The SVE patches made registers longer by increasing NumVecElemPerVecReg,
but the GDB XML was not updated to account for that, and as a result GDB
connections were failing with:
Remote 'g' packet reply is too long
This commit introduces NumVecElemPerSimdVecReg which counts only the SIMD
register sizes to get it back working. SVE GDB support is not added here.
Change-Id: I4191b9f1999ae02b0308863db4cc9b5b16a27d6d
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20468
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Non-temporal quadword/double-quadword move instructions.
This change ignores the non-temporal hint and instructions are
implemented to send cacheable request to memory.
This would have some "performance" impact (i.e. having some cache
pollution) to get better "correctness" in behavior.
Change-Id: I2052ac0970f61a54bafb7332762debcb7103202d
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20288
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Fixes the definition of the RISC-V GDB register cache. The latest
version, of RISC-V gdb, commit c3eb4078520dad8234ffd7fbf893ac0da23ad3c8,
appears to only accept the 32 integer registers + the PC in the 'g'
packet.
This functions with the Linux toolchain (riscv64-unknown-linux-gnu-*),
but works best with the Newlib toolchain (riscv64-unknown-elf-*).
Change-Id: Ie35ea89a45870fb634e6c68236261bde27c86e41
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20028
Reviewed-by: Alec Roelke <alec.roelke@gmail.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Alec Roelke <alec.roelke@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
The previous implementation left the registers unmodified which is
technically correct since there is no defined behavior in that case or
a fault to raise. That would make what happened when the following code
consumed the result unpredictable because it would depend on what junk
values were left in the registers. This was originally not a problem
since the space of supported functions were tightly packed, but someone
added a new function with a gap without adjusting this behavior.
This change makes CPUID zero out RAX, RBX, RCX, and RDX when it fails.
That should be more predictable and cause less flakey failures.
Change-Id: If6ffb17c2969d34aff1600c0ffc32333d0b9be44
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20168
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Pouya Fotouhi <pfotouhi@ucdavis.edu>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Adding LD/ST/SWP family of instructions, LD/ST include a set of
operations like ADD/CLR/EOR/SET/UMAX/UMIN/SMAX/SMIN
This commit includes:
+ Instruction decode
+ Instruction functional code
+ New set of skeletons for Ex/Com/Ini/Constructor and declaration.
Change-Id: Ieea8d4256807e004d2f8aca8f421b3df8d76b116
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19812
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Maintainer: Giacomo Travaglini <giacomo.travaglini@arm.com>
The X86 local APIC doesn't actually use the pio_addr set in the config
and instead computes what address it will respond to based on the
initial ID of the CPU it's attached to. gem5's BasicPioDevice, which
the X86LocalApic class inherits from, does not provide a default value
for that parameter and will complain if *something* isn't set. The
value used, 0x2000000000000000, is a dummy value which is the base of
the region of the physical address space set aside for messages to
local APICs from the CPU and from other local APICs.
Also, the clock for the local APIC's timer is defined to be the bus
clock. The assumption seems to be that this has a 16:1 ratio with the
CPU clock, and I vaguely remember finding that that was more or less
unofficially true, even if it isn't necessary stringently defined to
be that.
Since we were already just assuming that that ratio was correct and
always setting up the local APICs clock that way, we can do that in
the X86LocalApic class definition and remove some special x86 specific
setup that we'd otherwise need for the x86 version of the Interrupt
class. If that's not correct, it can still be overridden somewhere else
in the config.
Change-Id: I50e84f899f44b1191c2ad79d05803b44f07001f9
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19968
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
ARMv8.1-PAN adds a new bit to PSTATE. When the value of this PAN state
bit is 1, any privileged data access from EL1 or EL2 to a virtual memory
address that is accessible at EL0 generates a Permission fault.
This feature is mandatory in ARMv8.1 implementations.
This feature is supported in AArch64 and AArch32 states.
The ID_AA64MMFR1_EL1.PAN, ID_MMFR3_EL1.PAN, and ID_MMFR3.PAN fields
identify the support for ARMv8.1-PAN.
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Change-Id: I94a76311711739dd2394c72944d88ba9321fd159
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19729
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
MSR <pstatefield>, #imm is used for setting a PSTATE field using an
immediate. Current implementation has the following flaws:
* There is no base MSR immediate definition: all the existing
PSTATE fields have a different class definition
* Those implementation make use of a generic data64 base class
which results in a wrong disassembly (pstate register is printed as an
integer register).
This patch is fixing this by defining a new base class (MiscRegImmOp64)
and new related templates. In this way, we aim to ease addition of new
PSTATE fields (in ARMv8.x)
Change-Id: I71b630ff32abe1b105bbb3ab5781c6589b67d419
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19728
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
If a page table walk is squashed, the walker state is being deleted
in the squash code. If there are in flight requests, the deleted
walker state values may be clobbered, leading to undefined behavior.
This adds a squashed boolean to the walker state which is set if a
walk is squashed while requests are still in flight. When packets
for the in flight request return, we check if the walk was squashed
and return that the walk is complete once the number of in flight
requests reaches zero. The walker state is then freed by the PTW.
Change-Id: I57a64b1548b83a8a9e8441fc9d6f33e9842df2b3
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19568
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>
According to the armarm:
ARMv8.1-HPD introduces the facility to disable the hierarchical
attributes, APTable, PXNTable, and UXNTable, in the translation tables.
This disable has no effect on the NSTable bit. This feature is
mandatory in ARMv8.1 implementations.
This feature is added only to the VMSAv8-64 translation regimes. ARMv8.2
extends this to the AArch32 translation regimes, see ARMv8.2-AA32HPD.
The ID_AA64MMFR1_EL1.HPDS field identifies the support for ARMv8.1-HPD.
Change-Id: Ibbf589b82f2c1e4437b43252f8f633e8f6fb0b80
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Ciro Santilli <ciro.santilli@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19610
Tested-by: kokoro <noreply+kokoro@google.com>
Normally, a translation will start via translateTiming/functional
which will check if the miscRegs have been updated and if so,
will update the TLB state accordingly. However, in a 2 stage
system, if there is a hit in stage 1, the resulting IPA will be
sent to the S2-TLB for translation via a getTE() function call
(via the stage2_lookup object). This will cause the state of the
S2-TLB to be out of sync.
Change-Id: I117e4032fc76d7d31f4f999887b5573a7e5811e6
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/14995
Tested-by: kokoro <noreply+kokoro@google.com>
Maintainer: Giacomo Travaglini <giacomo.travaglini@arm.com>
This is trying to fix the bug that arises when a memory exception
is generated during a fp flavoured load (A memory load targeting
a SIMD & FP register).
With the previous template a fault was not stopping the register
value to be modified (wrong)
if (fault == NoFault) {
fault = readMemAtomic(xc, traceData, EA, Mem, memAccessFlags);
%(memacc_code)s;
}
if (fault == NoFault) {
%(op_wb)s;
}
The patch introduces a Load64FpExecute template which is moving the
register write (memacc_code) just before the op_wb
Change-Id: I1c89c525dfa7a4ef489abe0872cd7baacdd6ce3c
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19228
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
