Index: src/scrypt.m
==================================================================
--- src/scrypt.m
+++ src/scrypt.m
@@ -30,11 +30,11 @@
 of_salsa20_8_core(uint32_t buffer[16])
 {
 	uint32_t tmp[16];
 
 	for (uint_fast8_t i = 0; i < 16; i++)
-		tmp[i] = buffer[i];
+		tmp[i] = OF_BSWAP32_IF_BE(buffer[i]);
 
 	for (uint_fast8_t i = 0; i < 8; i+= 2) {
 		tmp[ 4] ^= OF_ROL(tmp[ 0] + tmp[12],  7);
 		tmp[ 8] ^= OF_ROL(tmp[ 4] + tmp[ 0],  9);
 		tmp[12] ^= OF_ROL(tmp[ 8] + tmp[ 4], 13);
@@ -68,11 +68,12 @@
 		tmp[14] ^= OF_ROL(tmp[13] + tmp[12], 13);
 		tmp[15] ^= OF_ROL(tmp[14] + tmp[13], 18);
 	}
 
 	for (uint_fast8_t i = 0; i < 16; i++)
-		buffer[i] += tmp[i];
+		buffer[i] = OF_BSWAP32_IF_BE(OF_BSWAP32_IF_BE(buffer[i]) +
+		    tmp[i]);
 
 	of_explicit_memset(tmp, 0, sizeof(tmp));
 }
 
 void
@@ -121,11 +122,12 @@
 		memcpy(tmp2 + i * 32 * blockSize, tmp, 128 * blockSize);
 		of_scrypt_block_mix(tmp, tmp2 + i * 32 * blockSize, blockSize);
 	}
 
 	for (size_t i = 0; i < costFactor; i++) {
-		uint32_t j = tmp[(2 * blockSize - 1) * 16] & (costFactor - 1);
+		uint32_t j = OF_BSWAP32_IF_BE(tmp[(2 * blockSize - 1) * 16]) &
+		    (costFactor - 1);
 
 		for (size_t k = 0; k < 32 * blockSize; k++)
 			tmp[k] ^= tmp2[j * 32 * blockSize + k];
 
 		of_scrypt_block_mix(buffer, tmp, blockSize);