d0p1/ack
1
0
Fork 0
Code Issues Pull requests Packages Projects Releases Wiki Activity
6879 commits 2 branches 0 tags 43 MiB
Commit graph

29 commits

Author SHA1 Message Date
David Given d20b63dc94 The register allocator is really a pass, so arrange the code like one. 2016-10-05 23:55:38 +02:00
David Given 21034c0d65 No, dammit, for register allocation I need to walk the blocks in *dominance* 
order. Since the dominance tree has changed when I fiddled with the graph, I
need to recompute it, so factor it out of the SSA pass. Code is uglier than I'd
like but at least the RET statement goes last in the generated code now.
 2016-10-05 23:52:54 +02:00
David Given 88fb231d6e Better constraint syntax; mcgg now passes register usage information up to mcg; 
mcg can track individual hop inputs and outputs (needed for live range
analysis!); the register allocator now puts the basic blocks into the right
order in preparation for live range analysis.
 2016-10-05 22:56:25 +02:00
David Given 92502901a7 Better management of register data. Add struct hreg. 2016-10-05 21:00:28 +02:00
David Given ac62c34e19 Add a pass to do critical edge splitting. 2016-10-04 23:42:00 +02:00
David Given 288ee56203 Get quite a long way towards basic output-register equality constraints (needed 
to make special nodes like NOP work properly). Realise that the way I'm dealing
with the instruction selector is all wrong; I need to physically copy chunks of
tree to give to burg (so I can terminate them correctly).
 2016-10-02 23:25:54 +02:00
David Given c079e97492 Perform SSA conversion of locals. Much, *much* better code now, at least 
inasmuch as it looks better before register allocation. Basic blocks now know
their own successors and predecessors (after a certain point in the IR
processing).
 2016-10-02 17:50:34 +02:00
David Given b298c27c63 Refactor mcg.h as it's getting a bit big; keep track of register variables. 2016-10-02 00:30:33 +02:00
David Given a3cfe6047f More rigorous dealing of IR groups; no need for is_generated and is_root any 
more (but now passes are required to set IR roots properly when changing
instructions).
 2016-10-01 22:58:29 +02:00
David Given 21898f784a We're going to need some type inference after all, I think. Let's do a little 
for now and see how it goes.
 2016-10-01 19:10:22 +02:00
David Given 4a3a9a98dc It doesn't really make a lot of sense to have BURG nonterminal names different 
to register classes, so combine them. Refactor the map code.
 2016-10-01 12:17:14 +02:00
David Given 3a973a19f3 Move fatal(), warning() and aprintf() into the new data module (because they're 
really useful).
 2016-09-30 19:10:30 +02:00
David Given a0131fdb47 You know what, the type inference stuff is a complete red herring. What this 
actually needs is a more intelligent register allocator. So, remove the type
inference.
 2016-09-29 19:58:02 +02:00
David Given 4572f1b774 Actually, I don't need vregs: hops work just as well. Particularly if I 
restructure things so that I don't need to walk the blasted ir / burg tree
every time I look at an instruction.
 2016-09-27 23:38:47 +02:00
David Given e77c5164cf Fleshed out hops and vregs. The result is almost looking like code now --- 
uncanny.
 2016-09-27 00:19:45 +02:00
David Given c4b8e00ae2 Revamp the array module not to use nasty macros any more. Slightly more verbose 
to use, but definitely cleaner.
 2016-09-26 22:48:58 +02:00
David Given cc176e5183 Keep more data around about ir instructions. Implement a half-baked type 
inference routine to propagate information about floats up the tree, so we know
whether to put floats into special registers as early as possible.
 2016-09-26 22:12:46 +02:00
David Given 416b13fd76 Start factoring out the hardware op code. 2016-09-25 23:29:59 +02:00
David Given 67eb21d428 Rename struct insn to struct em (throughout). 2016-09-25 12:29:03 +02:00
David Given 629e0ddfc6 Some instruction selection is now happening. 2016-09-24 22:46:08 +02:00
David Given bb9aa030a5 Procedure compilation now happens after the entire EM file has been read in (so 
that we can look inside data blocks which might be defined in the future...
sigh, csa and csb). csa and csb no longer generate invalid IR.
 2016-09-24 01:04:00 +02:00
David Given ed67d427c9 Replaced the block splicer with a trivial block eliminator (which rewrites 
jumps to blocks which contain only a jump). Don't bother storing the bb graph
in the ir nodes; we can find it on demand by walking the tree instead ---
slower, but much easier to understand and more robust. Added a terrible map
library.
 2016-09-23 23:59:15 +02:00
David Given f8bbf9e87d Each pass now lives in its own source file; much cleaner. 2016-09-23 21:07:16 +02:00
David Given 9077baa850 Add a bodged in algorithm for converting basic block communication from stacked 
variables to SSA. Also add dead block removal and block splicing. IR code is
much better now.
 2016-09-22 23:19:29 +02:00
David Given 6a74cb2e11 Tracing cleanup. Simplified the IR code. Some more opcodes. 2016-09-22 00:15:48 +02:00
David Given 36d7d1ee4e Create hacky fake basic blocks for data fragments, used to track which 
instruction labels descriptor blocks refer to; this allows csa and csb to know
where they're going.
 2016-09-20 00:19:39 +02:00
David Given 6ce2495aeb Store the EM code up front and build the basic block graph *before* 
generating the IR code. Lots more IR code.
 2016-09-19 23:06:59 +02:00
David Given 176cd7365c Archival checking of the half-written IR treebuilder. 2016-09-18 23:24:54 +02:00
David Given 24380e2a93 Abstract out the EM reader; skeleton of the tree builder. 2016-09-18 00:02:16 +02:00