#Gsdx v0.1.16 avx or sse2 code

+ + Jussi Kivilinna + +Дџix wrong code execution in Poly1305 ARM/NEON implementation. + + Werner Koch + + mpi: Abort on division by zero also in _gcry_mpi_tdiv_qr. + (_gcry_mpi_invm): Use mpi_invm_odd for usual odd cases. + (mpi_invm_generic): Rename from _gcry_mpi_invm. + + mpi: Constant time mpi_inv with some conditions. + * cipher/ecc-ecdsa.c (_gcry_ecc_ecdsa_sign): Likewise. + + NIIBE Yutaka + +ДќSA,ECDSA: Fix use of mpi_invm. + + mpi: Fix return value of mpi_invm_generic. + + NIIBE Yutaka + + mpi: Fix the return value of mpi_invm_generic. + + NIIBE Yutaka + + mpi: Fix flags in mpi_copy for opaque MPI. + + Werner Koch + + Post release updates. + + NIIBE Yutaka + + mpi: Support opaque MPI with gcry_mpi_print. + + Jussi Kivilinna + + random/jitterentropy: fix USE_JENT = JENT_USES_GETTIME code path. + + commit 9c42c068c37411bfda9851b1bd60e927cb81ec7c + * Makefile.am (sign-release): Read variabales from user configuration. + + Werner Koch + + build: Allow customization of the signing key. omp simd reduction) is a more complicated story, but your expectation of performance proportional to vector width may be reasonable.- a/third_party/libgcrypt/ChangeLog +++ b/third_party/libgcrypt/ChangeLog -1,8 +1,119 + Werner Koch + + Release 1.8.7. For float/single complex arithmetic, they produce promotion to double to protect range, which may often be the right choice, even though it gives up part of the potential vector speedup. In case you have vectorized double complex arithmetic, those no-limited-range options silently produce non-vector expansion of abs(), sqrt(), and divide. Boosting unrolling increases size of generated code, which may be the reason why compilers require you to call it out specifically (and I suggest it's important to limit the unrolling).

That's reasonably similar in effect to icc options -xHost -O2 -unroll=4. If your ground rules limit the use of gcc verbosity, you have no satisfactory way out (typical options: -march=native -O3 -funroll-loops -param max-unroll-times=4 -ffast-math -fno-cx-limited-range). Gcc default is no unrolling (beyond what is implicit in vectorization). In a few cases, AVX2 instructions may even lose performance compared with earlier AVX, when there is no unrolling. If you switch to a compiler which supports unrolling in addition to vectorization, a 50% gain from SSE to AVX2 might be a reasonable target. I've seen no difference in performance with VS2012 between SSE2 and AVX when both vectorize. In the 'LCD'/vectors benchmark suite, the performance gain from VS2012 AVX-128 to VS2015 AVX2-256 (when both cases vectorize) is typically the 40% you mention. As andysem hinted, AVX2 performance depends more on optimized unrolling (at least when comparing SSE2 and AVX2 on a CPU which supports both).

Gsdx v0.1.16 avx or sse2