Alan Modra [Thu, 6 Jun 2024 22:57:31 +0000 (08:27 +0930)]
Re: Yet another ecoff fuzzed object fix
In commit 6fc018e9e593 I replaced the fdr_ptr csym check against the
header isymMax count with a check against bfd symcount. In fact, both
checks are needed. The isymMax check sanity checks accesses against
the external sym array, the symcount one against the internal array.
* ecoff.c (_bfd_ecoff_slurp_symbol_table): Reinstate fdr_ptr
csym check against isymMax.
Szabolcs Nagy [Mon, 3 Jun 2024 16:20:32 +0000 (17:20 +0100)]
aarch64: Fix DT_RELR support with discarded sections
In case of discarded sections, via /DISCARD/ or .gnu.linkonce,
relr relocation accounting was wrong. This broke building linux.
The issue was that the *_relocate_section logic was copied to
record_relr_non_got_relocs to find the relative relocs that can
be packed, however *_relocate_section is not called on sections
that are discarded, while record_relr_non_got_relocs is called
for all input sections. The fix is to filter out the discarded
sections with the same logic that is used to count non-GOT
relocs in *_late_size_sections for local symbols earlier.
Use the discarded_section helper in both cases to clarify the
intent and handle all corner-cases consistently.
GOT relocations are affected too if all sections are discarded
that reference the GOT entry of a particular symbol, however
this can cause unused GOT entries independently of DT_RELR, and
the only difference with DT_RELR is that a relative reloc may be
emitted instead of a R_AARCH64_NONE for the unused GOT entry
which is acceptable. A proper fix would require redoing the GOT
refcounting after we know the discarded sections, see bug 31850.
I rewrote the rules for building the man pages. While doing this I
accidentally switched from using MAN2POD5 to MAN2POD1 for generating
the file gdbinit.5.
Johan Sternerup [Sat, 1 Jun 2024 16:16:32 +0000 (18:16 +0200)]
DAP: Handle "stepOut" request in outermost frame
Previously a "stepOut" request when in the outermost frame would result
in a sucessful response even though gdb internally would reject the
associated "finish" request, which means no stoppedEvent would ever be
sent back to the client. Thus the client would believe the inferior was
still running and as a consequence reject subsequent "next" and "stepIn"
requests from the user.
The solution is to execute the underlying finish command as a background
command, i.e. `finish &`. If we're in the outermost frame an exception
will be raised immediately, which we can now capture and report back to
the client as success=False so then the absence of a `stopped` event is
no longer a problem.
We also make use of the `defer_stop_event` option to prevent a stop
event from reaching the client until the response has been sent.
Johan Sternerup [Sat, 1 Jun 2024 16:16:31 +0000 (18:16 +0200)]
DAP: Allow gdb exception in exec_and_log to propagate
This allows a request to specify that any gdb exception raised in
exec_and_log within the gdb thread to be propagated back to the DAP
thread (using the Canceller object as the orchestrator).
Johan Sternerup [Sat, 1 Jun 2024 16:16:30 +0000 (18:16 +0200)]
DAP: Allow for deferring stop events from gdb thread
The existing `send_event_later()` method allows commands processed on
the DAP thread to queue an event for execution until after the response
has been sent to the client.
We now introduce a corresponding method for use by the gdb thread. This
method `send_event_maybe_later()` will queue the event just like
`send_event_later()`, but only if it has been configured to do so by a
new @request option `defer_stop_events`. As the name implies the
functionality is currently only used for handling stop events.
arm: fix testsuite fallout on arm-elf and arm-nto from FPA removal
Removing FPA means that in some cases we default to 'no-fpu' in the
assembler when previously we would have picked FPA-format floating
numbers. This patch fixes the testsuite fallout on a couple of
targets that are affected by this change. Where possible we do this
by adding an option to set the floating-point format, but for bad-bss
we just skip the test.
Andrew Burgess [Wed, 5 Jun 2024 12:12:42 +0000 (13:12 +0100)]
opcodes/riscv: prevent future use of disassemble_info::fprintf_func
The previous commit removed a use of disassemble_info::fprintf_func
which had been added to the RISC-V disassembler after the disassembler
had been switched to use ::fprintf_styled_func, for styled output.
To prevent future mistakes, I propose adding a #define to rename
fprintf_func to something which does not exist. If this had been in
place then the before the previous commit libopcodes would have failed
to compile, like this:
../../src/opcodes/riscv-dis.c: In function ‘print_reg_list’:
../../src/opcodes/riscv-dis.c:229:7: error: ‘disassemble_info’ {aka ‘struct disassemble_info’} has no member named ‘please_use_fprintf_styled_func_instead’
229 | info->fprintf_func (info->stream, "%s", riscv_gpr_names[X_RA]);
| ^~
If this commit is accepted then I'll follow up with another commit
that adds the same #define to every disassembler that has been
converted to use styled output.
As the RISC-V disassembler is now fully styled, this commit should
make no difference at all.
Xiao Zeng [Thu, 6 Jun 2024 07:59:53 +0000 (15:59 +0800)]
RISC-V: Add support for Zvfbfwma extension
This implements the Zvfbfwma extension, as of version 1.0.
View detailed information in:
<https://github.com/riscv/riscv-isa-manual/blob/main/src/bfloat16.adoc#zvfbfwma---vector-bf16-widening-mul-add>
1 In spec: "Zvfbfwma requires the Zvfbfmin extension and the Zfbfmin extension."
1.1 In Embedded Processor: Zvfbfwma -> Zvfbfmin -> Zve32f
1.2 In Application Processor: Zvfbfwma -> Zvfbfmin -> V
1.3 In both scenarios, there are: Zvfbfwma -> Zfbfmin
2 Depending on different usage scenarios, the Zvfbfwma extension may
depend on 'V' or 'Zve32f'. This patch only implements dependencies in
scenario of Embedded Processor. This is consistent with the processing
strategy in Zvfbfmin. In scenario of Application Processor, it is
necessary to explicitly indicate the dependent 'V' extension.
Xiao Zeng [Thu, 6 Jun 2024 07:59:52 +0000 (15:59 +0800)]
RISC-V: Add support for Zvfbfmin extension
This implements the Zvfbfmin extension, as of version 1.0.
View detailed information in:
<https://github.com/riscv/riscv-isa-manual/blob/main/src/bfloat16.adoc#zvfbfmin---vector-bf16-converts>
Depending on different usage scenarios, the Zvfbfmin extension may
depend on 'V' or 'Zve32f'. This patch only implements dependencies
in scenario of Embedded Processor. In scenario of Application
Processor, it is necessary to explicitly indicate the dependent
'V' extension.
Xiao Zeng [Thu, 6 Jun 2024 07:59:51 +0000 (15:59 +0800)]
RISC-V: Add support for Zfbfmin extension
This implements the Zfbfmin extension, as of version 1.0.
View detailed information in:
<https://github.com/riscv/riscv-isa-manual/blob/main/src/bfloat16.adoc#zfbfmin---scalar-bf16-converts>
1 The Zfbfmin extension depend on 'F', and the FLH, FSH, FMV.X.H, and
FMV.H.X instructions as defined in the Zfh extension.
2 The Zfhmin extension includes the following instructions from the Zfh
extension: FLH, FSH, FMV.X.H, FMV.H.X... View detailed information in:
<https://github.com/riscv/riscv-isa-manual/blob/main/src/zfh.adoc>
3 Zfhmin extension depend on 'F'.
4 Simply put, just make Zfbfmin dependent on Zfhmin.
Perhaps in the future, we could propose making the FLH, FSH, FMV.X.H, and
FMV.H.X instructions an independent extension to achieve precise dependency
relationships for the Zfbfmin.
Alan Modra [Mon, 3 Jun 2024 23:30:57 +0000 (09:00 +0930)]
Re: aarch64: Add some DT_RELR ld tests
aarch64-elf fails these tests due to .rela.dyn being at a different
address to that expected, and due to the symbol table being different.
Unexpected symbol numbering results in a mismatch of reloc r_info
field, but these are shown decoded so the raw field doesn't really add
anything to the test.
* testsuite/ld-aarch64/relr-align.d: Accept any address for
.relr.dyn section. Don't match raw r_info field.
* testsuite/ld-aarch64/relr-data-shared.d: Likewise.
* testsuite/ld-aarch64/relr-got-shared.d: Likewise.
* testsuite/ld-aarch64/relr-text-shared.d: Likewise.
arm: remove disassembly support for the FPA co-processor
Remove the FPA support from the disassembler. This entails a couple
of testsuite fixes where we were (probably incorrectly) disassembling
a generic co-processor instruction using the legacy FPA opcodes.
Remove the command-line options to choose the FPA (or FPE - an
emulated FPA). From this point on it should be impossible to assemble
the old FPA instructions.
Change the cases where the default FPU was FPA to none. This should
ensure that any code that used settings to pick the floating-point
order will not silently produce a different output. The options that
explicitly set the FPA remain for the moment.
arm: redirect fp constant data directives through a wrapper
Assembler directives such as .float, or .double are handled by generic
code, but on Arm, their output can vary depeding on the type of FPU
begin targetted. When we remove FPA support we don't want to silently
generate different code for processors that previously defaulted to
the FPA, so redirect these directives through a wrapper function that
checks the FPU is enabled; we use the legacy -mno-fpu in the test to
catch this.
Also fix a few tests so that they won't start to fail on targets (eg
arm-wince-pe) where there is no default format for the FPU and we pick
this from the default processor type.
The logic here seems to be overly complex, so simplify it a bit. One
particular problem was that using the legacy -mno-fpu option was not
working properly, as this has all the feature bits set to zero causing
the code to then pick a different FPU as the default. Fix this by
only selecting an FPU as a fallback if the code has not otherwise
selected one: there was only one route by which this could happen.
This patch is really a pre-cursor to the following one where we want
to make no-fpu internally a fall-back position for some legacy
processors where previously we would have dropped back to the FPA.
From armv6 onwards a lot of cores started to come with a physical VFP
implementation; but many still did not and in some cases there are
both variants. For the cores that lacked a physical VFP we would fall
back to FPU_NONE if the platform/ABI did not mandate something else.
To make matters worse, FPU_NONE is internal state used to imply
soft-fpa (ie a mixed-endian double format), so any use of .double in
hand-written assembly is almost certainly generating incorrect output.
That's undesirable, all these cores should really default to a softvfp
model.
FPU_ARCH_VFP has always meant VFP floating-point format (natural FP
word order) but without any VFP instructions. But the name
FPU_ARCH_VFP is potentially confusing. This patch just changes the
name to make the meaning clearer.
Nelson Chu [Wed, 5 Jun 2024 09:55:33 +0000 (17:55 +0800)]
RISC-V: Tidy vendor core-v extension gas testcases
1. Combined testcases into one if they use same extention name.
2. Likewise for the fail testcases.
3. Renamed with x-cv prefix, just like what other vendors did.
gas/
* testsuite/gas/riscv/cv-alu-*: Combined and renamed to
x-cv-alu. Likewise for fail testcases, to x-cv-alu-fail*.
* testsuite/gas/riscv/cv-bi-*: Likewise, but renamed to
x-cv-bi and x-cv-bi-fail.
* testsuite/gas/riscv/cv-elw-*: Likewise, but renamed to
x-cv-elw and x-cv-elw-fail.
* testsuite/gas/riscv/cv-mac-*: Likewise, but renamed to
x-cv-mac and x-cv-mac-fail.
* testsuite/gas/riscv/cv-mem-*: Likewise, but renamed to
x-cv-mem and x-cv-mem-fail.
gas/ChangeLog:
* doc/c-riscv.texi: Note XCVmem as an additional ISA extension
for CORE-V.
* testsuite/gas/riscv/cv-mem-fail-march.d: New test.
* testsuite/gas/riscv/cv-mem-fail-march.l: New test.
* testsuite/gas/riscv/cv-mem-fail-march.s: New test.
* testsuite/gas/riscv/cv-mem-fail-operand-01.d: New test.
* testsuite/gas/riscv/cv-mem-fail-operand-01.l: New test.
* testsuite/gas/riscv/cv-mem-fail-operand-01.s: New test.
* testsuite/gas/riscv/cv-mem-fail-operand-02.d: New test.
* testsuite/gas/riscv/cv-mem-fail-operand-02.l: New test.
* testsuite/gas/riscv/cv-mem-fail-operand-02.s: New test.
* testsuite/gas/riscv/cv-mem-fail-operand-03.d: New test.
* testsuite/gas/riscv/cv-mem-fail-operand-03.l: New test.
* testsuite/gas/riscv/cv-mem-fail-operand-03.s: New test.
* testsuite/gas/riscv/cv-mem-fail-operand-04.d: New test.
* testsuite/gas/riscv/cv-mem-fail-operand-04.l: New test.
* testsuite/gas/riscv/cv-mem-fail-operand-04.s: New test.
* testsuite/gas/riscv/cv-mem-fail-operand-05.d: New test.
* testsuite/gas/riscv/cv-mem-fail-operand-05.l: New test.
* testsuite/gas/riscv/cv-mem-fail-operand-05.s: New test.
* testsuite/gas/riscv/cv-mem-lbpost.d: New test.
* testsuite/gas/riscv/cv-mem-lbpost.s: New test.
* testsuite/gas/riscv/cv-mem-lbrr.d: New test.
* testsuite/gas/riscv/cv-mem-lbrr.s: New test.
* testsuite/gas/riscv/cv-mem-lbrrpost.d: New test.
* testsuite/gas/riscv/cv-mem-lbrrpost.s: New test.
* testsuite/gas/riscv/cv-mem-lbupost.d: New test.
* testsuite/gas/riscv/cv-mem-lbupost.s: New test.
* testsuite/gas/riscv/cv-mem-lburr.d: New test.
* testsuite/gas/riscv/cv-mem-lburr.s: New test.
* testsuite/gas/riscv/cv-mem-lburrpost.d: New test.
* testsuite/gas/riscv/cv-mem-lburrpost.s: New test.
* testsuite/gas/riscv/cv-mem-lhpost.d: New test.
* testsuite/gas/riscv/cv-mem-lhpost.s: New test.
* testsuite/gas/riscv/cv-mem-lhrr.d: New test.
* testsuite/gas/riscv/cv-mem-lhrr.s: New test.
* testsuite/gas/riscv/cv-mem-lhrrpost.d: New test.
* testsuite/gas/riscv/cv-mem-lhrrpost.s: New test.
* testsuite/gas/riscv/cv-mem-lhupost.d: New test.
* testsuite/gas/riscv/cv-mem-lhupost.s: New test.
* testsuite/gas/riscv/cv-mem-lhurr.d: New test.
* testsuite/gas/riscv/cv-mem-lhurr.s: New test.
* testsuite/gas/riscv/cv-mem-lhurrpost.d: New test.
* testsuite/gas/riscv/cv-mem-lhurrpost.s: New test.
* testsuite/gas/riscv/cv-mem-lwpost.d: New test.
* testsuite/gas/riscv/cv-mem-lwpost.s: New test.
* testsuite/gas/riscv/cv-mem-lwrr.d: New test.
* testsuite/gas/riscv/cv-mem-lwrr.s: New test.
* testsuite/gas/riscv/cv-mem-lwrrpost.d: New test.
* testsuite/gas/riscv/cv-mem-lwrrpost.s: New test.
* testsuite/gas/riscv/cv-mem-sbpost.d: New test.
* testsuite/gas/riscv/cv-mem-sbpost.s: New test.
* testsuite/gas/riscv/cv-mem-sbrr.d: New test.
* testsuite/gas/riscv/cv-mem-sbrr.s: New test.
* testsuite/gas/riscv/cv-mem-sbrrpost.d: New test.
* testsuite/gas/riscv/cv-mem-sbrrpost.s: New test.
* testsuite/gas/riscv/cv-mem-shpost.d: New test.
* testsuite/gas/riscv/cv-mem-shpost.s: New test.
* testsuite/gas/riscv/cv-mem-shrr.d: New test.
* testsuite/gas/riscv/cv-mem-shrr.s: New test.
* testsuite/gas/riscv/cv-mem-shrrpost.d: New test.
* testsuite/gas/riscv/cv-mem-shrrpost.s: New test.
* testsuite/gas/riscv/cv-mem-swpost.d: New test.
* testsuite/gas/riscv/cv-mem-swpost.s: New test.
* testsuite/gas/riscv/cv-mem-swrr.d: New test.
* testsuite/gas/riscv/cv-mem-swrr.s: New test.
* testsuite/gas/riscv/cv-mem-swrrpost.d: New test.
* testsuite/gas/riscv/cv-mem-swrrpost.s: New test.
* testsuite/gas/riscv/march-help.l: Add xcvmem string.
include/ChangeLog:
* opcode/riscv-opc.h: Add corresponding MATCH and MASK macros
for XCVmem.
* opcode/riscv.h: Add corresponding EXTRACT and ENCODE macros
for XCVmem.
(enum riscv_insn_class): Add the XCVmem instruction class.
Contributors:
Mary Bennett <mary.bennett682@gmail.com>
Nandni Jamnadas <nandni.jamnadas@embecosm.com>
Pietra Ferreira <pietra.ferreira@embecosm.com>
Charlie Keaney
Jessica Mills
Craig Blackmore <craig.blackmore@embecosm.com>
Simon Cook <simon.cook@embecosm.com>
Jeremy Bennett <jeremy.bennett@embecosm.com>
Helene Chelin <helene.chelin@embecosm.com>
Nazareno Bruschi <nazareno.bruschi@embecosm.com>
Lin Sinan
include/ChangeLog:
* opcode/riscv-opc.h: Add corresponding MATCH and MASK
macros for XCVbi.
* opcode/riscv.h: Add corresponding EXTRACT and ENCODE macros
for XCVbi.
(enum riscv_insn_class): Add the XCVbi instruction class.
gas/ChangeLog:
* config/tc-riscv.c (validate_riscv_insn): Add the necessary
operands for the extension.
(riscv_ip): Likewise.
* doc/c-riscv.texi: Note XCVbi as an additional ISA extension
for CORE-V.
* testsuite/gas/riscv/cv-bi-beqimm.d: New test.
* testsuite/gas/riscv/cv-bi-beqimm.s: New test.
* testsuite/gas/riscv/cv-bi-bneimm.d: New test.
* testsuite/gas/riscv/cv-bi-bneimm.s: New test.
* testsuite/gas/riscv/cv-bi-fail-march.d: New test.
* testsuite/gas/riscv/cv-bi-fail-march.l: New test.
* testsuite/gas/riscv/cv-bi-fail-march.s: New test.
* testsuite/gas/riscv/cv-bi-fail-operand-01.d: New test.
* testsuite/gas/riscv/cv-bi-fail-operand-01.l: New test.
* testsuite/gas/riscv/cv-bi-fail-operand-01.s: New test.
* testsuite/gas/riscv/cv-bi-fail-operand-02.d: New test.
* testsuite/gas/riscv/cv-bi-fail-operand-02.l: New test.
* testsuite/gas/riscv/cv-bi-fail-operand-02.s: New test.
* testsuite/gas/riscv/cv-bi-fail-operand-03.d: New test.
* testsuite/gas/riscv/cv-bi-fail-operand-03.l: New test.
* testsuite/gas/riscv/cv-bi-fail-operand-03.s: New test.
* testsuite/gas/riscv/march-help.l: Add xcvbi string.
include/ChangeLog:
* opcode/riscv-opc.h: Add corresponding MATCH and MASK
macros for XCVbi.
* opcode/riscv.h: Add corresponding EXTRACT and ENCODE macros
for XCVbi.
(enum riscv_insn_class): Add the XCVbi instruction class.
opcodes/ChangeLog:
* riscv-dis.c (print_insn_args): Add disassembly for new operand.
* riscv-opc.c: Add XCVbi instructions.
* doc/c-riscv.texi: Note XCVelw as an additional ISA extension
for CORE-V.
* testsuite/gas/riscv/cv-elw-fail.d: New test.
* testsuite/gas/riscv/cv-elw-fail.l: New test.
* testsuite/gas/riscv/cv-elw-fail.s: New test.
* testsuite/gas/riscv/cv-elw-fail-march.d: New test.
* testsuite/gas/riscv/cv-elw-fail-march.l: New test.
* testsuite/gas/riscv/cv-elw-fail-march.s: New test.
* testsuite/gas/riscv/cv-elw-pass.d: New test.
* testsuite/gas/riscv/cv-elw-pass.s: New test.
* testsuite/gas/riscv/march-help.l: Add xcvelw string.
Andrew Burgess [Fri, 31 May 2024 21:03:06 +0000 (22:03 +0100)]
gdb/testsuite: remove trailing \r from rust_llvm_version result
I noticed that the value returned by rust_llvm_version had a trailing
carriage return. I don't think this is causing any problems right
now, but looking at the code I don't think this was the desired
behaviour.
The current code runs 'rustc --version --verbose', splits the output
at each '\n' and then loops over every line looking for the line that
contains the LLVM version.
There are two problems here. First, at the end of each captured line
we have '\r\n', so when we split the lines on '\n', each of the lines
will still end with a '\r' character.
Second, though we loop over the lines, when we try to compare the line
contents we actually compare the unsplit full output. Luckily this
still finds the match, but this renders the loop over lines redundant.
This commit makes two fixes:
1. I use regsub to convert all '\r\n' sequences to '\n'; now when we
split on '\n' the lines will not end in '\r'.
2. Within the loop over lines block I now check the line contents
rather than the unsplit full output; now we capture a value
without a trailing '\r'.
There's only one test (gdb.rust/simple.exp) that uses
rust_llvm_version, and it doesn't care if there's a trailing '\r' or
not, so this change should make no difference there.
Andrew Burgess [Tue, 21 May 2024 15:20:43 +0000 (16:20 +0100)]
gdb: more filename styling in remote.c and target.c
I spotted a few more places where we could apply filename styling in
remote.c and target.c. Other than the styling, there should be no
user visible changes after this commit.
Tom Tromey [Thu, 16 May 2024 13:58:07 +0000 (07:58 -0600)]
Return global scope from DAP scopes request
A co-worker requested that the DAP code emit a scope for global
variables. It's not really practical to do this for all globals, but
it seemed reasonable to do this for globals coming from the frame's
compilation unit. For Ada in particular, this is convenient as it
exposes package-scoped variables.
Tom Tromey [Wed, 15 May 2024 18:07:36 +0000 (12:07 -0600)]
Memoize gdb.Block and make them hashable
In subsequent patches, it's handy if gdb.Block is hashable, so it can
be stored in a set or a dictionary. However, doing this in a
straightforward way is not really possible, because a block isn't
truly immutable -- it can be invalidated. And, while this isn't a
real problem for my use case (in DAP the maps are only used during a
single stop), it seemed error-prone.
This patch instead takes the approach of using the gdb.Block's own
object identity to allow hashing. This seems fine because the
contents don't affect the hashing. In order for this to work, though,
the blocks have to be memoized -- two requests for the same block must
return the same object.
This also allows (actually, requires) the simplification of the
rich-compare method for blocks.
Reviewed-By: Alexandra Petlanova Hajkova <ahajkova@redhat.com>
Tom Tromey [Mon, 29 Apr 2024 18:57:11 +0000 (12:57 -0600)]
Put "source" into DAP scope
I noticed a FIXME comment in the DAP code about adding a "source"
field to a scope. This is easy to implement; I don't know why I
didn't do this originally.
Tom Tromey [Wed, 20 Mar 2024 19:24:33 +0000 (13:24 -0600)]
Make bcache more type-safe
The bcache uses memcpy to make copies of the data passed to it. In
C++, this is only safe for trivially-copyable types.
This patch changes bcache to require this property, and slightly
changes the API to make it easier to use when copying a single object.
It also makes the new 'insert' template methods return the correct
type.
Tom Tromey [Wed, 20 Mar 2024 19:42:36 +0000 (13:42 -0600)]
Some constification in psymtab
This patch changes some spots in psymtab.[ch] to use 'const'. This is
just preparation for a subsequent patch. Note that psymbols are
conceptually const, and since they were changed to be
objfile-indepdendent, they are truly never modified -- which is what
makes this patch reasonably short.
Andrew Burgess [Sat, 18 May 2024 09:50:23 +0000 (10:50 +0100)]
gdb/testsuite: tests for debug lookup within the sysroot
Add tests for looking up debug information within the sysroot via both
build-id and gnu_debuglink.
I wanted to ensure that the gnu_debuglink test couldn't make use of
build-ids, so I added the 'no-build-id' flag to gdb_compile.
As these tests rely on setting the sysroot, if I'm running a
dynamically linked executable, GDB will try to find all shared
libraries within the sysroot. This would mean I'd have to figure out
and copy all shared libraries the executable uses, certainly possible,
but a bit of a pain.
So instead, I've just compiled the test executable as a static binary.
Now there are no shared library dependencies.
I can now split the debug information out from the test binary, and
place it within the sysroot. When GDB is started and the executable
loaded, we can check that GDB is finding the debug information within
the sysroot.
Andrew Burgess [Sat, 18 May 2024 10:05:49 +0000 (11:05 +0100)]
gdb/testsuite: make gdb_gnu_strip_debug consistent
While writing a test I realised that the default behaviour of
gdb_gnu_strip_debug doesn't match its comment.
The comment says that the function takes a FILENAME, and splits the
file into FILENAME.stripped and FILENAME.debug, leaving FILENAME
unchanged. The comment says that a .gnu_debuglink will be added to
FILENAME.stripped.
However, this is not true, FILENAME.stripped is created, with no debug
information. FILENAME.debug is created containing the debug
information.
But, when adding the .gnu_debuglink we take FILENAME.stripped as the
input, and then overwrite FILENAME with the output. As a result,
FILENAME.stripped does not include a .gnu_debuglink, while FILENAME
contains the .gnu_debuglink and no debug information!
The users of gdb_gnu_strip_debug can be split into two groups, those
who are using the .gnu_debuglink, these tests are all written assuming
that FILENAME is updated.
Then there are some tests that only rely on gdb_gnu_strip_debug's
ability to split out the debug information, these tests are then going
to do a lookup based on the build-id, these tests don't require the
.gnu_debuglink. These tests use the FILENAME.stripped output file.
This all seems too confused to me.
As most uses of gdb_gnu_strip_debug assume that FILENAME is updated, I
propose that we just make that the actual, advertised behaviour of
this proc.
So now, gdb_gnu_strip_debug will take FILENAME, and will split the
debug information out into FILENAME.debug. The debug information will
then be stripped from FILENAME, and by default a .gnu_debuglink will
be added to FILENAME pointing to FILENAME.debug.
I've updated the two tests that actually relied on FILENAME.stripped
to instead just use FILENAME.
One of the tests was doing a build-id based lookup, but was still
allowing the .gnu_debuglink to be added to FILENAME, I've updated this
test to pass the no-debuglink flag to gdb_gnu_strip_debug, which stops
the .gnu_debuglink from being added.
All of the tests that call gdb_gnu_strip_debug still pass for me.
this commit uses silent-rules.mk to silence this output, now we'll
see:
GEN arch/.deps
GEN cli/.deps
GEN dwarf2/.deps
... etc ...
The recipe that currently generates these directories uses
mkinstalldirs, as I mention in commit 032e5e0c0c08977, mkinstalldirs
is deprecated and 'install-sh -d' should be used instead. This
silences the 'mkdir -p -- ...' part of the output.
There should be no change in what is actually built after this commit.
mengqinggang [Wed, 29 May 2024 06:50:39 +0000 (14:50 +0800)]
LoongArch: Make align symbol be in same section with alignment directive
R_LARCH_ALIGN (psABI v2.30) requires a symbol index. The symbol is only
created at the first time to handle alignment directive. This means that
all other sections may use this symbol. If the section of this symbol is
discarded, there may be problems. Search it in its own section.
Remove elf_backend_data.is_rela_normal() function added at commit daeda14191c.
A couple of the tests in the testsuite were at some point in the past
committed with DOS-style CRLF line endings. This potentially causes
email problems if the tests are touched in the middle of a large patch
series so convert them to standard Un*x line endings.
Hannes Domani [Mon, 3 Jun 2024 15:18:30 +0000 (17:18 +0200)]
Allow calling of convenience functions with python
As mentioned in PR13326, currently when you try to call a
convenience function with python, you get this error:
(gdb) py print(gdb.convenience_variable("_isvoid")(3))
Traceback (most recent call last):
File "<string>", line 1, in <module>
RuntimeError: Value is not callable (not TYPE_CODE_FUNC or TYPE_CODE_METHOD).
Error while executing Python code.
So this extends valpy_call to handle TYPE_CODE_INTERNAL_FUNCTION as
well, making this possible:
Tom de Vries [Mon, 3 Jun 2024 05:49:24 +0000 (07:49 +0200)]
[gdb/testsuite] Fix timeout in gdb.tui/resize-2.exp
When running test-case gdb.tui/resize-2.exp with taskset -c 0, I sometimes run
into:
...
tui disable^[[40;1H^M(gdb) PASS: $exp: tui disable
^M^[[K(gdb) FAIL: $exp: two prompt redisplays after resize (timeout)
...
The test-case does "Term::resize 24 80 0" while having the settings of an
earlier "Term::resize 40 90", so both dimensions change.
When TUI is enabled, we call Term::resize with wait_for_msg == 1, and the proc:
- calls stty to change one dimension,
- waits for the message (enabled by "maint set tui-resize-message on")
confirming the resize has happened,
- calls stty to change the other dimension, and again
- waits for the message confirming the resize has happened.
Since TUI is disabled, we call Term::resize with wait_for_msg == 0 because the
message is not printed, so stty is called twice, and afterwards we check for
the results of the two resizes, which is the test that is failing.
The problem is that not waiting for a response after each stty call opens up
the possibility of the responses being merged.
Fix this by calling Term::resize twice, changing one dimension at a time, and
waiting for a single prompt redisplay after each one.
Tested on x86_64-linux.
Reviewed-By: Alexandra Petlanova Hajkova <ahajkova@redhat.com>
PR testsuite/31822
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=31822
Kevin Buettner [Thu, 2 May 2024 01:37:58 +0000 (18:37 -0700)]
[gdb/testsuite] New test: gdb.base/errno.exp
Printing the value of 'errno' from GDB is sometimes problematic. The
situation has improved in recent years, though there are still
scenarios for which "print errno" doesn't work.
The test, gdb.base/errno.exp, introduced by this commit, tests whether
or not GDB can print errno using a binary compiled in the following
different ways:
- default: no switches aside from -g (and whatever else is added by the
testing framework)
- macros: macro info is included in the debuginfo; this is enabled by
using -g3 when using gcc or clang
- static: statically linked binary
- static-macros: statically linked binary w/ macro definitions included
in debuginfo
- pthreads: libpthread linked binary
- pthreads-macros: libpthread linked binary w/ macro definitions included
in debuginfo
- pthreads-static: Statically linked against libpthread
- pthreads-static-macros: Statically linked against libpthread w/ macro
definitions
For each of these, the test also creates a corefile, then loads the
corefile and attempts to print errno again.
Additionally, the test checks that a "masking" errno declared as a
local variable will print correctly.
On Linux, if the machine is missing glibc debuginfo (or you have
debuginfod disabled), it's likely you'll see:
(gdb) print errno
'errno' has unknown type; cast it to its declared type
But if you add a cast, the value of errno is often available:
(gdb) print (int) errno
$1 = 42
The test detects this situation along with several others and does
'setup_xfail' for tests that will almost certainly fail. It could be
argued that some of these ought to be KFAILs due to deficiencies in
GDB, but I'm not entirely certain which, if any, are fixable yet.
On Fedora 39, without glibc debuginfo, there are no failures, but
I do see the following XFAILS:
XFAIL: gdb.base/errno.exp: default: print errno
XFAIL: gdb.base/errno.exp: default: check errno value from corefile
XFAIL: gdb.base/errno.exp: macros: print errno
XFAIL: gdb.base/errno.exp: macros: print (int) errno
XFAIL: gdb.base/errno.exp: macros: check errno value from corefile
XFAIL: gdb.base/errno.exp: static: print errno
XFAIL: gdb.base/errno.exp: static: print (int) errno
XFAIL: gdb.base/errno.exp: static: check errno value from corefile
XFAIL: gdb.base/errno.exp: static-macros: print errno
XFAIL: gdb.base/errno.exp: static-macros: print (int) errno
XFAIL: gdb.base/errno.exp: static-macros: check errno value from corefile
XFAIL: gdb.base/errno.exp: pthreads: print errno
XFAIL: gdb.base/errno.exp: pthreads: check errno value from corefile
XFAIL: gdb.base/errno.exp: pthreads-macros: print errno
XFAIL: gdb.base/errno.exp: pthreads-macros: print (int) errno
XFAIL: gdb.base/errno.exp: pthreads-macros: check errno value from corefile
XFAIL: gdb.base/errno.exp: pthreads-static: print errno
XFAIL: gdb.base/errno.exp: pthreads-static: print (int) errno
XFAIL: gdb.base/errno.exp: pthreads-static: check errno value from corefile
XFAIL: gdb.base/errno.exp: pthreads-static-macros: print errno
XFAIL: gdb.base/errno.exp: pthreads-static-macros: print (int) errno
XFAIL: gdb.base/errno.exp: pthreads-static-macros: check errno value from corefile
On Fedora 39, with glibc debug info, but without libc.a (for static
linking), there are 2 XFAILs, 2 UNSUPPORTED tests, and 4 UNTESTED
tests.
So, even when testing in less than ideal conditions, either due to lack
of glibc debuginfo or lack of a libc to link against to make a static
binary, there are no failures.
With glibc debuginfo installed, on Fedora 38, Fedora 39, Fedora 40,
Fedora rawhide (41), and Ubuntu 22.04.1 LTS, I see these XFAILs:
XFAIL: gdb.base/errno.exp: macros: check errno value from corefile
XFAIL: gdb.base/errno.exp: static: print errno
XFAIL: gdb.base/errno.exp: static: print (int) errno
XFAIL: gdb.base/errno.exp: static: check errno value from corefile
XFAIL: gdb.base/errno.exp: static-macros: print errno
XFAIL: gdb.base/errno.exp: static-macros: print (int) errno
XFAIL: gdb.base/errno.exp: static-macros: check errno value from corefile
XFAIL: gdb.base/errno.exp: pthreads-macros: check errno value from corefile
XFAIL: gdb.base/errno.exp: pthreads-static: print errno
XFAIL: gdb.base/errno.exp: pthreads-static: print (int) errno
XFAIL: gdb.base/errno.exp: pthreads-static: check errno value from corefile
XFAIL: gdb.base/errno.exp: pthreads-static-macros: print errno
XFAIL: gdb.base/errno.exp: pthreads-static-macros: print (int) errno
XFAIL: gdb.base/errno.exp: pthreads-static-macros: check errno value from corefile
On FreeBSD 13.1, the total number of XFAILs are fewer, and could be
even better still if it had debug info for glibc:
XFAIL: gdb.base/errno.exp: default: print errno
XFAIL: gdb.base/errno.exp: default: check errno value from corefile
XFAIL: gdb.base/errno.exp: macros: print errno
XFAIL: gdb.base/errno.exp: macros: print (int) errno
XFAIL: gdb.base/errno.exp: macros: check errno value from corefile
XFAIL: gdb.base/errno.exp: static-macros: check errno value from corefile
XFAIL: gdb.base/errno.exp: pthreads: print errno
XFAIL: gdb.base/errno.exp: pthreads: check errno value from corefile
XFAIL: gdb.base/errno.exp: pthreads-macros: print errno
XFAIL: gdb.base/errno.exp: pthreads-macros: print (int) errno
XFAIL: gdb.base/errno.exp: pthreads-macros: check errno value from corefile
XFAIL: gdb.base/errno.exp: pthreads-static-macros: check errno value from corefile
Starting with glibc-2.34, most of the pthreads library has been
incorporated into libc, so finding thread-local variables using
libthread_db is possible for several scenarios in which it previously
wasn't. But, prior to this, accessing errno for the default scenario
was a problem. This is borne out by running this new test on Fedora
34, which uses glibc-2.33:
XFAIL: gdb.base/errno.exp: default: print errno
XFAIL: gdb.base/errno.exp: default: print (int) errno
XFAIL: gdb.base/errno.exp: default: check errno value from corefile
XFAIL: gdb.base/errno.exp: macros: check errno value from corefile
XFAIL: gdb.base/errno.exp: static: print errno
XFAIL: gdb.base/errno.exp: static: print (int) errno
XFAIL: gdb.base/errno.exp: static: check errno value from corefile
XFAIL: gdb.base/errno.exp: static-macros: print errno
XFAIL: gdb.base/errno.exp: static-macros: print (int) errno
XFAIL: gdb.base/errno.exp: static-macros: check errno value from corefile
XFAIL: gdb.base/errno.exp: pthreads-macros: check errno value from corefile
XFAIL: gdb.base/errno.exp: pthreads-static: print errno
XFAIL: gdb.base/errno.exp: pthreads-static: check errno value from corefile
XFAIL: gdb.base/errno.exp: pthreads-static-macros: print errno
XFAIL: gdb.base/errno.exp: pthreads-static-macros: print (int) errno
XFAIL: gdb.base/errno.exp: pthreads-static-macros: check errno value from corefile
In the v3 version of this test, Tom de Vries tested on openSUSE Leap
15.5 and found a number of cases which showed a FAIL instead of an
XFAIL. The v4 version of this test fixed those problems. On Leap
15.5, which uses glibc-2.31, with glibc debug info, I now see:
XFAIL: gdb.base/errno.exp: default: print errno
XFAIL: gdb.base/errno.exp: default: print (int) errno
XFAIL: gdb.base/errno.exp: default: check errno value from corefile
XFAIL: gdb.base/errno.exp: macros: check errno value from corefile
XFAIL: gdb.base/errno.exp: static: print errno
XFAIL: gdb.base/errno.exp: static: print (int) errno
XFAIL: gdb.base/errno.exp: static: check errno value from corefile
XFAIL: gdb.base/errno.exp: static-macros: check errno value from corefile
XFAIL: gdb.base/errno.exp: pthreads-macros: check errno value from corefile
XFAIL: gdb.base/errno.exp: pthreads-static: print errno
XFAIL: gdb.base/errno.exp: pthreads-static: print (int) errno
XFAIL: gdb.base/errno.exp: pthreads-static: check errno value from corefile
XFAIL: gdb.base/errno.exp: pthreads-static-macros: check errno value from corefile
On Leap 15.5, with glibc debuginfo missing, the results are a little
worse:
XFAIL: gdb.base/errno.exp: default: print errno
XFAIL: gdb.base/errno.exp: default: print (int) errno
XFAIL: gdb.base/errno.exp: default: check errno value from corefile
XFAIL: gdb.base/errno.exp: macros: print errno
XFAIL: gdb.base/errno.exp: macros: print (int) errno
XFAIL: gdb.base/errno.exp: macros: check errno value from corefile
XFAIL: gdb.base/errno.exp: static: print errno
XFAIL: gdb.base/errno.exp: static: print (int) errno
XFAIL: gdb.base/errno.exp: static: check errno value from corefile
XFAIL: gdb.base/errno.exp: static-macros: check errno value from corefile
XFAIL: gdb.base/errno.exp: pthreads: print errno
XFAIL: gdb.base/errno.exp: pthreads: check errno value from corefile
XFAIL: gdb.base/errno.exp: pthreads-macros: print errno
XFAIL: gdb.base/errno.exp: pthreads-macros: print (int) errno
XFAIL: gdb.base/errno.exp: pthreads-macros: check errno value from corefile
XFAIL: gdb.base/errno.exp: pthreads-static: print errno
XFAIL: gdb.base/errno.exp: pthreads-static: print (int) errno
XFAIL: gdb.base/errno.exp: pthreads-static: check errno value from corefile
XFAIL: gdb.base/errno.exp: pthreads-static-macros: check errno value from corefile
The v5 version of this test fixed failures when testing with
check-read1. (Thanks to Linaro CI for finding these.) I revised the
regular expressions being used so that the failures were eliminated,
but the results mentioned above have not changed.
The v6 version of this test fixes some nits pointed out by both Tom
de Vries and Pedro Alves. One of Pedro's suggestions was to rename the
test from check-errno.exp to errno.exp, so in v5, the name has
changed. Tom also noticed that there were failures when using
--target_board=native-extended-gdbserver. For v6, I've tested on 10
different Linux machines (F38, F39, F39 w/o glibc debuginfo, F39 w/o
static glibc, F40, rawhide, Ubuntu 22.04, Leap 15.5, Leap 15.5 w/o
glibc debuginfo, and Fedora 34) using "make check" and "make check-read1"
using target boards "unix", "native-extended-gdbserver", and
"native-gdbserver", with CC_FOR_TARGET set to both gcc and clang, for
a total of 12 distinct test runs on each machine. I've also tested the
native-only cases on FreeBSD. (Attempting to test against gdbserver
on FreeBSD resulted in hangs while running the test suite.)
The v7 version of this test simplifies the REs used in the uses of
gdb_test_multiple by adding -wrap and removing parts of the REs which
match the GDB prompt. In cases where there was a leading '.*', those
were removed too. Thanks to Pedro for explaining how to use -wrap.
So, bottom line, this test does not introduce any new failures on the
platforms on which I've tested, but the XFAILs are certainly unfortunate.
Some aren't fixable - e.g. when attempting to make a function call while
debugging a core file - but I think that some of them are. I'm using
this new test case as a starting point for investigating problems with
printing errno.
Co-Authored-By: Jan Kratochvil Approved-By: Tom de Vries <tdevries@suse.de>
gas, aarch64: Fixes in texi and tests following faminmax and lut changes
Making two cleanups that came out of the comments from my previous
patches:
1. Fixing `c-aarch64.texi` file so that the AArch64 architecture
extensions are ordered alphabetically.
2. Fixing faminmax test cases so that they follow the existing test
conventions.
Tom Tromey [Thu, 30 May 2024 16:39:17 +0000 (10:39 -0600)]
Move dwarf2_per_bfd::index_addrmap to mapped_gdb_index
dwarf2_per_bfd::index_addrmap is only used by the .gdb_index reader,
so this field can be moved to mapped_gdb_index instead. Then,
cooked_index_functions::find_per_cu can be removed in favor of a
method on the index object.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=31821 Approved-By: Simon Marchi <simon.marchi@efficios.com>
Szabolcs Nagy [Wed, 27 Mar 2024 15:38:06 +0000 (15:38 +0000)]
aarch64: Add DT_RELR support
The logic to decide if an input relocation for a symbol becomes a
particular kind of output relocation is one of the hard to maintain
parts of the bfd ld backend, since it is partially repeated across
elfNN_aarch64_check_relocs (where dynamic relocations are counted per
symbol and input section),
elfNN_aarch64_late_size_sections (where relocation sections are sized
and GOT offsets assigned),
elfNN_aarch64_relocate_section (where most relocations are applied and
output to a relocation section),
elfNN_aarch64_finish_dynamic_symbol (where some of the GOT
relocations are applied and output).
The DT_RELR support adds another layer to this complexity: after the
output relocation sections are sized, so all dynamic relocations are
accounted (in elfNN_aarch64_late_size_sections), we scan the symbols
and input relocations again to decide which ones become relative
relocations that can be packed. The sizes of the relocation sections
are updated accordingly. This logic must be consistent with the code
that applies the relocs later so each relative relocation is emitted
exactly once: either in .rela.* or packed in .relr.dyn.
Sizing of .relr.dyn is done via elfNN_aarch64_size_relative_relocs that
may be called repeatedly whenever the layout changes since an address
change can affect the size of the packed format. Then the final content
is emitted in elfNN_aarch64_finish_relative_relocs, that is called
after the layout is final and before relocations are applied and
emitted. These hooks are only called if DT_RELR is enabled.
We only pack relative relocs that are known to be aligned in the output
during .relr.dyn sizing, the potentially unaligned relative relocs are
emitted normally (in .rela.*, not packed), because the format requires
aligned addresses.
These run after template matching. Therefore it is quite pointless for
them to check all operands, when operand sizes matching across operands
is already known. Exit the loops early in such cases.
In check_byte_reg() also drop a long-stale part of a comment.
Felix Willgerodt [Tue, 21 May 2024 12:54:05 +0000 (14:54 +0200)]
gdb, doc: Fix AVX-512 documentation.
org.gnu.gdb.i386.avx512 adds k registers, but these aren't mentioned in the
docs yet. Fix that.
In addition the documentation describes xmm registers with an `h`
(e.g. xmm16h). I am assuming that we follow the register xml files here,
which don't have the h suffix. So this removes that as well.
was causing problems. Given a release tar file, an attempt to build
and install GDB would give an error like this:
[...]
TEXI2POD gdb.pod
cannot find GDBvn.texi at ../../../gdb-15.0.50.20240508/gdb/doc/../../etc/texi2pod.pl line 251, <GEN0> line 16.
make[5]: *** [Makefile:663: gdb.pod] Error 2
The problem here is how the man pages are built, and how they are
distributed within a release.
Within the development (git) tree, the man page files are not part of
the source tree, these files are built as needed. Within a release
tar file though, the man pages are included. The idea being that a
user can build and install GDB, including getting the man pages,
without having to install the tools needed to generate the man pages.
The man pages are generated in a two step process. First the .texi
file is processed with texi2pod to create a .pod file, then this .pod
file is processed to create the .1 or .5 man file.
Prior to the above commit these two steps were combined into a single
recipe, this meant that when a user performed a build/install from a
release tree all of the dependencies, as well as the final result,
were all present in the source tree, and so nothing needed to be
rebuilt.
However, the above commit split the two steps apart. Now we had a
separate rule for building the .pod files, and the .1/.5 man page
files depended on the relevant .pod file.
As the .pod files are not shipped in a GDB release, this meant that
one of the dependencies of the man page files was now missing. As a
result if a user tried to install from a release tree a rebuild of the
.pod files would be attempted, and if that succeeded then building the
man pages would follow that.
Unfortunately, building the .pod files would fail as the GDBvn.texi
file, though present in the source tree, was not present in the build
tree, which is where it is needed for the .pod file generation to
work.
To fix this, I propose merging the .pod creation and the .1/.5 man
page creation back into a single recipe. Having these two steps split
is probably the "cleaner" solution, but makes it harder for us to
achieve our goal of shipping the prebuilt man page files. I've added
a comment explaining what's going on (such a comment would have
prevented this mistake having been made in the first place).
One possibly weird thing here is that I have left both an
ECHO_TEXI2POD and a ECHO_TEXI2MAN in the rule $(MAN1S) and $(MAN5S)
recipes. This is 100% not going to break anything, these just print
two different progress messages while executing the recipes, but I'm
not sure if this is considered poor style or not. Maybe we're only
supposed to have a single ECHO_* per recipe?
Anyway, even if this is poor style, I figure it really is just a style
thing. We can tweak this later as needed. Otherwise, this commit
should fix the current issue blocking the next GDB release.
Szabolcs Nagy [Thu, 23 May 2024 17:16:57 +0000 (18:16 +0100)]
readelf: Fix symbol display for RELR relocs
Filter symbols before binary searching for the right symbol to display
for a given address, such that only displayable symbols are present and
at most one per address.
The current logic does not handle multiple symbols for the same address
well if some of them are empty, the selected symbol is not stable with
respect to an unrelated symbol table change and on aarch64 often mapping
symbols are displayed which is not useful.
Filtering solves these problems at the cost of a linear scan of the
sorted symbol table.
The heuristic to select the best symbol likely could be improved, this
patch aims to improve symbol display for RELR without complex logic
such that the output is useful and stable for ld tests.
Jan Beulich [Wed, 29 May 2024 08:03:00 +0000 (10:03 +0200)]
x86/Intel: warn about undue mnemonic suffixes
Except for very few insns mnemonic suffixes aren't permitted in Intel
syntax. Warn about such for now, indicating that they will be outright
refused down the road.
While fiddling with testcases to address fallout, drop a few things
which should never have been tested as valid Intel syntax.
Also add a previously missing line to simd-suffix.d.
Alan Modra [Tue, 28 May 2024 04:10:50 +0000 (13:40 +0930)]
PR31796, Internal error in write_function_pdata at obj-coff-seh
PR31796 is the result of lack of aarch64 support in obj-coff-seh.c.
Nick fixed this with commit 73c8603c3f. Make the seh support
consistently warn in future if some archictecture is missing, rather
than giving internal errors.
PR 31796
* config/obj-coff-seh.c (verify_target): New function.
(obj_coff_seh_handler, obj_coff_seh_endproc, obj_coff_seh_proc),
(obj_coff_seh_endprologue): Use it.
Dimitar Dimitrov [Tue, 21 May 2024 15:05:54 +0000 (18:05 +0300)]
ld: pru: Increase the default memory region sizes
The default memory region sizes for PRU were set somewhat arbitrarily to
the sizes of the most popular BeagleBone board with AM33x SoC. But the
PRU toolchain documentation has always instructed to use SoC-specific
spec files to override the defaults and set the correct memory sizes [1].
The small default memory sizes can cause IMEM memory region overflow
even for simple printf("Hello world") programs, as usually done by
Autotools checks. The stdio is simply too big to fit in 8K
instruction memory. This can confuse the check and lead to wrong
feature selection during configure [2].
Fix by bumping the default DMEM and IMEM memory sizes.
There is no need to backport this patch. Issue was caught with a
feature-rich newlib build used for daily CI. The release builds of the
PRU toolchain use stripped newlib configuration, which does not overflow
the IMEM region, even for 8K.
Introduces instructions for the SVE2 lut extension for AArch64. They are documented in the following links:
* luti2: https://developer.arm.com/documentation/ddi0602/2024-03/SVE-Instructions/LUTI2--Lookup-table-read-with-2-bit-indices-?lang=en
* luti4: https://developer.arm.com/documentation/ddi0602/2024-03/SVE-Instructions/LUTI4--Lookup-table-read-with-4-bit-indices-?lang=en
These instructions use new SVE2 vector operands. They are called
SVE_Zm1_23_INDEX, SVE_Zm2_22_INDEX, and Zm3_12_INDEX and they have
1 bit, 2 bit, and 3 bit indices respectively.
The lsb and width of these new operands are the same as many existing
operands but the convention is to give different names to fields that
serve different purpose so we introduced new fields in aarch64-opc.c
and aarch64-opc.h.
We made a design choice for the second operand of the halfword variant of
luti4 with two register tables. We could have either defined a new operand,
like SVE_Znx2, or we could have use the existing operand SVE_ZnxN. With
the new operand, we would need to implement constraints on register
lists based on either operand or opcode flag. With existing operand, we
could just existing constraint checks using opcode flag. We chose
the second approach and went with SVE_ZnxN and added opcode flag to
enforce lengths of vector register list operands. This way, we can reuse
the existing constraint check logic.
Introduces instructions for the Advanced SIMD lut extension for AArch64. They are documented in the following links:
* luti2: https://developer.arm.com/documentation/ddi0602/2024-03/SIMD-FP-Instructions/LUTI2--Lookup-table-read-with-2-bit-indices-?lang=en
* luti4: https://developer.arm.com/documentation/ddi0602/2024-03/SIMD-FP-Instructions/LUTI4--Lookup-table-read-with-4-bit-indices-?lang=en
These instructions needed definition of some new operands. We will first
discuss operands for the third operand of the instructions and then
discuss a vector register list operand needed for the second operand.
The third operands are vectors with bit indices and without type
qualifiers. They are called Em_INDEX1_14, Em_INDEX2_13, and Em_INDEX3_12
and they have 1 bit, 2 bit, and 3 bit indices respectively. For these
new operands, we defined new parsing case branch. The lsb and width of
these operands are the same as many existing but the convention is to
give different names to fields that serve different purpose so we
introduced new fields in aarch64-opc.c and aarch64-opc.h for these new
operands.
For the second operand of these instructions, we introduced a new
operand called LVn_LUT. This represents a vector register list with
stride 1. We defined new inserter and extractor for this new operand and
it is encoded in FLD_Rn. We are enforcing the number of registers in the
reglist using opcode flag rather than operand flag as this is what other
SIMD vector register list operands are doing. The disassembly also uses
opcode flag to print the correct number of registers.
Richard Earnshaw [Thu, 23 May 2024 15:25:51 +0000 (16:25 +0100)]
opcodes: add a .gitattributes file for aarch64 autogenerated file exceptions
The autogenerated files in opcodes use spaces for indentation.
Changing that would be a lot of work to little benefit, so add a local
override to the white-space rules, so patches apply cleanly.
projects / thirdparty / binutils-gdb.git / log
