David Given [Fri, 14 Oct 2016 21:17:06 +0000 (23:17 +0200)]
Factor out the register allocation routines to make them easier to deal with.
David Given [Fri, 14 Oct 2016 21:12:29 +0000 (23:12 +0200)]
Fix stupid issue where hop output registers were being overwritten, leading to
invalid SSA form.
David Given [Fri, 14 Oct 2016 20:17:02 +0000 (22:17 +0200)]
Output register equality constraints work.
David Given [Wed, 12 Oct 2016 21:12:53 +0000 (23:12 +0200)]
Make loads and stores in the table nicer; fix a place where it looked like it
was working but only accidentally.
David Given [Wed, 12 Oct 2016 20:58:46 +0000 (22:58 +0200)]
Keep track of register types as well as attributes --- the type being how we
find new registers when evicting values. Input constraints work (they were
being ignored before). Various bug fixing so they actually work.
David Given [Wed, 12 Oct 2016 19:50:12 +0000 (21:50 +0200)]
Add code to remove unused phis, converting to pruned SSA form, to avoid
confusing the register allocator later.
David Given [Tue, 11 Oct 2016 22:45:36 +0000 (00:45 +0200)]
Don't allow the same IR to be added to the sequence list more than once
(sometimes happens because op_dup, but makes no sense).
David Given [Tue, 11 Oct 2016 21:17:30 +0000 (23:17 +0200)]
Clean up the allocator a bit, in preparation for making it lots more
complicated; no semantic changes.
David Given [Mon, 10 Oct 2016 22:29:18 +0000 (00:29 +0200)]
A few more opcodes.
David Given [Mon, 10 Oct 2016 22:23:35 +0000 (00:23 +0200)]
A little fiddling with store instructions. The PowerPC is not friendly to
iburg.
David Given [Mon, 10 Oct 2016 22:12:11 +0000 (00:12 +0200)]
Refactored the way hops are rendered; add support for emitting code (although
with no prologue or epilogue yet).
David Given [Mon, 10 Oct 2016 21:19:46 +0000 (23:19 +0200)]
Register allocator now gets all the way through all of my test file without
crashing (albeit with register moves and swaps stubbed out). Correct code? Who
knows.
David Given [Mon, 10 Oct 2016 21:18:37 +0000 (23:18 +0200)]
D'oh, need multiple passes over the edge splitter in order to properly find all
cases.
David Given [Sun, 9 Oct 2016 20:04:20 +0000 (22:04 +0200)]
Calculate phi congruency groups; use them to solve the
importing-hreg-from-the-future problem (probably poorly).
David Given [Sun, 9 Oct 2016 20:03:44 +0000 (22:03 +0200)]
Add pmap_remove().
David Given [Sun, 9 Oct 2016 13:09:34 +0000 (15:09 +0200)]
The register allocator now makes a spirited attempt to honour register
attributes when allocating. Unfortunately, backward edges don't work (because
the limited def-use chain stuff doesn't work across basic blocks). Needs more
thought.
David Given [Sun, 9 Oct 2016 13:08:03 +0000 (15:08 +0200)]
Floating point promotion was broken since the IR float change. Fix.
David Given [Sun, 9 Oct 2016 12:45:13 +0000 (14:45 +0200)]
Add some more opcodes; rearrange the registers to be more PowerPC-friendly.
David Given [Sun, 9 Oct 2016 10:32:36 +0000 (12:32 +0200)]
Predicates can now take numeric arguments. The PowerPC predicates have been
turned into generic ones (as they'll be useful everywhere). Node arguments for
predicates require the '%' prefix for consistency. Hex numbers are permitted.
David Given [Sat, 8 Oct 2016 21:32:54 +0000 (23:32 +0200)]
Basic register allocation works!
David Given [Sat, 8 Oct 2016 21:32:13 +0000 (23:32 +0200)]
Replace pmap_get() with pmap_findleft() and pmap_findright(), to allow lookups
in both directions.
David Given [Sat, 8 Oct 2016 10:15:21 +0000 (12:15 +0200)]
Only allocate an output vreg if the instruction actually wants one.
David Given [Sat, 8 Oct 2016 09:35:33 +0000 (11:35 +0200)]
Instruction predicates work now.
David Given [Sat, 8 Oct 2016 09:07:28 +0000 (11:07 +0200)]
Minor cleanup.
David Given [Sat, 8 Oct 2016 08:21:24 +0000 (10:21 +0200)]
Fix bug where pushes were being placed in the wrong blocks.
David Given [Fri, 7 Oct 2016 22:21:23 +0000 (00:21 +0200)]
Overhaul of everything phi related; critical edge splitting now happens before
anything SSA happens; liveness calculations now look like they might be
working.
David Given [Fri, 7 Oct 2016 22:20:26 +0000 (00:20 +0200)]
Add array_{appendall,removeall,appendallu}.
David Given [Thu, 6 Oct 2016 19:34:21 +0000 (21:34 +0200)]
Refactor the cfg and dominance stuff to make it a lot nicer.
David Given [Thu, 6 Oct 2016 19:33:43 +0000 (21:33 +0200)]
We're not using 'allocates' any more; clean up.
David Given [Wed, 5 Oct 2016 21:55:38 +0000 (23:55 +0200)]
The register allocator is really a pass, so arrange the code like one.
David Given [Wed, 5 Oct 2016 21:55:04 +0000 (23:55 +0200)]
Warning fix.
David Given [Wed, 5 Oct 2016 21:52:54 +0000 (23:52 +0200)]
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.
David Given [Wed, 5 Oct 2016 21:24:29 +0000 (23:24 +0200)]
Allowing an input filename on the command line makes debuggers happy. (Then we
don't need to redirect stdin.)
David Given [Wed, 5 Oct 2016 20:56:25 +0000 (22:56 +0200)]
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.
David Given [Wed, 5 Oct 2016 19:07:29 +0000 (21:07 +0200)]
Made sure that all files end in vim magic.
David Given [Wed, 5 Oct 2016 19:00:28 +0000 (21:00 +0200)]
Better management of register data. Add struct hreg.
David Given [Tue, 4 Oct 2016 21:42:00 +0000 (23:42 +0200)]
Add a pass to do critical edge splitting.
David Given [Tue, 4 Oct 2016 21:28:16 +0000 (23:28 +0200)]
Added support for the op_bXX conditional branch instructions.
David Given [Tue, 4 Oct 2016 20:36:01 +0000 (22:36 +0200)]
Fix the horror of the startup code; now uses getopt and stuff and the debug
flags can be set as an option.
David Given [Tue, 4 Oct 2016 20:35:08 +0000 (22:35 +0200)]
Remove unused variable (reduce memory usage by 1/10).
David Given [Tue, 4 Oct 2016 19:58:31 +0000 (21:58 +0200)]
Bodge in enough phi support to let the instruction generator complete on basic
programs.
David Given [Tue, 4 Oct 2016 19:32:28 +0000 (21:32 +0200)]
Oops --- hadn't updated the nts array for the new child order.
David Given [Tue, 4 Oct 2016 19:29:03 +0000 (21:29 +0200)]
Don't allocate new vregs for REG and NOP --- a bit hacky, but suppresses stray
movs very effectively.
David Given [Mon, 3 Oct 2016 22:16:06 +0000 (00:16 +0200)]
Massive rewrite of how emitters and the instruction selector works, after I
realised that the existing approach wasn't working. Now, hopefully, tracks the
instruction trees generated during selection properly.
David Given [Mon, 3 Oct 2016 18:52:36 +0000 (20:52 +0200)]
Instruction selection now happens on a shadow tree, rather than on the IR tree
itself. Currently it's semantically the same but the implementation is cleaner.
David Given [Sun, 2 Oct 2016 21:25:54 +0000 (23:25 +0200)]
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).
David Given [Sun, 2 Oct 2016 19:51:25 +0000 (21:51 +0200)]
Come up with a syntax for register constraints.
David Given [Sun, 2 Oct 2016 15:50:34 +0000 (17:50 +0200)]
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).
David Given [Sun, 2 Oct 2016 15:24:31 +0000 (17:24 +0200)]
Add array push/pop; fix ghastly memory overrun bug.
David Given [Sun, 2 Oct 2016 14:08:46 +0000 (16:08 +0200)]
Add zrl opcode. Keep track of local sizes as well as offsets.
David Given [Sun, 2 Oct 2016 12:44:21 +0000 (14:44 +0200)]
Add inl and del opcodes.
David Given [Sat, 1 Oct 2016 22:30:33 +0000 (00:30 +0200)]
Refactor mcg.h as it's getting a bit big; keep track of register variables.
David Given [Sat, 1 Oct 2016 21:41:45 +0000 (23:41 +0200)]
Make betterer.
David Given [Sat, 1 Oct 2016 21:41:35 +0000 (23:41 +0200)]
Fix edge case where leftover pushes would occasionally cause infinite loops in
the analysis.
David Given [Sat, 1 Oct 2016 21:41:03 +0000 (23:41 +0200)]
Show expression trees correctly.
David Given [Sat, 1 Oct 2016 21:13:39 +0000 (23:13 +0200)]
Deal with malformed mes instructions emitted by ego.
David Given [Sat, 1 Oct 2016 20:58:29 +0000 (22:58 +0200)]
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).
David Given [Sat, 1 Oct 2016 17:10:22 +0000 (19:10 +0200)]
We're going to need some type inference after all, I think. Let's do a little
for now and see how it goes.
David Given [Sat, 1 Oct 2016 11:56:52 +0000 (13:56 +0200)]
Predicates work; we now have prefers and requires clauses. Predicates must be
functions. Not convinced that semantic types are actually working --- there are
still problems with earlier statements leaving things in the wrong registers.
David Given [Sat, 1 Oct 2016 10:17:14 +0000 (12:17 +0200)]
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.
David Given [Fri, 30 Sep 2016 17:10:30 +0000 (19:10 +0200)]
Move fatal(), warning() and aprintf() into the new data module (because they're
really useful).
David Given [Thu, 29 Sep 2016 20:32:43 +0000 (22:32 +0200)]
More properly keep track of register classes.
David Given [Thu, 29 Sep 2016 20:14:11 +0000 (22:14 +0200)]
Error check fragment rules which don't emit anything.
David Given [Thu, 29 Sep 2016 20:06:04 +0000 (22:06 +0200)]
Much better handling of fragments (no run-time code needed to distinguish them
from registers) and better handling of individual hops within a paragraph ---
no more ghastly hacks to try and distinguish the input from the output.
David Given [Thu, 29 Sep 2016 17:58:02 +0000 (19:58 +0200)]
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.
David Given [Wed, 28 Sep 2016 21:39:00 +0000 (23:39 +0200)]
You can tell whether an item already exists in the array when calling
array_appendu() now.
David Given [Tue, 27 Sep 2016 21:38:47 +0000 (23:38 +0200)]
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.
David Given [Mon, 26 Sep 2016 22:19:45 +0000 (00:19 +0200)]
Fleshed out hops and vregs. The result is almost looking like code now ---
uncanny.
David Given [Mon, 26 Sep 2016 21:03:04 +0000 (23:03 +0200)]
Move map into the data module.
David Given [Mon, 26 Sep 2016 20:48:58 +0000 (22:48 +0200)]
Revamp the array module not to use nasty macros any more. Slightly more verbose
to use, but definitely cleaner.
David Given [Mon, 26 Sep 2016 20:24:49 +0000 (22:24 +0200)]
Move the array library into the data module.
David Given [Mon, 26 Sep 2016 20:12:46 +0000 (22:12 +0200)]
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.
David Given [Sun, 25 Sep 2016 21:29:59 +0000 (23:29 +0200)]
Start factoring out the hardware op code.
David Given [Sun, 25 Sep 2016 20:17:14 +0000 (22:17 +0200)]
Sort of keep track of registers and register classes. Start walking the
generated instruction tree --- holy cow, they look like instructions!
David Given [Sun, 25 Sep 2016 15:14:54 +0000 (17:14 +0200)]
Collapse several rule arrays into one; actually generate the array properly.
David Given [Sun, 25 Sep 2016 10:29:03 +0000 (12:29 +0200)]
Rename struct insn to struct em (throughout).
David Given [Sun, 25 Sep 2016 10:18:39 +0000 (12:18 +0200)]
Stupid stringlist is stupid. Use a proper data structure, properly abstracted
out (so other things can use it).
David Given [Sun, 25 Sep 2016 09:49:51 +0000 (11:49 +0200)]
Rethink the way patterns are mapped to rules; generate emitters (probably
badly).
David Given [Sat, 24 Sep 2016 22:21:46 +0000 (00:21 +0200)]
We now record the code fragments to be emitted by each rule.
David Given [Sat, 24 Sep 2016 20:50:53 +0000 (22:50 +0200)]
Instruction selection is so important the file needs a longer name.
David Given [Sat, 24 Sep 2016 20:46:08 +0000 (22:46 +0200)]
Some instruction selection is now happening.
David Given [Sat, 24 Sep 2016 17:03:55 +0000 (19:03 +0200)]
More grammar changes.
David Given [Sat, 24 Sep 2016 16:31:35 +0000 (18:31 +0200)]
IR codes are now owned by mcgg; ir terminals are inserted into the table during
compilation (so you can refer to them).
David Given [Sat, 24 Sep 2016 15:20:40 +0000 (17:20 +0200)]
Crudely bolt on mcgg to mcg itself.
David Given [Sat, 24 Sep 2016 14:59:49 +0000 (16:59 +0200)]
Parameters are parsed with getopt. Simplify, constify.
David Given [Sat, 24 Sep 2016 11:33:59 +0000 (13:33 +0200)]
Change the predicate stuff to use costs instead; now you can use when clauses
on leaves. Remove an iburg premature optimisation (required for above).
David Given [Sat, 24 Sep 2016 11:08:17 +0000 (13:08 +0200)]
Lots of exploratory new grammar for instruction definitions and string and
fragment emission (none of which is hooked up to anything yet).
David Given [Sat, 24 Sep 2016 10:11:30 +0000 (12:11 +0200)]
Add support for labelled tree nodes.
David Given [Fri, 23 Sep 2016 23:09:32 +0000 (01:09 +0200)]
Fix some late-night typo bugs.
David Given [Fri, 23 Sep 2016 23:04:00 +0000 (01:04 +0200)]
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.
David Given [Fri, 23 Sep 2016 21:59:15 +0000 (23:59 +0200)]
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.
David Given [Fri, 23 Sep 2016 19:07:16 +0000 (21:07 +0200)]
Each pass now lives in its own source file; much cleaner.
David Given [Thu, 22 Sep 2016 21:19:29 +0000 (23:19 +0200)]
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.
David Given [Wed, 21 Sep 2016 22:15:48 +0000 (00:15 +0200)]
Tracing cleanup. Simplified the IR code. Some more opcodes.
David Given [Tue, 20 Sep 2016 22:43:10 +0000 (00:43 +0200)]
Massive grammar overhaul and refactor. Hacked in support for predicates, where
instructions can be turned on and off based on their parameters. New lexer
using a lexer. Now quite a lot of the way towards being a real instruction
selector.
David Given [Tue, 20 Sep 2016 19:00:16 +0000 (21:00 +0200)]
Run through clang-format.
David Given [Tue, 20 Sep 2016 18:46:45 +0000 (20:46 +0200)]
Strip out surplus files. Rewrite README.
David Given [Tue, 20 Sep 2016 18:44:51 +0000 (20:44 +0200)]
Import iburg.
David Given [Tue, 20 Sep 2016 18:37:16 +0000 (20:37 +0200)]
Import iburg.
David Given [Mon, 19 Sep 2016 22:19:39 +0000 (00:19 +0200)]
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.