diff --git a/crypto_kem/firesaber/META.yml b/crypto_kem/firesaber/META.yml
index def16e46..d1781339 100644
--- a/crypto_kem/firesaber/META.yml
+++ b/crypto_kem/firesaber/META.yml
@@ -14,9 +14,9 @@ principal-submitters:
   - Frederik Vercauteren
 implementations:
     - name: clean
-      version: https://github.com/KULeuven-COSIC/SABER/commit/509cc5ec3a7e12a751ccdd2ef5bd6e54e00bd350
+      version: https://github.com/KULeuven-COSIC/SABER/tree/509cc5ec3a7e12a751ccdd2ef5bd6e54e00bd350 via https://github.com/jschanck/package-pqclean/tree/b53a47b5/saber
     - name: avx2
-      version: https://github.com/KULeuven-COSIC/SABER/commit/509cc5ec3a7e12a751ccdd2ef5bd6e54e00bd350
+      version: https://github.com/KULeuven-COSIC/SABER/tree/509cc5ec3a7e12a751ccdd2ef5bd6e54e00bd350 via https://github.com/jschanck/package-pqclean/tree/b53a47b5/saber
       supported_platforms:
           - architecture: x86_64
             operating_systems:
diff --git a/crypto_kem/firesaber/avx2/SABER_indcpa.c b/crypto_kem/firesaber/avx2/SABER_indcpa.c
index ab017224..625a3f6b 100644
--- a/crypto_kem/firesaber/avx2/SABER_indcpa.c
+++ b/crypto_kem/firesaber/avx2/SABER_indcpa.c
@@ -66,7 +66,7 @@ static void GenSecret(uint16_t r[SABER_K][SABER_N], const uint8_t *seed) {
 }
 
 //********************************matrix-vector mul routines*****************************************************
-static void matrix_vector_mul(__m256i a1_avx_combined[NUM_POLY][NUM_POLY][AVX_N1], __m256i b_bucket[NUM_POLY][SCHB_N * 4], __m256i res_avx[NUM_POLY][AVX_N1], int isTranspose) {
+static void matrix_vector_mul(__m256i res_avx[NUM_POLY][AVX_N1], __m256i a1_avx_combined[NUM_POLY][NUM_POLY][AVX_N1], __m256i b_bucket[NUM_POLY][SCHB_N * 4], int isTranspose) {
     int64_t i, j;
 
     __m256i c_bucket[2 * SCM_SIZE * 4]; //Holds results for 9 Karatsuba at a time
@@ -86,7 +86,7 @@ static void matrix_vector_mul(__m256i a1_avx_combined[NUM_POLY][NUM_POLY][AVX_N1
 
 }
 
-static void vector_vector_mul(__m256i a_avx[NUM_POLY][AVX_N1], __m256i b_bucket[NUM_POLY][SCHB_N * 4], __m256i res_avx[AVX_N1]) {
+static void vector_vector_mul(__m256i res_avx[AVX_N1], __m256i a_avx[NUM_POLY][AVX_N1], __m256i b_bucket[NUM_POLY][SCHB_N * 4]) {
 
     int64_t i;
 
@@ -162,7 +162,7 @@ void PQCLEAN_FIRESABER_AVX2_indcpa_kem_keypair(uint8_t *pk, uint8_t *sk) {
     for (j = 0; j < NUM_POLY; j++) {
         TC_eval(sk_avx[j], b_bucket[j]);
     }
-    matrix_vector_mul(a_avx, b_bucket, res_avx, 1);// Matrix-vector multiplication; Matrix in transposed order
+    matrix_vector_mul(res_avx, a_avx, b_bucket, 1);// Matrix-vector multiplication; Matrix in transposed order
 
     // Now truncation
 
@@ -259,7 +259,7 @@ void PQCLEAN_FIRESABER_AVX2_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_DE
     for (j = 0; j < NUM_POLY; j++) {
         TC_eval(sk_avx[j], b_bucket[j]);
     }
-    matrix_vector_mul(a_avx, b_bucket, res_avx, 0);// Matrix-vector multiplication; Matrix in normal order
+    matrix_vector_mul(res_avx, a_avx, b_bucket, 0);// Matrix-vector multiplication; Matrix in normal order
 
     // Now truncation
 
@@ -302,7 +302,7 @@ void PQCLEAN_FIRESABER_AVX2_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_DE
 
     // vector-vector scalar multiplication with mod p
 
-    vector_vector_mul(pkcl_avx, b_bucket, vprime_avx);
+    vector_vector_mul(vprime_avx, pkcl_avx, b_bucket);
 
     // Computation of v'+h1
     for (i = 0; i < SABER_N / 16; i++) { //adding h1
@@ -392,7 +392,7 @@ void PQCLEAN_FIRESABER_AVX2_indcpa_kem_dec(uint8_t m[SABER_KEYBYTES], const uint
         TC_eval(sksv_avx[j], b_bucket[j]);
     }
 
-    vector_vector_mul(pksv_avx, b_bucket, v_avx);
+    vector_vector_mul(v_avx, pksv_avx, b_bucket);
 
     for (i = 0; i < SABER_N / 16; i++) {
         _mm256_maskstore_epi32 ((int *)(message_dec_unpacked + i * 16), _mm256_set1_epi32(-1), v_avx[i]);
diff --git a/crypto_kem/firesaber/clean/SABER_indcpa.c b/crypto_kem/firesaber/clean/SABER_indcpa.c
index 76156e79..342eb7ca 100644
--- a/crypto_kem/firesaber/clean/SABER_indcpa.c
+++ b/crypto_kem/firesaber/clean/SABER_indcpa.c
@@ -17,7 +17,7 @@ void PQCLEAN_FIRESABER_CLEAN_indcpa_kem_keypair(uint8_t pk[SABER_INDCPA_PUBLICKE
 
     uint8_t seed_A[SABER_SEEDBYTES];
     uint8_t seed_s[SABER_NOISE_SEEDBYTES];
-    int i, j;
+    size_t i, j;
 
     randombytes(seed_A, SABER_SEEDBYTES);
     shake128(seed_A, SABER_SEEDBYTES, seed_A, SABER_SEEDBYTES); // for not revealing system RNG state
@@ -45,7 +45,7 @@ void PQCLEAN_FIRESABER_CLEAN_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_D
     uint16_t vp[SABER_N] = {0};
     uint16_t mp[SABER_N];
     uint16_t b[SABER_L][SABER_N];
-    int i, j;
+    size_t i, j;
     const uint8_t *seed_A = pk + SABER_POLYVECCOMPRESSEDBYTES;
 
     PQCLEAN_FIRESABER_CLEAN_GenMatrix(A, seed_A);
@@ -77,7 +77,7 @@ void PQCLEAN_FIRESABER_CLEAN_indcpa_kem_dec(uint8_t m[SABER_KEYBYTES], const uin
     uint16_t b[SABER_L][SABER_N];
     uint16_t v[SABER_N] = {0};
     uint16_t cm[SABER_N];
-    int i;
+    size_t i;
 
     PQCLEAN_FIRESABER_CLEAN_BS2POLVECq(s, sk);
     PQCLEAN_FIRESABER_CLEAN_BS2POLVECp(b, ciphertext);
diff --git a/crypto_kem/firesaber/clean/kem.c b/crypto_kem/firesaber/clean/kem.c
index e94219a6..2ffe4e75 100644
--- a/crypto_kem/firesaber/clean/kem.c
+++ b/crypto_kem/firesaber/clean/kem.c
@@ -4,13 +4,12 @@
 #include "fips202.h"
 #include "randombytes.h"
 #include "verify.h"
+#include <stddef.h>
 #include <stdint.h>
-#include <stdio.h>
-#include <string.h>
 
 
 int PQCLEAN_FIRESABER_CLEAN_crypto_kem_keypair(uint8_t *pk, uint8_t *sk) {
-    int i;
+    size_t i;
 
     PQCLEAN_FIRESABER_CLEAN_indcpa_kem_keypair(pk, sk); // sk[0:SABER_INDCPA_SECRETKEYBYTES-1] <-- sk
     for (i = 0; i < SABER_INDCPA_PUBLICKEYBYTES; i++) {
@@ -48,7 +47,7 @@ int PQCLEAN_FIRESABER_CLEAN_crypto_kem_enc(uint8_t *c, uint8_t *k, const uint8_t
 }
 
 int PQCLEAN_FIRESABER_CLEAN_crypto_kem_dec(uint8_t *k, const uint8_t *c, const uint8_t *sk) {
-    int i;
+    size_t i;
     uint8_t fail;
     uint8_t cmp[SABER_BYTES_CCA_DEC];
     uint8_t buf[64];
diff --git a/crypto_kem/firesaber/clean/poly.c b/crypto_kem/firesaber/clean/poly.c
index c65175fe..c6d729ba 100644
--- a/crypto_kem/firesaber/clean/poly.c
+++ b/crypto_kem/firesaber/clean/poly.c
@@ -4,31 +4,31 @@
 #include "pack_unpack.h"
 #include "poly.h"
 #include "poly_mul.h"
-#include <stdio.h>
+#include <stddef.h>
 
 void PQCLEAN_FIRESABER_CLEAN_MatrixVectorMul(uint16_t res[SABER_L][SABER_N], const uint16_t A[SABER_L][SABER_L][SABER_N], const uint16_t s[SABER_L][SABER_N], int16_t transpose) {
-    int i, j;
+    size_t i, j;
     for (i = 0; i < SABER_L; i++) {
         for (j = 0; j < SABER_L; j++) {
             if (transpose == 1) {
-                PQCLEAN_FIRESABER_CLEAN_poly_mul_acc(A[j][i], s[j], res[i]);
+                PQCLEAN_FIRESABER_CLEAN_poly_mul_acc(res[i], A[j][i], s[j]);
             } else {
-                PQCLEAN_FIRESABER_CLEAN_poly_mul_acc(A[i][j], s[j], res[i]);
+                PQCLEAN_FIRESABER_CLEAN_poly_mul_acc(res[i], A[i][j], s[j]);
             }
         }
     }
 }
 
 void PQCLEAN_FIRESABER_CLEAN_InnerProd(uint16_t res[SABER_N], const uint16_t b[SABER_L][SABER_N], const uint16_t s[SABER_L][SABER_N]) {
-    int j;
+    size_t j;
     for (j = 0; j < SABER_L; j++) {
-        PQCLEAN_FIRESABER_CLEAN_poly_mul_acc(b[j], s[j], res);
+        PQCLEAN_FIRESABER_CLEAN_poly_mul_acc(res, b[j], s[j]);
     }
 }
 
 void PQCLEAN_FIRESABER_CLEAN_GenMatrix(uint16_t A[SABER_L][SABER_L][SABER_N], const uint8_t seed[SABER_SEEDBYTES]) {
     uint8_t buf[SABER_L * SABER_POLYVECBYTES];
-    int i;
+    size_t i;
 
     shake128(buf, sizeof(buf), seed, SABER_SEEDBYTES);
 
diff --git a/crypto_kem/firesaber/clean/poly_mul.c b/crypto_kem/firesaber/clean/poly_mul.c
index 27c92f29..6b527c21 100644
--- a/crypto_kem/firesaber/clean/poly_mul.c
+++ b/crypto_kem/firesaber/clean/poly_mul.c
@@ -11,13 +11,13 @@
 #define OVERFLOWING_MUL(X, Y) ((uint16_t)((uint32_t)(X) * (uint32_t)(Y)))
 
 #define KARATSUBA_N 64
-static void karatsuba_simple(const uint16_t *a_1, const uint16_t *b_1, uint16_t *result_final) {
+static void karatsuba_simple(uint16_t *result_final, const uint16_t *a_1, const uint16_t *b_1) {
     uint16_t d01[KARATSUBA_N / 2 - 1];
     uint16_t d0123[KARATSUBA_N / 2 - 1];
     uint16_t d23[KARATSUBA_N / 2 - 1];
     uint16_t result_d01[KARATSUBA_N - 1];
 
-    int32_t i, j;
+    size_t i, j;
 
     memset(result_d01, 0, (KARATSUBA_N - 1)*sizeof(uint16_t));
     memset(d01, 0, (KARATSUBA_N / 2 - 1)*sizeof(uint16_t));
@@ -110,7 +110,7 @@ static void karatsuba_simple(const uint16_t *a_1, const uint16_t *b_1, uint16_t
 
 
 
-static void toom_cook_4way (const uint16_t *a1, const uint16_t *b1, uint16_t *result) {
+static void toom_cook_4way (uint16_t *result, const uint16_t *a1, const uint16_t *b1) {
     uint16_t inv3 = 43691, inv9 = 36409, inv15 = 61167;
 
     uint16_t aw1[N_SB], aw2[N_SB], aw3[N_SB], aw4[N_SB], aw5[N_SB], aw6[N_SB], aw7[N_SB];
@@ -181,13 +181,13 @@ static void toom_cook_4way (const uint16_t *a1, const uint16_t *b1, uint16_t *re
 
     // MULTIPLICATION
 
-    karatsuba_simple(aw1, bw1, w1);
-    karatsuba_simple(aw2, bw2, w2);
-    karatsuba_simple(aw3, bw3, w3);
-    karatsuba_simple(aw4, bw4, w4);
-    karatsuba_simple(aw5, bw5, w5);
-    karatsuba_simple(aw6, bw6, w6);
-    karatsuba_simple(aw7, bw7, w7);
+    karatsuba_simple(w1, aw1, bw1);
+    karatsuba_simple(w2, aw2, bw2);
+    karatsuba_simple(w3, aw3, bw3);
+    karatsuba_simple(w4, aw4, bw4);
+    karatsuba_simple(w5, aw5, bw5);
+    karatsuba_simple(w6, aw6, bw6);
+    karatsuba_simple(w7, aw7, bw7);
 
     // INTERPOLATION
     for (i = 0; i < N_SB_RES; ++i) {
@@ -229,11 +229,11 @@ static void toom_cook_4way (const uint16_t *a1, const uint16_t *b1, uint16_t *re
 }
 
 /* res += a*b */
-void PQCLEAN_FIRESABER_CLEAN_poly_mul_acc(const uint16_t a[SABER_N], const uint16_t b[SABER_N], uint16_t res[SABER_N]) {
+void PQCLEAN_FIRESABER_CLEAN_poly_mul_acc(uint16_t res[SABER_N], const uint16_t a[SABER_N], const uint16_t b[SABER_N]) {
     uint16_t c[2 * SABER_N] = {0};
-    int i;
+    size_t i;
 
-    toom_cook_4way(a, b, c);
+    toom_cook_4way(c, a, b);
 
     /* reduction */
     for (i = SABER_N; i < 2 * SABER_N; i++) {
diff --git a/crypto_kem/firesaber/clean/poly_mul.h b/crypto_kem/firesaber/clean/poly_mul.h
index e554d60c..b6911577 100644
--- a/crypto_kem/firesaber/clean/poly_mul.h
+++ b/crypto_kem/firesaber/clean/poly_mul.h
@@ -3,7 +3,7 @@
 #include "SABER_params.h"
 #include <stdint.h>
 
-void PQCLEAN_FIRESABER_CLEAN_poly_mul_acc(const uint16_t a[SABER_N], const uint16_t b[SABER_N], uint16_t res[SABER_N]);
+void PQCLEAN_FIRESABER_CLEAN_poly_mul_acc(uint16_t res[SABER_N], const uint16_t a[SABER_N], const uint16_t b[SABER_N]);
 
 
 #endif
diff --git a/crypto_kem/lightsaber/META.yml b/crypto_kem/lightsaber/META.yml
index 1cc06c9a..7e1dd2eb 100644
--- a/crypto_kem/lightsaber/META.yml
+++ b/crypto_kem/lightsaber/META.yml
@@ -14,9 +14,9 @@ principal-submitters:
   - Frederik Vercauteren
 implementations:
     - name: clean
-      version: https://github.com/KULeuven-COSIC/SABER/commit/509cc5ec3a7e12a751ccdd2ef5bd6e54e00bd350
+      version: https://github.com/KULeuven-COSIC/SABER/tree/509cc5ec3a7e12a751ccdd2ef5bd6e54e00bd350 via https://github.com/jschanck/package-pqclean/tree/b53a47b5/saber
     - name: avx2
-      version: https://github.com/KULeuven-COSIC/SABER/commit/509cc5ec3a7e12a751ccdd2ef5bd6e54e00bd350
+      version: https://github.com/KULeuven-COSIC/SABER/tree/509cc5ec3a7e12a751ccdd2ef5bd6e54e00bd350 via https://github.com/jschanck/package-pqclean/tree/b53a47b5/saber
       supported_platforms:
           - architecture: x86_64
             operating_systems:
diff --git a/crypto_kem/lightsaber/avx2/SABER_indcpa.c b/crypto_kem/lightsaber/avx2/SABER_indcpa.c
index 3270a8c9..47f760e9 100644
--- a/crypto_kem/lightsaber/avx2/SABER_indcpa.c
+++ b/crypto_kem/lightsaber/avx2/SABER_indcpa.c
@@ -66,7 +66,7 @@ static void GenSecret(uint16_t r[SABER_K][SABER_N], const uint8_t *seed) {
 }
 
 //********************************matrix-vector mul routines*****************************************************
-static void matrix_vector_mul(__m256i a1_avx_combined[NUM_POLY][NUM_POLY][AVX_N1], __m256i b_bucket[NUM_POLY][SCHB_N * 4], __m256i res_avx[NUM_POLY][AVX_N1], int isTranspose) {
+static void matrix_vector_mul(__m256i res_avx[NUM_POLY][AVX_N1], __m256i a1_avx_combined[NUM_POLY][NUM_POLY][AVX_N1], __m256i b_bucket[NUM_POLY][SCHB_N * 4], int isTranspose) {
     int64_t i, j;
 
     __m256i c_bucket[2 * SCM_SIZE * 4]; //Holds results for 9 Karatsuba at a time
@@ -86,7 +86,7 @@ static void matrix_vector_mul(__m256i a1_avx_combined[NUM_POLY][NUM_POLY][AVX_N1
 
 }
 
-static void vector_vector_mul(__m256i a_avx[NUM_POLY][AVX_N1], __m256i b_bucket[NUM_POLY][SCHB_N * 4], __m256i res_avx[AVX_N1]) {
+static void vector_vector_mul(__m256i res_avx[AVX_N1], __m256i a_avx[NUM_POLY][AVX_N1], __m256i b_bucket[NUM_POLY][SCHB_N * 4]) {
 
     int64_t i;
 
@@ -162,7 +162,7 @@ void PQCLEAN_LIGHTSABER_AVX2_indcpa_kem_keypair(uint8_t *pk, uint8_t *sk) {
     for (j = 0; j < NUM_POLY; j++) {
         TC_eval(sk_avx[j], b_bucket[j]);
     }
-    matrix_vector_mul(a_avx, b_bucket, res_avx, 1);// Matrix-vector multiplication; Matrix in transposed order
+    matrix_vector_mul(res_avx, a_avx, b_bucket, 1);// Matrix-vector multiplication; Matrix in transposed order
 
     // Now truncation
 
@@ -259,7 +259,7 @@ void PQCLEAN_LIGHTSABER_AVX2_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_D
     for (j = 0; j < NUM_POLY; j++) {
         TC_eval(sk_avx[j], b_bucket[j]);
     }
-    matrix_vector_mul(a_avx, b_bucket, res_avx, 0);// Matrix-vector multiplication; Matrix in normal order
+    matrix_vector_mul(res_avx, a_avx, b_bucket, 0);// Matrix-vector multiplication; Matrix in normal order
 
     // Now truncation
 
@@ -302,7 +302,7 @@ void PQCLEAN_LIGHTSABER_AVX2_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_D
 
     // vector-vector scalar multiplication with mod p
 
-    vector_vector_mul(pkcl_avx, b_bucket, vprime_avx);
+    vector_vector_mul(vprime_avx, pkcl_avx, b_bucket);
 
     // Computation of v'+h1
     for (i = 0; i < SABER_N / 16; i++) { //adding h1
@@ -392,7 +392,7 @@ void PQCLEAN_LIGHTSABER_AVX2_indcpa_kem_dec(uint8_t m[SABER_KEYBYTES], const uin
         TC_eval(sksv_avx[j], b_bucket[j]);
     }
 
-    vector_vector_mul(pksv_avx, b_bucket, v_avx);
+    vector_vector_mul(v_avx, pksv_avx, b_bucket);
 
     for (i = 0; i < SABER_N / 16; i++) {
         _mm256_maskstore_epi32 ((int *)(message_dec_unpacked + i * 16), _mm256_set1_epi32(-1), v_avx[i]);
diff --git a/crypto_kem/lightsaber/clean/SABER_indcpa.c b/crypto_kem/lightsaber/clean/SABER_indcpa.c
index 4b9cb150..9dcdfb93 100644
--- a/crypto_kem/lightsaber/clean/SABER_indcpa.c
+++ b/crypto_kem/lightsaber/clean/SABER_indcpa.c
@@ -17,7 +17,7 @@ void PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_keypair(uint8_t pk[SABER_INDCPA_PUBLICK
 
     uint8_t seed_A[SABER_SEEDBYTES];
     uint8_t seed_s[SABER_NOISE_SEEDBYTES];
-    int i, j;
+    size_t i, j;
 
     randombytes(seed_A, SABER_SEEDBYTES);
     shake128(seed_A, SABER_SEEDBYTES, seed_A, SABER_SEEDBYTES); // for not revealing system RNG state
@@ -45,7 +45,7 @@ void PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_
     uint16_t vp[SABER_N] = {0};
     uint16_t mp[SABER_N];
     uint16_t b[SABER_L][SABER_N];
-    int i, j;
+    size_t i, j;
     const uint8_t *seed_A = pk + SABER_POLYVECCOMPRESSEDBYTES;
 
     PQCLEAN_LIGHTSABER_CLEAN_GenMatrix(A, seed_A);
@@ -77,7 +77,7 @@ void PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_dec(uint8_t m[SABER_KEYBYTES], const ui
     uint16_t b[SABER_L][SABER_N];
     uint16_t v[SABER_N] = {0};
     uint16_t cm[SABER_N];
-    int i;
+    size_t i;
 
     PQCLEAN_LIGHTSABER_CLEAN_BS2POLVECq(s, sk);
     PQCLEAN_LIGHTSABER_CLEAN_BS2POLVECp(b, ciphertext);
diff --git a/crypto_kem/lightsaber/clean/kem.c b/crypto_kem/lightsaber/clean/kem.c
index eb9353b1..d0a67736 100644
--- a/crypto_kem/lightsaber/clean/kem.c
+++ b/crypto_kem/lightsaber/clean/kem.c
@@ -4,13 +4,12 @@
 #include "fips202.h"
 #include "randombytes.h"
 #include "verify.h"
+#include <stddef.h>
 #include <stdint.h>
-#include <stdio.h>
-#include <string.h>
 
 
 int PQCLEAN_LIGHTSABER_CLEAN_crypto_kem_keypair(uint8_t *pk, uint8_t *sk) {
-    int i;
+    size_t i;
 
     PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_keypair(pk, sk); // sk[0:SABER_INDCPA_SECRETKEYBYTES-1] <-- sk
     for (i = 0; i < SABER_INDCPA_PUBLICKEYBYTES; i++) {
@@ -48,7 +47,7 @@ int PQCLEAN_LIGHTSABER_CLEAN_crypto_kem_enc(uint8_t *c, uint8_t *k, const uint8_
 }
 
 int PQCLEAN_LIGHTSABER_CLEAN_crypto_kem_dec(uint8_t *k, const uint8_t *c, const uint8_t *sk) {
-    int i;
+    size_t i;
     uint8_t fail;
     uint8_t cmp[SABER_BYTES_CCA_DEC];
     uint8_t buf[64];
diff --git a/crypto_kem/lightsaber/clean/poly.c b/crypto_kem/lightsaber/clean/poly.c
index 1c1e22cc..9bb55afe 100644
--- a/crypto_kem/lightsaber/clean/poly.c
+++ b/crypto_kem/lightsaber/clean/poly.c
@@ -4,31 +4,31 @@
 #include "pack_unpack.h"
 #include "poly.h"
 #include "poly_mul.h"
-#include <stdio.h>
+#include <stddef.h>
 
 void PQCLEAN_LIGHTSABER_CLEAN_MatrixVectorMul(uint16_t res[SABER_L][SABER_N], const uint16_t A[SABER_L][SABER_L][SABER_N], const uint16_t s[SABER_L][SABER_N], int16_t transpose) {
-    int i, j;
+    size_t i, j;
     for (i = 0; i < SABER_L; i++) {
         for (j = 0; j < SABER_L; j++) {
             if (transpose == 1) {
-                PQCLEAN_LIGHTSABER_CLEAN_poly_mul_acc(A[j][i], s[j], res[i]);
+                PQCLEAN_LIGHTSABER_CLEAN_poly_mul_acc(res[i], A[j][i], s[j]);
             } else {
-                PQCLEAN_LIGHTSABER_CLEAN_poly_mul_acc(A[i][j], s[j], res[i]);
+                PQCLEAN_LIGHTSABER_CLEAN_poly_mul_acc(res[i], A[i][j], s[j]);
             }
         }
     }
 }
 
 void PQCLEAN_LIGHTSABER_CLEAN_InnerProd(uint16_t res[SABER_N], const uint16_t b[SABER_L][SABER_N], const uint16_t s[SABER_L][SABER_N]) {
-    int j;
+    size_t j;
     for (j = 0; j < SABER_L; j++) {
-        PQCLEAN_LIGHTSABER_CLEAN_poly_mul_acc(b[j], s[j], res);
+        PQCLEAN_LIGHTSABER_CLEAN_poly_mul_acc(res, b[j], s[j]);
     }
 }
 
 void PQCLEAN_LIGHTSABER_CLEAN_GenMatrix(uint16_t A[SABER_L][SABER_L][SABER_N], const uint8_t seed[SABER_SEEDBYTES]) {
     uint8_t buf[SABER_L * SABER_POLYVECBYTES];
-    int i;
+    size_t i;
 
     shake128(buf, sizeof(buf), seed, SABER_SEEDBYTES);
 
diff --git a/crypto_kem/lightsaber/clean/poly_mul.c b/crypto_kem/lightsaber/clean/poly_mul.c
index 5e37a024..c7f5c424 100644
--- a/crypto_kem/lightsaber/clean/poly_mul.c
+++ b/crypto_kem/lightsaber/clean/poly_mul.c
@@ -11,13 +11,13 @@
 #define OVERFLOWING_MUL(X, Y) ((uint16_t)((uint32_t)(X) * (uint32_t)(Y)))
 
 #define KARATSUBA_N 64
-static void karatsuba_simple(const uint16_t *a_1, const uint16_t *b_1, uint16_t *result_final) {
+static void karatsuba_simple(uint16_t *result_final, const uint16_t *a_1, const uint16_t *b_1) {
     uint16_t d01[KARATSUBA_N / 2 - 1];
     uint16_t d0123[KARATSUBA_N / 2 - 1];
     uint16_t d23[KARATSUBA_N / 2 - 1];
     uint16_t result_d01[KARATSUBA_N - 1];
 
-    int32_t i, j;
+    size_t i, j;
 
     memset(result_d01, 0, (KARATSUBA_N - 1)*sizeof(uint16_t));
     memset(d01, 0, (KARATSUBA_N / 2 - 1)*sizeof(uint16_t));
@@ -110,7 +110,7 @@ static void karatsuba_simple(const uint16_t *a_1, const uint16_t *b_1, uint16_t
 
 
 
-static void toom_cook_4way (const uint16_t *a1, const uint16_t *b1, uint16_t *result) {
+static void toom_cook_4way (uint16_t *result, const uint16_t *a1, const uint16_t *b1) {
     uint16_t inv3 = 43691, inv9 = 36409, inv15 = 61167;
 
     uint16_t aw1[N_SB], aw2[N_SB], aw3[N_SB], aw4[N_SB], aw5[N_SB], aw6[N_SB], aw7[N_SB];
@@ -181,13 +181,13 @@ static void toom_cook_4way (const uint16_t *a1, const uint16_t *b1, uint16_t *re
 
     // MULTIPLICATION
 
-    karatsuba_simple(aw1, bw1, w1);
-    karatsuba_simple(aw2, bw2, w2);
-    karatsuba_simple(aw3, bw3, w3);
-    karatsuba_simple(aw4, bw4, w4);
-    karatsuba_simple(aw5, bw5, w5);
-    karatsuba_simple(aw6, bw6, w6);
-    karatsuba_simple(aw7, bw7, w7);
+    karatsuba_simple(w1, aw1, bw1);
+    karatsuba_simple(w2, aw2, bw2);
+    karatsuba_simple(w3, aw3, bw3);
+    karatsuba_simple(w4, aw4, bw4);
+    karatsuba_simple(w5, aw5, bw5);
+    karatsuba_simple(w6, aw6, bw6);
+    karatsuba_simple(w7, aw7, bw7);
 
     // INTERPOLATION
     for (i = 0; i < N_SB_RES; ++i) {
@@ -229,11 +229,11 @@ static void toom_cook_4way (const uint16_t *a1, const uint16_t *b1, uint16_t *re
 }
 
 /* res += a*b */
-void PQCLEAN_LIGHTSABER_CLEAN_poly_mul_acc(const uint16_t a[SABER_N], const uint16_t b[SABER_N], uint16_t res[SABER_N]) {
+void PQCLEAN_LIGHTSABER_CLEAN_poly_mul_acc(uint16_t res[SABER_N], const uint16_t a[SABER_N], const uint16_t b[SABER_N]) {
     uint16_t c[2 * SABER_N] = {0};
-    int i;
+    size_t i;
 
-    toom_cook_4way(a, b, c);
+    toom_cook_4way(c, a, b);
 
     /* reduction */
     for (i = SABER_N; i < 2 * SABER_N; i++) {
diff --git a/crypto_kem/lightsaber/clean/poly_mul.h b/crypto_kem/lightsaber/clean/poly_mul.h
index 0d5cf6ed..5ec233bb 100644
--- a/crypto_kem/lightsaber/clean/poly_mul.h
+++ b/crypto_kem/lightsaber/clean/poly_mul.h
@@ -3,7 +3,7 @@
 #include "SABER_params.h"
 #include <stdint.h>
 
-void PQCLEAN_LIGHTSABER_CLEAN_poly_mul_acc(const uint16_t a[SABER_N], const uint16_t b[SABER_N], uint16_t res[SABER_N]);
+void PQCLEAN_LIGHTSABER_CLEAN_poly_mul_acc(uint16_t res[SABER_N], const uint16_t a[SABER_N], const uint16_t b[SABER_N]);
 
 
 #endif
diff --git a/crypto_kem/saber/META.yml b/crypto_kem/saber/META.yml
index 50250180..87187702 100644
--- a/crypto_kem/saber/META.yml
+++ b/crypto_kem/saber/META.yml
@@ -14,9 +14,9 @@ principal-submitters:
   - Frederik Vercauteren
 implementations:
     - name: clean
-      version: https://github.com/KULeuven-COSIC/SABER/commit/509cc5ec3a7e12a751ccdd2ef5bd6e54e00bd350
+      version: https://github.com/KULeuven-COSIC/SABER/tree/509cc5ec3a7e12a751ccdd2ef5bd6e54e00bd350 via https://github.com/jschanck/package-pqclean/tree/b53a47b5/saber
     - name: avx2
-      version: https://github.com/KULeuven-COSIC/SABER/commit/509cc5ec3a7e12a751ccdd2ef5bd6e54e00bd350
+      version: https://github.com/KULeuven-COSIC/SABER/tree/509cc5ec3a7e12a751ccdd2ef5bd6e54e00bd350 via https://github.com/jschanck/package-pqclean/tree/b53a47b5/saber
       supported_platforms:
           - architecture: x86_64
             operating_systems:
diff --git a/crypto_kem/saber/avx2/SABER_indcpa.c b/crypto_kem/saber/avx2/SABER_indcpa.c
index d16a7a06..5515c610 100644
--- a/crypto_kem/saber/avx2/SABER_indcpa.c
+++ b/crypto_kem/saber/avx2/SABER_indcpa.c
@@ -66,7 +66,7 @@ static void GenSecret(uint16_t r[SABER_K][SABER_N], const uint8_t *seed) {
 }
 
 //********************************matrix-vector mul routines*****************************************************
-static void matrix_vector_mul(__m256i a1_avx_combined[NUM_POLY][NUM_POLY][AVX_N1], __m256i b_bucket[NUM_POLY][SCHB_N * 4], __m256i res_avx[NUM_POLY][AVX_N1], int isTranspose) {
+static void matrix_vector_mul(__m256i res_avx[NUM_POLY][AVX_N1], __m256i a1_avx_combined[NUM_POLY][NUM_POLY][AVX_N1], __m256i b_bucket[NUM_POLY][SCHB_N * 4], int isTranspose) {
     int64_t i, j;
 
     __m256i c_bucket[2 * SCM_SIZE * 4]; //Holds results for 9 Karatsuba at a time
@@ -86,7 +86,7 @@ static void matrix_vector_mul(__m256i a1_avx_combined[NUM_POLY][NUM_POLY][AVX_N1
 
 }
 
-static void vector_vector_mul(__m256i a_avx[NUM_POLY][AVX_N1], __m256i b_bucket[NUM_POLY][SCHB_N * 4], __m256i res_avx[AVX_N1]) {
+static void vector_vector_mul(__m256i res_avx[AVX_N1], __m256i a_avx[NUM_POLY][AVX_N1], __m256i b_bucket[NUM_POLY][SCHB_N * 4]) {
 
     int64_t i;
 
@@ -162,7 +162,7 @@ void PQCLEAN_SABER_AVX2_indcpa_kem_keypair(uint8_t *pk, uint8_t *sk) {
     for (j = 0; j < NUM_POLY; j++) {
         TC_eval(sk_avx[j], b_bucket[j]);
     }
-    matrix_vector_mul(a_avx, b_bucket, res_avx, 1);// Matrix-vector multiplication; Matrix in transposed order
+    matrix_vector_mul(res_avx, a_avx, b_bucket, 1);// Matrix-vector multiplication; Matrix in transposed order
 
     // Now truncation
 
@@ -259,7 +259,7 @@ void PQCLEAN_SABER_AVX2_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_DEC],
     for (j = 0; j < NUM_POLY; j++) {
         TC_eval(sk_avx[j], b_bucket[j]);
     }
-    matrix_vector_mul(a_avx, b_bucket, res_avx, 0);// Matrix-vector multiplication; Matrix in normal order
+    matrix_vector_mul(res_avx, a_avx, b_bucket, 0);// Matrix-vector multiplication; Matrix in normal order
 
     // Now truncation
 
@@ -302,7 +302,7 @@ void PQCLEAN_SABER_AVX2_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_DEC],
 
     // vector-vector scalar multiplication with mod p
 
-    vector_vector_mul(pkcl_avx, b_bucket, vprime_avx);
+    vector_vector_mul(vprime_avx, pkcl_avx, b_bucket);
 
     // Computation of v'+h1
     for (i = 0; i < SABER_N / 16; i++) { //adding h1
@@ -392,7 +392,7 @@ void PQCLEAN_SABER_AVX2_indcpa_kem_dec(uint8_t m[SABER_KEYBYTES], const uint8_t
         TC_eval(sksv_avx[j], b_bucket[j]);
     }
 
-    vector_vector_mul(pksv_avx, b_bucket, v_avx);
+    vector_vector_mul(v_avx, pksv_avx, b_bucket);
 
     for (i = 0; i < SABER_N / 16; i++) {
         _mm256_maskstore_epi32 ((int *)(message_dec_unpacked + i * 16), _mm256_set1_epi32(-1), v_avx[i]);
diff --git a/crypto_kem/saber/clean/SABER_indcpa.c b/crypto_kem/saber/clean/SABER_indcpa.c
index c36f02ea..23325749 100644
--- a/crypto_kem/saber/clean/SABER_indcpa.c
+++ b/crypto_kem/saber/clean/SABER_indcpa.c
@@ -17,7 +17,7 @@ void PQCLEAN_SABER_CLEAN_indcpa_kem_keypair(uint8_t pk[SABER_INDCPA_PUBLICKEYBYT
 
     uint8_t seed_A[SABER_SEEDBYTES];
     uint8_t seed_s[SABER_NOISE_SEEDBYTES];
-    int i, j;
+    size_t i, j;
 
     randombytes(seed_A, SABER_SEEDBYTES);
     shake128(seed_A, SABER_SEEDBYTES, seed_A, SABER_SEEDBYTES); // for not revealing system RNG state
@@ -45,7 +45,7 @@ void PQCLEAN_SABER_CLEAN_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_DEC],
     uint16_t vp[SABER_N] = {0};
     uint16_t mp[SABER_N];
     uint16_t b[SABER_L][SABER_N];
-    int i, j;
+    size_t i, j;
     const uint8_t *seed_A = pk + SABER_POLYVECCOMPRESSEDBYTES;
 
     PQCLEAN_SABER_CLEAN_GenMatrix(A, seed_A);
@@ -77,7 +77,7 @@ void PQCLEAN_SABER_CLEAN_indcpa_kem_dec(uint8_t m[SABER_KEYBYTES], const uint8_t
     uint16_t b[SABER_L][SABER_N];
     uint16_t v[SABER_N] = {0};
     uint16_t cm[SABER_N];
-    int i;
+    size_t i;
 
     PQCLEAN_SABER_CLEAN_BS2POLVECq(s, sk);
     PQCLEAN_SABER_CLEAN_BS2POLVECp(b, ciphertext);
diff --git a/crypto_kem/saber/clean/kem.c b/crypto_kem/saber/clean/kem.c
index ed8e3ac7..6a7f20c4 100644
--- a/crypto_kem/saber/clean/kem.c
+++ b/crypto_kem/saber/clean/kem.c
@@ -4,13 +4,12 @@
 #include "fips202.h"
 #include "randombytes.h"
 #include "verify.h"
+#include <stddef.h>
 #include <stdint.h>
-#include <stdio.h>
-#include <string.h>
 
 
 int PQCLEAN_SABER_CLEAN_crypto_kem_keypair(uint8_t *pk, uint8_t *sk) {
-    int i;
+    size_t i;
 
     PQCLEAN_SABER_CLEAN_indcpa_kem_keypair(pk, sk); // sk[0:SABER_INDCPA_SECRETKEYBYTES-1] <-- sk
     for (i = 0; i < SABER_INDCPA_PUBLICKEYBYTES; i++) {
@@ -48,7 +47,7 @@ int PQCLEAN_SABER_CLEAN_crypto_kem_enc(uint8_t *c, uint8_t *k, const uint8_t *pk
 }
 
 int PQCLEAN_SABER_CLEAN_crypto_kem_dec(uint8_t *k, const uint8_t *c, const uint8_t *sk) {
-    int i;
+    size_t i;
     uint8_t fail;
     uint8_t cmp[SABER_BYTES_CCA_DEC];
     uint8_t buf[64];
diff --git a/crypto_kem/saber/clean/poly.c b/crypto_kem/saber/clean/poly.c
index f0403ccf..2c44e962 100644
--- a/crypto_kem/saber/clean/poly.c
+++ b/crypto_kem/saber/clean/poly.c
@@ -4,31 +4,31 @@
 #include "pack_unpack.h"
 #include "poly.h"
 #include "poly_mul.h"
-#include <stdio.h>
+#include <stddef.h>
 
 void PQCLEAN_SABER_CLEAN_MatrixVectorMul(uint16_t res[SABER_L][SABER_N], const uint16_t A[SABER_L][SABER_L][SABER_N], const uint16_t s[SABER_L][SABER_N], int16_t transpose) {
-    int i, j;
+    size_t i, j;
     for (i = 0; i < SABER_L; i++) {
         for (j = 0; j < SABER_L; j++) {
             if (transpose == 1) {
-                PQCLEAN_SABER_CLEAN_poly_mul_acc(A[j][i], s[j], res[i]);
+                PQCLEAN_SABER_CLEAN_poly_mul_acc(res[i], A[j][i], s[j]);
             } else {
-                PQCLEAN_SABER_CLEAN_poly_mul_acc(A[i][j], s[j], res[i]);
+                PQCLEAN_SABER_CLEAN_poly_mul_acc(res[i], A[i][j], s[j]);
             }
         }
     }
 }
 
 void PQCLEAN_SABER_CLEAN_InnerProd(uint16_t res[SABER_N], const uint16_t b[SABER_L][SABER_N], const uint16_t s[SABER_L][SABER_N]) {
-    int j;
+    size_t j;
     for (j = 0; j < SABER_L; j++) {
-        PQCLEAN_SABER_CLEAN_poly_mul_acc(b[j], s[j], res);
+        PQCLEAN_SABER_CLEAN_poly_mul_acc(res, b[j], s[j]);
     }
 }
 
 void PQCLEAN_SABER_CLEAN_GenMatrix(uint16_t A[SABER_L][SABER_L][SABER_N], const uint8_t seed[SABER_SEEDBYTES]) {
     uint8_t buf[SABER_L * SABER_POLYVECBYTES];
-    int i;
+    size_t i;
 
     shake128(buf, sizeof(buf), seed, SABER_SEEDBYTES);
 
diff --git a/crypto_kem/saber/clean/poly_mul.c b/crypto_kem/saber/clean/poly_mul.c
index 0655383b..686960dc 100644
--- a/crypto_kem/saber/clean/poly_mul.c
+++ b/crypto_kem/saber/clean/poly_mul.c
@@ -11,13 +11,13 @@
 #define OVERFLOWING_MUL(X, Y) ((uint16_t)((uint32_t)(X) * (uint32_t)(Y)))
 
 #define KARATSUBA_N 64
-static void karatsuba_simple(const uint16_t *a_1, const uint16_t *b_1, uint16_t *result_final) {
+static void karatsuba_simple(uint16_t *result_final, const uint16_t *a_1, const uint16_t *b_1) {
     uint16_t d01[KARATSUBA_N / 2 - 1];
     uint16_t d0123[KARATSUBA_N / 2 - 1];
     uint16_t d23[KARATSUBA_N / 2 - 1];
     uint16_t result_d01[KARATSUBA_N - 1];
 
-    int32_t i, j;
+    size_t i, j;
 
     memset(result_d01, 0, (KARATSUBA_N - 1)*sizeof(uint16_t));
     memset(d01, 0, (KARATSUBA_N / 2 - 1)*sizeof(uint16_t));
@@ -110,7 +110,7 @@ static void karatsuba_simple(const uint16_t *a_1, const uint16_t *b_1, uint16_t
 
 
 
-static void toom_cook_4way (const uint16_t *a1, const uint16_t *b1, uint16_t *result) {
+static void toom_cook_4way (uint16_t *result, const uint16_t *a1, const uint16_t *b1) {
     uint16_t inv3 = 43691, inv9 = 36409, inv15 = 61167;
 
     uint16_t aw1[N_SB], aw2[N_SB], aw3[N_SB], aw4[N_SB], aw5[N_SB], aw6[N_SB], aw7[N_SB];
@@ -181,13 +181,13 @@ static void toom_cook_4way (const uint16_t *a1, const uint16_t *b1, uint16_t *re
 
     // MULTIPLICATION
 
-    karatsuba_simple(aw1, bw1, w1);
-    karatsuba_simple(aw2, bw2, w2);
-    karatsuba_simple(aw3, bw3, w3);
-    karatsuba_simple(aw4, bw4, w4);
-    karatsuba_simple(aw5, bw5, w5);
-    karatsuba_simple(aw6, bw6, w6);
-    karatsuba_simple(aw7, bw7, w7);
+    karatsuba_simple(w1, aw1, bw1);
+    karatsuba_simple(w2, aw2, bw2);
+    karatsuba_simple(w3, aw3, bw3);
+    karatsuba_simple(w4, aw4, bw4);
+    karatsuba_simple(w5, aw5, bw5);
+    karatsuba_simple(w6, aw6, bw6);
+    karatsuba_simple(w7, aw7, bw7);
 
     // INTERPOLATION
     for (i = 0; i < N_SB_RES; ++i) {
@@ -229,11 +229,11 @@ static void toom_cook_4way (const uint16_t *a1, const uint16_t *b1, uint16_t *re
 }
 
 /* res += a*b */
-void PQCLEAN_SABER_CLEAN_poly_mul_acc(const uint16_t a[SABER_N], const uint16_t b[SABER_N], uint16_t res[SABER_N]) {
+void PQCLEAN_SABER_CLEAN_poly_mul_acc(uint16_t res[SABER_N], const uint16_t a[SABER_N], const uint16_t b[SABER_N]) {
     uint16_t c[2 * SABER_N] = {0};
-    int i;
+    size_t i;
 
-    toom_cook_4way(a, b, c);
+    toom_cook_4way(c, a, b);
 
     /* reduction */
     for (i = SABER_N; i < 2 * SABER_N; i++) {
diff --git a/crypto_kem/saber/clean/poly_mul.h b/crypto_kem/saber/clean/poly_mul.h
index e0f10043..82140f5b 100644
--- a/crypto_kem/saber/clean/poly_mul.h
+++ b/crypto_kem/saber/clean/poly_mul.h
@@ -3,7 +3,7 @@
 #include "SABER_params.h"
 #include <stdint.h>
 
-void PQCLEAN_SABER_CLEAN_poly_mul_acc(const uint16_t a[SABER_N], const uint16_t b[SABER_N], uint16_t res[SABER_N]);
+void PQCLEAN_SABER_CLEAN_poly_mul_acc(uint16_t res[SABER_N], const uint16_t a[SABER_N], const uint16_t b[SABER_N]);
 
 
 #endif