76be7b54ed
compiler.cfg.linear-scan.*: removed start and end from live-interval-state
...
They weren't needed because start>> was always equal to ranges min and
end>> ranges max
2016-04-01 19:16:35 +02:00
fb45bc278f
compiler.cfg.linear-scan.live-intervals: the reg-class slot can be
...
removed
it's always the same as vreg>> rep-of reg-class-of so no need to store
that data twice
2015-09-22 08:51:04 +02:00
68c6654763
compiler.cfg.linear-scan.*: move words for splitting and intersecting ranges to the compiler.cfg.linear-scan.ranges vocab
2015-09-22 08:51:04 +02:00
8c8605d11a
compiler.cfg.linear-scan.*: more tests and docs for allocation and state
2015-06-16 18:28:32 -07:00
78fe8bf1dc
compiler.cfg.linear-scan.*: dont use a hashtable here, it fixes register
...
allocation so it's always in the same order as machine-registers
2015-06-16 18:28:32 -07:00
c0645ecafe
compiler.cfg.linear-scan.*: fix help-lint errors
2015-06-14 16:44:37 +02:00
418438ab13
compiler.cfg.linear-scan.*: change compute-live-intervals so that it
...
returns only one sequence, it makes some words a little simpler
2015-06-14 16:44:37 +02:00
167d1598b3
compiler.cfg.linear-scan.allocation.*: lots more docs and some word simplifications
2015-04-29 09:31:58 -07:00
d6c1c075ff
compiler.cfg.linear-scan.allocation: using slurp-heap
2014-12-15 20:48:34 +01:00
35d00da765
compiler.cfg.linear-scan.allocation.*: maybe a fix for #1227
...
The register allocator pops from the unhandled-intervals and
unhandled-sync-points heaps. If two live intervals share the same
start slot, then they can be popped in either order leading to
differently generated code. So instead use one
heap (unhandled-min-heap) and both the start and end slot as the heap
key to resolve tie-breakers
2014-12-15 20:44:53 +01:00
c810677b47
compiler.cfg.linear-scan.*: refactoring to simplify init-allocator and get rid of init-unhandled
2014-12-15 20:44:06 +01:00
fab478482e
compiler.cfg.linear-scan.allocation: refactoring to simplify the (allocate-registers) recursion
2014-12-15 20:44:06 +01:00
c65877104e
compiler: cleanup usings.
2014-12-13 16:10:21 -08:00
5f377c0849
compiler: Rename value-info, live-interval
2013-03-23 22:56:47 -07:00
94f6ac98de
compiler.cfg.linear-scan: fix bad interaction between inactive intervals and sync points
2010-07-29 21:06:48 -04:00
adc9db3ea8
compiler.cfg.linear-scan.allocation: fix wrong order
2010-07-19 06:46:50 -04:00
b776a92546
compiler.cfg.linear-scan: fix bad interaction between split position calculation and register-clobbering instructions
2010-07-18 17:38:52 -04:00
e27adb2830
compiler: re-architect low-level optimizer to allow more than one output value per instruction
2010-07-13 07:40:14 -04:00
5b48cd2a63
FFI rewrite part 5: return value boxing and callback parameter boxing now uses vregs; simplify return value unboxing
2010-05-16 03:43:23 -04:00
eb802208d1
FFI rewrite part 4: parameter and return value unboxing redesign
2010-05-11 19:11:44 -04:00
982e704626
compiler.cfg.linear-scan: clean up clobber-insn handling
2010-05-07 18:22:35 -04:00
43fc230c69
compiler.cfg.linear-scan: cleanups
2010-05-03 17:34:18 -04:00
95ff5ffe51
New GC checks work in progress
2010-05-03 17:34:16 -04:00
51b13ce373
compiler.cfg.linear-scan: don't insert a _reload if the register is going to be overwritten anyway
2010-05-03 17:34:14 -04:00
299b5b0f6c
filter-here -> filter!
2009-10-28 00:44:05 -05:00
705b4ab5c3
compiler.cfg.linear-scan: fix partial sync point logic in case where dst == src, and clean up spilling code
2009-09-27 19:28:20 -05:00
0db01f6d5f
compiler.cfg.linear-scan now supports partial sync-points where all registers are spilled; taking advantage of this, there are new trigonometric intrinsics which yield a 2x performance boost on benchmark.struct-arrays and a 25% boost on benchmark.partial-sums
2009-08-30 04:52:01 -05:00
370f4c081d
compiler.cfg: convert code into two-operand form before SSA destruction; SSA destruction now operates on a relaxed SSA form where multiple defs of the same vreg are allowed, but only within a single basic block. This makes linear scan's coalescing redundant, allowing it to be removed completely
2009-08-05 18:57:46 -05:00
791fbe4003
compiler.cfg.linear-scan: fix case where a register can be made available for only a part of a live interval's lifetime, but there are no more usages after the split location. This case never came up until global stack analysis, at which point it started to be exercised on x86-32
2009-07-30 02:13:30 -05:00
5a64290386
compiler.cfg.linear-scan: more code cleanups, and working on split-to-fit algorithm
2009-07-09 00:20:03 -05:00
4507bdcbc0
compiler.cfg.linear-scan: code cleanups
2009-07-08 23:28:28 -05:00
0bcf4ce535
compiler.cfg.linear-scan: Re-implement spilling, add some additional runtime assertions, simplify assignment pass since it doesn't have to track spill slots anymore; just assume a live value that's not in active or inactive is spilled
2009-07-07 03:28:55 -05:00
a19f22ec82
compiler.cfg.linear-scan: debugging spilling and resolve pass
2009-07-01 17:41:07 -05:00
578b0126fc
compiler.cfg.linear-scan: compute-free-pos was broken in the case where more than one inactive interval had the same physical register assigned
2009-06-30 16:07:58 -05:00
c53aca6016
compiler.cfg.linear-scan: fix coalescing to take lifetime holes into account
2009-06-23 21:32:51 -05:00
5c912504d7
compiler.cfg.linear-scan: untangle add-active/delete-active/add-handled calls in spilling, replace 'sort-values last' with 'alist-max' in compiler.utilities
2009-06-19 18:28:30 -05:00
2f15ac3c8f
compiler.cfg.linear-scan: Fix a couple of bugs
2009-06-19 03:42:42 -05:00
1f5f8393c3
compiler.cfg.linear-scan: Further progress on inactive interval handling
2009-06-17 15:27:20 -05:00
d0f6a7d048
Split up compiler.cfg.linear-scan.allocation into a number of sub-vocabularies; start work on compiler.cfg.linear-scan.resolve; start work on inactive interval splitting
2009-06-11 17:55:14 -05:00
70b75b8dea
Linear scan: implement live range intersection
2009-06-06 01:04:12 -05:00
04c72d0a72
Linear scan: correctly compute live spill slots
2009-06-05 18:06:47 -05:00
81b2a39070
compiler.cfg.linear-scan: re-do interval splitting to operate on live ranges; add inactive set processing
2009-06-04 18:53:02 -05:00
37e278ed02
Use [ ] [ ] map-reduce instead of unclip [ ] reduce
2009-04-06 17:50:54 -05:00
669548e62e
Clean up some usages of tuck, and add Joe's curried cleave/spread/apply combinators to kernel vocabulary
2009-02-02 13:43:54 -06:00
1af3b8ed65
If a register is spilled and the reload location is also a copy, we chicken out and don't coalesce. This is suboptimal but it's not clear to me how to implement it cleanly, and SSA graph coloring will solve this problem eventually anyway
2008-11-04 00:32:56 -06:00
8df1aba71d
Coalescing
2008-11-02 01:49:57 -06:00
14d8696f40
Oops, don't mix register classes in active set
2008-10-19 03:34:42 -05:00
c0d89b061e
Fixing register allocator prspilling
2008-10-19 01:10:21 -05:00
cf46a832e7
Debugging register allocator and inline allocation
2008-10-08 23:42:53 -05:00
7b6d9c4c4f
Debugging new codegen
2008-10-07 20:00:38 -05:00
Björn Lindqvist
John Benediktsson
Doug Coleman
Slava Pestov
Slava Pestov
Slava Pestov