MasterDuke17
MasterDuke17
I'm experimenting with using DwarFS for a very similar use case as yours (a build of each Rakudo commit). However, given there are several new commits every day, creating a...
that uses Barrett reduction to speed up stringifying large integers. Some benchmarking, done with a lightly modified `demo/timing.c`: ``` CLK_PER_SEC == 3600313344 mp_torad of 2**1 => 22224156/sec, 162 cycles, 0.0000...
For background, I starting looking at this when a profile of a Perl 6 program that calculated Ackerman numbers showed that 90% of the time was spent stringifying a big...
With the recent change to make using the mimalloc allocator the default, it was discovered that it does such a good job that the FSA doesn't provide any extra benefit....
``` src/profiler/telemeh.c:156:5: warning: misaligned atomic operation may incur significant performance penalty; the expected alignment (8 bytes) exceeds the actual alignment (4 bytes) [-Watomic-alignment] MVM_incr(&intervalIDCounter); ^ src/moar.h:325:24: note: expanded from macro...
The Mersenne Twister is now a bit outdated and there are alternatives that are both faster and have better statistical properties. `my num $r; my int $i := 0; my...
Happens with `--no-mimalloc` also ``` [dan@alexandria perl6]$ MVM_SPESH_DISABLE=1 valgrind ./install/bin/raku -e 'my $t = ("errors.md.tmpl" xx 1_000).join(" foo "); race for ^100_000 { my $page = $t; $page ~~ s:g[...
This makes all the build and code changes necessary to use GMP for our big integer library instead of LibTomMath. Uses the ShiftMediaProject fork of GMP because it adds support...
This will fix it complaining about `%h` and `%z` being invalid sprintf directives.