LDXP
Load exclusive pair of registers
This instruction derives an address from a base register
value, loads two 32-bit words or two
64-bit doublewords from memory,
and writes them to
two registers.
For information on single-copy atomicity and alignment requirements,
see Requirements for single-copy atomicity and
Alignment of data accesses.
The PE marks the physical address being accessed as an exclusive access.
This exclusive access mark is checked by Store Exclusive instructions. See
Synchronization and semaphores.
For information about addressing modes, see
Load/Store addressing modes.
For information about the CONSTRAINED UNPREDICTABLE behavior of this instruction, see Architectural Constraints on UNPREDICTABLE behaviors, and particularly LDXP.
If PSTATE.DIT is 1, the timing of this instruction is insensitive to the value of the data being loaded or stored.
1
0
0
1
0
0
0
0
1
1
(1)
(1)
(1)
(1)
(1)
0
0
LDXP <Wt1>, <Wt2>, [<Xn|SP>{, #0}]
1
LDXP <Xt1>, <Xt2>, [<Xn|SP>{, #0}]
constant integer t = UInt(Rt);
constant integer t2 = UInt(Rt2);
constant integer n = UInt(Rn);
constant integer elsize = 32 << UInt(sz);
constant integer datasize = elsize * 2;
constant boolean acqrel = FALSE;
constant boolean tagchecked = n != 31;
boolean rt_unknown = FALSE;
if t == t2 then
constant Constraint c = ConstrainUnpredictable(Unpredictable_LDPOVERLAP);
assert c IN {Constraint_UNKNOWN, Constraint_UNDEF, Constraint_NOP};
case c of
when Constraint_UNKNOWN rt_unknown = TRUE; // result is UNKNOWN
when Constraint_UNDEF UNDEFINED;
when Constraint_NOP ExecuteAsNOP();
<Wt1>
Is the 32-bit name of the first general-purpose register to be transferred, encoded in the "Rt" field.
<Wt2>
Is the 32-bit name of the second general-purpose register to be transferred, encoded in the "Rt2" field.
<Xn|SP>
Is the 64-bit name of the general-purpose base register or stack pointer, encoded in the "Rn" field.
<Xt1>
Is the 64-bit name of the first general-purpose register to be transferred, encoded in the "Rt" field.
<Xt2>
Is the 64-bit name of the second general-purpose register to be transferred, encoded in the "Rt2" field.
bits(64) address;
bits(datasize) data;
constant integer dbytes = datasize DIV 8;
constant boolean privileged = PSTATE.EL != EL0;
constant AccessDescriptor accdesc = CreateAccDescExLDST(MemOp_LOAD, acqrel, tagchecked,
privileged);
if n == 31 then
CheckSPAlignment();
address = SP[];
else
address = X[n, 64];
// Tell the Exclusives monitors to record a sequence of one or more atomic
// memory reads from virtual address range [address, address+dbytes-1].
// The Exclusives monitor will only be set if all the reads are from the
// same dbytes-aligned physical address, to allow for the possibility of
// an atomicity break if the translation is changed between reads.
AArch64.SetExclusiveMonitors(address, dbytes);
if rt_unknown then
// ConstrainedUNPREDICTABLE case
X[t, datasize] = bits(datasize) UNKNOWN; // In this case t = t2
elsif elsize == 32 then
// 32-bit load exclusive pair (atomic)
data = Mem[address, dbytes, accdesc];
if BigEndian(accdesc.acctype) then
X[t, elsize] = data<elsize +: elsize>;
X[t2, elsize] = data<0 +: elsize>;
else
X[t, elsize] = data<0 +: elsize>;
X[t2, elsize] = data<elsize +: elsize>;
else // elsize == 64
// 64-bit load exclusive pair (not atomic), but must be 128-bit aligned
if !IsAligned(address, dbytes) then
AArch64.Abort(address, AlignmentFault(accdesc));
constant bits(64) address2 = AddressIncrement(address, 8, accdesc);
X[t, 64] = Mem[address , 8, accdesc];
X[t2, 64] = Mem[address2, 8, accdesc];