Implementation notes: amd64, samba, crypto_sign/lattisigns512

Computer: samba
Microarchitecture: amd64; Skylake (506e3)
Architecture: amd64
CPU ID: GenuineIntel-000506e3-bfebfbff
SUPERCOP version: 20240107
Operation: crypto_sign
Primitive: lattisigns512

Time	Object size	Test size	Implementation	Compiler	Benchmark date	SUPERCOP version
486106	52652 0 68	88677 812 1920	`T:avx`	`gcc_-march=native_-mtune=native_-O3_-fomit-frame-pointer_-fwrapv_-fPIC_-fPIE`	20231214	20231212
510409	37734 0 68	71533 812 1920	`T:avx`	`gcc_-march=native_-mtune=native_-O2_-fomit-frame-pointer_-fwrapv_-fPIC_-fPIE`	20231214	20231212
524752	41067 0 56	77723 852 1856	`T:avx`	`clang_-march=native_-O2_-fomit-frame-pointer_-fwrapv_-Qunused-arguments_-fPIC_-fPIE`	20231214	20231212
534752	41707 0 56	78611 852 1888	`T:avx`	`clang_-march=native_-O3_-fomit-frame-pointer_-fwrapv_-Qunused-arguments_-fPIC_-fPIE`	20231214	20231212
536703	36420 0 68	68573 804 1888	`T:avx`	`gcc_-march=native_-mtune=native_-Os_-fomit-frame-pointer_-fwrapv_-fPIC_-fPIE`	20231214	20231212
547101	36379 0 56	69877 844 1888	`T:avx`	`clang_-march=native_-Os_-fomit-frame-pointer_-fwrapv_-Qunused-arguments_-fPIC_-fPIE`	20231214	20231212
561022	36520 0 56	69307 852 1824	`T:avx`	`clang_-march=native_-O_-fomit-frame-pointer_-fwrapv_-Qunused-arguments_-fPIC_-fPIE`	20231214	20231212
565322	37199 0 68	70597 812 1920	`T:avx`	`gcc_-march=native_-mtune=native_-O_-fomit-frame-pointer_-fwrapv_-fPIC_-fPIE`	20231214	20231212

Compiler output

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

ntt_transform.c: ntt_transform.c:27:11: error: always_inline function '_mm256_set_pd' requires target feature 'avx', but would be inlined into function 'ntt_transform' that is compiled without support for 'avx'
ntt_transform.c: vpinv = _mm256_set_pd(PARAM_APPROX_P_INVERSE, PARAM_APPROX_P_INVERSE, PARAM_APPROX_P_INVERSE, PARAM_APPROX_P_INVERSE);
ntt_transform.c: ^
ntt_transform.c: ntt_transform.c:27:11: error: AVX vector return of type '__m256d' (vector of 4 'double' values) without 'avx' enabled changes the ABI
ntt_transform.c: ntt_transform.c:28:11: error: always_inline function '_mm256_set_pd' requires target feature 'avx', but would be inlined into function 'ntt_transform' that is compiled without support for 'avx'
ntt_transform.c: vp = _mm256_set_pd(8383489., 8383489., 8383489., 8383489.);
ntt_transform.c: ^
ntt_transform.c: ntt_transform.c:28:11: error: AVX vector return of type '__m256d' (vector of 4 'double' values) without 'avx' enabled changes the ABI
ntt_transform.c: ntt_transform.c:32:10: error: always_inline function '_mm256_load_pd' requires target feature 'avx', but would be inlined into function 'ntt_transform' that is compiled without support for 'avx'
ntt_transform.c: vo10 = _mm256_load_pd(o+pos);
ntt_transform.c: ^
ntt_transform.c: ntt_transform.c:32:10: error: AVX vector return of type '__m256d' (vector of 4 'double' values) without 'avx' enabled changes the ABI
ntt_transform.c: ntt_transform.c:33:10: error: always_inline function '_mm256_load_pd' requires target feature 'avx', but would be inlined into function 'ntt_transform' that is compiled without support for 'avx'
ntt_transform.c: vo20 = _mm256_load_pd(o+pos+4);
ntt_transform.c: ^
ntt_transform.c: ntt_transform.c:33:10: error: AVX vector return of type '__m256d' (vector of 4 'double' values) without 'avx' enabled changes the ABI
ntt_transform.c: ntt_transform.c:34:10: error: always_inline function '_mm256_load_pd' requires target feature 'avx', but would be inlined into function 'ntt_transform' that is compiled without support for 'avx'
ntt_transform.c: neg2 = _mm256_load_pd(_neg2);
ntt_transform.c: ^
ntt_transform.c: ntt_transform.c:34:10: error: AVX vector return of type '__m256d' (vector of 4 'double' values) without 'avx' enabled changes the ABI
ntt_transform.c: ntt_transform.c:35:10: error: always_inline function '_mm256_load_pd' requires target feature 'avx', but would be inlined into function 'ntt_transform' that is compiled without support for 'avx'
ntt_transform.c: neg4 = _mm256_load_pd(_neg4);
ntt_transform.c: ^
ntt_transform.c: ntt_transform.c:35:10: error: AVX vector return of type '__m256d' (vector of 4 'double' values) without 'avx' enabled changes the ABI
ntt_transform.c: ntt_transform.c:41:11: error: always_inline function '_mm256_load_pd' requires target feature 'avx', but would be inlined into function 'ntt_transform' that is compiled without support for 'avx'
ntt_transform.c: ...

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

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