Implementation notes: amd64, hunsnivy, crypto_sign/rainbow6b

Computer: hunsnivy
Microarchitecture: amd64; Ivy Bridge+AES (306a9)
Architecture: amd64
CPU ID: GenuineIntel-000306a9-bfebfbff
SUPERCOP version: 20240107
Operation: crypto_sign
Primitive: rainbow6b

Time	Object size	Test size	Implementation	Compiler	Benchmark date	SUPERCOP version
20599494	82187 0 1049652	101620 916 1051384	`T:ref`	`clang_-march=native_-O2_-fomit-frame-pointer_-fwrapv_-Qunused-arguments_-fPIC_-fPIE`	20231222	20231212
20620940	82187 0 1049652	101620 916 1051384	`T:portable`	`clang_-march=native_-O2_-fomit-frame-pointer_-fwrapv_-Qunused-arguments_-fPIC_-fPIE`	20231222	20231212
20641270	110328 0 1049652	131428 916 1051384	`T:portable`	`clang_-march=native_-O3_-fomit-frame-pointer_-fwrapv_-Qunused-arguments_-fPIC_-fPIE`	20231222	20231212
20675698	110328 0 1049652	131396 916 1051384	`T:ref`	`clang_-march=native_-O3_-fomit-frame-pointer_-fwrapv_-Qunused-arguments_-fPIC_-fPIE`	20231222	20231212

Test failure

Implementation: T:portable
Security model: timingleaks
Compiler: clang -march=native -O -fomit-frame-pointer -fwrapv -Qunused-arguments -fPIC -fPIE

error 142
Alarm clock

Number of similar (compiler,implementation) pairs: 14, namely:

Compiler	Implementations
clang -march=native -O -fomit-frame-pointer -fwrapv -Qunused-arguments -fPIC -fPIE	T:portable
clang -march=native -Os -fomit-frame-pointer -fwrapv -Qunused-arguments -fPIC -fPIE	T:portable
clang -mcpu=native -O3 -fomit-frame-pointer -fwrapv -Qunused-arguments -fPIC -fPIE	T:portable
gcc -march=native -mtune=native -O2 -fomit-frame-pointer -fwrapv -fPIC -fPIE	T:portable
gcc -march=native -mtune=native -O3 -fomit-frame-pointer -fwrapv -fPIC -fPIE	T:portable
gcc -march=native -mtune=native -O -fomit-frame-pointer -fwrapv -fPIC -fPIE	T:portable
gcc -march=native -mtune=native -Os -fomit-frame-pointer -fwrapv -fPIC -fPIE	T:portable
clang -march=native -O -fomit-frame-pointer -fwrapv -Qunused-arguments -fPIC -fPIE	T:ref
clang -march=native -Os -fomit-frame-pointer -fwrapv -Qunused-arguments -fPIC -fPIE	T:ref
clang -mcpu=native -O3 -fomit-frame-pointer -fwrapv -Qunused-arguments -fPIC -fPIE	T:ref
gcc -march=native -mtune=native -O2 -fomit-frame-pointer -fwrapv -fPIC -fPIE	T:ref
gcc -march=native -mtune=native -O3 -fomit-frame-pointer -fwrapv -fPIC -fPIE	T:ref
gcc -march=native -mtune=native -O -fomit-frame-pointer -fwrapv -fPIC -fPIE	T:ref
gcc -march=native -mtune=native -Os -fomit-frame-pointer -fwrapv -fPIC -fPIE	T:ref

Compiler output

Implementation: T:avx2
Security model: timingleaks
Compiler: clang -march=native -O2 -fomit-frame-pointer -fwrapv -Qunused-arguments -fPIC -fPIE

rainbow.c: In file included from rainbow.c:4:
rainbow.c: In file included from ./rainbow.h:7:
rainbow.c: In file included from ./blas.h:17:
rainbow.c: ./blas_avx2.h:274:36: error: always_inline function '_mm256_maddubs_epi16' requires target feature 'avx2', but would be inlined into function 'gf31mat_prod_avx2' that is compiled without support for 'avx2'
rainbow.c: r[j] = _mm256_add_epi16( r[j] , _mm256_maddubs_epi16( tmp , bi ) );
rainbow.c: ^
rainbow.c: ./blas_avx2.h:274:11: error: always_inline function '_mm256_add_epi16' requires target feature 'avx2', but would be inlined into function 'gf31mat_prod_avx2' that is compiled without support for 'avx2'
rainbow.c: r[j] = _mm256_add_epi16( r[j] , _mm256_maddubs_epi16( tmp , bi ) );
rainbow.c: ^
rainbow.c: ./blas_avx2.h:287:10: error: always_inline function '_mm256_packs_epi16' requires target feature 'avx2', but would be inlined into function 'gf31mat_prod_avx2' that is compiled without support for 'avx2'
rainbow.c: r[j] = _mm256_packs_epi16( r[j*2] , r[j*2+1] );
rainbow.c: ^
rainbow.c: ./blas_avx2.h:288:10: error: '__builtin_ia32_permdi256' needs target feature avx2
rainbow.c: r[j] = _mm256_permute4x64_epi64( r[j] , 0xd8 ); // 3,1,2,0
rainbow.c: ^
rainbow.c: /usr/lib/llvm-11/lib/clang/11.0.1/include/avx2intrin.h:818:12: note: expanded from macro '_mm256_permute4x64_epi64'
rainbow.c: (__m256i)__builtin_ia32_permdi256((__v4di)(__m256i)(V), (int)(M))
rainbow.c: ^
rainbow.c: In file included from rainbow.c:4:
rainbow.c: In file included from ./rainbow.h:7:
rainbow.c: In file included from ./blas.h:17:
rainbow.c: ./blas_avx2.h:112:36: error: always_inline function '_mm256_srli_epi16' requires target feature 'avx2', but would be inlined into function '_gf31v_reduce_u16_avx2' that is compiled without support for 'avx2'
rainbow.c: return _mm256_add_epi16( a&mask , _mm256_srli_epi16(a,5) );
rainbow.c: ^
rainbow.c: ./blas_avx2.h:112:9: error: always_inline function '_mm256_add_epi16' requires target feature 'avx2', but would be inlined into function '_gf31v_reduce_u16_avx2' that is compiled without support for 'avx2'
rainbow.c: ...

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

Compiler	Implementations
clang -march=native -O2 -fomit-frame-pointer -fwrapv -Qunused-arguments -fPIC -fPIE	T:avx2
clang -march=native -O3 -fomit-frame-pointer -fwrapv -Qunused-arguments -fPIC -fPIE	T:avx2
clang -march=native -O -fomit-frame-pointer -fwrapv -Qunused-arguments -fPIC -fPIE	T:avx2
clang -march=native -Os -fomit-frame-pointer -fwrapv -Qunused-arguments -fPIC -fPIE	T:avx2

Compiler output

Implementation: T:avx2
Security model: timingleaks
Compiler: clang -mcpu=native -O3 -fomit-frame-pointer -fwrapv -Qunused-arguments -fPIC -fPIE

rainbow.c: In file included from rainbow.c:4:
rainbow.c: In file included from ./rainbow.h:7:
rainbow.c: In file included from ./blas.h:17:
rainbow.c: ./blas_avx2.h:265:39: error: always_inline function '_mm256_setzero_si256' requires target feature 'avx', but would be inlined into function 'gf31mat_prod_avx2' that is compiled without support for 'avx'
rainbow.c: for(unsigned i=0;i<n_ymm;i++) r[i] = _mm256_setzero_si256();
rainbow.c: ^
rainbow.c: ./blas_avx2.h:265:39: error: AVX vector return of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI
rainbow.c: ./blas_avx2.h:270:16: error: always_inline function '_mm256_set1_epi16' requires target feature 'avx', but would be inlined into function 'gf31mat_prod_avx2' that is compiled without support for 'avx'
rainbow.c: __m256i bi = _mm256_set1_epi16( b_u16[i] );
rainbow.c: ^
rainbow.c: ./blas_avx2.h:270:16: error: AVX vector return of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI
rainbow.c: ./blas_avx2.h:273:18: error: always_inline function '_mm256_loadu_si256' requires target feature 'avx', but would be inlined into function 'gf31mat_prod_avx2' that is compiled without support for 'avx'
rainbow.c: __m256i tmp = _mm256_loadu_si256( (__m256i*)(mat+j*32) );
rainbow.c: ^
rainbow.c: ./blas_avx2.h:273:18: error: AVX vector return of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI
rainbow.c: ./blas_avx2.h:274:36: error: always_inline function '_mm256_maddubs_epi16' requires target feature 'avx2', but would be inlined into function 'gf31mat_prod_avx2' that is compiled without support for 'avx2'
rainbow.c: r[j] = _mm256_add_epi16( r[j] , _mm256_maddubs_epi16( tmp , bi ) );
rainbow.c: ^
rainbow.c: ./blas_avx2.h:274:36: error: AVX vector argument of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI
rainbow.c: ./blas_avx2.h:274:11: error: always_inline function '_mm256_add_epi16' requires target feature 'avx2', but would be inlined into function 'gf31mat_prod_avx2' that is compiled without support for 'avx2'
rainbow.c: r[j] = _mm256_add_epi16( r[j] , _mm256_maddubs_epi16( tmp , bi ) );
rainbow.c: ^
rainbow.c: ./blas_avx2.h:274:11: error: AVX vector argument of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI
rainbow.c: ./blas_avx2.h:282:10: warning: AVX vector argument of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI [-Wpsabi]
rainbow.c: r[j] = _gf31v_reduce_u16_avx2( r[j] );
rainbow.c: ...

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

Compiler	Implementations
clang -mcpu=native -O3 -fomit-frame-pointer -fwrapv -Qunused-arguments -fPIC -fPIE	T:avx2

Compiler output

Implementation: T:avx2
Security model: timingleaks
Compiler: gcc -march=native -mtune=native -O2 -fomit-frame-pointer -fwrapv -fPIC -fPIE

rainbow.c: In file included from /usr/lib/gcc/x86_64-linux-gnu/10/include/immintrin.h:53,
rainbow.c: from blas_avx2.h:8,
rainbow.c: from blas.h:17,
rainbow.c: from rainbow.h:7,
rainbow.c: from rainbow.c:4:
rainbow.c: blas_avx2.h: In function '_gf31v_reduce_u16_avx2':
rainbow.c: /usr/lib/gcc/x86_64-linux-gnu/10/include/avx2intrin.h:112:1: error: inlining failed in call to 'always_inline' '_mm256_add_epi16': target specific option mismatch
rainbow.c: 112 | _mm256_add_epi16 (__m256i __A, __m256i __B)
rainbow.c: | ^~~~~~~~~~~~~~~~
rainbow.c: In file included from blas.h:17,
rainbow.c: from rainbow.h:7,
rainbow.c: from rainbow.c:4:
rainbow.c: blas_avx2.h:112:9: note: called from here
rainbow.c: 112 | return _mm256_add_epi16( a&mask , _mm256_srli_epi16(a,5) );
rainbow.c: | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
rainbow.c: In file included from /usr/lib/gcc/x86_64-linux-gnu/10/include/immintrin.h:53,
rainbow.c: from blas_avx2.h:8,
rainbow.c: from blas.h:17,
rainbow.c: from rainbow.h:7,
rainbow.c: from rainbow.c:4:
rainbow.c: /usr/lib/gcc/x86_64-linux-gnu/10/include/avx2intrin.h:759:1: error: inlining failed in call to 'always_inline' '_mm256_srli_epi16': target specific option mismatch
rainbow.c: 759 | _mm256_srli_epi16 (__m256i __A, int __B)
rainbow.c: | ^~~~~~~~~~~~~~~~~
rainbow.c: In file included from blas.h:17,
rainbow.c: from rainbow.h:7,
rainbow.c: ...

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

Compiler	Implementations
gcc -march=native -mtune=native -O2 -fomit-frame-pointer -fwrapv -fPIC -fPIE	T:avx2
gcc -march=native -mtune=native -O3 -fomit-frame-pointer -fwrapv -fPIC -fPIE	T:avx2
gcc -march=native -mtune=native -O -fomit-frame-pointer -fwrapv -fPIC -fPIE	T:avx2
gcc -march=native -mtune=native -Os -fomit-frame-pointer -fwrapv -fPIC -fPIE	T:avx2