Mask elements
Each element index maps to one predicate bit; this instruction operates on mask bits, not SEW-wide data elements.
VMXNOR.MM
RISC-V VMXNOR.MM Instruction Details
Instruction ManualOPMVV (OP-V)Vector mask XNOR: vd.mask[i] = ~(vs2.mask[i] ^ vs1.mask[i])
Instruction Syntax
vmxnor.mm vd, vs2, vs1
Operand Breakdown
vd: destination mask register receiving a one-bit Boolean result per element.
vs2: first source mask register; element i supplies vs2.mask[i].
vs1: second source mask register; element i supplies vs1.mask[i]. `.mm` mask logical instructions are always unmasked and are not controlled by v0.t; vd can be any mask register, and only a result written or moved into v0 is later consumed through v0.t by ordinary masked vector instructions.
VVector OperationsMask Logic
Instruction Behavior
VMXNOR.MM is an RVV vector mask XNOR instruction. The `.mm` suffix means the sources and destination are mask-register operands; each mask element is one bit, and the instruction computes vd.mask[i] = ~(vs2.mask[i] ^ vs1.mask[i]) by element index. These instructions are always unmasked, vm=0 encodings are reserved, and mask-tail bits are always tail-agnostic and have no deterministic value software may rely on.
VMXNOR.MM Decode And Execute Animation
Starts from V-extension encoding fields, then shows XNOR mask-logical field decode, mask-bit reads, per-lane execution, and vd writeback.
vd
vs2
vs1
vmxnor.mm
,
,
Execution Context
V-format mask logical encoding
31..26
25
24..20
19..15
14..12
11..7
6..0
lane
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
v8
1
0
1
1
0
1
0
0
1
0
1
1
0
1
0
0
XNOR
~^
~^
~^
~^
~^
~^
~^
~^
~^
~^
~^
~^
~^
~^
~^
~^
v16
1
1
0
0
1
1
0
0
1
1
0
0
1
1
0
0
v0
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
This animation fixes vstart=0 and models vd.mask[i] = ~(vs2.mask[i] ^ vs1.mask[i]) only for the shown VL body elements. Source rows show the original operand values used by the model; the specification permits vd to overlap either source. The final vd appears at Writeback, and successful completion leaves vstart at zero. The unshown mask tail is always tail-agnostic and may retain old bits, become 1s, or contain a result allowed by the specification, so the animation does not present a deterministic tail value.
Quick Understanding & Search Notes
VMXNOR.MM is the vector mask XNOR member of the RVV mask-register logical instructions, used for predicate composition. The `.mm` suffix means vd, vs2, and vs1 have mask-register semantics; each element is one bit, independent of the current data SEW.
Operand semantics: vd can be any mask register and receives the result mask; vs2 and vs1 provide the two source masks. For element i, the Boolean formula is vd.mask[i] = ~(vs2.mask[i] ^ vs1.mask[i]).
Each mask-register element is one bit, and every operand always occupies one vector register regardless of vlmul without changing vlmul; vd may be the same register as either source.
Mask logical instructions are always unmasked; they do not use v0.t to limit this instruction's execution, and vm=0 encodings are reserved. vd can be any mask register; only a result written or moved into v0 is later consumed through v0.t by ordinary masked vector instructions.
It selects lanes where the two predicates are equal; a result bit of 1 means both source masks are true or both are false at that lane. C semantic analogy: maskC = (a > 0) == (b < 10); generate two comparison masks, then apply XNOR bit by bit.
Typical workflow: compare instructions such as vmsgt/vmslt/vmseq first generate predicate masks, VMXNOR.MM then combines them with XNOR, and, when the composed result is in v0, a later maskable vector instruction consumes that mask through v0.t.
Execution boundary: destination bits below vstart remain unchanged, body bits from vstart through vl-1 are evaluated, and successful completion resets vstart to zero. If vstart >= vl (including vl=0), no destination bits are updated. The mask tail after vl is always tail-agnostic regardless of vta; it may retain old bits, become 1s, or contain a result allowed by the specification, so software cannot depend on a deterministic tail value.
Vector Execution Context
When reading VMXNOR.MM, focus on bit-by-bit predicate composition in mask registers. .mm: sources and destination are mask registers, operating on one predicate bit per element.
.mm operands
vd can be any mask register receiving the result mask, while vs2 and vs1 provide the two source masks combined bit by bit at the same element index.
v0.t relation
Mask logical instructions are always unmasked; only a result written or moved into v0 is later consumed through v0.t by ordinary masked vector instructions.
Common Usage Scenarios
Predicate condition composition
Understand this scenario with real code like «vmxnor.mm v1, v2, v3».
Mask logical family comparison
Understand this scenario with real code like «vmxnor.mm v1, v2, v3».
Mask construction before masked execution
Understand this scenario with real code like «vmxnor.mm v1, v2, v3».
Pre-Use Checklist
Syntax Check
- Confirm the current instruction format is OPMVV (OP-V).
- Confirm the operand order matches the example.
Semantic Check
- Ensure the destination register usage is compatible with the calling convention.
- Confirm this is not the lower-level form of a pseudo-instruction expansion.
Pitfalls / Common Confusions
Assembly operand order is vd, vs2, vs1; ANDN/ORN invert vs1.
These mask logical instructions are always unmasked and vm=0 encodings are reserved; do not treat v0.t as this instruction's execution mask.
They operate on one-bit mask elements, not ordinary SEW-wide vector data elements; do not confuse them with vand/vor/vxor.
FAQ
What does the `.mm` suffix mean?
It means a mask-mask form: vd, vs2, and vs1 are interpreted as mask registers, operating on one predicate bit per element.
What is the Boolean formula for VMXNOR.MM?
The formula is vd.mask[i] = ~(vs2.mask[i] ^ vs1.mask[i]).
Is it controlled by the v0 mask?
No. Vector mask logical instructions are always unmasked and vm=0 encodings are reserved; vd can be any mask register. Only a result written or moved into v0 is later consumed through v0.t by ordinary masked vector instructions.
What does a VMXNOR.MM result bit of 1 mean?
It means the two source masks are equal at that lane: both 1 or both 0.