Implementation notes: x86, thoth, crypto_hash/cubehash512

Computer: thoth
Architecture: x86
CPU ID: AuthenticAMD-00000622-0183f9ff
SUPERCOP version: 20160806
Operation: crypto_hash
Primitive: cubehash512

Time	Implementation	Compiler	Benchmark date	SUPERCOP version
72903	`unrolled5`	`clang -march=native -O3 -fomit-frame-pointer -fwrapv -Qunused-arguments`	20160726	20160724
73583	`unrolled3`	`clang -march=native -O3 -fomit-frame-pointer -fwrapv -Qunused-arguments`	20160726	20160724
73744	`unrolled3`	`gcc -march=native -mtune=native -O3 -fomit-frame-pointer -fwrapv`	20160726	20160724
74968	`unrolled5`	`gcc -march=native -mtune=native -O2 -fomit-frame-pointer -fwrapv`	20160726	20160724
75007	`unrolled4`	`clang -march=native -O3 -fomit-frame-pointer -fwrapv -Qunused-arguments`	20160726	20160724
76105	`unrolled3`	`gcc -march=native -mtune=native -O2 -fomit-frame-pointer -fwrapv`	20160726	20160724
77482	`unrolled5`	`gcc -march=native -mtune=native -O3 -fomit-frame-pointer -fwrapv`	20160726	20160724
78319	`unrolled3`	`gcc -march=native -mtune=native -Os -fomit-frame-pointer -fwrapv`	20160726	20160724
78831	`unrolled4`	`gcc -march=native -mtune=native -O3 -fomit-frame-pointer -fwrapv`	20160726	20160724
79600	`unrolled5`	`gcc -march=native -mtune=native -Os -fomit-frame-pointer -fwrapv`	20160726	20160724
79678	`unrolled4`	`gcc -march=native -mtune=native -O2 -fomit-frame-pointer -fwrapv`	20160726	20160724
81680	`unrolled3`	`gcc -march=native -mtune=native -O -fomit-frame-pointer -fwrapv`	20160726	20160724
84717	`unrolled5`	`gcc -march=native -mtune=native -O -fomit-frame-pointer -fwrapv`	20160726	20160724
85933	`unrolled4`	`gcc -march=native -mtune=native -O -fomit-frame-pointer -fwrapv`	20160726	20160724
92794	`unrolled4`	`gcc -march=native -mtune=native -Os -fomit-frame-pointer -fwrapv`	20160726	20160724
102853	`x86`	`clang -march=native -O3 -fomit-frame-pointer -fwrapv -Qunused-arguments`	20160726	20160724
102853	`x86`	`gcc -march=native -mtune=native -O3 -fomit-frame-pointer -fwrapv`	20160726	20160724
103895	`x86`	`gcc -march=native -mtune=native -O2 -fomit-frame-pointer -fwrapv`	20160726	20160724
103906	`x86`	`gcc -march=native -mtune=native -O -fomit-frame-pointer -fwrapv`	20160726	20160724
103919	`x86`	`gcc -march=native -mtune=native -Os -fomit-frame-pointer -fwrapv`	20160726	20160724
115883	`unrolled`	`gcc -march=native -mtune=native -O2 -fomit-frame-pointer -fwrapv`	20160726	20160724
121821	`unrolled`	`gcc -march=native -mtune=native -O -fomit-frame-pointer -fwrapv`	20160726	20160724
123801	`unrolled`	`gcc -march=native -mtune=native -O3 -fomit-frame-pointer -fwrapv`	20160726	20160724
128183	`simple`	`gcc -march=native -mtune=native -O3 -fomit-frame-pointer -fwrapv`	20160726	20160724
134605	`unrolled2`	`clang -march=native -O3 -fomit-frame-pointer -fwrapv -Qunused-arguments`	20160726	20160724
138053	`unrolled2`	`gcc -march=native -mtune=native -O -fomit-frame-pointer -fwrapv`	20160726	20160724
138675	`unrolled2`	`gcc -march=native -mtune=native -O2 -fomit-frame-pointer -fwrapv`	20160726	20160724
146846	`unrolled2`	`gcc -march=native -mtune=native -O3 -fomit-frame-pointer -fwrapv`	20160726	20160724
147480	`unrolled`	`clang -march=native -O3 -fomit-frame-pointer -fwrapv -Qunused-arguments`	20160726	20160724
269642	`simple`	`clang -march=native -O3 -fomit-frame-pointer -fwrapv -Qunused-arguments`	20160726	20160724
305279	`unrolled`	`gcc -march=native -mtune=native -Os -fomit-frame-pointer -fwrapv`	20160726	20160724
330332	`unrolled2`	`gcc -march=native -mtune=native -Os -fomit-frame-pointer -fwrapv`	20160726	20160724
618228	`simple`	`gcc -march=native -mtune=native -O -fomit-frame-pointer -fwrapv`	20160726	20160724
629850	`simple`	`gcc -march=native -mtune=native -O2 -fomit-frame-pointer -fwrapv`	20160726	20160724
812015	`simple`	`gcc -march=native -mtune=native -Os -fomit-frame-pointer -fwrapv`	20160726	20160724

Test failure

Implementation: crypto_hash/cubehash512/x86xmm
Compiler: clang -march=native -O3 -fomit-frame-pointer -fwrapv -Qunused-arguments

error 111
crypto_hash does not handle m=h overlap

Number of similar (compiler,implementation) pairs: 5, namely:

Compiler	Implementations
clang -march=native -O3 -fomit-frame-pointer -fwrapv -Qunused-arguments	x86xmm
gcc -march=native -mtune=native -O2 -fomit-frame-pointer -fwrapv	x86xmm
gcc -march=native -mtune=native -O3 -fomit-frame-pointer -fwrapv	x86xmm
gcc -march=native -mtune=native -O -fomit-frame-pointer -fwrapv	x86xmm
gcc -march=native -mtune=native -Os -fomit-frame-pointer -fwrapv	x86xmm

Compiler output

Implementation: crypto_hash/cubehash512/emmintrin4
Compiler: clang -march=native -O3 -fomit-frame-pointer -fwrapv -Qunused-arguments

cubehash.c: cubehash.c:101:19: error: always_inline function '_mm_set_epi32' requires target feature 'sse2', but would be inlined into function 'Init' that is compiled without support for 'sse2'
cubehash.c: state->gt;x[0] = _mm_set_epi32(0xc8237df7,0x992b7520,0x18f45926,0x781f814f);
cubehash.c: ^
cubehash.c: cubehash.c:102:19: error: always_inline function '_mm_set_epi32' requires target feature 'sse2', but would be inlined into function 'Init' that is compiled without support for 'sse2'
cubehash.c: state->gt;x[1] = _mm_set_epi32(0x947c6147,0x51916982,0x7b0075ff,0xe4e3ba88);
cubehash.c: ^
cubehash.c: cubehash.c:103:19: error: always_inline function '_mm_set_epi32' requires target feature 'sse2', but would be inlined into function 'Init' that is compiled without support for 'sse2'
cubehash.c: state->gt;x[2] = _mm_set_epi32(0x912e1aca,0xb6e17224,0x4d197eb5,0x9dc06f0a);
cubehash.c: ^
cubehash.c: cubehash.c:104:19: error: always_inline function '_mm_set_epi32' requires target feature 'sse2', but would be inlined into function 'Init' that is compiled without support for 'sse2'
cubehash.c: state->gt;x[3] = _mm_set_epi32(0x20d4074f,0xf0fcf7c8,0xd9efd0ec,0x5270f5e1);
cubehash.c: ^
cubehash.c: cubehash.c:105:19: error: always_inline function '_mm_set_epi32' requires target feature 'sse2', but would be inlined into function 'Init' that is compiled without support for 'sse2'
cubehash.c: state->gt;x[4] = _mm_set_epi32(0xaf1c332a,0x017c189c,0xf4839313,0x15547fee);
cubehash.c: ^
cubehash.c: cubehash.c:106:19: error: always_inline function '_mm_set_epi32' requires target feature 'sse2', but would be inlined into function 'Init' that is compiled without support for 'sse2'
cubehash.c: state->gt;x[5] = _mm_set_epi32(0xb6d3d055,0x5bd87a43,0x84997eec,0xde4d7c8c);
cubehash.c: ^
cubehash.c: cubehash.c:107:19: error: always_inline function '_mm_set_epi32' requires target feature 'sse2', but would be inlined into function 'Init' that is compiled without support for 'sse2'
cubehash.c: state->gt;x[6] = _mm_set_epi32(0x4bf12b94,0xd9d6ca35,0x2b8cb0a6,0x3ae247b0);
cubehash.c: ^
cubehash.c: cubehash.c:108:19: error: always_inline function '_mm_set_epi32' requires target feature 'sse2', but would be inlined into function 'Init' that is compiled without support for 'sse2'
cubehash.c: state->gt;x[7] = _mm_set_epi32(0x520c709b,0x7e70e613,0x62fb84ad,0x97f33a51);
cubehash.c: ^
cubehash.c: cubehash.c:110:19: error: always_inline function '_mm_set_epi32' requires target feature 'sse2', but would be inlined into function 'Init' that is compiled without support for 'sse2'
cubehash.c: ...

Number of similar (compiler,implementation) pairs: 1, namely:

Compiler	Implementations
clang -march=native -O3 -fomit-frame-pointer -fwrapv -Qunused-arguments	emmintrin4

Compiler output

Implementation: crypto_hash/cubehash512/mmintrin
Compiler: clang -march=native -O3 -fomit-frame-pointer -fwrapv -Qunused-arguments

hash.c: hash.c:148:13: error: '__builtin_ia32_pshufw' needs target feature sse
hash.c: x10 = _mm_shuffle_pi16(x10,0x4e);
hash.c: ^
hash.c: /usr/lib/llvm-3.8/bin/../lib/clang/3.8.0/include/xmmintrin.h:757:10: note: expanded from macro '_mm_shuffle_pi16'
hash.c: (__m64)__builtin_ia32_pshufw((__v4hi)(__m64)(a), (n)); })
hash.c: ^
hash.c: hash.c:149:13: error: '__builtin_ia32_pshufw' needs target feature sse
hash.c: x12 = _mm_shuffle_pi16(x12,0x4e);
hash.c: ^
hash.c: /usr/lib/llvm-3.8/bin/../lib/clang/3.8.0/include/xmmintrin.h:757:10: note: expanded from macro '_mm_shuffle_pi16'
hash.c: (__m64)__builtin_ia32_pshufw((__v4hi)(__m64)(a), (n)); })
hash.c: ^
hash.c: hash.c:150:13: error: '__builtin_ia32_pshufw' needs target feature sse
hash.c: x14 = _mm_shuffle_pi16(x14,0x4e);
hash.c: ^
hash.c: /usr/lib/llvm-3.8/bin/../lib/clang/3.8.0/include/xmmintrin.h:757:10: note: expanded from macro '_mm_shuffle_pi16'
hash.c: (__m64)__builtin_ia32_pshufw((__v4hi)(__m64)(a), (n)); })
hash.c: ^
hash.c: hash.c:151:13: error: '__builtin_ia32_pshufw' needs target feature sse
hash.c: x16 = _mm_shuffle_pi16(x16,0x4e);
hash.c: ^
hash.c: /usr/lib/llvm-3.8/bin/../lib/clang/3.8.0/include/xmmintrin.h:757:10: note: expanded from macro '_mm_shuffle_pi16'
hash.c: (__m64)__builtin_ia32_pshufw((__v4hi)(__m64)(a), (n)); })
hash.c: ^
hash.c: hash.c:152:13: error: '__builtin_ia32_pshufw' needs target feature sse
hash.c: ...

Number of similar (compiler,implementation) pairs: 1, namely:

Compiler	Implementations
clang -march=native -O3 -fomit-frame-pointer -fwrapv -Qunused-arguments	mmintrin

Compiler output

Implementation: crypto_hash/cubehash512/emmintrin5
Compiler: clang -march=native -O3 -fomit-frame-pointer -fwrapv -Qunused-arguments

hash.c: hash.c:23:8: error: always_inline function '_mm_set_epi32' requires target feature 'sse2', but would be inlined into function 'crypto_hash_cubehash512_emmintrin5' that is compiled without support for 'sse2'
hash.c: x0 = _mm_set_epi32(0xc8237df7,0x992b7520,0x18f45926,0x781f814f);
hash.c: ^
hash.c: hash.c:24:8: error: always_inline function '_mm_set_epi32' requires target feature 'sse2', but would be inlined into function 'crypto_hash_cubehash512_emmintrin5' that is compiled without support for 'sse2'
hash.c: x1 = _mm_set_epi32(0x947c6147,0x51916982,0x7b0075ff,0xe4e3ba88);
hash.c: ^
hash.c: hash.c:25:8: error: always_inline function '_mm_set_epi32' requires target feature 'sse2', but would be inlined into function 'crypto_hash_cubehash512_emmintrin5' that is compiled without support for 'sse2'
hash.c: x2 = _mm_set_epi32(0x912e1aca,0xb6e17224,0x4d197eb5,0x9dc06f0a);
hash.c: ^
hash.c: hash.c:26:8: error: always_inline function '_mm_set_epi32' requires target feature 'sse2', but would be inlined into function 'crypto_hash_cubehash512_emmintrin5' that is compiled without support for 'sse2'
hash.c: x3 = _mm_set_epi32(0x20d4074f,0xf0fcf7c8,0xd9efd0ec,0x5270f5e1);
hash.c: ^
hash.c: hash.c:27:8: error: always_inline function '_mm_set_epi32' requires target feature 'sse2', but would be inlined into function 'crypto_hash_cubehash512_emmintrin5' that is compiled without support for 'sse2'
hash.c: x4 = _mm_set_epi32(0xaf1c332a,0x017c189c,0xf4839313,0x15547fee);
hash.c: ^
hash.c: hash.c:28:8: error: always_inline function '_mm_set_epi32' requires target feature 'sse2', but would be inlined into function 'crypto_hash_cubehash512_emmintrin5' that is compiled without support for 'sse2'
hash.c: x5 = _mm_set_epi32(0xb6d3d055,0x5bd87a43,0x84997eec,0xde4d7c8c);
hash.c: ^
hash.c: hash.c:29:8: error: always_inline function '_mm_set_epi32' requires target feature 'sse2', but would be inlined into function 'crypto_hash_cubehash512_emmintrin5' that is compiled without support for 'sse2'
hash.c: x6 = _mm_set_epi32(0x4bf12b94,0xd9d6ca35,0x2b8cb0a6,0x3ae247b0);
hash.c: ^
hash.c: hash.c:30:8: error: always_inline function '_mm_set_epi32' requires target feature 'sse2', but would be inlined into function 'crypto_hash_cubehash512_emmintrin5' that is compiled without support for 'sse2'
hash.c: x7 = _mm_set_epi32(0x520c709b,0x7e70e613,0x62fb84ad,0x97f33a51);
hash.c: ^
hash.c: hash.c:36:10: error: always_inline function '_mm_xor_si128' requires target feature 'sse2', but would be inlined into function 'crypto_hash_cubehash512_emmintrin5' that is compiled without support for 'sse2'
hash.c: ...

Number of similar (compiler,implementation) pairs: 1, namely:

Compiler	Implementations
clang -march=native -O3 -fomit-frame-pointer -fwrapv -Qunused-arguments	emmintrin5

Compiler output

Implementation: crypto_hash/cubehash512/emmintrin4
Compiler: gcc -march=native -mtune=native -O2 -fomit-frame-pointer -fwrapv

cubehash.c: cubehash.c: In function 'Init':
cubehash.c: cubehash.c:101:17: warning: SSE vector return without SSE enabled changes the ABI [-Wpsabi]
cubehash.c: state->gt;x[0] = _mm_set_epi32(0xc8237df7,0x992b7520,0x18f45926,0x781f814f);
cubehash.c: ^
cubehash.c: In file included from cubehash.h:14:0,
cubehash.c: from cubehash.c:9:
cubehash.c: cubehash.c: In function 'transform':
cubehash.c: /usr/lib/gcc/i686-linux-gnu/5/include/emmintrin.h:1415:1: error: inlining failed in call to always_inline '_mm_shuffle_epi32': target specific option mismatch
cubehash.c: _mm_shuffle_epi32 (__m128i __A, const int __mask)
cubehash.c: ^
cubehash.c: cubehash.c:76:8: error: called from here
cubehash.c: x7 = _mm_shuffle_epi32(x7,0xb1);
cubehash.c: ^
cubehash.c: In file included from cubehash.h:14:0,
cubehash.c: from cubehash.c:9:
cubehash.c: /usr/lib/gcc/i686-linux-gnu/5/include/emmintrin.h:1415:1: error: inlining failed in call to always_inline '_mm_shuffle_epi32': target specific option mismatch
cubehash.c: _mm_shuffle_epi32 (__m128i __A, const int __mask)
cubehash.c: ^
cubehash.c: cubehash.c:75:8: error: called from here
cubehash.c: x6 = _mm_shuffle_epi32(x6,0xb1);
cubehash.c: ^
cubehash.c: In file included from cubehash.h:14:0,
cubehash.c: from cubehash.c:9:
cubehash.c: /usr/lib/gcc/i686-linux-gnu/5/include/emmintrin.h:1415:1: error: inlining failed in call to always_inline '_mm_shuffle_epi32': target specific option mismatch
cubehash.c: _mm_shuffle_epi32 (__m128i __A, const int __mask)
cubehash.c: ...

Number of similar (compiler,implementation) pairs: 4, namely:

Compiler	Implementations
gcc -march=native -mtune=native -O2 -fomit-frame-pointer -fwrapv	emmintrin4
gcc -march=native -mtune=native -O3 -fomit-frame-pointer -fwrapv	emmintrin4
gcc -march=native -mtune=native -O -fomit-frame-pointer -fwrapv	emmintrin4
gcc -march=native -mtune=native -Os -fomit-frame-pointer -fwrapv	emmintrin4

Compiler output

Implementation: crypto_hash/cubehash512/mmintrin
Compiler: gcc -march=native -mtune=native -O2 -fomit-frame-pointer -fwrapv

hash.c: In file included from hash.c:3:0:
hash.c: hash.c: In function 'crypto_hash_cubehash512_mmintrin':
hash.c: /usr/lib/gcc/i686-linux-gnu/5/include/xmmintrin.h:1139:1: error: inlining failed in call to always_inline '_mm_shuffle_pi16': target specific option mismatch
hash.c: _mm_shuffle_pi16 (__m64 __A, int const __N)
hash.c: ^
hash.c: hash.c:155:11: error: called from here
hash.c: x1e = _mm_shuffle_pi16(x1e,0x4e);
hash.c: ^
hash.c: In file included from hash.c:3:0:
hash.c: /usr/lib/gcc/i686-linux-gnu/5/include/xmmintrin.h:1139:1: error: inlining failed in call to always_inline '_mm_shuffle_pi16': target specific option mismatch
hash.c: _mm_shuffle_pi16 (__m64 __A, int const __N)
hash.c: ^
hash.c: hash.c:154:11: error: called from here
hash.c: x1c = _mm_shuffle_pi16(x1c,0x4e);
hash.c: ^
hash.c: In file included from hash.c:3:0:
hash.c: /usr/lib/gcc/i686-linux-gnu/5/include/xmmintrin.h:1139:1: error: inlining failed in call to always_inline '_mm_shuffle_pi16': target specific option mismatch
hash.c: _mm_shuffle_pi16 (__m64 __A, int const __N)
hash.c: ^
hash.c: hash.c:153:11: error: called from here
hash.c: x1a = _mm_shuffle_pi16(x1a,0x4e);
hash.c: ^
hash.c: In file included from hash.c:3:0:
hash.c: /usr/lib/gcc/i686-linux-gnu/5/include/xmmintrin.h:1139:1: error: inlining failed in call to always_inline '_mm_shuffle_pi16': target specific option mismatch
hash.c: _mm_shuffle_pi16 (__m64 __A, int const __N)
hash.c: ...

Number of similar (compiler,implementation) pairs: 4, namely:

Compiler	Implementations
gcc -march=native -mtune=native -O2 -fomit-frame-pointer -fwrapv	mmintrin
gcc -march=native -mtune=native -O3 -fomit-frame-pointer -fwrapv	mmintrin
gcc -march=native -mtune=native -O -fomit-frame-pointer -fwrapv	mmintrin
gcc -march=native -mtune=native -Os -fomit-frame-pointer -fwrapv	mmintrin

Compiler output

Implementation: crypto_hash/cubehash512/emmintrin5
Compiler: gcc -march=native -mtune=native -O2 -fomit-frame-pointer -fwrapv

hash.c: hash.c: In function 'crypto_hash_cubehash512_emmintrin5':
hash.c: hash.c:23:6: warning: SSE vector return without SSE enabled changes the ABI [-Wpsabi]
hash.c: x0 = _mm_set_epi32(0xc8237df7,0x992b7520,0x18f45926,0x781f814f);
hash.c: ^
hash.c: In file included from hash.c:2:0:
hash.c: /usr/lib/gcc/i686-linux-gnu/5/include/emmintrin.h:597:1: error: inlining failed in call to always_inline '_mm_set_epi32': target specific option mismatch
hash.c: _mm_set_epi32 (int __q3, int __q2, int __q1, int __q0)
hash.c: ^
hash.c: hash.c:30:6: error: called from here
hash.c: x7 = _mm_set_epi32(0x520c709b,0x7e70e613,0x62fb84ad,0x97f33a51);
hash.c: ^
hash.c: In file included from hash.c:2:0:
hash.c: /usr/lib/gcc/i686-linux-gnu/5/include/emmintrin.h:597:1: error: inlining failed in call to always_inline '_mm_set_epi32': target specific option mismatch
hash.c: _mm_set_epi32 (int __q3, int __q2, int __q1, int __q0)
hash.c: ^
hash.c: hash.c:29:6: error: called from here
hash.c: x6 = _mm_set_epi32(0x4bf12b94,0xd9d6ca35,0x2b8cb0a6,0x3ae247b0);
hash.c: ^
hash.c: In file included from hash.c:2:0:
hash.c: /usr/lib/gcc/i686-linux-gnu/5/include/emmintrin.h:597:1: error: inlining failed in call to always_inline '_mm_set_epi32': target specific option mismatch
hash.c: _mm_set_epi32 (int __q3, int __q2, int __q1, int __q0)
hash.c: ^
hash.c: hash.c:28:6: error: called from here
hash.c: x5 = _mm_set_epi32(0xb6d3d055,0x5bd87a43,0x84997eec,0xde4d7c8c);
hash.c: ^
hash.c: ...

Number of similar (compiler,implementation) pairs: 4, namely:

Compiler	Implementations
gcc -march=native -mtune=native -O2 -fomit-frame-pointer -fwrapv	emmintrin5
gcc -march=native -mtune=native -O3 -fomit-frame-pointer -fwrapv	emmintrin5
gcc -march=native -mtune=native -O -fomit-frame-pointer -fwrapv	emmintrin5
gcc -march=native -mtune=native -Os -fomit-frame-pointer -fwrapv	emmintrin5