Builds fail on AMD without flags

[KeeJef]

Builds still failing with fastadd errors if make is not passed

CFLAGS='-march=haswell -mfma' CXXFLAGS='-march=haswell -mfma'

We should include some kind of fix for this since AMD architecture is very common and we don't want people to have to go searching for the flags just to build Lokinet

In file included from /usr/lib/gcc/x86_64-linux-gnu/7/include/immintrin.h:85:0, from /home/dev/loki-network/crypto/libntrup/src/avx/mult.c:3: /home/dev/loki-network/crypto/libntrup/src/avx/mult.c: In function ‘fastadd’: /usr/lib/gcc/x86_64-linux-gnu/7/include/fmaintrin.h:63:1: error: inlining failed in call to always_inline ‘_mm256_fmadd_ps’: target specific option mismatch _mm256_fmadd_ps (__m256 __A, __m256 __B, __m256 __C) ^~~~~~~~~~~~~~~ /home/dev/loki-network/crypto/libntrup/src/avx/mult.c:128:10: note: called from here return _mm256_fmadd_ps(y, v1_float, x); ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ In file included from /usr/lib/gcc/x86_64-linux-gnu/7/include/immintrin.h:85:0, from /home/dev/loki-network/crypto/libntrup/src/avx/mult.c:3: /usr/lib/gcc/x86_64-linux-gnu/7/include/fmaintrin.h:63:1: error: inlining failed in call to always_inline ‘_mm256_fmadd_ps’: target specific option mismatch _mm256_fmadd_ps (__m256 __A, __m256 __B, __m256 __C) ^~~~~~~~~~~~~~~ /home/dev/loki-network/crypto/libntrup/src/avx/mult.c:128:10: note: called from here return _mm256_fmadd_ps(y, v1_float, x); ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

This issue is a replication of #92

[despair86]

from the wikipedia page on FMA3/FMA4:

Zen: WikiChip's testing shows FMA4 still appears to work (under the conditions of the tests) despite not being officially supported and not even reported by CPUID. 
This has also been confirmed by Agner.[5] But other tests gave wrong results.[6] 
AMD Official Web Site FMA4 Support Note ZEN CPUs = AMD ThreadRipper 1900x, R7 Pro 1800, 1700, R5 Pro 1600, 1500, R3 Pro 1300, 1200, R3 2200G, R5 2400G.[7][8][9]

GCC and clang go on what CPUID returns for FPU feature flags

[KeeJef]

We should work out how to do CPU detection to rectify this bug its really annoying :(

[KeeJef]

confirming that #138 fixes this for me

[neuroscr]

@KeeJef you mean AMD_RYZEN_HACK makes it compile for you?

I'd like to keep the issue open until we can actually get a build-time test, if that's possible.

[KeeJef]

Yes AMD_RYZEN_HACK makes it compile without needing the above flags

[despair86]

ye

that's because it fills in those flags for you -march=native -mavx -mavx2 -mfma -mfpmath=sse (yes that last one is there because multilib....there's probably 6 people that are going to throw the -m32 switch in there)