diff options
author | Torbjorn Granlund <torbjorng@google.com> | 2015-10-19 13:57:58 +0200 |
---|---|---|
committer | Torbjorn Granlund <torbjorng@google.com> | 2015-10-19 13:57:58 +0200 |
commit | 59b93b18d37a10a6eb31547304eab2ef454e9c23 (patch) | |
tree | 8b8c9933d445d4ea67786f05600cb7ebe7abc2a2 /NEWS | |
parent | b9acd91382edd6ca5a49f81ec92035e168a6c80a (diff) | |
download | gmp-59b93b18d37a10a6eb31547304eab2ef454e9c23.tar.gz |
Some NEWS clarifications.
Diffstat (limited to 'NEWS')
-rw-r--r-- | NEWS | 9 |
1 files changed, 6 insertions, 3 deletions
@@ -19,8 +19,11 @@ Changes between GMP version 6.0.* and 6.1.0 * Various problems related to precision for mpf have been fixed. + * Fixed incorrect stack alignment in calls from assembly code. + SPEEDUPS - * Speedup for Intel Broadwell. + * Speedup for Intel Broadwell and Skylake though assembly code making use of + new ADX instructions. * Square root is now faster when the remainder is not needed. Also the speed to compute the k-th root improved, for small sizes. @@ -37,14 +40,14 @@ Changes between GMP version 6.0.* and 6.1.0 * Improved arm64 support. - * Support for more arm processors. + * Support for more 32-bit arm processors. * Support compilation with clang/llvm for more platforms. Caution: GMP triggers mis-compilation bugs in clang for many platforms, such as arm, x86 (32-bit and 64-bit), powerpc, mips. * Support for AVX-less modern x86 CPUs, either because the CPU vendor chose - to disable it, or because the running kernel lacks AVX context switch + to disable AVX, or because the running kernel lacks AVX context switch support. * Stack usage trimmed; we believe 512 KiB is now sufficient for any GMP call, |