gold/
PR gold/18780
* arm.cc (Target_arm::do_make_data_plt): Choose PLT generator based
on value of --long-plt flag.
(Output_data_plt_arm_standard::do_get_plt_entry_size): Moved to
Output_data_plt_arm_short.
(Output_data_plt_arm_standard::do_fill_plt_entry): Likewise.
(Output_data_plt_arm_standard::plt_entry): Likewise.
(Output_data_plt_arm_standard::do_fill_first_plt_entry): Fix
variable reference.
(Output_data_plt_arm_short): New class.
(Output_data_plt_arm_short::do_fill_plt_entry): Error out on too large
PLT offsets instead of asserting.
(Output_data_plt_arm_long): New class.
* options.h (General_options): Define --long-plt flag.
Note: this applies on top of:
[PATCH] Remove support for LinuxThreads and vendor 2.4 kernels w/ backported NPTL
https://sourceware.org/ml/gdb-patches/2015-12/msg00214.html
We try to avoid using libthread_db.so to list threads in the inferior
when debugging live processes, but the code that decides whether to
use it decides incorrectly if you have more than one inferior, and the
current inferior doesn't have execution yet. The result is visible
as:
(gdb) add-inferior
Added inferior 2
(gdb) inferior 2
[Switching to inferior 2 [<null>] (<noexec>)]
(gdb) info inferiors
Num Description Executable
1 process 15397 /home/pedro/gdb/tests/threads
* 2 <null>
(gdb) info threads
Cannot find new threads: generic error
(gdb)
Fix this by checking whether each inferior has execution rather than
just the current inferior.
By moving the core updating to linux-nat.c's update_thread_list
implementation, this also ends up fixing the
lwp-last-seen-running-on-core updating in the case we're debugging a
program that uses raw clone rather than pthreads, as linux-thread-db.c
isn't pushed in the target stack in that scenario.
Tested on x86_64 Fedora 20.
gdb/ChangeLog:
2015-12-17 Pedro Alves <palves@redhat.com>
PR threads/19354
* linux-nat.c (linux_nat_update_thread_list): Update process cores
each lwp was last seen running on here.
* linux-thread-db.c (update_thread_core): Delete.
(thread_db_update_thread_list_td_ta_thr_iter): Rename to ...
(thread_db_update_thread_list): ... this. Skip inferiors with
execution. Also call the target beneath.
(thread_db_update_thread_list): Delete.
gdb/testsuite/ChangeLog:
2015-12-17 Pedro Alves <palves@redhat.com>
PR threads/19354
* gdb.multi/info-threads.exp: New file.
Since we now rely on PTRACE_EVENT_CLONE being available (added in
Linux 2.5.46), we're relying on NPTL.
This commit removes the support for older LinuxThreads, as well as the
workarounds for vendor 2.4 kernels with NPTL backported.
- Rely on tkill being available.
- Assume gdb doesn't get cancel signals.
- Remove code that checks the LinuxThreads restart and cancel signals
in the inferior.
- Assume that __WALL is available.
- Assume that non-leader threads report WIFEXITED.
- Thus, no longer need to send signal 0 to check whether threads are
still alive.
- Update comments throughout.
Tested on x86_64 Fedora 20, native and gdbserver.
gdb/ChangeLog:
* configure.ac: Remove tkill checks.
* configure, config.in: Regenerate.
* linux-nat.c: Remove HAVE_TKILL_SYSCALL check. Update top level
comments.
(linux_nat_post_attach_wait): Remove 'cloned' parameter. Use
__WALL.
(attach_proc_task_lwp_callback): Don't set the cloned flag.
(linux_nat_attach): Adjust.
(kill_lwp): Remove HAVE_TKILL_SYSCALL check. No longer fall back
to 'kill'.
(linux_handle_extended_wait): Use __WALL. Don't set the cloned
flag.
(wait_lwp): Use __WALL. Update comments.
(running_callback, stop_and_resume_callback): Delete.
(linux_nat_filter_event): Don't stop and resume all lwps. Don't
check if the event LWP has previously exited.
(check_zombie_leaders): Update comments.
(linux_nat_wait_1): Use __WALL.
(kill_wait_callback): Don't handle clone processes separately.
Use __WALL instead.
(linux_thread_alive): Delete.
(linux_nat_thread_alive): Return true as long as the LWP is in the
LWP list.
(linux_nat_update_thread_list): Assume the kernel supports
PTRACE_EVENT_CLONE.
(get_signo): Delete.
(lin_thread_get_thread_signals): Remove LinuxThreads references.
No longer check __pthread_sig_restart / __pthread_sig_cancel in
the inferior.
* linux-nat.h (struct lwp_info) <cloned>: Delete field.
* linux-thread-db.c: Update comments.
(_initialize_thread_db): Remove LinuxThreads references.
* nat/linux-waitpid.c (my_waitpid): No longer emulate __WALL.
Pass down flags unmodified.
* linux-waitpid.h (my_waitpid): Update documentation.
gdb/gdbserver/ChangeLog:
* linux-low.c (linux_kill_one_lwp): Remove references to
LinuxThreads.
(kill_lwp): Remove HAVE_TKILL_SYSCALL check. No longer fall back
to 'kill'.
(linux_init_signals): Delete.
(initialize_low): Adjust.
* thread-db.c (thread_db_init): Remove LinuxThreads reference.
Hi,
AddressSanitizer reports an error like this,
(gdb) PASS: gdb.base/call-ar-st.exp: continue to tbreak9
print print_long_arg_list(a, b, c, d, e, f, *struct1, *struct2, *struct3, *struct4, *flags, *flags_combo, *three_char, *five_char, *int_char_combo, *d1, *d2, *d3, *f1, *f2, *f3)
=================================================================
==6236==ERROR: AddressSanitizer: heap-buffer-overflow on address 0x60200008eb50 at pc 0x89e432 bp 0x7fffa3df9080 sp 0x7fffa3df9078
READ of size 5 at 0x60200008eb50 thread T0
#0 0x89e431 in memory_xfer_partial gdb/target.c:1264
#1 0x89e6c7 in target_xfer_partial gdb/target.c:1320
#2 0x89f267 in target_write_partial gdb/target.c:1595^M
#3 0x8a014b in target_write_with_progress gdb/target.c:1889^M
#4 0x8a0262 in target_write gdb/target.c:1914^M
#5 0x89ee59 in target_write_memory gdb/target.c:1492^M
#6 0x9a1c74 in write_memory gdb/corefile.c:393^M
#7 0x467ea5 in aarch64_push_dummy_call gdb/aarch64-tdep.c:1388
The problem is that an instance of stack_item_t is created to adjust
stack for alignment, the item.len is correct, but item.data is buf,
which is wrong, because item.len can be greater than the length of
buf. This patch sets item.data to NULL, and only update sp (no
inferior memory writes on stack for this item).
gdb:
2015-12-17 Yao Qi <yao.qi@linaro.org>
* aarch64-tdep.c (struct stack_item_t): Update comments.
(pass_on_stack): Set item.data to NULL.
(aarch64_push_dummy_call): Call write_memory if si->data
isn't NULL.
There is currently a problem in the way in which we produce
build attributes for simple assembler files that have armv8-a
instructions.
In these case we need to generate TAG_ISA_THUMB_Use to be Thumb-2
and set the architecture profile to be 'A' rather than not
setting architecture profile to be 'A' and setting TAG_ISA_THUMB_Use
to be Thumb-1.
This is a pre-requisite for any v8-m patches that have been posted.
arm-none-eabi gas testsuite run. no regressions.
2015-12-17 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
* gas/config/tc-arm.c (aeabi_set_public_attributes): Adjust
TAG_ARCH_profile for armv8-a.
* gas/testsuite/gas/arm/armv8a-automatic-hlt.d: New test.
* gas/testsuite/gas/arm/armv8a-automatic-hlt.s: New test.
* gas/testsuite/gas/arm/armv8a-automatic-lda.d: New test.
* gas/testsuite/gas/arm/armv8a-automatic-lda.s: New test.
2015-12-16 Mickael Guene <mickael.guene@st.com>
bfd/
* bfd-in2.h: Regenerate.
* reloc.c: Add new relocations.
* libbfd.h (bfd_reloc_code_real_names): Add new relocations
display names.
* elf32-arm.c (elf32_arm_howto_table_1): Add HOWTO for new
relocations.
(elf32_arm_reloc_map): Add bfd/arm mapping for new relocations.
(elf32_arm_final_link_relocate): Implement new relocations
resolution.
gas/
* doc/c-arm.texi: Add documentation about new directives
* config/tc-arm.c (group_reloc_table): Add mapping between gas
syntax and new relocations.
(do_t_add_sub): Keep new relocations for add operand.
(do_t_mov_cmp): Keep new relocations for mov operand.
(insns): Use 'shifter operand with possible group relocation'
operand parse code for movs operand.
(md_apply_fix): Implement mov and add encoding when new
relocations on them.
(tc_gen_reloc): Add new relocations.
(arm_fix_adjustable): Since offset has a limited range ([0:255])
we disable adjust_reloc_syms() for new relocations.
gas/testsuite/
* gas/arm/adds-thumb1-reloc-local.d: New
* gas/arm/adds-thumb1-reloc-local.s: New
* gas/arm/movs-thumb1-reloc-local.d: New
* gas/arm/movs-thumb1-reloc-local.s: New
include/
* elf/arm.h: Add new arm relocations.
ld/testsuite/
* ld-arm/arm-elf.exp (armelftests_common): Add new relocations
tests.
* ld-arm/thumb1-adds.d: New
* ld-arm/thumb1-adds.s: New
* ld-arm/thumb1-movs.d: New
* ld-arm/thumb1-movs.s: New
Ref: https://sourceware.org/ml/gdb/2015-12/msg00024.html
We have code in configure.ac that tries to detect whether the compiler
supports each warning and suppress it if not, but that doesn't work
with "-Wno-" options, because gcc doesn't error out for
-Wno-unknown-warning unless other diagnostics are being produced.
See https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html.
Handle this by checking whether -Wfoo works when we actually want
-Wno-foo.
gdb/ChangeLog:
2015-12-16 Pedro Alves <palves@redhat.com>
* configure.ac (compiler warning flags): When testing a
-Wno-foo option, check whether -Wfoo works instead.
* configure: Regenerate.
gdb/gdbserver/ChangeLog:
2015-12-16 Pedro Alves <palves@redhat.com>
* configure.ac (compiler warning flags): When testing a
-Wno-foo option, check whether -Wfoo works instead.
* configure: Regenerate.
Ref: https://sourceware.org/ml/gdb/2015-12/msg00014.html
Fixes the build in C++ mode with g++ 4.4:
gdb/btrace.h: In function ‘size_t VEC_btrace_insn_s_embedded_size(int)’:
gdb/btrace.h:84: error: invalid access to non-static data member ‘VEC_btrace_insn_s::vec’ of NULL object
gdb/btrace.h:84: error: (perhaps the ‘offsetof’ macro was used incorrectly)
gdb/btrace.h: In function ‘VEC_btrace_insn_s* VEC_btrace_insn_s_alloc(int)’:
gdb/btrace.h:84: error: invalid access to non-static data member ‘VEC_btrace_insn_s::vec’ of NULL object
gdb/btrace.h:84: error: (perhaps the ‘offsetof’ macro was used incorrectly)
gdb/btrace.h: In function ‘VEC_btrace_insn_s* VEC_btrace_insn_s_copy(VEC_btrace_insn_s*)’:
gdb/btrace.h:84: error: invalid access to non-static data member ‘VEC_btrace_insn_s::vec’ of NULL object
gdb/btrace.h:84: error: (perhaps the ‘offsetof’ macro was used incorrectly)
gdb/btrace.h: In function ‘VEC_btrace_insn_s* VEC_btrace_insn_s_merge(VEC_btrace_insn_s*, VEC_btrace_insn_s*)’:
gdb/btrace.h:84: error: invalid access to non-static data member ‘VEC_btrace_insn_s::vec’ of NULL object
gdb/btrace.h:84: error: (perhaps the ‘offsetof’ macro was used incorrectly)
gdb/btrace.h: In function ‘int VEC_btrace_insn_s_reserve(VEC_btrace_insn_s**, int, const char*, unsigned int)’:
gdb/btrace.h:84: error: invalid access to non-static data member ‘VEC_btrace_insn_s::vec’ of NULL object
gdb/btrace.h:84: error: (perhaps the ‘offsetof’ macro was used incorrectly)
gdb/ChangeLog:
2015-12-16 Pedro Alves <palves@redhat.com>
* common/vec.h (vec_offset): New macro.
(DEF_VEC_ALLOC_FUNC_I, DEF_VEC_ALLOC_FUNC_O): Use it instead of
offsetof.
To support thumb1 execute-only code we need to support four new
relocations (R_ARM_THM_ALU_ABS_G0_NC, R_ARM_THM_ALU_ABS_G1_NC,
R_ARM_THM_ALU_ABS_G2_NC and R_ARM_THM_ALU_ABS_G3_NC).
These relocations allow the static linker to finalize construction
of symbol address.
Typical sequence of code to get address of the symbol foo is then
the following :
movs r3, #:upper8_15:#foo
lsls r3, #8
adds r3, #:upper0_7:#foo
lsls r3, #8
adds r3, #:lower8_15:#foo
lsls r3, #8
adds r3, #:lower0_7:#foo
This will give following sequence of text and relocations after
assembly :
4: 2300 movs r3, #0
4: R_ARM_THM_ALU_ABS_G3_NC foo
6: 021b lsls r3, r3, #8
8: 3300 adds r3, #0
8: R_ARM_THM_ALU_ABS_G2_NC foo
a: 021b lsls r3, r3, #8
c: 3300 adds r3, #0
c: R_ARM_THM_ALU_ABS_G1_NC foo
e: 021b lsls r3, r3, #8
10: 3300 adds r3, #0
10: R_ARM_THM_ALU_ABS_G0_NC foo
ARMv8.1 includes CRC as a required extension but the +crc feature isn't
enabled by -march=armv8.1-a as it should be. This patch fixes that.
opcode/include
2015-12-15 Matthew Wahab <matthew.wahab@arm.com>
* arm.h (ARM_ARCH_V8_1A): Add the CRC_EXT_ARMV8 co-processor
feature macro.
(ARM_ARCH_V8_2A): Likewise.
Change-Id: Id1fe0e6fa51dede19d61e1fd08e68628ea1b1e9e
Some tests are skipped on aarch64 unexpectedly because arg0exp isn't
set. This patch is to set arg0exp to "$x0" for aarch64.
gdb/testsuite:
2015-12-15 Yao Qi <yao.qi@linaro.org>
* gdb.trace/ftrace.exp: Set arg0exp to "$x0" if target
is aarch64*-*-*.
The re-writing logic in _bfd_coff_final_link() overwrote the ".file"
part of the symbol table entry, due to not coping with the auxiliary
entry generated in all cases.
Note that while I would have wanted to add a test case,
(a) I didn't spot any one testing the base functionality here, and
(b) I wasn't able to figure out proper conditionals to use in e.g.
ld-elf/elf.exp to check for the necessary PE/PE+ support (which
varies by target).
Fix occurrences of left-shifting negative constants in C code.
sim/arm/ChangeLog:
* thumbemu.c (handle_T2_insn): Fix left shift of negative value.
* armemu.c (handle_v6_insn): Likewise.
sim/avr/ChangeLog:
* interp.c (sign_ext): Fix left shift of negative value.
sim/mips/ChangeLog:
* micromips.igen (process_isa_mode): Fix left shift of negative
value.
sim/msp430/ChangeLog:
* msp430-sim.c (get_op, put_op): Fix left shift of negative value.
sim/v850/ChangeLog:
* simops.c (v850_bins): Fix left shift of negative value.
* aarch64/simulator.c (system_get): New function. Provides read
access to the dczid system register.
(do_mrs): New function - implements the MRS instruction.
(dexSystem): Call do_mrs for the MRS instruction. Halt on
unimplemented system instructions.
This patch implements documentation updates for target remote mode fork and
exec events. A summary of the rationale for the changes made here:
* Connecting to a remote target -- explain that the two protocols exist.
* Connecting in target remote mode -- explain invoking gdbserver for target
remote mode, and move remote-specific text from original "Connecting to a
remote target" section.
* Connecting in target extended-remote mode -- promote this section from
"Using the gdbserver Program | Running gdbserver | Multi-Process Mode for
gdbserver". Put it next to the target remote mode section.
* Host and target files -- collect paragraphs dealing with how to locate
symbol files from original sections "Connecting to a remote target" and
"Using the gdbserver program | Connecting to gdbserver".
* Steps for connecting to a remote target -- used to be "Using the
gdbserver program | Connecting to gdbserver"
* Remote connection commands -- used to be the bulk of "Connecting to a
remote target". Added "target extended-remote" commands and information.
gdb/ChangeLog:
* NEWS: Announce fork and exec event support for target remote.
gdb/doc/ChangeLog:
* gdb.texinfo (Forks): Correct Linux kernel version where
fork and exec events are supported, add notes about support
of these events in target remote mode.
(Connecting): Reorganize and clarify distinctions between
target remote, extended-remote, and multiprocess.
Reorganize related text from separate sections into new
sections.
(Server): Note effects of target extended-remote mode.
Delete section on Multi-Process Mode for gdbserver.
Move some text to "Connecting" node.
This patch implements support for fork and exec events with target remote
mode Linux targets. For such targets with Linux kernels 2.5.46 and later,
this enables follow-fork-mode, detach-on-fork and fork and exec
catchpoints.
The changes required to implement this included:
* Don't exit from gdbserver if there are still active inferiors.
* Allow changing the active process in remote mode.
* Enable fork and exec events in remote mode.
* Print "Ending remote debugging" only when disconnecting.
* Combine remote_kill and extended_remote_kill into a single function
that can handle the multiple inferior case for target remote. Also,
the same thing for remote_mourn and extended_remote_mourn.
* Enable process-style ptids in target remote.
* Remove restriction on multiprocess mode in target remote.
gdb/gdbserver/ChangeLog:
* server.c (process_serial_event): Don't exit from gdbserver
in remote mode if there are still active inferiors.
gdb/ChangeLog:
* inferior.c (number_of_live_inferiors): New function.
(have_live_inferiors): Use number_of_live_inferiors in place
of duplicate code.
* inferior.h (number_of_live_inferiors): Declare new function.
* remote.c (set_general_process): Remove restriction on target
remote mode.
(remote_query_supported): Likewise.
(remote_detach_1): Exit in target remote mode only when there
is just one live inferior left.
(remote_disconnect): Unpush the target directly instead of
calling remote_mourn.
(remote_kill): Rewrite function to handle both target remote
and extended-remote. Call remote_kill_k.
(remote_kill_k): New function.
(extended_remote_kill): Delete function.
(remote_mourn, extended_remote_mourn): Combine functions into
one, remote_mourn, and enable extended functionality for target
remote.
(remote_pid_to_str): Enable "process" style ptid string for
target remote.
(remote_supports_multi_process): Remove restriction on target
remote mode.
This patch updates tests for fork and exec events in target remote mode.
In the majority of cases this was a simple matter of removing some code
that disabled the test for target remote. In a few cases the test needed
to be disabled; in those cases the gdb_protocol was checked instead of
using the [is_remote target] etc.
In a couple of cases we needed to use clean_restart, since target remote
doesn't support the run command, and in one case we had to modify an expect
expression to allow for a "multiprocess-style" ptid.
Tested with the patch that implemented target remote mode fork and exec
event support.
gdb/testsuite/ChangeLog:
* gdb.base/execl-update-breakpoints.exp (main): Enable for target
remote.
* gdb.base/foll-exec-mode.exp (main): Disable for target remote.
* gdb.base/foll-exec.exp (main): Enable for target remote.
* gdb.base/foll-fork.exp (main): Likewise.
* gdb.base/foll-vfork.exp (main): Likewise.
* gdb.base/multi-forks.exp (main): Likewise, and use clean_restart.
(proc continue_to_exit_bp_loc): Use clean_restart.
* gdb.base/pie-execl.exp (main): Disable for target remote.
* gdb.base/watch-vfork.exp (main): Enable for target remote.
* gdb.mi/mi-nsthrexec.exp (main): Likewise.
* gdb.threads/execl.exp (main): Likewise.
* gdb.threads/fork-child-threads.exp (main): Likewise.
* gdb.threads/fork-plus-threads.exp (main): Disable for target
remote.
* gdb.threads/fork-thread-pending.exp (main): Enable for target
remote.
* gdb.threads/linux-dp.exp (check_philosopher_stack): Allow
pid.tid style ptids, instead of just tid.
* gdb.threads/thread-execl.exp (main): Enable for target remote.
* gdb.threads/watchpoint-fork.exp (main): Likewise.
* gdb.trace/report.exp (use_collected_data): Allow pid.tid style
ptids, instead of just tid.
ARMv8.2 adds 16-bit floating point operations as an optional extension
to the floating point and Adv.SIMD support. This patch extends
instructions in the group Adv.SIMD Scalar Shift By Immediate to support
FP16, making this support available when +simd+fp16 is enabled.
The extended instructions are: SCVTF, FCVTZS, UCVTF and FCVTZU.
The general form for these instructions is
<OP> <Hd>, <Hs>, #<imm>
gas/testsuite/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* gas/aarch64/advsimd-fp16.d: Update expected output.
* gas/aarch64/advsimd-fp16: Add tests for Adv.SIMD scalar shift
by immediate instructions.
opcodes/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* aarch64-asm-2.c: Regenerate.
* aarch64-dis-2.c: Regenerate.
* aarch64-opc-2.c: Regenerate.
* aarch64-tbl.h (QL_SSHIFT_H): New.
(aarch64_opcode_table): Add fp16 versions of scvtf, fcvtzs, ucvtf
and fcvtzu to the Adv.SIMD scalar shift by immediate group.
Change-Id: I40506496f52dd96909e7344f243b38a1870df7ff
ARMv8.2 adds 16-bit floating point operations as an optional extension
to the floating point and Adv.SIMD support. This patch extends
instructions in the group Adv.SIMD Shift By Immediate to support FP16,
making this support available when +simd+fp16 is enabled.
The new instructions legal make some uses of the 4h vector type that had
been invalid. This patch adjusts a test that checks for these uses.
The extended instructions are: SCVTF, FCVTZS, UCVTF and FCVTZU.
The general form for these instructions is
<OP> <Vd>.<T>, <Vs>.<T>, #<imm>
where T is 4h or 8h.
gas/testsuite/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* gas/aarch64/advsimd-fp16.d: Update expected output.
* gas/aarch64/advsimd-fp16.s: Add tests for Adv.SIMD across lanes
instructions.
* gas/aarch64/illegal.d: Update expected output.
* gas/aarch64/illegal.s: Replace tests for illegal use of 'h'
specifier.
opcodes/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* aarch64-asm-2.c: Regenerate.
* aarch64-dis-2.c: Regenerate.
* aarch64-opc-2.c: Regenerate.
* aarch64-tbl.h (QL_VSHIFT_H): New.
(aarch64_opcode_table): Add fp16 versions of scvtf, fcvtzs, ucvtf
and fcvtzu to the Adv.SIMD shift by immediate group.
Change-Id: I3480f63883d54db46562573185da6982f2365ee8
ARMv8.2 adds 16-bit floating point operations as an optional extension
to the floating point and Adv.SIMD support. This patch adds FP16
instructions to the group Adv.SIMD Scalar Pairwise, making them
available when +simd+fp16 is enabled.
The instructions added are: FMAXNMP, FADDP, FMAXP, FMINNMP and FMINP
The general form for these instructions is
<OP> <Hd>, <V>.<T>
where T is 4h or 8h.
gas/testsuite/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* gas/aarch64/advsimd-fp16.d: Update expected output.
* gas/aarch64/advsimd-fp16.s: Add tests for Adv.SIMD Scalar
Pairwise instructions.
opcodes/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* aarch64-asm-2.c: Regenerate.
* aarch64-dis-2.c: Regenerate.
* aarch64-opc-2.c: Regenerate.
* aarch64-tbl.h (QL_SISD_PAIR_H): New.
(aarch64_opcode_table): Add fp16 versions of fmaxnmp, faddp,
fmaxp, fminnmp, fminp to the Adv.SIMD scalar pairwise group.
Change-Id: I19937ede3441b66dd0f940269ece895b17d3c345
ARMv8.2 adds 16-bit floating point operations as an optional extension
to the floating point and Adv.SIMD support. The FP16 additions to the
scalar pairwise group introduce a new vector type, 2H. This patch adds
support for this vector type to binutils.
The patch adds a new operand qualifier to the enum
aarch64.h:aarch64_opnd_qualifier. This interferes with the calculation
used by aarch64-dis.c:get_vreg_qualifier_from_value, called when
decoding an instruction. Since the new vector type is only used in FP16
scalar pairwise instructions which do not require the function, this
patch adjusts the function to ignore the new qualifier.
gas/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* config/tc-aarch64.c (parse_neon_type_for_operand): Adjust to
take into account new vector type 2H.
(vectype_to_qualifier): Likewise.
include/opcode/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* aarch64.h (enum aarch64_opnd_qualifier): Add
AARCH64_OPND_QLF_V_2H.
opcodes/
2015-12-14 Matthew Wahab <matthew.wahab@arm.coM>
* aarch64-dis.c (get_vreg_qualifier_from_value): Update comment
and adjust calculation to ignore qualifier for type 2H.
* aarch64-opc.c (aarch64_opnd_qualifier): Add "2H".
Change-Id: Idf9a3694732962c80fde04f08c7304de9164f126
ARMv8.2 adds 16-bit floating point operations as an optional extension
to the floating point and Adv.SIMD support. The FP16 additions to the
scalar pairwise group introduce a new vector type. This patch reworks
code in the assembler to allow the addition of the new type.
The new vector type requires the addtion of a new operand qualifier to
the enum aarch64_opnd_qualifier which is defined
include/opcodes/aarch64.h, in the group prefixed by AARCH64_OPN_QLF_V_.
The correctness of the GAS utility function
tc-aarch64.c:vectype_to_qualifier is heavily dependent on the number and
ordering of this group. In particular, it makes assumptions about the
positions of the members of the group that are not true if a qualifier
for type 2H is added before the qualifier for 4H.
This patch reworks the function to weaken its assumptions, making it
calculate positions in the group from the type (B, H, S, D, Q) and
register width.
gas/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* config/tc-aarch64.c (vectype_to_qualifier): Calculate operand
qualifier from per-type base and offet.
Change-Id: I95535864e342a6dec46f69d2696b3900a008f0b1
ARMv8.2 adds 16-bit floating point operations as an optional extension
to the floating point and Adv.SIMD support. This patch adds an FP16
instruction to the group Adv.SIMD Modified Immediate, making it
available when +simd+fp16 is enabled.
The instruction added is: FMOV.
The form of this instructions is
<OP> <Hd>, #<imm>
gas/testsuite/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* gas/aarch64/advsimd-fp16.d: Update expected output.
* gas/aarch64/advsimd-fp16.s: Add tests for Adv.SIMD modified immediate
instructions.
opcodes/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* aarch64-asm-2.c: Regenerate.
* aarch64-dis-2.c: Regenerate.
* aarch64-opc-2.c: Regenerate.
* aarch64-tbl.h (QL_SIMD_IMM_H): New.
(aarch64_opcode_table): Add fp16 version of fmov to the Adv.SIMD
modified immediate group.
Change-Id: Ic66af44c494e6a53fb1cf01c372cdc62d12643e2
ARMv8.2 adds 16-bit floating point operations as an optional extension
to the floating point and Adv.SIMD support. This patch adds FP16
instructions to the group Adv.SIMD Across Lanes, making them available
when +simd+fp16 is enabled.
The instructions added are: FMAXNMV, FMAXV, FMINNMV and FMINV.
The general form for these instructions is
<OP> <Hd>, <V>.<T>
where T is 4h or 8h.
The new instructions valid make uses of the 8H and 4H that were
previously illegal. The patch adjusts a test for illegal uses of vector
types to take this into account.
gas/testsuite/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* gas/aarch64/advsimd-fp16.d: Update expected output.
* gas/aarch64/advsimd-fp16.s: Add tests for Adv.SIMD across lanes
instructions.
* gas/aarch64/illegal.d: Update expected output.
* gas/aarch64/illegal.s: Replace test for illegal use of 'h'
specifier.
opcodes/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* aarch64-asm-2.c: Regenerate.
* aarch64-dis-2.c: Regenerate.
* aarch64-opc-2.c: Regenerate.
* aarch64-tbl.h (QL_XLANES_FP_H): New.
(aarch64_opcode_table): Add fp16 versions of fmaxnmv, fmaxv,
fminnmv, fminv to the Adv.SIMD across lanes group.
Change-Id: Ib9a47e867f55e0272c2446eb7e16837503d2f94c
ARMv8.2 adds 16-bit floating point operations as an optional extension
to the floating point and Adv.SIMD support. This patch adds FP16
instructions to the group Scalar Indexed Element, making them available
when +simd+fp16 is enabled.
The instructions added are: FMLA, FMLS, FMUL and FMULX.
The general form for these instructions is
<OP> <Hd>, <Hs>, <V>.h[<idx>]
gas/testsuite/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* gas/aarch64/advsimd-fp16.d: Update expected output.
* gas/aarch64/advsimd-fp16.s: Add tests for scalar indexed element
instructions.
opcodes/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* aarch64-asm-2.c: Regenerate.
* aarch64-dis-2.c: Regenerate.
* aarch64-opc-2.c: Regenerate.
* aarch64-tbl.h (aarch64_opcode_table): Add fp16 versions of fmla,
fmls, fmul and fmulx to the scalar indexed element group.
Change-Id: I6a4ee20a9ae1019b89d0fd05da55222f267c5627
ARMv8.2 adds 16-bit floating point operations as an optional extension
to the floating point and Adv.SIMD support. This patch adds FP16
instructions to the group Vector Indexed Element, making them available
when +simd+fp16 is enabled.
The instructions added are: FMLA, FMLS, FMUL and FMULX.
The general form for these instructions is
<OP> <V>.<T>, <V>.<T>, <V>.h[<idx>]
where T is 4h or 8h
gas/testsuite/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* gas/aarch64/advsimd-fp16.d: Update expected output.
* gas/aarch64/advsimd-fp16.s: Add tests for vector indexed element
instructions.
opcodes/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* aarch64-asm-2.c: Regenerate.
* aarch64-dis-2.c: Regenerate.
* aarch64-opc-2.c: Regenerate.
* aarch64-tbl.h (QL_ELEMENT_FP_H): New.
(aarch64_opcode_table): Add fp16 versions of fmla, fmls, fmul and
fmulx to the vector indexed element group.
Change-Id: Ib70cd4eaa6ea2938f84ac41f31d72644dbb0ceb4
ARMv8.2 adds 16-bit floating point operations as an optional extension
to the floating point and Adv.SIMD support. This patch adds FP16
instructions to the group Scalar Two Register Misc, making them
available when +simd+fp16 is enabled.
The instructions added are: FCMGT, FCMGE, FCMEQ, FCMLE, FCMLT, FCVTNS,
FCVTNU, FCVTPS, FCVTPU, FCVTMS, FCVTMU, FCVTZS, FCVTZU, FCVTAS, FCVTAU,
SCVTF, UCVTF, FRECPE, FRSQRTE and FRECPX.
The general form for these instructions is
<OP> <Hd>, <Hs>
or
<OP> <Hd>, <Hs>, #0.0
Tested the series for aarch64-none-linux-gnu with cross-compiled
check-binutils and check-gas.
gas/testsuite/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* gas/aarch64/advsimd-fp16.d: Update expected output.
* gas/aarch64/advsimd-fp16.s: Add tests for scalar two register misc.
instructions.
opcodes/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* aarch64-asm-2.c: Regenerate.
* aarch64-dis-2.c: Regenerate.
* aarch64-opc-2.c: Regenerate.
* aarch64-tbl.h (QL_SISD_FCMP_H_0): new.
(QL_S_2SAMEH): New.
(aarch64_opcode_table): Add fp16 versions of fcvtns, fcvtms,
fcvtas, scvtf, fcmgt, fcmeq, fcmlt, fcvtps, fcvtzs, frecpe,
frecpx, fcvtnu, fcvtmu, fcvtau, ucvtf, fcmge, fcmle, fcvtpu,
fcvtzu and frsqrte to the scalar two register misc. group.
Change-Id: I19b25baae33027ce1bade68cc8dc47a4321d045c
ARMv8.2 adds 16-bit floating point operations as an optional extension
to the floating point and Adv.SIMD support. This patch adds FP16
instructions to the group Vector Two Register Misc, making them
available when +simd+fp16 is enabled.
The instructions added are: FCMGT, FCMGE, FCMEQ, FCMLE, FCMLT, FABS,
FNEG, FRINTN, FRINTA, FRINTP, FRINTM, FRINTX, FRINTZ, FRINTI, FCVTNS,
FCVTNU, FCVTPS, FCVTPU, FCVTMS, FCVTMU, FCVTZS, FCVTZU, FCVTAS, FCVTAU,
SCVTF, UCVTF, FRECPE, FRSQRTE and FSQRT.
The general form for these instructions is
<OP> <Vd>.<T>, <Vs>.<T>
where T is 4h or 8h.
gas/testsuite/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* gas/aarch64/advsimd-fp16.d: Update expected output.
* gas/aarch64/advsimd-fp16.s: Add tests for vector two register misc.
instructions.
opcodes/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* aarch64-asm-2.c: Regenerate.
* aarch64-dis-2.c: Regenerate.
* aarch64-opc-2.c: Regenerate.
* aarch64-tbl.h (QL_V2SAMEH): New.
(aarch64_opcode_table): Add fp16 versions of frintn, frintm,
fcvtns, fcvtms, fcvtas, scvtf, fcmgt, fcmeq, fcmlt, fabs, frintp,
frintz, fcvtps, fcvtzs, frecpe, frinta, frintx, fcvtnu, fcvtmu,
fcvtau, ucvtf, fcmge, fcmle, fneg, frinti, fcvtpu, fcvtzu, frsqrte
and fsqrt to the vector register misc. group.
Change-Id: I0267511a7f7ea14247504d29fe4752e84c9af9ad
ARMv8.2 adds 16-bit floating point operations as an optional extension
to the floating point and Adv.SIMD support. This patch adds FP16
instructions to the group Scalar Three Register Same, making them
available when +simd+fp16 is enabled.
The instructions added are: FABD, FMULX, FCMEQ, FCMGE, FCMGT, FACGE,
FACGT, FRECPS and FRSQRTS.
The general form for these instructions is
<OP> <Hd>, <Hs>, <Hm>
gas/testsuite/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* gas/aarch64/advsimd-fp16.d: Update expected output.
* gas/aarch64/advsimd-fp16.s: Add tests for scalar three register same
instructions.
opcodes/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* aarch64-asm-2.c: Regenerate.
* aarch64-dis-2.c: Regenerate.
* aarch64-opc-2.c: Regenerate.
* aarch64-tbl.h (aarch64_opcode_table): Add fp16 versions of
fmulx, fcmeq, frecps, frsqrts, fcmge, facge, fabd, fcmgt and
facgt to the scalar three same group.
Change-Id: I155eb8d7c1e9a7c89d691d7e4aae83be51ff1238
ARMv8.2 adds 16-bit floating point operations as an optional extension
to floating point and Adv.SIMD support. This patch adds FP16
instructions to the group Vector Three Register Same, making them
available when +simd+fp16 is enabled.
The instructions added are: FMAXNM, FMAXNMP, FNMINNM, FMINNMP, FMLA,
FMLS, FADD, FADDP, FSUB, FABD, FMULX, FMUL, FCMEQ, FCMGE, FCMGT, FACGE,
FACGT, FMAX, FMAXP, FMIN, FMINP, FRECPS, FDIV and FRSQRTS.
The general form for these instructions is
<OP> <Vd>.<T>, <Vs>.<T>, <Vm>.<T>
where T is 4h or 8h.
gas/testsuite/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* gas/aarch64/advsimd-fp16.d: New.
* gas/aarch64/advsimd-fp16.s: New.
opcodes/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* aarch64-asm-2.c: Regenerate.
* aarch64-dis-2.c: Regenerate.
* aarch64-opc-2.c: Regenerate.
* aarch64-tbl.h (QL_V3SAMEH): New.
(aarch64_opcode_table): Add fp16 versions of fmaxnm, fmla, fadd,
fmulx, fcmeq, fmax, frecps, fminnm, fmls, fsub, fmin, frsqrts,
fmaxnmp, faddp, fmul, fcmge, facge, fmaxp, fdiv, fminnmp, fabd,
fcmgt, facgt and fminp to the vector three same group.
Change-Id: I3f1c5fe82ca73f7a17fe5329cf2b0de03c94328c
ARMv8.2 adds 16-bit floating point operations as an optional extension
to floating point and Adv.SIMD support. This patch set adds the 16-bit
Adv.SIMD vector and scalar instructions to binutils, making them
available when both +simd and +fp16 architecture extensions are enabled.
The series also adds support for a new vector type, 2H, used by the FP16
scalar pairwise instructions.
The patches in this series:
- Add a FP16 Adv.SIMD feature macro for use by the encoding/decoding
routines.
- Add FP16 instructions in the group Vector Three Register Same.
- Add FP16 instructions in the group Scalar Three Register Same.
- Add FP16 instructions in the group Vector Two Register Misc.
- Add FP16 instructions in the group Scalar Two Register Misc.
- Add FP16 instructions in the group Vector Indexed Element.
- Add FP16 instructions in the group Scalar Indexed Element.
- Add FP16 instructions in the group Adv.SIMD Across Lanes.
- Add FP16 instructions in the group Adv.SIMD Modified Immediate.
- Rework some code for handling vector types to weaken its assumptions
about available vector-types.
- Add support for the 2H vector type.
- Add FP16 instructions in the group Adv.SIMD Scalar Pairwise.
- Add FP16 instructions in the group Adv.SIMD Shift By Immediate.
- Add a FP16 instructions in the group Adv.SIMD Scalar Shift By
Immediate.
This patch adds the feature macro SIMD_F16 to the AArch64
encoding/decoding routines. It is used to decide when the new
instructions are available to the assembler and is true when both +simd
and +fp16 are selected.
opcodes/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* aarch64-tbl.h (aarch64_feature_simd_f16): New.
(SIMD_F16): New.
Change-Id: Iee5a37928418f15e51dfaa927b24cafef7295e8f
A mistake with rebasing the ARMv8.2 AT instruction patch left this part
+ /* AT S1E1RP, AT S1E1WP. Values are from aarch64_sys_regs_at. */
+ if ((reg->value == CPENS (0, C7, C9, 0)
+ || reg->value == CPENS (0, C7, C9, 1))
+ && !AARCH64_CPU_HAS_FEATURE (features, AARCH64_FEATURE_V8_2))
+ return FALSE;
in aarch64_pstatefield_supported_p rather than in
aarch64_sys_ins_reg_supported_p, where it was supposed to be.
The patch adding support for id_aa64mmfr2_el1, also had the effect of
removing a conditional branch in aarch64_sys_reg_supported_p.
The effect of both of these is to suppress an error if some ARMv8.2
system registers are used with the wrong -march settings.
This patch fixes these mistakes.
opcodes/
2015-12-14 Matthew Wahab <matthew.wahab@arm.com>
* aarch64-opc.c (aarch64_sys_reg_supported_p): Add mistakenly
removed statement.
(aarch64_pstatefield_supported_p): Move feature checks for AT
registers ..
(aarch64_sys_ins_reg_supported_p): .. to here.
Change-Id: I48783d118eaaf0f3312e8b08a8340ef7af4e36a4
Fixes a couple of places where we access the length field of the type
structure directly, rather than using the TYPE_LENGTH macro.
gdb/ChangeLog:
* i386-tdep.c (i386_mpx_info_bounds): Use TYPE_LENGTH.
(i386_mpx_set_bounds): Likewise.
* solib-darwin.c (darwin_load_image_infos): Likewise.
(darwin_solib_read_all_image_info_addr): Likewise.
