diff --git a/internal/sm2ec/p256_asm_s390x.s b/internal/sm2ec/p256_asm_s390x.s
index a82bc55..cd9a6d8 100644
--- a/internal/sm2ec/p256_asm_s390x.s
+++ b/internal/sm2ec/p256_asm_s390x.s
@@ -328,7 +328,6 @@ loop_select:
 
 #define ZER   V6
 #define SEL1  V7
-#define SEL2  V8
 #define CAR1  V9
 #define CAR2  V10
 #define RED1  V11
@@ -345,8 +344,7 @@ TEXT ·p256FromMont(SB), NOSPLIT, $0
 	MOVD  $p256<>+0x00(SB), CPOOL
 	VL    16(CPOOL), PL
 	VL    0(CPOOL), PH
-	VL    48(CPOOL), SEL2
-	VL    64(CPOOL), SEL1
+	VL    48(CPOOL), SEL1
 
 	VL   (0*16)(x_ptr), T0
 	VPDI $0x4, T0, T0, T0
@@ -449,7 +447,6 @@ TEXT ·p256FromMont(SB), NOSPLIT, $0
 #undef TT1
 #undef ZER
 #undef SEL1
-#undef SEL2
 #undef CAR1
 #undef CAR2
 #undef RED1
diff --git a/internal/sm2ec/sm2p256_asm_s390x_test.go b/internal/sm2ec/sm2p256_asm_s390x_test.go
index b0baef8..aacc61f 100644
--- a/internal/sm2ec/sm2p256_asm_s390x_test.go
+++ b/internal/sm2ec/sm2p256_asm_s390x_test.go
@@ -66,7 +66,7 @@ func testP256OrderReduce(v, expected *big.Int, t *testing.T) {
 	fromBig((*[4]uint64)(val), v)
 	p256OrdReduce(val)
 	if ordElmToBigInt(val).Cmp(expected) != 0 {
-		t.Errorf("p256OrdReduce failed for %x", v.Bytes())
+		t.Errorf("p256OrdReduce failed for %x, expected %x", v.Bytes(), expected.Bytes())
 	}
 }
 
@@ -75,9 +75,12 @@ func TestP256OrderReduce(t *testing.T) {
 	for i := 0; i < 20; i++ {
 		bigVal := big.NewInt(int64(i))
 		testP256OrderReduce(bigVal, bigVal, t)
-		bigVal = new(big.Int).Sub(p, big.NewInt(int64(i)))
-		testP256OrderReduce(bigVal, bigVal, t)
 		bigVal = new(big.Int).Add(p, big.NewInt(int64(i)))
 		testP256OrderReduce(bigVal, big.NewInt(int64(i)), t)
 	}
+	testP256OrderReduce(p, big.NewInt(0), t)
+	for i := 1; i < 20; i++ {
+		bigVal := new(big.Int).Sub(p, big.NewInt(int64(i)))
+		testP256OrderReduce(bigVal, bigVal, t)
+	}
 }