Activity for Tomas Rokicki
-
Tomas Rokicki committed [e61492] on Code
Made random fill use C++11's Mersenne Twister for random fill, rather
-
Fix a bug where setbit on very large universes in 8-bit hashlife trashed
-
Don't permit magnification to go below 2^(-2^20).
-
Restore bailIfCalculating() everywhere we think it was, and if we can
-
Fix a critical bug that occurs when parsing a bigint from string
-
Set the stack size on Windows to a default of 32MB (the normal Windows
-
Move localdeltaforward down to tiles in qlifealgo, and only set
-
Fix qlife for an edge condition that happens with some torus neighborhoods
-
Howdy! My first reaction is, wow, if you're going to support all of that new functionality, is there any reason to make it at all compatible? Golly will reject the file anyway when it sees stuff it doesn't recognize (not from the version line but it's pretty strict about the remaining format). In particular, if you're going to permit JSON fragments, why not just have the whole file be JSON? (I'm not personally a fan of JSON, but for many uses it's better than foisting another textual format on the...
-
Updated Rubik's cube solver; it should now initialize almost twice
-
Eliminated spurious noise, likely from an applied patch, that
-
Fixed division operators as part of converting Python2 to Python3.
-
Initial changes for Python2 to Python3 changes. Ran 2to3 over all
-
Fixed a problem where soft interrupts can update the generation
-
Oops; had this open in an editor; should have been part of the last
-
Make the hashlife algorithms not abort with bad increment by
-
Made a very small change to qlifealgo; the long in tiles should
-
First attempt to fix a performance bug where changing the rule
-
Use fast popcount in golly on Unix and Mac (really when compiling with
-
Updated help to be consistent with new circular larger than life rule
-
Updated code to take r, not r^2, in rule spec for circular neighborhoods.
-
Fix fencepost error found by Andrew in parsing circular LtL rules.
-
Added a note to the change list.
-
Changed NS to NC in LTL rules.
-
Added 'NumRandomizedCellStates' to lifealgo to make random fills work better.
-
Added circular neighborhoods to larger than life. They are slower than
-
Tiny tiny performance improvement in getbounds.
-
Removed an unneeded variable.
-
Probably a tiny bit faster; shorter.
-
A bit more rearrangement for simplicity.
-
Shorter and probably slightly faster script.
-
Swapped x and z; x is slightly faster to compute (for the moment).
-
Slightly faster 3D?
-
Fix an issue where if the zeronodes are not as deep as the root, drawing
-
Added a missing memset to clear the buffer before we render.
-
Removed conditions in rendering leaves for ghbase 1:1 case.
-
If we load a super-large RLE that exceeds memory constraints, we can
-
Faster getcells for ghashbase; may want to carry this over to
-
Fixed initialization of alloced.
-
Extend getcells to ltl.
-
Added getcells() to lifealgo and implemented it for quicklife,
-
Fix an issue with writing the macrocell format that was introduced
-
Put hperf into util and removed hperf source files.
-
Fixed the makefiles to not mention hperf. Added hperf to the
-
Added a "soft" interrupt to the hash-based algorithms triggered when
-
Initialized some fields we weren't initializing before.
-
Added frames per second to stats.
-
Added some more stats; separated hperf into its own source file;
-
Experimental start at more status updates; they are laughably bad for
-
Fix another bug where during ensurehash the root was changing.
-
Fix a crash in 32-bit builds caused by my population changes.
-
Added a comment.
-
Use of placement new means I should also use direct destructor
-
Fixed a bug I introduced in my popcount speedup.
-
Moved the timing changes from cmdline/bgolly to gollybase/util so
-
Changed the population counting somewhat to improve speed.
-
Made bgolly echo its arguments.
-
Eliminate extra space in GC message.
-
Added information on what the resize size is to the status message.
-
Improve the GC message to include more accurate free values and not let the free node count go negative.
-
Make things compile on all platforms by using
-
A better way to change increments after a GC (and when there is
-
Only increment halvesdone when we are about to assign the result;
-
Added a cast for the prefetch command in platform.h
-
First attempt at making timing work with Windows.
-
Added two options to bgolly: -b (benchmark/show timestamps), -T
-
Slight tweak; minor performance improvement, small decrease in
-
Change prefetch to use intrinsics rather than gcc's builtin.
-
Eliminate some compiler warnings on windows by introducing
-
Went back to a more conservative prefetch.
-
Complete the load factor improvement change.
-
This is the fastest of three prefetch options I've tried.
-
Added main memory prefetching to Golly. You can turn this off
-
Added code to explicitly do a garbage collection before any
-
Added a load factor variable to tweak the hash; set it to a reasonable
-
Rewrote the hashtable routines in hlifealgo and ghashbase to use
-
Fixed an issue when setting (or getting!) cell values on very big universes.
-
Fixed a crash that occurred when rendering during updates; we need
-
Protect against undefined behavior if strchr re...
-
Fix two bugs in golly when you try to set bits ...
-
Made ruletreealgo check tree levels as it loads...
-
Add a conditional on the drift prevention calcu...
-
Fixed a selection paste viewport positioning is...
-
Fixed contains again to not count partial pixels.
-
Changed default rounding in viewport back to wh...
-
More work on positioning.
-
Eliminated a couple of unnecessary memory alloc...
-
Changed the way fit and reposition work, so if ...
-
Undid my "centering" fix because it doesn't act...
-
Merge branch 'master' of ssh://git.code.sf.net/...
-
Improve rounding when "centering" a pattern.
-
Fix oddgen lowerRightPixel() for qlife. Now ed...
-
Fix a rounding error in algo->lowerRightPixel()...
-
Merge branch 'master' of ssh://git.code.sf.net/...
-
Fix a bug in qlifedraw() that sometimes makes p...
-
Initialize liferules better, so hlifealgo doesn...
1
