Changed speed benchmark for i31 to a 521-bit modulus.
[BearSSL] / test / test_speed.c
index a09aa04..0e885db 100644 (file)
@@ -175,17 +175,24 @@ test_speed_ ## fname(void) \
 static void \
 test_speed_ ## fname(void) \
 { \
+       br_chacha20_run bc; \
        unsigned char key[32]; \
        unsigned char buf[8192]; \
        unsigned char iv[12]; \
        int i; \
        long num; \
  \
+       bc = br_ ## fname ## _get(); \
+       if (bc == 0) { \
+               printf("%-30s UNAVAILABLE\n", #Name); \
+               fflush(stdout); \
+               return; \
+       } \
        memset(key, 'T', sizeof key); \
        memset(buf, 'P', sizeof buf); \
        memset(iv, 'X', sizeof iv); \
        for (i = 0; i < 10; i ++) { \
-               br_ ## fname ## _run(key, iv, i, buf, sizeof buf); \
+               bc(key, iv, i, buf, sizeof buf); \
        } \
        num = 10; \
        for (;;) { \
@@ -195,8 +202,7 @@ test_speed_ ## fname(void) \
  \
                begin = clock(); \
                for (k = num; k > 0; k --) { \
-                       br_ ## fname ## _run(key, iv, \
-                               (uint32_t)k, buf, sizeof buf); \
+                       bc(key, iv, (uint32_t)k, buf, sizeof buf); \
                } \
                end = clock(); \
                tt = (double)(end - begin) / CLOCKS_PER_SEC; \
@@ -223,15 +229,20 @@ SPEED_HASH(SHA-512, sha512)
 #define br_aes_big_cbcenc_get_vtable()     (&br_aes_big_cbcenc_vtable)
 #define br_aes_big_cbcdec_get_vtable()     (&br_aes_big_cbcdec_vtable)
 #define br_aes_big_ctr_get_vtable()        (&br_aes_big_ctr_vtable)
+#define br_aes_big_ctrcbc_get_vtable()     (&br_aes_big_ctrcbc_vtable)
 #define br_aes_small_cbcenc_get_vtable()   (&br_aes_small_cbcenc_vtable)
 #define br_aes_small_cbcdec_get_vtable()   (&br_aes_small_cbcdec_vtable)
 #define br_aes_small_ctr_get_vtable()      (&br_aes_small_ctr_vtable)
+#define br_aes_small_ctrcbc_get_vtable()   (&br_aes_small_ctrcbc_vtable)
 #define br_aes_ct_cbcenc_get_vtable()      (&br_aes_ct_cbcenc_vtable)
 #define br_aes_ct_cbcdec_get_vtable()      (&br_aes_ct_cbcdec_vtable)
 #define br_aes_ct_ctr_get_vtable()         (&br_aes_ct_ctr_vtable)
+#define br_aes_ct_ctrcbc_get_vtable()      (&br_aes_ct_ctrcbc_vtable)
 #define br_aes_ct64_cbcenc_get_vtable()    (&br_aes_ct64_cbcenc_vtable)
 #define br_aes_ct64_cbcdec_get_vtable()    (&br_aes_ct64_cbcdec_vtable)
 #define br_aes_ct64_ctr_get_vtable()       (&br_aes_ct64_ctr_vtable)
+#define br_aes_ct64_ctrcbc_get_vtable()    (&br_aes_ct64_ctrcbc_vtable)
+#define br_chacha20_ct_get()               (&br_chacha20_ct_run)
 
 #define SPEED_AES(iname) \
 SPEED_BLOCKCIPHER_CBC(AES-128 CBC encrypt (iname), aes128_ ## iname ## _cbcenc, aes_ ## iname, 16, enc) \
@@ -265,7 +276,8 @@ SPEED_BLOCKCIPHER_CBC(3DES CBC decrypt (iname), 3des_ ## iname ## _cbcdec, des_
 SPEED_DES(tab)
 SPEED_DES(ct)
 
-SPEED_CHACHA20(ChaCha20, chacha20_ct)
+SPEED_CHACHA20(ChaCha20 (ct), chacha20_ct)
+SPEED_CHACHA20(ChaCha20 (sse2), chacha20_sse2)
 
 static void
 test_speed_ghash_inner(char *name, br_ghash gh)
@@ -435,6 +447,168 @@ test_speed_poly1305_i15(void)
        test_speed_poly1305_inner("Poly1305 (i15)", &br_poly1305_i15_run);
 }
 
+static void
+test_speed_eax_inner(char *name,
+       const br_block_ctrcbc_class *vt, size_t key_len)
+{
+       unsigned char buf[8192], key[32], nonce[16], aad[16], tag[16];
+       int i;
+       long num;
+       br_aes_gen_ctrcbc_keys ac;
+       br_eax_context ec;
+
+       if (vt == NULL) {
+               printf("%-30s UNAVAILABLE\n", name);
+               fflush(stdout);
+               return;
+       }
+       memset(key, 'K', key_len);
+       memset(nonce, 'N', sizeof nonce);
+       memset(aad, 'A', sizeof aad);
+       memset(buf, 'T', sizeof buf);
+       for (i = 0; i < 10; i ++) {
+               vt->init(&ac.vtable, key, key_len);
+               br_eax_init(&ec, &ac.vtable);
+               br_eax_reset(&ec, nonce, sizeof nonce);
+               br_eax_aad_inject(&ec, aad, sizeof aad);
+               br_eax_flip(&ec);
+               br_eax_run(&ec, 1, buf, sizeof buf);
+               br_eax_get_tag(&ec, tag);
+       }
+       num = 10;
+       for (;;) {
+               clock_t begin, end;
+               double tt;
+               long k;
+
+               begin = clock();
+               for (k = num; k > 0; k --) {
+                       vt->init(&ac.vtable, key, key_len);
+                       br_eax_init(&ec, &ac.vtable);
+                       br_eax_reset(&ec, nonce, sizeof nonce);
+                       br_eax_aad_inject(&ec, aad, sizeof aad);
+                       br_eax_flip(&ec);
+                       br_eax_run(&ec, 1, buf, sizeof buf);
+                       br_eax_get_tag(&ec, tag);
+               }
+               end = clock();
+               tt = (double)(end - begin) / CLOCKS_PER_SEC;
+               if (tt >= 2.0) {
+                       printf("%-30s %8.2f MB/s\n", name,
+                               ((double)sizeof buf) * (double)num
+                               / (tt * 1000000.0));
+                       fflush(stdout);
+                       return;
+               }
+               num <<= 1;
+       }
+}
+
+#define SPEED_EAX(Algo, algo, keysize, impl) \
+static void \
+test_speed_eax_ ## algo ## keysize ## _ ## impl(void) \
+{ \
+       test_speed_eax_inner("EAX " #Algo "-" #keysize "(" #impl ")", \
+               br_ ## algo ## _ ## impl ##  _ctrcbc_get_vtable() \
+               , (keysize) >> 3); \
+}
+
+SPEED_EAX(AES, aes, 128, big)
+SPEED_EAX(AES, aes, 128, small)
+SPEED_EAX(AES, aes, 128, ct)
+SPEED_EAX(AES, aes, 128, ct64)
+SPEED_EAX(AES, aes, 128, x86ni)
+SPEED_EAX(AES, aes, 128, pwr8)
+SPEED_EAX(AES, aes, 192, big)
+SPEED_EAX(AES, aes, 192, small)
+SPEED_EAX(AES, aes, 192, ct)
+SPEED_EAX(AES, aes, 192, ct64)
+SPEED_EAX(AES, aes, 192, x86ni)
+SPEED_EAX(AES, aes, 192, pwr8)
+SPEED_EAX(AES, aes, 256, big)
+SPEED_EAX(AES, aes, 256, small)
+SPEED_EAX(AES, aes, 256, ct)
+SPEED_EAX(AES, aes, 256, ct64)
+SPEED_EAX(AES, aes, 256, x86ni)
+SPEED_EAX(AES, aes, 256, pwr8)
+
+static void
+test_speed_shake_inner(int security_level)
+{
+       unsigned char buf[8192];
+       br_shake_context sc;
+       int i;
+       long num;
+
+       memset(buf, 'D', sizeof buf);
+       br_shake_init(&sc, security_level);
+       for (i = 0; i < 10; i ++) {
+               br_shake_inject(&sc, buf, sizeof buf);
+       }
+       num = 10;
+       for (;;) {
+               clock_t begin, end;
+               double tt;
+               long k;
+
+               begin = clock();
+               for (k = num; k > 0; k --) {
+                       br_shake_inject(&sc, buf, sizeof buf);
+               }
+               end = clock();
+               tt = (double)(end - begin) / CLOCKS_PER_SEC;
+               if (tt >= 2.0) {
+                       printf("SHAKE%-3d (inject)              %8.2f MB/s\n",
+                               security_level,
+                               ((double)sizeof buf) * (double)num
+                               / (tt * 1000000.0));
+                       fflush(stdout);
+                       break;
+               }
+               num <<= 1;
+       }
+
+       br_shake_flip(&sc);
+       for (i = 0; i < 10; i ++) {
+               br_shake_produce(&sc, buf, sizeof buf);
+       }
+
+       num = 10;
+       for (;;) {
+               clock_t begin, end;
+               double tt;
+               long k;
+
+               begin = clock();
+               for (k = num; k > 0; k --) {
+                       br_shake_produce(&sc, buf, sizeof buf);
+               }
+               end = clock();
+               tt = (double)(end - begin) / CLOCKS_PER_SEC;
+               if (tt >= 2.0) {
+                       printf("SHAKE%-3d (produce)             %8.2f MB/s\n",
+                               security_level,
+                               ((double)sizeof buf) * (double)num
+                               / (tt * 1000000.0));
+                       fflush(stdout);
+                       break;
+               }
+               num <<= 1;
+       }
+}
+
+static void
+test_speed_shake128(void)
+{
+       test_speed_shake_inner(128);
+}
+
+static void
+test_speed_shake256(void)
+{
+       test_speed_shake_inner(256);
+}
+
 static const unsigned char RSA_N[] = {
        0xE9, 0xF2, 0x4A, 0x2F, 0x96, 0xDF, 0x0A, 0x23,
        0x01, 0x85, 0xF1, 0x2C, 0xB2, 0xA8, 0xEF, 0x23,
@@ -585,11 +759,16 @@ static const br_rsa_private_key RSA_SK = {
 
 static void
 test_speed_rsa_inner(char *name,
-       br_rsa_public fpub, br_rsa_private fpriv)
+       br_rsa_public fpub, br_rsa_private fpriv, br_rsa_keygen kgen)
 {
        unsigned char tmp[sizeof RSA_N];
        int i;
        long num;
+       /*
+       br_hmac_drbg_context rng;
+       */
+       br_aesctr_drbg_context rng;
+       const br_block_ctr_class *ictr;
 
        memset(tmp, 'R', sizeof tmp);
        tmp[0] = 0;
@@ -643,27 +822,96 @@ test_speed_rsa_inner(char *name,
                }
                num <<= 1;
        }
+
+       if (kgen == 0) {
+               printf("%-30s KEYGEN UNAVAILABLE\n", name);
+               fflush(stdout);
+               return;
+       }
+       /*
+       br_hmac_drbg_init(&rng, &br_sha256_vtable, "RSA keygen seed", 15);
+       */
+       ictr = br_aes_x86ni_ctr_get_vtable();
+       if (ictr == NULL) {
+               ictr = br_aes_pwr8_ctr_get_vtable();
+               if (ictr == NULL) {
+#if BR_64
+                       ictr = &br_aes_ct64_ctr_vtable;
+#else
+                       ictr = &br_aes_ct_ctr_vtable;
+#endif
+               }
+       }
+       br_aesctr_drbg_init(&rng, ictr, "RSA keygen seed", 15);
+
+       num = 10;
+       for (;;) {
+               clock_t begin, end;
+               double tt;
+               long k;
+
+               begin = clock();
+               for (k = num; k > 0; k --) {
+                       br_rsa_private_key sk;
+                       unsigned char kbuf[BR_RSA_KBUF_PRIV_SIZE(1024)];
+
+                       kgen(&rng.vtable, &sk, kbuf, NULL, NULL, 1024, 0);
+               }
+               end = clock();
+               tt = (double)(end - begin) / CLOCKS_PER_SEC;
+               if (tt >= 10.0) {
+                       printf("%-30s %8.2f kgen[1024]/s\n", name,
+                               (double)num / tt);
+                       fflush(stdout);
+                       break;
+               }
+               num <<= 1;
+       }
+
+       num = 10;
+       for (;;) {
+               clock_t begin, end;
+               double tt;
+               long k;
+
+               begin = clock();
+               for (k = num; k > 0; k --) {
+                       br_rsa_private_key sk;
+                       unsigned char kbuf[BR_RSA_KBUF_PRIV_SIZE(2048)];
+
+                       kgen(&rng.vtable, &sk, kbuf, NULL, NULL, 2048, 0);
+               }
+               end = clock();
+               tt = (double)(end - begin) / CLOCKS_PER_SEC;
+               if (tt >= 10.0) {
+                       printf("%-30s %8.2f kgen[2048]/s\n", name,
+                               (double)num / tt);
+                       fflush(stdout);
+                       break;
+               }
+               num <<= 1;
+       }
 }
 
 static void
 test_speed_rsa_i15(void)
 {
        test_speed_rsa_inner("RSA i15",
-               &br_rsa_i15_public, &br_rsa_i15_private);
+               &br_rsa_i15_public, &br_rsa_i15_private, &br_rsa_i15_keygen);
 }
 
 static void
 test_speed_rsa_i31(void)
 {
        test_speed_rsa_inner("RSA i31",
-               &br_rsa_i31_public, &br_rsa_i31_private);
+               &br_rsa_i31_public, &br_rsa_i31_private, &br_rsa_i31_keygen);
 }
 
 static void
 test_speed_rsa_i32(void)
 {
        test_speed_rsa_inner("RSA i32",
-               &br_rsa_i32_public, &br_rsa_i32_private);
+               &br_rsa_i32_public, &br_rsa_i32_private, 0);
 }
 
 static void
@@ -671,11 +919,13 @@ test_speed_rsa_i62(void)
 {
        br_rsa_public pub;
        br_rsa_private priv;
+       br_rsa_keygen kgen;
 
        pub = br_rsa_i62_public_get();
        priv = br_rsa_i62_private_get();
+       kgen = br_rsa_i62_keygen_get();
        if (pub) {
-               test_speed_rsa_inner("RSA i62", pub, priv);
+               test_speed_rsa_inner("RSA i62", pub, priv, kgen);
        } else {
                printf("%-30s UNAVAILABLE\n", "RSA i62");
        }
@@ -979,14 +1229,20 @@ static void
 test_speed_i31(void)
 {
        static const unsigned char bp[] = {
-               0xFF, 0xFF, 0xFF, 0xFF, 0x00, 0x00, 0x00, 0x00,
+               /* A 521-bit prime integer (order of the P-521 curve). */
+               0x01, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+               0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
                0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
-               0xBC, 0xE6, 0xFA, 0xAD, 0xA7, 0x17, 0x9E, 0x84,
-               0xF3, 0xB9, 0xCA, 0xC2, 0xFC, 0x63, 0x25, 0x51
+               0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+               0xFF, 0xFA, 0x51, 0x86, 0x87, 0x83, 0xBF, 0x2F,
+               0x96, 0x6B, 0x7F, 0xCC, 0x01, 0x48, 0xF7, 0x09,
+               0xA5, 0xD0, 0x3B, 0xB5, 0xC9, 0xB8, 0x89, 0x9C,
+               0x47, 0xAE, 0xBB, 0x6F, 0xB7, 0x1E, 0x91, 0x38,
+               0x64, 0x09
        };
 
        unsigned char tmp[60 + sizeof bp];
-       uint32_t p[10], x[10], y[10], z[10], p0i;
+       uint32_t p[20], x[20], y[20], z[20], uu[60], p0i;
        int i;
        long num;
 
@@ -1068,6 +1324,30 @@ test_speed_i31(void)
                }
                num <<= 1;
        }
+
+       for (i = 0; i < 10; i ++) {
+               br_i31_moddiv(x, y, p, p0i, uu);
+       }
+       num = 10;
+       for (;;) {
+               clock_t begin, end;
+               double tt;
+               long k;
+
+               begin = clock();
+               for (k = num; k > 0; k --) {
+                       br_i31_moddiv(x, y, p, p0i, uu);
+               }
+               end = clock();
+               tt = (double)(end - begin) / CLOCKS_PER_SEC;
+               if (tt >= 2.0) {
+                       printf("%-30s %8.2f ops/s\n", "i31 moddiv",
+                               (double)num / tt);
+                       fflush(stdout);
+                       break;
+               }
+               num <<= 1;
+       }
 }
 
 #if 0
@@ -1279,6 +1559,7 @@ static const struct {
        STU(3des_ct_cbcdec),
 
        STU(chacha20_ct),
+       STU(chacha20_sse2),
 
        STU(ghash_ctmul),
        STU(ghash_ctmul32),
@@ -1291,6 +1572,28 @@ static const struct {
        STU(poly1305_ctmulq),
        STU(poly1305_i15),
 
+       STU(eax_aes128_big),
+       STU(eax_aes192_big),
+       STU(eax_aes256_big),
+       STU(eax_aes128_small),
+       STU(eax_aes192_small),
+       STU(eax_aes256_small),
+       STU(eax_aes128_ct),
+       STU(eax_aes192_ct),
+       STU(eax_aes256_ct),
+       STU(eax_aes128_ct64),
+       STU(eax_aes192_ct64),
+       STU(eax_aes256_ct64),
+       STU(eax_aes128_x86ni),
+       STU(eax_aes192_x86ni),
+       STU(eax_aes256_x86ni),
+       STU(eax_aes128_pwr8),
+       STU(eax_aes192_pwr8),
+       STU(eax_aes256_pwr8),
+
+       STU(shake128),
+       STU(shake256),
+
        STU(rsa_i15),
        STU(rsa_i31),
        STU(rsa_i32),