When using Clang, use it also for linking (compatibility with core FreeBSD systems).

[BearSSL] / src / inner.h

diff --git a/src/inner.h b/src/inner.h
index 2bb8192..3cafa42 100644 (file)
--- a/src/inner.h
+++ b/src/inner.h
@@ -105,58 +105,238 @@
 #endif
 #endif
 
 #endif
 #endif
 

+/*
+ * Set BR_LOMUL on platforms where it makes sense.
+ */
+#ifndef BR_LOMUL
+#if BR_ARMEL_CORTEX_GCC
+#define BR_LOMUL   1
+#endif
+#endif
+
+/*
+ * Determine whether x86 AES instructions are understood by the compiler.
+ */
+#ifndef BR_AES_X86NI
+#if (__i386__ || __x86_64__) \
+       && ((__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8)) \
+           || (__clang_major__ > 3 \
+               || (__clang_major__ == 3 && __clang_minor__ >= 7)))
+#define BR_AES_X86NI   1
+#elif (_M_IX86 || _M_X64) && (_MSC_VER >= 1700)
+#define BR_AES_X86NI   1
+#endif
+#endif
+
+/*
+ * If we use x86 AES instruction, determine the compiler brand.
+ */
+#if BR_AES_X86NI
+#ifndef BR_AES_X86NI_GCC
+#if __GNUC__
+#define BR_AES_X86NI_GCC   1
+#endif
+#endif
+#ifndef BR_AES_X86NI_MSC
+#if _MSC_VER >= 1700
+#define BR_AES_X86NI_MSC   1
+#endif
+#endif
+#endif
+
+/*
+ * A macro to tag a function with a "target" attribute (for GCC and Clang).
+ */
+#if BR_AES_X86NI_GCC
+#define BR_TARGET(x)   __attribute__((target(x)))
+#else
+#define BR_TARGET(x)
+#endif
+
+/*
+ * GCC versions from 4.4 to 4.8 (inclusive) must use a special #pragma
+ * to activate extra opcodes before including the relevant intrinsic
+ * headers. But these don't work with Clang (which does not need them
+ * either).
+ */
+#if BR_AES_X86NI_GCC && !defined BR_AES_X86NI_GCC_OLD
+#if __GNUC__ == 4 && __GNUC_MINOR__ >= 4 && __GNUC_MINOR__ <= 8 && !__clang__
+#define BR_AES_X86NI_GCC_OLD   1
+#endif
+#endif
+
+/*
+ * POWER8 crypto support. We rely on compiler macros for the
+ * architecture, since we do not have a reliable, simple way to detect
+ * the required support at runtime (we could try running an opcode, and
+ * trapping the exception or signal on illegal instruction, but this
+ * induces some non-trivial OS dependencies that we would prefer to
+ * avoid if possible).
+ */
+#ifndef BR_POWER8
+#if __GNUC__ && ((_ARCH_PWR8 || _ARCH_PPC) && __CRYPTO__)
+#define BR_POWER8   1
+#endif
+#endif
+
+/*
+ * Detect endinanness on POWER8.
+ */
+#if BR_POWER8
+#if defined BR_POWER8_LE
+#undef BR_POWER8_BE
+#if BR_POWER8_LE
+#define BR_POWER8_BE   0
+#else
+#define BR_POWER8_BE   1
+#endif
+#elif defined BR_POWER8_BE
+#undef BR_POWER8_LE
+#if BR_POWER8_BE
+#define BR_POWER8_LE   0
+#else
+#define BR_POWER8_LE   1
+#endif
+#else
+#if __LITTLE_ENDIAN__
+#define BR_POWER8_LE   1
+#define BR_POWER8_BE   0
+#else
+#define BR_POWER8_LE   0
+#define BR_POWER8_BE   1
+#endif
+#endif
+#endif
+
+/*
+ * Detect support for 128-bit integers.
+ */
+#if !defined BR_INT128 && !defined BR_UMUL128
+#ifdef __SIZEOF_INT128__
+#define BR_INT128    1
+#elif _M_X64
+#define BR_UMUL128   1
+#endif
+#endif
+
+/*
+ * Detect support for unaligned accesses with known endianness.
+ *
+ *  x86 (both 32-bit and 64-bit) is little-endian and allows unaligned
+ *  accesses.
+ *
+ *  POWER/PowerPC allows unaligned accesses when big-endian. POWER8 and
+ *  later also allow unaligned accesses when little-endian.
+ */
+#if !defined BR_LE_UNALIGNED && !defined BR_BE_UNALIGNED
+
+#if __i386 || __i386__ || __x86_64__ || _M_IX86 || _M_X64
+#define BR_LE_UNALIGNED   1
+#elif BR_POWER8_BE
+#define BR_BE_UNALIGNED   1
+#elif BR_POWER8_LE
+#define BR_LE_UNALIGNED   1
+#elif (__powerpc__ || __powerpc64__ || _M_PPC || _ARCH_PPC || _ARCH_PPC64) \
+       && __BIG_ENDIAN__
+#define BR_BE_UNALIGNED   1
+#endif
+
+#endif
+

 /* ==================================================================== */
 /*
  * Encoding/decoding functions.
  *
  * 32-bit and 64-bit decoding, both little-endian and big-endian, is
 /* ==================================================================== */
 /*
  * Encoding/decoding functions.
  *
  * 32-bit and 64-bit decoding, both little-endian and big-endian, is

- * implemented with the inline functions below. These functions are
- * generic: they don't depend on the architecture natural endianness,
- * and they can handle unaligned accesses. Optimized versions for some
- * specific architectures may be implemented at a later time.
- */
+ * implemented with the inline functions below.
+ *
+ * When allowed by some compile-time options (autodetected or provided),
+ * optimised code is used, to perform direct memory access when the
+ * underlying architecture supports it, both for endianness and
+ * alignment. This, however, may trigger strict aliasing issues; the
+ * code below uses unions to perform (supposedly) safe type punning.
+ * Since the C aliasing rules are relatively complex and were amended,
+ * or at least re-explained with different phrasing, in all successive
+ * versions of the C standard, it is always a bit risky to bet that any
+ * specific version of a C compiler got it right, for some notion of
+ * "right".
+ */
+
+typedef union {
+       uint16_t u;
+       unsigned char b[sizeof(uint16_t)];
+} br_union_u16;
+
+typedef union {
+       uint32_t u;
+       unsigned char b[sizeof(uint32_t)];
+} br_union_u32;
+
+typedef union {
+       uint64_t u;
+       unsigned char b[sizeof(uint64_t)];
+} br_union_u64;

 
 static inline void
 br_enc16le(void *dst, unsigned x)
 {
 
 static inline void
 br_enc16le(void *dst, unsigned x)
 {

+#if BR_LE_UNALIGNED
+       ((br_union_u16 *)dst)->u = x;
+#else

        unsigned char *buf;
 
        buf = dst;
        buf[0] = (unsigned char)x;
        buf[1] = (unsigned char)(x >> 8);
        unsigned char *buf;
 
        buf = dst;
        buf[0] = (unsigned char)x;
        buf[1] = (unsigned char)(x >> 8);

+#endif

 }
 
 static inline void
 br_enc16be(void *dst, unsigned x)
 {
 }
 
 static inline void
 br_enc16be(void *dst, unsigned x)
 {

+#if BR_BE_UNALIGNED
+       ((br_union_u16 *)dst)->u = x;
+#else

        unsigned char *buf;
 
        buf = dst;
        buf[0] = (unsigned char)(x >> 8);
        buf[1] = (unsigned char)x;
        unsigned char *buf;
 
        buf = dst;
        buf[0] = (unsigned char)(x >> 8);
        buf[1] = (unsigned char)x;

+#endif

 }
 
 static inline unsigned
 br_dec16le(const void *src)
 {
 }
 
 static inline unsigned
 br_dec16le(const void *src)
 {

+#if BR_LE_UNALIGNED
+       return ((const br_union_u16 *)src)->u;
+#else

        const unsigned char *buf;
 
        buf = src;
        return (unsigned)buf[0] | ((unsigned)buf[1] << 8);
        const unsigned char *buf;
 
        buf = src;
        return (unsigned)buf[0] | ((unsigned)buf[1] << 8);

+#endif

 }
 
 static inline unsigned
 br_dec16be(const void *src)
 {
 }
 
 static inline unsigned
 br_dec16be(const void *src)
 {

+#if BR_BE_UNALIGNED
+       return ((const br_union_u16 *)src)->u;
+#else

        const unsigned char *buf;
 
        buf = src;
        return ((unsigned)buf[0] << 8) | (unsigned)buf[1];
        const unsigned char *buf;
 
        buf = src;
        return ((unsigned)buf[0] << 8) | (unsigned)buf[1];

+#endif

 }
 
 static inline void
 br_enc32le(void *dst, uint32_t x)
 {
 }
 
 static inline void
 br_enc32le(void *dst, uint32_t x)
 {

+#if BR_LE_UNALIGNED
+       ((br_union_u32 *)dst)->u = x;
+#else

        unsigned char *buf;
 
        buf = dst;
        unsigned char *buf;
 
        buf = dst;


@@ -164,11 +344,15 @@ br_enc32le(void *dst, uint32_t x)
        buf[1] = (unsigned char)(x >> 8);
        buf[2] = (unsigned char)(x >> 16);
        buf[3] = (unsigned char)(x >> 24);
        buf[1] = (unsigned char)(x >> 8);
        buf[2] = (unsigned char)(x >> 16);
        buf[3] = (unsigned char)(x >> 24);

+#endif

 }
 
 static inline void
 br_enc32be(void *dst, uint32_t x)
 {
 }
 
 static inline void
 br_enc32be(void *dst, uint32_t x)
 {

+#if BR_BE_UNALIGNED
+       ((br_union_u32 *)dst)->u = x;
+#else

        unsigned char *buf;
 
        buf = dst;
        unsigned char *buf;
 
        buf = dst;


@@ -176,11 +360,15 @@ br_enc32be(void *dst, uint32_t x)
        buf[1] = (unsigned char)(x >> 16);
        buf[2] = (unsigned char)(x >> 8);
        buf[3] = (unsigned char)x;
        buf[1] = (unsigned char)(x >> 16);
        buf[2] = (unsigned char)(x >> 8);
        buf[3] = (unsigned char)x;

+#endif

 }
 
 static inline uint32_t
 br_dec32le(const void *src)
 {
 }
 
 static inline uint32_t
 br_dec32le(const void *src)
 {

+#if BR_LE_UNALIGNED
+       return ((const br_union_u32 *)src)->u;
+#else

        const unsigned char *buf;
 
        buf = src;
        const unsigned char *buf;
 
        buf = src;


@@ -188,11 +376,15 @@ br_dec32le(const void *src)
                | ((uint32_t)buf[1] << 8)
                | ((uint32_t)buf[2] << 16)
                | ((uint32_t)buf[3] << 24);
                | ((uint32_t)buf[1] << 8)
                | ((uint32_t)buf[2] << 16)
                | ((uint32_t)buf[3] << 24);

+#endif

 }
 
 static inline uint32_t
 br_dec32be(const void *src)
 {
 }
 
 static inline uint32_t
 br_dec32be(const void *src)
 {

+#if BR_BE_UNALIGNED
+       return ((const br_union_u32 *)src)->u;
+#else

        const unsigned char *buf;
 
        buf = src;
        const unsigned char *buf;
 
        buf = src;


@@ -200,46 +392,63 @@ br_dec32be(const void *src)
                | ((uint32_t)buf[1] << 16)
                | ((uint32_t)buf[2] << 8)
                | (uint32_t)buf[3];
                | ((uint32_t)buf[1] << 16)
                | ((uint32_t)buf[2] << 8)
                | (uint32_t)buf[3];

+#endif

 }
 
 static inline void
 br_enc64le(void *dst, uint64_t x)
 {
 }
 
 static inline void
 br_enc64le(void *dst, uint64_t x)
 {

+#if BR_LE_UNALIGNED
+       ((br_union_u64 *)dst)->u = x;
+#else

        unsigned char *buf;
 
        buf = dst;
        br_enc32le(buf, (uint32_t)x);
        br_enc32le(buf + 4, (uint32_t)(x >> 32));
        unsigned char *buf;
 
        buf = dst;
        br_enc32le(buf, (uint32_t)x);
        br_enc32le(buf + 4, (uint32_t)(x >> 32));

+#endif

 }
 
 static inline void
 br_enc64be(void *dst, uint64_t x)
 {
 }
 
 static inline void
 br_enc64be(void *dst, uint64_t x)
 {

+#if BR_BE_UNALIGNED
+       ((br_union_u64 *)dst)->u = x;
+#else

        unsigned char *buf;
 
        buf = dst;
        br_enc32be(buf, (uint32_t)(x >> 32));
        br_enc32be(buf + 4, (uint32_t)x);
        unsigned char *buf;
 
        buf = dst;
        br_enc32be(buf, (uint32_t)(x >> 32));
        br_enc32be(buf + 4, (uint32_t)x);

+#endif

 }
 
 static inline uint64_t
 br_dec64le(const void *src)
 {
 }
 
 static inline uint64_t
 br_dec64le(const void *src)
 {

+#if BR_LE_UNALIGNED
+       return ((const br_union_u64 *)src)->u;
+#else

        const unsigned char *buf;
 
        buf = src;
        return (uint64_t)br_dec32le(buf)
                | ((uint64_t)br_dec32le(buf + 4) << 32);
        const unsigned char *buf;
 
        buf = src;
        return (uint64_t)br_dec32le(buf)
                | ((uint64_t)br_dec32le(buf + 4) << 32);

+#endif

 }
 
 static inline uint64_t
 br_dec64be(const void *src)
 {
 }
 
 static inline uint64_t
 br_dec64be(const void *src)
 {

+#if BR_BE_UNALIGNED
+       return ((const br_union_u64 *)src)->u;
+#else

        const unsigned char *buf;
 
        buf = src;
        return ((uint64_t)br_dec32be(buf) << 32)
                | (uint64_t)br_dec32be(buf + 4);
        const unsigned char *buf;
 
        buf = src;
        return ((uint64_t)br_dec32be(buf) << 32)
                | (uint64_t)br_dec32be(buf + 4);

+#endif

 }
 
 /*
 }
 
 /*


@@ -1062,6 +1271,25 @@ void br_i31_from_monty(uint32_t *x, const uint32_t *m, uint32_t m0i);
 void br_i31_modpow(uint32_t *x, const unsigned char *e, size_t elen,
        const uint32_t *m, uint32_t m0i, uint32_t *t1, uint32_t *t2);
 
 void br_i31_modpow(uint32_t *x, const unsigned char *e, size_t elen,
        const uint32_t *m, uint32_t m0i, uint32_t *t1, uint32_t *t2);
 

+/*
+ * Compute a modular exponentiation. x[] MUST be an integer modulo m[]
+ * (same announced bit length, lower value). m[] MUST be odd. The
+ * exponent is in big-endian unsigned notation, over 'elen' bytes. The
+ * "m0i" parameter is equal to -(1/m0) mod 2^31, where m0 is the least
+ * significant value word of m[] (this works only if m[] is an odd
+ * integer). The tmp[] array is used for temporaries, and has size
+ * 'twlen' words; it must be large enough to accommodate at least two
+ * temporary values with the same size as m[] (including the leading
+ * "bit length" word). If there is room for more temporaries, then this
+ * function may use the extra room for window-based optimisation,
+ * resulting in faster computations.
+ *
+ * Returned value is 1 on success, 0 on error. An error is reported if
+ * the provided tmp[] array is too short.
+ */
+uint32_t br_i31_modpow_opt(uint32_t *x, const unsigned char *e, size_t elen,
+       const uint32_t *m, uint32_t m0i, uint32_t *tmp, size_t twlen);
+

 /*
  * Compute d+a*b, result in d. The initial announced bit length of d[]
  * MUST match that of a[]. The d[] array MUST be large enough to
 /*
  * Compute d+a*b, result in d. The initial announced bit length of d[]
  * MUST match that of a[]. The d[] array MUST be large enough to


@@ -1077,6 +1305,10 @@ void br_i31_mulacc(uint32_t *d, const uint32_t *a, const uint32_t *b);
 
 /* ==================================================================== */
 
 
 /* ==================================================================== */
 

+/*
+ * FIXME: document "i15" functions.
+ */
+

 static inline void
 br_i15_zero(uint16_t *x, uint16_t bit_len)
 {
 static inline void
 br_i15_zero(uint16_t *x, uint16_t bit_len)
 {


@@ -1125,6 +1357,9 @@ void br_i15_reduce(uint16_t *x, const uint16_t *a, const uint16_t *m);
 
 void br_i15_mulacc(uint16_t *d, const uint16_t *a, const uint16_t *b);
 
 
 void br_i15_mulacc(uint16_t *d, const uint16_t *a, const uint16_t *b);
 

+uint32_t br_i62_modpow_opt(uint32_t *x31, const unsigned char *e, size_t elen,
+       const uint32_t *m31, uint32_t m0i31, uint64_t *tmp, size_t twlen);
+

 /* ==================================================================== */
 
 static inline size_t
 /* ==================================================================== */
 
 static inline size_t


@@ -1424,6 +1659,40 @@ unsigned br_aes_ct64_keysched(uint64_t *comp_skey,
 void br_aes_ct64_skey_expand(uint64_t *skey,
        unsigned num_rounds, const uint64_t *comp_skey);
 
 void br_aes_ct64_skey_expand(uint64_t *skey,
        unsigned num_rounds, const uint64_t *comp_skey);
 

+/*
+ * Test support for AES-NI opcodes.
+ */
+int br_aes_x86ni_supported(void);
+
+/*
+ * AES key schedule, using x86 AES-NI instructions. This yields the
+ * subkeys in the encryption direction. Number of rounds is returned.
+ * Key size MUST be 16, 24 or 32 bytes; otherwise, 0 is returned.
+ */
+unsigned br_aes_x86ni_keysched_enc(unsigned char *skni,
+       const void *key, size_t len);
+
+/*
+ * AES key schedule, using x86 AES-NI instructions. This yields the
+ * subkeys in the decryption direction. Number of rounds is returned.
+ * Key size MUST be 16, 24 or 32 bytes; otherwise, 0 is returned.
+ */
+unsigned br_aes_x86ni_keysched_dec(unsigned char *skni,
+       const void *key, size_t len);
+
+/*
+ * Test support for AES POWER8 opcodes.
+ */
+int br_aes_pwr8_supported(void);
+
+/*
+ * AES key schedule, using POWER8 instructions. This yields the
+ * subkeys in the encryption direction. Number of rounds is returned.
+ * Key size MUST be 16, 24 or 32 bytes; otherwise, 0 is returned.
+ */
+unsigned br_aes_pwr8_keysched(unsigned char *skni,
+       const void *key, size_t len);
+

 /* ==================================================================== */
 /*
  * RSA.
 /* ==================================================================== */
 /*
  * RSA.


@@ -1701,4 +1970,85 @@ int br_ssl_choose_hash(unsigned bf);
 
 /* ==================================================================== */
 
 
 /* ==================================================================== */
 

+/*
+ * PowerPC / POWER assembly stuff. The special BR_POWER_ASM_MACROS macro
+ * must be defined before including this file; this is done by source
+ * files that use some inline assembly for PowerPC / POWER machines.
+ */
+
+#if BR_POWER_ASM_MACROS
+
+#define lxvw4x(xt, ra, rb)        lxvw4x_(xt, ra, rb)
+#define stxvw4x(xt, ra, rb)       stxvw4x_(xt, ra, rb)
+
+#define bdnz(foo)                 bdnz_(foo)
+#define beq(foo)                  beq_(foo)
+
+#define li(rx, value)             li_(rx, value)
+#define addi(rx, ra, imm)         addi_(rx, ra, imm)
+#define cmpldi(rx, imm)           cmpldi_(rx, imm)
+#define mtctr(rx)                 mtctr_(rx)
+#define vspltb(vrt, vrb, uim)     vspltb_(vrt, vrb, uim)
+#define vspltw(vrt, vrb, uim)     vspltw_(vrt, vrb, uim)
+#define vspltisb(vrt, imm)        vspltisb_(vrt, imm)
+#define vspltisw(vrt, imm)        vspltisw_(vrt, imm)
+#define vrlw(vrt, vra, vrb)       vrlw_(vrt, vra, vrb)
+#define vsbox(vrt, vra)           vsbox_(vrt, vra)
+#define vxor(vrt, vra, vrb)       vxor_(vrt, vra, vrb)
+#define vand(vrt, vra, vrb)       vand_(vrt, vra, vrb)
+#define vsro(vrt, vra, vrb)       vsro_(vrt, vra, vrb)
+#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 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)
+#define vsrw(vrt, vra, vrb)       vsrw_(vrt, vra, vrb)
+#define vcipher(vt, va, vb)       vcipher_(vt, va, vb)
+#define vcipherlast(vt, va, vb)   vcipherlast_(vt, va, vb)
+#define vncipher(vt, va, vb)      vncipher_(vt, va, vb)
+#define vncipherlast(vt, va, vb)  vncipherlast_(vt, va, vb)
+#define vperm(vt, va, vb, vc)     vperm_(vt, va, vb, vc)
+#define vpmsumd(vt, va, vb)       vpmsumd_(vt, va, vb)
+#define xxpermdi(vt, va, vb, d)   xxpermdi_(vt, va, vb, d)
+
+#define lxvw4x_(xt, ra, rb)       "\tlxvw4x\t" #xt "," #ra "," #rb "\n"
+#define stxvw4x_(xt, ra, rb)      "\tstxvw4x\t" #xt "," #ra "," #rb "\n"
+
+#define label(foo)                #foo "%=:\n"
+#define bdnz_(foo)                "\tbdnz\t" #foo "%=\n"
+#define beq_(foo)                 "\tbeq\t" #foo "%=\n"
+
+#define li_(rx, value)            "\tli\t" #rx "," #value "\n"
+#define addi_(rx, ra, imm)        "\taddi\t" #rx "," #ra "," #imm "\n"
+#define cmpldi_(rx, imm)          "\tcmpldi\t" #rx "," #imm "\n"
+#define mtctr_(rx)                "\tmtctr\t" #rx "\n"
+#define vspltb_(vrt, vrb, uim)    "\tvspltb\t" #vrt "," #vrb "," #uim "\n"
+#define vspltw_(vrt, vrb, uim)    "\tvspltw\t" #vrt "," #vrb "," #uim "\n"
+#define vspltisb_(vrt, imm)       "\tvspltisb\t" #vrt "," #imm "\n"
+#define vspltisw_(vrt, imm)       "\tvspltisw\t" #vrt "," #imm "\n"
+#define vrlw_(vrt, vra, vrb)      "\tvrlw\t" #vrt "," #vra "," #vrb "\n"
+#define vsbox_(vrt, vra)          "\tvsbox\t" #vrt "," #vra "\n"
+#define vxor_(vrt, vra, vrb)      "\tvxor\t" #vrt "," #vra "," #vrb "\n"
+#define vand_(vrt, vra, vrb)      "\tvand\t" #vrt "," #vra "," #vrb "\n"
+#define vsro_(vrt, vra, vrb)      "\tvsro\t" #vrt "," #vra "," #vrb "\n"
+#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 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"
+#define vsrw_(vrt, vra, vrb)      "\tvsrw\t" #vrt "," #vra "," #vrb "\n"
+#define vcipher_(vt, va, vb)      "\tvcipher\t" #vt "," #va "," #vb "\n"
+#define vcipherlast_(vt, va, vb)  "\tvcipherlast\t" #vt "," #va "," #vb "\n"
+#define vncipher_(vt, va, vb)     "\tvncipher\t" #vt "," #va "," #vb "\n"
+#define vncipherlast_(vt, va, vb) "\tvncipherlast\t" #vt "," #va "," #vb "\n"
+#define vperm_(vt, va, vb, vc)    "\tvperm\t" #vt "," #va "," #vb "," #vc "\n"
+#define vpmsumd_(vt, va, vb)      "\tvpmsumd\t" #vt "," #va "," #vb "\n"
+#define xxpermdi_(vt, va, vb, d)  "\txxpermdi\t" #vt "," #va "," #vb "," #d "\n"
+
+#endif
+
+/* ==================================================================== */
+

 #endif
 #endif