diff --git a/src/cmd/compile/internal/loong64/ssa.go b/src/cmd/compile/internal/loong64/ssa.go
index 1b1fdfdc71..7cdf5637f2 100644
--- a/src/cmd/compile/internal/loong64/ssa.go
+++ b/src/cmd/compile/internal/loong64/ssa.go
@@ -468,6 +468,8 @@ func ssaGenValue(s *ssagen.State, v *ssa.Value) {
 		p.To.Sym = ssagen.BoundsCheckFunc[v.AuxInt]
 		s.UseArgs(16) // space used in callee args area by assembly stubs
 	case ssa.OpLOONG64LoweredAtomicLoad8, ssa.OpLOONG64LoweredAtomicLoad32, ssa.OpLOONG64LoweredAtomicLoad64:
+		// MOVB	(Rarg0), Rout
+		// DBAR	0x14
 		as := loong64.AMOVV
 		switch v.Op {
 		case ssa.OpLOONG64LoweredAtomicLoad8:
@@ -475,13 +477,15 @@ func ssaGenValue(s *ssagen.State, v *ssa.Value) {
 		case ssa.OpLOONG64LoweredAtomicLoad32:
 			as = loong64.AMOVW
 		}
-		s.Prog(loong64.ADBAR)
 		p := s.Prog(as)
 		p.From.Type = obj.TYPE_MEM
 		p.From.Reg = v.Args[0].Reg()
 		p.To.Type = obj.TYPE_REG
 		p.To.Reg = v.Reg0()
-		s.Prog(loong64.ADBAR)
+		p1 := s.Prog(loong64.ADBAR)
+		p1.From.Type = obj.TYPE_CONST
+		p1.From.Offset = 0x14
+
 	case ssa.OpLOONG64LoweredAtomicStore8, ssa.OpLOONG64LoweredAtomicStore32, ssa.OpLOONG64LoweredAtomicStore64:
 		as := loong64.AMOVV
 		switch v.Op {
diff --git a/src/internal/runtime/atomic/atomic_loong64.s b/src/internal/runtime/atomic/atomic_loong64.s
index 1812cb95fd..9bed8654c8 100644
--- a/src/internal/runtime/atomic/atomic_loong64.s
+++ b/src/internal/runtime/atomic/atomic_loong64.s
@@ -319,38 +319,30 @@ TEXT ·Oruintptr(SB), NOSPLIT, $0-24
 // uint32 internal∕runtime∕atomic·Load(uint32 volatile* ptr)
 TEXT ·Load(SB),NOSPLIT|NOFRAME,$0-12
 	MOVV	ptr+0(FP), R19
-	DBAR
 	MOVWU	0(R19), R19
-	DBAR
+	DBAR	$0x14	// LoadAcquire barrier
 	MOVW	R19, ret+8(FP)
 	RET
 
 // uint8 internal∕runtime∕atomic·Load8(uint8 volatile* ptr)
 TEXT ·Load8(SB),NOSPLIT|NOFRAME,$0-9
 	MOVV	ptr+0(FP), R19
-	DBAR
 	MOVBU	0(R19), R19
-	DBAR
+	DBAR	$0x14
 	MOVB	R19, ret+8(FP)
 	RET
 
 // uint64 internal∕runtime∕atomic·Load64(uint64 volatile* ptr)
 TEXT ·Load64(SB),NOSPLIT|NOFRAME,$0-16
 	MOVV	ptr+0(FP), R19
-	DBAR
 	MOVV	0(R19), R19
-	DBAR
+	DBAR	$0x14
 	MOVV	R19, ret+8(FP)
 	RET
 
 // void *internal∕runtime∕atomic·Loadp(void *volatile *ptr)
 TEXT ·Loadp(SB),NOSPLIT|NOFRAME,$0-16
-	MOVV	ptr+0(FP), R19
-	DBAR
-	MOVV	0(R19), R19
-	DBAR
-	MOVV	R19, ret+8(FP)
-	RET
+	JMP     ·Load64(SB)
 
 // uint32 internal∕runtime∕atomic·LoadAcq(uint32 volatile* ptr)
 TEXT ·LoadAcq(SB),NOSPLIT|NOFRAME,$0-12
diff --git a/src/internal/runtime/atomic/bench_test.go b/src/internal/runtime/atomic/bench_test.go
index 798431cf72..6e3f14cbe4 100644
--- a/src/internal/runtime/atomic/bench_test.go
+++ b/src/internal/runtime/atomic/bench_test.go
@@ -43,6 +43,14 @@ func BenchmarkAtomicStore(b *testing.B) {
 	}
 }
 
+func BenchmarkAtomicLoad8(b *testing.B) {
+	var x uint8
+	sink = &x
+	for i := 0; i < b.N; i++ {
+		atomic.Load8(&x)
+	}
+}
+
 func BenchmarkAnd8(b *testing.B) {
 	var x [512]uint8 // give byte its own cache line
 	sink = &x