movntss
Move Non-Temporal Scalar Single
MOVNTSS m32, xmm1
Stores scalar float bypassing cache (AMD SSE4a).
Details
Stores the low 32-bit scalar single-precision floating-point value from an XMM register to a 32-bit memory location, bypassing the cache hierarchy using non-temporal semantics. The instruction issues a weakly-ordered write that avoids polluting the L1/L2 caches, useful for streaming writes. No arithmetic flags are affected; this is a memory store only.
Pseudocode Operation
[dest] ← src[31:0];
Example
MOVNTSS [rbp-4], xmm1
Encoding
Binary Layout
F3
+0
0F
+1
2B
+2
Operands
-
dest
32-bit memory operand -
src
128-bit XMM SIMD register
Reference (AMD APM)
Description
Single-Precision Floating-Point
Stores one single-precision floating-point value from an XMM register to a 32-bit memory location.
This instruction indicates to the processor that the data is non-temporal, and is unlikely to be used again soon. The processor treats the store as a write-combining memory write, which minimizes cache pollution.
The diagram below illustrates the operation of this instruction:
mem32
XMM register
31 0 127 31 0 copy
Flags Affected
None 220 [AMDMOVNTSSPublic Use] Instruction Reference