Implementation notes: amd64, margaux, crypto_stream/chacha12

Test failure

error 111

Compiler output

chacha.c: In file included from chacha.c:103:
chacha.c: ./u4.h:122:7: error: always_inline function '_mm_shuffle_epi8' requires target feature 'ssse3', but would be inlined into function 'crypto_stream_chacha12_dolbeau_amd64_avx2_constbranchindex_ECRYPT_encrypt_bytes' that is compiled without support for 'ssse3'
chacha.c:       VEC4_QUARTERROUND( 0, 4, 8,12);
chacha.c:       ^
chacha.c: ./u4.h:17:36: note: expanded from macro 'VEC4_QUARTERROUND'
chacha.c: #define VEC4_QUARTERROUND(a,b,c,d) VEC4_QUARTERROUND_SHUFFLE(a,b,c,d)
chacha.c:                                    ^
chacha.c: ./u4.h:12:86: note: expanded from macro 'VEC4_QUARTERROUND_SHUFFLE'
chacha.c:    x_##a = _mm_add_epi32(x_##a, x_##b); t_##a = _mm_xor_si128(x_##d, x_##a); x_##d = _mm_shuffle_epi8(t_##a, rot16); \
chacha.c:                                                                                      ^
chacha.c: ./u4.h:122:7: error: always_inline function '_mm_shuffle_epi8' requires target feature 'ssse3', but would be inlined into function 'crypto_stream_chacha12_dolbeau_amd64_avx2_constbranchindex_ECRYPT_encrypt_bytes' that is compiled without support for 'ssse3'
chacha.c: ./u4.h:17:36: note: expanded from macro 'VEC4_QUARTERROUND'
chacha.c: #define VEC4_QUARTERROUND(a,b,c,d) VEC4_QUARTERROUND_SHUFFLE(a,b,c,d)
chacha.c:                                    ^
chacha.c: ./u4.h:14:86: note: expanded from macro 'VEC4_QUARTERROUND_SHUFFLE'
chacha.c:    x_##a = _mm_add_epi32(x_##a, x_##b); t_##a = _mm_xor_si128(x_##d, x_##a); x_##d = _mm_shuffle_epi8(t_##a, rot8); \
chacha.c:                                                                                      ^
chacha.c: ./u4.h:123:7: error: always_inline function '_mm_shuffle_epi8' requires target feature 'ssse3', but would be inlined into function 'crypto_stream_chacha12_dolbeau_amd64_avx2_constbranchindex_ECRYPT_encrypt_bytes' that is compiled without support for 'ssse3'
chacha.c:       VEC4_QUARTERROUND( 1, 5, 9,13);
chacha.c:       ^
chacha.c: ./u4.h:17:36: note: expanded from macro 'VEC4_QUARTERROUND'
chacha.c: #define VEC4_QUARTERROUND(a,b,c,d) VEC4_QUARTERROUND_SHUFFLE(a,b,c,d)
chacha.c:                                    ^
chacha.c: ./u4.h:12:86: note: expanded from macro 'VEC4_QUARTERROUND_SHUFFLE'
chacha.c:    x_##a = _mm_add_epi32(x_##a, x_##b); t_##a = _mm_xor_si128(x_##d, x_##a); x_##d = _mm_shuffle_epi8(t_##a, rot16); \
chacha.c: ...

Compiler output

chacha.c: In file included from chacha.c:11:
chacha.c: /usr/lib/llvm-14/lib/clang/14.0.0/include/arm_neon.h:28:2: error: "NEON intrinsics not available with the soft-float ABI. Please use -mfloat-abi=softfp or -mfloat-abi=hard"
chacha.c: #error "NEON intrinsics not available with the soft-float ABI. Please use -mfloat-abi=softfp or -mfloat-abi=hard"
chacha.c:  ^
chacha.c: 1 error generated.

Compiler output

chacha.c: chacha.c:11:10: fatal error: arm_neon.h: No such file or directory
chacha.c:    11 | #include <arm_neon.h>
chacha.c:       |          ^~~~~~~~~~~~
chacha.c: compilation terminated.

Compiler output

chacha.c: In file included from chacha.c:1:
chacha.c: e/chacha.c: In function 'crypto_stream_chacha12_e_merged_constbranchindex_ECRYPT_encrypt_bytes':
chacha.c: e/chacha.c:179:46: warning: 'ctarget' may be used uninitialized in this function [-Wmaybe-uninitialized]
chacha.c:   179 |         for (i = 0;i < bytes;++i) ctarget[i] = c[i];
chacha.c:       |                                   ~~~~~~~~~~~^~~~~~

Compiler output

stream.c: stream.c:126:2: error: -- Implementation supports only microarchitectures with support for Advanced Vector Extensions (AVX2 or AVX512).
stream.c: #error -- Implementation supports only microarchitectures with support for Advanced Vector Extensions (AVX2 or AVX512).
stream.c:  ^
stream.c: 1 error generated.

Compiler output

stream.c: stream.c:126:2: error: #error -- Implementation supports only microarchitectures with support for Advanced Vector Extensions (AVX2 or AVX512).
stream.c:   126 | #error -- Implementation supports only microarchitectures with support for Advanced Vector Extensions (AVX2 or AVX512).
stream.c:       |  ^~~~~

Compiler output

stream.c: stream.c:56:43: error: passing '__m128i *' to parameter of incompatible type '__m128i' (vector of 2 'long long' values); dereference with *
stream.c:     __m256i s0 = _mm_broadcastsi128_si256((__m128i *)sigma);
stream.c:                                           ^~~~~~~~~~~~~~~~
stream.c:                                           *
stream.c: /usr/lib/llvm-14/lib/clang/14.0.0/include/avx2intrin.h:743:65: note: expanded from macro '_mm_broadcastsi128_si256'
stream.c: #define _mm_broadcastsi128_si256(X) _mm256_broadcastsi128_si256(X)
stream.c:                                                                 ^
stream.c: /usr/lib/llvm-14/lib/clang/14.0.0/include/avx2intrin.h:738:37: note: passing argument to parameter '__X' here
stream.c: _mm256_broadcastsi128_si256(__m128i __X)
stream.c:                                     ^
stream.c: 1 error generated.

Compiler output

stream.c: stream.c: In function 'crypto_stream_chacha12_krovetz_avx2_constbranchindex_xor':
stream.c: stream.c:58:13: warning: AVX vector return without AVX enabled changes the ABI [-Wpsabi]
stream.c:    58 |     __m256i s0 = _mm256_broadcastsi128_si256(*(__m128i *)sigma);
stream.c:       |             ^~
stream.c: In file included from /usr/lib/gcc/x86_64-linux-gnu/11/include/immintrin.h:47,
stream.c:                  from stream.c:8:
stream.c: /usr/lib/gcc/x86_64-linux-gnu/11/include/avx2intrin.h:574:1: error: inlining failed in call to 'always_inline' '_mm256_or_si256': target specific option mismatch
stream.c:   574 | _mm256_or_si256 (__m256i __A, __m256i __B)
stream.c:       | ^~~~~~~~~~~~~~~
stream.c: stream.c:63:18: note: called from here
stream.c:    63 |     __m256i s3 = _mm256_or_si256(
stream.c:       |                  ^~~~~~~~~~~~~~~~
stream.c:    64 |         _mm256_slli_si256(_mm256_broadcastq_epi64(*(__m128i *)n), 8),
stream.c:       |         ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
stream.c:    65 |         _mm256_set_epi32(0,0,0,1,0,0,0,0)
stream.c:       |         ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
stream.c:    66 |     );
stream.c:       |     ~
stream.c: In file included from /usr/lib/gcc/x86_64-linux-gnu/11/include/immintrin.h:47,
stream.c:                  from stream.c:8:
stream.c: /usr/lib/gcc/x86_64-linux-gnu/11/include/avx2intrin.h:655:1: error: inlining failed in call to 'always_inline' '_mm256_slli_si256': target specific option mismatch
stream.c:   655 | _mm256_slli_si256 (__m256i __A, const int __N)
stream.c:       | ^~~~~~~~~~~~~~~~~
stream.c: stream.c:63:18: note: called from here
stream.c:    63 |     __m256i s3 = _mm256_or_si256(
stream.c: ...

Compiler output

stream.c: stream.c:133:45: warning: unused variable 'key' [-Wunused-variable]
stream.c:     __attribute__ ((aligned (16))) unsigned key[8], nonce[2];
stream.c:                                             ^
stream.c: stream.c:133:53: warning: unused variable 'nonce' [-Wunused-variable]
stream.c:     __attribute__ ((aligned (16))) unsigned key[8], nonce[2];
stream.c:                                                     ^
stream.c: 2 warnings generated.

Compiler output

stream.c: stream.c: In function 'crypto_stream_chacha12_krovetz_vec128_constbranchindex_xor':
stream.c: stream.c:133:53: warning: unused variable 'nonce' [-Wunused-variable]
stream.c:   133 |     __attribute__ ((aligned (16))) unsigned key[8], nonce[2];
stream.c:       |                                                     ^~~~~
stream.c: stream.c:133:45: warning: unused variable 'key' [-Wunused-variable]
stream.c:   133 |     __attribute__ ((aligned (16))) unsigned key[8], nonce[2];
stream.c:       |                                             ^~~

Passed TIMECOP

TIMECOP iterations: 10