Dmitry Selyutin [Sat, 3 Sep 2022 13:16:05 +0000 (16:16 +0300)]
power_insn: drop pack/unpack bit
Dmitry Selyutin [Sat, 3 Sep 2022 11:16:30 +0000 (14:16 +0300)]
power_insn: rename normal field to simple
Dmitry Selyutin [Sat, 3 Sep 2022 11:08:47 +0000 (14:08 +0300)]
power_insn: decouple IMM/IDX LD/ST modes
Dmitry Selyutin [Sat, 3 Sep 2022 10:09:30 +0000 (13:09 +0300)]
power_insn: support operands check
Dmitry Selyutin [Sat, 3 Sep 2022 10:06:05 +0000 (13:06 +0300)]
power_insn: canonicalize Rc field name
Dmitry Selyutin [Fri, 2 Sep 2022 22:11:32 +0000 (01:11 +0300)]
power_insn: support mode selector
Dmitry Selyutin [Fri, 2 Sep 2022 22:04:48 +0000 (01:04 +0300)]
power_insn: support normal mode
Dmitry Selyutin [Fri, 2 Sep 2022 21:20:16 +0000 (00:20 +0300)]
power_insn: support LD/ST indexed mode
Dmitry Selyutin [Fri, 2 Sep 2022 21:17:39 +0000 (00:17 +0300)]
power_insn: simplify fields
Dmitry Selyutin [Fri, 2 Sep 2022 21:01:47 +0000 (00:01 +0300)]
power_insn: decrease LDSTMode class nesting
Dmitry Selyutin [Fri, 2 Sep 2022 20:46:10 +0000 (23:46 +0300)]
power_insn: support LD/ST immediate mode
Dmitry Selyutin [Fri, 2 Sep 2022 16:56:39 +0000 (19:56 +0300)]
power_insn: make RM class public
Dmitry Selyutin [Fri, 2 Sep 2022 16:34:21 +0000 (19:34 +0300)]
power_insn: remap RM immediately
Dmitry Selyutin [Fri, 2 Sep 2022 21:34:29 +0000 (00:34 +0300)]
power_fields: allow slicing mappings
Dmitry Selyutin [Fri, 2 Sep 2022 16:34:11 +0000 (19:34 +0300)]
power_fields: allow slicing fields
Dmitry Selyutin [Fri, 2 Sep 2022 15:04:32 +0000 (18:04 +0300)]
power_fields: create arrays from Array class
Dmitry Selyutin [Fri, 2 Sep 2022 21:48:01 +0000 (00:48 +0300)]
sv_binutils: shorten and simplify the output
Luke Kenneth Casson Leighton [Sat, 3 Sep 2022 18:38:53 +0000 (19:38 +0100)]
create list of opcodes by dict entry
Luke Kenneth Casson Leighton [Sat, 3 Sep 2022 18:32:34 +0000 (19:32 +0100)]
correct table header
Luke Kenneth Casson Leighton [Sat, 3 Sep 2022 18:30:08 +0000 (19:30 +0100)]
use opcode directly
Luke Kenneth Casson Leighton [Sat, 3 Sep 2022 18:07:39 +0000 (19:07 +0100)]
divpoint 2 to match v3.1
Luke Kenneth Casson Leighton [Sat, 3 Sep 2022 18:05:45 +0000 (19:05 +0100)]
create correct divpoint to make match against v3.0 Appendix C
Luke Kenneth Casson Leighton [Sat, 3 Sep 2022 17:43:29 +0000 (18:43 +0100)]
complete markdown table
Luke Kenneth Casson Leighton [Sat, 3 Sep 2022 17:00:54 +0000 (18:00 +0100)]
correct table-matching
Luke Kenneth Casson Leighton [Sat, 3 Sep 2022 16:48:09 +0000 (17:48 +0100)]
enumeration almost there
Luke Kenneth Casson Leighton [Sat, 3 Sep 2022 16:37:47 +0000 (17:37 +0100)]
MSB0-order, xomask 31-start
Luke Kenneth Casson Leighton [Sat, 3 Sep 2022 16:33:46 +0000 (17:33 +0100)]
add power_table.py start of creating markdown Appendix D tables
Luke Kenneth Casson Leighton [Sat, 3 Sep 2022 12:45:12 +0000 (13:45 +0100)]
add svshape2 offset test demonstrating RA being offset by one
Luke Kenneth Casson Leighton [Sat, 3 Sep 2022 12:13:55 +0000 (13:13 +0100)]
Revert "add inv option to svshape2 (only 1 bit)"
This reverts commit
77a4f7104968859385e0b8117ea74041cbe6e436.
the reason is that the inclusion of an inv bit can only be
done by reducing the range of "offset", from 0-15 to 0-7.
as this is the *only way* to get at elements on EXTRA2-encoding
it is considered "unwise"
Luke Kenneth Casson Leighton [Sat, 3 Sep 2022 11:59:06 +0000 (12:59 +0100)]
add inv option to svshape2 (only 1 bit)
https://bugs.libre-soc.org/show_bug.cgi?id=911
there is precious space: an sv.svshape2 can add more bits later
Luke Kenneth Casson Leighton [Sat, 3 Sep 2022 11:06:00 +0000 (12:06 +0100)]
update sv_analysis to create separate SVMode.LDST_IDX from SVMode.LDST_IMM
to help distinguish the two SVP64.RM LD/ST types
Jacob Lifshay [Sat, 3 Sep 2022 01:50:01 +0000 (18:50 -0700)]
fix test_caller_svshape2.py
Jacob Lifshay [Sat, 3 Sep 2022 01:44:08 +0000 (18:44 -0700)]
format code
Luke Kenneth Casson Leighton [Fri, 2 Sep 2022 20:46:53 +0000 (21:46 +0100)]
add test_caller_svshape2.py and make corrections to csv and fields.txt
yx needed to be SVM2-form and was in a different bitposition
offs needed to be 6:9 not 6:10
XO was off-by-one in minor_22.csv
Luke Kenneth Casson Leighton [Fri, 2 Sep 2022 17:44:22 +0000 (18:44 +0100)]
add first svshape2 pseudocode, based on svindex
it is near-identical but is actually back to svshape in terms of
the SHAPE bits set. bits 18-20 of SVSHAPEn are set in "Matrix" mode
not "Indexed" mode though.
Luke Kenneth Casson Leighton [Fri, 2 Sep 2022 17:41:52 +0000 (18:41 +0100)]
add svshape2 to ISACaller
first recognising the persistence mode bit, second as not an illegal op
Luke Kenneth Casson Leighton [Fri, 2 Sep 2022 15:33:01 +0000 (16:33 +0100)]
add svshape2 to sv/trans/svp64.py
https://bugs.libre-soc.org/show_bug.cgi?id=911
Luke Kenneth Casson Leighton [Fri, 2 Sep 2022 15:29:11 +0000 (16:29 +0100)]
add svshape2 to list of instructions in power_enums.py
Luke Kenneth Casson Leighton [Fri, 2 Sep 2022 15:25:52 +0000 (16:25 +0100)]
whoops bit 25 is sk not vf in svshape2. matches with svindex
Luke Kenneth Casson Leighton [Fri, 2 Sep 2022 15:16:47 +0000 (16:16 +0100)]
disallow reserved SVrm values in svshape, svp64.py
Luke Kenneth Casson Leighton [Fri, 2 Sep 2022 14:43:17 +0000 (15:43 +0100)]
add svshape2 (stub pseudocode) fields, Form, and CSV file minor_22.csv
https://bugs.libre-soc.org/show_bug.cgi?id=911
also added missing SVI-Form (svindex) SVd and rmm which are exactly
the same. svshape2 is weird, it is a hybrid of svindex and svshapes
Luke Kenneth Casson Leighton [Fri, 2 Sep 2022 14:40:59 +0000 (15:40 +0100)]
add explicit 13 patterns for svshape which make a hole for svshape2
svshape2 will take up 2 out of the 16 4-bit patterns so svshape has
to explicitly list them. this takes advantage of the new feature
added by ghostmansd: opcode merging, commit
d5ec553e768
Luke Kenneth Casson Leighton [Fri, 2 Sep 2022 14:37:50 +0000 (15:37 +0100)]
shuffle down numbering after SVM to make room for SVM2
power_enums.py Form needs to make space for SVM2
(and svshape is no longer temporary)
Luke Kenneth Casson Leighton [Fri, 2 Sep 2022 12:32:45 +0000 (13:32 +0100)]
add fix of out_sel in power_decoder.py formal proof
but there are others. basically the list of enums needs to be
automatically listed and put into the matches
Luke Kenneth Casson Leighton [Fri, 2 Sep 2022 12:19:25 +0000 (13:19 +0100)]
fix RCOE.RC_ONLY in formal test_decoder2.py
Dmitry Selyutin [Fri, 2 Sep 2022 08:43:06 +0000 (11:43 +0300)]
power_insn: drop custom Record representation
Dmitry Selyutin [Thu, 1 Sep 2022 19:55:01 +0000 (22:55 +0300)]
power_insn: support opcode merging
Dmitry Selyutin [Thu, 1 Sep 2022 14:50:54 +0000 (17:50 +0300)]
power_insn: support AA matching
Dmitry Selyutin [Thu, 1 Sep 2022 13:05:36 +0000 (16:05 +0300)]
power_insn: support LK matching
Dmitry Selyutin [Thu, 1 Sep 2022 12:55:39 +0000 (15:55 +0300)]
power_insn: refactor name matching algorithm
Dmitry Selyutin [Thu, 1 Sep 2022 15:04:24 +0000 (18:04 +0300)]
power_insn: refactor databases composition
Dmitry Selyutin [Thu, 1 Sep 2022 14:29:07 +0000 (17:29 +0300)]
power_insn: refactor operands
Dmitry Selyutin [Thu, 1 Sep 2022 12:40:16 +0000 (15:40 +0300)]
power_insn: introduce binutils-like representation
Dmitry Selyutin [Thu, 1 Sep 2022 12:32:38 +0000 (15:32 +0300)]
pysvp64dis: refactor instruction loading
Dmitry Selyutin [Thu, 1 Sep 2022 12:32:06 +0000 (15:32 +0300)]
power_fields: allow comparing references
Dmitry Selyutin [Thu, 1 Sep 2022 12:02:14 +0000 (15:02 +0300)]
power_insn: disable disassembly for prefixed instructions
Dmitry Selyutin [Thu, 1 Sep 2022 11:54:11 +0000 (14:54 +0300)]
power_insn: switch target_addr to real fields
Dmitry Selyutin [Wed, 31 Aug 2022 10:45:32 +0000 (13:45 +0300)]
power_insn: support GPR and FPR operands
Dmitry Selyutin [Wed, 31 Aug 2022 10:36:48 +0000 (13:36 +0300)]
power_insn: support target_addr operands
Dmitry Selyutin [Wed, 31 Aug 2022 10:13:44 +0000 (13:13 +0300)]
power_insn: hide operand classes
Dmitry Selyutin [Wed, 31 Aug 2022 08:37:31 +0000 (11:37 +0300)]
power_insn: introduce operand disassembly
Dmitry Selyutin [Tue, 30 Aug 2022 19:32:05 +0000 (22:32 +0300)]
pysvp64dis: disassemble word instruction operands
Dmitry Selyutin [Tue, 30 Aug 2022 19:06:00 +0000 (22:06 +0300)]
power_insn: support operands
Jacob Lifshay [Fri, 2 Sep 2022 06:34:11 +0000 (23:34 -0700)]
make tests pass again
Jacob Lifshay [Fri, 2 Sep 2022 06:33:39 +0000 (23:33 -0700)]
add formal_test_temp to .gitignore
Jacob Lifshay [Fri, 2 Sep 2022 06:32:02 +0000 (23:32 -0700)]
format code
Jacob Lifshay [Fri, 2 Sep 2022 06:06:56 +0000 (23:06 -0700)]
rename proof_decoder*.py -> test_decoder*.py so it gets run by pytest
Luke Kenneth Casson Leighton [Thu, 1 Sep 2022 22:56:59 +0000 (23:56 +0100)]
ghostmansd found that extswsli is incorrectly declared
as extswsli RA,RS,SH as XS-Form, where a quick check shows that
there *is* no field "SH" in XS-Form. there *is* however a field "sh"
in XS-Form and the pseudocode *correctly* uses it.
that then led to a review of all the other uses of SH and sh,
which also led to spotting that ME and MB are *also* mis-used
(MD-Form and MDS-Form) and need lower-casing in the operand
declaration but are *also* correct in the pseudo-code
`
Luke Kenneth Casson Leighton [Thu, 1 Sep 2022 17:11:28 +0000 (18:11 +0100)]
remove hard-coded list of operations in DecodeOE which
had been bugging the hell out of me for some time
Luke Kenneth Casson Leighton [Thu, 1 Sep 2022 17:10:55 +0000 (18:10 +0100)]
missed rlwm* in conversion to RC_ONLY
Luke Kenneth Casson Leighton [Thu, 1 Sep 2022 16:16:20 +0000 (17:16 +0100)]
update CSV files marking those instructions that are RC-only as such
there were actually some bugs here, entries that had not been added to
DecodeOE such as (all!) fp operations (which have no OE=1 variant)
and the new AV abs/min/max instructions
Luke Kenneth Casson Leighton [Thu, 1 Sep 2022 15:54:39 +0000 (16:54 +0100)]
drat have to use RCOE.RC not RCOE.RC_OE for now otherwise
all CSV entries have to change
Luke Kenneth Casson Leighton [Thu, 1 Sep 2022 15:51:06 +0000 (16:51 +0100)]
add missing case name RC_OE RC_ONLY in power_insn.py
Luke Kenneth Casson Leighton [Thu, 1 Sep 2022 15:49:29 +0000 (16:49 +0100)]
rename RC to RCOE in power_insns.py
Luke Kenneth Casson Leighton [Thu, 1 Sep 2022 15:47:42 +0000 (16:47 +0100)]
rename FLAGS to RCOE
Luke Kenneth Casson Leighton [Thu, 1 Sep 2022 15:42:14 +0000 (16:42 +0100)]
begin rename of RC to FLAGS and add RC_OE/RC_ONLY
Luke Kenneth Casson Leighton [Thu, 1 Sep 2022 15:36:49 +0000 (16:36 +0100)]
rename RC to FLAGS
Luke Kenneth Casson Leighton [Thu, 1 Sep 2022 15:34:15 +0000 (16:34 +0100)]
begin doing RC_OE / RC_ONLY
Dmitry Selyutin [Thu, 1 Sep 2022 14:40:39 +0000 (17:40 +0300)]
svbranch.mdwn: replace target_addr with BD
Dmitry Selyutin [Thu, 1 Sep 2022 14:20:02 +0000 (17:20 +0300)]
pagereader: skip empty dynamic and static operands
Jacob Lifshay [Thu, 1 Sep 2022 08:17:09 +0000 (01:17 -0700)]
move fsins/fcoss to fptrans.mdwn -- they are transcendental not SV instructions
Jacob Lifshay [Thu, 1 Sep 2022 07:48:48 +0000 (00:48 -0700)]
don't install recommended packages in CI
this should save a bunch of time
Jacob Lifshay [Thu, 1 Sep 2022 07:38:38 +0000 (00:38 -0700)]
remove support for unnamed arguments for @_custom_insns functions
unnamed arguments support was broken anyway, it would call with
`fields` last rather than first, due to how partial() works.
Jacob Lifshay [Wed, 31 Aug 2022 04:24:09 +0000 (21:24 -0700)]
silence default LogKind for CI, to avoid pytest running out of memory due to storing stdout
(cherry picked from commit
bec04597888ebf0ca61b852b63cf1efa41ed5f79)
Jacob Lifshay [Tue, 30 Aug 2022 07:51:16 +0000 (00:51 -0700)]
remove dead code, sv.svstep and sv.fcoss are now handled by CUSTOM_INSNS
(cherry picked from commit
cfd0215fc6f716317eb22d04d965028eb794965b)
Jacob Lifshay [Tue, 30 Aug 2022 07:50:28 +0000 (00:50 -0700)]
switch ci tests to verbose
(cherry picked from commit
4053433d420915d9cb777f62fdc10d0b29178dda)
Jacob Lifshay [Tue, 30 Aug 2022 07:32:21 +0000 (00:32 -0700)]
make sv.instr use CUSTOM_INSNS by default for assembling instr
(cherry picked from commit
9be3b61b6a3037fdcf087f692e9145678e571a35)
Jacob Lifshay [Tue, 30 Aug 2022 06:14:31 +0000 (23:14 -0700)]
use a decorator for constructing CUSTOM_INSNS
(cherry picked from commit
63fd4ebc03ea0a7d51a1cf8d215affe45e0f0b33)
Jacob Lifshay [Tue, 30 Aug 2022 06:12:52 +0000 (23:12 -0700)]
fix comment location
(cherry picked from commit
2ba62128ffb197e41a9a7b518a915ebe839d8a71)
Jacob Lifshay [Tue, 30 Aug 2022 05:53:52 +0000 (22:53 -0700)]
format code
(cherry picked from commit
61faa7c3c443a260eaeb58b111ebd464a2031e06)
Jacob Lifshay [Thu, 1 Sep 2022 06:38:43 +0000 (23:38 -0700)]
Revert "silence default LogKind for CI, to avoid pytest running out of memory due to storing stdout"
lkcl rebased my commit, mushing multiple changes into one commit in the process
This reverts commit
aea0a4d637b04e54a9e942883e5e31dd338ca3f2.
Jacob Lifshay [Wed, 31 Aug 2022 04:54:39 +0000 (21:54 -0700)]
test less cases of utf-8 validation, to avoid taking forever
Jacob Lifshay [Wed, 31 Aug 2022 04:24:09 +0000 (21:24 -0700)]
silence default LogKind for CI, to avoid pytest running out of memory due to storing stdout
Luke Kenneth Casson Leighton [Wed, 31 Aug 2022 13:33:56 +0000 (14:33 +0100)]
target_addr in b and bc pseudo-code has no corresponding
field in Section 1.6. actually, target_addr=LI for b and target_addr=BI
for bc
therefore make the pseudo-code actually do exactly that
Dmitry Selyutin [Tue, 30 Aug 2022 18:41:05 +0000 (21:41 +0300)]
bcd.mdwn: fix cbcdtd operands
Dmitry Selyutin [Tue, 30 Aug 2022 18:36:36 +0000 (21:36 +0300)]
bcd.mdwn: fix cdtbcd operands
Dmitry Selyutin [Tue, 30 Aug 2022 18:29:51 +0000 (21:29 +0300)]
fixedlogical.mdwn: fix bpermd operands
Dmitry Selyutin [Tue, 30 Aug 2022 12:27:35 +0000 (15:27 +0300)]
pagereader: validate input
Luke Kenneth Casson Leighton [Tue, 30 Aug 2022 16:14:25 +0000 (17:14 +0100)]
remove space from arguments in popcnt, should not have been there
Luke Kenneth Casson Leighton [Tue, 30 Aug 2022 13:21:14 +0000 (14:21 +0100)]
Revert "format code"
This reverts commit
61faa7c3c443a260eaeb58b111ebd464a2031e06.