Merge branch 'libmath-merge'
Turn off fancy stuff when we have an old assembler.
Regenerate netbsd bootstrap
Merge remote-tracking branch 'npnth/libmath'
Don't specify .comm alignment.
libjson: allow square brackets in strings
Make std.writeall api slightly nicer.
Implement most of expm1. Errors in huge numbers.
Merge branch 'master' of /git/ori/mc
Extract slices recursively.
We don't do PIE yet, let's deal with it later.
Add a special exit stack to our thread library.
Fix OpenBSD 6.1 specialization.
Put _s in the right spot in ASM files.
Subnormalize special-case floats in fltXYassem.
Implement Horner's Scheme for polynomial evaluation.
Boilerplate FP control word code for ASM round.
Explicitly type some variables.
Merge remote-tracking branch 'ori/master' into HEAD
Don't mutate the traits on builtin types.
Remove stray ',' from error message.
Merge trait specs if provided.
Add missing libcrypto dependency
Add support for ragged RSA numbers.
Add first tested RSA implementation.
Add big-endian constructor for ctbig.
Break out some fpmath functions to utililty file.
Remove dead code/comments.
Add a nontrivial modpow test.
Make constant time ints package local.
Add support for `-T` option to compile in test mode.
Ensure that we initialize the environment fully.
Comment out unused tests (oops)
Add mising file to build list.
Add constant time bigint ops.
Remove suprious free. We don't want to mutate the targets.
Allow 'make release' to override version.
Improve comments on the regex interpreter.
Make output buffers bigger.
Clean up and remove edge cases in strfind.
Handle subtests correctly.
Handle zero length needles.
Improve test runner test filtering.
Add support for OpenBSD's MAP_STACK mmap flags, use it in libthread
Add OpenBSD 6.3 system call support.
Fix the cpufeatures... er... features on plan 9.
Fix the cpufeatures... er... features on plan 9.
Even more subnormal detection in flt32fromflt64.
Drop "fpmath-" prefix, it's redundant
Implement assembly version of fused multiply-add
Do not use packed SIMD instructions for only one argument
Correct off-by-one for subnormals when converting flt64 to flt32
Merge branch 'libmath-asm' into libmath
Merge remote-tracking branch 'ori/master' into libmath
Replace fma32 with simpler, cleaner algorithm from musl
Correct sign handling in fma32 when one of x, y is NaN
Fix fma64 for subnormal results arising from normal inputs, start fma32
Allow CPU feature detection by mbld, and add "sse4" tag
Add -T to use *only* the passed sytem tags.
Allow intfmt to handle uint64s with {p=0,w=64,b=2}
Fix corrupted merge of 'autoexpr'
Check both digits for < Kmin
Add benchmark for bigint multiplication.
Implement karatsuba multiplication.
Add incomplete constant time bigint code.
Implement Kahan and Priest summation
Merge branch 'autoexpr' into libmath
Allow CPU feature detection by mbld, and add "sse4" tag
Add asm implementations for trunc & co.
Make floor & friends slower, but more readable
Reorder arguments/results of fltXY{explode/assem}
Make fltXYexplode and fltXYassem inverses of each other
Implement a few simple floating-point functions
Reinit nexttid on every run.
Add constant time 'min()' and 'max()'