1 files changed, 36 insertions, 12 deletions
diff --git a/README.md b/README.md
index 4ec5eff..d829af4 100644
--- a/README.md
+++ b/README.md
@@ -30,30 +30,54 @@ for benchmarks.
 ## TODO:
-### Documentation and interface
+### Coordinates
-* inline some documentation as comments in cube.h or cube.c
-* README.md (maybe convert to txt?) becomes the reference documentation
-### More features
+* [done] eo
+* co
+* ep
+* epsep
+* cp
+* cpsep
+* cphtr
-* move() that takes a string (alg) as input
+What about symcoord?
-* coordinates: co, eo, epsep, cpsep_sym, cocpsep_sym, cphtr_sym, cocphtr_sym
 ### Solving
-* Fixed depth
+All solving functions take a cube and some parameters as input.
-* pruning tables (1 bit per entry + fallback)
-* Takes as parameters the amount of memory to use and a FILE for the tables
+* Depth [uint, <= 20]: all solvers work at fixed depth. The caller
-* Use multi-move (up to 4/5 moves at once)
+  implementation can implement an A* search.
+* Full [bool]: if false, stop at first solution found, otherwise
+  find all solutions at that depth.
+* Table [uint8_t *]: table with all the necessare pre-computed info.
+  The table can be generated with a companion function, but reading
+  from and writing to file is delegated to the caller implementation.
+Implement the following solvers:
+* Slow: basic solver without any table.
+* H48: one-bit-per-entry table + fallback, 48 symmetries and so on.
+  See planner.
+* nxopt31: mostly for comparison.
+* other nxopt solvers: make generic and take the type as parameter.
+* Step solver: take a coordinate function and a moveset as a parameter.
 ### cube.h changes
 * Consider removing zerocube() from the api
 * prefix public functions with nissy_ or something similar
+* move() that takes a string (alg) as input
+### Documentation and interface
-### Future optimizations
+* inline some documentation as comments in source code
+* README.md (maybe convert to txt?) becomes the reference documentation
+### Optimizations
+* Trans: don't do full compose, for some trans composing perm is enough.
+  Split out sumco() as a separate function and refactor, optimize.
+* Use multi-move (up to 4/5 moves at once)
 * CO is the worst part of moving, transforming and inverting. Try basing
  everything on representing the cube without CO and apply it only at the
  end to check that it is actually solved.

diff --git a/README.md b/README.md index 4ec5eff..d829af4 100644 --- a/README.md +++ b/README.md
@@ -30,30 +30,54 @@ for benchmarks.
30		30
31	## TODO:	31	## TODO:
32		32
33	### Documentation and interface	33	### Coordinates
34
35	* inline some documentation as comments in cube.h or cube.c
36	* README.md (maybe convert to txt?) becomes the reference documentation
37		34
38	### More features	35	* [done] eo
		36	* co
		37	* ep
		38	* epsep
		39	* cp
		40	* cpsep
		41	* cphtr
39		42
40	* move() that takes a string (alg) as input	43	What about symcoord?
41	* coordinates: co, eo, epsep, cpsep_sym, cocpsep_sym, cphtr_sym, cocphtr_sym
42		44
43	### Solving	45	### Solving
44		46
45	* Fixed depth	47	All solving functions take a cube and some parameters as input.
46	* pruning tables (1 bit per entry + fallback)	48
47	* Takes as parameters the amount of memory to use and a FILE for the tables	49	* Depth [uint, <= 20]: all solvers work at fixed depth. The caller
48	* Use multi-move (up to 4/5 moves at once)	50	implementation can implement an A* search.
		51	* Full [bool]: if false, stop at first solution found, otherwise
		52	find all solutions at that depth.
		53	* Table [uint8_t *]: table with all the necessare pre-computed info.
		54	The table can be generated with a companion function, but reading
		55	from and writing to file is delegated to the caller implementation.
		56
		57	Implement the following solvers:
		58	* Slow: basic solver without any table.
		59	* H48: one-bit-per-entry table + fallback, 48 symmetries and so on.
		60	See planner.
		61	* nxopt31: mostly for comparison.
		62	* other nxopt solvers: make generic and take the type as parameter.
		63	* Step solver: take a coordinate function and a moveset as a parameter.
49		64
50	### cube.h changes	65	### cube.h changes
51		66
52	* Consider removing zerocube() from the api	67	* Consider removing zerocube() from the api
53	* prefix public functions with nissy_ or something similar	68	* prefix public functions with nissy_ or something similar
		69	* move() that takes a string (alg) as input
		70
		71	### Documentation and interface
54		72
55	### Future optimizations	73	* inline some documentation as comments in source code
		74	* README.md (maybe convert to txt?) becomes the reference documentation
		75
		76	### Optimizations
56		77
		78	* Trans: don't do full compose, for some trans composing perm is enough.
		79	Split out sumco() as a separate function and refactor, optimize.
		80	* Use multi-move (up to 4/5 moves at once)
57	* CO is the worst part of moving, transforming and inverting. Try basing	81	* CO is the worst part of moving, transforming and inverting. Try basing
58	everything on representing the cube without CO and apply it only at the	82	everything on representing the cube without CO and apply it only at the
59	end to check that it is actually solved.	83	end to check that it is actually solved.