diff --git a/include/mcl/fp_tower.hpp b/include/mcl/fp_tower.hpp
index f4c4343..c236428 100644
--- a/include/mcl/fp_tower.hpp
+++ b/include/mcl/fp_tower.hpp
@@ -19,8 +19,9 @@ struct BnT {
 		static inline size_t getUnitSize() { return Fp::op_.N * 2; }
 		void dump() const
 		{
-			for (size_t i = 0; i < getUnitSize(); i++) {
-				printf("%016llx ", (long long)v_[i]);
+			const size_t n = getUnitSize();
+			for (size_t i = 0; i < n; i++) {
+				printf("%016llx ", (long long)v_[n - 1 - i]);
 			}
 			printf("\n");
 		}
diff --git a/src/fp.cpp b/src/fp.cpp
index a0a3995..3f11bd5 100644
--- a/src/fp.cpp
+++ b/src/fp.cpp
@@ -138,7 +138,6 @@ struct OpeFunc {
 	*/
 	static inline void fpDbl_addPC(Unit *z, const Unit *x, const Unit *y, const Unit *p)
 	{
-puts("BBB");
 		Unit ret[N * 2 + 2]; // not N + 1
 		Unit pDbl[N * 2];
 		mpz_t mz, mx, my, mp;
diff --git a/src/short.txt b/src/short.txt
index b340f79..75c0ad2 100644
--- a/src/short.txt
+++ b/src/short.txt
@@ -56,7 +56,8 @@ entry:
 	%t0 = add i$(b2u) %x1, %y1 ; x + y = [H:L]
 	%t1 = trunc i$(b2u) %t0 to i$(bit) ; t1 = L
 	store i$(bit) %t1, i$(bit)* %pz ; store L
-	%t2 = shl i$(b2u) %t0, $(bit)
+
+	%t2 = lshr i$(b2u) %t0, $(bit)
 	%H = trunc i$(b2u) %t2 to i$(bu)
 
 	%p = load i$(bit)* %pp
diff --git a/test/fp_tower_test.cpp b/test/fp_tower_test.cpp
index ab74fb1..091508e 100644
--- a/test/fp_tower_test.cpp
+++ b/test/fp_tower_test.cpp
@@ -92,7 +92,6 @@ void testFpDbl()
 				mcl::Gmp::addMod(mo, mx, my, mp);
 				z.getMpz(mz);
 				CYBOZU_TEST_EQUAL(mz, mo);
-
 				mcl::Gmp::subMod(mo, mx, my, mp);
 				FpDbl::sub(z, x, y);
 				z.getMpz(mz);
@@ -105,8 +104,8 @@ void testFpDbl()
 void test(const char *p)
 {
 	printf("prime=%s\n", p);
-//	Fp::setModulo(p);
-	Fp::setModulo(p, 0, mcl::fp::FP_LLVM); // QQQ
+	Fp::setModulo(p);
+//	Fp::setModulo(p, 0, mcl::fp::FP_LLVM); // QQQ
 	testFp2();
 	if (Fp::getBitSize() <= 256) {
 		testFpDbl();