Since they're shorter to encode, the 0xa0...0xa3 encodings are preferred
for moves between accumulator and absolute address outside of 64-bit
mode. With HLE release semantics this encoding is unsupported though,
with the assembler raising an error. The operation is valid though, we
merely need to pick the longer encoding in that case.
The wrong placement of the Load attribute in the templates prevented
this from working. The disassembler also didn't handle this consistently
with other similar dual-encoding insns.
While many templates allowing multiple suitably matching XMM/YMM/ZMM
operand sizes can be folded, a few need to be split in order to not
wrongly accept "xmmword ptr" operands when only XMM registers are
permitted (and memory operands are more narrow). Add a test case
validating this.
Since the addition of pseudo prefixes changed how the scrubber treats
'{', we need to explicitly strip whitespace in check_VecOperations ().
* config/tc-i386.c (check_VecOperations): Strip whitespace.
* testsuite/gas/i386/optimize-1.s: Add whitespaces before
{%k7} and {z},
* testsuite/gas/i386/x86-64-optimize-2.s: Likewise.
We can optimize AVX512 instructions with EVEX128 only if AVX512VL is
enabled:
1. Instruction is an AVX512VL instruction. Or
2. AVX512VL is enabled explicitly by -march=+avx512vl/".arch .avx512vl".
We should optimize EVEX instructions with EVEX128 encoding when pseudo
{evex} prefix is used.
* config/tc-i386.c (set_cpu_arch): Set cpu_arch_isa_flags.
(md_parse_option): Likewise.
(optimize_encoding): Check i.tm.cpu_flags and cpu_arch_isa_flags
for cpuavx512vl instead of cpu_arch_flags. Optimize EVEX with
EVEX128 when EVEX encoding is required.
* testsuite/gas/i386/i386.exp: Run optimize-4, optimize-5,
x86-64-optimize-5 and x86-64-optimize-6.
* testsuite/gas/i386/optimize-1.d: Updated.
* testsuite/gas/i386/x86-64-optimize-2.d: Likewise.
* testsuite/gas/i386/optimize-4.d: New file.
* testsuite/gas/i386/optimize-4.s: Likewise.
* testsuite/gas/i386/optimize-5.d: Likewise.
* testsuite/gas/i386/optimize-5.s: Likewise.
* testsuite/gas/i386/x86-64-optimize-5.d: Likewise.
* testsuite/gas/i386/x86-64-optimize-5.s: Likewise.
* testsuite/gas/i386/x86-64-optimize-6.d: Likewise.
* testsuite/gas/i386/x86-64-optimize-6.s: Likewise.
"clr reg" is an alias of "xor reg, reg". We can encode "clr reg64" as
"xor reg32, reg32".
gas/
* config/tc-i386.c (optimize_encoding): Also encode "clr reg64"
as "xor reg32, reg32".
* testsuite/gas/i386/x86-64-optimize-1.s: Add "clr reg64" tests.
* testsuite/gas/i386/x86-64-optimize-1.d: Updated.
opcodes/
* i386-opc.tbl: Add Optimize to clr.
* i386-tbl.h: Regenerated.
The differences between some of the register and memory forms of the
same insn often don't really require the templates to be separate. For
example, Disp8MemShift is simply irrelevant to register forms. Fold
these as far as possible, and also fold register-only forms. Further
folding is possible, but needs other prereq work done first.
A note regarding EVEXDYN: This is intended to be used only when no other
properties of the template would make is_evex_encoding() return true. In
all "normal" cases I think it is preferable to omit this indicator, to
keep the table half way readable.
Their memory forms were bogusly using VexLWP instead of VexNDD. Adjust
VexNDD handling to cope with these, allowing their register and memory
forms to be folded.
They aren't really useful (anymore?): The conflicting operand size check
isn't applicable to any insn validly using respective memory operand
sizes (and if they're used wrongly, another error would result), and the
logic in process_suffix() can be easily changed to work without them.
While re-structuring conditionals in process_suffix() also drop the
CMPXCHG8B special case in favor of a NoRex64 attribute in the opcode
table.
Some BMI/BMI2 insns allow their middle operands to be a memory one. In
such a case, matching register types between operands 0 and 1 as well as
1 and 2 won't help - operands 0 and 2 also need to be checked.
Make more obvious what the success and failure paths are, and in
particular that what used to be at the "skip" label can't be reached
by what used to be straight line code.
Just like for the AVX/AES and AVX/PCLMUL combinations, AVX/GFN,
AVX512F/GFNI, AVX512F/VAES, and AVX512F/PCLMUL need special handling to
deal with the pair of required checks specified in the templates.
fsub/fsubr/fsubp/fsubrp as well as fdiv/fdivr/fdivp/fdivrp disassembly
should match (a) the Intel SDM and (b) respective input fed to gas (both
of course with the exception of when we intentionally convert bogus
insns, accompanied by a warning).
Drop "second": For one there's no other source register (the other
source operand is in memory), and in Intel syntax such numbering would
also be wrong.
Take the opportunity and also
- properly place declarations ahead of statements
- use %u format for unsigned int arguments
- fix indentation
This requires a change to ModR/M handling: Recording of displacement
types must not discard operand size information. Change the respective
code to alter only .disp<N>.
Oops, not tested well enough. -mpower9 sets all the PPC_OPCODE_POWERn
for n <= 9.
* config/tc-ppc.c (ppc_handle_align): Correct last patch. Really
don't emit a group terminating nop for power9. Simplify cpu
tests.
Power9 doesn't have a group terminating nop, so we may as well emit a
normal nop for power9. Not that it matters a great deal, I believe
ori 2,2,0 will be treated exactly as ori 0,0,0 by the hardware.
* config/tc-ppc.c (ppc_handle_align): Don't emit a group
terminating nop for power9.
xcoff (32-bit) objdump accepted but ignored -M options unless
-mpowerpc was also given. This patch fixes that, leaving the default
as -Mpwr for xcoff. I've also enabled more tests for xcoff targets.
binutils/
* configure.ac: Add objdump_private_desc_xcoff for rs6000.
* configure: Regenerate.
gas/
* testsuite/gas/ppc/aix.exp: Run for rs6000 too.
* testsuite/gas/ppc/ppc.exp: Run more tests for non-ELF targets.
* testsuite/gas/ppc/machine.d: Don't run for PE targets.
opcodes/
* disassemble.c (disassembler): Use bfd_arch_powerpc entry for
bfd_arch_rs6000.
* disassemble.h (print_insn_rs6000): Delete.
* ppc-dis.c (powerpc_init_dialect): Handle rs6000.
(disassemble_init_powerpc): Call powerpc_init_dialect for rs6000.
(print_insn_rs6000): Delete.
Commit 4d354d8b8932d450c77fd52f3501662827523014 introduced a NULL
pointer dereference by replacing a pointer assignment by a pointer
dereference assignment without adding a NULL pointer check. This patch
fixes it.
2018-03-02 Thomas Preud'homme <thomas.preudhomme@arm.com>
gas/
* config/tc-arm.c (md_begin): Add NULL pointer check before
dereferencing march_ext_opt.
I've always found the code in ARM backend of gas to control what
CPU/architecture and FPU are selected by the user and to support
autodetection of features complex and confusing. Chief among the
issues I have with that code is the lack of comments to explain
the meaning of the various variables. This patch addresses that
and much more:
- add comments to explain meaning of all arm_feature_set variables
- keep track of currently selected CPU, extensions and FPU in a separate
set of new variables
- make naming of variable more consistent
- remove dead code
- simplify handling of extensions
The overall approach is as follows:
* restrict m*_opt variable to hold the feature bits of the
corresponding mcpu/march/mfpu command-line options
* record selected CPU, extensions and FPU in new selected_* during
md_begin
* whenever a .cpu/.arch/.arch_extension/.fpu directive is met, update
the corresponding selected_* variables (eg. selected_arch, then
selected_cpu for a .cpu or .arch directive) and then finally
cpu_variant from them
* pass extension feature set pointer by value to arm_parse_extension
since it's only ever called from arm_parse_cpu and arm_parse_arch
which allocate the extension feature set themselves
* likewise, remove allocation from s_arm_arch_extension since the use
of arm_feature_set structure for selected_ext rather than a pointer
alleviate the need for it
* in autodetection mode, only set all CPU fits in cpu_variant but leave
selected_cpu* variables unset
* in md_begin, remove dead "else if" to set a default FPU when no FPU
was selected. Setting a default FPU based on CPU as did the code
before it turn dead should be based on the default FPU field of the
CPU and architecture table as will be done in a separate patch. Logic
is wrong anyway since it sets VFP2 as default FPU for Armv6-M and
Armv7-M
Hopefully that should be enough to understand the change but if not feel
free to ask questions about the patch. While I believe the new code is
easier to understand, it remains complex and the old one was even more
complex so the change is difficult to understand.
2018-03-01 Thomas Preud'homme <thomas.preudhomme@arm.com>
gas/
* config/tc-arm.c (cpu_variant, arm_arch_used, thumb_arch_used,
legacy_cpu, legacy_fpu, mcpu_cpu_opt, dyn_mcpu_ext_opt,
mcpu_fpu_opt, march_cpu_opt, dyn_march_ext_opt, march_fpu_opt,
mfpu_opt, object_arch, selected_cpu): Comment meaning of variables.
(dyn_mcpu_ext_opt): Also rename into ...
(mcpu_ext_opt): This.
(dyn_march_ext_opt): Also rename into ...
(march_ext_opt): This.
(object_arch): Also rename into ...
(selected_object_arch): This and make it a plain arm_feature_set
structure.
(selected_arch, selected_ext, selected_fpu): New static variables.
(mark_feature_used): Fix comments, feature is marked as used iff it is
currently allowed.
(do_bx): Adapt to change in name and type of object_arch.
(md_begin): Set selected_arch rather than mcpu_cpu_opt, selected_ext
rather than dyn_mcpu_ext_opt and selected_fpu rather than mfpu_opt.
Remove dead code to set default FPU if architecture version is greater
than 5. Set all CPU bits of cpu_variant directly in autodection
leaving mcpu_cpu_opt, selected_arch and selected_cpu unset.
(arm_parse_extension): Take extension feature set pointer parameter by
value rather than by pointer. Remove allocation code. Adapt code
accordingly.
(arm_parse_cpu): Adapt to variable renaming and changes in
arm_parse_extension () signature.
(arm_parse_arch): Likewise.
(aeabi_set_public_attributes): Also set selected_arch and selected_ext
in addition to selected_cpu. Set flags_arch and flags_ext from them
instead of selected_cpu. Adapt to variables renaming and type change.
(arm_md_post_relax): Adapt to variable renaming.
(s_arm_cpu): Set selcted_cpu_cpu and selected_ext instead of
mcpu_cpu_opt and dyn_mcpu_ext_opt. Set selected_cpu from them and
cpu_variant from selected_cpu and selected_fpu.
(s_arm_arch): Likewise.
(s_arm_object_arch): Adapt to variable renaming.
(s_arm_arch_extension): Use ARM_CPU_IS_ANY instead of checking feature
set against arm_any. Check selected_arch rather than *mcpu_cpu_opt.
Set selected_ext rather than *dyn_mcpu_ext_opt and remove allocation
code.
(s_arm_fpu): Set selected_fpu instead of mfpu_opt. Set all CPU feature
bits if in autodetection mode.
On x86, some instructions have alternate shorter encodings:
1. When the upper 32 bits of destination registers of
andq $imm31, %r64
testq $imm31, %r64
xorq %r64, %r64
subq %r64, %r64
known to be zero, we can encode them without the REX_W bit:
andl $imm31, %r32
testl $imm31, %r32
xorl %r32, %r32
subl %r32, %r32
This optimization is enabled with -O, -O2 and -Os.
2. Since 0xb0 mov with 32-bit destination registers zero-extends 32-bit
immediate to 64-bit destination register, we can use it to encode 64-bit
mov with 32-bit immediates. This optimization is enabled with -O, -O2
and -Os.
3. Since the upper bits of destination registers of VEX128 and EVEX128
instructions are extended to zero, if all bits of destination registers
of AVX256 or AVX512 instructions are zero, we can use VEX128 or EVEX128
encoding to encode AVX256 or AVX512 instructions. When 2 source
registers are identical, AVX256 and AVX512 andn and xor instructions:
VOP %reg, %reg, %dest_reg
can be encoded with
VOP128 %reg, %reg, %dest_reg
This optimization is enabled with -O2 and -Os.
4. 16-bit, 32-bit and 64-bit register tests with immediate may be
encoded as 8-bit register test with immediate. This optimization is
enabled with -Os.
This patch does:
1. Add {nooptimize} pseudo prefix to disable instruction size
optimization.
2. Add optimize to i386_opcode_modifier to tell assembler that encoding
of an instruction may be optimized.
gas/
PR gas/22871
* NEWS: Mention -O[2|s].
* config/tc-i386.c (_i386_insn): Add no_optimize.
(optimize): New.
(optimize_for_space): Likewise.
(fits_in_imm7): New function.
(fits_in_imm31): Likewise.
(optimize_encoding): Likewise.
(md_assemble): Call optimize_encoding to optimize encoding.
(parse_insn): Handle {nooptimize}.
(md_shortopts): Append "O::".
(md_parse_option): Handle -On.
* doc/c-i386.texi: Document -O0, -O, -O1, -O2 and -Os as well
as {nooptimize}.
* testsuite/gas/cfi/cfi-x86_64.d: Pass -O0 to assembler.
* testsuite/gas/i386/ilp32/cfi/cfi-x86_64.d: Likewise.
* testsuite/gas/i386/i386.exp: Run optimize-1, optimize-2,
optimize-3, x86-64-optimize-1, x86-64-optimize-2,
x86-64-optimize-3 and x86-64-optimize-4.
* testsuite/gas/i386/optimize-1.d: New file.
* testsuite/gas/i386/optimize-1.s: Likewise.
* testsuite/gas/i386/optimize-2.d: Likewise.
* testsuite/gas/i386/optimize-2.s: Likewise.
* testsuite/gas/i386/optimize-3.d: Likewise.
* testsuite/gas/i386/optimize-3.s: Likewise.
* testsuite/gas/i386/x86-64-optimize-1.s: Likewise.
* testsuite/gas/i386/x86-64-optimize-1.d: Likewise.
* testsuite/gas/i386/x86-64-optimize-2.d: Likewise.
* testsuite/gas/i386/x86-64-optimize-2.s: Likewise.
* testsuite/gas/i386/x86-64-optimize-3.d: Likewise.
* testsuite/gas/i386/x86-64-optimize-3.s: Likewise.
* testsuite/gas/i386/x86-64-optimize-4.d: Likewise.
* testsuite/gas/i386/x86-64-optimize-4.s: Likewise.
opcodes/
PR gas/22871
* i386-gen.c (opcode_modifiers): Add Optimize.
* i386-opc.h (Optimize): New enum.
(i386_opcode_modifier): Add optimize.
* i386-opc.tbl: Add "Optimize" to "mov $imm, reg",
"sub reg, reg/mem", "test $imm, acc", "test $imm, reg/mem",
"and $imm, acc", "and $imm, reg/mem", "xor reg, reg/mem",
"movq $imm, reg" and AVX256 and AVX512 versions of vandnps,
vandnpd, vpandn, vpandnd, vpandnq, vxorps, vxorpd, vpxor,
vpxord and vpxorq.
* i386-tbl.h: Regenerated.
This patch makes GAS emit a warning when trying to assemble the Armv8.2
FP16 instructions VMOVX and VINS with condition codes. The Armv8-A
Reference Manual specifies these instructions without conditional codes
and says that if they are found in an IT block that they are CONSTRAINED
UNPREDICABLE.
gas/ChangeLog:
2018-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
* config/tc-arm.c (do_neon_movhf): If conditional error out when in arm
mode and emit warning in thumb mode.
* testsuite/gas/arm/armv8-2-fp16-scalar-bad.s: Add new tests.
* testsuite/gas/arm/armv8-2-fp16-scalar-bad.l: Idem.
