RJ Swedlow
|
37d6648dc5
|
more info detailed on the divide table
|
2013-04-11 18:46:08 -05:00 |
|
RJ Swedlow
|
0db41038ad
|
forwards compatibility when compiling as C++
|
2013-04-10 19:06:04 -05:00 |
|
RJ Swedlow
|
11a7ece6ee
|
finished back-porting perfect FP vector divide precision by MAME
|
2013-04-10 01:06:59 -05:00 |
|
RJ Swedlow
|
d745ab4143
|
umm on second thought doesn't really hurt if I do it this way
|
2013-04-09 00:19:58 -05:00 |
|
RJ Swedlow
|
46e9bef4d1
|
test ROM lookup fixing 11-bit-FP-precise vector divides
|
2013-04-09 00:08:54 -05:00 |
|
RJ Swedlow
|
d7e768d0c7
|
alternative mechanism to hack Intel negate operation
|
2013-04-08 23:56:57 -05:00 |
|
RJ Swedlow
|
e0016f5c65
|
cut to one-dimensional branch trees via arithmetic clamping
|
2013-04-06 22:20:34 -05:00 |
|
RJ Swedlow
|
e5f3f4d911
|
beta implementation of the rare VMACQ opcode
|
2013-04-04 00:57:50 -05:00 |
|
RJ Swedlow
|
18af0cf45c
|
faster, direct unsigned clamping (F3DZEX and F3DEX)
|
2013-03-28 22:54:17 -05:00 |
|
RJ Swedlow
|
412bc6f93d
|
more reliable 64-bit accumulator superset to fetch sign
|
2013-03-28 22:46:19 -05:00 |
|
RJ Swedlow
|
4d19a56d8b
|
patched potential security glitch on accumulator overflow
|
2013-03-26 13:17:07 -05:00 |
|
RJ Swedlow
|
23c9c8ffd9
|
handle a RSP corner case when negating -32768
|
2013-03-25 18:38:15 -05:00 |
|
RJ Swedlow
|
43bea1275c
|
added some support and info on reserved ops
|
2013-03-24 10:50:26 -05:00 |
|
RJ Swedlow
|
8220bdde3c
|
NEW: cut branch hits with 16-bit slice-spanning across bytes
|
2013-03-22 14:43:38 -05:00 |
|
RJ Swedlow
|
fecc3c37ed
|
deprecated control VU_OVERRIDE_WEIRD_ELEMENT
|
2013-03-21 10:04:12 -05:00 |
|
RJ Swedlow
|
6d31df178e
|
static condition codes for unsigned clamp
|
2013-03-21 10:02:14 -05:00 |
|
RJ Swedlow
|
c768ab4d4e
|
finished collapsing scalar element decoder mess into look-up
|
2013-03-20 22:43:38 -05:00 |
|
RJ Swedlow
|
5c72eb8015
|
fixed corrupt scalar element decoder when VS ^ VT < 0
|
2013-03-20 22:42:29 -05:00 |
|
RJ Swedlow
|
c2e5d0f21f
|
restored maintenance template to fast-decode scalar elements
|
2013-03-20 22:42:12 -05:00 |
|
RJ Swedlow
|
ad3f780817
|
major VU rewrite: destroyed artificial union index (arrays for real vectors)
|
2013-03-20 17:42:54 -05:00 |
|
RJ Swedlow
|
d602acbe45
|
temporary overload fix for C++ preprocessor on old FP model
|
2013-03-05 12:20:52 -06:00 |
|
RJ Swedlow
|
bf01b28b59
|
static single-case clamp
|
2013-03-04 19:17:59 -06:00 |
|
RJ Swedlow
|
259f5ecf75
|
consistency for template match-ups on remaining ops + faster VMRG
|
2013-02-24 23:20:42 -06:00 |
|
RJ Swedlow
|
069c293506
|
linear, branch-cut typical compare selects
|
2013-02-24 20:23:21 -06:00 |
|
RJ Swedlow
|
5a5ae47889
|
quick simplify computational logical vector template
|
2013-02-23 23:16:49 -06:00 |
|
RJ Swedlow
|
c125163163
|
revoke const table attribute to fix GNU/ANSI C warning
|
2013-02-23 22:58:14 -06:00 |
|
RJ Swedlow
|
87de0b6b1e
|
cut branches in vector computational adds (more static)
|
2013-02-23 22:57:20 -06:00 |
|
RJ Swedlow
|
c06cd5f1ac
|
alphebetize header files per vector instruction
|
2013-02-22 23:57:41 -06:00 |
|
RJ Swedlow
|
4de51ad13a
|
allocate signed 48-bit register sub-string access for x86_64
|
2013-02-22 23:45:52 -06:00 |
|
RJ Swedlow
|
7c29a1acc3
|
simplified slice buffer management for multiply-accumulates
|
2013-02-22 23:44:45 -06:00 |
|
RJ Swedlow
|
297c244451
|
faster 32-bit access for checking slice overflow
|
2013-02-22 20:37:55 -06:00 |
|
RJ Swedlow
|
0ba5935eac
|
decompress function pointer type conversion to const void methods
|
2013-02-22 20:11:50 -06:00 |
|
RJ Swedlow
|
daa7d7b156
|
cleanup excess clamping code and vector element multiply/shifts
|
2013-02-22 20:02:48 -06:00 |
|
RJ Swedlow
|
87b6922279
|
use a constant table of pointers to constant functions
|
2013-02-22 11:12:47 -06:00 |
|
RJ Swedlow
|
73df38b684
|
quick (but critical!!) rewrites to RSP divide gateway structure
|
2013-02-21 21:06:31 -06:00 |
|
RJ Swedlow
|
ecdb56667b
|
fix bad accumulator source buffer reads (SM64 weird vs mask)
|
2013-02-20 22:34:53 -06:00 |
|
RJ Swedlow
|
8e0e360059
|
help template for inline loop unrolling
|
2013-02-19 13:59:10 -06:00 |
|
RJ Swedlow
|
4416c2de6f
|
confirmed accumulator write-back ammendment for divides
|
2013-02-17 22:56:13 -06:00 |
|
RJ Swedlow
|
8a98d1f3b4
|
expanded hacked algorithm into static field decoders
|
2013-02-16 21:49:18 -06:00 |
|
RJ Swedlow
|
01650478a5
|
merge gateway static check path inversion into a single check
|
2013-02-15 12:37:14 -06:00 |
|
RJ Swedlow
|
840cc8bc4c
|
more static set-on-condition, fixed possible emu bug holding VCO mask
|
2013-02-14 20:59:46 -06:00 |
|
RJ Swedlow
|
651b8dd576
|
fix Resident Evil prologue cinnema (VCC set should be cond.)
|
2013-02-13 23:29:43 -06:00 |
|
RJ Swedlow
|
f6b20b2d9a
|
zero-based comparison is more direct
|
2013-02-13 20:33:08 -06:00 |
|
RJ Swedlow
|
e3edd268eb
|
fixed edge condition where vs + vt was -1, more static flow
|
2013-02-13 18:42:37 -06:00 |
|
RJ Swedlow
|
a8521a1bc4
|
expanded element decoder with upgraded NOT on dual-bit-check
|
2013-02-13 12:16:18 -06:00 |
|
RJ Swedlow
|
86448f9325
|
fix wrong changed register indirection target (read VCO, not VCC)
|
2013-02-12 17:54:22 -06:00 |
|
RJ Swedlow
|
bd759b08d0
|
expand element decoding and rewrote notes on accuracy
|
2013-02-12 12:14:32 -06:00 |
|
RJ Swedlow
|
7606d2b89c
|
vector condition flags array split into direct reg control
|
2013-02-11 22:40:28 -06:00 |
|
RJ Swedlow
|
b901b55e33
|
simplified slice element decoder
|
2013-02-10 20:12:55 -06:00 |
|
RJ Swedlow
|
4294cefe2e
|
rewrote vector accumulator access modes
|
2013-02-10 17:00:57 -06:00 |
|