1 # AArch32 VFP instruction descriptions (conditional insns)
3 # Copyright (c) 2019 Linaro, Ltd
5 # This library is free software; you can redistribute it and/or
6 # modify it under the terms of the GNU Lesser General Public
7 # License as published by the Free Software Foundation; either
8 # version 2 of the License, or (at your option) any later version.
10 # This library is distributed in the hope that it will be useful,
11 # but WITHOUT ANY WARRANTY; without even the implied warranty of
12 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 # Lesser General Public License for more details.
15 # You should have received a copy of the GNU Lesser General Public
16 # License along with this library; if not, see <http://www.gnu.org/licenses/>.
19 # This file is processed by scripts/decodetree.py
21 # Encodings for the conditional VFP instructions are here:
22 # generally anything matching A32
23 # cccc 11.. .... .... .... 101. .... ....
25 # 1110 110. .... .... .... 101. .... ....
26 # 1110 1110 .... .... .... 101. .... ....
27 # (but those patterns might also cover some Neon instructions,
28 # which do not live in this file.)
30 # VFP registers have an odd encoding with a four-bit field
31 # and a one-bit field which are assembled in different orders
32 # depending on whether the register is double or single precision.
33 # Each individual instruction function must do the checks for
34 # "double register selected but CPU does not have double support"
35 # and "double register number has bit 4 set but CPU does not
36 # support D16-D31" (which should UNDEF).
47 # VMOV scalar to general-purpose register; note that this does
48 # include some Neon cases.
49 VMOV_to_gp ---- 1110 u:1 1. 1 .... rt:4 1011 ... 1 0000 \
50 vn=%vn_dp size=0 index=%vmov_idx_b
51 VMOV_to_gp ---- 1110 u:1 0. 1 .... rt:4 1011 ..1 1 0000 \
52 vn=%vn_dp size=1 index=%vmov_idx_h
53 VMOV_to_gp ---- 1110 0 0 index:1 1 .... rt:4 1011 .00 1 0000 \
56 VMOV_from_gp ---- 1110 0 1. 0 .... rt:4 1011 ... 1 0000 \
57 vn=%vn_dp size=0 index=%vmov_idx_b
58 VMOV_from_gp ---- 1110 0 0. 0 .... rt:4 1011 ..1 1 0000 \
59 vn=%vn_dp size=1 index=%vmov_idx_h
60 VMOV_from_gp ---- 1110 0 0 index:1 0 .... rt:4 1011 .00 1 0000 \
63 VDUP ---- 1110 1 b:1 q:1 0 .... rt:4 1011 . 0 e:1 1 0000 \
66 VMSR_VMRS ---- 1110 111 l:1 reg:4 rt:4 1010 0001 0000
67 VMOV_single ---- 1110 000 l:1 .... rt:4 1010 . 001 0000 \
70 VMOV_64_sp ---- 1100 010 op:1 rt2:4 rt:4 1010 00.1 .... \
72 VMOV_64_dp ---- 1100 010 op:1 rt2:4 rt:4 1011 00.1 .... \
75 # Note that the half-precision variants of VLDR and VSTR are
76 # not part of this decodetree at all because they have bits [9:8] == 0b01
77 VLDR_VSTR_sp ---- 1101 u:1 .0 l:1 rn:4 .... 1010 imm:8 \
79 VLDR_VSTR_dp ---- 1101 u:1 .0 l:1 rn:4 .... 1011 imm:8 \
82 # We split the load/store multiple up into two patterns to avoid
83 # overlap with other insns in the "Advanced SIMD load/store and 64-bit move"
85 # P=0 U=0 W=0 is 64-bit VMOV
86 # P=1 W=0 is VLDR/VSTR
88 # leaving P=0 U=1 W=x and P=1 U=0 W=1 for load/store multiple.
89 # These include FSTM/FLDM.
90 VLDM_VSTM_sp ---- 1100 1 . w:1 l:1 rn:4 .... 1010 imm:8 \
92 VLDM_VSTM_dp ---- 1100 1 . w:1 l:1 rn:4 .... 1011 imm:8 \
95 VLDM_VSTM_sp ---- 1101 0.1 l:1 rn:4 .... 1010 imm:8 \
97 VLDM_VSTM_dp ---- 1101 0.1 l:1 rn:4 .... 1011 imm:8 \
100 # 3-register VFP data-processing; bits [23,21:20,6] identify the operation.
101 VMLA_sp ---- 1110 0.00 .... .... 1010 .0.0 .... \
102 vm=%vm_sp vn=%vn_sp vd=%vd_sp
103 VMLA_dp ---- 1110 0.00 .... .... 1011 .0.0 .... \
104 vm=%vm_dp vn=%vn_dp vd=%vd_dp
106 VMLS_sp ---- 1110 0.00 .... .... 1010 .1.0 .... \
107 vm=%vm_sp vn=%vn_sp vd=%vd_sp
108 VMLS_dp ---- 1110 0.00 .... .... 1011 .1.0 .... \
109 vm=%vm_dp vn=%vn_dp vd=%vd_dp
111 VNMLS_sp ---- 1110 0.01 .... .... 1010 .0.0 .... \
112 vm=%vm_sp vn=%vn_sp vd=%vd_sp
113 VNMLS_dp ---- 1110 0.01 .... .... 1011 .0.0 .... \
114 vm=%vm_dp vn=%vn_dp vd=%vd_dp
116 VNMLA_sp ---- 1110 0.01 .... .... 1010 .1.0 .... \
117 vm=%vm_sp vn=%vn_sp vd=%vd_sp
118 VNMLA_dp ---- 1110 0.01 .... .... 1011 .1.0 .... \
119 vm=%vm_dp vn=%vn_dp vd=%vd_dp
121 VMUL_sp ---- 1110 0.10 .... .... 1010 .0.0 .... \
122 vm=%vm_sp vn=%vn_sp vd=%vd_sp
123 VMUL_dp ---- 1110 0.10 .... .... 1011 .0.0 .... \
124 vm=%vm_dp vn=%vn_dp vd=%vd_dp
126 VNMUL_sp ---- 1110 0.10 .... .... 1010 .1.0 .... \
127 vm=%vm_sp vn=%vn_sp vd=%vd_sp
128 VNMUL_dp ---- 1110 0.10 .... .... 1011 .1.0 .... \
129 vm=%vm_dp vn=%vn_dp vd=%vd_dp
131 VADD_sp ---- 1110 0.11 .... .... 1010 .0.0 .... \
132 vm=%vm_sp vn=%vn_sp vd=%vd_sp
133 VADD_dp ---- 1110 0.11 .... .... 1011 .0.0 .... \
134 vm=%vm_dp vn=%vn_dp vd=%vd_dp
136 VSUB_sp ---- 1110 0.11 .... .... 1010 .1.0 .... \
137 vm=%vm_sp vn=%vn_sp vd=%vd_sp
138 VSUB_dp ---- 1110 0.11 .... .... 1011 .1.0 .... \
139 vm=%vm_dp vn=%vn_dp vd=%vd_dp
141 VDIV_sp ---- 1110 1.00 .... .... 1010 .0.0 .... \
142 vm=%vm_sp vn=%vn_sp vd=%vd_sp
143 VDIV_dp ---- 1110 1.00 .... .... 1011 .0.0 .... \
144 vm=%vm_dp vn=%vn_dp vd=%vd_dp
146 VFM_sp ---- 1110 1.01 .... .... 1010 . o2:1 . 0 .... \
147 vm=%vm_sp vn=%vn_sp vd=%vd_sp o1=1
148 VFM_dp ---- 1110 1.01 .... .... 1011 . o2:1 . 0 .... \
149 vm=%vm_dp vn=%vn_dp vd=%vd_dp o1=1
150 VFM_sp ---- 1110 1.10 .... .... 1010 . o2:1 . 0 .... \
151 vm=%vm_sp vn=%vn_sp vd=%vd_sp o1=2
152 VFM_dp ---- 1110 1.10 .... .... 1011 . o2:1 . 0 .... \
153 vm=%vm_dp vn=%vn_dp vd=%vd_dp o1=2
155 VMOV_imm_sp ---- 1110 1.11 imm4h:4 .... 1010 0000 imm4l:4 \
157 VMOV_imm_dp ---- 1110 1.11 imm4h:4 .... 1011 0000 imm4l:4 \
160 VMOV_reg_sp ---- 1110 1.11 0000 .... 1010 01.0 .... \
162 VMOV_reg_dp ---- 1110 1.11 0000 .... 1011 01.0 .... \
165 VABS_sp ---- 1110 1.11 0000 .... 1010 11.0 .... \
167 VABS_dp ---- 1110 1.11 0000 .... 1011 11.0 .... \
170 VNEG_sp ---- 1110 1.11 0001 .... 1010 01.0 .... \
172 VNEG_dp ---- 1110 1.11 0001 .... 1011 01.0 .... \
175 VSQRT_sp ---- 1110 1.11 0001 .... 1010 11.0 .... \
177 VSQRT_dp ---- 1110 1.11 0001 .... 1011 11.0 .... \
180 VCMP_sp ---- 1110 1.11 010 z:1 .... 1010 e:1 1.0 .... \
182 VCMP_dp ---- 1110 1.11 010 z:1 .... 1011 e:1 1.0 .... \
185 # VCVTT and VCVTB from f16: Vd format depends on size bit; Vm is always vm_sp
186 VCVT_f32_f16 ---- 1110 1.11 0010 .... 1010 t:1 1.0 .... \
188 VCVT_f64_f16 ---- 1110 1.11 0010 .... 1011 t:1 1.0 .... \
191 # VCVTB and VCVTT to f16: Vd format is always vd_sp; Vm format depends on size bit
192 VCVT_f16_f32 ---- 1110 1.11 0011 .... 1010 t:1 1.0 .... \
194 VCVT_f16_f64 ---- 1110 1.11 0011 .... 1011 t:1 1.0 .... \
197 VRINTR_sp ---- 1110 1.11 0110 .... 1010 01.0 .... \
199 VRINTR_dp ---- 1110 1.11 0110 .... 1011 01.0 .... \
202 VRINTZ_sp ---- 1110 1.11 0110 .... 1010 11.0 .... \
204 VRINTZ_dp ---- 1110 1.11 0110 .... 1011 11.0 .... \
207 VRINTX_sp ---- 1110 1.11 0111 .... 1010 01.0 .... \
209 VRINTX_dp ---- 1110 1.11 0111 .... 1011 01.0 .... \
212 # VCVT between single and double: Vm precision depends on size; Vd is its reverse
213 VCVT_sp ---- 1110 1.11 0111 .... 1010 11.0 .... \
215 VCVT_dp ---- 1110 1.11 0111 .... 1011 11.0 .... \
218 # VCVT from integer to floating point: Vm always single; Vd depends on size
219 VCVT_int_sp ---- 1110 1.11 1000 .... 1010 s:1 1.0 .... \
221 VCVT_int_dp ---- 1110 1.11 1000 .... 1011 s:1 1.0 .... \
224 # VJCVT is always dp to sp
225 VJCVT ---- 1110 1.11 1001 .... 1011 11.0 .... \
228 # VCVT between floating-point and fixed-point. The immediate value
229 # is in the same format as a Vm single-precision register number.
230 # We assemble bits 18 (op), 16 (u) and 7 (sx) into a single opc field
231 # for the convenience of the trans_VCVT_fix functions.
232 %vcvt_fix_op 18:1 16:1 7:1
233 VCVT_fix_sp ---- 1110 1.11 1.1. .... 1010 .1.0 .... \
234 vd=%vd_sp imm=%vm_sp opc=%vcvt_fix_op
235 VCVT_fix_dp ---- 1110 1.11 1.1. .... 1011 .1.0 .... \
236 vd=%vd_dp imm=%vm_sp opc=%vcvt_fix_op