Implementation notes: amd64, saber214, crypto_encode/653x3

Computer: saber214
Microarchitecture: amd64; Bulldozer (600f20)
Architecture: amd64
CPU ID: AuthenticAMD-00600f20-1789c3f5
SUPERCOP version: 20240625
Operation: crypto_encode
Primitive: 653x3
TimeObject sizeTest sizeImplementationCompilerBenchmark dateSUPERCOP version
1941778 0 013726 776 800refgcc_-march=native_-mtune=native_-O3_-fwrapv_-fPIC_-fPIE_-gdwarf-4_-Wall2024062620240625
377459 0 011276 816 728refclang_-march=native_-O2_-fwrapv_-Qunused-arguments_-fPIC_-fPIE_-gdwarf-4_-Wall2024062620240625
377459 0 013668 816 744refclang_-march=native_-O3_-fwrapv_-Qunused-arguments_-fPIC_-fPIE_-gdwarf-4_-Wall2024062620240625
888169 0 012084 816 728refclang_-mcpu=native_-O3_-fwrapv_-Qunused-arguments_-fPIC_-fPIE_-gdwarf-4_-Wall2024062620240625
948123 0 09684 816 728refclang_-march=native_-O_-fwrapv_-Qunused-arguments_-fPIC_-fPIE_-gdwarf-4_-Wall2024062620240625
1034178 0 010678 776 800refgcc_-march=native_-mtune=native_-O2_-fwrapv_-fPIC_-fPIE_-gdwarf-4_-Wall2024062620240625
1039169 0 010429 768 800refgcc_-march=native_-mtune=native_-O_-fwrapv_-fPIC_-fPIE_-gdwarf-4_-Wall2024062620240625
1062151 0 09433 752 768refgcc_-march=native_-mtune=native_-Os_-fwrapv_-fPIC_-fPIE_-gdwarf-4_-Wall2024062620240625
1227106 0 09278 808 728refclang_-march=native_-Os_-fwrapv_-Qunused-arguments_-fPIC_-fPIE_-gdwarf-4_-Wall2024062620240625

Compiler output

encode.c: encode.c:33:18: error: always_inline function '_mm256_packus_epi16' requires target feature 'avx2', but would be inlined into function 'crypto_encode_653x3_avx_constbranchindex' that is compiled without support for 'avx2'
encode.c:     __m256i a0 = _mm256_packus_epi16(f0&lobytes,f1&lobytes);
encode.c:                  ^
encode.c: encode.c:36:38: error: always_inline function '_mm256_srli_epi16' requires target feature 'avx2', but would be inlined into function 'crypto_encode_653x3_avx_constbranchindex' that is compiled without support for 'avx2'
encode.c:     __m256i a1 = _mm256_packus_epi16(_mm256_srli_epi16(f0,8),_mm256_srli_epi16(f1,8));
encode.c:                                      ^
encode.c: encode.c:36:62: error: always_inline function '_mm256_srli_epi16' requires target feature 'avx2', but would be inlined into function 'crypto_encode_653x3_avx_constbranchindex' that is compiled without support for 'avx2'
encode.c:     __m256i a1 = _mm256_packus_epi16(_mm256_srli_epi16(f0,8),_mm256_srli_epi16(f1,8));
encode.c:                                                              ^
encode.c: encode.c:36:18: error: always_inline function '_mm256_packus_epi16' requires target feature 'avx2', but would be inlined into function 'crypto_encode_653x3_avx_constbranchindex' that is compiled without support for 'avx2'
encode.c:     __m256i a1 = _mm256_packus_epi16(_mm256_srli_epi16(f0,8),_mm256_srli_epi16(f1,8));
encode.c:                  ^
encode.c: encode.c:38:18: error: always_inline function '_mm256_packus_epi16' requires target feature 'avx2', but would be inlined into function 'crypto_encode_653x3_avx_constbranchindex' that is compiled without support for 'avx2'
encode.c:     __m256i a2 = _mm256_packus_epi16(f2&lobytes,f3&lobytes);
encode.c:                  ^
encode.c: encode.c:39:38: error: always_inline function '_mm256_srli_epi16' requires target feature 'avx2', but would be inlined into function 'crypto_encode_653x3_avx_constbranchindex' that is compiled without support for 'avx2'
encode.c:     __m256i a3 = _mm256_packus_epi16(_mm256_srli_epi16(f2,8),_mm256_srli_epi16(f3,8));
encode.c:                                      ^
encode.c: encode.c:39:62: error: always_inline function '_mm256_srli_epi16' requires target feature 'avx2', but would be inlined into function 'crypto_encode_653x3_avx_constbranchindex' that is compiled without support for 'avx2'
encode.c:     __m256i a3 = _mm256_packus_epi16(_mm256_srli_epi16(f2,8),_mm256_srli_epi16(f3,8));
encode.c:                                                              ^
encode.c: encode.c:39:18: error: always_inline function '_mm256_packus_epi16' requires target feature 'avx2', but would be inlined into function 'crypto_encode_653x3_avx_constbranchindex' that is compiled without support for 'avx2'
encode.c:     __m256i a3 = _mm256_packus_epi16(_mm256_srli_epi16(f2,8),_mm256_srli_epi16(f3,8));
encode.c:                  ^
encode.c: encode.c:41:29: error: always_inline function '_mm256_slli_epi16' requires target feature 'avx2', but would be inlined into function 'crypto_encode_653x3_avx_constbranchindex' that is compiled without support for 'avx2'
encode.c: ...
Number of similar (implementation,compiler) pairs: 4, namely:
ImplementationCompiler
avxclang -march=native -O2 -fwrapv -Qunused-arguments -fPIC -fPIE -gdwarf-4 -Wall (Ubuntu_Clang_14.0.0)
avxclang -march=native -O3 -fwrapv -Qunused-arguments -fPIC -fPIE -gdwarf-4 -Wall (Ubuntu_Clang_14.0.0)
avxclang -march=native -O -fwrapv -Qunused-arguments -fPIC -fPIE -gdwarf-4 -Wall (Ubuntu_Clang_14.0.0)
avxclang -march=native -Os -fwrapv -Qunused-arguments -fPIC -fPIE -gdwarf-4 -Wall (Ubuntu_Clang_14.0.0)

Compiler output

encode.c: encode.c:26:18: error: always_inline function '_mm256_loadu_si256' requires target feature 'avx', but would be inlined into function 'crypto_encode_653x3_avx_constbranchindex' that is compiled without support for 'avx'
encode.c:     __m256i f0 = _mm256_loadu_si256((const __m256i *) (f+0));
encode.c:                  ^
encode.c: encode.c:26:18: error: AVX vector return of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI
encode.c: encode.c:27:18: error: always_inline function '_mm256_loadu_si256' requires target feature 'avx', but would be inlined into function 'crypto_encode_653x3_avx_constbranchindex' that is compiled without support for 'avx'
encode.c:     __m256i f1 = _mm256_loadu_si256((const __m256i *) (f+32));
encode.c:                  ^
encode.c: encode.c:27:18: error: AVX vector return of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI
encode.c: encode.c:28:18: error: always_inline function '_mm256_loadu_si256' requires target feature 'avx', but would be inlined into function 'crypto_encode_653x3_avx_constbranchindex' that is compiled without support for 'avx'
encode.c:     __m256i f2 = _mm256_loadu_si256((const __m256i *) (f+64));
encode.c:                  ^
encode.c: encode.c:28:18: error: AVX vector return of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI
encode.c: encode.c:29:18: error: always_inline function '_mm256_loadu_si256' requires target feature 'avx', but would be inlined into function 'crypto_encode_653x3_avx_constbranchindex' that is compiled without support for 'avx'
encode.c:     __m256i f3 = _mm256_loadu_si256((const __m256i *) (f+96));
encode.c:                  ^
encode.c: encode.c:29:18: error: AVX vector return of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI
encode.c: encode.c:33:18: error: always_inline function '_mm256_packus_epi16' requires target feature 'avx2', but would be inlined into function 'crypto_encode_653x3_avx_constbranchindex' that is compiled without support for 'avx2'
encode.c:     __m256i a0 = _mm256_packus_epi16(f0&lobytes,f1&lobytes);
encode.c:                  ^
encode.c: encode.c:33:18: error: AVX vector argument of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI
encode.c: encode.c:36:38: error: always_inline function '_mm256_srli_epi16' requires target feature 'avx2', but would be inlined into function 'crypto_encode_653x3_avx_constbranchindex' that is compiled without support for 'avx2'
encode.c:     __m256i a1 = _mm256_packus_epi16(_mm256_srli_epi16(f0,8),_mm256_srli_epi16(f1,8));
encode.c:                                      ^
encode.c: encode.c:36:38: error: AVX vector argument of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI
encode.c: encode.c:36:62: error: always_inline function '_mm256_srli_epi16' requires target feature 'avx2', but would be inlined into function 'crypto_encode_653x3_avx_constbranchindex' that is compiled without support for 'avx2'
encode.c: ...
Number of similar (implementation,compiler) pairs: 1, namely:
ImplementationCompiler
avxclang -mcpu=native -O3 -fwrapv -Qunused-arguments -fPIC -fPIE -gdwarf-4 -Wall (Ubuntu_Clang_14.0.0)

Compiler output

encode.c: In file included from /usr/lib/gcc/x86_64-linux-gnu/11/include/immintrin.h:47,
encode.c:                  from encode.c:1:
encode.c: encode.c: In function 'crypto_encode_653x3_avx_constbranchindex':
encode.c: /usr/lib/gcc/x86_64-linux-gnu/11/include/avx2intrin.h:105:1: error: inlining failed in call to 'always_inline' '_mm256_add_epi8': target specific option mismatch
encode.c:   105 | _mm256_add_epi8 (__m256i __A, __m256i __B)
encode.c:       | ^~~~~~~~~~~~~~~
encode.c: encode.c:54:10: note: called from here
encode.c:    54 |     b0 = _mm256_add_epi8(b0,_mm256_set1_epi8(85));
encode.c:       |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
encode.c: In file included from /usr/lib/gcc/x86_64-linux-gnu/11/include/immintrin.h:47,
encode.c:                  from encode.c:1:
encode.c: /usr/lib/gcc/x86_64-linux-gnu/11/include/avx2intrin.h:1044:1: error: inlining failed in call to 'always_inline' '_mm256_permutevar8x32_epi32': target specific option mismatch
encode.c:  1044 | _mm256_permutevar8x32_epi32 (__m256i __X, __m256i __Y)
encode.c:       | ^~~~~~~~~~~~~~~~~~~~~~~~~~~
encode.c: encode.c:52:10: note: called from here
encode.c:    52 |     b0 = _mm256_permutevar8x32_epi32(b0,_mm256_set_epi32(7,3,6,2,5,1,4,0));
encode.c:       |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
encode.c: In file included from /usr/lib/gcc/x86_64-linux-gnu/11/include/immintrin.h:47,
encode.c:                  from encode.c:1:
encode.c: /usr/lib/gcc/x86_64-linux-gnu/11/include/avx2intrin.h:105:1: error: inlining failed in call to 'always_inline' '_mm256_add_epi8': target specific option mismatch
encode.c:   105 | _mm256_add_epi8 (__m256i __A, __m256i __B)
encode.c:       | ^~~~~~~~~~~~~~~
encode.c: encode.c:50:10: note: called from here
encode.c:    50 |     b0 = _mm256_add_epi8(b0,_mm256_slli_epi16(b2&_mm256_set1_epi8(15),4));
encode.c:       |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
encode.c: ...
Number of similar (implementation,compiler) pairs: 4, namely:
ImplementationCompiler
avxgcc -march=native -mtune=native -O2 -fwrapv -fPIC -fPIE -gdwarf-4 -Wall (11.4.0)
avxgcc -march=native -mtune=native -O3 -fwrapv -fPIC -fPIE -gdwarf-4 -Wall (11.4.0)
avxgcc -march=native -mtune=native -O -fwrapv -fPIC -fPIE -gdwarf-4 -Wall (11.4.0)
avxgcc -march=native -mtune=native -Os -fwrapv -fPIC -fPIE -gdwarf-4 -Wall (11.4.0)

TIMECOP error (can be valgrind bug)

Process terminating with default action of signal 4 (SIGILL)
 Illegal opcode at address 0x10A54F
   at 0x...: crypto_encode_653x3_ref_constbranchindex (encode.c:15)
   by 0x...: test (try.c:100)
   by 0x...: main (try-anything.c:345)
Number of similar (implementation,compiler) pairs: 1, namely:
ImplementationCompiler
refclang -march=native -O2 -fwrapv -Qunused-arguments -fPIC -fPIE -gdwarf-4 -Wall (Ubuntu_Clang_14.0.0)

TIMECOP error (can be valgrind bug)

Process terminating with default action of signal 4 (SIGILL)
 Illegal opcode at address 0x10A5EF
   at 0x...: crypto_encode_653x3_ref_constbranchindex (encode.c:15)
   by 0x...: test (try.c:100)
   by 0x...: main (try-anything.c:345)
Number of similar (implementation,compiler) pairs: 1, namely:
ImplementationCompiler
refclang -march=native -O3 -fwrapv -Qunused-arguments -fPIC -fPIE -gdwarf-4 -Wall (Ubuntu_Clang_14.0.0)

TIMECOP error (can be valgrind bug)

Process terminating with default action of signal 4 (SIGILL)
 Illegal opcode at address 0x109C3D
   at 0x...: salsa20.part.0 (try-anything.c:102)
   by 0x...: salsa20 (try-anything.c:85)
   by 0x...: testvector (try-anything.c:124)
   by 0x...: input_prepare (try-anything.c:162)
   by 0x...: test (try.c:95)
   by 0x...: main (try-anything.c:345)
Number of similar (implementation,compiler) pairs: 1, namely:
ImplementationCompiler
refgcc -march=native -mtune=native -O2 -fwrapv -fPIC -fPIE -gdwarf-4 -Wall (11.4.0)

TIMECOP error (can be valgrind bug)

Process terminating with default action of signal 4 (SIGILL)
 Illegal opcode at address 0x10988A
   at 0x...: core (try-anything.c:53)
   by 0x...: salsa20.part.0 (try-anything.c:89)
   by 0x...: salsa20 (try-anything.c:85)
   by 0x...: testvector (try-anything.c:124)
   by 0x...: input_prepare (try-anything.c:162)
   by 0x...: test (try.c:95)
   by 0x...: main (try-anything.c:345)
Number of similar (implementation,compiler) pairs: 1, namely:
ImplementationCompiler
refgcc -march=native -mtune=native -O3 -fwrapv -fPIC -fPIE -gdwarf-4 -Wall (11.4.0)

Passed TIMECOP

TIMECOP iterations: 10
Number of similar (implementation,compiler) pairs: 5, namely:
ImplementationCompiler
refclang -march=native -O -fwrapv -Qunused-arguments -fPIC -fPIE -gdwarf-4 -Wall (Ubuntu_Clang_14.0.0)
refclang -march=native -Os -fwrapv -Qunused-arguments -fPIC -fPIE -gdwarf-4 -Wall (Ubuntu_Clang_14.0.0)
refclang -mcpu=native -O3 -fwrapv -Qunused-arguments -fPIC -fPIE -gdwarf-4 -Wall (Ubuntu_Clang_14.0.0)
refgcc -march=native -mtune=native -O -fwrapv -fPIC -fPIE -gdwarf-4 -Wall (11.4.0)
refgcc -march=native -mtune=native -Os -fwrapv -fPIC -fPIE -gdwarf-4 -Wall (11.4.0)