Glibc 2.22 Will Have NaCl Support For ARMv7-A, Libmvec Vector Math Library
One of the features coming for glibc 2.22 is a port to Google's Native Client for ARMv7-A while separately there is also a new vector math library (libmvec) for OMP4.
Roland McGrath of Google contributed a port of Native Client (NaCl) for running on ARMv7-A with this next release. Glibc had been ported to NaCl for x86 architectures for some years now while with the next release it's getting support for ARMv7-A to ease the process of running GNU software via this Google sandboxing system on ARM hardware. Roland finished committing it on Tuesday.
While on the topic of glibc 2.22, this upcoming version will also bring a new vector math library named libmvec with vectorized x86_64 implementations of cos, cosf, sin, sinf, sincos, sincosf, log, logf, exp, expf, pow, and powf. Libmvec aims to suit the SIMD construct needs of the OpenMP 4.0 specification. Libmvec is used automatically when on GCC 4.9 or newer and -fopenmp -ffast-math are passed for the -O1 optimization level and higher. More details via this Wiki page.
Other glibc 2.22 changes will include a PowerPC/PowerPC64 optimization for TLS, a new fmemopen implementation that aims for POSIX compliance, and various bug/security fixes.
The glibc 2.22 release should be coming up very soon and when it happens you can expect to read about it on Phoronix.
Roland McGrath of Google contributed a port of Native Client (NaCl) for running on ARMv7-A with this next release. Glibc had been ported to NaCl for x86 architectures for some years now while with the next release it's getting support for ARMv7-A to ease the process of running GNU software via this Google sandboxing system on ARM hardware. Roland finished committing it on Tuesday.
While on the topic of glibc 2.22, this upcoming version will also bring a new vector math library named libmvec with vectorized x86_64 implementations of cos, cosf, sin, sinf, sincos, sincosf, log, logf, exp, expf, pow, and powf. Libmvec aims to suit the SIMD construct needs of the OpenMP 4.0 specification. Libmvec is used automatically when on GCC 4.9 or newer and -fopenmp -ffast-math are passed for the -O1 optimization level and higher. More details via this Wiki page.
Other glibc 2.22 changes will include a PowerPC/PowerPC64 optimization for TLS, a new fmemopen implementation that aims for POSIX compliance, and various bug/security fixes.
The glibc 2.22 release should be coming up very soon and when it happens you can expect to read about it on Phoronix.
3 Comments