FMOV (general)

Floating-point Move to or from general-purpose register without conversion. This instruction transfers the contents of a SIMD&FP register to a general-purpose register, or the contents of a general-purpose register to a SIMD&FP register.

Depending on the settings in the CPACR_EL1, CPTR_EL2, and CPTR_EL3 registers, and the current Security state and Exception level, an attempt to execute the instruction might be trapped.

313029282726252423222120191817161514131211109876543210
sf0011110ftype10x11x000000RnRd
rmodeopcode

Half-precision to 32-bit (sf == 0 && ftype == 11 && rmode == 00 && opcode == 110)
(FEAT_FP16)

FMOV <Wd>, <Hn>

Half-precision to 64-bit (sf == 1 && ftype == 11 && rmode == 00 && opcode == 110)
(FEAT_FP16)

FMOV <Xd>, <Hn>

32-bit to half-precision (sf == 0 && ftype == 11 && rmode == 00 && opcode == 111)
(FEAT_FP16)

FMOV <Hd>, <Wn>

32-bit to single-precision (sf == 0 && ftype == 00 && rmode == 00 && opcode == 111)

FMOV <Sd>, <Wn>

Single-precision to 32-bit (sf == 0 && ftype == 00 && rmode == 00 && opcode == 110)

FMOV <Wd>, <Sn>

64-bit to half-precision (sf == 1 && ftype == 11 && rmode == 00 && opcode == 111)
(FEAT_FP16)

FMOV <Hd>, <Xn>

64-bit to double-precision (sf == 1 && ftype == 01 && rmode == 00 && opcode == 111)

FMOV <Dd>, <Xn>

64-bit to top half of 128-bit (sf == 1 && ftype == 10 && rmode == 01 && opcode == 111)

FMOV <Vd>.D[1], <Xn>

Double-precision to 64-bit (sf == 1 && ftype == 01 && rmode == 00 && opcode == 110)

FMOV <Xd>, <Dn>

Top half of 128-bit to 64-bit (sf == 1 && ftype == 10 && rmode == 01 && opcode == 110)

FMOV <Xd>, <Vn>.D[1]

integer d = UInt(Rd); integer n = UInt(Rn); integer intsize = if sf == '1' then 64 else 32; integer fltsize; FPConvOp op; FPRounding rounding; boolean unsigned; integer part; case ftype of when '00' fltsize = 32; when '01' fltsize = 64; when '10' if opcode<2:1>:rmode != '11 01' then UNDEFINED; fltsize = 128; when '11' if HaveFP16Ext() then fltsize = 16; else UNDEFINED; case opcode<2:1>:rmode of when '00 xx' // FCVT[NPMZ][US] rounding = FPDecodeRounding(rmode); unsigned = (opcode<0> == '1'); op = FPConvOp_CVT_FtoI; when '01 00' // [US]CVTF rounding = FPRoundingMode(FPCR[]); unsigned = (opcode<0> == '1'); op = FPConvOp_CVT_ItoF; when '10 00' // FCVTA[US] rounding = FPRounding_TIEAWAY; unsigned = (opcode<0> == '1'); op = FPConvOp_CVT_FtoI; when '11 00' // FMOV if fltsize != 16 && fltsize != intsize then UNDEFINED; op = if opcode<0> == '1' then FPConvOp_MOV_ItoF else FPConvOp_MOV_FtoI; part = 0; when '11 01' // FMOV D[1] if intsize != 64 || fltsize != 128 then UNDEFINED; op = if opcode<0> == '1' then FPConvOp_MOV_ItoF else FPConvOp_MOV_FtoI; part = 1; fltsize = 64; // size of D[1] is 64 when '11 11' // FJCVTZS if !HaveFJCVTZSExt() then UNDEFINED; rounding = FPRounding_ZERO; unsigned = (opcode<0> == '1'); op = FPConvOp_CVT_FtoI_JS; otherwise UNDEFINED;

Assembler Symbols

<Dd>

Is the 64-bit name of the SIMD&FP destination register, encoded in the "Rd" field.

<Hd>

Is the 16-bit name of the SIMD&FP destination register, encoded in the "Rd" field.

<Sd>

Is the 32-bit name of the SIMD&FP destination register, encoded in the "Rd" field.

<Wn>

Is the 32-bit name of the general-purpose source register, encoded in the "Rn" field.

<Vd>

Is the name of the SIMD&FP destination register, encoded in the "Rd" field.

<Xn>

Is the 64-bit name of the general-purpose source register, encoded in the "Rn" field.

<Wd>

Is the 32-bit name of the general-purpose destination register, encoded in the "Rd" field.

<Sn>

Is the 32-bit name of the SIMD&FP source register, encoded in the "Rn" field.

<Xd>

Is the 64-bit name of the general-purpose destination register, encoded in the "Rd" field.

<Vn>

Is the name of the SIMD&FP source register, encoded in the "Rn" field.

<Hn>

Is the 16-bit name of the SIMD&FP source register, encoded in the "Rn" field.

<Dn>

Is the 64-bit name of the SIMD&FP source register, encoded in the "Rn" field.

Operation

if op == FPConvOp_CVT_FtoI_JS then CheckFPAdvSIMDEnabled64(); else CheckFPEnabled64(); FPCRType fpcr = FPCR[]; boolean merge = IsMerging(fpcr); integer fsize = if op == FPConvOp_CVT_ItoF && merge then 128 else fltsize; bits(fsize) fltval; bits(intsize) intval; case op of when FPConvOp_CVT_FtoI fltval = V[n]; intval = FPToFixed(fltval, 0, unsigned, fpcr, rounding); X[d] = intval; when FPConvOp_CVT_ItoF intval = X[n]; fltval = if merge then V[d] else Zeros(); Elem[fltval, 0, fltsize] = FixedToFP(intval, 0, unsigned, fpcr, rounding); V[d] = fltval; when FPConvOp_MOV_FtoI fltval = Vpart[n,part]; intval = ZeroExtend(fltval, intsize); X[d] = intval; when FPConvOp_MOV_ItoF intval = X[n]; fltval = intval<fsize-1:0>; Vpart[d,part] = fltval; when FPConvOp_CVT_FtoI_JS bit Z; fltval = V[n]; (intval, Z) = FPToFixedJS(fltval, fpcr, TRUE); PSTATE.<N,Z,C,V> = '0':Z:'00'; X[d] = intval;

Operational information

If FEAT_SME is implemented and the PE is in Streaming SVE mode, then any subsequent instruction which is dependent on the general-purpose register written by this instruction might be significantly delayed.


Internal version only: isa v33.11seprel, AdvSIMD v29.05, pseudocode v2021-09_rel, sve v2021-09_rc3d ; Build timestamp: 2021-10-06T11:41

Copyright © 2010-2021 Arm Limited or its affiliates. All rights reserved. This document is Non-Confidential.