValue386_OpGeq8U_0(v)
+	case OpGetCallerPC:
+		return rewriteValue386_OpGetCallerPC_0(v)
 	case OpGetClosurePtr:
 		return rewriteValue386_OpGetClosurePtr_0(v)
 	case OpGetG:
@@ -15052,6 +15054,15 @@ func rewriteValue386_OpGeq8U_0(v *Value) bool {
 		return true
 	}
 }
+func rewriteValue386_OpGetCallerPC_0(v *Value) bool {
+	// match: (GetCallerPC)
+	// cond:
+	// result: (LoweredGetCallerPC)
+	for {
+		v.reset(Op386LoweredGetCallerPC)
+		return true
+	}
+}
 func rewriteValue386_OpGetClosurePtr_0(v *Value) bool {
 	// match: (GetClosurePtr)
 	// cond:

src/cmd/compile/internal/ssa/rewriteAMD64.go

@@ -563,6 +563,8 @@ func rewriteValueAMD64(v *Value) bool {
 		return rewriteValueAMD64_OpGeq8_0(v)
 	case OpGeq8U:
 		return rewriteValueAMD64_OpGeq8U_0(v)
+	case OpGetCallerPC:
+		return rewriteValueAMD64_OpGetCallerPC_0(v)
 	case OpGetClosurePtr:
 		return rewriteValueAMD64_OpGetClosurePtr_0(v)
 	case OpGetG:
@@ -40851,6 +40853,15 @@ func rewriteValueAMD64_OpGeq8U_0(v *Value) bool {
 		return true
 	}
 }
+func rewriteValueAMD64_OpGetCallerPC_0(v *Value) bool {
+	// match: (GetCallerPC)
+	// cond:
+	// result: (LoweredGetCallerPC)
+	for {
+		v.reset(OpAMD64LoweredGetCallerPC)
+		return true
+	}
+}
 func rewriteValueAMD64_OpGetClosurePtr_0(v *Value) bool {
 	// match: (GetClosurePtr)
 	// cond:

src/cmd/compile/internal/x86/ssa.go

@@ -666,6 +666,15 @@ func ssaGenValue(s *gc.SSAGenState, v *ssa.Value) {
 			q.To.Type = obj.TYPE_REG
 			q.To.Reg = r
 		}
+
+	case ssa.Op386LoweredGetCallerPC:
+		p := s.Prog(x86.AMOVL)
+		p.From.Type = obj.TYPE_MEM
+		p.From.Offset = -4 // PC is stored 4 bytes below first parameter.
+		p.From.Name = obj.NAME_PARAM
+		p.To.Type = obj.TYPE_REG
+		p.To.Reg = v.Reg()
+
 	case ssa.Op386CALLstatic, ssa.Op386CALLclosure, ssa.Op386CALLinter:
 		s.Call(v)
 	case ssa.Op386NEGL,

src/runtime/asm_386.s

@@ -849,12 +849,6 @@ TEXT runtime·stackcheck(SB), NOSPLIT, $0-0
 	INT	$3
 	RET
 
-TEXT runtime·getcallerpc(SB),NOSPLIT,$4-8
-	MOVL	argp+0(FP),AX		// addr of first arg
-	MOVL	-4(AX),AX		// get calling pc
-	MOVL	AX, ret+4(FP)
-	RET
-
 // func cputicks() int64
 TEXT runtime·cputicks(SB),NOSPLIT,$0-8
 	CMPB	runtime·support_sse2(SB), $1

src/runtime/asm_amd64.s

@@ -833,12 +833,6 @@ TEXT runtime·stackcheck(SB), NOSPLIT, $0-0
 	INT	$3
 	RET
 
-TEXT runtime·getcallerpc(SB),NOSPLIT,$8-16
-	MOVQ	argp+0(FP),AX		// addr of first arg
-	MOVQ	-8(AX),AX		// get calling pc
-	MOVQ	AX, ret+8(FP)
-	RET
-
 // func cputicks() int64
 TEXT runtime·cputicks(SB),NOSPLIT,$0-0
 	CMPB	runtime·lfenceBeforeRdtsc(SB), $1

src/runtime/asm_amd64p32.s

@@ -559,12 +559,6 @@ TEXT runtime·stackcheck(SB), NOSPLIT, $0-0
 	MOVL	0, AX
 	RET
 
-TEXT runtime·getcallerpc(SB),NOSPLIT,$8-12
-	MOVL	argp+0(FP),AX		// addr of first arg
-	MOVL	-8(AX),AX		// get calling pc
-	MOVL	AX, ret+8(FP)
-	RET
-
 // int64 runtime·cputicks(void)
 TEXT runtime·cputicks(SB),NOSPLIT,$0-0
 	RDTSC

src/runtime/stubs.go

@@ -200,7 +200,9 @@ func publicationBarrier()
 // getcallersp returns the stack pointer (SP) of its caller's caller.
 // For both, the argp must be a pointer to the caller's first function argument.
 // The implementation may or may not use argp, depending on
-// the architecture.
+// the architecture. The implementation may be a compiler
+// intrinsic; there is not necessarily code implementing this
+// on every platform.
 //
 // For example:
 //