Age | Commit message (Collapse) | Author |
|
This adds a debugging hook for optionally calling a "maymorestack"
function in the prologue of any function that might call morestack
(whether it does at run time or not). The maymorestack function will
let us improve lock checking and add debugging modes that stress
function preemption and stack growth.
Passes toolstash-check -all (except on js/wasm, where toolstash
appears to be broken)
Fixes #48297.
Change-Id: I27197947482b329af75dafb9971fc0d3a52eaf31
Reviewed-on: https://go-review.googlesource.com/c/go/+/359795
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
|
|
The X3 (aka GP) register will potentially be loaded with the __global_pointer$ symbol
during program start up (usually by the dynamic linker). As such, non-Go code may depend
on the contents of GP and calculate offsets based on it, including code called via cgo
and signal handlers installed by non-Go code. As such, stop using the X3 register so
that there are fewer issues interacting between Go and non-Go code.
While here remove the X4 (TP) name from the assembler such that any references must
use the 'TP' name. This should reduce the likelihood of accidental use (like we do
for the 'g' register). The same applies for X3 (GP) when the -shared flag is given.
Updates #47100
Change-Id: I72e82b5ca3f80c46a781781345ca0432a4111b74
Reviewed-on: https://go-review.googlesource.com/c/go/+/351859
Trust: Joel Sing <joel@sing.id.au>
Run-TryBot: Joel Sing <joel@sing.id.au>
Reviewed-by: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
|
|
CR bits and CR fields should be treated separately. Some instructions
modify an entire CR, a CR field, or a single CR bit.
Add a new argument class for CR bits, and teach the assembler the
recognize them as names like CR0LT or CR2SO, and update the CR
bit logic instructions to use them. They will no longer accept
register field (CRn) type arguments.
Fixes #46422
Change-Id: Iaba127d88abada0c2a49b8d3b07a976180565ae4
Reviewed-on: https://go-review.googlesource.com/c/go/+/357774
Run-TryBot: Paul Murphy <murp@ibm.com>
Trust: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
|
|
Fix "MOVW CRx, Rx" and "MOVFL Rx, constant", The FXM field was not
encoded correctly.
Generate mtocrf instead of mtcrf when a CRx argument is used. This
form is much faster.
Simplify several conditional statements which test if the register
argument is REG_CR or one of REG_CRx if the tested argument is known
to be matched as C_CREG. Likewise, a4 is (the From3 arg) is always
TYPE_NONE in the existing optab entries for type_ 69.
Change-Id: I3a4749b1cbfdfab6a2616586ae59e932e01dae50
Reviewed-on: https://go-review.googlesource.com/c/go/+/352789
Reviewed-by: Cherry Mui <cherryyz@google.com>
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>
|
|
The double suffix doesn't seem to serve any purpose, and we can keep
the old spelling as a backwards compatible alias in cmd/asm.
Change-Id: I3f01fc7249fb093ac1b25bd75c1cb9f39b8f62a9
Reviewed-on: https://go-review.googlesource.com/c/go/+/354700
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
|
|
riscv64
CALL and JMP on riscv64 are currently implemented as an AUIPC+JALR pair. This means
that every call requires two instructions and makes use of the REG_TMP register,
even when the symbol would be directly reachable via a single JAL instruction.
Add support for call trampolines - CALL and JMP are now implemented as a single JAL
instruction, with the linker generating trampolines in the case where the symbol is
not reachable (more than +/-1MiB from the JAL instruction), is an unknown symbol or
does not yet have an address assigned. Each trampoline contains an AUIPC+JALR pair,
which the relocation is applied to.
Due to the limited reachability of the JAL instruction, combined with the way that
the Go linker currently assigns symbol addresses, there are cases where a call is to
a symbol that has no address currently assigned. In this situation we have to assume
that a trampoline will be required, however we can patch this up during relocation,
potentially calling directly instead. This means that we will end up with trampolines
that are unused. In the case of the Go binary, there are around 3,500 trampolines of
which approximately 2,300 are unused (around 9200 bytes of machine instructions).
Overall, this removes over 72,000 AUIPC instructions from the Go binary.
Change-Id: I2d9ecfb85dfc285c7729a3cd0b3a77b6f6c98be0
Reviewed-on: https://go-review.googlesource.com/c/go/+/345051
Trust: Joel Sing <joel@sing.id.au>
Run-TryBot: Joel Sing <joel@sing.id.au>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
|
|
The current Go assembler encodes "MOVK $(0<<16|32|48), Rd" as the
same binary with "MOVK $0, Rd", but for arm64 move constant instructions
MOVK, MOVN and MOVZ, "op $0, Rd" and "op $(0<<16|32|48), Rd" have
different semantics. In order not to change the way the assembler
frontend parses constants, this patch adds a check for the zero shifts.
Change-Id: Ia844c419ce49f63605b549e3a2e98d9075dd1cf1
Reviewed-on: https://go-review.googlesource.com/c/go/+/275812
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: fannie zhang <Fannie.Zhang@arm.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
|
|
Likewise, reorder register numbers such that extended mnemonics which
use FPR arguments can be transparently encoded as a VSR argument for
the move to/from VSR class of instructions. Specifically, ensure the
following holds for all FPx and VRx constants: FPRx & 63 == x, and
VRx & 63 == x + 32.
This simplifies encoding machine instructions, and likewise helps
ppc64 assembly writers to avoid hokey workarounds when switching from
vector to vector-scalar register notation. Notably, many VSX
instructions are limited to vector operands due to encoding
restrictions.
Secondly, this explicitly rejects dubious usages of the m[tf]vsr
family of instructions which had previously been accepted.
* Reject two GPR arguments for non-MTVSRDD opcodes. These
have no defined behavior today, and may set RFU bits. e.g
MTVSRD R1, R2, VS1
* Reject FPR destinations for MTVSRDD, and only accept with two GPR
arguments. This copies two GPR values into either half of a VSR. e.g
MTVSRDD R1, R2, F1
MTVSRDD R1, F1
Change-Id: If13dd88c3791d1892dbd18ef0e34675a5285fff9
Reviewed-on: https://go-review.googlesource.com/c/go/+/342929
Run-TryBot: Paul Murphy <murp@ibm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Lynn Boger <laboger@linux.vnet.ibm.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
|
|
Also, add the FABSS and FABSD pseudo instructions to the assembler.
The compiler could use FSGNJX[SD] directly but there doesn't seem
to be much advantage to doing so and the pseudo instructions are
easier to understand.
Change-Id: Ie8825b8aa8773c69cc4f07a32ef04abf4061d80d
Reviewed-on: https://go-review.googlesource.com/c/go/+/348989
Trust: Michael Munday <mike.munday@lowrisc.org>
Run-TryBot: Michael Munday <mike.munday@lowrisc.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Joel Sing <joel@sing.id.au>
|
|
The current assembler uses REGTMP as a temporary destination register,
when optimizing one instruction into a multi-instruction sequence. But
in some cases, when the source register is REGTMP, this behavior is
incorrect.
For example:
ADD $0x1234567, R27, R3
The current assembler encodes it into
MOVD $17767, R27
MOVK $(291<<16), R27
ADD R27, R27, R3
It is illegal to overwrite R27. This CL adds the related checks.
Add test cases.
Change-Id: I0af373d9fd23d8f067c093778dd4cc76748faf38
Reviewed-on: https://go-review.googlesource.com/c/go/+/344689
Reviewed-by: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: fannie zhang <Fannie.Zhang@arm.com>
|
|
Rewrite and simplify the rewriteMOV function in preparation for eliminating it
entirely. Improve some error messages in the process.
Change-Id: Id9a77be5174d46cc23651930c2e9068ee6555690
Reviewed-on: https://go-review.googlesource.com/c/go/+/344458
Trust: Joel Sing <joel@sing.id.au>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
|
|
Loading of constants that are 12 bits or smaller is currently performed using a single
ADDIW instruction, while constants between 13 bits and 32 bits are loaded using a
LUI+ADDIW pair.
Instead, use a single ADDI instruction for the 12 bits or smaller case - this
translates to the LI pseudo-instruction, making objdump more readable and giving:
11c7c: fff00293 li t0,-1
11c80: 00000313 li t1,0
Rather than:
11c7c: fff0029b addiw t0,zero,-1
11c80: 0000031b sext.w t1,zero
In the case where a constant exceeds 12 bits, an LUI instruction is required,
however if the lower 12 bits are zero, the ADDIW instruction can be omitted.
The same applies to the case where immediate splitting is performed for other
immediate instructions.
This removes around 900 instructions from the Go binary.
Change-Id: Id6c77774b3b429fa525da018a6926b85df838a2f
Reviewed-on: https://go-review.googlesource.com/c/go/+/344457
Trust: Joel Sing <joel@sing.id.au>
Run-TryBot: Joel Sing <joel@sing.id.au>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
|
|
Rather than rewriting the obj.Prog for a immediate instructions that need
splitting, generate the appropriate machine instruction sequence directly.
Change-Id: Ie90f0e2a98f97a29281e445c4c3b0c47b793ef4d
Reviewed-on: https://go-review.googlesource.com/c/go/+/344453
Trust: Joel Sing <joel@sing.id.au>
Reviewed-by: Cherry Mui <cherryyz@google.com>
|
|
Rather than rewriting the obj.Prog for a MOV pseudo-instruction targeting
a memory to register load, generate the appropriate machine instruction
sequence directly.
Change-Id: I4c7292ba00f576ec71d4842b6ff27a8ce6db0650
Reviewed-on: https://go-review.googlesource.com/c/go/+/344454
Trust: Joel Sing <joel@sing.id.au>
Reviewed-by: Cherry Mui <cherryyz@google.com>
|
|
Rather than rewriting the obj.Prog for a MOV pseudo-instruction targeting
a constant to register load, generate the appropriate machine instruction
sequence directly.
Change-Id: I38e62f282b39be2a0a241f32280d306558d49b44
Reviewed-on: https://go-review.googlesource.com/c/go/+/344450
Trust: Joel Sing <joel@sing.id.au>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
|
|
Add support to the assembler for F[N]M{ADD,SUB}[SD] instructions.
Argument order is:
OP RS1, RS2, RS3, RD
Also, add support for the FMA intrinsic to the compiler. Automatic
FMA matching is left to a future CL.
Change-Id: I47166c7393b2ab6bfc2e42aa8c1a8997c3a071b3
Reviewed-on: https://go-review.googlesource.com/c/go/+/293030
Trust: Michael Munday <mike.munday@lowrisc.org>
Run-TryBot: Michael Munday <mike.munday@lowrisc.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Joel Sing <joel@sing.id.au>
|
|
Rather than jumping and branching to a label (the offset for which changes
when instructions are added or removed), use PC-relative offsets. This
reduces unnecessary churn in the instruction encodings.
Change-Id: I6816ce939eeabdf828039f59f4f29260eb1ac8da
Reviewed-on: https://go-review.googlesource.com/c/go/+/344449
Trust: Joel Sing <joel@sing.id.au>
Run-TryBot: Joel Sing <joel@sing.id.au>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Munday <mike.munday@lowrisc.org>
|
|
Add more error tests for riscv64 assembly. Also avoid a panic when one of
these error conditions is hit.
Change-Id: If5d913894facbd67f7d014eab745da77c8c66ab0
Reviewed-on: https://go-review.googlesource.com/c/go/+/344228
Trust: Joel Sing <joel@sing.id.au>
Run-TryBot: Joel Sing <joel@sing.id.au>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Meng Zhuo <mzh@golangcn.org>
|
|
Operands of memory type do not support SXTB extension. This CL adds this
missing check.
Change-Id: I1fa438dd314fc8aeb889637079cc67b538e83a89
Reviewed-on: https://go-review.googlesource.com/c/go/+/342769
Reviewed-by: eric fang <eric.fang@arm.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Run-TryBot: eric fang <eric.fang@arm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Michael Knyszek <mknyszek@google.com>
|
|
registers
When using the FP or SP pseudo-register to load or store, pre-index and post-index formats
are not supported because the RSP and pseudo registers are not allowed to be modified in this
way. This CL deletes the related entries in optab and adds a few test cases.
Change-Id: Ie30d27d0e7b959242f0e6298b950489669d07989
Reviewed-on: https://go-review.googlesource.com/c/go/+/342770
Reviewed-by: eric fang <eric.fang@arm.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Run-TryBot: eric fang <eric.fang@arm.com>
Trust: Michael Knyszek <mknyszek@google.com>
|
|
MOVD $C_AACON2(Rf), Rt is encoded as ADD $C_AACON2_high_12_bits, Rf, REGTMP +
ADD $C_AACON2_low_12_bits, REGTMP, Rt. Actually REGTMP is not necessary here,
we can use Rt directly, so it becomes ADD $C_AACON2_high_12_bits, Rf, Rt +
ADD $C_AACON2_low_12_bits, Rt, Rt.
Change-Id: I90b7718b5fb0ab9f3ea28511f42946a6bdccfef3
Reviewed-on: https://go-review.googlesource.com/c/go/+/329751
Reviewed-by: eric fang <eric.fang@arm.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Trust: eric fang <eric.fang@arm.com>
Run-TryBot: eric fang <eric.fang@arm.com>
|
|
Some arm64 instructions accept ZR as its destination register, such as MOVD,
AND, ADD etc. although it doesn't seem to make much sense, but we should
make sure the encoding is correct. However there exists some encoding mistakes
in the current assembler, they are:
1, 'MOVD $1, ZR' is incorrectly encoded as 'MOVD $1, ZR' + '0x00000000'.
2, 'AND $1, R2, ZR' is incorrectly encoded as 'MOVD $1, R27' + 'AND R27, R2, ZR' +
'0x00000000'.
3, 'AND $1, ZR' is incorrectly encoded as 'AND $1, ZR, RSP'.
Obviously the first two encoding errors can cause SIGILL, and the third one will
rewrite RSP.
At the same time, I found some weird encodings but they don't cause errors.
4, 'MOVD $0x0001000100010001, ZR' is encoded as 'MOVW $1, ZR' + 'MOVKW $(1<<16), ZR'.
5, 'AND $0x0001000100010001, R2, ZR' is encoded as 'MOVD $1, R27' + 'MOVK $(1<<16), R27' +
'MOVK $(1<<32), R27'.
Some of these issues also apply to 32-bit versions of these instructions.
These problems are not very complicated, and are basically caused by the improper
adaptation of the class of the constant to the entry in the optab. But the relationship
between these constant classes is a bit complicated, so I don't know how to deal with
issue 4 and 5, because they won't cause errors, so this CL didn't deal with them.
This CL fixed the first three issues.
Issue 1:
before: 'MOVD $1, ZR' => 'MOVD $1, ZR' + '0x00000000'.
after: 'MOVD $1, ZR' => 'MOVD $1, ZR'.
Issue 2:
before: 'AND $1, R2, ZR' => 'MOVD $1, R27' + 'AND R27, R2, ZR' + '0x00000000'.
after: 'AND $1, R2, ZR' => 'ORR $1, ZR, R27' + 'AND R27, R2, ZR'.
Issue 3:
before: 'AND $1, ZR' => 'AND $1, ZR, RSP'.
after: 'AND $1, ZR' => 'ORR $1, ZR, R27' + 'AND R27, ZR, ZR'.
Change-Id: I3c889079229f847b863ad56c88966be12d947202
Reviewed-on: https://go-review.googlesource.com/c/go/+/329750
Reviewed-by: eric fang <eric.fang@arm.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Trust: eric fang <eric.fang@arm.com>
Run-TryBot: eric fang <eric.fang@arm.com>
TryBot-Result: Go Bot <gobot@golang.org>
|
|
For arm64 CMP, ADD and other similar extended register instructions,
if there is no extension, the default extion is LSL<<0, but the default
encoding value (the value of 'option' field) of 32-bit instruction and
64-bit instruction is different, 32-bit is 2 and 64-bit is 3. But the
current assembler incorrectly encodes the value of 32-bit instruction
to 3. This CL fixes this error.
Change-Id: I0e09af2c9c5047a4ed2db7d1183290283db9c31c
Reviewed-on: https://go-review.googlesource.com/c/go/+/329749
Reviewed-by: eric fang <eric.fang@arm.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Run-TryBot: eric fang <eric.fang@arm.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
Trust: Dmitri Shuralyov <dmitshur@golang.org>
|
|
On ARM64, instruction like "MOVD (R1)(R2*1), R3" is accepted and
assembles correctly with Go 1.16, but errors out on tip with
"arm64 doesn't support scaled register format", since CL 289589.
"MOVD (R1)(R2), R3" is the preferred form. But the *1 form works
before and assembles correctly. Keep supporting it.
Fixes #46766.
Change-Id: I0f7fd71fa87ea698919a936b6c68aa5a91afd486
Reviewed-on: https://go-review.googlesource.com/c/go/+/328229
Trust: Cherry Mui <cherryyz@google.com>
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: eric fang <eric.fang@arm.com>
|
|
instructions
According to the armv8-a reference manual, conditions AL and NV are not allowed
for instructions CINC, CINV, CNEG, CSET and CSETM. This CL adds this check and
the corresponding test cases.
Change-Id: Icb496b7b13a353f41491f2de4d939a5cd88abb04
Reviewed-on: https://go-review.googlesource.com/c/go/+/317912
Reviewed-by: eric fang <eric.fang@arm.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Trust: eric fang <eric.fang@arm.com>
Run-TryBot: eric fang <eric.fang@arm.com>
TryBot-Result: Go Bot <gobot@golang.org>
|
|
Recently some tsan tests were enabled on ppc64le which had not
been enabled before. This resulted in failures on systems with
tsan available, and while debugging it was determined that
there were other issues related to the use of signals with cgo.
Signals were not being forwarded within programs linked against
libtsan because the nocgo sigaction was being called for ppc64le
with or without cgo. Adding callCgoSigaction and calling that
allows signals to be registered so that signal forwarding works.
For linux-ppc64 and aix-ppc64, this won't change. On linux-ppc64
there is no cgo. I can't test aix-ppc64 so those owners can enable
it if they want.
In reviewing comments about sigtramp in sys_linux_arm64 it was
noted that a previous issue in arm64 due to missing callee save
registers could also be a problem on ppc64x, so code was added
to save and restore those.
Also, the use of R31 as a temp register in some cases caused an
issue since it is a nonvolatile register in C and was being clobbered
in cases where the C code expected it to be valid. The code sequences to
load these addresses were changed to avoid the use of R31 when loading
such an address.
To get around a vet error, the stubs_ppc64x.go file in runtime
was split into stubs_ppc64.go and stubs_ppc64le.go.
Updates #45040
Change-Id: Ia4ecff950613cbe1b89471790b1d3819d5b5cfb9
Reviewed-on: https://go-review.googlesource.com/c/go/+/306369
Trust: Lynn Boger <laboger@linux.vnet.ibm.com>
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Carlos Eduardo Seo <carlos.seo@linaro.org>
|
|
The current code lacks a check on whether the register and shift/extension
combination is valid, for example the follow instructions also compiles.
ADD F1<<1, R1, R3
ADD V1<<1, R1, R3
MOVW (R9)(F8.SXTW<<2), R19
VST1 R4.D[1], (R0)
Actually only general registers can perform shift operations, and element
and arrangement extensions are only applicable to vector registers. This
CL adds a check for the register and shift/extension combination on arm64.
Change-Id: I93dd9343e92a66899cba8eaf4e0ac5430e94692b
Reviewed-on: https://go-review.googlesource.com/c/go/+/312571
Trust: eric fang <eric.fang@arm.com>
Reviewed-by: eric fang <eric.fang@arm.com>
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: eric fang <eric.fang@arm.com>
TryBot-Result: Go Bot <gobot@golang.org>
|
|
The error message of checking whether the offset value of load/store
instruction is out of range is wrong. The right range is [-256, 255], not
[-255, 254]. The CL fixes it.
Change-Id: Ia342957f1f6bcec65eceb45944221d3972641bed
Reviewed-on: https://go-review.googlesource.com/c/go/+/313891
Reviewed-by: eric fang <eric.fang@arm.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: eric fang <eric.fang@arm.com>
Run-TryBot: eric fang <eric.fang@arm.com>
|
|
The code for generating a long constant versus generating an address
(either via a relocation, or known offset) should be handled in the
same place.
Resolve this by classifying memory arguments as C_LACON (a long
address constant) instead of C_LCON (a long constant).
Likewise, reorder AMOVD/AMOVW optab entries to keep similar
classifications near each other. An extra optab entry for
DWORD is also added to continue handling C_LACON arguments
correctly.
Change-Id: I5ce28400492a071f615125a9b8d260826f1600d7
Reviewed-on: https://go-review.googlesource.com/c/go/+/312296
Run-TryBot: Paul Murphy <murp@ibm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>
Reviewed-by: Carlos Eduardo Seo <carlos.seo@linaro.org>
Trust: Carlos Eduardo Seo <carlos.seo@linaro.org>
|
|
The cmd/link check of the objabi header was a bit lax because
historically the assembler has not included the full version string.
And the assembler didn't do that because it didn't have access to it:
that was buried inside the compiler.
But now that we have cmd/internal/objabi, all the tools have full
access to the expected string, and they can use it, which simplifies
the cmd/link consistency check.
Do that.
Change-Id: I33bd2f9d36c373cc3c32ff02ec6368365088b011
Reviewed-on: https://go-review.googlesource.com/c/go/+/312030
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
The extended opcode field (XO) is generated incorrectly. OPVCC
assumes an X-form like layout for the XO field. MD-form insns
also have an XO field, but it is both smaller and in a different
bit position.
This hasn't been noticed since const1 == 0 matches as a register
argument instead of a constant, thus it is unlikely anyone has
attempted to assemble this instruction with a non-zero shift
argument.
Likewise, update all other MD-form instructions using OPVCC
to use the new OPMD function.
Change-Id: Id81fa2727fb701431911a05492c2038415ad0a4d
Reviewed-on: https://go-review.googlesource.com/c/go/+/310851
Run-TryBot: Paul Murphy <murp@ibm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Lynn Boger <laboger@linux.vnet.ibm.com>
Reviewed-by: Carlos Eduardo Seo <carlos.seo@linaro.org>
|
|
Instructions such as ADD, SUB, CMP do not support ROR shift operations,
but we have not checked this at present. This CL adds this check.
Change-Id: Icac461f61ad6ddb60886a59ba34dddd29df1cc0f
Reviewed-on: https://go-review.googlesource.com/c/go/+/310035
Reviewed-by: eric fang <eric.fang@arm.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: eric fang <eric.fang@arm.com>
Run-TryBot: eric fang <eric.fang@arm.com>
TryBot-Result: Go Bot <gobot@golang.org>
|
|
The go/build package needs access to this configuration,
so move it into a new package available to the standard library.
Change-Id: I868a94148b52350c76116451f4ad9191246adcff
Reviewed-on: https://go-review.googlesource.com/c/go/+/310731
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Jay Conrod <jayconrod@google.com>
|
|
This requires consolidating the register move operations into a
single case entry in asmout. These moves are also used to
sign/zero-extend register values.
Combine the three asmout cases for register moves. This allows
AMOVWZ and AMOVW to be handled with the same optab entries.
Likewise, remove the diagnostic error for non-zero constant
loads into R0 using the register move operations, it is not
possible to match this asmout case with a non-zero constant.
Finally, fix the load constant 0 via "MOV[BHW]{,Z} $0, Rx".
These now generate "li Rx, $0" instead of a zero/sign-extend
of the contents of R0.
Change-Id: Ia4a263661582f10feda27ee21cb121e05ea931dc
Reviewed-on: https://go-review.googlesource.com/c/go/+/308190
Run-TryBot: Paul Murphy <murp@ibm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>
Reviewed-by: Carlos Eduardo Seo <carlos.seo@linaro.org>
Trust: Carlos Eduardo Seo <carlos.seo@linaro.org>
|
|
According the armv8-a specification, the destination register of the ADDS/ADDSW/
SUBS/SUBSW instructions can not be RSP, the current implementation does not
check this and encodes this wrong instruction format as a CMN instruction. This
CL adds a check and test cases for this situation.
Change-Id: I92cc2f8e17dbda70f0dce8fddf1ca6d5d7730589
Reviewed-on: https://go-review.googlesource.com/c/go/+/309989
Reviewed-by: eric fang <eric.fang@arm.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: eric fang <eric.fang@arm.com>
Run-TryBot: eric fang <eric.fang@arm.com>
TryBot-Result: Go Bot <gobot@golang.org>
|
|
Change the assembler to enforce the requirement that ABIInternal
functions need to be NOSPLIT. At the moment all of the assembly
routines in the runtime that are defined as ABIInternal also
happen to be NOSPLIT, but this CL makes it mandatory.
Updates #40724.
Change-Id: Ief80d22de1782edb44b798fcde9aab8a93548722
Reviewed-on: https://go-review.googlesource.com/c/go/+/309789
Trust: Than McIntosh <thanm@google.com>
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
Currently there's knowledge about the list of GOEXPERIMENTs in a few
different places. This CL introduces a new package and consolidates
the list into one place: the internal/goexperiment.Flags struct type.
This package gives us a central place to document the experiments as
well as the GOEXPERIMENT environment variable itself. It will also
give us a place to put built-time constants derived from the enabled
experiments.
Now the objabi package constructs experiment names by reflecting over
this struct type rather than having a separate list of these names
(this is similar to how the compiler handles command-line flags and
debug options). We also expose a better-typed API to the toolchain for
propagating enabled experiments.
Change-Id: I06e026712b59fe2bd7cd11a869aedb48ffe5a4b7
Reviewed-on: https://go-review.googlesource.com/c/go/+/307817
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
|
|
Currently, the objabi.Experiment fields use Go-standard CamelCase, the
GOEXPERIMENT environment variable flags and build tags use all
lowercase, and the asm macros use upper-case with underscores.
This CL makes asm use the lowercase names for macros so there is one
less spelling, e.g., GOEXPERIMENT_regabiargs. This also makes them
consistent with the GOOS_* and GOARCH_* macros, which also use lower
case.
Change-Id: I305cd89af5e8cd1a89cc148746c034bcfd76db3c
Reviewed-on: https://go-review.googlesource.com/c/go/+/307816
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
|
|
"VMOV Vn.<T>[index], Vn" is equivalent to "VDUP Vn.<T>[index], Vn", and
the latter has a higher priority in the disassembler than the former.
But the assembler doesn't support to encode this combination of VDUP,
this leads to an inconsistency between assembler and disassembler.
For example, if we assemble "VMOV V20.S[0], V20" to hex then decode it,
we'll get "VDUP V20.S[0], V20".
VMOV V20.S[0], V20 -> 9406045e -> VDUP V20.S[0], V20 -> error
But we cannot assemble this VDUP again.
Similar reason for "VDUP Rn, Vd.<T>". This CL completes the support for
VDUP.
This patch is a copy of CL 276092. Co-authored-by: JunchenLi
<junchen.li@arm.com>
Change-Id: I8f8d86cf1911d5b16bb40d189f1dc34b24416aaf
Reviewed-on: https://go-review.googlesource.com/c/go/+/302929
Trust: fannie zhang <Fannie.Zhang@arm.com>
Run-TryBot: fannie zhang <Fannie.Zhang@arm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
This separates GOEXPERIMENT=regabi into five sub-experiments:
regabiwrappers, regabig, regabireflect, regabidefer, and regabiargs.
Setting GOEXPERIMENT=regabi now implies the working subset of these
(currently, regabiwrappers, regabig, and regabireflect).
This simplifies testing, helps derisk the register ABI project,
and will also help with performance comparisons.
This replaces the -abiwrap flag to the compiler and linker with
the regabiwrappers experiment.
As part of this, regabiargs now enables registers for all calls
in the compiler. Previously, this was statically disabled in
regabiEnabledForAllCompilation, but now that we can control it
independently, this isn't necessary.
For #40724.
Change-Id: I5171e60cda6789031f2ef034cc2e7c5d62459122
Reviewed-on: https://go-review.googlesource.com/c/go/+/302070
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: David Chase <drchase@google.com>
|
|
riscv64
Follow what MIPS does and load >32-bit constants from memory using two instructions,
rather than generating a four to six instruction sequence. This removes more than 2,500
instructions from the Go binary. This also makes it possible to load >32-bit constants
via a single assembly instruction, if required.
Change-Id: Ie679a0754071e6d8c52fe0d027f00eb241b3a758
Reviewed-on: https://go-review.googlesource.com/c/go/+/302609
Trust: Joel Sing <joel@sing.id.au>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
The assember uses R15 as scratch space when assembling global variable
references in dynamically linked code. If the assembly code uses the
clobbered value of R15, report an error. The user is probably expecting
some other value in that register.
Getting rid of the R15 use isn't very practical (we could save a
register to a field in the G maybe, but that gets cumbersome).
Fixes #43661
Change-Id: I43f848a3d8b8a28931ec733386b85e6e9a42d8ff
Reviewed-on: https://go-review.googlesource.com/c/go/+/283474
Trust: Keith Randall <khr@golang.org>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
This CL encodes:
ROTR rd, rt, sa
ROTRV rd, rt, rs
=> ROTR (SCON|REG), (REG,)? REG
DROTR rd, rt, sa
DROTR32 rd, rt, sa
DROTRV rd, rt, rs
=> ROTRV (SCON|REG), (REG,)? REG
Note: ROTRV will handle const over 32
Ref: The MIPS64® Instruction Set Reference Manual Revision 6.05
Change-Id: Ibe69f999b83eb43843d088cf1ac5a13c995269a5
Reviewed-on: https://go-review.googlesource.com/c/go/+/280114
Trust: Meng Zhuo <mzh@golangcn.org>
Run-TryBot: Meng Zhuo <mzh@golangcn.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
Refer to ARM reference manual, like add(extended register) instructions,
the extension is encoded in the "option" field. If "Rd" or "Rn" is
RSP and "option" is "010" then LSL is preferred. Therefore, the instrution
"add Rm<<imm, RSP, RSP" or "add Rm<<imm RSP" is valid and can be encoded
as add(extended register) instruction.
But the current assembler can not handle like "op R1<<1, RSP, RSP"
instructions, this patch adds the support.
Because MVN(extended register) does not exist, remove it.
Add test cases.
Change-Id: I968749d75c6b93a4f297b39c73cc292e6b1035ad
Reviewed-on: https://go-review.googlesource.com/c/go/+/284900
Trust: fannie zhang <Fannie.Zhang@arm.com>
Run-TryBot: fannie zhang <Fannie.Zhang@arm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
This CL changes GOEXPERIMENT to act like other GO[CONFIG] environment
variables. Namely, that it can be set at make.bash time to provide a
default value used by the toolchain, but then can be manually set when
running either cmd/go or the individual tools (compiler, assembler,
linker).
For example, it's now possible to test rsc.io/tmp/fieldtrack by simply
running:
GOEXPERIMENT=fieldtrack go test -gcflags=-l rsc.io/tmp/fieldtrack \
-ldflags=-k=rsc.io/tmp/fieldtrack.tracked
without needing to re-run make.bash. (-gcflags=-l is needed because
the compiler's inlining abilities have improved, so calling a function
with a for loop is no longer sufficient to suppress inlining.)
Fixes #42681.
Change-Id: I2cf8995d5d0d05f6785a2ee1d3b54b2cfb3331ca
Reviewed-on: https://go-review.googlesource.com/c/go/+/300991
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
|
|
These instructions are actually 5 argument opcodes as specified
by the ISA. Prior to this patch, the MB and ME arguments were
merged into a single bitmask operand to workaround the limitations
of the ppc64 assembler backend.
This limitation no longer exists. Thus, we can pass operands for
these opcodes without having to merge the MB and ME arguments in
the assembler frontend or compiler backend.
Likewise, support for 4 operand variants is unchanged.
Change-Id: Ib086774f3581edeaadfd2190d652aaaa8a90daeb
Reviewed-on: https://go-review.googlesource.com/c/go/+/298750
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>
Reviewed-by: Carlos Eduardo Seo <carlos.seo@linaro.org>
Trust: Carlos Eduardo Seo <carlos.seo@linaro.org>
|
|
Arm64 doesn't have scaled register format, such as (R1*2), (R1)(R2*3),
but currently the assembler doesn't report an error for such kind of
instruction operand format. This CL disables the scaled register
operand format for arm64 and reports an error if this kind of instruction
format is seen.
With this CL, the assembler won't print (R1)(R2) as (R1)(R2*1), so that
we can make the assembly test simpler.
Change-Id: I6d7569065597215be4c767032a63648d2ad16fed
Reviewed-on: https://go-review.googlesource.com/c/go/+/289589
Trust: eric fang <eric.fang@arm.com>
Run-TryBot: eric fang <eric.fang@arm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: eric fang <eric.fang@arm.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
Constant of BITCON type can be moved into RSP by MOVD or MOVW instructions
directly, this CL enables this format of these two instructions.
For 32-bit ADDWop instructions with constant, rewrite the high 32-bit
to be a repetition of the low 32-bit, just as ANDWop instructions do,
so that we can optimize ADDW $bitcon, Rn, Rt as:
MOVW $bitcon, Rtmp
ADDW Rtmp, Rn, Rt
The original code is:
MOVZ $bitcon_low, Rtmp
MOVK $bitcon_high,Rtmp
ADDW Rtmp, Rn, Rt
Change-Id: I30e71972bcfd6470a8b6e6ffbacaee79d523805a
Reviewed-on: https://go-review.googlesource.com/c/go/+/289649
Trust: eric fang <eric.fang@arm.com>
Run-TryBot: eric fang <eric.fang@arm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: eric fang <eric.fang@arm.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
This CL adds support for arm64 fp&simd instructions VUMAX and VUMIN.
Fixes #42326
Change-Id: I3757ba165dc31ce1ce70f3b06a9e5b94c14d2ab9
Reviewed-on: https://go-review.googlesource.com/c/go/+/271497
Trust: eric fang <eric.fang@arm.com>
Run-TryBot: eric fang <eric.fang@arm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: fannie zhang <Fannie.Zhang@arm.com>
Reviewed-by: eric fang <eric.fang@arm.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
This CL adds assembly support for 128-bit FLDPQ and FSTPQ instructions.
This CL also deletes some wrong pre/post-indexed LDP and STP instructions,
such as {ALDP, C_UAUTO4K, C_NONE, C_NONE, C_PAIR, 74, 8, REGSP, 0, C_XPRE},
because when the offset type is C_UAUTO4K, pre and post don't work.
Change-Id: Ifd901d4440eb06eb9e86c9dd17518749fdf32848
Reviewed-on: https://go-review.googlesource.com/c/go/+/273668
Trust: eric fang <eric.fang@arm.com>
Run-TryBot: eric fang <eric.fang@arm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: eric fang <eric.fang@arm.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|