Does Linaro's GCC 4.9 include Crypto extensions and intrinsics?

[Jeffrey Walton]

Please forgive my ignorance. I'm working on a Pine64 dev-board Pine64
supplies Linaro's GCC 4.9.2 toolchain.

I am catching a compile error, and I am trying to determine why.

Does Linaro's GCC 4.9 provide AES and SHA intrinsics?

**********

$ uname -a
Linux pine64 3.10.102-2-pine64-longsleep #66 SMP PREEMPT Sat Jul 16
10:53:13 CEST 2016 aarch64 GNU/Linux

$ gcc --version
gcc (Debian/Linaro 4.9.2-10) 4.9.2
Copyright (C) 2014 Free Software Foundation, Inc.

**********

$ CFLAGS="-DMBEDTLS_HAVE_ARMV8A_CE=1" make DEBUG=1 V=1
CC aes_armv8a_ce.c
aes_armv8a_ce.c: In function 'mbedtls_armv8a_ce_aes_crypt_ecb':
aes_armv8a_ce.c:65:4: warning: implicit declaration of function
'vaeseq_u8' [-Wimplicit-function-declaration]
state_vec = vaeseq_u8( state_vec, roundkey_vec );
^
aes_armv8a_ce.c:65:14: error: incompatible types when assigning to
type 'uint8x16_t' from type 'int'
state_vec = vaeseq_u8( state_vec, roundkey_vec );
^
aes_armv8a_ce.c:67:4: warning: implicit declaration of function
'vaesmcq_u8' [-Wimplicit-function-declaration]
state_vec = vaesmcq_u8( state_vec );
^
aes_armv8a_ce.c:67:14: error: incompatible types when assigning to
type 'uint8x16_t' from type 'int'
state_vec = vaesmcq_u8( state_vec );
^
aes_armv8a_ce.c:74:13: error: incompatible types when assigning to
type 'uint8x16_t' from type 'int'
state_vec = vaeseq_u8( state_vec, roundkey_vec );
^
aes_armv8a_ce.c:84:4: warning: implicit declaration of function
'vaesdq_u8' [-Wimplicit-function-declaration]
state_vec = vaesdq_u8( state_vec, roundkey_vec );
^
aes_armv8a_ce.c:84:14: error: incompatible types when assigning to
type 'uint8x16_t' from type 'int'
state_vec = vaesdq_u8( state_vec, roundkey_vec );
^
aes_armv8a_ce.c:86:4: warning: implicit declaration of function
'vaesimcq_u8' [-Wimplicit-function-declaration]
state_vec = vaesimcq_u8( state_vec );
^
aes_armv8a_ce.c:86:14: error: incompatible types when assigning to
type 'uint8x16_t' from type 'int'
state_vec = vaesimcq_u8( state_vec );
^
aes_armv8a_ce.c:93:13: error: incompatible types when assigning to
type 'uint8x16_t' from type 'int'
state_vec = vaesdq_u8( state_vec, roundkey_vec );
^
aes_armv8a_ce.c: In function 'mbedtls_armv8a_ce_gcm_mult':
aes_armv8a_ce.c:138:2: warning: implicit declaration of function
'vmull_high_p64' [-Wimplicit-function-declaration]
r1 = (uint8x16_t)vmull_high_p64( (poly64x2_t)a_p, (poly64x2_t)b_p );
^
aes_armv8a_ce.c:138:2: error: can't convert between vector values of
different size
aes_armv8a_ce.c:141:2: error: can't convert between vector values of
different size
t0 = (uint8x16_t)vmull_high_p64( (poly64x2_t)a_p, (poly64x2_t)t0 );
^
aes_armv8a_ce.c:150:2: error: can't convert between vector values of
different size
t0 = (uint8x16_t)vmull_high_p64( (poly64x2_t)r1, (poly64x2_t)p );
^
Makefile:170: recipe for target 'aes_armv8a_ce.o' failed
make[1]: *** [aes_armv8a_ce.o] Error 1
Makefile:17: recipe for target 'lib' failed
make: *** [lib] Error 2

**********

CC sha1_armv8a_ce.c
sha1_armv8a_ce.c: In function 'mbedtls_armv8a_ce_sha1_process':
sha1_armv8a_ce.c:99:2: warning: implicit declaration of function
'vsha1h_u32' [-Wimplicit-function-declaration]
e1 = vsha1h_u32( a );
^
sha1_armv8a_ce.c:100:2: warning: implicit declaration of function
'vsha1cq_u32' [-Wimplicit-function-declaration]
abcd = vsha1cq_u32( abcd, e, wk0 ); /* 0 */
^
sha1_armv8a_ce.c:100:7: error: incompatible types when assigning to
type 'uint32x4_t' from type 'int'
abcd = vsha1cq_u32( abcd, e, wk0 ); /* 0 */
^
sha1_armv8a_ce.c:102:2: warning: implicit declaration of function
'vsha1su0q_u32' [-Wimplicit-function-declaration]
w0 = vsha1su0q_u32( w0, w1, w2 );
^
sha1_armv8a_ce.c:102:5: error: incompatible types when assigning to
type 'uint32x4_t' from type 'int'
w0 = vsha1su0q_u32( w0, w1, w2 );
^
sha1_armv8a_ce.c:106:7: error: incompatible types when assigning to
type 'uint32x4_t' from type 'int'
abcd = vsha1cq_u32( abcd, e1, wk1 ); /* 1 */
^
sha1_armv8a_ce.c:108:2: warning: implicit declaration of function
'vsha1su1q_u32' [-Wimplicit-function-declaration]
w0 = vsha1su1q_u32( w0, w3 );
^
sha1_armv8a_ce.c:108:5: error: incompatible types when assigning to
type 'uint32x4_t' from type 'int'
w0 = vsha1su1q_u32( w0, w3 );
^
sha1_armv8a_ce.c:109:5: error: incompatible types when assigning to
type 'uint32x4_t' from type 'int'
w1 = vsha1su0q_u32( w1, w2, w3 );
^
...

**********

$ grep -IR vaeseq_u8 /usr/include
/usr/include/clang/3.5.0/include/arm_neon.h:__ai uint8x16_t
vaeseq_u8(uint8x16_t __p0, uint8x16_t __p1) {
/usr/include/clang/3.5.0/include/arm_neon.h:__ai uint8x16_t
vaeseq_u8(uint8x16_t __p0, uint8x16_t __p1) {
/usr/include/clang/3.5/include/arm_neon.h:__ai uint8x16_t
vaeseq_u8(uint8x16_t __p0, uint8x16_t __p1) {
/usr/include/clang/3.5/include/arm_neon.h:__ai uint8x16_t
vaeseq_u8(uint8x16_t __p0, uint8x16_t __p1) {
$

[Jim Wilson]

Post by Jeffrey Walton
Does Linaro's GCC 4.9 provide AES and SHA intrinsics?

Crypto is an optional feature that is not enabled by default when
selecting the armv8-a architecture. You either need to specify a
processor that implements the crypto extension, or else you need to
specify the crypto extension when specifying the architecture. The
latter is probably easier. For instance, you can do
-march=armv8-a+crypto to enable the crypto support. You can also
enable crc support by adding "+crc", separately, or in combination
with +crypto.

Post by Jeffrey Walton
$ grep -IR vaeseq_u8 /usr/include
/usr/include/clang/3.5.0/include/arm_neon.h:__ai uint8x16_t
vaeseq_u8(uint8x16_t __p0, uint8x16_t __p1) {
...

GCC specific header files are normally not in /usr/include. On my
Ubuntu 14.04 system, it is
/usr/lib/gcc/aarch64-linux-gnu/4.8/include/arm_neon.h
and vaeseq_u8 is in there, but only enabled when crypto support is enabled.

Jim

[Jeffrey Walton]

Thanks guys, that was it.

I did not realize the build system was omitting the flags.

On Fri, Jan 13, 2017 at 2:23 AM, Pinski, Andrew

Can you try -march=armv8+crypto ?
-----Original Message-----
Sent: Thursday, January 12, 2017 10:56 PM
Subject: Does Linaro's GCC 4.9 include Crypto extensions and intrinsics?
Please forgive my ignorance. I'm working on a Pine64 dev-board Pine64 supplies Linaro's GCC 4.9.2 toolchain.
I am catching a compile error, and I am trying to determine why.
Does Linaro's GCC 4.9 provide AES and SHA intrinsics?