Some small performance improvements on 32-bit architectures.

[BearSSL] / src / inner.h

diff --git a/src/inner.h b/src/inner.h
index d7e4091..21d996e 100644 (file)
--- a/src/inner.h
+++ b/src/inner.h
@@ -114,6 +114,8 @@
 #define BR_64   1
 #elif defined(__x86_64__) || defined(_M_X64)
 #define BR_64   1
+#elif defined(__aarch64__) || defined(_M_ARM64)
+#define BR_64   1
 #endif
 #endif
 
@@ -1943,6 +1945,27 @@ uint32_t br_rsa_pkcs1_sig_unpad(const unsigned char *sig, size_t sig_len,
        const unsigned char *hash_oid, size_t hash_len,
        unsigned char *hash_out);
 
+/*
+ * Apply proper PSS padding. The 'x' buffer is output only: it
+ * receives the value that is to be exponentiated.
+ */
+uint32_t br_rsa_pss_sig_pad(const br_prng_class **rng,
+       const br_hash_class *hf_data, const br_hash_class *hf_mgf1,
+       const unsigned char *hash, size_t salt_len,
+       uint32_t n_bitlen, unsigned char *x);
+
+/*
+ * Check PSS padding. The provided value is the one _after_
+ * the modular exponentiation; it is modified by this function.
+ * This function infers the signature length from the public key
+ * size, i.e. it assumes that this has already been verified (as
+ * part of the exponentiation).
+ */
+uint32_t br_rsa_pss_sig_unpad(
+       const br_hash_class *hf_data, const br_hash_class *hf_mgf1,
+       const unsigned char *hash, size_t salt_len,
+       const br_rsa_public_key *pk, unsigned char *x);
+
 /*
  * Apply OAEP padding. Returned value is the actual padded string length,
  * or zero on error.
@@ -2337,6 +2360,7 @@ int br_ssl_choose_hash(unsigned bf);
 #define stxvw4x(xt, ra, rb)       stxvw4x_(xt, ra, rb)
 
 #define bdnz(foo)                 bdnz_(foo)
+#define bdz(foo)                  bdz_(foo)
 #define beq(foo)                  beq_(foo)
 
 #define li(rx, value)             li_(rx, value)
@@ -2355,6 +2379,7 @@ int br_ssl_choose_hash(unsigned bf);
 #define vsl(vrt, vra, vrb)        vsl_(vrt, vra, vrb)
 #define vsldoi(vt, va, vb, sh)    vsldoi_(vt, va, vb, sh)
 #define vsr(vrt, vra, vrb)        vsr_(vrt, vra, vrb)
+#define vaddcuw(vrt, vra, vrb)    vaddcuw_(vrt, vra, vrb)
 #define vadduwm(vrt, vra, vrb)    vadduwm_(vrt, vra, vrb)
 #define vsububm(vrt, vra, vrb)    vsububm_(vrt, vra, vrb)
 #define vsubuwm(vrt, vra, vrb)    vsubuwm_(vrt, vra, vrb)
@@ -2372,6 +2397,7 @@ int br_ssl_choose_hash(unsigned bf);
 
 #define label(foo)                #foo "%=:\n"
 #define bdnz_(foo)                "\tbdnz\t" #foo "%=\n"
+#define bdz_(foo)                 "\tbdz\t" #foo "%=\n"
 #define beq_(foo)                 "\tbeq\t" #foo "%=\n"
 
 #define li_(rx, value)            "\tli\t" #rx "," #value "\n"
@@ -2390,6 +2416,7 @@ int br_ssl_choose_hash(unsigned bf);
 #define vsl_(vrt, vra, vrb)       "\tvsl\t" #vrt "," #vra "," #vrb "\n"
 #define vsldoi_(vt, va, vb, sh)   "\tvsldoi\t" #vt "," #va "," #vb "," #sh "\n"
 #define vsr_(vrt, vra, vrb)       "\tvsr\t" #vrt "," #vra "," #vrb "\n"
+#define vaddcuw_(vrt, vra, vrb)   "\tvaddcuw\t" #vrt "," #vra "," #vrb "\n"
 #define vadduwm_(vrt, vra, vrb)   "\tvadduwm\t" #vrt "," #vra "," #vrb "\n"
 #define vsububm_(vrt, vra, vrb)   "\tvsububm\t" #vrt "," #vra "," #vrb "\n"
 #define vsubuwm_(vrt, vra, vrb)   "\tvsubuwm\t" #vrt "," #vra "," #vrb "\n"
@@ -2444,8 +2471,8 @@ int br_ssl_choose_hash(unsigned bf);
 #else
 #define BR_TARGETS_X86_UP \
        _Pragma("GCC target(\"sse2,ssse3,sse4.1,aes,pclmul\")")
-#endif
 #define BR_TARGETS_X86_DOWN
+#endif
 #pragma GCC diagnostic ignored "-Wpsabi"
 #endif