Implementation notes: amd64, wintermute, crypto_stream/chacha8

Computer: wintermute
Architecture: amd64
CPU ID: GenuineIntel-000306c3-bfebfbff
SUPERCOP version: 20140505
Operation: crypto_stream
Primitive: chacha8

Time	Implementation	Compiler	Benchmark date	SUPERCOP version
920	`krovetz/avx2`	`clang -O3 -mavx2`	20140508	20140505
940	`moon/avx2/64`	`clang -O3`	20140508	20140505
940	`moon/avx2/64`	`clang -O3 -mavx`	20140508	20140505
944	`moon/avx2/64`	`cc`	20140508	20140505
944	`moon/avx2/64`	`clang -O3 -mavx2`	20140508	20140505
1676	`moon/avx/64`	`cc`	20140508	20140505
1676	`moon/avx/64`	`clang -O3`	20140508	20140505
1676	`moon/avx/64`	`clang -O3 -mavx`	20140508	20140505
1676	`moon/avx/64`	`clang -O3 -mavx2`	20140508	20140505
1700	`moon/ssse3/64`	`clang -O3`	20140508	20140505
1700	`moon/ssse3/64`	`clang -O3 -mavx`	20140508	20140505
1704	`moon/ssse3/64`	`clang -O3 -mavx2`	20140508	20140505
1708	`moon/ssse3/64`	`cc`	20140508	20140505
1812	`krovetz/vec128`	`clang -O3 -mavx2`	20140508	20140505
1816	`krovetz/vec128`	`clang -O3 -mavx`	20140508	20140505
1924	`krovetz/vec128`	`clang -O3`	20140508	20140505
2040	`moon/sse2/64`	`clang -O3 -mavx`	20140508	20140505
2040	`moon/sse2/64`	`clang -O3 -mavx2`	20140508	20140505
2044	`moon/sse2/64`	`clang -O3`	20140508	20140505
2052	`moon/sse2/64`	`cc`	20140508	20140505
4216	`e/amd64-3`	`clang -O3`	20140508	20140505
4224	`e/amd64-3`	`clang -O3 -mavx`	20140508	20140505
4232	`e/amd64-3`	`clang -O3 -mavx2`	20140508	20140505
4236	`e/amd64-3`	`cc`	20140508	20140505
4876	`e/merged`	`clang -O3 -mavx`	20140508	20140505
4880	`e/merged`	`clang -O3`	20140508	20140505
4880	`e/merged`	`clang -O3 -mavx2`	20140508	20140505
4992	`e/regs`	`clang -O3`	20140508	20140505
5052	`e/regs`	`clang -O3 -mavx2`	20140508	20140505
5060	`e/regs`	`clang -O3 -mavx`	20140508	20140505
5320	`e/ref`	`clang -O3 -mavx2`	20140508	20140505
5848	`e/ref`	`clang -O3 -mavx`	20140508	20140505
6000	`e/ref`	`clang -O3`	20140508	20140505
11624	`krovetz/vec128`	`cc`	20140508	20140505
24772	`e/merged`	`cc`	20140508	20140505
39416	`e/regs`	`cc`	20140508	20140505
45628	`e/ref`	`cc`	20140508	20140505

Test failure

Implementation: crypto_stream/chacha8/e/amd64-xmm6
Compiler: cc

error 111

Number of similar (compiler,implementation) pairs: 24, namely:

Compiler	Implementations
cc	e/amd64-xmm6 moon/avx/32 moon/avx2/32 moon/ssse3/32 moon/xop/32 moon/xop/64
clang -O3	e/amd64-xmm6 moon/avx/32 moon/avx2/32 moon/ssse3/32 moon/xop/32 moon/xop/64
clang -O3 -mavx	e/amd64-xmm6 moon/avx/32 moon/avx2/32 moon/ssse3/32 moon/xop/32 moon/xop/64
clang -O3 -mavx2	e/amd64-xmm6 moon/avx/32 moon/avx2/32 moon/ssse3/32 moon/xop/32 moon/xop/64

Compiler output

Implementation: crypto_stream/chacha8/amd64-ssse3
Compiler: cc

api.c: In file included from api.c:1:
api.c: /home/dch/tmp/supercop-20140505/supercop-data/wintermute/amd64/include/estream-convert-api.h:7:10: fatal error: 'e/ecrypt-sync.h' file not found
api.c: #include "e/ecrypt-sync.h"
api.c: ^
api.c: 1 error generated.

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

Compiler	Implementations
cc	amd64-ssse3
clang -O3	amd64-ssse3
clang -O3 -mavx	amd64-ssse3
clang -O3 -mavx2	amd64-ssse3

Compiler output

Implementation: crypto_stream/chacha8/e/amd64-3
Compiler: cc

chacha.s: cc: warning: argument unused during compilation: '-D SUPERCOP'
chacha.s: cc: warning: argument unused during compilation: '-I .'
chacha.s: cc: warning: argument unused during compilation: '-I /home/dch/tmp/supercop-20140505/supercop-data/wintermute/amd64/include'

Number of similar (compiler,implementation) pairs: 2, namely:

Compiler	Implementations
cc	e/amd64-3 e/amd64-xmm6

Compiler output

Implementation: crypto_stream/chacha8/goll_gueron
Compiler: cc

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.

Number of similar (compiler,implementation) pairs: 3, namely:

Compiler	Implementations
cc	goll_gueron
clang -O3	goll_gueron
clang -O3 -mavx	goll_gueron

Compiler output

Implementation: crypto_stream/chacha8/krovetz/avx2
Compiler: cc

stream.c: stream.c:54:5: error: use of undeclared identifier '__m256i'
stream.c: __m256i v0,v1,v2,v3,v4,v5,v6,v7,v8,v9,v10,v11;
stream.c: ^
stream.c: stream.c:56:5: error: use of undeclared identifier '__m256i'
stream.c: __m256i s0 = _mm_broadcastsi128_si256((__m128i *)sigma);
stream.c: ^
stream.c: stream.c:60:5: error: use of undeclared identifier '__m256i'
stream.c: __m256i s1 = _mm256_loadu_si256((__m256i *)k);
stream.c: ^
stream.c: stream.c:61:5: error: use of undeclared identifier '__m256i'
stream.c: __m256i s2 = _mm256_permute2x128_si256(s1,s1,0x11);
stream.c: ^
stream.c: stream.c:62:5: error: use of undeclared identifier 's1'
stream.c: s1 = _mm256_permute2x128_si256(s1,s1,0x00);
stream.c: ^
stream.c: stream.c:62:10: warning: implicit declaration of function '_mm256_permute2x128_si256' is invalid in C99 [-Wimplicit-function-declaration]
stream.c: s1 = _mm256_permute2x128_si256(s1,s1,0x00);
stream.c: ^
stream.c: stream.c:62:36: error: use of undeclared identifier 's1'
stream.c: s1 = _mm256_permute2x128_si256(s1,s1,0x00);
stream.c: ^
stream.c: stream.c:63:5: error: use of undeclared identifier '__m256i'
stream.c: __m256i s3 = _mm256_or_si256(
stream.c: ^
stream.c: stream.c:68:9: error: use of undeclared identifier 'v8'
stream.c: ...

Number of similar (compiler,implementation) pairs: 2, namely:

Compiler	Implementations
cc	krovetz/avx2
clang -O3	krovetz/avx2

Compiler output

Implementation: crypto_stream/chacha8/e/amd64-3
Compiler: clang -O3

chacha.s: clang: warning: argument unused during compilation: '-D SUPERCOP'
chacha.s: clang: warning: argument unused during compilation: '-I .'
chacha.s: clang: warning: argument unused during compilation: '-I /home/dch/tmp/supercop-20140505/supercop-data/wintermute/amd64/include'

Number of similar (compiler,implementation) pairs: 2, namely:

Compiler	Implementations
clang -O3	e/amd64-3 e/amd64-xmm6

Compiler output

Implementation: crypto_stream/chacha8/e/amd64-3
Compiler: clang -O3 -mavx

chacha.s: clang: warning: argument unused during compilation: '-mavx'
chacha.s: clang: warning: argument unused during compilation: '-D SUPERCOP'
chacha.s: clang: warning: argument unused during compilation: '-I .'
chacha.s: clang: warning: argument unused during compilation: '-I /home/dch/tmp/supercop-20140505/supercop-data/wintermute/amd64/include'

Number of similar (compiler,implementation) pairs: 2, namely:

Compiler	Implementations
clang -O3 -mavx	e/amd64-3 e/amd64-xmm6

Compiler output

Implementation: crypto_stream/chacha8/krovetz/avx2
Compiler: clang -O3 -mavx

stream.c: stream.c:56:18: warning: implicit declaration of function '_mm_broadcastsi128_si256' is invalid in C99 [-Wimplicit-function-declaration]
stream.c: __m256i s0 = _mm_broadcastsi128_si256((__m128i *)sigma);
stream.c: ^
stream.c: stream.c:56:13: error: initializing '__m256i' with an expression of incompatible type 'int'
stream.c: __m256i s0 = _mm_broadcastsi128_si256((__m128i *)sigma);
stream.c: ^ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
stream.c: stream.c:61:18: warning: implicit declaration of function '_mm256_permute2x128_si256' is invalid in C99 [-Wimplicit-function-declaration]
stream.c: __m256i s2 = _mm256_permute2x128_si256(s1,s1,0x11);
stream.c: ^
stream.c: stream.c:61:13: error: initializing '__m256i' with an expression of incompatible type 'int'
stream.c: __m256i s2 = _mm256_permute2x128_si256(s1,s1,0x11);
stream.c: ^ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
stream.c: stream.c:62:8: error: assigning to '__m256i' from incompatible type 'int'
stream.c: s1 = _mm256_permute2x128_si256(s1,s1,0x00);
stream.c: ^ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
stream.c: stream.c:63:18: warning: implicit declaration of function '_mm256_or_si256' is invalid in C99 [-Wimplicit-function-declaration]
stream.c: __m256i s3 = _mm256_or_si256(
stream.c: ^
stream.c: stream.c:64:9: warning: implicit declaration of function '_mm256_slli_si256' is invalid in C99 [-Wimplicit-function-declaration]
stream.c: _mm256_slli_si256(_mm256_broadcastq_epi64(*(__m128i *)n), 8),
stream.c: ^
stream.c: stream.c:64:27: warning: implicit declaration of function '_mm256_broadcastq_epi64' is invalid in C99 [-Wimplicit-function-declaration]
stream.c: _mm256_slli_si256(_mm256_broadcastq_epi64(*(__m128i *)n), 8),
stream.c: ^
stream.c: stream.c:63:13: error: initializing '__m256i' with an expression of incompatible type 'int'
stream.c: ...

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

Compiler	Implementations
clang -O3 -mavx	krovetz/avx2

Compiler output

Implementation: crypto_stream/chacha8/e/amd64-3
Compiler: clang -O3 -mavx2

chacha.s: clang: warning: argument unused during compilation: '-mavx2'
chacha.s: clang: warning: argument unused during compilation: '-D SUPERCOP'
chacha.s: clang: warning: argument unused during compilation: '-I .'
chacha.s: clang: warning: argument unused during compilation: '-I /home/dch/tmp/supercop-20140505/supercop-data/wintermute/amd64/include'

Number of similar (compiler,implementation) pairs: 2, namely:

Compiler	Implementations
clang -O3 -mavx2	e/amd64-3 e/amd64-xmm6

Compiler output

Implementation: crypto_stream/chacha8/goll_gueron
Compiler: clang -O3 -mavx2

stream.c: stream.c:302:15: warning: implicit declaration of function '_mm256_broadcastsi128_si256' is invalid in C99 [-Wimplicit-function-declaration]
stream.c: d0 = (vec256)_mm256_broadcastsi128_si256(*(__m128i*)chacha_const);
stream.c: ^
stream.c: stream.c:302:7: error: invalid conversion between vector type 'vec256' and integer type 'int' of different size
stream.c: d0 = (vec256)_mm256_broadcastsi128_si256(*(__m128i*)chacha_const);
stream.c: ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
stream.c: stream.c:303:7: error: invalid conversion between vector type 'vec256' and integer type 'int' of different size
stream.c: d1 = (vec256)_mm256_broadcastsi128_si256(((__m128i*)key)[0]);
stream.c: ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
stream.c: stream.c:304:7: error: invalid conversion between vector type 'vec256' and integer type 'int' of different size
stream.c: d2 = (vec256)_mm256_broadcastsi128_si256(((__m128i*)key)[1]);
stream.c: ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
stream.c: stream.c:305:7: error: invalid conversion between vector type '__m256i' and integer type 'int' of different size
stream.c: d3 = ADD256_64(_mm256_broadcastsi128_si256((__m128i)s3), _mm256_set_epi64x(0,1,0,0));
stream.c: ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
stream.c: stream.c:38:49: note: expanded from macro 'ADD256_64'
stream.c: #define ADD256_64(a,b) (vec256)_mm256_add_epi64((__m256i)a, (__m256i)b)
stream.c: ^~~~~~~~~~
stream.c: 1 warning and 4 errors generated.

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

Compiler	Implementations
clang -O3 -mavx2	goll_gueron