Commit graph

132 commits

Author SHA1 Message Date
Bill Currie
3d8ee5df43 [qfcc] Ensure ops on globals occur before return
This fixes the return-postop test, and covers calls, too.
2022-02-03 16:33:42 +09:00
Bill Currie
a64f91129f [qfcc] Give lea its own statement type
This makes it much easier to check (and more robust to name changes),
allowing for effectively killing the node to which the variable being
addressed is attached. This fixes the incorrect address being used for
va_list, which is what caused double-alias to fail.
2022-02-02 18:55:01 +09:00
Bill Currie
93840d9892 [qfcc] Handle Ruamoko's call return destination
Since the call instruction in the Ruamoko ISA specifies the destination
of the return value of the called function, it is much like any
expression type instruction in that the def referenced by its c operand
is both defined and killed by the instruction. However, unlike other
instructions, it really has many pseudo-operands: the arguments placed
on the stack. The problem is that when one of the arguments is also the
destination of the return value, the dags code wants to use the stack
argument as it was the last use of the real argument. Thus, instead of
using the value of the child node for the result, use the value label
attached to the call node (there should be only one such label).

This fixes iterfunc, typedef, zerolinker and vkgen when optimizing. Now
all but the double tests and return postop tests pass (and the retun
postop test is not related to the Ruamoko ISA, so fails either way).
2022-02-01 21:46:28 +09:00
Bill Currie
e195dba626 [qfcc] Do not unalias temporary operands
I don't remember why I did this originally, but it causes the dags code
to lose the offset temp alias when accessing fields on structural temps
(known to be the case for vectors (temp-component.r), and I seem to
remember having problems with structs).
2022-01-30 14:02:13 +09:00
Bill Currie
5cf7924352 [qfcc] Improve dags handling of auxiliary use ops
The aux use ops need to be counted and given nodes explicitly as they
may refer to defs that are not accessed by other statements other than
by aliases, and those aliases need to be marked live as well as the used
def.
2022-01-24 16:55:55 +09:00
Bill Currie
616a52efb5 [qfcc] Implement flow analysis for Ruamoko calls
Thanks to the use/def/kill lists attached to statements for pseudo-ops,
it turned out to be a lot easier to implement flow analysis (and thus
dags processing) than I expected. I suspect I should go back and make
the old call code use them too, and probably several other places, as
that will greatly simplify the edge setting.
2022-01-21 17:14:10 +09:00
Bill Currie
2b7a8387e7 [qfcc] Get void return statements working
My little test program now builds with the Ruamoko ISA :)

    void cp (int *dst, int *src, int count)
    {
	while (count--) {
	    *dst++ = *src++;
	}
    }

Calls are broken (unimplemented), and non-void returns are not likely to
work either (only partially implemented).
2022-01-20 18:40:55 +09:00
Bill Currie
a5a8017220 [qfcc] Switch internal statement format to ruamoko
For the most part, it wasn't too bad as it's just a rotation of the
operands for some instructions (store, assign, branch), but dealing with
all the direct accesses to specific operands was a small pain. I am very
glad I made all those automated tests :)
2022-01-20 13:08:05 +09:00
Bill Currie
a4ebd6aa58 [gamecode] Fix a few missed opcode renames
if and ifnot became ifnz and ifz, and return_v lost its tail (it was
always redundant, except in dags, and that's fixed with a pointer check).
2022-01-20 13:07:57 +09:00
Bill Currie
143030fec4 [gamecode] Use text for all v6p opcode names
This makes the v6p instruction table consistent with the ruamoko
instruction table, and clears up some of the ugliness with the load,
store, and assign instructions (. .= and = are now spelled out). I think
I'd still prefer an enum code (faster) but at least this is more
readable.
2022-01-20 09:26:01 +09:00
Bill Currie
afd1eb775b [gamecode] Rename ev_pointer to ev_ptr
Rather short (no worse than ev_int, though) but more consistency is
usually a good thing.
2022-01-18 14:36:06 +09:00
Bill Currie
44c48feb56 [qfcc] Add attached node's parents to source edges
The assignment to the node's variable must come after any uses of that
node, which the node's parent set indicates. In the swap test, this was
not a problem as the node had no parents, and in the link order test, it
just happened(?) to work.
2021-12-28 09:23:14 +09:00
Bill Currie
8433905015 [qfcc] Extend reachable to the label node's parents
While using just the label node's reachable set was sufficient for a
simple swap (t = a; a = b; b = t;), it is not sufficient for
read-before-write dependencies such as found in linked-list building:

    { o = array[ind]; o.next = obj; obj = o; }

The assignment to o.next uses obj, but that use is hidden because obj's
reachable nodes does not include o thus assigning o to obj causes the
array dereference to be assigned directly to obj and thus o.next winds
up pointing to o instead of whatever obj was. The parent nodes of obj's
node are its users, so any new assigned to obj must come after those
parents as well as any node reachable by obj's node.

Fixes a runaway loop error when adding a frikbot to the server.
2021-12-27 14:25:55 +09:00
Bill Currie
f903211362 [qfcc] Make operand union anonymous
This one was easy enough and gets rid of that little o. wart.
2021-12-25 12:40:24 +09:00
Bill Currie
a3aebc983b [qfcc] Add dependency edges for moved labels
When moving an identifier label from one node to another, the first node
must be evaluated before the second node, which the edge guarantees.
However, code for swapping two variables

    t = a; a = b; b = t;

creates a dependency cycle. The solution is to create a new leaf node
for the source operand of the assignment. This fixes the swap.r test
without pessimizing postop code.

This takes care of the core problem in #3, but there is still room for
improvement in that the load/store can be combined into a move.
2021-06-29 14:42:16 +09:00
Bill Currie
76b3bedb72 [qfcc] Revert "Kill dag leaf nodes on assignment."
This reverts commit 2fcda44ab0.

Killing the node is not the correcgt answer as it blocks many
optimization opportunities. The correct answer is adding edges to
describe the temporal dependencies. Of course, this breaks the swap.r
test.
2021-06-29 12:09:35 +09:00
Bill Currie
5291cfb03d [qfcc] Keep track of reachable dag nodes
In order to correctly handle swap-style code

    { t = a; a = b; b = t; }

edges need to be created for each of the assignments moving an
identifier lable, but the dag must remain acyclic (the above example
wants to create a cycle). Having the reachable nodes recorded makes
checking for potential loops a quick operation.
2021-06-29 09:41:03 +09:00
Bill Currie
0d1fad12f0 [qfcc] Add some comments 2021-06-29 09:12:57 +09:00
Bill Currie
d50f2c3145 [qfcc] Correctly check for constant op
Identifiers can be constants. I don't remember quite what it fixed other
than some bogus kill relations in the dags (which might have caused
issues later).
2021-06-28 20:25:26 +09:00
Bill Currie
0c1699fdbb Fix a pile of double semicolons 2021-01-09 20:42:23 +09:00
Bill Currie
6d5ffa9f8e [build] Move to non-recursive make
There's still some cleanup to do, but everything seems to be working
nicely: `make -j` works, `make distcheck` passes. There is probably
plenty of bitrot in the package directories (RPM, debian), though.

The vc project files have been removed since those versions are way out
of date and quakeforge is pretty much dependent on gcc now anyway.

Most of the old Makefile.am files  are now Makemodule.am.  This should
allow for new Makefile.am files that allow local building (to be added
on an as-needed bases).  The current remaining Makefile.am files are for
standalone sub-projects.a

The installable bins are currently built in the top-level build
directory. This may change if the clutter gets to be too much.

While this does make a noticeable difference in build times, the main
reason for the switch was to take care of the growing dependency issues:
now it's possible to build tools for code generation (eg, using qfcc and
ruamoko programs for code-gen).
2020-06-25 11:35:37 +09:00
Bill Currie
17bb408b57 [qfcc] Clean up stray dags edges
Killed nodes can leave stray (dangling) edges that cause some confusion
in the dot graphs and may cause problems later on down the track, so
ensure there are no dangling edges.
2020-04-01 16:06:45 +09:00
Bill Currie
199b3e82d9 [qfcc] Add killer node to replacement node's edges
When an operand refers to a killed node, it needs to be evaluated AFTER
the killing node is evaluated. This fixes the double-alias bug.
2020-04-01 16:05:26 +09:00
Bill Currie
4c79c9ffaf [qfcc] Do a dags pre-pass to give operands leafs
The reason double-alias fails is when the double assignment occurs, the
int operands don't yet have leaf nodes and thus the nodes cannot be
killed. This doesn't fix the bug.
2020-04-01 16:03:12 +09:00
Bill Currie
0293d335d0 [qfcc] Mark known source def live for movep
This fixes ivar-struct-return (and qwaq).
2020-03-18 00:02:36 +09:00
Bill Currie
578bf9a16f [qfcc] Set dag node value for movep
This fixes compilation of all tests. However, structptr still fails.
2020-03-17 22:54:27 +09:00
Bill Currie
77806f4b1b [qfcc] Get dag code generation mostly working
There's an ICE in return-ivar, but assignchain passes let alone builds.
2020-03-17 22:35:36 +09:00
Bill Currie
16bda66785 [qfcc] Add more statement types for move/memset
They ease the statement checks between assign/move/memset and the
pointer versions (don't need all those strcmps)
2020-03-17 21:39:49 +09:00
Bill Currie
dec2e6249e [qfcc] Increase flow operand count to 5
MOVEP instructions have up to 5 operands: 2 pointers, the count, and 0-2
referenced variables (when known).
2020-03-17 21:24:12 +09:00
Bill Currie
a0c28a5ac5 [qfcc] Support pointers to temp operands
This is necessary for correctly taking the address of operands.
2020-03-16 14:24:47 +09:00
Bill Currie
2f07d9a310 [qfcc] Improve accuracy of some more diagnostics 2020-03-16 10:42:18 +09:00
Bill Currie
51a30de9c5 [qfcc] Print accurate linenos for more ICEs 2020-03-14 16:51:54 +09:00
Bill Currie
4a8854d9ed [qfcc] Add expression tracking to operands
Not much uses it yet, but it will make for better diagnostics.
2020-03-11 12:51:34 +09:00
Bill Currie
bcf75b541a [qfcc] Build movep dest pointer correctly
This fixes the mangled pointer in struct-init-param.r.
2020-03-08 17:40:38 +09:00
Bill Currie
b996fb7aa4 Make operand->type actual type instead of low-level
And clean up the resulting mess. This fixes struct copy, but uncovers
another bug :/
2019-06-17 23:38:34 +09:00
Bill Currie
d6d3027411 Mark the correct operand as live
This fixes vecexpr (and possibly other cases).
2019-06-16 19:21:02 +09:00
Bill Currie
6e21c3ae2e Treat func statements similarly to flow statements
func statements need their operands marked live like flow statements do
because usage is more indirect.
2019-06-16 19:20:21 +09:00
Bill Currie
db4a7a139e Use the alias code when making vars live
Not sure the live forcing flag is needed anymore (need to test).
2019-06-16 19:17:45 +09:00
Bill Currie
2977c145d0 Clean up dag live alias code a little
Mainly, this makes it possible to reuse the alias code.
2019-06-16 19:17:01 +09:00
Bill Currie
c40f4194e9 Use tempop_visit_all for flow and dags
Fixes t3 of vecexpr, but t2 is broken (lost first assignment).
2019-06-16 16:56:39 +09:00
Bill Currie
3c4903245a Fix some curly space 2019-06-16 16:55:54 +09:00
Bill Currie
fa69aeef0f Improve handling of temp aliases
This makes all tests pass when not optimizing. More work needs to be
done in dags.
2019-06-12 00:37:02 +09:00
Bill Currie
cc27949a34 Ensure pointer values always have a type
The dags generator was creating a pointer value with no type which
caused print_statement to segfault.
2019-06-10 23:52:39 +09:00
Bill Currie
78e0a8dc52 Support assigning non-constant vector expressions. 2018-10-12 22:05:17 +09:00
Bill Currie
34bcf7faab Do a pure/const/noreturn/format attribute pass.
I always wanted these, but as gcc now provides warnings for functions that
could do with such attributes, finding all the functions is much easier.
2018-10-09 12:42:21 +09:00
Bill Currie
2fcda44ab0 Kill dag leaf nodes on assignment.
Fixes the failing swap test caused by a's original value being used (via t)
after being written.
2018-08-23 20:07:22 +09:00
Bill Currie
ca0b03687f Change the naming of ALLOC's free-list.
Rather than prefixing free_ to the supplied name, suffix _freelist to the
supplied name. The biggest advantage of this is it allows the free-list to
be a structure member. It also cleans up the name-space a little.
2013-03-08 22:16:31 +09:00
Bill Currie
e27d7cbd2d Handle alloca "correctly".
Use AC_FUNC_ALLOCA and the #ifdef mess suggested by the autoconf docs
(hidden in qfalloca.h).
2013-01-22 21:02:50 +09:00
Bill Currie
412fb4aacd Fix a static out of bounds error.
I really must compile with optimizations more often :P
2012-12-26 11:51:40 +09:00
Bill Currie
386a729209 Remove dead nodes from dags.
Dead nodes are those that generate unused values (unassigned leaf nodes,
expressions or destinationless move(p) nodes). The revoval is done by the
flow analysis code (via the dags code) so that any pre and post removal
flow analysis and manipulation may be done (eg, available expressions).
2012-12-25 17:07:58 +09:00