1 //==- SystemZInstrVector.td - SystemZ Vector instructions ------*- tblgen-*-==//
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
7 //===----------------------------------------------------------------------===//
9 //===----------------------------------------------------------------------===//
11 //===----------------------------------------------------------------------===//
13 let Predicates = [FeatureVector] in {
15 def VLR : UnaryVRRa<"vlr", 0xE756, null_frag, v128any, v128any>;
16 def VLR32 : UnaryAliasVRR<null_frag, v32sb, v32sb>;
17 def VLR64 : UnaryAliasVRR<null_frag, v64db, v64db>;
19 // Load GR from VR element.
20 def VLGV : BinaryVRScGeneric<"vlgv", 0xE721>;
21 def VLGVB : BinaryVRSc<"vlgvb", 0xE721, null_frag, v128b, 0>;
22 def VLGVH : BinaryVRSc<"vlgvh", 0xE721, null_frag, v128h, 1>;
23 def VLGVF : BinaryVRSc<"vlgvf", 0xE721, null_frag, v128f, 2>;
24 def VLGVG : BinaryVRSc<"vlgvg", 0xE721, z_vector_extract, v128g, 3>;
26 // Load VR element from GR.
27 def VLVG : TernaryVRSbGeneric<"vlvg", 0xE722>;
28 def VLVGB : TernaryVRSb<"vlvgb", 0xE722, z_vector_insert,
29 v128b, v128b, GR32, 0>;
30 def VLVGH : TernaryVRSb<"vlvgh", 0xE722, z_vector_insert,
31 v128h, v128h, GR32, 1>;
32 def VLVGF : TernaryVRSb<"vlvgf", 0xE722, z_vector_insert,
33 v128f, v128f, GR32, 2>;
34 def VLVGG : TernaryVRSb<"vlvgg", 0xE722, z_vector_insert,
35 v128g, v128g, GR64, 3>;
37 // Load VR from GRs disjoint.
38 def VLVGP : BinaryVRRf<"vlvgp", 0xE762, z_join_dwords, v128g>;
39 def VLVGP32 : BinaryAliasVRRf<GR32>;
42 // Extractions always assign to the full GR64, even if the element would
43 // fit in the lower 32 bits. Sub-i64 extracts therefore need to take a
44 // subreg of the result.
45 class VectorExtractSubreg<ValueType type, Instruction insn>
46 : Pat<(i32 (z_vector_extract (type VR128:$vec), shift12only:$index)),
47 (EXTRACT_SUBREG (insn VR128:$vec, shift12only:$index), subreg_l32)>;
49 def : VectorExtractSubreg<v16i8, VLGVB>;
50 def : VectorExtractSubreg<v8i16, VLGVH>;
51 def : VectorExtractSubreg<v4i32, VLGVF>;
53 //===----------------------------------------------------------------------===//
54 // Immediate instructions
55 //===----------------------------------------------------------------------===//
57 let Predicates = [FeatureVector] in {
58 let isAsCheapAsAMove = 1, isMoveImm = 1, isReMaterializable = 1 in {
60 // Generate byte mask.
61 def VZERO : InherentVRIa<"vzero", 0xE744, 0>;
62 def VONE : InherentVRIa<"vone", 0xE744, 0xffff>;
63 def VGBM : UnaryVRIa<"vgbm", 0xE744, z_byte_mask, v128b, imm32zx16_timm>;
66 def VGM : BinaryVRIbGeneric<"vgm", 0xE746>;
67 def VGMB : BinaryVRIb<"vgmb", 0xE746, z_rotate_mask, v128b, 0>;
68 def VGMH : BinaryVRIb<"vgmh", 0xE746, z_rotate_mask, v128h, 1>;
69 def VGMF : BinaryVRIb<"vgmf", 0xE746, z_rotate_mask, v128f, 2>;
70 def VGMG : BinaryVRIb<"vgmg", 0xE746, z_rotate_mask, v128g, 3>;
72 // Replicate immediate.
73 def VREPI : UnaryVRIaGeneric<"vrepi", 0xE745, imm32sx16>;
74 def VREPIB : UnaryVRIa<"vrepib", 0xE745, z_replicate, v128b, imm32sx16_timm, 0>;
75 def VREPIH : UnaryVRIa<"vrepih", 0xE745, z_replicate, v128h, imm32sx16_timm, 1>;
76 def VREPIF : UnaryVRIa<"vrepif", 0xE745, z_replicate, v128f, imm32sx16_timm, 2>;
77 def VREPIG : UnaryVRIa<"vrepig", 0xE745, z_replicate, v128g, imm32sx16_timm, 3>;
80 // Load element immediate.
82 // We want these instructions to be used ahead of VLVG* where possible.
83 // However, VLVG* takes a variable BD-format index whereas VLEI takes
84 // a plain immediate index. This means that VLVG* has an extra "base"
85 // register operand and is 3 units more complex. Bumping the complexity
86 // of the VLEI* instructions by 4 means that they are strictly better
87 // than VLVG* in cases where both forms match.
88 let AddedComplexity = 4 in {
89 def VLEIB : TernaryVRIa<"vleib", 0xE740, z_vector_insert,
90 v128b, v128b, imm32sx16trunc, imm32zx4>;
91 def VLEIH : TernaryVRIa<"vleih", 0xE741, z_vector_insert,
92 v128h, v128h, imm32sx16trunc, imm32zx3>;
93 def VLEIF : TernaryVRIa<"vleif", 0xE743, z_vector_insert,
94 v128f, v128f, imm32sx16, imm32zx2>;
95 def VLEIG : TernaryVRIa<"vleig", 0xE742, z_vector_insert,
96 v128g, v128g, imm64sx16, imm32zx1>;
100 //===----------------------------------------------------------------------===//
102 //===----------------------------------------------------------------------===//
104 let Predicates = [FeatureVector] in {
106 defm VL : UnaryVRXAlign<"vl", 0xE706>;
108 // Load to block boundary. The number of loaded bytes is only known
109 // at run time. The instruction is really polymorphic, but v128b matches
110 // the return type of the associated intrinsic.
111 def VLBB : BinaryVRX<"vlbb", 0xE707, int_s390_vlbb, v128b, 0>;
113 // Load count to block boundary.
115 def LCBB : InstRXE<0xE727, (outs GR32:$R1),
116 (ins bdxaddr12only:$XBD2, imm32zx4:$M3),
117 "lcbb\t$R1, $XBD2, $M3",
118 [(set GR32:$R1, (int_s390_lcbb bdxaddr12only:$XBD2,
119 imm32zx4_timm:$M3))]>;
121 // Load with length. The number of loaded bytes is only known at run time.
122 def VLL : BinaryVRSb<"vll", 0xE737, int_s390_vll, 0>;
125 defm VLM : LoadMultipleVRSaAlign<"vlm", 0xE736>;
127 // Load and replicate
128 def VLREP : UnaryVRXGeneric<"vlrep", 0xE705>;
129 def VLREPB : UnaryVRX<"vlrepb", 0xE705, z_replicate_loadi8, v128b, 1, 0>;
130 def VLREPH : UnaryVRX<"vlreph", 0xE705, z_replicate_loadi16, v128h, 2, 1>;
131 def VLREPF : UnaryVRX<"vlrepf", 0xE705, z_replicate_loadi32, v128f, 4, 2>;
132 def VLREPG : UnaryVRX<"vlrepg", 0xE705, z_replicate_loadi64, v128g, 8, 3>;
133 def : Pat<(v4f32 (z_replicate_loadf32 bdxaddr12only:$addr)),
134 (VLREPF bdxaddr12only:$addr)>;
135 def : Pat<(v2f64 (z_replicate_loadf64 bdxaddr12only:$addr)),
136 (VLREPG bdxaddr12only:$addr)>;
138 // Use VLREP to load subvectors. These patterns use "12pair" because
139 // LEY and LDY offer full 20-bit displacement fields. It's often better
140 // to use those instructions rather than force a 20-bit displacement
141 // into a GPR temporary.
143 def VL32 : UnaryAliasVRX<load, v32sb, bdxaddr12pair>;
144 def VL64 : UnaryAliasVRX<load, v64db, bdxaddr12pair>;
147 // Load logical element and zero.
148 def VLLEZ : UnaryVRXGeneric<"vllez", 0xE704>;
149 def VLLEZB : UnaryVRX<"vllezb", 0xE704, z_vllezi8, v128b, 1, 0>;
150 def VLLEZH : UnaryVRX<"vllezh", 0xE704, z_vllezi16, v128h, 2, 1>;
151 def VLLEZF : UnaryVRX<"vllezf", 0xE704, z_vllezi32, v128f, 4, 2>;
152 def VLLEZG : UnaryVRX<"vllezg", 0xE704, z_vllezi64, v128g, 8, 3>;
153 def : Pat<(z_vllezf32 bdxaddr12only:$addr),
154 (VLLEZF bdxaddr12only:$addr)>;
155 def : Pat<(z_vllezf64 bdxaddr12only:$addr),
156 (VLLEZG bdxaddr12only:$addr)>;
157 let Predicates = [FeatureVectorEnhancements1] in {
158 def VLLEZLF : UnaryVRX<"vllezlf", 0xE704, z_vllezli32, v128f, 4, 6>;
159 def : Pat<(z_vllezlf32 bdxaddr12only:$addr),
160 (VLLEZLF bdxaddr12only:$addr)>;
164 def VLEB : TernaryVRX<"vleb", 0xE700, z_vlei8, v128b, v128b, 1, imm32zx4>;
165 def VLEH : TernaryVRX<"vleh", 0xE701, z_vlei16, v128h, v128h, 2, imm32zx3>;
166 def VLEF : TernaryVRX<"vlef", 0xE703, z_vlei32, v128f, v128f, 4, imm32zx2>;
167 def VLEG : TernaryVRX<"vleg", 0xE702, z_vlei64, v128g, v128g, 8, imm32zx1>;
168 def : Pat<(z_vlef32 (v4f32 VR128:$val), bdxaddr12only:$addr, imm32zx2:$index),
169 (VLEF VR128:$val, bdxaddr12only:$addr, imm32zx2:$index)>;
170 def : Pat<(z_vlef64 (v2f64 VR128:$val), bdxaddr12only:$addr, imm32zx1:$index),
171 (VLEG VR128:$val, bdxaddr12only:$addr, imm32zx1:$index)>;
174 def VGEF : TernaryVRV<"vgef", 0xE713, 4, imm32zx2>;
175 def VGEG : TernaryVRV<"vgeg", 0xE712, 8, imm32zx1>;
178 let Predicates = [FeatureVectorPackedDecimal] in {
179 // Load rightmost with length. The number of loaded bytes is only known
180 // at run time. Note that while the instruction will accept immediate
181 // lengths larger that 15 at runtime, those will always result in a trap,
182 // so we never emit them here.
183 def VLRL : BinaryVSI<"vlrl", 0xE635, null_frag, 0>;
184 def VLRLR : BinaryVRSd<"vlrlr", 0xE637, int_s390_vlrl, 0>;
185 def : Pat<(int_s390_vlrl imm32zx4:$len, bdaddr12only:$addr),
186 (VLRL bdaddr12only:$addr, imm32zx4:$len)>;
189 // Use replicating loads if we're inserting a single element into an
190 // undefined vector. This avoids a false dependency on the previous
191 // register contents.
192 multiclass ReplicatePeephole<Instruction vlrep, ValueType vectype,
193 SDPatternOperator load, ValueType scalartype> {
194 def : Pat<(vectype (z_vector_insert
195 (undef), (scalartype (load bdxaddr12only:$addr)), 0)),
196 (vlrep bdxaddr12only:$addr)>;
197 def : Pat<(vectype (scalar_to_vector
198 (scalartype (load bdxaddr12only:$addr)))),
199 (vlrep bdxaddr12only:$addr)>;
201 defm : ReplicatePeephole<VLREPB, v16i8, anyextloadi8, i32>;
202 defm : ReplicatePeephole<VLREPH, v8i16, anyextloadi16, i32>;
203 defm : ReplicatePeephole<VLREPF, v4i32, load, i32>;
204 defm : ReplicatePeephole<VLREPG, v2i64, load, i64>;
205 defm : ReplicatePeephole<VLREPF, v4f32, load, f32>;
206 defm : ReplicatePeephole<VLREPG, v2f64, load, f64>;
208 //===----------------------------------------------------------------------===//
210 //===----------------------------------------------------------------------===//
212 let Predicates = [FeatureVector] in {
214 defm VST : StoreVRXAlign<"vst", 0xE70E>;
216 // Store with length. The number of stored bytes is only known at run time.
217 def VSTL : StoreLengthVRSb<"vstl", 0xE73F, int_s390_vstl, 0>;
220 defm VSTM : StoreMultipleVRSaAlign<"vstm", 0xE73E>;
223 def VSTEB : StoreBinaryVRX<"vsteb", 0xE708, z_vstei8, v128b, 1, imm32zx4>;
224 def VSTEH : StoreBinaryVRX<"vsteh", 0xE709, z_vstei16, v128h, 2, imm32zx3>;
225 def VSTEF : StoreBinaryVRX<"vstef", 0xE70B, z_vstei32, v128f, 4, imm32zx2>;
226 def VSTEG : StoreBinaryVRX<"vsteg", 0xE70A, z_vstei64, v128g, 8, imm32zx1>;
227 def : Pat<(z_vstef32 (v4f32 VR128:$val), bdxaddr12only:$addr,
229 (VSTEF VR128:$val, bdxaddr12only:$addr, imm32zx2:$index)>;
230 def : Pat<(z_vstef64 (v2f64 VR128:$val), bdxaddr12only:$addr,
232 (VSTEG VR128:$val, bdxaddr12only:$addr, imm32zx1:$index)>;
234 // Use VSTE to store subvectors. These patterns use "12pair" because
235 // STEY and STDY offer full 20-bit displacement fields. It's often better
236 // to use those instructions rather than force a 20-bit displacement
237 // into a GPR temporary.
238 let mayStore = 1 in {
239 def VST32 : StoreAliasVRX<store, v32sb, bdxaddr12pair>;
240 def VST64 : StoreAliasVRX<store, v64db, bdxaddr12pair>;
244 def VSCEF : StoreBinaryVRV<"vscef", 0xE71B, 4, imm32zx2>;
245 def VSCEG : StoreBinaryVRV<"vsceg", 0xE71A, 8, imm32zx1>;
248 let Predicates = [FeatureVectorPackedDecimal] in {
249 // Store rightmost with length. The number of stored bytes is only known
250 // at run time. Note that while the instruction will accept immediate
251 // lengths larger that 15 at runtime, those will always result in a trap,
252 // so we never emit them here.
253 def VSTRL : StoreLengthVSI<"vstrl", 0xE63D, null_frag, 0>;
254 def VSTRLR : StoreLengthVRSd<"vstrlr", 0xE63F, int_s390_vstrl, 0>;
255 def : Pat<(int_s390_vstrl VR128:$val, imm32zx4:$len, bdaddr12only:$addr),
256 (VSTRL VR128:$val, bdaddr12only:$addr, imm32zx4:$len)>;
259 //===----------------------------------------------------------------------===//
261 //===----------------------------------------------------------------------===//
263 let Predicates = [FeatureVectorEnhancements2] in {
264 // Load byte-reversed elements.
265 def VLBR : UnaryVRXGeneric<"vlbr", 0xE606>;
266 def VLBRH : UnaryVRX<"vlbrh", 0xE606, z_loadbswap, v128h, 16, 1>;
267 def VLBRF : UnaryVRX<"vlbrf", 0xE606, z_loadbswap, v128f, 16, 2>;
268 def VLBRG : UnaryVRX<"vlbrg", 0xE606, z_loadbswap, v128g, 16, 3>;
269 def VLBRQ : UnaryVRX<"vlbrq", 0xE606, null_frag, v128q, 16, 4>;
271 // Load elements reversed.
272 def VLER : UnaryVRXGeneric<"vler", 0xE607>;
273 def VLERH : UnaryVRX<"vlerh", 0xE607, z_loadeswap, v128h, 16, 1>;
274 def VLERF : UnaryVRX<"vlerf", 0xE607, z_loadeswap, v128f, 16, 2>;
275 def VLERG : UnaryVRX<"vlerg", 0xE607, z_loadeswap, v128g, 16, 3>;
276 def : Pat<(v4f32 (z_loadeswap bdxaddr12only:$addr)),
277 (VLERF bdxaddr12only:$addr)>;
278 def : Pat<(v2f64 (z_loadeswap bdxaddr12only:$addr)),
279 (VLERG bdxaddr12only:$addr)>;
280 def : Pat<(v16i8 (z_loadeswap bdxaddr12only:$addr)),
281 (VLBRQ bdxaddr12only:$addr)>;
283 // Load byte-reversed element.
284 def VLEBRH : TernaryVRX<"vlebrh", 0xE601, z_vlebri16, v128h, v128h, 2, imm32zx3>;
285 def VLEBRF : TernaryVRX<"vlebrf", 0xE603, z_vlebri32, v128f, v128f, 4, imm32zx2>;
286 def VLEBRG : TernaryVRX<"vlebrg", 0xE602, z_vlebri64, v128g, v128g, 8, imm32zx1>;
288 // Load byte-reversed element and zero.
289 def VLLEBRZ : UnaryVRXGeneric<"vllebrz", 0xE604>;
290 def VLLEBRZH : UnaryVRX<"vllebrzh", 0xE604, z_vllebrzi16, v128h, 2, 1>;
291 def VLLEBRZF : UnaryVRX<"vllebrzf", 0xE604, z_vllebrzi32, v128f, 4, 2>;
292 def VLLEBRZG : UnaryVRX<"vllebrzg", 0xE604, z_vllebrzi64, v128g, 8, 3>;
293 def VLLEBRZE : UnaryVRX<"vllebrze", 0xE604, z_vllebrzli32, v128f, 4, 6>;
294 def : InstAlias<"lerv\t$V1, $XBD2",
295 (VLLEBRZE VR128:$V1, bdxaddr12only:$XBD2), 0>;
296 def : InstAlias<"ldrv\t$V1, $XBD2",
297 (VLLEBRZG VR128:$V1, bdxaddr12only:$XBD2), 0>;
299 // Load byte-reversed element and replicate.
300 def VLBRREP : UnaryVRXGeneric<"vlbrrep", 0xE605>;
301 def VLBRREPH : UnaryVRX<"vlbrreph", 0xE605, z_replicate_loadbswapi16, v128h, 2, 1>;
302 def VLBRREPF : UnaryVRX<"vlbrrepf", 0xE605, z_replicate_loadbswapi32, v128f, 4, 2>;
303 def VLBRREPG : UnaryVRX<"vlbrrepg", 0xE605, z_replicate_loadbswapi64, v128g, 8, 3>;
305 // Store byte-reversed elements.
306 def VSTBR : StoreVRXGeneric<"vstbr", 0xE60E>;
307 def VSTBRH : StoreVRX<"vstbrh", 0xE60E, z_storebswap, v128h, 16, 1>;
308 def VSTBRF : StoreVRX<"vstbrf", 0xE60E, z_storebswap, v128f, 16, 2>;
309 def VSTBRG : StoreVRX<"vstbrg", 0xE60E, z_storebswap, v128g, 16, 3>;
310 def VSTBRQ : StoreVRX<"vstbrq", 0xE60E, null_frag, v128q, 16, 4>;
312 // Store elements reversed.
313 def VSTER : StoreVRXGeneric<"vster", 0xE60F>;
314 def VSTERH : StoreVRX<"vsterh", 0xE60F, z_storeeswap, v128h, 16, 1>;
315 def VSTERF : StoreVRX<"vsterf", 0xE60F, z_storeeswap, v128f, 16, 2>;
316 def VSTERG : StoreVRX<"vsterg", 0xE60F, z_storeeswap, v128g, 16, 3>;
317 def : Pat<(z_storeeswap (v4f32 VR128:$val), bdxaddr12only:$addr),
318 (VSTERF VR128:$val, bdxaddr12only:$addr)>;
319 def : Pat<(z_storeeswap (v2f64 VR128:$val), bdxaddr12only:$addr),
320 (VSTERG VR128:$val, bdxaddr12only:$addr)>;
321 def : Pat<(z_storeeswap (v16i8 VR128:$val), bdxaddr12only:$addr),
322 (VSTBRQ VR128:$val, bdxaddr12only:$addr)>;
324 // Store byte-reversed element.
325 def VSTEBRH : StoreBinaryVRX<"vstebrh", 0xE609, z_vstebri16, v128h, 2, imm32zx3>;
326 def VSTEBRF : StoreBinaryVRX<"vstebrf", 0xE60B, z_vstebri32, v128f, 4, imm32zx2>;
327 def VSTEBRG : StoreBinaryVRX<"vstebrg", 0xE60A, z_vstebri64, v128g, 8, imm32zx1>;
328 def : InstAlias<"sterv\t$V1, $XBD2",
329 (VSTEBRF VR128:$V1, bdxaddr12only:$XBD2, 0), 0>;
330 def : InstAlias<"stdrv\t$V1, $XBD2",
331 (VSTEBRG VR128:$V1, bdxaddr12only:$XBD2, 0), 0>;
334 //===----------------------------------------------------------------------===//
335 // Selects and permutes
336 //===----------------------------------------------------------------------===//
338 let Predicates = [FeatureVector] in {
340 def VMRH: BinaryVRRcGeneric<"vmrh", 0xE761>;
341 def VMRHB : BinaryVRRc<"vmrhb", 0xE761, z_merge_high, v128b, v128b, 0>;
342 def VMRHH : BinaryVRRc<"vmrhh", 0xE761, z_merge_high, v128h, v128h, 1>;
343 def VMRHF : BinaryVRRc<"vmrhf", 0xE761, z_merge_high, v128f, v128f, 2>;
344 def VMRHG : BinaryVRRc<"vmrhg", 0xE761, z_merge_high, v128g, v128g, 3>;
345 def : BinaryRRWithType<VMRHF, VR128, z_merge_high, v4f32>;
346 def : BinaryRRWithType<VMRHG, VR128, z_merge_high, v2f64>;
349 def VMRL: BinaryVRRcGeneric<"vmrl", 0xE760>;
350 def VMRLB : BinaryVRRc<"vmrlb", 0xE760, z_merge_low, v128b, v128b, 0>;
351 def VMRLH : BinaryVRRc<"vmrlh", 0xE760, z_merge_low, v128h, v128h, 1>;
352 def VMRLF : BinaryVRRc<"vmrlf", 0xE760, z_merge_low, v128f, v128f, 2>;
353 def VMRLG : BinaryVRRc<"vmrlg", 0xE760, z_merge_low, v128g, v128g, 3>;
354 def : BinaryRRWithType<VMRLF, VR128, z_merge_low, v4f32>;
355 def : BinaryRRWithType<VMRLG, VR128, z_merge_low, v2f64>;
358 def VPERM : TernaryVRRe<"vperm", 0xE78C, z_permute, v128b, v128b>;
360 // Permute doubleword immediate.
361 def VPDI : TernaryVRRc<"vpdi", 0xE784, z_permute_dwords, v128g, v128g>;
364 let Predicates = [FeatureVectorEnhancements1] in
365 def VBPERM : BinaryVRRc<"vbperm", 0xE785, int_s390_vbperm, v128g, v128b>;
368 def VREP: BinaryVRIcGeneric<"vrep", 0xE74D>;
369 def VREPB : BinaryVRIc<"vrepb", 0xE74D, z_splat, v128b, v128b, 0>;
370 def VREPH : BinaryVRIc<"vreph", 0xE74D, z_splat, v128h, v128h, 1>;
371 def VREPF : BinaryVRIc<"vrepf", 0xE74D, z_splat, v128f, v128f, 2>;
372 def VREPG : BinaryVRIc<"vrepg", 0xE74D, z_splat, v128g, v128g, 3>;
373 def : Pat<(v4f32 (z_splat VR128:$vec, imm32zx16_timm:$index)),
374 (VREPF VR128:$vec, imm32zx16:$index)>;
375 def : Pat<(v2f64 (z_splat VR128:$vec, imm32zx16_timm:$index)),
376 (VREPG VR128:$vec, imm32zx16:$index)>;
379 def VSEL : TernaryVRRe<"vsel", 0xE78D, null_frag, v128any, v128any>;
382 //===----------------------------------------------------------------------===//
383 // Widening and narrowing
384 //===----------------------------------------------------------------------===//
386 let Predicates = [FeatureVector] in {
388 def VPK : BinaryVRRcGeneric<"vpk", 0xE794>;
389 def VPKH : BinaryVRRc<"vpkh", 0xE794, z_pack, v128b, v128h, 1>;
390 def VPKF : BinaryVRRc<"vpkf", 0xE794, z_pack, v128h, v128f, 2>;
391 def VPKG : BinaryVRRc<"vpkg", 0xE794, z_pack, v128f, v128g, 3>;
394 def VPKS : BinaryVRRbSPairGeneric<"vpks", 0xE797>;
395 defm VPKSH : BinaryVRRbSPair<"vpksh", 0xE797, int_s390_vpksh, z_packs_cc,
397 defm VPKSF : BinaryVRRbSPair<"vpksf", 0xE797, int_s390_vpksf, z_packs_cc,
399 defm VPKSG : BinaryVRRbSPair<"vpksg", 0xE797, int_s390_vpksg, z_packs_cc,
402 // Pack saturate logical.
403 def VPKLS : BinaryVRRbSPairGeneric<"vpkls", 0xE795>;
404 defm VPKLSH : BinaryVRRbSPair<"vpklsh", 0xE795, int_s390_vpklsh, z_packls_cc,
406 defm VPKLSF : BinaryVRRbSPair<"vpklsf", 0xE795, int_s390_vpklsf, z_packls_cc,
408 defm VPKLSG : BinaryVRRbSPair<"vpklsg", 0xE795, int_s390_vpklsg, z_packls_cc,
411 // Sign-extend to doubleword.
412 def VSEG : UnaryVRRaGeneric<"vseg", 0xE75F>;
413 def VSEGB : UnaryVRRa<"vsegb", 0xE75F, z_vsei8, v128g, v128g, 0>;
414 def VSEGH : UnaryVRRa<"vsegh", 0xE75F, z_vsei16, v128g, v128g, 1>;
415 def VSEGF : UnaryVRRa<"vsegf", 0xE75F, z_vsei32, v128g, v128g, 2>;
416 def : Pat<(z_vsei8_by_parts (v16i8 VR128:$src)), (VSEGB VR128:$src)>;
417 def : Pat<(z_vsei16_by_parts (v8i16 VR128:$src)), (VSEGH VR128:$src)>;
418 def : Pat<(z_vsei32_by_parts (v4i32 VR128:$src)), (VSEGF VR128:$src)>;
421 def VUPH : UnaryVRRaGeneric<"vuph", 0xE7D7>;
422 def VUPHB : UnaryVRRa<"vuphb", 0xE7D7, z_unpack_high, v128h, v128b, 0>;
423 def VUPHH : UnaryVRRa<"vuphh", 0xE7D7, z_unpack_high, v128f, v128h, 1>;
424 def VUPHF : UnaryVRRa<"vuphf", 0xE7D7, z_unpack_high, v128g, v128f, 2>;
426 // Unpack logical high.
427 def VUPLH : UnaryVRRaGeneric<"vuplh", 0xE7D5>;
428 def VUPLHB : UnaryVRRa<"vuplhb", 0xE7D5, z_unpackl_high, v128h, v128b, 0>;
429 def VUPLHH : UnaryVRRa<"vuplhh", 0xE7D5, z_unpackl_high, v128f, v128h, 1>;
430 def VUPLHF : UnaryVRRa<"vuplhf", 0xE7D5, z_unpackl_high, v128g, v128f, 2>;
433 def VUPL : UnaryVRRaGeneric<"vupl", 0xE7D6>;
434 def VUPLB : UnaryVRRa<"vuplb", 0xE7D6, z_unpack_low, v128h, v128b, 0>;
435 def VUPLHW : UnaryVRRa<"vuplhw", 0xE7D6, z_unpack_low, v128f, v128h, 1>;
436 def VUPLF : UnaryVRRa<"vuplf", 0xE7D6, z_unpack_low, v128g, v128f, 2>;
438 // Unpack logical low.
439 def VUPLL : UnaryVRRaGeneric<"vupll", 0xE7D4>;
440 def VUPLLB : UnaryVRRa<"vupllb", 0xE7D4, z_unpackl_low, v128h, v128b, 0>;
441 def VUPLLH : UnaryVRRa<"vupllh", 0xE7D4, z_unpackl_low, v128f, v128h, 1>;
442 def VUPLLF : UnaryVRRa<"vupllf", 0xE7D4, z_unpackl_low, v128g, v128f, 2>;
445 //===----------------------------------------------------------------------===//
446 // Instantiating generic operations for specific types.
447 //===----------------------------------------------------------------------===//
449 multiclass GenericVectorOps<ValueType type, ValueType inttype> {
450 let Predicates = [FeatureVector] in {
451 def : Pat<(type (load bdxaddr12only:$addr)),
452 (VL bdxaddr12only:$addr)>;
453 def : Pat<(store (type VR128:$src), bdxaddr12only:$addr),
454 (VST VR128:$src, bdxaddr12only:$addr)>;
455 def : Pat<(type (vselect (inttype VR128:$x), VR128:$y, VR128:$z)),
456 (VSEL VR128:$y, VR128:$z, VR128:$x)>;
457 def : Pat<(type (vselect (inttype (z_vnot VR128:$x)), VR128:$y, VR128:$z)),
458 (VSEL VR128:$z, VR128:$y, VR128:$x)>;
462 defm : GenericVectorOps<v16i8, v16i8>;
463 defm : GenericVectorOps<v8i16, v8i16>;
464 defm : GenericVectorOps<v4i32, v4i32>;
465 defm : GenericVectorOps<v2i64, v2i64>;
466 defm : GenericVectorOps<v4f32, v4i32>;
467 defm : GenericVectorOps<v2f64, v2i64>;
469 //===----------------------------------------------------------------------===//
470 // Integer arithmetic
471 //===----------------------------------------------------------------------===//
473 let Predicates = [FeatureVector] in {
474 let isCommutable = 1 in {
476 def VA : BinaryVRRcGeneric<"va", 0xE7F3>;
477 def VAB : BinaryVRRc<"vab", 0xE7F3, add, v128b, v128b, 0>;
478 def VAH : BinaryVRRc<"vah", 0xE7F3, add, v128h, v128h, 1>;
479 def VAF : BinaryVRRc<"vaf", 0xE7F3, add, v128f, v128f, 2>;
480 def VAG : BinaryVRRc<"vag", 0xE7F3, add, v128g, v128g, 3>;
481 def VAQ : BinaryVRRc<"vaq", 0xE7F3, int_s390_vaq, v128q, v128q, 4>;
484 let isCommutable = 1 in {
485 // Add compute carry.
486 def VACC : BinaryVRRcGeneric<"vacc", 0xE7F1>;
487 def VACCB : BinaryVRRc<"vaccb", 0xE7F1, int_s390_vaccb, v128b, v128b, 0>;
488 def VACCH : BinaryVRRc<"vacch", 0xE7F1, int_s390_vacch, v128h, v128h, 1>;
489 def VACCF : BinaryVRRc<"vaccf", 0xE7F1, int_s390_vaccf, v128f, v128f, 2>;
490 def VACCG : BinaryVRRc<"vaccg", 0xE7F1, int_s390_vaccg, v128g, v128g, 3>;
491 def VACCQ : BinaryVRRc<"vaccq", 0xE7F1, int_s390_vaccq, v128q, v128q, 4>;
494 def VAC : TernaryVRRdGeneric<"vac", 0xE7BB>;
495 def VACQ : TernaryVRRd<"vacq", 0xE7BB, int_s390_vacq, v128q, v128q, 4>;
497 // Add with carry compute carry.
498 def VACCC : TernaryVRRdGeneric<"vaccc", 0xE7B9>;
499 def VACCCQ : TernaryVRRd<"vacccq", 0xE7B9, int_s390_vacccq, v128q, v128q, 4>;
503 let isCommutable = 1 in
504 def VN : BinaryVRRc<"vn", 0xE768, null_frag, v128any, v128any>;
506 // And with complement.
507 def VNC : BinaryVRRc<"vnc", 0xE769, null_frag, v128any, v128any>;
509 let isCommutable = 1 in {
511 def VAVG : BinaryVRRcGeneric<"vavg", 0xE7F2>;
512 def VAVGB : BinaryVRRc<"vavgb", 0xE7F2, int_s390_vavgb, v128b, v128b, 0>;
513 def VAVGH : BinaryVRRc<"vavgh", 0xE7F2, int_s390_vavgh, v128h, v128h, 1>;
514 def VAVGF : BinaryVRRc<"vavgf", 0xE7F2, int_s390_vavgf, v128f, v128f, 2>;
515 def VAVGG : BinaryVRRc<"vavgg", 0xE7F2, int_s390_vavgg, v128g, v128g, 3>;
518 def VAVGL : BinaryVRRcGeneric<"vavgl", 0xE7F0>;
519 def VAVGLB : BinaryVRRc<"vavglb", 0xE7F0, int_s390_vavglb, v128b, v128b, 0>;
520 def VAVGLH : BinaryVRRc<"vavglh", 0xE7F0, int_s390_vavglh, v128h, v128h, 1>;
521 def VAVGLF : BinaryVRRc<"vavglf", 0xE7F0, int_s390_vavglf, v128f, v128f, 2>;
522 def VAVGLG : BinaryVRRc<"vavglg", 0xE7F0, int_s390_vavglg, v128g, v128g, 3>;
526 def VCKSM : BinaryVRRc<"vcksm", 0xE766, int_s390_vcksm, v128f, v128f>;
528 // Count leading zeros.
529 def VCLZ : UnaryVRRaGeneric<"vclz", 0xE753>;
530 def VCLZB : UnaryVRRa<"vclzb", 0xE753, ctlz, v128b, v128b, 0>;
531 def VCLZH : UnaryVRRa<"vclzh", 0xE753, ctlz, v128h, v128h, 1>;
532 def VCLZF : UnaryVRRa<"vclzf", 0xE753, ctlz, v128f, v128f, 2>;
533 def VCLZG : UnaryVRRa<"vclzg", 0xE753, ctlz, v128g, v128g, 3>;
535 // Count trailing zeros.
536 def VCTZ : UnaryVRRaGeneric<"vctz", 0xE752>;
537 def VCTZB : UnaryVRRa<"vctzb", 0xE752, cttz, v128b, v128b, 0>;
538 def VCTZH : UnaryVRRa<"vctzh", 0xE752, cttz, v128h, v128h, 1>;
539 def VCTZF : UnaryVRRa<"vctzf", 0xE752, cttz, v128f, v128f, 2>;
540 def VCTZG : UnaryVRRa<"vctzg", 0xE752, cttz, v128g, v128g, 3>;
542 let isCommutable = 1 in {
544 let Predicates = [FeatureVectorEnhancements1] in
545 def VNX : BinaryVRRc<"vnx", 0xE76C, null_frag, v128any, v128any>;
548 def VX : BinaryVRRc<"vx", 0xE76D, null_frag, v128any, v128any>;
551 // Galois field multiply sum.
552 def VGFM : BinaryVRRcGeneric<"vgfm", 0xE7B4>;
553 def VGFMB : BinaryVRRc<"vgfmb", 0xE7B4, int_s390_vgfmb, v128h, v128b, 0>;
554 def VGFMH : BinaryVRRc<"vgfmh", 0xE7B4, int_s390_vgfmh, v128f, v128h, 1>;
555 def VGFMF : BinaryVRRc<"vgfmf", 0xE7B4, int_s390_vgfmf, v128g, v128f, 2>;
556 def VGFMG : BinaryVRRc<"vgfmg", 0xE7B4, int_s390_vgfmg, v128q, v128g, 3>;
558 // Galois field multiply sum and accumulate.
559 def VGFMA : TernaryVRRdGeneric<"vgfma", 0xE7BC>;
560 def VGFMAB : TernaryVRRd<"vgfmab", 0xE7BC, int_s390_vgfmab, v128h, v128b, 0>;
561 def VGFMAH : TernaryVRRd<"vgfmah", 0xE7BC, int_s390_vgfmah, v128f, v128h, 1>;
562 def VGFMAF : TernaryVRRd<"vgfmaf", 0xE7BC, int_s390_vgfmaf, v128g, v128f, 2>;
563 def VGFMAG : TernaryVRRd<"vgfmag", 0xE7BC, int_s390_vgfmag, v128q, v128g, 3>;
566 def VLC : UnaryVRRaGeneric<"vlc", 0xE7DE>;
567 def VLCB : UnaryVRRa<"vlcb", 0xE7DE, z_vneg, v128b, v128b, 0>;
568 def VLCH : UnaryVRRa<"vlch", 0xE7DE, z_vneg, v128h, v128h, 1>;
569 def VLCF : UnaryVRRa<"vlcf", 0xE7DE, z_vneg, v128f, v128f, 2>;
570 def VLCG : UnaryVRRa<"vlcg", 0xE7DE, z_vneg, v128g, v128g, 3>;
573 def VLP : UnaryVRRaGeneric<"vlp", 0xE7DF>;
574 def VLPB : UnaryVRRa<"vlpb", 0xE7DF, abs, v128b, v128b, 0>;
575 def VLPH : UnaryVRRa<"vlph", 0xE7DF, abs, v128h, v128h, 1>;
576 def VLPF : UnaryVRRa<"vlpf", 0xE7DF, abs, v128f, v128f, 2>;
577 def VLPG : UnaryVRRa<"vlpg", 0xE7DF, abs, v128g, v128g, 3>;
579 let isCommutable = 1 in {
581 def VMX : BinaryVRRcGeneric<"vmx", 0xE7FF>;
582 def VMXB : BinaryVRRc<"vmxb", 0xE7FF, null_frag, v128b, v128b, 0>;
583 def VMXH : BinaryVRRc<"vmxh", 0xE7FF, null_frag, v128h, v128h, 1>;
584 def VMXF : BinaryVRRc<"vmxf", 0xE7FF, null_frag, v128f, v128f, 2>;
585 def VMXG : BinaryVRRc<"vmxg", 0xE7FF, null_frag, v128g, v128g, 3>;
588 def VMXL : BinaryVRRcGeneric<"vmxl", 0xE7FD>;
589 def VMXLB : BinaryVRRc<"vmxlb", 0xE7FD, null_frag, v128b, v128b, 0>;
590 def VMXLH : BinaryVRRc<"vmxlh", 0xE7FD, null_frag, v128h, v128h, 1>;
591 def VMXLF : BinaryVRRc<"vmxlf", 0xE7FD, null_frag, v128f, v128f, 2>;
592 def VMXLG : BinaryVRRc<"vmxlg", 0xE7FD, null_frag, v128g, v128g, 3>;
595 let isCommutable = 1 in {
597 def VMN : BinaryVRRcGeneric<"vmn", 0xE7FE>;
598 def VMNB : BinaryVRRc<"vmnb", 0xE7FE, null_frag, v128b, v128b, 0>;
599 def VMNH : BinaryVRRc<"vmnh", 0xE7FE, null_frag, v128h, v128h, 1>;
600 def VMNF : BinaryVRRc<"vmnf", 0xE7FE, null_frag, v128f, v128f, 2>;
601 def VMNG : BinaryVRRc<"vmng", 0xE7FE, null_frag, v128g, v128g, 3>;
604 def VMNL : BinaryVRRcGeneric<"vmnl", 0xE7FC>;
605 def VMNLB : BinaryVRRc<"vmnlb", 0xE7FC, null_frag, v128b, v128b, 0>;
606 def VMNLH : BinaryVRRc<"vmnlh", 0xE7FC, null_frag, v128h, v128h, 1>;
607 def VMNLF : BinaryVRRc<"vmnlf", 0xE7FC, null_frag, v128f, v128f, 2>;
608 def VMNLG : BinaryVRRc<"vmnlg", 0xE7FC, null_frag, v128g, v128g, 3>;
611 let isCommutable = 1 in {
612 // Multiply and add low.
613 def VMAL : TernaryVRRdGeneric<"vmal", 0xE7AA>;
614 def VMALB : TernaryVRRd<"vmalb", 0xE7AA, z_muladd, v128b, v128b, 0>;
615 def VMALHW : TernaryVRRd<"vmalhw", 0xE7AA, z_muladd, v128h, v128h, 1>;
616 def VMALF : TernaryVRRd<"vmalf", 0xE7AA, z_muladd, v128f, v128f, 2>;
618 // Multiply and add high.
619 def VMAH : TernaryVRRdGeneric<"vmah", 0xE7AB>;
620 def VMAHB : TernaryVRRd<"vmahb", 0xE7AB, int_s390_vmahb, v128b, v128b, 0>;
621 def VMAHH : TernaryVRRd<"vmahh", 0xE7AB, int_s390_vmahh, v128h, v128h, 1>;
622 def VMAHF : TernaryVRRd<"vmahf", 0xE7AB, int_s390_vmahf, v128f, v128f, 2>;
624 // Multiply and add logical high.
625 def VMALH : TernaryVRRdGeneric<"vmalh", 0xE7A9>;
626 def VMALHB : TernaryVRRd<"vmalhb", 0xE7A9, int_s390_vmalhb, v128b, v128b, 0>;
627 def VMALHH : TernaryVRRd<"vmalhh", 0xE7A9, int_s390_vmalhh, v128h, v128h, 1>;
628 def VMALHF : TernaryVRRd<"vmalhf", 0xE7A9, int_s390_vmalhf, v128f, v128f, 2>;
630 // Multiply and add even.
631 def VMAE : TernaryVRRdGeneric<"vmae", 0xE7AE>;
632 def VMAEB : TernaryVRRd<"vmaeb", 0xE7AE, int_s390_vmaeb, v128h, v128b, 0>;
633 def VMAEH : TernaryVRRd<"vmaeh", 0xE7AE, int_s390_vmaeh, v128f, v128h, 1>;
634 def VMAEF : TernaryVRRd<"vmaef", 0xE7AE, int_s390_vmaef, v128g, v128f, 2>;
636 // Multiply and add logical even.
637 def VMALE : TernaryVRRdGeneric<"vmale", 0xE7AC>;
638 def VMALEB : TernaryVRRd<"vmaleb", 0xE7AC, int_s390_vmaleb, v128h, v128b, 0>;
639 def VMALEH : TernaryVRRd<"vmaleh", 0xE7AC, int_s390_vmaleh, v128f, v128h, 1>;
640 def VMALEF : TernaryVRRd<"vmalef", 0xE7AC, int_s390_vmalef, v128g, v128f, 2>;
642 // Multiply and add odd.
643 def VMAO : TernaryVRRdGeneric<"vmao", 0xE7AF>;
644 def VMAOB : TernaryVRRd<"vmaob", 0xE7AF, int_s390_vmaob, v128h, v128b, 0>;
645 def VMAOH : TernaryVRRd<"vmaoh", 0xE7AF, int_s390_vmaoh, v128f, v128h, 1>;
646 def VMAOF : TernaryVRRd<"vmaof", 0xE7AF, int_s390_vmaof, v128g, v128f, 2>;
648 // Multiply and add logical odd.
649 def VMALO : TernaryVRRdGeneric<"vmalo", 0xE7AD>;
650 def VMALOB : TernaryVRRd<"vmalob", 0xE7AD, int_s390_vmalob, v128h, v128b, 0>;
651 def VMALOH : TernaryVRRd<"vmaloh", 0xE7AD, int_s390_vmaloh, v128f, v128h, 1>;
652 def VMALOF : TernaryVRRd<"vmalof", 0xE7AD, int_s390_vmalof, v128g, v128f, 2>;
655 let isCommutable = 1 in {
657 def VMH : BinaryVRRcGeneric<"vmh", 0xE7A3>;
658 def VMHB : BinaryVRRc<"vmhb", 0xE7A3, int_s390_vmhb, v128b, v128b, 0>;
659 def VMHH : BinaryVRRc<"vmhh", 0xE7A3, int_s390_vmhh, v128h, v128h, 1>;
660 def VMHF : BinaryVRRc<"vmhf", 0xE7A3, int_s390_vmhf, v128f, v128f, 2>;
662 // Multiply logical high.
663 def VMLH : BinaryVRRcGeneric<"vmlh", 0xE7A1>;
664 def VMLHB : BinaryVRRc<"vmlhb", 0xE7A1, int_s390_vmlhb, v128b, v128b, 0>;
665 def VMLHH : BinaryVRRc<"vmlhh", 0xE7A1, int_s390_vmlhh, v128h, v128h, 1>;
666 def VMLHF : BinaryVRRc<"vmlhf", 0xE7A1, int_s390_vmlhf, v128f, v128f, 2>;
669 def VML : BinaryVRRcGeneric<"vml", 0xE7A2>;
670 def VMLB : BinaryVRRc<"vmlb", 0xE7A2, mul, v128b, v128b, 0>;
671 def VMLHW : BinaryVRRc<"vmlhw", 0xE7A2, mul, v128h, v128h, 1>;
672 def VMLF : BinaryVRRc<"vmlf", 0xE7A2, mul, v128f, v128f, 2>;
675 def VME : BinaryVRRcGeneric<"vme", 0xE7A6>;
676 def VMEB : BinaryVRRc<"vmeb", 0xE7A6, int_s390_vmeb, v128h, v128b, 0>;
677 def VMEH : BinaryVRRc<"vmeh", 0xE7A6, int_s390_vmeh, v128f, v128h, 1>;
678 def VMEF : BinaryVRRc<"vmef", 0xE7A6, int_s390_vmef, v128g, v128f, 2>;
680 // Multiply logical even.
681 def VMLE : BinaryVRRcGeneric<"vmle", 0xE7A4>;
682 def VMLEB : BinaryVRRc<"vmleb", 0xE7A4, int_s390_vmleb, v128h, v128b, 0>;
683 def VMLEH : BinaryVRRc<"vmleh", 0xE7A4, int_s390_vmleh, v128f, v128h, 1>;
684 def VMLEF : BinaryVRRc<"vmlef", 0xE7A4, int_s390_vmlef, v128g, v128f, 2>;
687 def VMO : BinaryVRRcGeneric<"vmo", 0xE7A7>;
688 def VMOB : BinaryVRRc<"vmob", 0xE7A7, int_s390_vmob, v128h, v128b, 0>;
689 def VMOH : BinaryVRRc<"vmoh", 0xE7A7, int_s390_vmoh, v128f, v128h, 1>;
690 def VMOF : BinaryVRRc<"vmof", 0xE7A7, int_s390_vmof, v128g, v128f, 2>;
692 // Multiply logical odd.
693 def VMLO : BinaryVRRcGeneric<"vmlo", 0xE7A5>;
694 def VMLOB : BinaryVRRc<"vmlob", 0xE7A5, int_s390_vmlob, v128h, v128b, 0>;
695 def VMLOH : BinaryVRRc<"vmloh", 0xE7A5, int_s390_vmloh, v128f, v128h, 1>;
696 def VMLOF : BinaryVRRc<"vmlof", 0xE7A5, int_s390_vmlof, v128g, v128f, 2>;
699 // Multiply sum logical.
700 let Predicates = [FeatureVectorEnhancements1], isCommutable = 1 in {
701 def VMSL : QuaternaryVRRdGeneric<"vmsl", 0xE7B8>;
702 def VMSLG : QuaternaryVRRd<"vmslg", 0xE7B8, int_s390_vmslg,
703 v128q, v128g, v128g, v128q, 3>;
707 let Predicates = [FeatureVectorEnhancements1], isCommutable = 1 in
708 def VNN : BinaryVRRc<"vnn", 0xE76E, null_frag, v128any, v128any>;
711 let isCommutable = 1 in
712 def VNO : BinaryVRRc<"vno", 0xE76B, null_frag, v128any, v128any>;
713 def : InstAlias<"vnot\t$V1, $V2", (VNO VR128:$V1, VR128:$V2, VR128:$V2), 0>;
716 let isCommutable = 1 in
717 def VO : BinaryVRRc<"vo", 0xE76A, null_frag, v128any, v128any>;
719 // Or with complement.
720 let Predicates = [FeatureVectorEnhancements1] in
721 def VOC : BinaryVRRc<"voc", 0xE76F, null_frag, v128any, v128any>;
724 def VPOPCT : UnaryVRRaGeneric<"vpopct", 0xE750>;
725 def : Pat<(v16i8 (z_popcnt VR128:$x)), (VPOPCT VR128:$x, 0)>;
726 let Predicates = [FeatureVectorEnhancements1] in {
727 def VPOPCTB : UnaryVRRa<"vpopctb", 0xE750, ctpop, v128b, v128b, 0>;
728 def VPOPCTH : UnaryVRRa<"vpopcth", 0xE750, ctpop, v128h, v128h, 1>;
729 def VPOPCTF : UnaryVRRa<"vpopctf", 0xE750, ctpop, v128f, v128f, 2>;
730 def VPOPCTG : UnaryVRRa<"vpopctg", 0xE750, ctpop, v128g, v128g, 3>;
733 // Element rotate left logical (with vector shift amount).
734 def VERLLV : BinaryVRRcGeneric<"verllv", 0xE773>;
735 def VERLLVB : BinaryVRRc<"verllvb", 0xE773, int_s390_verllvb,
737 def VERLLVH : BinaryVRRc<"verllvh", 0xE773, int_s390_verllvh,
739 def VERLLVF : BinaryVRRc<"verllvf", 0xE773, int_s390_verllvf,
741 def VERLLVG : BinaryVRRc<"verllvg", 0xE773, int_s390_verllvg,
744 // Element rotate left logical (with scalar shift amount).
745 def VERLL : BinaryVRSaGeneric<"verll", 0xE733>;
746 def VERLLB : BinaryVRSa<"verllb", 0xE733, int_s390_verllb, v128b, v128b, 0>;
747 def VERLLH : BinaryVRSa<"verllh", 0xE733, int_s390_verllh, v128h, v128h, 1>;
748 def VERLLF : BinaryVRSa<"verllf", 0xE733, int_s390_verllf, v128f, v128f, 2>;
749 def VERLLG : BinaryVRSa<"verllg", 0xE733, int_s390_verllg, v128g, v128g, 3>;
751 // Element rotate and insert under mask.
752 def VERIM : QuaternaryVRIdGeneric<"verim", 0xE772>;
753 def VERIMB : QuaternaryVRId<"verimb", 0xE772, int_s390_verimb, v128b, v128b, 0>;
754 def VERIMH : QuaternaryVRId<"verimh", 0xE772, int_s390_verimh, v128h, v128h, 1>;
755 def VERIMF : QuaternaryVRId<"verimf", 0xE772, int_s390_verimf, v128f, v128f, 2>;
756 def VERIMG : QuaternaryVRId<"verimg", 0xE772, int_s390_verimg, v128g, v128g, 3>;
758 // Element shift left (with vector shift amount).
759 def VESLV : BinaryVRRcGeneric<"veslv", 0xE770>;
760 def VESLVB : BinaryVRRc<"veslvb", 0xE770, z_vshl, v128b, v128b, 0>;
761 def VESLVH : BinaryVRRc<"veslvh", 0xE770, z_vshl, v128h, v128h, 1>;
762 def VESLVF : BinaryVRRc<"veslvf", 0xE770, z_vshl, v128f, v128f, 2>;
763 def VESLVG : BinaryVRRc<"veslvg", 0xE770, z_vshl, v128g, v128g, 3>;
765 // Element shift left (with scalar shift amount).
766 def VESL : BinaryVRSaGeneric<"vesl", 0xE730>;
767 def VESLB : BinaryVRSa<"veslb", 0xE730, z_vshl_by_scalar, v128b, v128b, 0>;
768 def VESLH : BinaryVRSa<"veslh", 0xE730, z_vshl_by_scalar, v128h, v128h, 1>;
769 def VESLF : BinaryVRSa<"veslf", 0xE730, z_vshl_by_scalar, v128f, v128f, 2>;
770 def VESLG : BinaryVRSa<"veslg", 0xE730, z_vshl_by_scalar, v128g, v128g, 3>;
772 // Element shift right arithmetic (with vector shift amount).
773 def VESRAV : BinaryVRRcGeneric<"vesrav", 0xE77A>;
774 def VESRAVB : BinaryVRRc<"vesravb", 0xE77A, z_vsra, v128b, v128b, 0>;
775 def VESRAVH : BinaryVRRc<"vesravh", 0xE77A, z_vsra, v128h, v128h, 1>;
776 def VESRAVF : BinaryVRRc<"vesravf", 0xE77A, z_vsra, v128f, v128f, 2>;
777 def VESRAVG : BinaryVRRc<"vesravg", 0xE77A, z_vsra, v128g, v128g, 3>;
779 // Element shift right arithmetic (with scalar shift amount).
780 def VESRA : BinaryVRSaGeneric<"vesra", 0xE73A>;
781 def VESRAB : BinaryVRSa<"vesrab", 0xE73A, z_vsra_by_scalar, v128b, v128b, 0>;
782 def VESRAH : BinaryVRSa<"vesrah", 0xE73A, z_vsra_by_scalar, v128h, v128h, 1>;
783 def VESRAF : BinaryVRSa<"vesraf", 0xE73A, z_vsra_by_scalar, v128f, v128f, 2>;
784 def VESRAG : BinaryVRSa<"vesrag", 0xE73A, z_vsra_by_scalar, v128g, v128g, 3>;
786 // Element shift right logical (with vector shift amount).
787 def VESRLV : BinaryVRRcGeneric<"vesrlv", 0xE778>;
788 def VESRLVB : BinaryVRRc<"vesrlvb", 0xE778, z_vsrl, v128b, v128b, 0>;
789 def VESRLVH : BinaryVRRc<"vesrlvh", 0xE778, z_vsrl, v128h, v128h, 1>;
790 def VESRLVF : BinaryVRRc<"vesrlvf", 0xE778, z_vsrl, v128f, v128f, 2>;
791 def VESRLVG : BinaryVRRc<"vesrlvg", 0xE778, z_vsrl, v128g, v128g, 3>;
793 // Element shift right logical (with scalar shift amount).
794 def VESRL : BinaryVRSaGeneric<"vesrl", 0xE738>;
795 def VESRLB : BinaryVRSa<"vesrlb", 0xE738, z_vsrl_by_scalar, v128b, v128b, 0>;
796 def VESRLH : BinaryVRSa<"vesrlh", 0xE738, z_vsrl_by_scalar, v128h, v128h, 1>;
797 def VESRLF : BinaryVRSa<"vesrlf", 0xE738, z_vsrl_by_scalar, v128f, v128f, 2>;
798 def VESRLG : BinaryVRSa<"vesrlg", 0xE738, z_vsrl_by_scalar, v128g, v128g, 3>;
801 def VSL : BinaryVRRc<"vsl", 0xE774, int_s390_vsl, v128b, v128b>;
803 // Shift left by byte.
804 def VSLB : BinaryVRRc<"vslb", 0xE775, int_s390_vslb, v128b, v128b>;
806 // Shift left double by byte.
807 def VSLDB : TernaryVRId<"vsldb", 0xE777, z_shl_double, v128b, v128b, 0>;
808 def : Pat<(int_s390_vsldb VR128:$x, VR128:$y, imm32zx8_timm:$z),
809 (VSLDB VR128:$x, VR128:$y, imm32zx8:$z)>;
811 // Shift left double by bit.
812 let Predicates = [FeatureVectorEnhancements2] in
813 def VSLD : TernaryVRId<"vsld", 0xE786, int_s390_vsld, v128b, v128b, 0>;
815 // Shift right arithmetic.
816 def VSRA : BinaryVRRc<"vsra", 0xE77E, int_s390_vsra, v128b, v128b>;
818 // Shift right arithmetic by byte.
819 def VSRAB : BinaryVRRc<"vsrab", 0xE77F, int_s390_vsrab, v128b, v128b>;
821 // Shift right logical.
822 def VSRL : BinaryVRRc<"vsrl", 0xE77C, int_s390_vsrl, v128b, v128b>;
824 // Shift right logical by byte.
825 def VSRLB : BinaryVRRc<"vsrlb", 0xE77D, int_s390_vsrlb, v128b, v128b>;
827 // Shift right double by bit.
828 let Predicates = [FeatureVectorEnhancements2] in
829 def VSRD : TernaryVRId<"vsrd", 0xE787, int_s390_vsrd, v128b, v128b, 0>;
832 def VS : BinaryVRRcGeneric<"vs", 0xE7F7>;
833 def VSB : BinaryVRRc<"vsb", 0xE7F7, sub, v128b, v128b, 0>;
834 def VSH : BinaryVRRc<"vsh", 0xE7F7, sub, v128h, v128h, 1>;
835 def VSF : BinaryVRRc<"vsf", 0xE7F7, sub, v128f, v128f, 2>;
836 def VSG : BinaryVRRc<"vsg", 0xE7F7, sub, v128g, v128g, 3>;
837 def VSQ : BinaryVRRc<"vsq", 0xE7F7, int_s390_vsq, v128q, v128q, 4>;
839 // Subtract compute borrow indication.
840 def VSCBI : BinaryVRRcGeneric<"vscbi", 0xE7F5>;
841 def VSCBIB : BinaryVRRc<"vscbib", 0xE7F5, int_s390_vscbib, v128b, v128b, 0>;
842 def VSCBIH : BinaryVRRc<"vscbih", 0xE7F5, int_s390_vscbih, v128h, v128h, 1>;
843 def VSCBIF : BinaryVRRc<"vscbif", 0xE7F5, int_s390_vscbif, v128f, v128f, 2>;
844 def VSCBIG : BinaryVRRc<"vscbig", 0xE7F5, int_s390_vscbig, v128g, v128g, 3>;
845 def VSCBIQ : BinaryVRRc<"vscbiq", 0xE7F5, int_s390_vscbiq, v128q, v128q, 4>;
847 // Subtract with borrow indication.
848 def VSBI : TernaryVRRdGeneric<"vsbi", 0xE7BF>;
849 def VSBIQ : TernaryVRRd<"vsbiq", 0xE7BF, int_s390_vsbiq, v128q, v128q, 4>;
851 // Subtract with borrow compute borrow indication.
852 def VSBCBI : TernaryVRRdGeneric<"vsbcbi", 0xE7BD>;
853 def VSBCBIQ : TernaryVRRd<"vsbcbiq", 0xE7BD, int_s390_vsbcbiq,
856 // Sum across doubleword.
857 def VSUMG : BinaryVRRcGeneric<"vsumg", 0xE765>;
858 def VSUMGH : BinaryVRRc<"vsumgh", 0xE765, z_vsum, v128g, v128h, 1>;
859 def VSUMGF : BinaryVRRc<"vsumgf", 0xE765, z_vsum, v128g, v128f, 2>;
861 // Sum across quadword.
862 def VSUMQ : BinaryVRRcGeneric<"vsumq", 0xE767>;
863 def VSUMQF : BinaryVRRc<"vsumqf", 0xE767, z_vsum, v128q, v128f, 2>;
864 def VSUMQG : BinaryVRRc<"vsumqg", 0xE767, z_vsum, v128q, v128g, 3>;
867 def VSUM : BinaryVRRcGeneric<"vsum", 0xE764>;
868 def VSUMB : BinaryVRRc<"vsumb", 0xE764, z_vsum, v128f, v128b, 0>;
869 def VSUMH : BinaryVRRc<"vsumh", 0xE764, z_vsum, v128f, v128h, 1>;
872 // Instantiate the bitwise ops for type TYPE.
873 multiclass BitwiseVectorOps<ValueType type> {
874 let Predicates = [FeatureVector] in {
875 def : Pat<(type (and VR128:$x, VR128:$y)), (VN VR128:$x, VR128:$y)>;
876 def : Pat<(type (and VR128:$x, (z_vnot VR128:$y))),
877 (VNC VR128:$x, VR128:$y)>;
878 def : Pat<(type (or VR128:$x, VR128:$y)), (VO VR128:$x, VR128:$y)>;
879 def : Pat<(type (xor VR128:$x, VR128:$y)), (VX VR128:$x, VR128:$y)>;
880 def : Pat<(type (or (and VR128:$x, VR128:$z),
881 (and VR128:$y, (z_vnot VR128:$z)))),
882 (VSEL VR128:$x, VR128:$y, VR128:$z)>;
883 def : Pat<(type (z_vnot (or VR128:$x, VR128:$y))),
884 (VNO VR128:$x, VR128:$y)>;
885 def : Pat<(type (z_vnot VR128:$x)), (VNO VR128:$x, VR128:$x)>;
887 let Predicates = [FeatureVectorEnhancements1] in {
888 def : Pat<(type (z_vnot (xor VR128:$x, VR128:$y))),
889 (VNX VR128:$x, VR128:$y)>;
890 def : Pat<(type (z_vnot (and VR128:$x, VR128:$y))),
891 (VNN VR128:$x, VR128:$y)>;
892 def : Pat<(type (or VR128:$x, (z_vnot VR128:$y))),
893 (VOC VR128:$x, VR128:$y)>;
897 defm : BitwiseVectorOps<v16i8>;
898 defm : BitwiseVectorOps<v8i16>;
899 defm : BitwiseVectorOps<v4i32>;
900 defm : BitwiseVectorOps<v2i64>;
902 // Instantiate additional patterns for absolute-related expressions on
903 // type TYPE. LC is the negate instruction for TYPE and LP is the absolute
905 multiclass IntegerAbsoluteVectorOps<ValueType type, Instruction lc,
906 Instruction lp, int shift> {
907 let Predicates = [FeatureVector] in {
908 def : Pat<(type (vselect (type (z_vicmph_zero VR128:$x)),
909 (z_vneg VR128:$x), VR128:$x)),
911 def : Pat<(type (vselect (type (z_vnot (z_vicmph_zero VR128:$x))),
912 VR128:$x, (z_vneg VR128:$x))),
914 def : Pat<(type (vselect (type (z_vicmpl_zero VR128:$x)),
915 VR128:$x, (z_vneg VR128:$x))),
917 def : Pat<(type (vselect (type (z_vnot (z_vicmpl_zero VR128:$x))),
918 (z_vneg VR128:$x), VR128:$x)),
920 def : Pat<(type (or (and (z_vsra_by_scalar VR128:$x, (i32 shift)),
922 (and (z_vnot (z_vsra_by_scalar VR128:$x, (i32 shift))),
925 def : Pat<(type (or (and (z_vsra_by_scalar VR128:$x, (i32 shift)),
927 (and (z_vnot (z_vsra_by_scalar VR128:$x, (i32 shift))),
928 (z_vneg VR128:$x)))),
933 defm : IntegerAbsoluteVectorOps<v16i8, VLCB, VLPB, 7>;
934 defm : IntegerAbsoluteVectorOps<v8i16, VLCH, VLPH, 15>;
935 defm : IntegerAbsoluteVectorOps<v4i32, VLCF, VLPF, 31>;
936 defm : IntegerAbsoluteVectorOps<v2i64, VLCG, VLPG, 63>;
938 // Instantiate minimum- and maximum-related patterns for TYPE. CMPH is the
939 // signed or unsigned "set if greater than" comparison instruction and
940 // MIN and MAX are the associated minimum and maximum instructions.
941 multiclass IntegerMinMaxVectorOps<ValueType type, SDPatternOperator cmph,
942 Instruction min, Instruction max> {
943 let Predicates = [FeatureVector] in {
944 def : Pat<(type (vselect (cmph VR128:$x, VR128:$y), VR128:$x, VR128:$y)),
945 (max VR128:$x, VR128:$y)>;
946 def : Pat<(type (vselect (cmph VR128:$x, VR128:$y), VR128:$y, VR128:$x)),
947 (min VR128:$x, VR128:$y)>;
948 def : Pat<(type (vselect (z_vnot (cmph VR128:$x, VR128:$y)),
949 VR128:$x, VR128:$y)),
950 (min VR128:$x, VR128:$y)>;
951 def : Pat<(type (vselect (z_vnot (cmph VR128:$x, VR128:$y)),
952 VR128:$y, VR128:$x)),
953 (max VR128:$x, VR128:$y)>;
958 defm : IntegerMinMaxVectorOps<v16i8, z_vicmph, VMNB, VMXB>;
959 defm : IntegerMinMaxVectorOps<v8i16, z_vicmph, VMNH, VMXH>;
960 defm : IntegerMinMaxVectorOps<v4i32, z_vicmph, VMNF, VMXF>;
961 defm : IntegerMinMaxVectorOps<v2i64, z_vicmph, VMNG, VMXG>;
964 defm : IntegerMinMaxVectorOps<v16i8, z_vicmphl, VMNLB, VMXLB>;
965 defm : IntegerMinMaxVectorOps<v8i16, z_vicmphl, VMNLH, VMXLH>;
966 defm : IntegerMinMaxVectorOps<v4i32, z_vicmphl, VMNLF, VMXLF>;
967 defm : IntegerMinMaxVectorOps<v2i64, z_vicmphl, VMNLG, VMXLG>;
969 //===----------------------------------------------------------------------===//
970 // Integer comparison
971 //===----------------------------------------------------------------------===//
973 let Predicates = [FeatureVector] in {
976 def VEC : CompareVRRaGeneric<"vec", 0xE7DB>;
977 def VECB : CompareVRRa<"vecb", 0xE7DB, null_frag, v128b, 0>;
978 def VECH : CompareVRRa<"vech", 0xE7DB, null_frag, v128h, 1>;
979 def VECF : CompareVRRa<"vecf", 0xE7DB, null_frag, v128f, 2>;
980 def VECG : CompareVRRa<"vecg", 0xE7DB, null_frag, v128g, 3>;
983 // Element compare logical.
985 def VECL : CompareVRRaGeneric<"vecl", 0xE7D9>;
986 def VECLB : CompareVRRa<"veclb", 0xE7D9, null_frag, v128b, 0>;
987 def VECLH : CompareVRRa<"veclh", 0xE7D9, null_frag, v128h, 1>;
988 def VECLF : CompareVRRa<"veclf", 0xE7D9, null_frag, v128f, 2>;
989 def VECLG : CompareVRRa<"veclg", 0xE7D9, null_frag, v128g, 3>;
993 def VCEQ : BinaryVRRbSPairGeneric<"vceq", 0xE7F8>;
994 defm VCEQB : BinaryVRRbSPair<"vceqb", 0xE7F8, z_vicmpe, z_vicmpes,
996 defm VCEQH : BinaryVRRbSPair<"vceqh", 0xE7F8, z_vicmpe, z_vicmpes,
998 defm VCEQF : BinaryVRRbSPair<"vceqf", 0xE7F8, z_vicmpe, z_vicmpes,
1000 defm VCEQG : BinaryVRRbSPair<"vceqg", 0xE7F8, z_vicmpe, z_vicmpes,
1004 def VCH : BinaryVRRbSPairGeneric<"vch", 0xE7FB>;
1005 defm VCHB : BinaryVRRbSPair<"vchb", 0xE7FB, z_vicmph, z_vicmphs,
1007 defm VCHH : BinaryVRRbSPair<"vchh", 0xE7FB, z_vicmph, z_vicmphs,
1009 defm VCHF : BinaryVRRbSPair<"vchf", 0xE7FB, z_vicmph, z_vicmphs,
1011 defm VCHG : BinaryVRRbSPair<"vchg", 0xE7FB, z_vicmph, z_vicmphs,
1014 // Compare high logical.
1015 def VCHL : BinaryVRRbSPairGeneric<"vchl", 0xE7F9>;
1016 defm VCHLB : BinaryVRRbSPair<"vchlb", 0xE7F9, z_vicmphl, z_vicmphls,
1018 defm VCHLH : BinaryVRRbSPair<"vchlh", 0xE7F9, z_vicmphl, z_vicmphls,
1020 defm VCHLF : BinaryVRRbSPair<"vchlf", 0xE7F9, z_vicmphl, z_vicmphls,
1022 defm VCHLG : BinaryVRRbSPair<"vchlg", 0xE7F9, z_vicmphl, z_vicmphls,
1027 def VTM : CompareVRRa<"vtm", 0xE7D8, z_vtm, v128b, 0>;
1030 //===----------------------------------------------------------------------===//
1031 // Floating-point arithmetic
1032 //===----------------------------------------------------------------------===//
1034 // See comments in SystemZInstrFP.td for the suppression flags and
1036 multiclass VectorRounding<Instruction insn, TypedReg tr> {
1037 def : FPConversion<insn, any_frint, tr, tr, 0, 0>;
1038 def : FPConversion<insn, any_fnearbyint, tr, tr, 4, 0>;
1039 def : FPConversion<insn, any_ffloor, tr, tr, 4, 7>;
1040 def : FPConversion<insn, any_fceil, tr, tr, 4, 6>;
1041 def : FPConversion<insn, any_ftrunc, tr, tr, 4, 5>;
1042 def : FPConversion<insn, any_fround, tr, tr, 4, 1>;
1045 let Predicates = [FeatureVector] in {
1047 let Uses = [FPC], mayRaiseFPException = 1, isCommutable = 1 in {
1048 def VFA : BinaryVRRcFloatGeneric<"vfa", 0xE7E3>;
1049 def VFADB : BinaryVRRc<"vfadb", 0xE7E3, any_fadd, v128db, v128db, 3, 0>;
1050 def WFADB : BinaryVRRc<"wfadb", 0xE7E3, any_fadd, v64db, v64db, 3, 8, 0,
1052 let Predicates = [FeatureVectorEnhancements1] in {
1053 def VFASB : BinaryVRRc<"vfasb", 0xE7E3, any_fadd, v128sb, v128sb, 2, 0>;
1054 def WFASB : BinaryVRRc<"wfasb", 0xE7E3, any_fadd, v32sb, v32sb, 2, 8, 0,
1056 def WFAXB : BinaryVRRc<"wfaxb", 0xE7E3, any_fadd, v128xb, v128xb, 4, 8>;
1060 // Convert from fixed.
1061 let Uses = [FPC], mayRaiseFPException = 1 in {
1062 def VCDG : TernaryVRRaFloatGeneric<"vcdg", 0xE7C3>;
1063 def VCDGB : TernaryVRRa<"vcdgb", 0xE7C3, null_frag, v128db, v128g, 3, 0>;
1064 def WCDGB : TernaryVRRa<"wcdgb", 0xE7C3, null_frag, v64db, v64g, 3, 8>;
1066 def : FPConversion<VCDGB, any_sint_to_fp, v128db, v128g, 0, 0>;
1067 let Predicates = [FeatureVectorEnhancements2] in {
1068 let Uses = [FPC], mayRaiseFPException = 1 in {
1069 let isAsmParserOnly = 1 in
1070 def VCFPS : TernaryVRRaFloatGeneric<"vcfps", 0xE7C3>;
1071 def VCEFB : TernaryVRRa<"vcefb", 0xE7C3, null_frag, v128sb, v128g, 2, 0>;
1072 def WCEFB : TernaryVRRa<"wcefb", 0xE7C3, null_frag, v32sb, v32f, 2, 8>;
1074 def : FPConversion<VCEFB, any_sint_to_fp, v128sb, v128f, 0, 0>;
1077 // Convert from logical.
1078 let Uses = [FPC], mayRaiseFPException = 1 in {
1079 def VCDLG : TernaryVRRaFloatGeneric<"vcdlg", 0xE7C1>;
1080 def VCDLGB : TernaryVRRa<"vcdlgb", 0xE7C1, null_frag, v128db, v128g, 3, 0>;
1081 def WCDLGB : TernaryVRRa<"wcdlgb", 0xE7C1, null_frag, v64db, v64g, 3, 8>;
1083 def : FPConversion<VCDLGB, any_uint_to_fp, v128db, v128g, 0, 0>;
1084 let Predicates = [FeatureVectorEnhancements2] in {
1085 let Uses = [FPC], mayRaiseFPException = 1 in {
1086 let isAsmParserOnly = 1 in
1087 def VCFPL : TernaryVRRaFloatGeneric<"vcfpl", 0xE7C1>;
1088 def VCELFB : TernaryVRRa<"vcelfb", 0xE7C1, null_frag, v128sb, v128g, 2, 0>;
1089 def WCELFB : TernaryVRRa<"wcelfb", 0xE7C1, null_frag, v32sb, v32f, 2, 8>;
1091 def : FPConversion<VCELFB, any_uint_to_fp, v128sb, v128f, 0, 0>;
1094 // Convert to fixed.
1095 let Uses = [FPC], mayRaiseFPException = 1 in {
1096 def VCGD : TernaryVRRaFloatGeneric<"vcgd", 0xE7C2>;
1097 def VCGDB : TernaryVRRa<"vcgdb", 0xE7C2, null_frag, v128g, v128db, 3, 0>;
1098 def WCGDB : TernaryVRRa<"wcgdb", 0xE7C2, null_frag, v64g, v64db, 3, 8>;
1100 // Rounding mode should agree with SystemZInstrFP.td.
1101 def : FPConversion<VCGDB, any_fp_to_sint, v128g, v128db, 0, 5>;
1102 let Predicates = [FeatureVectorEnhancements2] in {
1103 let Uses = [FPC], mayRaiseFPException = 1 in {
1104 let isAsmParserOnly = 1 in
1105 def VCSFP : TernaryVRRaFloatGeneric<"vcsfp", 0xE7C2>;
1106 def VCFEB : TernaryVRRa<"vcfeb", 0xE7C2, null_frag, v128sb, v128g, 2, 0>;
1107 def WCFEB : TernaryVRRa<"wcfeb", 0xE7C2, null_frag, v32sb, v32f, 2, 8>;
1109 // Rounding mode should agree with SystemZInstrFP.td.
1110 def : FPConversion<VCFEB, any_fp_to_sint, v128f, v128sb, 0, 5>;
1113 // Convert to logical.
1114 let Uses = [FPC], mayRaiseFPException = 1 in {
1115 def VCLGD : TernaryVRRaFloatGeneric<"vclgd", 0xE7C0>;
1116 def VCLGDB : TernaryVRRa<"vclgdb", 0xE7C0, null_frag, v128g, v128db, 3, 0>;
1117 def WCLGDB : TernaryVRRa<"wclgdb", 0xE7C0, null_frag, v64g, v64db, 3, 8>;
1119 // Rounding mode should agree with SystemZInstrFP.td.
1120 def : FPConversion<VCLGDB, any_fp_to_uint, v128g, v128db, 0, 5>;
1121 let Predicates = [FeatureVectorEnhancements2] in {
1122 let Uses = [FPC], mayRaiseFPException = 1 in {
1123 let isAsmParserOnly = 1 in
1124 def VCLFP : TernaryVRRaFloatGeneric<"vclfp", 0xE7C0>;
1125 def VCLFEB : TernaryVRRa<"vclfeb", 0xE7C0, null_frag, v128sb, v128g, 2, 0>;
1126 def WCLFEB : TernaryVRRa<"wclfeb", 0xE7C0, null_frag, v32sb, v32f, 2, 8>;
1128 // Rounding mode should agree with SystemZInstrFP.td.
1129 def : FPConversion<VCLFEB, any_fp_to_uint, v128f, v128sb, 0, 5>;
1133 let Uses = [FPC], mayRaiseFPException = 1 in {
1134 def VFD : BinaryVRRcFloatGeneric<"vfd", 0xE7E5>;
1135 def VFDDB : BinaryVRRc<"vfddb", 0xE7E5, any_fdiv, v128db, v128db, 3, 0>;
1136 def WFDDB : BinaryVRRc<"wfddb", 0xE7E5, any_fdiv, v64db, v64db, 3, 8, 0,
1138 let Predicates = [FeatureVectorEnhancements1] in {
1139 def VFDSB : BinaryVRRc<"vfdsb", 0xE7E5, any_fdiv, v128sb, v128sb, 2, 0>;
1140 def WFDSB : BinaryVRRc<"wfdsb", 0xE7E5, any_fdiv, v32sb, v32sb, 2, 8, 0,
1142 def WFDXB : BinaryVRRc<"wfdxb", 0xE7E5, any_fdiv, v128xb, v128xb, 4, 8>;
1147 let Uses = [FPC], mayRaiseFPException = 1 in {
1148 def VFI : TernaryVRRaFloatGeneric<"vfi", 0xE7C7>;
1149 def VFIDB : TernaryVRRa<"vfidb", 0xE7C7, int_s390_vfidb, v128db, v128db, 3, 0>;
1150 def WFIDB : TernaryVRRa<"wfidb", 0xE7C7, null_frag, v64db, v64db, 3, 8>;
1152 defm : VectorRounding<VFIDB, v128db>;
1153 defm : VectorRounding<WFIDB, v64db>;
1154 let Predicates = [FeatureVectorEnhancements1] in {
1155 let Uses = [FPC], mayRaiseFPException = 1 in {
1156 def VFISB : TernaryVRRa<"vfisb", 0xE7C7, int_s390_vfisb, v128sb, v128sb, 2, 0>;
1157 def WFISB : TernaryVRRa<"wfisb", 0xE7C7, null_frag, v32sb, v32sb, 2, 8>;
1158 def WFIXB : TernaryVRRa<"wfixb", 0xE7C7, null_frag, v128xb, v128xb, 4, 8>;
1160 defm : VectorRounding<VFISB, v128sb>;
1161 defm : VectorRounding<WFISB, v32sb>;
1162 defm : VectorRounding<WFIXB, v128xb>;
1166 let Uses = [FPC], mayRaiseFPException = 1 in {
1167 def VLDE : UnaryVRRaFloatGeneric<"vlde", 0xE7C4>;
1168 def VLDEB : UnaryVRRa<"vldeb", 0xE7C4, z_any_vextend, v128db, v128sb, 2, 0>;
1169 def WLDEB : UnaryVRRa<"wldeb", 0xE7C4, any_fpextend, v64db, v32sb, 2, 8, 0,
1172 let Predicates = [FeatureVectorEnhancements1] in {
1173 let Uses = [FPC], mayRaiseFPException = 1 in {
1174 let isAsmParserOnly = 1 in {
1175 def VFLL : UnaryVRRaFloatGeneric<"vfll", 0xE7C4>;
1176 def VFLLS : UnaryVRRa<"vflls", 0xE7C4, null_frag, v128db, v128sb, 2, 0>;
1177 def WFLLS : UnaryVRRa<"wflls", 0xE7C4, null_frag, v64db, v32sb, 2, 8>;
1179 def WFLLD : UnaryVRRa<"wflld", 0xE7C4, any_fpextend, v128xb, v64db, 3, 8>;
1181 def : Pat<(f128 (any_fpextend (f32 VR32:$src))),
1182 (WFLLD (WLDEB VR32:$src))>;
1186 let Uses = [FPC], mayRaiseFPException = 1 in {
1187 def VLED : TernaryVRRaFloatGeneric<"vled", 0xE7C5>;
1188 def VLEDB : TernaryVRRa<"vledb", 0xE7C5, null_frag, v128sb, v128db, 3, 0>;
1189 def WLEDB : TernaryVRRa<"wledb", 0xE7C5, null_frag, v32sb, v64db, 3, 8>;
1191 def : Pat<(v4f32 (z_any_vround (v2f64 VR128:$src))), (VLEDB VR128:$src, 0, 0)>;
1192 def : FPConversion<WLEDB, any_fpround, v32sb, v64db, 0, 0>;
1193 let Predicates = [FeatureVectorEnhancements1] in {
1194 let Uses = [FPC], mayRaiseFPException = 1 in {
1195 let isAsmParserOnly = 1 in {
1196 def VFLR : TernaryVRRaFloatGeneric<"vflr", 0xE7C5>;
1197 def VFLRD : TernaryVRRa<"vflrd", 0xE7C5, null_frag, v128sb, v128db, 3, 0>;
1198 def WFLRD : TernaryVRRa<"wflrd", 0xE7C5, null_frag, v32sb, v64db, 3, 8>;
1200 def WFLRX : TernaryVRRa<"wflrx", 0xE7C5, null_frag, v64db, v128xb, 4, 8>;
1202 def : FPConversion<WFLRX, any_fpround, v64db, v128xb, 0, 0>;
1203 def : Pat<(f32 (any_fpround (f128 VR128:$src))),
1204 (WLEDB (WFLRX VR128:$src, 0, 3), 0, 0)>;
1208 multiclass VectorMax<Instruction insn, TypedReg tr> {
1209 def : FPMinMax<insn, any_fmaxnum, tr, 4>;
1210 def : FPMinMax<insn, any_fmaximum, tr, 1>;
1212 let Predicates = [FeatureVectorEnhancements1] in {
1213 let Uses = [FPC], mayRaiseFPException = 1, isCommutable = 1 in {
1214 def VFMAX : TernaryVRRcFloatGeneric<"vfmax", 0xE7EF>;
1215 def VFMAXDB : TernaryVRRcFloat<"vfmaxdb", 0xE7EF, int_s390_vfmaxdb,
1216 v128db, v128db, 3, 0>;
1217 def WFMAXDB : TernaryVRRcFloat<"wfmaxdb", 0xE7EF, null_frag,
1218 v64db, v64db, 3, 8>;
1219 def VFMAXSB : TernaryVRRcFloat<"vfmaxsb", 0xE7EF, int_s390_vfmaxsb,
1220 v128sb, v128sb, 2, 0>;
1221 def WFMAXSB : TernaryVRRcFloat<"wfmaxsb", 0xE7EF, null_frag,
1222 v32sb, v32sb, 2, 8>;
1223 def WFMAXXB : TernaryVRRcFloat<"wfmaxxb", 0xE7EF, null_frag,
1224 v128xb, v128xb, 4, 8>;
1226 defm : VectorMax<VFMAXDB, v128db>;
1227 defm : VectorMax<WFMAXDB, v64db>;
1228 defm : VectorMax<VFMAXSB, v128sb>;
1229 defm : VectorMax<WFMAXSB, v32sb>;
1230 defm : VectorMax<WFMAXXB, v128xb>;
1234 multiclass VectorMin<Instruction insn, TypedReg tr> {
1235 def : FPMinMax<insn, any_fminnum, tr, 4>;
1236 def : FPMinMax<insn, any_fminimum, tr, 1>;
1238 let Predicates = [FeatureVectorEnhancements1] in {
1239 let Uses = [FPC], mayRaiseFPException = 1, isCommutable = 1 in {
1240 def VFMIN : TernaryVRRcFloatGeneric<"vfmin", 0xE7EE>;
1241 def VFMINDB : TernaryVRRcFloat<"vfmindb", 0xE7EE, int_s390_vfmindb,
1242 v128db, v128db, 3, 0>;
1243 def WFMINDB : TernaryVRRcFloat<"wfmindb", 0xE7EE, null_frag,
1244 v64db, v64db, 3, 8>;
1245 def VFMINSB : TernaryVRRcFloat<"vfminsb", 0xE7EE, int_s390_vfminsb,
1246 v128sb, v128sb, 2, 0>;
1247 def WFMINSB : TernaryVRRcFloat<"wfminsb", 0xE7EE, null_frag,
1248 v32sb, v32sb, 2, 8>;
1249 def WFMINXB : TernaryVRRcFloat<"wfminxb", 0xE7EE, null_frag,
1250 v128xb, v128xb, 4, 8>;
1252 defm : VectorMin<VFMINDB, v128db>;
1253 defm : VectorMin<WFMINDB, v64db>;
1254 defm : VectorMin<VFMINSB, v128sb>;
1255 defm : VectorMin<WFMINSB, v32sb>;
1256 defm : VectorMin<WFMINXB, v128xb>;
1260 let Uses = [FPC], mayRaiseFPException = 1, isCommutable = 1 in {
1261 def VFM : BinaryVRRcFloatGeneric<"vfm", 0xE7E7>;
1262 def VFMDB : BinaryVRRc<"vfmdb", 0xE7E7, any_fmul, v128db, v128db, 3, 0>;
1263 def WFMDB : BinaryVRRc<"wfmdb", 0xE7E7, any_fmul, v64db, v64db, 3, 8, 0,
1265 let Predicates = [FeatureVectorEnhancements1] in {
1266 def VFMSB : BinaryVRRc<"vfmsb", 0xE7E7, any_fmul, v128sb, v128sb, 2, 0>;
1267 def WFMSB : BinaryVRRc<"wfmsb", 0xE7E7, any_fmul, v32sb, v32sb, 2, 8, 0,
1269 def WFMXB : BinaryVRRc<"wfmxb", 0xE7E7, any_fmul, v128xb, v128xb, 4, 8>;
1273 // Multiply and add.
1274 let Uses = [FPC], mayRaiseFPException = 1, isCommutable = 1 in {
1275 def VFMA : TernaryVRReFloatGeneric<"vfma", 0xE78F>;
1276 def VFMADB : TernaryVRRe<"vfmadb", 0xE78F, any_fma, v128db, v128db, 0, 3>;
1277 def WFMADB : TernaryVRRe<"wfmadb", 0xE78F, any_fma, v64db, v64db, 8, 3,
1279 let Predicates = [FeatureVectorEnhancements1] in {
1280 def VFMASB : TernaryVRRe<"vfmasb", 0xE78F, any_fma, v128sb, v128sb, 0, 2>;
1281 def WFMASB : TernaryVRRe<"wfmasb", 0xE78F, any_fma, v32sb, v32sb, 8, 2,
1283 def WFMAXB : TernaryVRRe<"wfmaxb", 0xE78F, any_fma, v128xb, v128xb, 8, 4>;
1287 // Multiply and subtract.
1288 let Uses = [FPC], mayRaiseFPException = 1, isCommutable = 1 in {
1289 def VFMS : TernaryVRReFloatGeneric<"vfms", 0xE78E>;
1290 def VFMSDB : TernaryVRRe<"vfmsdb", 0xE78E, any_fms, v128db, v128db, 0, 3>;
1291 def WFMSDB : TernaryVRRe<"wfmsdb", 0xE78E, any_fms, v64db, v64db, 8, 3,
1293 let Predicates = [FeatureVectorEnhancements1] in {
1294 def VFMSSB : TernaryVRRe<"vfmssb", 0xE78E, any_fms, v128sb, v128sb, 0, 2>;
1295 def WFMSSB : TernaryVRRe<"wfmssb", 0xE78E, any_fms, v32sb, v32sb, 8, 2,
1297 def WFMSXB : TernaryVRRe<"wfmsxb", 0xE78E, any_fms, v128xb, v128xb, 8, 4>;
1301 // Negative multiply and add.
1302 let Uses = [FPC], mayRaiseFPException = 1, isCommutable = 1,
1303 Predicates = [FeatureVectorEnhancements1] in {
1304 def VFNMA : TernaryVRReFloatGeneric<"vfnma", 0xE79F>;
1305 def VFNMADB : TernaryVRRe<"vfnmadb", 0xE79F, any_fnma, v128db, v128db, 0, 3>;
1306 def WFNMADB : TernaryVRRe<"wfnmadb", 0xE79F, any_fnma, v64db, v64db, 8, 3>;
1307 def VFNMASB : TernaryVRRe<"vfnmasb", 0xE79F, any_fnma, v128sb, v128sb, 0, 2>;
1308 def WFNMASB : TernaryVRRe<"wfnmasb", 0xE79F, any_fnma, v32sb, v32sb, 8, 2>;
1309 def WFNMAXB : TernaryVRRe<"wfnmaxb", 0xE79F, any_fnma, v128xb, v128xb, 8, 4>;
1312 // Negative multiply and subtract.
1313 let Uses = [FPC], mayRaiseFPException = 1, isCommutable = 1,
1314 Predicates = [FeatureVectorEnhancements1] in {
1315 def VFNMS : TernaryVRReFloatGeneric<"vfnms", 0xE79E>;
1316 def VFNMSDB : TernaryVRRe<"vfnmsdb", 0xE79E, any_fnms, v128db, v128db, 0, 3>;
1317 def WFNMSDB : TernaryVRRe<"wfnmsdb", 0xE79E, any_fnms, v64db, v64db, 8, 3>;
1318 def VFNMSSB : TernaryVRRe<"vfnmssb", 0xE79E, any_fnms, v128sb, v128sb, 0, 2>;
1319 def WFNMSSB : TernaryVRRe<"wfnmssb", 0xE79E, any_fnms, v32sb, v32sb, 8, 2>;
1320 def WFNMSXB : TernaryVRRe<"wfnmsxb", 0xE79E, any_fnms, v128xb, v128xb, 8, 4>;
1323 // Perform sign operation.
1324 def VFPSO : BinaryVRRaFloatGeneric<"vfpso", 0xE7CC>;
1325 def VFPSODB : BinaryVRRa<"vfpsodb", 0xE7CC, null_frag, v128db, v128db, 3, 0>;
1326 def WFPSODB : BinaryVRRa<"wfpsodb", 0xE7CC, null_frag, v64db, v64db, 3, 8>;
1327 let Predicates = [FeatureVectorEnhancements1] in {
1328 def VFPSOSB : BinaryVRRa<"vfpsosb", 0xE7CC, null_frag, v128sb, v128sb, 2, 0>;
1329 def WFPSOSB : BinaryVRRa<"wfpsosb", 0xE7CC, null_frag, v32sb, v32sb, 2, 8>;
1330 def WFPSOXB : BinaryVRRa<"wfpsoxb", 0xE7CC, null_frag, v128xb, v128xb, 4, 8>;
1334 def VFLCDB : UnaryVRRa<"vflcdb", 0xE7CC, fneg, v128db, v128db, 3, 0, 0>;
1335 def WFLCDB : UnaryVRRa<"wflcdb", 0xE7CC, fneg, v64db, v64db, 3, 8, 0>;
1336 let Predicates = [FeatureVectorEnhancements1] in {
1337 def VFLCSB : UnaryVRRa<"vflcsb", 0xE7CC, fneg, v128sb, v128sb, 2, 0, 0>;
1338 def WFLCSB : UnaryVRRa<"wflcsb", 0xE7CC, fneg, v32sb, v32sb, 2, 8, 0>;
1339 def WFLCXB : UnaryVRRa<"wflcxb", 0xE7CC, fneg, v128xb, v128xb, 4, 8, 0>;
1343 def VFLNDB : UnaryVRRa<"vflndb", 0xE7CC, fnabs, v128db, v128db, 3, 0, 1>;
1344 def WFLNDB : UnaryVRRa<"wflndb", 0xE7CC, fnabs, v64db, v64db, 3, 8, 1>;
1345 let Predicates = [FeatureVectorEnhancements1] in {
1346 def VFLNSB : UnaryVRRa<"vflnsb", 0xE7CC, fnabs, v128sb, v128sb, 2, 0, 1>;
1347 def WFLNSB : UnaryVRRa<"wflnsb", 0xE7CC, fnabs, v32sb, v32sb, 2, 8, 1>;
1348 def WFLNXB : UnaryVRRa<"wflnxb", 0xE7CC, fnabs, v128xb, v128xb, 4, 8, 1>;
1352 def VFLPDB : UnaryVRRa<"vflpdb", 0xE7CC, fabs, v128db, v128db, 3, 0, 2>;
1353 def WFLPDB : UnaryVRRa<"wflpdb", 0xE7CC, fabs, v64db, v64db, 3, 8, 2>;
1354 let Predicates = [FeatureVectorEnhancements1] in {
1355 def VFLPSB : UnaryVRRa<"vflpsb", 0xE7CC, fabs, v128sb, v128sb, 2, 0, 2>;
1356 def WFLPSB : UnaryVRRa<"wflpsb", 0xE7CC, fabs, v32sb, v32sb, 2, 8, 2>;
1357 def WFLPXB : UnaryVRRa<"wflpxb", 0xE7CC, fabs, v128xb, v128xb, 4, 8, 2>;
1361 let Uses = [FPC], mayRaiseFPException = 1 in {
1362 def VFSQ : UnaryVRRaFloatGeneric<"vfsq", 0xE7CE>;
1363 def VFSQDB : UnaryVRRa<"vfsqdb", 0xE7CE, any_fsqrt, v128db, v128db, 3, 0>;
1364 def WFSQDB : UnaryVRRa<"wfsqdb", 0xE7CE, any_fsqrt, v64db, v64db, 3, 8, 0,
1366 let Predicates = [FeatureVectorEnhancements1] in {
1367 def VFSQSB : UnaryVRRa<"vfsqsb", 0xE7CE, any_fsqrt, v128sb, v128sb, 2, 0>;
1368 def WFSQSB : UnaryVRRa<"wfsqsb", 0xE7CE, any_fsqrt, v32sb, v32sb, 2, 8, 0,
1370 def WFSQXB : UnaryVRRa<"wfsqxb", 0xE7CE, any_fsqrt, v128xb, v128xb, 4, 8>;
1375 let Uses = [FPC], mayRaiseFPException = 1 in {
1376 def VFS : BinaryVRRcFloatGeneric<"vfs", 0xE7E2>;
1377 def VFSDB : BinaryVRRc<"vfsdb", 0xE7E2, any_fsub, v128db, v128db, 3, 0>;
1378 def WFSDB : BinaryVRRc<"wfsdb", 0xE7E2, any_fsub, v64db, v64db, 3, 8, 0,
1380 let Predicates = [FeatureVectorEnhancements1] in {
1381 def VFSSB : BinaryVRRc<"vfssb", 0xE7E2, any_fsub, v128sb, v128sb, 2, 0>;
1382 def WFSSB : BinaryVRRc<"wfssb", 0xE7E2, any_fsub, v32sb, v32sb, 2, 8, 0,
1384 def WFSXB : BinaryVRRc<"wfsxb", 0xE7E2, any_fsub, v128xb, v128xb, 4, 8>;
1388 // Test data class immediate.
1389 let Defs = [CC] in {
1390 def VFTCI : BinaryVRIeFloatGeneric<"vftci", 0xE74A>;
1391 def VFTCIDB : BinaryVRIe<"vftcidb", 0xE74A, z_vftci, v128g, v128db, 3, 0>;
1392 def WFTCIDB : BinaryVRIe<"wftcidb", 0xE74A, null_frag, v64g, v64db, 3, 8>;
1393 let Predicates = [FeatureVectorEnhancements1] in {
1394 def VFTCISB : BinaryVRIe<"vftcisb", 0xE74A, z_vftci, v128f, v128sb, 2, 0>;
1395 def WFTCISB : BinaryVRIe<"wftcisb", 0xE74A, null_frag, v32f, v32sb, 2, 8>;
1396 def WFTCIXB : BinaryVRIe<"wftcixb", 0xE74A, null_frag, v128q, v128xb, 4, 8>;
1401 //===----------------------------------------------------------------------===//
1402 // Floating-point comparison
1403 //===----------------------------------------------------------------------===//
1405 let Predicates = [FeatureVector] in {
1407 let Uses = [FPC], mayRaiseFPException = 1, Defs = [CC] in {
1408 def WFC : CompareVRRaFloatGeneric<"wfc", 0xE7CB>;
1409 def WFCDB : CompareVRRa<"wfcdb", 0xE7CB, z_any_fcmp, v64db, 3, "cdbr">;
1410 let Predicates = [FeatureVectorEnhancements1] in {
1411 def WFCSB : CompareVRRa<"wfcsb", 0xE7CB, z_any_fcmp, v32sb, 2, "cebr">;
1412 def WFCXB : CompareVRRa<"wfcxb", 0xE7CB, z_any_fcmp, v128xb, 4>;
1416 // Compare and signal scalar.
1417 let Uses = [FPC], mayRaiseFPException = 1, Defs = [CC] in {
1418 def WFK : CompareVRRaFloatGeneric<"wfk", 0xE7CA>;
1419 def WFKDB : CompareVRRa<"wfkdb", 0xE7CA, z_strict_fcmps, v64db, 3, "kdbr">;
1420 let Predicates = [FeatureVectorEnhancements1] in {
1421 def WFKSB : CompareVRRa<"wfksb", 0xE7CA, z_strict_fcmps, v32sb, 2, "kebr">;
1422 def WFKXB : CompareVRRa<"wfkxb", 0xE7CA, z_strict_fcmps, v128xb, 4>;
1427 let Uses = [FPC], mayRaiseFPException = 1 in {
1428 def VFCE : BinaryVRRcSPairFloatGeneric<"vfce", 0xE7E8>;
1429 defm VFCEDB : BinaryVRRcSPair<"vfcedb", 0xE7E8, z_any_vfcmpe, z_vfcmpes,
1430 v128g, v128db, 3, 0>;
1431 defm WFCEDB : BinaryVRRcSPair<"wfcedb", 0xE7E8, null_frag, null_frag,
1433 let Predicates = [FeatureVectorEnhancements1] in {
1434 defm VFCESB : BinaryVRRcSPair<"vfcesb", 0xE7E8, z_any_vfcmpe, z_vfcmpes,
1435 v128f, v128sb, 2, 0>;
1436 defm WFCESB : BinaryVRRcSPair<"wfcesb", 0xE7E8, null_frag, null_frag,
1438 defm WFCEXB : BinaryVRRcSPair<"wfcexb", 0xE7E8, null_frag, null_frag,
1439 v128q, v128xb, 4, 8>;
1443 // Compare and signal equal.
1444 let Uses = [FPC], mayRaiseFPException = 1,
1445 Predicates = [FeatureVectorEnhancements1] in {
1446 defm VFKEDB : BinaryVRRcSPair<"vfkedb", 0xE7E8, z_strict_vfcmpes, null_frag,
1447 v128g, v128db, 3, 4>;
1448 defm WFKEDB : BinaryVRRcSPair<"wfkedb", 0xE7E8, null_frag, null_frag,
1449 v64g, v64db, 3, 12>;
1450 defm VFKESB : BinaryVRRcSPair<"vfkesb", 0xE7E8, z_strict_vfcmpes, null_frag,
1451 v128f, v128sb, 2, 4>;
1452 defm WFKESB : BinaryVRRcSPair<"wfkesb", 0xE7E8, null_frag, null_frag,
1453 v32f, v32sb, 2, 12>;
1454 defm WFKEXB : BinaryVRRcSPair<"wfkexb", 0xE7E8, null_frag, null_frag,
1455 v128q, v128xb, 4, 12>;
1459 let Uses = [FPC], mayRaiseFPException = 1 in {
1460 def VFCH : BinaryVRRcSPairFloatGeneric<"vfch", 0xE7EB>;
1461 defm VFCHDB : BinaryVRRcSPair<"vfchdb", 0xE7EB, z_any_vfcmph, z_vfcmphs,
1462 v128g, v128db, 3, 0>;
1463 defm WFCHDB : BinaryVRRcSPair<"wfchdb", 0xE7EB, null_frag, null_frag,
1465 let Predicates = [FeatureVectorEnhancements1] in {
1466 defm VFCHSB : BinaryVRRcSPair<"vfchsb", 0xE7EB, z_any_vfcmph, z_vfcmphs,
1467 v128f, v128sb, 2, 0>;
1468 defm WFCHSB : BinaryVRRcSPair<"wfchsb", 0xE7EB, null_frag, null_frag,
1470 defm WFCHXB : BinaryVRRcSPair<"wfchxb", 0xE7EB, null_frag, null_frag,
1471 v128q, v128xb, 4, 8>;
1475 // Compare and signal high.
1476 let Uses = [FPC], mayRaiseFPException = 1,
1477 Predicates = [FeatureVectorEnhancements1] in {
1478 defm VFKHDB : BinaryVRRcSPair<"vfkhdb", 0xE7EB, z_strict_vfcmphs, null_frag,
1479 v128g, v128db, 3, 4>;
1480 defm WFKHDB : BinaryVRRcSPair<"wfkhdb", 0xE7EB, null_frag, null_frag,
1481 v64g, v64db, 3, 12>;
1482 defm VFKHSB : BinaryVRRcSPair<"vfkhsb", 0xE7EB, z_strict_vfcmphs, null_frag,
1483 v128f, v128sb, 2, 4>;
1484 defm WFKHSB : BinaryVRRcSPair<"wfkhsb", 0xE7EB, null_frag, null_frag,
1485 v32f, v32sb, 2, 12>;
1486 defm WFKHXB : BinaryVRRcSPair<"wfkhxb", 0xE7EB, null_frag, null_frag,
1487 v128q, v128xb, 4, 12>;
1490 // Compare high or equal.
1491 let Uses = [FPC], mayRaiseFPException = 1 in {
1492 def VFCHE : BinaryVRRcSPairFloatGeneric<"vfche", 0xE7EA>;
1493 defm VFCHEDB : BinaryVRRcSPair<"vfchedb", 0xE7EA, z_any_vfcmphe, z_vfcmphes,
1494 v128g, v128db, 3, 0>;
1495 defm WFCHEDB : BinaryVRRcSPair<"wfchedb", 0xE7EA, null_frag, null_frag,
1497 let Predicates = [FeatureVectorEnhancements1] in {
1498 defm VFCHESB : BinaryVRRcSPair<"vfchesb", 0xE7EA, z_any_vfcmphe, z_vfcmphes,
1499 v128f, v128sb, 2, 0>;
1500 defm WFCHESB : BinaryVRRcSPair<"wfchesb", 0xE7EA, null_frag, null_frag,
1502 defm WFCHEXB : BinaryVRRcSPair<"wfchexb", 0xE7EA, null_frag, null_frag,
1503 v128q, v128xb, 4, 8>;
1507 // Compare and signal high or equal.
1508 let Uses = [FPC], mayRaiseFPException = 1,
1509 Predicates = [FeatureVectorEnhancements1] in {
1510 defm VFKHEDB : BinaryVRRcSPair<"vfkhedb", 0xE7EA, z_strict_vfcmphes, null_frag,
1511 v128g, v128db, 3, 4>;
1512 defm WFKHEDB : BinaryVRRcSPair<"wfkhedb", 0xE7EA, null_frag, null_frag,
1513 v64g, v64db, 3, 12>;
1514 defm VFKHESB : BinaryVRRcSPair<"vfkhesb", 0xE7EA, z_strict_vfcmphes, null_frag,
1515 v128f, v128sb, 2, 4>;
1516 defm WFKHESB : BinaryVRRcSPair<"wfkhesb", 0xE7EA, null_frag, null_frag,
1517 v32f, v32sb, 2, 12>;
1518 defm WFKHEXB : BinaryVRRcSPair<"wfkhexb", 0xE7EA, null_frag, null_frag,
1519 v128q, v128xb, 4, 12>;
1523 //===----------------------------------------------------------------------===//
1525 //===----------------------------------------------------------------------===//
1527 def : Pat<(v16i8 (bitconvert (v8i16 VR128:$src))), (v16i8 VR128:$src)>;
1528 def : Pat<(v16i8 (bitconvert (v4i32 VR128:$src))), (v16i8 VR128:$src)>;
1529 def : Pat<(v16i8 (bitconvert (v2i64 VR128:$src))), (v16i8 VR128:$src)>;
1530 def : Pat<(v16i8 (bitconvert (v4f32 VR128:$src))), (v16i8 VR128:$src)>;
1531 def : Pat<(v16i8 (bitconvert (v2f64 VR128:$src))), (v16i8 VR128:$src)>;
1532 def : Pat<(v16i8 (bitconvert (f128 VR128:$src))), (v16i8 VR128:$src)>;
1534 def : Pat<(v8i16 (bitconvert (v16i8 VR128:$src))), (v8i16 VR128:$src)>;
1535 def : Pat<(v8i16 (bitconvert (v4i32 VR128:$src))), (v8i16 VR128:$src)>;
1536 def : Pat<(v8i16 (bitconvert (v2i64 VR128:$src))), (v8i16 VR128:$src)>;
1537 def : Pat<(v8i16 (bitconvert (v4f32 VR128:$src))), (v8i16 VR128:$src)>;
1538 def : Pat<(v8i16 (bitconvert (v2f64 VR128:$src))), (v8i16 VR128:$src)>;
1539 def : Pat<(v8i16 (bitconvert (f128 VR128:$src))), (v8i16 VR128:$src)>;
1541 def : Pat<(v4i32 (bitconvert (v16i8 VR128:$src))), (v4i32 VR128:$src)>;
1542 def : Pat<(v4i32 (bitconvert (v8i16 VR128:$src))), (v4i32 VR128:$src)>;
1543 def : Pat<(v4i32 (bitconvert (v2i64 VR128:$src))), (v4i32 VR128:$src)>;
1544 def : Pat<(v4i32 (bitconvert (v4f32 VR128:$src))), (v4i32 VR128:$src)>;
1545 def : Pat<(v4i32 (bitconvert (v2f64 VR128:$src))), (v4i32 VR128:$src)>;
1546 def : Pat<(v4i32 (bitconvert (f128 VR128:$src))), (v4i32 VR128:$src)>;
1548 def : Pat<(v2i64 (bitconvert (v16i8 VR128:$src))), (v2i64 VR128:$src)>;
1549 def : Pat<(v2i64 (bitconvert (v8i16 VR128:$src))), (v2i64 VR128:$src)>;
1550 def : Pat<(v2i64 (bitconvert (v4i32 VR128:$src))), (v2i64 VR128:$src)>;
1551 def : Pat<(v2i64 (bitconvert (v4f32 VR128:$src))), (v2i64 VR128:$src)>;
1552 def : Pat<(v2i64 (bitconvert (v2f64 VR128:$src))), (v2i64 VR128:$src)>;
1553 def : Pat<(v2i64 (bitconvert (f128 VR128:$src))), (v2i64 VR128:$src)>;
1555 def : Pat<(v4f32 (bitconvert (v16i8 VR128:$src))), (v4f32 VR128:$src)>;
1556 def : Pat<(v4f32 (bitconvert (v8i16 VR128:$src))), (v4f32 VR128:$src)>;
1557 def : Pat<(v4f32 (bitconvert (v4i32 VR128:$src))), (v4f32 VR128:$src)>;
1558 def : Pat<(v4f32 (bitconvert (v2i64 VR128:$src))), (v4f32 VR128:$src)>;
1559 def : Pat<(v4f32 (bitconvert (v2f64 VR128:$src))), (v4f32 VR128:$src)>;
1560 def : Pat<(v4f32 (bitconvert (f128 VR128:$src))), (v4f32 VR128:$src)>;
1562 def : Pat<(v2f64 (bitconvert (v16i8 VR128:$src))), (v2f64 VR128:$src)>;
1563 def : Pat<(v2f64 (bitconvert (v8i16 VR128:$src))), (v2f64 VR128:$src)>;
1564 def : Pat<(v2f64 (bitconvert (v4i32 VR128:$src))), (v2f64 VR128:$src)>;
1565 def : Pat<(v2f64 (bitconvert (v2i64 VR128:$src))), (v2f64 VR128:$src)>;
1566 def : Pat<(v2f64 (bitconvert (v4f32 VR128:$src))), (v2f64 VR128:$src)>;
1567 def : Pat<(v2f64 (bitconvert (f128 VR128:$src))), (v2f64 VR128:$src)>;
1569 def : Pat<(f128 (bitconvert (v16i8 VR128:$src))), (f128 VR128:$src)>;
1570 def : Pat<(f128 (bitconvert (v8i16 VR128:$src))), (f128 VR128:$src)>;
1571 def : Pat<(f128 (bitconvert (v4i32 VR128:$src))), (f128 VR128:$src)>;
1572 def : Pat<(f128 (bitconvert (v2i64 VR128:$src))), (f128 VR128:$src)>;
1573 def : Pat<(f128 (bitconvert (v4f32 VR128:$src))), (f128 VR128:$src)>;
1574 def : Pat<(f128 (bitconvert (v2f64 VR128:$src))), (f128 VR128:$src)>;
1576 //===----------------------------------------------------------------------===//
1577 // Replicating scalars
1578 //===----------------------------------------------------------------------===//
1580 // Define patterns for replicating a scalar GR32 into a vector of type TYPE.
1581 // INDEX is 8 minus the element size in bytes.
1582 class VectorReplicateScalar<ValueType type, Instruction insn, bits<16> index>
1583 : Pat<(type (z_replicate GR32:$scalar)),
1584 (insn (VLVGP32 GR32:$scalar, GR32:$scalar), index)>;
1586 def : VectorReplicateScalar<v16i8, VREPB, 7>;
1587 def : VectorReplicateScalar<v8i16, VREPH, 3>;
1588 def : VectorReplicateScalar<v4i32, VREPF, 1>;
1590 // i64 replications are just a single instruction.
1591 def : Pat<(v2i64 (z_replicate GR64:$scalar)),
1592 (VLVGP GR64:$scalar, GR64:$scalar)>;
1594 //===----------------------------------------------------------------------===//
1595 // Floating-point insertion and extraction
1596 //===----------------------------------------------------------------------===//
1598 // Moving 32-bit values between GPRs and FPRs can be done using VLVGF
1600 let Predicates = [FeatureVector] in {
1601 def LEFR : UnaryAliasVRS<VR32, GR32>;
1602 def LFER : UnaryAliasVRS<GR64, VR32>;
1603 def : Pat<(f32 (bitconvert (i32 GR32:$src))), (LEFR GR32:$src)>;
1604 def : Pat<(i32 (bitconvert (f32 VR32:$src))),
1605 (EXTRACT_SUBREG (LFER VR32:$src), subreg_l32)>;
1608 // Floating-point values are stored in element 0 of the corresponding
1609 // vector register. Scalar to vector conversion is just a subreg and
1610 // scalar replication can just replicate element 0 of the vector register.
1611 multiclass ScalarToVectorFP<Instruction vrep, ValueType vt, RegisterOperand cls,
1612 SubRegIndex subreg> {
1613 def : Pat<(vt (scalar_to_vector cls:$scalar)),
1614 (INSERT_SUBREG (vt (IMPLICIT_DEF)), cls:$scalar, subreg)>;
1615 def : Pat<(vt (z_replicate cls:$scalar)),
1616 (vrep (INSERT_SUBREG (vt (IMPLICIT_DEF)), cls:$scalar,
1619 defm : ScalarToVectorFP<VREPF, v4f32, FP32, subreg_h32>;
1620 defm : ScalarToVectorFP<VREPG, v2f64, FP64, subreg_h64>;
1622 // Match v2f64 insertions. The AddedComplexity counters the 3 added by
1623 // TableGen for the base register operand in VLVG-based integer insertions
1624 // and ensures that this version is strictly better.
1625 let AddedComplexity = 4 in {
1626 def : Pat<(z_vector_insert (v2f64 VR128:$vec), FP64:$elt, 0),
1627 (VPDI (INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FP64:$elt,
1628 subreg_h64), VR128:$vec, 1)>;
1629 def : Pat<(z_vector_insert (v2f64 VR128:$vec), FP64:$elt, 1),
1630 (VPDI VR128:$vec, (INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FP64:$elt,
1634 // We extract floating-point element X by replicating (for elements other
1635 // than 0) and then taking a high subreg. The AddedComplexity counters the
1636 // 3 added by TableGen for the base register operand in VLGV-based integer
1637 // extractions and ensures that this version is strictly better.
1638 let AddedComplexity = 4 in {
1639 def : Pat<(f32 (z_vector_extract (v4f32 VR128:$vec), 0)),
1640 (EXTRACT_SUBREG VR128:$vec, subreg_h32)>;
1641 def : Pat<(f32 (z_vector_extract (v4f32 VR128:$vec), imm32zx2:$index)),
1642 (EXTRACT_SUBREG (VREPF VR128:$vec, imm32zx2:$index), subreg_h32)>;
1644 def : Pat<(f64 (z_vector_extract (v2f64 VR128:$vec), 0)),
1645 (EXTRACT_SUBREG VR128:$vec, subreg_h64)>;
1646 def : Pat<(f64 (z_vector_extract (v2f64 VR128:$vec), imm32zx1:$index)),
1647 (EXTRACT_SUBREG (VREPG VR128:$vec, imm32zx1:$index), subreg_h64)>;
1650 //===----------------------------------------------------------------------===//
1651 // Support for 128-bit floating-point values in vector registers
1652 //===----------------------------------------------------------------------===//
1654 let Predicates = [FeatureVectorEnhancements1] in {
1655 def : Pat<(f128 (load bdxaddr12only:$addr)),
1656 (VL bdxaddr12only:$addr)>;
1657 def : Pat<(store (f128 VR128:$src), bdxaddr12only:$addr),
1658 (VST VR128:$src, bdxaddr12only:$addr)>;
1660 def : Pat<(f128 fpimm0), (VZERO)>;
1661 def : Pat<(f128 fpimmneg0), (WFLNXB (VZERO))>;
1664 //===----------------------------------------------------------------------===//
1665 // String instructions
1666 //===----------------------------------------------------------------------===//
1668 let Predicates = [FeatureVector] in {
1669 defm VFAE : TernaryOptVRRbSPairGeneric<"vfae", 0xE782>;
1670 defm VFAEB : TernaryOptVRRbSPair<"vfaeb", 0xE782, int_s390_vfaeb,
1671 z_vfae_cc, v128b, v128b, 0>;
1672 defm VFAEH : TernaryOptVRRbSPair<"vfaeh", 0xE782, int_s390_vfaeh,
1673 z_vfae_cc, v128h, v128h, 1>;
1674 defm VFAEF : TernaryOptVRRbSPair<"vfaef", 0xE782, int_s390_vfaef,
1675 z_vfae_cc, v128f, v128f, 2>;
1676 defm VFAEZB : TernaryOptVRRbSPair<"vfaezb", 0xE782, int_s390_vfaezb,
1677 z_vfaez_cc, v128b, v128b, 0, 2>;
1678 defm VFAEZH : TernaryOptVRRbSPair<"vfaezh", 0xE782, int_s390_vfaezh,
1679 z_vfaez_cc, v128h, v128h, 1, 2>;
1680 defm VFAEZF : TernaryOptVRRbSPair<"vfaezf", 0xE782, int_s390_vfaezf,
1681 z_vfaez_cc, v128f, v128f, 2, 2>;
1683 defm VFEE : BinaryExtraVRRbSPairGeneric<"vfee", 0xE780>;
1684 defm VFEEB : BinaryExtraVRRbSPair<"vfeeb", 0xE780, int_s390_vfeeb,
1685 z_vfee_cc, v128b, v128b, 0>;
1686 defm VFEEH : BinaryExtraVRRbSPair<"vfeeh", 0xE780, int_s390_vfeeh,
1687 z_vfee_cc, v128h, v128h, 1>;
1688 defm VFEEF : BinaryExtraVRRbSPair<"vfeef", 0xE780, int_s390_vfeef,
1689 z_vfee_cc, v128f, v128f, 2>;
1690 defm VFEEZB : BinaryVRRbSPair<"vfeezb", 0xE780, int_s390_vfeezb,
1691 z_vfeez_cc, v128b, v128b, 0, 2>;
1692 defm VFEEZH : BinaryVRRbSPair<"vfeezh", 0xE780, int_s390_vfeezh,
1693 z_vfeez_cc, v128h, v128h, 1, 2>;
1694 defm VFEEZF : BinaryVRRbSPair<"vfeezf", 0xE780, int_s390_vfeezf,
1695 z_vfeez_cc, v128f, v128f, 2, 2>;
1697 defm VFENE : BinaryExtraVRRbSPairGeneric<"vfene", 0xE781>;
1698 defm VFENEB : BinaryExtraVRRbSPair<"vfeneb", 0xE781, int_s390_vfeneb,
1699 z_vfene_cc, v128b, v128b, 0>;
1700 defm VFENEH : BinaryExtraVRRbSPair<"vfeneh", 0xE781, int_s390_vfeneh,
1701 z_vfene_cc, v128h, v128h, 1>;
1702 defm VFENEF : BinaryExtraVRRbSPair<"vfenef", 0xE781, int_s390_vfenef,
1703 z_vfene_cc, v128f, v128f, 2>;
1704 defm VFENEZB : BinaryVRRbSPair<"vfenezb", 0xE781, int_s390_vfenezb,
1705 z_vfenez_cc, v128b, v128b, 0, 2>;
1706 defm VFENEZH : BinaryVRRbSPair<"vfenezh", 0xE781, int_s390_vfenezh,
1707 z_vfenez_cc, v128h, v128h, 1, 2>;
1708 defm VFENEZF : BinaryVRRbSPair<"vfenezf", 0xE781, int_s390_vfenezf,
1709 z_vfenez_cc, v128f, v128f, 2, 2>;
1711 defm VISTR : UnaryExtraVRRaSPairGeneric<"vistr", 0xE75C>;
1712 defm VISTRB : UnaryExtraVRRaSPair<"vistrb", 0xE75C, int_s390_vistrb,
1713 z_vistr_cc, v128b, v128b, 0>;
1714 defm VISTRH : UnaryExtraVRRaSPair<"vistrh", 0xE75C, int_s390_vistrh,
1715 z_vistr_cc, v128h, v128h, 1>;
1716 defm VISTRF : UnaryExtraVRRaSPair<"vistrf", 0xE75C, int_s390_vistrf,
1717 z_vistr_cc, v128f, v128f, 2>;
1719 defm VSTRC : QuaternaryOptVRRdSPairGeneric<"vstrc", 0xE78A>;
1720 defm VSTRCB : QuaternaryOptVRRdSPair<"vstrcb", 0xE78A, int_s390_vstrcb,
1721 z_vstrc_cc, v128b, v128b, 0>;
1722 defm VSTRCH : QuaternaryOptVRRdSPair<"vstrch", 0xE78A, int_s390_vstrch,
1723 z_vstrc_cc, v128h, v128h, 1>;
1724 defm VSTRCF : QuaternaryOptVRRdSPair<"vstrcf", 0xE78A, int_s390_vstrcf,
1725 z_vstrc_cc, v128f, v128f, 2>;
1726 defm VSTRCZB : QuaternaryOptVRRdSPair<"vstrczb", 0xE78A, int_s390_vstrczb,
1727 z_vstrcz_cc, v128b, v128b, 0, 2>;
1728 defm VSTRCZH : QuaternaryOptVRRdSPair<"vstrczh", 0xE78A, int_s390_vstrczh,
1729 z_vstrcz_cc, v128h, v128h, 1, 2>;
1730 defm VSTRCZF : QuaternaryOptVRRdSPair<"vstrczf", 0xE78A, int_s390_vstrczf,
1731 z_vstrcz_cc, v128f, v128f, 2, 2>;
1734 let Predicates = [FeatureVectorEnhancements2] in {
1735 defm VSTRS : TernaryExtraVRRdGeneric<"vstrs", 0xE78B>;
1736 defm VSTRSB : TernaryExtraVRRd<"vstrsb", 0xE78B,
1737 z_vstrs_cc, v128b, v128b, 0>;
1738 defm VSTRSH : TernaryExtraVRRd<"vstrsh", 0xE78B,
1739 z_vstrs_cc, v128b, v128h, 1>;
1740 defm VSTRSF : TernaryExtraVRRd<"vstrsf", 0xE78B,
1741 z_vstrs_cc, v128b, v128f, 2>;
1742 let Defs = [CC] in {
1743 def VSTRSZB : TernaryVRRd<"vstrszb", 0xE78B,
1744 z_vstrsz_cc, v128b, v128b, 0, 2>;
1745 def VSTRSZH : TernaryVRRd<"vstrszh", 0xE78B,
1746 z_vstrsz_cc, v128b, v128h, 1, 2>;
1747 def VSTRSZF : TernaryVRRd<"vstrszf", 0xE78B,
1748 z_vstrsz_cc, v128b, v128f, 2, 2>;
1752 //===----------------------------------------------------------------------===//
1753 // NNP assist instructions
1754 //===----------------------------------------------------------------------===//
1756 let Predicates = [FeatureVector, FeatureNNPAssist] in {
1757 let Uses = [FPC], mayRaiseFPException = 1 in
1758 def VCFN : UnaryVRRaFloatGeneric<"vcfn", 0xE65D>;
1759 def : Pat<(int_s390_vcfn VR128:$x, imm32zx4_timm:$m),
1760 (VCFN VR128:$x, 1, imm32zx4:$m)>;
1762 let Uses = [FPC], mayRaiseFPException = 1 in
1763 def VCLFNL : UnaryVRRaFloatGeneric<"vclfnl", 0xE65E>;
1764 def : Pat<(int_s390_vclfnls VR128:$x, imm32zx4_timm:$m),
1765 (VCLFNL VR128:$x, 2, imm32zx4:$m)>;
1767 let Uses = [FPC], mayRaiseFPException = 1 in
1768 def VCLFNH : UnaryVRRaFloatGeneric<"vclfnh", 0xE656>;
1769 def : Pat<(int_s390_vclfnhs VR128:$x, imm32zx4_timm:$m),
1770 (VCLFNH VR128:$x, 2, imm32zx4:$m)>;
1772 let Uses = [FPC], mayRaiseFPException = 1 in
1773 def VCNF : UnaryVRRaFloatGeneric<"vcnf", 0xE655>;
1774 def : Pat<(int_s390_vcnf VR128:$x, imm32zx4_timm:$m),
1775 (VCNF VR128:$x, imm32zx4:$m, 1)>;
1777 let Uses = [FPC], mayRaiseFPException = 1 in
1778 def VCRNF : BinaryVRRcFloatGeneric<"vcrnf", 0xE675>;
1779 def : Pat<(int_s390_vcrnfs VR128:$x, VR128:$y, imm32zx4_timm:$m),
1780 (VCRNF VR128:$x, VR128:$y, imm32zx4:$m, 2)>;
1783 //===----------------------------------------------------------------------===//
1784 // Packed-decimal instructions
1785 //===----------------------------------------------------------------------===//
1787 let Predicates = [FeatureVectorPackedDecimal] in {
1788 def VLIP : BinaryVRIh<"vlip", 0xE649>;
1790 def VPKZ : BinaryVSI<"vpkz", 0xE634, null_frag, 0>;
1791 def VUPKZ : StoreLengthVSI<"vupkz", 0xE63C, null_frag, 0>;
1793 let Defs = [CC] in {
1794 let Predicates = [FeatureVectorPackedDecimalEnhancement] in {
1795 def VCVBOpt : TernaryVRRi<"vcvb", 0xE650, GR32>;
1796 def VCVBGOpt : TernaryVRRi<"vcvbg", 0xE652, GR64>;
1798 def VCVB : BinaryVRRi<"vcvb", 0xE650, GR32>;
1799 def VCVBG : BinaryVRRi<"vcvbg", 0xE652, GR64>;
1800 def VCVD : TernaryVRIi<"vcvd", 0xE658, GR32>;
1801 def VCVDG : TernaryVRIi<"vcvdg", 0xE65A, GR64>;
1803 def VAP : QuaternaryVRIf<"vap", 0xE671>;
1804 def VSP : QuaternaryVRIf<"vsp", 0xE673>;
1806 def VMP : QuaternaryVRIf<"vmp", 0xE678>;
1807 def VMSP : QuaternaryVRIf<"vmsp", 0xE679>;
1809 def VDP : QuaternaryVRIf<"vdp", 0xE67A>;
1810 def VRP : QuaternaryVRIf<"vrp", 0xE67B>;
1811 def VSDP : QuaternaryVRIf<"vsdp", 0xE67E>;
1813 def VSRP : QuaternaryVRIg<"vsrp", 0xE659>;
1814 def VPSOP : QuaternaryVRIg<"vpsop", 0xE65B>;
1816 def VTP : TestVRRg<"vtp", 0xE65F>;
1817 def VCP : CompareVRRh<"vcp", 0xE677>;
1821 let Predicates = [FeatureVectorPackedDecimalEnhancement2] in {
1822 def VSCHP : BinaryExtraVRRbGeneric<"vschp", 0xE674>;
1823 def VSCHSP : BinaryExtraVRRb<"vschsp", 0xE674, 2>;
1824 def VSCHDP : BinaryExtraVRRb<"vschdp", 0xE674, 3>;
1825 def VSCHXP : BinaryExtraVRRb<"vschxp", 0xE674, 4>;
1827 def VSCSHP : BinaryVRRb<"vscshp", 0xE67C, null_frag, v128b, v128b>;
1829 def VCSPH : TernaryVRRj<"vcsph", 0xE67D>;
1832 def VCLZDP : BinaryVRRk<"vclzdp", 0xE651>;
1835 def VSRPR : QuaternaryVRIf<"vsrpr", 0xE672>;
1837 let Defs = [CC] in {
1838 def VPKZR : QuaternaryVRIf<"vpkzr", 0xE670>;
1839 def VUPKZH : BinaryVRRk<"vupkzh", 0xE654>;
1840 def VUPKZL : BinaryVRRk<"vupkzl", 0xE65C>;