1 //===-- SystemZInstrInfo.td - General SystemZ 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 def IsTargetXPLINK64 : Predicate<"Subtarget->isTargetXPLINK64()">;
10 def IsTargetELF : Predicate<"Subtarget->isTargetELF()">;
12 //===----------------------------------------------------------------------===//
14 //===----------------------------------------------------------------------===//
16 // The callseq_start node requires the hasSideEffects flag, even though these
17 // instructions are noops on SystemZ.
18 let hasNoSchedulingInfo = 1, hasSideEffects = 1 in {
19 def ADJCALLSTACKDOWN : Pseudo<(outs), (ins i64imm:$amt1, i64imm:$amt2),
20 [(callseq_start timm:$amt1, timm:$amt2)]>;
21 def ADJCALLSTACKUP : Pseudo<(outs), (ins i64imm:$amt1, i64imm:$amt2),
22 [(callseq_end timm:$amt1, timm:$amt2)]>;
25 // Takes as input the value of the stack pointer after a dynamic allocation
26 // has been made. Sets the output to the address of the dynamically-
27 // allocated area itself, skipping the outgoing arguments.
29 // This expands to an LA or LAY instruction. We restrict the offset
30 // to the range of LA and keep the LAY range in reserve for when
31 // the size of the outgoing arguments is added.
32 def ADJDYNALLOC : Pseudo<(outs GR64:$dst), (ins dynalloc12only:$src),
33 [(set GR64:$dst, dynalloc12only:$src)]>;
35 let Defs = [R15D, CC], Uses = [R15D], hasNoSchedulingInfo = 1,
36 usesCustomInserter = 1 in
37 def PROBED_ALLOCA : Pseudo<(outs GR64:$dst),
38 (ins GR64:$oldSP, GR64:$space),
39 [(set GR64:$dst, (z_probed_alloca GR64:$oldSP, GR64:$space))]>;
41 let Defs = [R1D, R15D, CC], Uses = [R15D], hasNoSchedulingInfo = 1,
43 def PROBED_STACKALLOC : Pseudo<(outs), (ins i64imm:$stacksize), []>;
45 //===----------------------------------------------------------------------===//
46 // Branch instructions
47 //===----------------------------------------------------------------------===//
49 // Conditional branches.
50 let isBranch = 1, isTerminator = 1, Uses = [CC] in {
51 // It's easier for LLVM to handle these branches in their raw BRC/BRCL form
52 // with the condition-code mask being the first operand. It seems friendlier
53 // to use mnemonic forms like JE and JLH when writing out the assembly though.
54 let isCodeGenOnly = 1 in {
55 // An assembler extended mnemonic for BRC.
56 def BRC : CondBranchRI <"j#", 0xA74, z_br_ccmask>;
57 // An assembler extended mnemonic for BRCL. (The extension is "G"
58 // rather than "L" because "JL" is "Jump if Less".)
59 def BRCL : CondBranchRIL<"jg#", 0xC04>;
60 let isIndirectBranch = 1 in {
61 def BC : CondBranchRX<"b#", 0x47>;
62 def BCR : CondBranchRR<"b#r", 0x07>;
63 def BIC : CondBranchRXY<"bi#", 0xe347>,
64 Requires<[FeatureMiscellaneousExtensions2]>;
68 // Allow using the raw forms directly from the assembler (and occasional
69 // special code generation needs) as well.
70 def BRCAsm : AsmCondBranchRI <"brc", 0xA74>;
71 def BRCLAsm : AsmCondBranchRIL<"brcl", 0xC04>;
72 let isIndirectBranch = 1 in {
73 def BCAsm : AsmCondBranchRX<"bc", 0x47>;
74 def BCRAsm : AsmCondBranchRR<"bcr", 0x07>;
75 def BICAsm : AsmCondBranchRXY<"bic", 0xe347>,
76 Requires<[FeatureMiscellaneousExtensions2]>;
79 // Define AsmParser extended mnemonics for each general condition-code mask
80 // (integer or floating-point)
81 foreach V = [ "E", "NE", "H", "NH", "L", "NL", "HE", "NHE", "LE", "NLE",
82 "Z", "NZ", "P", "NP", "M", "NM", "LH", "NLH", "O", "NO" ] in {
83 def JAsm#V : FixedCondBranchRI <CV<V>, "j#", 0xA74>;
84 def JGAsm#V : FixedCondBranchRIL<CV<V>, "j{g|l}#", 0xC04>;
85 let isIndirectBranch = 1 in {
86 def BAsm#V : FixedCondBranchRX <CV<V>, "b#", 0x47>;
87 def BRAsm#V : FixedCondBranchRR <CV<V>, "b#r", 0x07>;
88 def BIAsm#V : FixedCondBranchRXY<CV<V>, "bi#", 0xe347>,
89 Requires<[FeatureMiscellaneousExtensions2]>;
94 // Unconditional branches. These are in fact simply variants of the
95 // conditional branches with the condition mask set to "always".
96 let isBranch = 1, isTerminator = 1, isBarrier = 1 in {
97 def J : FixedCondBranchRI <CondAlways, "j", 0xA74, br>;
98 def JG : FixedCondBranchRIL<CondAlways, "j{g|lu}", 0xC04>;
99 let isIndirectBranch = 1 in {
100 def B : FixedCondBranchRX<CondAlways, "b", 0x47>;
101 def BR : FixedCondBranchRR<CondAlways, "br", 0x07, brind>;
102 def BI : FixedCondBranchRXY<CondAlways, "bi", 0xe347, brind>,
103 Requires<[FeatureMiscellaneousExtensions2]>;
107 // NOPs. These are again variants of the conditional branches, with the
108 // condition mask set to "never". NOP_bare can't be an InstAlias since it
109 // would need R0D hard coded which is not part of ADDR64BitRegClass.
110 def NOP : InstAlias<"nop\t$XBD", (BCAsm 0, bdxaddr12only:$XBD), 0>;
111 let isAsmParserOnly = 1, hasNoSchedulingInfo = 1, M1 = 0, XBD2 = 0 in
112 def NOP_bare : InstRXb<0x47,(outs), (ins), "nop", []>;
113 def NOPR : InstAlias<"nopr\t$R", (BCRAsm 0, GR64:$R), 0>;
114 def NOPR_bare : InstAlias<"nopr", (BCRAsm 0, R0D), 0>;
116 // An alias of BRC 0, label
117 def JNOP : InstAlias<"jnop\t$RI2", (BRCAsm 0, brtarget16:$RI2), 0>;
119 // An alias of BRCL 0, label
120 // jgnop on att ; jlnop on hlasm
121 def JGNOP : InstAlias<"{jgnop|jlnop}\t$RI2", (BRCLAsm 0, brtarget32:$RI2), 0>;
123 // Fused compare-and-branch instructions.
125 // These instructions do not use or clobber the condition codes.
126 // We nevertheless pretend that the relative compare-and-branch
127 // instructions clobber CC, so that we can lower them to separate
128 // comparisons and BRCLs if the branch ends up being out of range.
129 let isBranch = 1, isTerminator = 1 in {
130 // As for normal branches, we handle these instructions internally in
131 // their raw CRJ-like form, but use assembly macros like CRJE when writing
132 // them out. Using the *Pair multiclasses, we also create the raw forms.
134 defm CRJ : CmpBranchRIEbPair<"crj", 0xEC76, GR32>;
135 defm CGRJ : CmpBranchRIEbPair<"cgrj", 0xEC64, GR64>;
136 defm CIJ : CmpBranchRIEcPair<"cij", 0xEC7E, GR32, imm32sx8>;
137 defm CGIJ : CmpBranchRIEcPair<"cgij", 0xEC7C, GR64, imm64sx8>;
138 defm CLRJ : CmpBranchRIEbPair<"clrj", 0xEC77, GR32>;
139 defm CLGRJ : CmpBranchRIEbPair<"clgrj", 0xEC65, GR64>;
140 defm CLIJ : CmpBranchRIEcPair<"clij", 0xEC7F, GR32, imm32zx8>;
141 defm CLGIJ : CmpBranchRIEcPair<"clgij", 0xEC7D, GR64, imm64zx8>;
143 let isIndirectBranch = 1 in {
144 defm CRB : CmpBranchRRSPair<"crb", 0xECF6, GR32>;
145 defm CGRB : CmpBranchRRSPair<"cgrb", 0xECE4, GR64>;
146 defm CIB : CmpBranchRISPair<"cib", 0xECFE, GR32, imm32sx8>;
147 defm CGIB : CmpBranchRISPair<"cgib", 0xECFC, GR64, imm64sx8>;
148 defm CLRB : CmpBranchRRSPair<"clrb", 0xECF7, GR32>;
149 defm CLGRB : CmpBranchRRSPair<"clgrb", 0xECE5, GR64>;
150 defm CLIB : CmpBranchRISPair<"clib", 0xECFF, GR32, imm32zx8>;
151 defm CLGIB : CmpBranchRISPair<"clgib", 0xECFD, GR64, imm64zx8>;
154 // Define AsmParser mnemonics for each integer condition-code mask.
155 foreach V = [ "E", "H", "L", "HE", "LE", "LH",
156 "NE", "NH", "NL", "NHE", "NLE", "NLH" ] in {
158 def CRJAsm#V : FixedCmpBranchRIEb<ICV<V>, "crj", 0xEC76, GR32>;
159 def CGRJAsm#V : FixedCmpBranchRIEb<ICV<V>, "cgrj", 0xEC64, GR64>;
160 def CIJAsm#V : FixedCmpBranchRIEc<ICV<V>, "cij", 0xEC7E, GR32,
162 def CGIJAsm#V : FixedCmpBranchRIEc<ICV<V>, "cgij", 0xEC7C, GR64,
164 def CLRJAsm#V : FixedCmpBranchRIEb<ICV<V>, "clrj", 0xEC77, GR32>;
165 def CLGRJAsm#V : FixedCmpBranchRIEb<ICV<V>, "clgrj", 0xEC65, GR64>;
166 def CLIJAsm#V : FixedCmpBranchRIEc<ICV<V>, "clij", 0xEC7F, GR32,
168 def CLGIJAsm#V : FixedCmpBranchRIEc<ICV<V>, "clgij", 0xEC7D, GR64,
171 let isIndirectBranch = 1 in {
172 def CRBAsm#V : FixedCmpBranchRRS<ICV<V>, "crb", 0xECF6, GR32>;
173 def CGRBAsm#V : FixedCmpBranchRRS<ICV<V>, "cgrb", 0xECE4, GR64>;
174 def CIBAsm#V : FixedCmpBranchRIS<ICV<V>, "cib", 0xECFE, GR32,
176 def CGIBAsm#V : FixedCmpBranchRIS<ICV<V>, "cgib", 0xECFC, GR64,
178 def CLRBAsm#V : FixedCmpBranchRRS<ICV<V>, "clrb", 0xECF7, GR32>;
179 def CLGRBAsm#V : FixedCmpBranchRRS<ICV<V>, "clgrb", 0xECE5, GR64>;
180 def CLIBAsm#V : FixedCmpBranchRIS<ICV<V>, "clib", 0xECFF, GR32,
182 def CLGIBAsm#V : FixedCmpBranchRIS<ICV<V>, "clgib", 0xECFD, GR64,
188 // Decrement a register and branch if it is nonzero. These don't clobber CC,
189 // but we might need to split long relative branches into sequences that do.
190 let isBranch = 1, isTerminator = 1 in {
192 def BRCT : BranchUnaryRI<"brct", 0xA76, GR32>;
193 def BRCTG : BranchUnaryRI<"brctg", 0xA77, GR64>;
195 // This doesn't need to clobber CC since we never need to split it.
196 def BRCTH : BranchUnaryRIL<"brcth", 0xCC6, GRH32>,
197 Requires<[FeatureHighWord]>;
199 def BCT : BranchUnaryRX<"bct", 0x46,GR32>;
200 def BCTR : BranchUnaryRR<"bctr", 0x06, GR32>;
201 def BCTG : BranchUnaryRXY<"bctg", 0xE346, GR64>;
202 def BCTGR : BranchUnaryRRE<"bctgr", 0xB946, GR64>;
205 let isBranch = 1, isTerminator = 1 in {
207 def BRXH : BranchBinaryRSI<"brxh", 0x84, GR32>;
208 def BRXLE : BranchBinaryRSI<"brxle", 0x85, GR32>;
209 def BRXHG : BranchBinaryRIEe<"brxhg", 0xEC44, GR64>;
210 def BRXLG : BranchBinaryRIEe<"brxlg", 0xEC45, GR64>;
212 def BXH : BranchBinaryRS<"bxh", 0x86, GR32>;
213 def BXLE : BranchBinaryRS<"bxle", 0x87, GR32>;
214 def BXHG : BranchBinaryRSY<"bxhg", 0xEB44, GR64>;
215 def BXLEG : BranchBinaryRSY<"bxleg", 0xEB45, GR64>;
218 //===----------------------------------------------------------------------===//
220 //===----------------------------------------------------------------------===//
222 // Unconditional trap.
223 let hasCtrlDep = 1, hasSideEffects = 1 in
224 def Trap : Alias<4, (outs), (ins), [(trap)]>;
227 let hasCtrlDep = 1, Uses = [CC], hasSideEffects = 1 in
228 def CondTrap : Alias<4, (outs), (ins cond4:$valid, cond4:$R1), []>;
230 // Fused compare-and-trap instructions.
231 let hasCtrlDep = 1, hasSideEffects = 1 in {
232 // These patterns work the same way as for compare-and-branch.
233 defm CRT : CmpBranchRRFcPair<"crt", 0xB972, GR32>;
234 defm CGRT : CmpBranchRRFcPair<"cgrt", 0xB960, GR64>;
235 defm CLRT : CmpBranchRRFcPair<"clrt", 0xB973, GR32>;
236 defm CLGRT : CmpBranchRRFcPair<"clgrt", 0xB961, GR64>;
237 defm CIT : CmpBranchRIEaPair<"cit", 0xEC72, GR32, imm32sx16>;
238 defm CGIT : CmpBranchRIEaPair<"cgit", 0xEC70, GR64, imm64sx16>;
239 defm CLFIT : CmpBranchRIEaPair<"clfit", 0xEC73, GR32, imm32zx16>;
240 defm CLGIT : CmpBranchRIEaPair<"clgit", 0xEC71, GR64, imm64zx16>;
241 let Predicates = [FeatureMiscellaneousExtensions] in {
242 defm CLT : CmpBranchRSYbPair<"clt", 0xEB23, GR32>;
243 defm CLGT : CmpBranchRSYbPair<"clgt", 0xEB2B, GR64>;
246 foreach V = [ "E", "H", "L", "HE", "LE", "LH",
247 "NE", "NH", "NL", "NHE", "NLE", "NLH" ] in {
248 def CRTAsm#V : FixedCmpBranchRRFc<ICV<V>, "crt", 0xB972, GR32>;
249 def CGRTAsm#V : FixedCmpBranchRRFc<ICV<V>, "cgrt", 0xB960, GR64>;
250 def CLRTAsm#V : FixedCmpBranchRRFc<ICV<V>, "clrt", 0xB973, GR32>;
251 def CLGRTAsm#V : FixedCmpBranchRRFc<ICV<V>, "clgrt", 0xB961, GR64>;
252 def CITAsm#V : FixedCmpBranchRIEa<ICV<V>, "cit", 0xEC72, GR32,
254 def CGITAsm#V : FixedCmpBranchRIEa<ICV<V>, "cgit", 0xEC70, GR64,
256 def CLFITAsm#V : FixedCmpBranchRIEa<ICV<V>, "clfit", 0xEC73, GR32,
258 def CLGITAsm#V : FixedCmpBranchRIEa<ICV<V>, "clgit", 0xEC71, GR64,
260 let Predicates = [FeatureMiscellaneousExtensions] in {
261 def CLTAsm#V : FixedCmpBranchRSYb<ICV<V>, "clt", 0xEB23, GR32>;
262 def CLGTAsm#V : FixedCmpBranchRSYb<ICV<V>, "clgt", 0xEB2B, GR64>;
267 //===----------------------------------------------------------------------===//
268 // Call and return instructions
269 //===----------------------------------------------------------------------===//
271 // Define the general form of the call instructions for the asm parser.
272 // These instructions don't hard-code %r14 as the return address register.
273 let isCall = 1, Defs = [CC] in {
274 def BRAS : CallRI <"bras", 0xA75>;
275 def BRASL : CallRIL<"brasl", 0xC05>;
276 def BAS : CallRX <"bas", 0x4D>;
277 def BASR : CallRR <"basr", 0x0D>;
281 let Predicates = [IsTargetXPLINK64] in {
282 let isCall = 1, Defs = [R7D, CC], Uses = [FPC] in {
283 def CallBRASL_XPLINK64 : Alias<8, (outs), (ins pcrel32:$I2, variable_ops),
284 [(z_call pcrel32:$I2)]>;
285 def CallBASR_XPLINK64 : Alias<4, (outs), (ins ADDR64:$R2, variable_ops),
286 [(z_call ADDR64:$R2)]>;
292 let Predicates = [IsTargetELF] in {
293 let isCall = 1, Defs = [R14D, CC], Uses = [FPC] in {
294 def CallBRASL : Alias<6, (outs), (ins pcrel32:$I2, variable_ops),
295 [(z_call pcrel32:$I2)]>;
296 def CallBASR : Alias<2, (outs), (ins ADDR64:$R2, variable_ops),
297 [(z_call ADDR64:$R2)]>;
300 // TLS calls. These will be lowered into a call to __tls_get_offset,
301 // with an extra relocation specifying the TLS symbol.
302 let isCall = 1, Defs = [R14D, CC] in {
303 def TLS_GDCALL : Alias<6, (outs), (ins tlssym:$I2, variable_ops),
304 [(z_tls_gdcall tglobaltlsaddr:$I2)]>;
305 def TLS_LDCALL : Alias<6, (outs), (ins tlssym:$I2, variable_ops),
306 [(z_tls_ldcall tglobaltlsaddr:$I2)]>;
310 // Sibling calls. Indirect sibling calls must be via R6 for XPLink,
312 let isCall = 1, isTerminator = 1, isReturn = 1, isBarrier = 1 in {
313 def CallJG : Alias<6, (outs), (ins pcrel32:$I2),
314 [(z_sibcall pcrel32:$I2)]>;
315 def CallBR : Alias<2, (outs), (ins ADDR64:$R2),
316 [(z_sibcall ADDR64:$R2)]>;
319 // Conditional sibling calls.
320 let CCMaskFirst = 1, isCall = 1, isTerminator = 1, isReturn = 1 in {
321 def CallBRCL : Alias<6, (outs), (ins cond4:$valid, cond4:$R1,
323 def CallBCR : Alias<2, (outs), (ins cond4:$valid, cond4:$R1,
327 // Fused compare and conditional sibling calls.
328 let isCall = 1, isTerminator = 1, isReturn = 1 in {
329 def CRBCall : Alias<6, (outs), (ins GR32:$R1, GR32:$R2, cond4:$M3, ADDR64:$R4), []>;
330 def CGRBCall : Alias<6, (outs), (ins GR64:$R1, GR64:$R2, cond4:$M3, ADDR64:$R4), []>;
331 def CIBCall : Alias<6, (outs), (ins GR32:$R1, imm32sx8:$I2, cond4:$M3, ADDR64:$R4), []>;
332 def CGIBCall : Alias<6, (outs), (ins GR64:$R1, imm64sx8:$I2, cond4:$M3, ADDR64:$R4), []>;
333 def CLRBCall : Alias<6, (outs), (ins GR32:$R1, GR32:$R2, cond4:$M3, ADDR64:$R4), []>;
334 def CLGRBCall : Alias<6, (outs), (ins GR64:$R1, GR64:$R2, cond4:$M3, ADDR64:$R4), []>;
335 def CLIBCall : Alias<6, (outs), (ins GR32:$R1, imm32zx8:$I2, cond4:$M3, ADDR64:$R4), []>;
336 def CLGIBCall : Alias<6, (outs), (ins GR64:$R1, imm64zx8:$I2, cond4:$M3, ADDR64:$R4), []>;
339 // A return instruction (br %r14) for ELF and (b 2 %r7) for XPLink.
340 let isReturn = 1, isTerminator = 1, isBarrier = 1, hasCtrlDep = 1 in
341 def Return : Alias<2, (outs), (ins), [(z_retflag)]>;
343 // A conditional return instruction (bcr <cond>, %r14).
344 let isReturn = 1, isTerminator = 1, hasCtrlDep = 1, CCMaskFirst = 1, Uses = [CC] in
345 def CondReturn : Alias<2, (outs), (ins cond4:$valid, cond4:$R1), []>;
347 // Fused compare and conditional returns.
348 let isReturn = 1, isTerminator = 1, hasCtrlDep = 1 in {
349 def CRBReturn : Alias<6, (outs), (ins GR32:$R1, GR32:$R2, cond4:$M3), []>;
350 def CGRBReturn : Alias<6, (outs), (ins GR64:$R1, GR64:$R2, cond4:$M3), []>;
351 def CIBReturn : Alias<6, (outs), (ins GR32:$R1, imm32sx8:$I2, cond4:$M3), []>;
352 def CGIBReturn : Alias<6, (outs), (ins GR64:$R1, imm64sx8:$I2, cond4:$M3), []>;
353 def CLRBReturn : Alias<6, (outs), (ins GR32:$R1, GR32:$R2, cond4:$M3), []>;
354 def CLGRBReturn : Alias<6, (outs), (ins GR64:$R1, GR64:$R2, cond4:$M3), []>;
355 def CLIBReturn : Alias<6, (outs), (ins GR32:$R1, imm32zx8:$I2, cond4:$M3), []>;
356 def CLGIBReturn : Alias<6, (outs), (ins GR64:$R1, imm64zx8:$I2, cond4:$M3), []>;
359 //===----------------------------------------------------------------------===//
360 // Select instructions
361 //===----------------------------------------------------------------------===//
363 def Select32 : SelectWrapper<i32, GR32>,
364 Requires<[FeatureNoLoadStoreOnCond]>;
365 def Select64 : SelectWrapper<i64, GR64>,
366 Requires<[FeatureNoLoadStoreOnCond]>;
368 // We don't define 32-bit Mux stores if we don't have STOCFH, because the
369 // low-only STOC should then always be used if possible.
370 defm CondStore8Mux : CondStores<GRX32, nonvolatile_truncstorei8,
371 nonvolatile_anyextloadi8, bdxaddr20only>,
372 Requires<[FeatureHighWord]>;
373 defm CondStore16Mux : CondStores<GRX32, nonvolatile_truncstorei16,
374 nonvolatile_anyextloadi16, bdxaddr20only>,
375 Requires<[FeatureHighWord]>;
376 defm CondStore32Mux : CondStores<GRX32, simple_store,
377 simple_load, bdxaddr20only>,
378 Requires<[FeatureLoadStoreOnCond2]>;
379 defm CondStore8 : CondStores<GR32, nonvolatile_truncstorei8,
380 nonvolatile_anyextloadi8, bdxaddr20only>;
381 defm CondStore16 : CondStores<GR32, nonvolatile_truncstorei16,
382 nonvolatile_anyextloadi16, bdxaddr20only>;
383 defm CondStore32 : CondStores<GR32, simple_store,
384 simple_load, bdxaddr20only>;
386 defm : CondStores64<CondStore8, CondStore8Inv, nonvolatile_truncstorei8,
387 nonvolatile_anyextloadi8, bdxaddr20only>;
388 defm : CondStores64<CondStore16, CondStore16Inv, nonvolatile_truncstorei16,
389 nonvolatile_anyextloadi16, bdxaddr20only>;
390 defm : CondStores64<CondStore32, CondStore32Inv, nonvolatile_truncstorei32,
391 nonvolatile_anyextloadi32, bdxaddr20only>;
392 defm CondStore64 : CondStores<GR64, simple_store,
393 simple_load, bdxaddr20only>;
395 //===----------------------------------------------------------------------===//
397 //===----------------------------------------------------------------------===//
400 def LR : UnaryRR <"lr", 0x18, null_frag, GR32, GR32>;
401 def LGR : UnaryRRE<"lgr", 0xB904, null_frag, GR64, GR64>;
403 let Defs = [CC], CCValues = 0xE, CompareZeroCCMask = 0xE in {
404 def LTR : UnaryRR <"ltr", 0x12, null_frag, GR32, GR32>;
405 def LTGR : UnaryRRE<"ltgr", 0xB902, null_frag, GR64, GR64>;
408 let usesCustomInserter = 1, hasNoSchedulingInfo = 1 in
409 def PAIR128 : Pseudo<(outs GR128:$dst), (ins GR64:$hi, GR64:$lo), []>;
412 let isAsCheapAsAMove = 1, isMoveImm = 1, isReMaterializable = 1 in {
413 // 16-bit sign-extended immediates. LHIMux expands to LHI or IIHF,
414 // deopending on the choice of register.
415 def LHIMux : UnaryRIPseudo<bitconvert, GRX32, imm32sx16>,
416 Requires<[FeatureHighWord]>;
417 def LHI : UnaryRI<"lhi", 0xA78, bitconvert, GR32, imm32sx16>;
418 def LGHI : UnaryRI<"lghi", 0xA79, bitconvert, GR64, imm64sx16>;
420 // Other 16-bit immediates.
421 def LLILL : UnaryRI<"llill", 0xA5F, bitconvert, GR64, imm64ll16>;
422 def LLILH : UnaryRI<"llilh", 0xA5E, bitconvert, GR64, imm64lh16>;
423 def LLIHL : UnaryRI<"llihl", 0xA5D, bitconvert, GR64, imm64hl16>;
424 def LLIHH : UnaryRI<"llihh", 0xA5C, bitconvert, GR64, imm64hh16>;
426 // 32-bit immediates.
427 def LGFI : UnaryRIL<"lgfi", 0xC01, bitconvert, GR64, imm64sx32>;
428 def LLILF : UnaryRIL<"llilf", 0xC0F, bitconvert, GR64, imm64lf32>;
429 def LLIHF : UnaryRIL<"llihf", 0xC0E, bitconvert, GR64, imm64hf32>;
433 let canFoldAsLoad = 1, SimpleBDXLoad = 1, mayLoad = 1 in {
434 // Expands to L, LY or LFH, depending on the choice of register.
435 def LMux : UnaryRXYPseudo<"l", load, GRX32, 4>,
436 Requires<[FeatureHighWord]>;
437 defm L : UnaryRXPair<"l", 0x58, 0xE358, load, GR32, 4>;
438 def LFH : UnaryRXY<"lfh", 0xE3CA, load, GRH32, 4>,
439 Requires<[FeatureHighWord]>;
440 def LG : UnaryRXY<"lg", 0xE304, load, GR64, 8>;
442 // These instructions are split after register allocation, so we don't
443 // want a custom inserter.
444 let Has20BitOffset = 1, HasIndex = 1, Is128Bit = 1 in {
445 def L128 : Pseudo<(outs GR128:$dst), (ins bdxaddr20only128:$src),
446 [(set GR128:$dst, (load bdxaddr20only128:$src))]>;
449 let Defs = [CC], CCValues = 0xE, CompareZeroCCMask = 0xE in {
450 def LT : UnaryRXY<"lt", 0xE312, load, GR32, 4>;
451 def LTG : UnaryRXY<"ltg", 0xE302, load, GR64, 8>;
454 let canFoldAsLoad = 1 in {
455 def LRL : UnaryRILPC<"lrl", 0xC4D, aligned_load, GR32>;
456 def LGRL : UnaryRILPC<"lgrl", 0xC48, aligned_load, GR64>;
459 // Load and zero rightmost byte.
460 let Predicates = [FeatureLoadAndZeroRightmostByte] in {
461 def LZRF : UnaryRXY<"lzrf", 0xE33B, null_frag, GR32, 4>;
462 def LZRG : UnaryRXY<"lzrg", 0xE32A, null_frag, GR64, 8>;
463 def : Pat<(and (i32 (load bdxaddr20only:$src)), 0xffffff00),
464 (LZRF bdxaddr20only:$src)>;
465 def : Pat<(and (i64 (load bdxaddr20only:$src)), 0xffffffffffffff00),
466 (LZRG bdxaddr20only:$src)>;
470 let Predicates = [FeatureLoadAndTrap], hasSideEffects = 1 in {
471 def LAT : UnaryRXY<"lat", 0xE39F, null_frag, GR32, 4>;
472 def LFHAT : UnaryRXY<"lfhat", 0xE3C8, null_frag, GRH32, 4>;
473 def LGAT : UnaryRXY<"lgat", 0xE385, null_frag, GR64, 8>;
477 let SimpleBDXStore = 1, mayStore = 1 in {
478 // Expands to ST, STY or STFH, depending on the choice of register.
479 def STMux : StoreRXYPseudo<store, GRX32, 4>,
480 Requires<[FeatureHighWord]>;
481 defm ST : StoreRXPair<"st", 0x50, 0xE350, store, GR32, 4>;
482 def STFH : StoreRXY<"stfh", 0xE3CB, store, GRH32, 4>,
483 Requires<[FeatureHighWord]>;
484 def STG : StoreRXY<"stg", 0xE324, store, GR64, 8>;
486 // These instructions are split after register allocation, so we don't
487 // want a custom inserter.
488 let Has20BitOffset = 1, HasIndex = 1, Is128Bit = 1 in {
489 def ST128 : Pseudo<(outs), (ins GR128:$src, bdxaddr20only128:$dst),
490 [(store GR128:$src, bdxaddr20only128:$dst)]>;
493 def STRL : StoreRILPC<"strl", 0xC4F, aligned_store, GR32>;
494 def STGRL : StoreRILPC<"stgrl", 0xC4B, aligned_store, GR64>;
496 // 8-bit immediate stores to 8-bit fields.
497 defm MVI : StoreSIPair<"mvi", 0x92, 0xEB52, truncstorei8, imm32zx8trunc>;
499 // 16-bit immediate stores to 16-, 32- or 64-bit fields.
500 def MVHHI : StoreSIL<"mvhhi", 0xE544, truncstorei16, imm32sx16trunc>;
501 def MVHI : StoreSIL<"mvhi", 0xE54C, store, imm32sx16>;
502 def MVGHI : StoreSIL<"mvghi", 0xE548, store, imm64sx16>;
504 // Memory-to-memory moves.
505 let mayLoad = 1, mayStore = 1 in
506 defm MVC : MemorySS<"mvc", 0xD2, z_mvc, z_mvc_loop>;
507 let mayLoad = 1, mayStore = 1, Defs = [CC] in {
508 def MVCL : SideEffectBinaryMemMemRR<"mvcl", 0x0E, GR128, GR128>;
509 def MVCLE : SideEffectTernaryMemMemRS<"mvcle", 0xA8, GR128, GR128>;
510 def MVCLU : SideEffectTernaryMemMemRSY<"mvclu", 0xEB8E, GR128, GR128>;
514 let Predicates = [FeatureMiscellaneousExtensions3],
515 mayLoad = 1, mayStore = 1, Uses = [R0L] in
516 def MVCRL : SideEffectBinarySSE<"mvcrl", 0xE50A>;
519 let mayLoad = 1, mayStore = 1, Defs = [CC] in
520 defm MVST : StringRRE<"mvst", 0xB255, z_stpcpy>;
522 //===----------------------------------------------------------------------===//
523 // Conditional move instructions
524 //===----------------------------------------------------------------------===//
526 let Predicates = [FeatureMiscellaneousExtensions3], Uses = [CC] in {
528 let isCommutable = 1 in {
529 // Expands to SELR or SELFHR or a branch-and-move sequence,
530 // depending on the choice of registers.
531 def SELRMux : CondBinaryRRFaPseudo<"MUXselr", GRX32, GRX32, GRX32>;
532 defm SELFHR : CondBinaryRRFaPair<"selfhr", 0xB9C0, GRH32, GRH32, GRH32>;
533 defm SELR : CondBinaryRRFaPair<"selr", 0xB9F0, GR32, GR32, GR32>;
534 defm SELGR : CondBinaryRRFaPair<"selgr", 0xB9E3, GR64, GR64, GR64>;
537 // Define AsmParser extended mnemonics for each general condition-code mask.
538 foreach V = [ "E", "NE", "H", "NH", "L", "NL", "HE", "NHE", "LE", "NLE",
539 "Z", "NZ", "P", "NP", "M", "NM", "LH", "NLH", "O", "NO" ] in {
540 def SELRAsm#V : FixedCondBinaryRRFa<CV<V>, "selr", 0xB9F0,
542 def SELFHRAsm#V : FixedCondBinaryRRFa<CV<V>, "selfhr", 0xB9C0,
543 GRH32, GRH32, GRH32>;
544 def SELGRAsm#V : FixedCondBinaryRRFa<CV<V>, "selgr", 0xB9E3,
549 let Predicates = [FeatureLoadStoreOnCond2], Uses = [CC] in {
550 // Load immediate on condition. Matched via DAG pattern and created
551 // by the PeepholeOptimizer via FoldImmediate.
553 // Expands to LOCHI or LOCHHI, depending on the choice of register.
554 def LOCHIMux : CondBinaryRIEPseudo<GRX32, imm32sx16>;
555 defm LOCHHI : CondBinaryRIEPair<"lochhi", 0xEC4E, GRH32, imm32sx16>;
556 defm LOCHI : CondBinaryRIEPair<"lochi", 0xEC42, GR32, imm32sx16>;
557 defm LOCGHI : CondBinaryRIEPair<"locghi", 0xEC46, GR64, imm64sx16>;
559 // Move register on condition. Matched via DAG pattern and
560 // created by early if-conversion.
561 let isCommutable = 1 in {
562 // Expands to LOCR or LOCFHR or a branch-and-move sequence,
563 // depending on the choice of registers.
564 def LOCRMux : CondBinaryRRFPseudo<"MUXlocr", GRX32, GRX32>;
565 defm LOCFHR : CondBinaryRRFPair<"locfhr", 0xB9E0, GRH32, GRH32>;
568 // Load on condition. Matched via DAG pattern.
569 // Expands to LOC or LOCFH, depending on the choice of register.
570 defm LOCMux : CondUnaryRSYPseudoAndMemFold<"MUXloc", simple_load, GRX32, 4>;
571 defm LOCFH : CondUnaryRSYPair<"locfh", 0xEBE0, simple_load, GRH32, 4>;
573 // Store on condition. Expanded from CondStore* pseudos.
574 // Expands to STOC or STOCFH, depending on the choice of register.
575 def STOCMux : CondStoreRSYPseudo<GRX32, 4>;
576 defm STOCFH : CondStoreRSYPair<"stocfh", 0xEBE1, GRH32, 4>;
578 // Define AsmParser extended mnemonics for each general condition-code mask.
579 foreach V = [ "E", "NE", "H", "NH", "L", "NL", "HE", "NHE", "LE", "NLE",
580 "Z", "NZ", "P", "NP", "M", "NM", "LH", "NLH", "O", "NO" ] in {
581 def LOCHIAsm#V : FixedCondBinaryRIE<CV<V>, "lochi", 0xEC42, GR32,
583 def LOCGHIAsm#V : FixedCondBinaryRIE<CV<V>, "locghi", 0xEC46, GR64,
585 def LOCHHIAsm#V : FixedCondBinaryRIE<CV<V>, "lochhi", 0xEC4E, GRH32,
587 def LOCFHRAsm#V : FixedCondBinaryRRF<CV<V>, "locfhr", 0xB9E0, GRH32, GRH32>;
588 def LOCFHAsm#V : FixedCondUnaryRSY<CV<V>, "locfh", 0xEBE0, GRH32, 4>;
589 def STOCFHAsm#V : FixedCondStoreRSY<CV<V>, "stocfh", 0xEBE1, GRH32, 4>;
593 let Predicates = [FeatureLoadStoreOnCond], Uses = [CC] in {
594 // Move register on condition. Matched via DAG pattern and
595 // created by early if-conversion.
596 let isCommutable = 1 in {
597 defm LOCR : CondBinaryRRFPair<"locr", 0xB9F2, GR32, GR32>;
598 defm LOCGR : CondBinaryRRFPair<"locgr", 0xB9E2, GR64, GR64>;
601 // Load on condition. Matched via DAG pattern.
602 defm LOC : CondUnaryRSYPair<"loc", 0xEBF2, simple_load, GR32, 4>;
603 defm LOCG : CondUnaryRSYPairAndMemFold<"locg", 0xEBE2, simple_load, GR64, 8>;
605 // Store on condition. Expanded from CondStore* pseudos.
606 defm STOC : CondStoreRSYPair<"stoc", 0xEBF3, GR32, 4>;
607 defm STOCG : CondStoreRSYPair<"stocg", 0xEBE3, GR64, 8>;
609 // Define AsmParser extended mnemonics for each general condition-code mask.
610 foreach V = [ "E", "NE", "H", "NH", "L", "NL", "HE", "NHE", "LE", "NLE",
611 "Z", "NZ", "P", "NP", "M", "NM", "LH", "NLH", "O", "NO" ] in {
612 def LOCRAsm#V : FixedCondBinaryRRF<CV<V>, "locr", 0xB9F2, GR32, GR32>;
613 def LOCGRAsm#V : FixedCondBinaryRRF<CV<V>, "locgr", 0xB9E2, GR64, GR64>;
614 def LOCAsm#V : FixedCondUnaryRSY<CV<V>, "loc", 0xEBF2, GR32, 4>;
615 def LOCGAsm#V : FixedCondUnaryRSY<CV<V>, "locg", 0xEBE2, GR64, 8>;
616 def STOCAsm#V : FixedCondStoreRSY<CV<V>, "stoc", 0xEBF3, GR32, 4>;
617 def STOCGAsm#V : FixedCondStoreRSY<CV<V>, "stocg", 0xEBE3, GR64, 8>;
620 //===----------------------------------------------------------------------===//
622 //===----------------------------------------------------------------------===//
624 // Note that putting these before zero extensions mean that we will prefer
625 // them for anyextload*. There's not really much to choose between the two
626 // either way, but signed-extending loads have a short LH and a long LHY,
627 // while zero-extending loads have only the long LLH.
629 //===----------------------------------------------------------------------===//
631 // 32-bit extensions from registers.
632 def LBR : UnaryRRE<"lbr", 0xB926, sext8, GR32, GR32>;
633 def LHR : UnaryRRE<"lhr", 0xB927, sext16, GR32, GR32>;
635 // 64-bit extensions from registers.
636 def LGBR : UnaryRRE<"lgbr", 0xB906, sext8, GR64, GR64>;
637 def LGHR : UnaryRRE<"lghr", 0xB907, sext16, GR64, GR64>;
638 def LGFR : UnaryRRE<"lgfr", 0xB914, sext32, GR64, GR32>;
640 let Defs = [CC], CCValues = 0xE, CompareZeroCCMask = 0xE in
641 def LTGFR : UnaryRRE<"ltgfr", 0xB912, null_frag, GR64, GR32>;
643 // Match 32-to-64-bit sign extensions in which the source is already
644 // in a 64-bit register.
645 def : Pat<(sext_inreg GR64:$src, i32),
646 (LGFR (EXTRACT_SUBREG GR64:$src, subreg_l32))>;
648 // 32-bit extensions from 8-bit memory. LBMux expands to LB or LBH,
649 // depending on the choice of register.
650 def LBMux : UnaryRXYPseudo<"lb", asextloadi8, GRX32, 1>,
651 Requires<[FeatureHighWord]>;
652 def LB : UnaryRXY<"lb", 0xE376, asextloadi8, GR32, 1>;
653 def LBH : UnaryRXY<"lbh", 0xE3C0, asextloadi8, GRH32, 1>,
654 Requires<[FeatureHighWord]>;
656 // 32-bit extensions from 16-bit memory. LHMux expands to LH or LHH,
657 // depending on the choice of register.
658 def LHMux : UnaryRXYPseudo<"lh", asextloadi16, GRX32, 2>,
659 Requires<[FeatureHighWord]>;
660 defm LH : UnaryRXPair<"lh", 0x48, 0xE378, asextloadi16, GR32, 2>;
661 def LHH : UnaryRXY<"lhh", 0xE3C4, asextloadi16, GRH32, 2>,
662 Requires<[FeatureHighWord]>;
663 def LHRL : UnaryRILPC<"lhrl", 0xC45, aligned_asextloadi16, GR32>;
665 // 64-bit extensions from memory.
666 def LGB : UnaryRXY<"lgb", 0xE377, asextloadi8, GR64, 1>;
667 def LGH : UnaryRXY<"lgh", 0xE315, asextloadi16, GR64, 2>;
668 def LGF : UnaryRXY<"lgf", 0xE314, asextloadi32, GR64, 4>;
669 def LGHRL : UnaryRILPC<"lghrl", 0xC44, aligned_asextloadi16, GR64>;
670 def LGFRL : UnaryRILPC<"lgfrl", 0xC4C, aligned_asextloadi32, GR64>;
671 let Defs = [CC], CCValues = 0xE, CompareZeroCCMask = 0xE in
672 def LTGF : UnaryRXY<"ltgf", 0xE332, asextloadi32, GR64, 4>;
674 //===----------------------------------------------------------------------===//
676 //===----------------------------------------------------------------------===//
678 // 32-bit extensions from registers.
680 // Expands to LLCR or RISB[LH]G, depending on the choice of registers.
681 def LLCRMux : UnaryRRPseudo<"llcr", zext8, GRX32, GRX32>,
682 Requires<[FeatureHighWord]>;
683 def LLCR : UnaryRRE<"llcr", 0xB994, zext8, GR32, GR32>;
684 // Expands to LLHR or RISB[LH]G, depending on the choice of registers.
685 def LLHRMux : UnaryRRPseudo<"llhr", zext16, GRX32, GRX32>,
686 Requires<[FeatureHighWord]>;
687 def LLHR : UnaryRRE<"llhr", 0xB995, zext16, GR32, GR32>;
689 // 64-bit extensions from registers.
690 def LLGCR : UnaryRRE<"llgcr", 0xB984, zext8, GR64, GR64>;
691 def LLGHR : UnaryRRE<"llghr", 0xB985, zext16, GR64, GR64>;
692 def LLGFR : UnaryRRE<"llgfr", 0xB916, zext32, GR64, GR32>;
694 // Match 32-to-64-bit zero extensions in which the source is already
695 // in a 64-bit register.
696 def : Pat<(and GR64:$src, 0xffffffff),
697 (LLGFR (EXTRACT_SUBREG GR64:$src, subreg_l32))>;
699 // 32-bit extensions from 8-bit memory. LLCMux expands to LLC or LLCH,
700 // depending on the choice of register.
701 def LLCMux : UnaryRXYPseudo<"llc", azextloadi8, GRX32, 1>,
702 Requires<[FeatureHighWord]>;
703 def LLC : UnaryRXY<"llc", 0xE394, azextloadi8, GR32, 1>;
704 def LLCH : UnaryRXY<"llch", 0xE3C2, azextloadi8, GRH32, 1>,
705 Requires<[FeatureHighWord]>;
707 // 32-bit extensions from 16-bit memory. LLHMux expands to LLH or LLHH,
708 // depending on the choice of register.
709 def LLHMux : UnaryRXYPseudo<"llh", azextloadi16, GRX32, 2>,
710 Requires<[FeatureHighWord]>;
711 def LLH : UnaryRXY<"llh", 0xE395, azextloadi16, GR32, 2>;
712 def LLHH : UnaryRXY<"llhh", 0xE3C6, azextloadi16, GRH32, 2>,
713 Requires<[FeatureHighWord]>;
714 def LLHRL : UnaryRILPC<"llhrl", 0xC42, aligned_azextloadi16, GR32>;
716 // 64-bit extensions from memory.
717 def LLGC : UnaryRXY<"llgc", 0xE390, azextloadi8, GR64, 1>;
718 def LLGH : UnaryRXY<"llgh", 0xE391, azextloadi16, GR64, 2>;
719 def LLGF : UnaryRXY<"llgf", 0xE316, azextloadi32, GR64, 4>;
720 def LLGHRL : UnaryRILPC<"llghrl", 0xC46, aligned_azextloadi16, GR64>;
721 def LLGFRL : UnaryRILPC<"llgfrl", 0xC4E, aligned_azextloadi32, GR64>;
723 // 31-to-64-bit zero extensions.
724 def LLGTR : UnaryRRE<"llgtr", 0xB917, null_frag, GR64, GR64>;
725 def LLGT : UnaryRXY<"llgt", 0xE317, null_frag, GR64, 4>;
726 def : Pat<(and GR64:$src, 0x7fffffff),
728 def : Pat<(and (i64 (azextloadi32 bdxaddr20only:$src)), 0x7fffffff),
729 (LLGT bdxaddr20only:$src)>;
731 // Load and zero rightmost byte.
732 let Predicates = [FeatureLoadAndZeroRightmostByte] in {
733 def LLZRGF : UnaryRXY<"llzrgf", 0xE33A, null_frag, GR64, 4>;
734 def : Pat<(and (i64 (azextloadi32 bdxaddr20only:$src)), 0xffffff00),
735 (LLZRGF bdxaddr20only:$src)>;
739 let Predicates = [FeatureLoadAndTrap], hasSideEffects = 1 in {
740 def LLGFAT : UnaryRXY<"llgfat", 0xE39D, null_frag, GR64, 4>;
741 def LLGTAT : UnaryRXY<"llgtat", 0xE39C, null_frag, GR64, 4>;
744 // Extend GR64s to GR128s.
745 let usesCustomInserter = 1, hasNoSchedulingInfo = 1 in
746 def ZEXT128 : Pseudo<(outs GR128:$dst), (ins GR64:$src), []>;
748 //===----------------------------------------------------------------------===//
750 //===----------------------------------------------------------------------===//
752 // Use subregs to populate the "don't care" bits in a 32-bit to 64-bit anyext.
753 def : Pat<(i64 (anyext GR32:$src)),
754 (INSERT_SUBREG (i64 (IMPLICIT_DEF)), GR32:$src, subreg_l32)>;
756 // Extend GR64s to GR128s.
757 let usesCustomInserter = 1, hasNoSchedulingInfo = 1 in
758 def AEXT128 : Pseudo<(outs GR128:$dst), (ins GR64:$src), []>;
760 //===----------------------------------------------------------------------===//
762 //===----------------------------------------------------------------------===//
764 // Truncations of 64-bit registers to 32-bit registers.
765 def : Pat<(i32 (trunc GR64:$src)),
766 (EXTRACT_SUBREG GR64:$src, subreg_l32)>;
768 // Truncations of 32-bit registers to 8-bit memory. STCMux expands to
769 // STC, STCY or STCH, depending on the choice of register.
770 def STCMux : StoreRXYPseudo<truncstorei8, GRX32, 1>,
771 Requires<[FeatureHighWord]>;
772 defm STC : StoreRXPair<"stc", 0x42, 0xE372, truncstorei8, GR32, 1>;
773 def STCH : StoreRXY<"stch", 0xE3C3, truncstorei8, GRH32, 1>,
774 Requires<[FeatureHighWord]>;
776 // Truncations of 32-bit registers to 16-bit memory. STHMux expands to
777 // STH, STHY or STHH, depending on the choice of register.
778 def STHMux : StoreRXYPseudo<truncstorei16, GRX32, 1>,
779 Requires<[FeatureHighWord]>;
780 defm STH : StoreRXPair<"sth", 0x40, 0xE370, truncstorei16, GR32, 2>;
781 def STHH : StoreRXY<"sthh", 0xE3C7, truncstorei16, GRH32, 2>,
782 Requires<[FeatureHighWord]>;
783 def STHRL : StoreRILPC<"sthrl", 0xC47, aligned_truncstorei16, GR32>;
785 // Truncations of 64-bit registers to memory.
786 defm : StoreGR64Pair<STC, STCY, truncstorei8>;
787 defm : StoreGR64Pair<STH, STHY, truncstorei16>;
788 def : StoreGR64PC<STHRL, aligned_truncstorei16>;
789 defm : StoreGR64Pair<ST, STY, truncstorei32>;
790 def : StoreGR64PC<STRL, aligned_truncstorei32>;
792 // Store characters under mask -- not (yet) used for codegen.
793 defm STCM : StoreBinaryRSPair<"stcm", 0xBE, 0xEB2D, GR32, 0>;
794 def STCMH : StoreBinaryRSY<"stcmh", 0xEB2C, GRH32, 0>;
796 //===----------------------------------------------------------------------===//
797 // Multi-register moves
798 //===----------------------------------------------------------------------===//
800 // Multi-register loads.
801 defm LM : LoadMultipleRSPair<"lm", 0x98, 0xEB98, GR32>;
802 def LMG : LoadMultipleRSY<"lmg", 0xEB04, GR64>;
803 def LMH : LoadMultipleRSY<"lmh", 0xEB96, GRH32>;
804 def LMD : LoadMultipleSSe<"lmd", 0xEF, GR64>;
806 // Multi-register stores.
807 defm STM : StoreMultipleRSPair<"stm", 0x90, 0xEB90, GR32>;
808 def STMG : StoreMultipleRSY<"stmg", 0xEB24, GR64>;
809 def STMH : StoreMultipleRSY<"stmh", 0xEB26, GRH32>;
811 //===----------------------------------------------------------------------===//
813 //===----------------------------------------------------------------------===//
815 // Byte-swapping register moves.
816 def LRVR : UnaryRRE<"lrvr", 0xB91F, bswap, GR32, GR32>;
817 def LRVGR : UnaryRRE<"lrvgr", 0xB90F, bswap, GR64, GR64>;
819 // Byte-swapping loads.
820 def LRVH : UnaryRXY<"lrvh", 0xE31F, z_loadbswap16, GR32, 2>;
821 def LRV : UnaryRXY<"lrv", 0xE31E, z_loadbswap32, GR32, 4>;
822 def LRVG : UnaryRXY<"lrvg", 0xE30F, z_loadbswap64, GR64, 8>;
824 // Byte-swapping stores.
825 def STRVH : StoreRXY<"strvh", 0xE33F, z_storebswap16, GR32, 2>;
826 def STRV : StoreRXY<"strv", 0xE33E, z_storebswap32, GR32, 4>;
827 def STRVG : StoreRXY<"strvg", 0xE32F, z_storebswap64, GR64, 8>;
829 // Byte-swapping memory-to-memory moves.
830 let mayLoad = 1, mayStore = 1 in
831 def MVCIN : SideEffectBinarySSa<"mvcin", 0xE8>;
833 //===----------------------------------------------------------------------===//
834 // Load address instructions
835 //===----------------------------------------------------------------------===//
837 // Load BDX-style addresses.
838 let isAsCheapAsAMove = 1, isReMaterializable = 1 in
839 defm LA : LoadAddressRXPair<"la", 0x41, 0xE371, bitconvert>;
841 // Load a PC-relative address. There's no version of this instruction
842 // with a 16-bit offset, so there's no relaxation.
843 let isAsCheapAsAMove = 1, isMoveImm = 1, isReMaterializable = 1 in
844 def LARL : LoadAddressRIL<"larl", 0xC00, bitconvert>;
846 // Load the Global Offset Table address. This will be lowered into a
847 // larl $R1, _GLOBAL_OFFSET_TABLE_
849 def GOT : Alias<6, (outs GR64:$R1), (ins),
850 [(set GR64:$R1, (global_offset_table))]>;
852 //===----------------------------------------------------------------------===//
853 // Absolute and Negation
854 //===----------------------------------------------------------------------===//
857 let CCValues = 0xF, CompareZeroCCMask = 0x8 in {
858 def LPR : UnaryRR <"lpr", 0x10, abs, GR32, GR32>;
859 def LPGR : UnaryRRE<"lpgr", 0xB900, abs, GR64, GR64>;
861 let CCValues = 0xE, CompareZeroCCMask = 0xE in
862 def LPGFR : UnaryRRE<"lpgfr", 0xB910, null_frag, GR64, GR32>;
864 defm : SXU<abs, LPGFR>;
867 let CCValues = 0xF, CompareZeroCCMask = 0x8 in {
868 def LNR : UnaryRR <"lnr", 0x11, z_inegabs, GR32, GR32>;
869 def LNGR : UnaryRRE<"lngr", 0xB901, z_inegabs, GR64, GR64>;
871 let CCValues = 0xE, CompareZeroCCMask = 0xE in
872 def LNGFR : UnaryRRE<"lngfr", 0xB911, null_frag, GR64, GR32>;
874 defm : SXU<z_inegabs, LNGFR>;
877 let CCValues = 0xF, CompareZeroCCMask = 0x8 in {
878 def LCR : UnaryRR <"lcr", 0x13, ineg, GR32, GR32>;
879 def LCGR : UnaryRRE<"lcgr", 0xB903, ineg, GR64, GR64>;
881 let CCValues = 0xE, CompareZeroCCMask = 0xE in
882 def LCGFR : UnaryRRE<"lcgfr", 0xB913, null_frag, GR64, GR32>;
884 defm : SXU<ineg, LCGFR>;
886 //===----------------------------------------------------------------------===//
888 //===----------------------------------------------------------------------===//
890 let isCodeGenOnly = 1 in
891 defm IC32 : BinaryRXPair<"ic", 0x43, 0xE373, inserti8, GR32, azextloadi8, 1>;
892 defm IC : BinaryRXPair<"ic", 0x43, 0xE373, inserti8, GR64, azextloadi8, 1>;
894 defm : InsertMem<"inserti8", IC32, GR32, azextloadi8, bdxaddr12pair>;
895 defm : InsertMem<"inserti8", IC32Y, GR32, azextloadi8, bdxaddr20pair>;
897 defm : InsertMem<"inserti8", IC, GR64, azextloadi8, bdxaddr12pair>;
898 defm : InsertMem<"inserti8", ICY, GR64, azextloadi8, bdxaddr20pair>;
900 // Insert characters under mask -- not (yet) used for codegen.
902 defm ICM : TernaryRSPair<"icm", 0xBF, 0xEB81, GR32, 0>;
903 def ICMH : TernaryRSY<"icmh", 0xEB80, GRH32, 0>;
906 // Insertions of a 16-bit immediate, leaving other bits unaffected.
907 // We don't have or_as_insert equivalents of these operations because
908 // OI is available instead.
910 // IIxMux expands to II[LH]x, depending on the choice of register.
911 def IILMux : BinaryRIPseudo<insertll, GRX32, imm32ll16>,
912 Requires<[FeatureHighWord]>;
913 def IIHMux : BinaryRIPseudo<insertlh, GRX32, imm32lh16>,
914 Requires<[FeatureHighWord]>;
915 def IILL : BinaryRI<"iill", 0xA53, insertll, GR32, imm32ll16>;
916 def IILH : BinaryRI<"iilh", 0xA52, insertlh, GR32, imm32lh16>;
917 def IIHL : BinaryRI<"iihl", 0xA51, insertll, GRH32, imm32ll16>;
918 def IIHH : BinaryRI<"iihh", 0xA50, insertlh, GRH32, imm32lh16>;
919 def IILL64 : BinaryAliasRI<insertll, GR64, imm64ll16>;
920 def IILH64 : BinaryAliasRI<insertlh, GR64, imm64lh16>;
921 def IIHL64 : BinaryAliasRI<inserthl, GR64, imm64hl16>;
922 def IIHH64 : BinaryAliasRI<inserthh, GR64, imm64hh16>;
924 // ...likewise for 32-bit immediates. For GR32s this is a general
925 // full-width move. (We use IILF rather than something like LLILF
926 // for 32-bit moves because IILF leaves the upper 32 bits of the
928 let isAsCheapAsAMove = 1, isMoveImm = 1, isReMaterializable = 1 in {
929 def IIFMux : UnaryRIPseudo<bitconvert, GRX32, uimm32>,
930 Requires<[FeatureHighWord]>;
931 def IILF : UnaryRIL<"iilf", 0xC09, bitconvert, GR32, uimm32>;
932 def IIHF : UnaryRIL<"iihf", 0xC08, bitconvert, GRH32, uimm32>;
934 def IILF64 : BinaryAliasRIL<insertlf, GR64, imm64lf32>;
935 def IIHF64 : BinaryAliasRIL<inserthf, GR64, imm64hf32>;
937 // An alternative model of inserthf, with the first operand being
938 // a zero-extended value.
939 def : Pat<(or (zext32 GR32:$src), imm64hf32:$imm),
940 (IIHF64 (INSERT_SUBREG (i64 (IMPLICIT_DEF)), GR32:$src, subreg_l32),
943 //===----------------------------------------------------------------------===//
945 //===----------------------------------------------------------------------===//
947 // Addition producing a signed overflow flag.
948 let Defs = [CC], CCValues = 0xF, CCIfNoSignedWrap = 1 in {
949 // Addition of a register.
950 let isCommutable = 1 in {
951 defm AR : BinaryRRAndK<"ar", 0x1A, 0xB9F8, z_sadd, GR32, GR32>;
952 defm AGR : BinaryRREAndK<"agr", 0xB908, 0xB9E8, z_sadd, GR64, GR64>;
954 def AGFR : BinaryRRE<"agfr", 0xB918, null_frag, GR64, GR32>;
956 // Addition to a high register.
957 def AHHHR : BinaryRRFa<"ahhhr", 0xB9C8, null_frag, GRH32, GRH32, GRH32>,
958 Requires<[FeatureHighWord]>;
959 def AHHLR : BinaryRRFa<"ahhlr", 0xB9D8, null_frag, GRH32, GRH32, GR32>,
960 Requires<[FeatureHighWord]>;
962 // Addition of signed 16-bit immediates.
963 defm AHIMux : BinaryRIAndKPseudo<"ahimux", z_sadd, GRX32, imm32sx16>;
964 defm AHI : BinaryRIAndK<"ahi", 0xA7A, 0xECD8, z_sadd, GR32, imm32sx16>;
965 defm AGHI : BinaryRIAndK<"aghi", 0xA7B, 0xECD9, z_sadd, GR64, imm64sx16>;
967 // Addition of signed 32-bit immediates.
968 def AFIMux : BinaryRIPseudo<z_sadd, GRX32, simm32>,
969 Requires<[FeatureHighWord]>;
970 def AFI : BinaryRIL<"afi", 0xC29, z_sadd, GR32, simm32>;
971 def AIH : BinaryRIL<"aih", 0xCC8, z_sadd, GRH32, simm32>,
972 Requires<[FeatureHighWord]>;
973 def AGFI : BinaryRIL<"agfi", 0xC28, z_sadd, GR64, imm64sx32>;
975 // Addition of memory.
976 defm AH : BinaryRXPair<"ah", 0x4A, 0xE37A, z_sadd, GR32, asextloadi16, 2>;
977 defm A : BinaryRXPairAndPseudo<"a", 0x5A, 0xE35A, z_sadd, GR32, load, 4>;
978 def AGH : BinaryRXY<"agh", 0xE338, z_sadd, GR64, asextloadi16, 2>,
979 Requires<[FeatureMiscellaneousExtensions2]>;
980 def AGF : BinaryRXY<"agf", 0xE318, z_sadd, GR64, asextloadi32, 4>;
981 defm AG : BinaryRXYAndPseudo<"ag", 0xE308, z_sadd, GR64, load, 8>;
983 // Addition to memory.
984 def ASI : BinarySIY<"asi", 0xEB6A, add, imm32sx8>;
985 def AGSI : BinarySIY<"agsi", 0xEB7A, add, imm64sx8>;
987 defm : SXB<z_sadd, GR64, AGFR>;
989 // Addition producing a carry.
990 let Defs = [CC], CCValues = 0xF, IsLogical = 1 in {
991 // Addition of a register.
992 let isCommutable = 1 in {
993 defm ALR : BinaryRRAndK<"alr", 0x1E, 0xB9FA, z_uadd, GR32, GR32>;
994 defm ALGR : BinaryRREAndK<"algr", 0xB90A, 0xB9EA, z_uadd, GR64, GR64>;
996 def ALGFR : BinaryRRE<"algfr", 0xB91A, null_frag, GR64, GR32>;
998 // Addition to a high register.
999 def ALHHHR : BinaryRRFa<"alhhhr", 0xB9CA, null_frag, GRH32, GRH32, GRH32>,
1000 Requires<[FeatureHighWord]>;
1001 def ALHHLR : BinaryRRFa<"alhhlr", 0xB9DA, null_frag, GRH32, GRH32, GR32>,
1002 Requires<[FeatureHighWord]>;
1004 // Addition of signed 16-bit immediates.
1005 def ALHSIK : BinaryRIE<"alhsik", 0xECDA, z_uadd, GR32, imm32sx16>,
1006 Requires<[FeatureDistinctOps]>;
1007 def ALGHSIK : BinaryRIE<"alghsik", 0xECDB, z_uadd, GR64, imm64sx16>,
1008 Requires<[FeatureDistinctOps]>;
1010 // Addition of unsigned 32-bit immediates.
1011 def ALFI : BinaryRIL<"alfi", 0xC2B, z_uadd, GR32, uimm32>;
1012 def ALGFI : BinaryRIL<"algfi", 0xC2A, z_uadd, GR64, imm64zx32>;
1014 // Addition of signed 32-bit immediates.
1015 def ALSIH : BinaryRIL<"alsih", 0xCCA, null_frag, GRH32, simm32>,
1016 Requires<[FeatureHighWord]>;
1018 // Addition of memory.
1019 defm AL : BinaryRXPairAndPseudo<"al", 0x5E, 0xE35E, z_uadd, GR32, load, 4>;
1020 def ALGF : BinaryRXY<"algf", 0xE31A, z_uadd, GR64, azextloadi32, 4>;
1021 defm ALG : BinaryRXYAndPseudo<"alg", 0xE30A, z_uadd, GR64, load, 8>;
1023 // Addition to memory.
1024 def ALSI : BinarySIY<"alsi", 0xEB6E, null_frag, imm32sx8>;
1025 def ALGSI : BinarySIY<"algsi", 0xEB7E, null_frag, imm64sx8>;
1027 defm : ZXB<z_uadd, GR64, ALGFR>;
1029 // Addition producing and using a carry.
1030 let Defs = [CC], Uses = [CC], CCValues = 0xF, IsLogical = 1 in {
1031 // Addition of a register.
1032 def ALCR : BinaryRRE<"alcr", 0xB998, z_addcarry, GR32, GR32>;
1033 def ALCGR : BinaryRRE<"alcgr", 0xB988, z_addcarry, GR64, GR64>;
1035 // Addition of memory.
1036 def ALC : BinaryRXY<"alc", 0xE398, z_addcarry, GR32, load, 4>;
1037 def ALCG : BinaryRXY<"alcg", 0xE388, z_addcarry, GR64, load, 8>;
1040 // Addition that does not modify the condition code.
1041 def ALSIHN : BinaryRIL<"alsihn", 0xCCB, null_frag, GRH32, simm32>,
1042 Requires<[FeatureHighWord]>;
1045 //===----------------------------------------------------------------------===//
1047 //===----------------------------------------------------------------------===//
1049 // Subtraction producing a signed overflow flag.
1050 let Defs = [CC], CCValues = 0xF, CompareZeroCCMask = 0x8,
1051 CCIfNoSignedWrap = 1 in {
1052 // Subtraction of a register.
1053 defm SR : BinaryRRAndK<"sr", 0x1B, 0xB9F9, z_ssub, GR32, GR32>;
1054 def SGFR : BinaryRRE<"sgfr", 0xB919, null_frag, GR64, GR32>;
1055 defm SGR : BinaryRREAndK<"sgr", 0xB909, 0xB9E9, z_ssub, GR64, GR64>;
1057 // Subtraction from a high register.
1058 def SHHHR : BinaryRRFa<"shhhr", 0xB9C9, null_frag, GRH32, GRH32, GRH32>,
1059 Requires<[FeatureHighWord]>;
1060 def SHHLR : BinaryRRFa<"shhlr", 0xB9D9, null_frag, GRH32, GRH32, GR32>,
1061 Requires<[FeatureHighWord]>;
1063 // Subtraction of memory.
1064 defm SH : BinaryRXPair<"sh", 0x4B, 0xE37B, z_ssub, GR32, asextloadi16, 2>;
1065 defm S : BinaryRXPairAndPseudo<"s", 0x5B, 0xE35B, z_ssub, GR32, load, 4>;
1066 def SGH : BinaryRXY<"sgh", 0xE339, z_ssub, GR64, asextloadi16, 2>,
1067 Requires<[FeatureMiscellaneousExtensions2]>;
1068 def SGF : BinaryRXY<"sgf", 0xE319, z_ssub, GR64, asextloadi32, 4>;
1069 defm SG : BinaryRXYAndPseudo<"sg", 0xE309, z_ssub, GR64, load, 8>;
1071 defm : SXB<z_ssub, GR64, SGFR>;
1073 // Subtracting an immediate is the same as adding the negated immediate.
1074 let AddedComplexity = 1 in {
1075 def : Pat<(z_ssub GR32:$src1, imm32sx16n:$src2),
1076 (AHIMux GR32:$src1, imm32sx16n:$src2)>,
1077 Requires<[FeatureHighWord]>;
1078 def : Pat<(z_ssub GR32:$src1, simm32n:$src2),
1079 (AFIMux GR32:$src1, simm32n:$src2)>,
1080 Requires<[FeatureHighWord]>;
1081 def : Pat<(z_ssub GR32:$src1, imm32sx16n:$src2),
1082 (AHI GR32:$src1, imm32sx16n:$src2)>;
1083 def : Pat<(z_ssub GR32:$src1, simm32n:$src2),
1084 (AFI GR32:$src1, simm32n:$src2)>;
1085 def : Pat<(z_ssub GR64:$src1, imm64sx16n:$src2),
1086 (AGHI GR64:$src1, imm64sx16n:$src2)>;
1087 def : Pat<(z_ssub GR64:$src1, imm64sx32n:$src2),
1088 (AGFI GR64:$src1, imm64sx32n:$src2)>;
1091 // And vice versa in one special case, where we need to load a
1092 // constant into a register in any case, but the negated constant
1093 // requires fewer instructions to load.
1094 def : Pat<(z_saddo GR64:$src1, imm64lh16n:$src2),
1095 (SGR GR64:$src1, (LLILH imm64lh16n:$src2))>;
1096 def : Pat<(z_saddo GR64:$src1, imm64lf32n:$src2),
1097 (SGR GR64:$src1, (LLILF imm64lf32n:$src2))>;
1099 // Subtraction producing a carry.
1100 let Defs = [CC], CCValues = 0x7, IsLogical = 1 in {
1101 // Subtraction of a register.
1102 defm SLR : BinaryRRAndK<"slr", 0x1F, 0xB9FB, z_usub, GR32, GR32>;
1103 def SLGFR : BinaryRRE<"slgfr", 0xB91B, null_frag, GR64, GR32>;
1104 defm SLGR : BinaryRREAndK<"slgr", 0xB90B, 0xB9EB, z_usub, GR64, GR64>;
1106 // Subtraction from a high register.
1107 def SLHHHR : BinaryRRFa<"slhhhr", 0xB9CB, null_frag, GRH32, GRH32, GRH32>,
1108 Requires<[FeatureHighWord]>;
1109 def SLHHLR : BinaryRRFa<"slhhlr", 0xB9DB, null_frag, GRH32, GRH32, GR32>,
1110 Requires<[FeatureHighWord]>;
1112 // Subtraction of unsigned 32-bit immediates.
1113 def SLFI : BinaryRIL<"slfi", 0xC25, z_usub, GR32, uimm32>;
1114 def SLGFI : BinaryRIL<"slgfi", 0xC24, z_usub, GR64, imm64zx32>;
1116 // Subtraction of memory.
1117 defm SL : BinaryRXPairAndPseudo<"sl", 0x5F, 0xE35F, z_usub, GR32, load, 4>;
1118 def SLGF : BinaryRXY<"slgf", 0xE31B, z_usub, GR64, azextloadi32, 4>;
1119 defm SLG : BinaryRXYAndPseudo<"slg", 0xE30B, z_usub, GR64, load, 8>;
1121 defm : ZXB<z_usub, GR64, SLGFR>;
1123 // Subtracting an immediate is the same as adding the negated immediate.
1124 let AddedComplexity = 1 in {
1125 def : Pat<(z_usub GR32:$src1, imm32sx16n:$src2),
1126 (ALHSIK GR32:$src1, imm32sx16n:$src2)>,
1127 Requires<[FeatureDistinctOps]>;
1128 def : Pat<(z_usub GR64:$src1, imm64sx16n:$src2),
1129 (ALGHSIK GR64:$src1, imm64sx16n:$src2)>,
1130 Requires<[FeatureDistinctOps]>;
1133 // And vice versa in one special case (but we prefer addition).
1134 def : Pat<(add GR64:$src1, imm64zx32n:$src2),
1135 (SLGFI GR64:$src1, imm64zx32n:$src2)>;
1137 // Subtraction producing and using a carry.
1138 let Defs = [CC], Uses = [CC], CCValues = 0xF, IsLogical = 1 in {
1139 // Subtraction of a register.
1140 def SLBR : BinaryRRE<"slbr", 0xB999, z_subcarry, GR32, GR32>;
1141 def SLBGR : BinaryRRE<"slbgr", 0xB989, z_subcarry, GR64, GR64>;
1143 // Subtraction of memory.
1144 def SLB : BinaryRXY<"slb", 0xE399, z_subcarry, GR32, load, 4>;
1145 def SLBG : BinaryRXY<"slbg", 0xE389, z_subcarry, GR64, load, 8>;
1149 //===----------------------------------------------------------------------===//
1151 //===----------------------------------------------------------------------===//
1153 let Defs = [CC] in {
1154 // ANDs of a register.
1155 let isCommutable = 1, CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1156 defm NR : BinaryRRAndK<"nr", 0x14, 0xB9F4, and, GR32, GR32>;
1157 defm NGR : BinaryRREAndK<"ngr", 0xB980, 0xB9E4, and, GR64, GR64>;
1160 let isConvertibleToThreeAddress = 1 in {
1161 // ANDs of a 16-bit immediate, leaving other bits unaffected.
1162 // The CC result only reflects the 16-bit field, not the full register.
1164 // NIxMux expands to NI[LH]x, depending on the choice of register.
1165 def NILMux : BinaryRIPseudo<and, GRX32, imm32ll16c>,
1166 Requires<[FeatureHighWord]>;
1167 def NIHMux : BinaryRIPseudo<and, GRX32, imm32lh16c>,
1168 Requires<[FeatureHighWord]>;
1169 def NILL : BinaryRI<"nill", 0xA57, and, GR32, imm32ll16c>;
1170 def NILH : BinaryRI<"nilh", 0xA56, and, GR32, imm32lh16c>;
1171 def NIHL : BinaryRI<"nihl", 0xA55, and, GRH32, imm32ll16c>;
1172 def NIHH : BinaryRI<"nihh", 0xA54, and, GRH32, imm32lh16c>;
1173 def NILL64 : BinaryAliasRI<and, GR64, imm64ll16c>;
1174 def NILH64 : BinaryAliasRI<and, GR64, imm64lh16c>;
1175 def NIHL64 : BinaryAliasRI<and, GR64, imm64hl16c>;
1176 def NIHH64 : BinaryAliasRI<and, GR64, imm64hh16c>;
1178 // ANDs of a 32-bit immediate, leaving other bits unaffected.
1179 // The CC result only reflects the 32-bit field, which means we can
1180 // use it as a zero indicator for i32 operations but not otherwise.
1181 let CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1182 // Expands to NILF or NIHF, depending on the choice of register.
1183 def NIFMux : BinaryRIPseudo<and, GRX32, uimm32>,
1184 Requires<[FeatureHighWord]>;
1185 def NILF : BinaryRIL<"nilf", 0xC0B, and, GR32, uimm32>;
1186 def NIHF : BinaryRIL<"nihf", 0xC0A, and, GRH32, uimm32>;
1188 def NILF64 : BinaryAliasRIL<and, GR64, imm64lf32c>;
1189 def NIHF64 : BinaryAliasRIL<and, GR64, imm64hf32c>;
1193 let CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1194 defm N : BinaryRXPairAndPseudo<"n", 0x54, 0xE354, and, GR32, load, 4>;
1195 defm NG : BinaryRXYAndPseudo<"ng", 0xE380, and, GR64, load, 8>;
1199 defm NI : BinarySIPair<"ni", 0x94, 0xEB54, null_frag, imm32zx8>;
1202 let mayLoad = 1, mayStore = 1 in
1203 defm NC : MemorySS<"nc", 0xD4, z_nc, z_nc_loop>;
1205 defm : RMWIByte<and, bdaddr12pair, NI>;
1206 defm : RMWIByte<and, bdaddr20pair, NIY>;
1208 //===----------------------------------------------------------------------===//
1210 //===----------------------------------------------------------------------===//
1212 let Defs = [CC] in {
1213 // ORs of a register.
1214 let isCommutable = 1, CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1215 defm OR : BinaryRRAndK<"or", 0x16, 0xB9F6, or, GR32, GR32>;
1216 defm OGR : BinaryRREAndK<"ogr", 0xB981, 0xB9E6, or, GR64, GR64>;
1219 // ORs of a 16-bit immediate, leaving other bits unaffected.
1220 // The CC result only reflects the 16-bit field, not the full register.
1222 // OIxMux expands to OI[LH]x, depending on the choice of register.
1223 def OILMux : BinaryRIPseudo<or, GRX32, imm32ll16>,
1224 Requires<[FeatureHighWord]>;
1225 def OIHMux : BinaryRIPseudo<or, GRX32, imm32lh16>,
1226 Requires<[FeatureHighWord]>;
1227 def OILL : BinaryRI<"oill", 0xA5B, or, GR32, imm32ll16>;
1228 def OILH : BinaryRI<"oilh", 0xA5A, or, GR32, imm32lh16>;
1229 def OIHL : BinaryRI<"oihl", 0xA59, or, GRH32, imm32ll16>;
1230 def OIHH : BinaryRI<"oihh", 0xA58, or, GRH32, imm32lh16>;
1231 def OILL64 : BinaryAliasRI<or, GR64, imm64ll16>;
1232 def OILH64 : BinaryAliasRI<or, GR64, imm64lh16>;
1233 def OIHL64 : BinaryAliasRI<or, GR64, imm64hl16>;
1234 def OIHH64 : BinaryAliasRI<or, GR64, imm64hh16>;
1236 // ORs of a 32-bit immediate, leaving other bits unaffected.
1237 // The CC result only reflects the 32-bit field, which means we can
1238 // use it as a zero indicator for i32 operations but not otherwise.
1239 let CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1240 // Expands to OILF or OIHF, depending on the choice of register.
1241 def OIFMux : BinaryRIPseudo<or, GRX32, uimm32>,
1242 Requires<[FeatureHighWord]>;
1243 def OILF : BinaryRIL<"oilf", 0xC0D, or, GR32, uimm32>;
1244 def OIHF : BinaryRIL<"oihf", 0xC0C, or, GRH32, uimm32>;
1246 def OILF64 : BinaryAliasRIL<or, GR64, imm64lf32>;
1247 def OIHF64 : BinaryAliasRIL<or, GR64, imm64hf32>;
1250 let CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1251 defm O : BinaryRXPairAndPseudo<"o", 0x56, 0xE356, or, GR32, load, 4>;
1252 defm OG : BinaryRXYAndPseudo<"og", 0xE381, or, GR64, load, 8>;
1256 defm OI : BinarySIPair<"oi", 0x96, 0xEB56, null_frag, imm32zx8>;
1259 let mayLoad = 1, mayStore = 1 in
1260 defm OC : MemorySS<"oc", 0xD6, z_oc, z_oc_loop>;
1262 defm : RMWIByte<or, bdaddr12pair, OI>;
1263 defm : RMWIByte<or, bdaddr20pair, OIY>;
1265 //===----------------------------------------------------------------------===//
1267 //===----------------------------------------------------------------------===//
1269 let Defs = [CC] in {
1270 // XORs of a register.
1271 let isCommutable = 1, CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1272 defm XR : BinaryRRAndK<"xr", 0x17, 0xB9F7, xor, GR32, GR32>;
1273 defm XGR : BinaryRREAndK<"xgr", 0xB982, 0xB9E7, xor, GR64, GR64>;
1276 // XORs of a 32-bit immediate, leaving other bits unaffected.
1277 // The CC result only reflects the 32-bit field, which means we can
1278 // use it as a zero indicator for i32 operations but not otherwise.
1279 let CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1280 // Expands to XILF or XIHF, depending on the choice of register.
1281 def XIFMux : BinaryRIPseudo<xor, GRX32, uimm32>,
1282 Requires<[FeatureHighWord]>;
1283 def XILF : BinaryRIL<"xilf", 0xC07, xor, GR32, uimm32>;
1284 def XIHF : BinaryRIL<"xihf", 0xC06, xor, GRH32, uimm32>;
1286 def XILF64 : BinaryAliasRIL<xor, GR64, imm64lf32>;
1287 def XIHF64 : BinaryAliasRIL<xor, GR64, imm64hf32>;
1290 let CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1291 defm X : BinaryRXPairAndPseudo<"x",0x57, 0xE357, xor, GR32, load, 4>;
1292 defm XG : BinaryRXYAndPseudo<"xg", 0xE382, xor, GR64, load, 8>;
1296 defm XI : BinarySIPair<"xi", 0x97, 0xEB57, null_frag, imm32zx8>;
1299 let mayLoad = 1, mayStore = 1 in
1300 defm XC : MemorySS<"xc", 0xD7, z_xc, z_xc_loop>;
1302 defm : RMWIByte<xor, bdaddr12pair, XI>;
1303 defm : RMWIByte<xor, bdaddr20pair, XIY>;
1305 //===----------------------------------------------------------------------===//
1306 // Combined logical operations
1307 //===----------------------------------------------------------------------===//
1309 let Predicates = [FeatureMiscellaneousExtensions3],
1311 // AND with complement.
1312 let CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1313 def NCRK : BinaryRRFa<"ncrk", 0xB9F5, andc, GR32, GR32, GR32>;
1314 def NCGRK : BinaryRRFa<"ncgrk", 0xB9E5, andc, GR64, GR64, GR64>;
1317 // OR with complement.
1318 let CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1319 def OCRK : BinaryRRFa<"ocrk", 0xB975, orc, GR32, GR32, GR32>;
1320 def OCGRK : BinaryRRFa<"ocgrk", 0xB965, orc, GR64, GR64, GR64>;
1324 let isCommutable = 1, CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1325 def NNRK : BinaryRRFa<"nnrk", 0xB974, nand, GR32, GR32, GR32>;
1326 def NNGRK : BinaryRRFa<"nngrk", 0xB964, nand, GR64, GR64, GR64>;
1330 let isCommutable = 1, CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1331 def NORK : BinaryRRFa<"nork", 0xB976, nor, GR32, GR32, GR32>;
1332 def NOGRK : BinaryRRFa<"nogrk", 0xB966, nor, GR64, GR64, GR64>;
1336 let isCommutable = 1, CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1337 def NXRK : BinaryRRFa<"nxrk", 0xB977, nxor, GR32, GR32, GR32>;
1338 def NXGRK : BinaryRRFa<"nxgrk", 0xB967, nxor, GR64, GR64, GR64>;
1342 //===----------------------------------------------------------------------===//
1344 //===----------------------------------------------------------------------===//
1346 // Multiplication of a register, setting the condition code. We prefer these
1347 // over MS(G)R if available, even though we cannot use the condition code,
1348 // since they are three-operand instructions.
1349 let Predicates = [FeatureMiscellaneousExtensions2],
1350 Defs = [CC], isCommutable = 1 in {
1351 def MSRKC : BinaryRRFa<"msrkc", 0xB9FD, mul, GR32, GR32, GR32>;
1352 def MSGRKC : BinaryRRFa<"msgrkc", 0xB9ED, mul, GR64, GR64, GR64>;
1355 // Multiplication of a register.
1356 let isCommutable = 1 in {
1357 def MSR : BinaryRRE<"msr", 0xB252, mul, GR32, GR32>;
1358 def MSGR : BinaryRRE<"msgr", 0xB90C, mul, GR64, GR64>;
1360 def MSGFR : BinaryRRE<"msgfr", 0xB91C, null_frag, GR64, GR32>;
1361 defm : SXB<mul, GR64, MSGFR>;
1363 // Multiplication of a signed 16-bit immediate.
1364 def MHI : BinaryRI<"mhi", 0xA7C, mul, GR32, imm32sx16>;
1365 def MGHI : BinaryRI<"mghi", 0xA7D, mul, GR64, imm64sx16>;
1367 // Multiplication of a signed 32-bit immediate.
1368 def MSFI : BinaryRIL<"msfi", 0xC21, mul, GR32, simm32>;
1369 def MSGFI : BinaryRIL<"msgfi", 0xC20, mul, GR64, imm64sx32>;
1371 // Multiplication of memory.
1372 defm MH : BinaryRXPair<"mh", 0x4C, 0xE37C, mul, GR32, asextloadi16, 2>;
1373 defm MS : BinaryRXPair<"ms", 0x71, 0xE351, mul, GR32, load, 4>;
1374 def MGH : BinaryRXY<"mgh", 0xE33C, mul, GR64, asextloadi16, 2>,
1375 Requires<[FeatureMiscellaneousExtensions2]>;
1376 def MSGF : BinaryRXY<"msgf", 0xE31C, mul, GR64, asextloadi32, 4>;
1377 def MSG : BinaryRXY<"msg", 0xE30C, mul, GR64, load, 8>;
1379 // Multiplication of memory, setting the condition code.
1380 let Predicates = [FeatureMiscellaneousExtensions2], Defs = [CC] in {
1381 defm MSC : BinaryRXYAndPseudo<"msc", 0xE353, null_frag, GR32, load, 4>;
1382 defm MSGC : BinaryRXYAndPseudo<"msgc", 0xE383, null_frag, GR64, load, 8>;
1385 // Multiplication of a register, producing two results.
1386 def MR : BinaryRR <"mr", 0x1C, null_frag, GR128, GR32>;
1387 def MGRK : BinaryRRFa<"mgrk", 0xB9EC, null_frag, GR128, GR64, GR64>,
1388 Requires<[FeatureMiscellaneousExtensions2]>;
1389 def MLR : BinaryRRE<"mlr", 0xB996, null_frag, GR128, GR32>;
1390 def MLGR : BinaryRRE<"mlgr", 0xB986, null_frag, GR128, GR64>;
1392 def : Pat<(z_smul_lohi GR64:$src1, GR64:$src2),
1393 (MGRK GR64:$src1, GR64:$src2)>;
1394 def : Pat<(z_umul_lohi GR64:$src1, GR64:$src2),
1395 (MLGR (AEXT128 GR64:$src1), GR64:$src2)>;
1397 // Multiplication of memory, producing two results.
1398 def M : BinaryRX <"m", 0x5C, null_frag, GR128, load, 4>;
1399 def MFY : BinaryRXY<"mfy", 0xE35C, null_frag, GR128, load, 4>;
1400 def MG : BinaryRXY<"mg", 0xE384, null_frag, GR128, load, 8>,
1401 Requires<[FeatureMiscellaneousExtensions2]>;
1402 def ML : BinaryRXY<"ml", 0xE396, null_frag, GR128, load, 4>;
1403 def MLG : BinaryRXY<"mlg", 0xE386, null_frag, GR128, load, 8>;
1405 def : Pat<(z_smul_lohi GR64:$src1, (i64 (load bdxaddr20only:$src2))),
1406 (MG (AEXT128 GR64:$src1), bdxaddr20only:$src2)>;
1407 def : Pat<(z_umul_lohi GR64:$src1, (i64 (load bdxaddr20only:$src2))),
1408 (MLG (AEXT128 GR64:$src1), bdxaddr20only:$src2)>;
1410 //===----------------------------------------------------------------------===//
1411 // Division and remainder
1412 //===----------------------------------------------------------------------===//
1414 let hasSideEffects = 1 in { // Do not speculatively execute.
1415 // Division and remainder, from registers.
1416 def DR : BinaryRR <"dr", 0x1D, null_frag, GR128, GR32>;
1417 def DSGFR : BinaryRRE<"dsgfr", 0xB91D, null_frag, GR128, GR32>;
1418 def DSGR : BinaryRRE<"dsgr", 0xB90D, null_frag, GR128, GR64>;
1419 def DLR : BinaryRRE<"dlr", 0xB997, null_frag, GR128, GR32>;
1420 def DLGR : BinaryRRE<"dlgr", 0xB987, null_frag, GR128, GR64>;
1422 // Division and remainder, from memory.
1423 def D : BinaryRX <"d", 0x5D, null_frag, GR128, load, 4>;
1424 def DSGF : BinaryRXY<"dsgf", 0xE31D, null_frag, GR128, load, 4>;
1425 def DSG : BinaryRXY<"dsg", 0xE30D, null_frag, GR128, load, 8>;
1426 def DL : BinaryRXY<"dl", 0xE397, null_frag, GR128, load, 4>;
1427 def DLG : BinaryRXY<"dlg", 0xE387, null_frag, GR128, load, 8>;
1429 def : Pat<(z_sdivrem GR64:$src1, GR32:$src2),
1430 (DSGFR (AEXT128 GR64:$src1), GR32:$src2)>;
1431 def : Pat<(z_sdivrem GR64:$src1, (i32 (load bdxaddr20only:$src2))),
1432 (DSGF (AEXT128 GR64:$src1), bdxaddr20only:$src2)>;
1433 def : Pat<(z_sdivrem GR64:$src1, GR64:$src2),
1434 (DSGR (AEXT128 GR64:$src1), GR64:$src2)>;
1435 def : Pat<(z_sdivrem GR64:$src1, (i64 (load bdxaddr20only:$src2))),
1436 (DSG (AEXT128 GR64:$src1), bdxaddr20only:$src2)>;
1438 def : Pat<(z_udivrem GR32:$src1, GR32:$src2),
1439 (DLR (ZEXT128 (INSERT_SUBREG (i64 (IMPLICIT_DEF)), GR32:$src1,
1440 subreg_l32)), GR32:$src2)>;
1441 def : Pat<(z_udivrem GR32:$src1, (i32 (load bdxaddr20only:$src2))),
1442 (DL (ZEXT128 (INSERT_SUBREG (i64 (IMPLICIT_DEF)), GR32:$src1,
1443 subreg_l32)), bdxaddr20only:$src2)>;
1444 def : Pat<(z_udivrem GR64:$src1, GR64:$src2),
1445 (DLGR (ZEXT128 GR64:$src1), GR64:$src2)>;
1446 def : Pat<(z_udivrem GR64:$src1, (i64 (load bdxaddr20only:$src2))),
1447 (DLG (ZEXT128 GR64:$src1), bdxaddr20only:$src2)>;
1449 //===----------------------------------------------------------------------===//
1451 //===----------------------------------------------------------------------===//
1453 // Logical shift left.
1454 defm SLL : BinaryRSAndK<"sll", 0x89, 0xEBDF, shiftop<shl>, GR32>;
1455 def SLLG : BinaryRSY<"sllg", 0xEB0D, shiftop<shl>, GR64>;
1456 def SLDL : BinaryRS<"sldl", 0x8D, null_frag, GR128>;
1458 // Arithmetic shift left.
1459 let Defs = [CC] in {
1460 defm SLA : BinaryRSAndK<"sla", 0x8B, 0xEBDD, null_frag, GR32>;
1461 def SLAG : BinaryRSY<"slag", 0xEB0B, null_frag, GR64>;
1462 def SLDA : BinaryRS<"slda", 0x8F, null_frag, GR128>;
1465 // Logical shift right.
1466 defm SRL : BinaryRSAndK<"srl", 0x88, 0xEBDE, shiftop<srl>, GR32>;
1467 def SRLG : BinaryRSY<"srlg", 0xEB0C, shiftop<srl>, GR64>;
1468 def SRDL : BinaryRS<"srdl", 0x8C, null_frag, GR128>;
1470 // Arithmetic shift right.
1471 let Defs = [CC], CCValues = 0xE, CompareZeroCCMask = 0xE in {
1472 defm SRA : BinaryRSAndK<"sra", 0x8A, 0xEBDC, shiftop<sra>, GR32>;
1473 def SRAG : BinaryRSY<"srag", 0xEB0A, shiftop<sra>, GR64>;
1474 def SRDA : BinaryRS<"srda", 0x8E, null_frag, GR128>;
1478 def RLL : BinaryRSY<"rll", 0xEB1D, shiftop<rotl>, GR32>;
1479 def RLLG : BinaryRSY<"rllg", 0xEB1C, shiftop<rotl>, GR64>;
1481 // Rotate second operand left and inserted selected bits into first operand.
1482 // These can act like 32-bit operands provided that the constant start and
1483 // end bits (operands 2 and 3) are in the range [32, 64).
1484 let Defs = [CC] in {
1485 let isCodeGenOnly = 1 in
1486 def RISBG32 : RotateSelectRIEf<"risbg", 0xEC55, GR32, GR32>;
1487 let CCValues = 0xE, CompareZeroCCMask = 0xE in
1488 def RISBG : RotateSelectRIEf<"risbg", 0xEC55, GR64, GR64>;
1491 // On zEC12 we have a variant of RISBG that does not set CC.
1492 let Predicates = [FeatureMiscellaneousExtensions] in
1493 def RISBGN : RotateSelectRIEf<"risbgn", 0xEC59, GR64, GR64>;
1495 // Forms of RISBG that only affect one word of the destination register.
1496 // They do not set CC.
1497 let Predicates = [FeatureHighWord] in {
1498 def RISBMux : RotateSelectRIEfPseudo<GRX32, GRX32>;
1499 def RISBLL : RotateSelectAliasRIEf<GR32, GR32>;
1500 def RISBLH : RotateSelectAliasRIEf<GR32, GRH32>;
1501 def RISBHL : RotateSelectAliasRIEf<GRH32, GR32>;
1502 def RISBHH : RotateSelectAliasRIEf<GRH32, GRH32>;
1503 def RISBLG : RotateSelectRIEf<"risblg", 0xEC51, GR32, GR64>;
1504 def RISBHG : RotateSelectRIEf<"risbhg", 0xEC5D, GRH32, GR64>;
1507 // Rotate second operand left and perform a logical operation with selected
1508 // bits of the first operand. The CC result only describes the selected bits,
1509 // so isn't useful for a full comparison against zero.
1510 let Defs = [CC] in {
1511 def RNSBG : RotateSelectRIEf<"rnsbg", 0xEC54, GR64, GR64>;
1512 def ROSBG : RotateSelectRIEf<"rosbg", 0xEC56, GR64, GR64>;
1513 def RXSBG : RotateSelectRIEf<"rxsbg", 0xEC57, GR64, GR64>;
1516 //===----------------------------------------------------------------------===//
1518 //===----------------------------------------------------------------------===//
1520 // Signed comparisons. We put these before the unsigned comparisons because
1521 // some of the signed forms have COMPARE AND BRANCH equivalents whereas none
1522 // of the unsigned forms do.
1523 let Defs = [CC], CCValues = 0xE in {
1524 // Comparison with a register.
1525 def CR : CompareRR <"cr", 0x19, z_scmp, GR32, GR32>;
1526 def CGFR : CompareRRE<"cgfr", 0xB930, null_frag, GR64, GR32>;
1527 def CGR : CompareRRE<"cgr", 0xB920, z_scmp, GR64, GR64>;
1529 // Comparison with a high register.
1530 def CHHR : CompareRRE<"chhr", 0xB9CD, null_frag, GRH32, GRH32>,
1531 Requires<[FeatureHighWord]>;
1532 def CHLR : CompareRRE<"chlr", 0xB9DD, null_frag, GRH32, GR32>,
1533 Requires<[FeatureHighWord]>;
1535 // Comparison with a signed 16-bit immediate. CHIMux expands to CHI or CIH,
1536 // depending on the choice of register.
1537 def CHIMux : CompareRIPseudo<z_scmp, GRX32, imm32sx16>,
1538 Requires<[FeatureHighWord]>;
1539 def CHI : CompareRI<"chi", 0xA7E, z_scmp, GR32, imm32sx16>;
1540 def CGHI : CompareRI<"cghi", 0xA7F, z_scmp, GR64, imm64sx16>;
1542 // Comparison with a signed 32-bit immediate. CFIMux expands to CFI or CIH,
1543 // depending on the choice of register.
1544 def CFIMux : CompareRIPseudo<z_scmp, GRX32, simm32>,
1545 Requires<[FeatureHighWord]>;
1546 def CFI : CompareRIL<"cfi", 0xC2D, z_scmp, GR32, simm32>;
1547 def CIH : CompareRIL<"cih", 0xCCD, z_scmp, GRH32, simm32>,
1548 Requires<[FeatureHighWord]>;
1549 def CGFI : CompareRIL<"cgfi", 0xC2C, z_scmp, GR64, imm64sx32>;
1551 // Comparison with memory.
1552 defm CH : CompareRXPair<"ch", 0x49, 0xE379, z_scmp, GR32, asextloadi16, 2>;
1553 def CMux : CompareRXYPseudo<z_scmp, GRX32, load, 4>,
1554 Requires<[FeatureHighWord]>;
1555 defm C : CompareRXPair<"c", 0x59, 0xE359, z_scmp, GR32, load, 4>;
1556 def CHF : CompareRXY<"chf", 0xE3CD, z_scmp, GRH32, load, 4>,
1557 Requires<[FeatureHighWord]>;
1558 def CGH : CompareRXY<"cgh", 0xE334, z_scmp, GR64, asextloadi16, 2>;
1559 def CGF : CompareRXY<"cgf", 0xE330, z_scmp, GR64, asextloadi32, 4>;
1560 def CG : CompareRXY<"cg", 0xE320, z_scmp, GR64, load, 8>;
1561 def CHRL : CompareRILPC<"chrl", 0xC65, z_scmp, GR32, aligned_asextloadi16>;
1562 def CRL : CompareRILPC<"crl", 0xC6D, z_scmp, GR32, aligned_load>;
1563 def CGHRL : CompareRILPC<"cghrl", 0xC64, z_scmp, GR64, aligned_asextloadi16>;
1564 def CGFRL : CompareRILPC<"cgfrl", 0xC6C, z_scmp, GR64, aligned_asextloadi32>;
1565 def CGRL : CompareRILPC<"cgrl", 0xC68, z_scmp, GR64, aligned_load>;
1567 // Comparison between memory and a signed 16-bit immediate.
1568 def CHHSI : CompareSIL<"chhsi", 0xE554, z_scmp, asextloadi16, imm32sx16>;
1569 def CHSI : CompareSIL<"chsi", 0xE55C, z_scmp, load, imm32sx16>;
1570 def CGHSI : CompareSIL<"cghsi", 0xE558, z_scmp, load, imm64sx16>;
1572 defm : SXB<z_scmp, GR64, CGFR>;
1574 // Unsigned comparisons.
1575 let Defs = [CC], CCValues = 0xE, IsLogical = 1 in {
1576 // Comparison with a register.
1577 def CLR : CompareRR <"clr", 0x15, z_ucmp, GR32, GR32>;
1578 def CLGFR : CompareRRE<"clgfr", 0xB931, null_frag, GR64, GR32>;
1579 def CLGR : CompareRRE<"clgr", 0xB921, z_ucmp, GR64, GR64>;
1581 // Comparison with a high register.
1582 def CLHHR : CompareRRE<"clhhr", 0xB9CF, null_frag, GRH32, GRH32>,
1583 Requires<[FeatureHighWord]>;
1584 def CLHLR : CompareRRE<"clhlr", 0xB9DF, null_frag, GRH32, GR32>,
1585 Requires<[FeatureHighWord]>;
1587 // Comparison with an unsigned 32-bit immediate. CLFIMux expands to CLFI
1588 // or CLIH, depending on the choice of register.
1589 def CLFIMux : CompareRIPseudo<z_ucmp, GRX32, uimm32>,
1590 Requires<[FeatureHighWord]>;
1591 def CLFI : CompareRIL<"clfi", 0xC2F, z_ucmp, GR32, uimm32>;
1592 def CLIH : CompareRIL<"clih", 0xCCF, z_ucmp, GRH32, uimm32>,
1593 Requires<[FeatureHighWord]>;
1594 def CLGFI : CompareRIL<"clgfi", 0xC2E, z_ucmp, GR64, imm64zx32>;
1596 // Comparison with memory.
1597 def CLMux : CompareRXYPseudo<z_ucmp, GRX32, load, 4>,
1598 Requires<[FeatureHighWord]>;
1599 defm CL : CompareRXPair<"cl", 0x55, 0xE355, z_ucmp, GR32, load, 4>;
1600 def CLHF : CompareRXY<"clhf", 0xE3CF, z_ucmp, GRH32, load, 4>,
1601 Requires<[FeatureHighWord]>;
1602 def CLGF : CompareRXY<"clgf", 0xE331, z_ucmp, GR64, azextloadi32, 4>;
1603 def CLG : CompareRXY<"clg", 0xE321, z_ucmp, GR64, load, 8>;
1604 def CLHRL : CompareRILPC<"clhrl", 0xC67, z_ucmp, GR32,
1605 aligned_azextloadi16>;
1606 def CLRL : CompareRILPC<"clrl", 0xC6F, z_ucmp, GR32,
1608 def CLGHRL : CompareRILPC<"clghrl", 0xC66, z_ucmp, GR64,
1609 aligned_azextloadi16>;
1610 def CLGFRL : CompareRILPC<"clgfrl", 0xC6E, z_ucmp, GR64,
1611 aligned_azextloadi32>;
1612 def CLGRL : CompareRILPC<"clgrl", 0xC6A, z_ucmp, GR64,
1615 // Comparison between memory and an unsigned 8-bit immediate.
1616 defm CLI : CompareSIPair<"cli", 0x95, 0xEB55, z_ucmp, azextloadi8, imm32zx8>;
1618 // Comparison between memory and an unsigned 16-bit immediate.
1619 def CLHHSI : CompareSIL<"clhhsi", 0xE555, z_ucmp, azextloadi16, imm32zx16>;
1620 def CLFHSI : CompareSIL<"clfhsi", 0xE55D, z_ucmp, load, imm32zx16>;
1621 def CLGHSI : CompareSIL<"clghsi", 0xE559, z_ucmp, load, imm64zx16>;
1623 defm : ZXB<z_ucmp, GR64, CLGFR>;
1625 // Memory-to-memory comparison.
1626 let mayLoad = 1, Defs = [CC] in {
1627 defm CLC : CompareMemorySS<"clc", 0xD5, z_clc, z_clc_loop>;
1628 def CLCL : SideEffectBinaryMemMemRR<"clcl", 0x0F, GR128, GR128>;
1629 def CLCLE : SideEffectTernaryMemMemRS<"clcle", 0xA9, GR128, GR128>;
1630 def CLCLU : SideEffectTernaryMemMemRSY<"clclu", 0xEB8F, GR128, GR128>;
1633 // String comparison.
1634 let mayLoad = 1, Defs = [CC] in
1635 defm CLST : StringRRE<"clst", 0xB25D, z_strcmp>;
1638 let Defs = [CC] in {
1639 // TMxMux expands to TM[LH]x, depending on the choice of register.
1640 def TMLMux : CompareRIPseudo<z_tm_reg, GRX32, imm32ll16>,
1641 Requires<[FeatureHighWord]>;
1642 def TMHMux : CompareRIPseudo<z_tm_reg, GRX32, imm32lh16>,
1643 Requires<[FeatureHighWord]>;
1644 def TMLL : CompareRI<"tmll", 0xA71, z_tm_reg, GR32, imm32ll16>;
1645 def TMLH : CompareRI<"tmlh", 0xA70, z_tm_reg, GR32, imm32lh16>;
1646 def TMHL : CompareRI<"tmhl", 0xA73, z_tm_reg, GRH32, imm32ll16>;
1647 def TMHH : CompareRI<"tmhh", 0xA72, z_tm_reg, GRH32, imm32lh16>;
1649 def TMLL64 : CompareAliasRI<z_tm_reg, GR64, imm64ll16>;
1650 def TMLH64 : CompareAliasRI<z_tm_reg, GR64, imm64lh16>;
1651 def TMHL64 : CompareAliasRI<z_tm_reg, GR64, imm64hl16>;
1652 def TMHH64 : CompareAliasRI<z_tm_reg, GR64, imm64hh16>;
1654 defm TM : CompareSIPair<"tm", 0x91, 0xEB51, z_tm_mem, anyextloadi8, imm32zx8>;
1657 def TML : InstAlias<"tml\t$R, $I", (TMLL GR32:$R, imm32ll16:$I), 0>;
1658 def TMH : InstAlias<"tmh\t$R, $I", (TMLH GR32:$R, imm32lh16:$I), 0>;
1660 // Compare logical characters under mask -- not (yet) used for codegen.
1661 let Defs = [CC] in {
1662 defm CLM : CompareRSPair<"clm", 0xBD, 0xEB21, GR32, 0>;
1663 def CLMH : CompareRSY<"clmh", 0xEB20, GRH32, 0>;
1666 //===----------------------------------------------------------------------===//
1667 // Prefetch and execution hint
1668 //===----------------------------------------------------------------------===//
1670 let mayLoad = 1, mayStore = 1 in {
1671 def PFD : PrefetchRXY<"pfd", 0xE336, z_prefetch>;
1672 def PFDRL : PrefetchRILPC<"pfdrl", 0xC62, z_prefetch>;
1675 let Predicates = [FeatureExecutionHint], hasSideEffects = 1 in {
1676 // Branch Prediction Preload
1677 def BPP : BranchPreloadSMI<"bpp", 0xC7>;
1678 def BPRP : BranchPreloadMII<"bprp", 0xC5>;
1680 // Next Instruction Access Intent
1681 def NIAI : SideEffectBinaryIE<"niai", 0xB2FA, imm32zx4, imm32zx4>;
1684 //===----------------------------------------------------------------------===//
1685 // Atomic operations
1686 //===----------------------------------------------------------------------===//
1688 // A serialization instruction that acts as a barrier for all memory
1689 // accesses, which expands to "bcr 14, 0".
1690 let hasSideEffects = 1 in
1691 def Serialize : Alias<2, (outs), (ins), []>;
1693 // A pseudo instruction that serves as a compiler barrier.
1694 let hasSideEffects = 1, hasNoSchedulingInfo = 1 in
1695 def MemBarrier : Pseudo<(outs), (ins), [(z_membarrier)]>;
1697 let Predicates = [FeatureInterlockedAccess1], Defs = [CC] in {
1698 def LAA : LoadAndOpRSY<"laa", 0xEBF8, atomic_load_add_32, GR32>;
1699 def LAAG : LoadAndOpRSY<"laag", 0xEBE8, atomic_load_add_64, GR64>;
1700 def LAAL : LoadAndOpRSY<"laal", 0xEBFA, null_frag, GR32>;
1701 def LAALG : LoadAndOpRSY<"laalg", 0xEBEA, null_frag, GR64>;
1702 def LAN : LoadAndOpRSY<"lan", 0xEBF4, atomic_load_and_32, GR32>;
1703 def LANG : LoadAndOpRSY<"lang", 0xEBE4, atomic_load_and_64, GR64>;
1704 def LAO : LoadAndOpRSY<"lao", 0xEBF6, atomic_load_or_32, GR32>;
1705 def LAOG : LoadAndOpRSY<"laog", 0xEBE6, atomic_load_or_64, GR64>;
1706 def LAX : LoadAndOpRSY<"lax", 0xEBF7, atomic_load_xor_32, GR32>;
1707 def LAXG : LoadAndOpRSY<"laxg", 0xEBE7, atomic_load_xor_64, GR64>;
1710 def ATOMIC_SWAPW : AtomicLoadWBinaryReg<z_atomic_swapw>;
1711 def ATOMIC_SWAP_32 : AtomicLoadBinaryReg32<atomic_swap_32>;
1712 def ATOMIC_SWAP_64 : AtomicLoadBinaryReg64<atomic_swap_64>;
1714 def ATOMIC_LOADW_AR : AtomicLoadWBinaryReg<z_atomic_loadw_add>;
1715 def ATOMIC_LOADW_AFI : AtomicLoadWBinaryImm<z_atomic_loadw_add, simm32>;
1716 let Predicates = [FeatureNoInterlockedAccess1] in {
1717 def ATOMIC_LOAD_AR : AtomicLoadBinaryReg32<atomic_load_add_32>;
1718 def ATOMIC_LOAD_AHI : AtomicLoadBinaryImm32<atomic_load_add_32, imm32sx16>;
1719 def ATOMIC_LOAD_AFI : AtomicLoadBinaryImm32<atomic_load_add_32, simm32>;
1720 def ATOMIC_LOAD_AGR : AtomicLoadBinaryReg64<atomic_load_add_64>;
1721 def ATOMIC_LOAD_AGHI : AtomicLoadBinaryImm64<atomic_load_add_64, imm64sx16>;
1722 def ATOMIC_LOAD_AGFI : AtomicLoadBinaryImm64<atomic_load_add_64, imm64sx32>;
1725 def ATOMIC_LOADW_SR : AtomicLoadWBinaryReg<z_atomic_loadw_sub>;
1726 def ATOMIC_LOAD_SR : AtomicLoadBinaryReg32<atomic_load_sub_32>;
1727 def ATOMIC_LOAD_SGR : AtomicLoadBinaryReg64<atomic_load_sub_64>;
1729 def ATOMIC_LOADW_NR : AtomicLoadWBinaryReg<z_atomic_loadw_and>;
1730 def ATOMIC_LOADW_NILH : AtomicLoadWBinaryImm<z_atomic_loadw_and, imm32lh16c>;
1731 let Predicates = [FeatureNoInterlockedAccess1] in {
1732 def ATOMIC_LOAD_NR : AtomicLoadBinaryReg32<atomic_load_and_32>;
1733 def ATOMIC_LOAD_NILL : AtomicLoadBinaryImm32<atomic_load_and_32,
1735 def ATOMIC_LOAD_NILH : AtomicLoadBinaryImm32<atomic_load_and_32,
1737 def ATOMIC_LOAD_NILF : AtomicLoadBinaryImm32<atomic_load_and_32, uimm32>;
1738 def ATOMIC_LOAD_NGR : AtomicLoadBinaryReg64<atomic_load_and_64>;
1739 def ATOMIC_LOAD_NILL64 : AtomicLoadBinaryImm64<atomic_load_and_64,
1741 def ATOMIC_LOAD_NILH64 : AtomicLoadBinaryImm64<atomic_load_and_64,
1743 def ATOMIC_LOAD_NIHL64 : AtomicLoadBinaryImm64<atomic_load_and_64,
1745 def ATOMIC_LOAD_NIHH64 : AtomicLoadBinaryImm64<atomic_load_and_64,
1747 def ATOMIC_LOAD_NILF64 : AtomicLoadBinaryImm64<atomic_load_and_64,
1749 def ATOMIC_LOAD_NIHF64 : AtomicLoadBinaryImm64<atomic_load_and_64,
1753 def ATOMIC_LOADW_OR : AtomicLoadWBinaryReg<z_atomic_loadw_or>;
1754 def ATOMIC_LOADW_OILH : AtomicLoadWBinaryImm<z_atomic_loadw_or, imm32lh16>;
1755 let Predicates = [FeatureNoInterlockedAccess1] in {
1756 def ATOMIC_LOAD_OR : AtomicLoadBinaryReg32<atomic_load_or_32>;
1757 def ATOMIC_LOAD_OILL : AtomicLoadBinaryImm32<atomic_load_or_32, imm32ll16>;
1758 def ATOMIC_LOAD_OILH : AtomicLoadBinaryImm32<atomic_load_or_32, imm32lh16>;
1759 def ATOMIC_LOAD_OILF : AtomicLoadBinaryImm32<atomic_load_or_32, uimm32>;
1760 def ATOMIC_LOAD_OGR : AtomicLoadBinaryReg64<atomic_load_or_64>;
1761 def ATOMIC_LOAD_OILL64 : AtomicLoadBinaryImm64<atomic_load_or_64, imm64ll16>;
1762 def ATOMIC_LOAD_OILH64 : AtomicLoadBinaryImm64<atomic_load_or_64, imm64lh16>;
1763 def ATOMIC_LOAD_OIHL64 : AtomicLoadBinaryImm64<atomic_load_or_64, imm64hl16>;
1764 def ATOMIC_LOAD_OIHH64 : AtomicLoadBinaryImm64<atomic_load_or_64, imm64hh16>;
1765 def ATOMIC_LOAD_OILF64 : AtomicLoadBinaryImm64<atomic_load_or_64, imm64lf32>;
1766 def ATOMIC_LOAD_OIHF64 : AtomicLoadBinaryImm64<atomic_load_or_64, imm64hf32>;
1769 def ATOMIC_LOADW_XR : AtomicLoadWBinaryReg<z_atomic_loadw_xor>;
1770 def ATOMIC_LOADW_XILF : AtomicLoadWBinaryImm<z_atomic_loadw_xor, uimm32>;
1771 let Predicates = [FeatureNoInterlockedAccess1] in {
1772 def ATOMIC_LOAD_XR : AtomicLoadBinaryReg32<atomic_load_xor_32>;
1773 def ATOMIC_LOAD_XILF : AtomicLoadBinaryImm32<atomic_load_xor_32, uimm32>;
1774 def ATOMIC_LOAD_XGR : AtomicLoadBinaryReg64<atomic_load_xor_64>;
1775 def ATOMIC_LOAD_XILF64 : AtomicLoadBinaryImm64<atomic_load_xor_64, imm64lf32>;
1776 def ATOMIC_LOAD_XIHF64 : AtomicLoadBinaryImm64<atomic_load_xor_64, imm64hf32>;
1779 def ATOMIC_LOADW_NRi : AtomicLoadWBinaryReg<z_atomic_loadw_nand>;
1780 def ATOMIC_LOADW_NILHi : AtomicLoadWBinaryImm<z_atomic_loadw_nand,
1782 def ATOMIC_LOAD_NRi : AtomicLoadBinaryReg32<atomic_load_nand_32>;
1783 def ATOMIC_LOAD_NILLi : AtomicLoadBinaryImm32<atomic_load_nand_32,
1785 def ATOMIC_LOAD_NILHi : AtomicLoadBinaryImm32<atomic_load_nand_32,
1787 def ATOMIC_LOAD_NILFi : AtomicLoadBinaryImm32<atomic_load_nand_32, uimm32>;
1788 def ATOMIC_LOAD_NGRi : AtomicLoadBinaryReg64<atomic_load_nand_64>;
1789 def ATOMIC_LOAD_NILL64i : AtomicLoadBinaryImm64<atomic_load_nand_64,
1791 def ATOMIC_LOAD_NILH64i : AtomicLoadBinaryImm64<atomic_load_nand_64,
1793 def ATOMIC_LOAD_NIHL64i : AtomicLoadBinaryImm64<atomic_load_nand_64,
1795 def ATOMIC_LOAD_NIHH64i : AtomicLoadBinaryImm64<atomic_load_nand_64,
1797 def ATOMIC_LOAD_NILF64i : AtomicLoadBinaryImm64<atomic_load_nand_64,
1799 def ATOMIC_LOAD_NIHF64i : AtomicLoadBinaryImm64<atomic_load_nand_64,
1802 def ATOMIC_LOADW_MIN : AtomicLoadWBinaryReg<z_atomic_loadw_min>;
1803 def ATOMIC_LOAD_MIN_32 : AtomicLoadBinaryReg32<atomic_load_min_32>;
1804 def ATOMIC_LOAD_MIN_64 : AtomicLoadBinaryReg64<atomic_load_min_64>;
1806 def ATOMIC_LOADW_MAX : AtomicLoadWBinaryReg<z_atomic_loadw_max>;
1807 def ATOMIC_LOAD_MAX_32 : AtomicLoadBinaryReg32<atomic_load_max_32>;
1808 def ATOMIC_LOAD_MAX_64 : AtomicLoadBinaryReg64<atomic_load_max_64>;
1810 def ATOMIC_LOADW_UMIN : AtomicLoadWBinaryReg<z_atomic_loadw_umin>;
1811 def ATOMIC_LOAD_UMIN_32 : AtomicLoadBinaryReg32<atomic_load_umin_32>;
1812 def ATOMIC_LOAD_UMIN_64 : AtomicLoadBinaryReg64<atomic_load_umin_64>;
1814 def ATOMIC_LOADW_UMAX : AtomicLoadWBinaryReg<z_atomic_loadw_umax>;
1815 def ATOMIC_LOAD_UMAX_32 : AtomicLoadBinaryReg32<atomic_load_umax_32>;
1816 def ATOMIC_LOAD_UMAX_64 : AtomicLoadBinaryReg64<atomic_load_umax_64>;
1818 def ATOMIC_CMP_SWAPW
1819 : Pseudo<(outs GR32:$dst), (ins bdaddr20only:$addr, GR32:$cmp, GR32:$swap,
1820 ADDR32:$bitshift, ADDR32:$negbitshift,
1823 (z_atomic_cmp_swapw bdaddr20only:$addr, GR32:$cmp, GR32:$swap,
1824 ADDR32:$bitshift, ADDR32:$negbitshift,
1825 uimm32:$bitsize))]> {
1829 let usesCustomInserter = 1;
1830 let hasNoSchedulingInfo = 1;
1834 let mayLoad = 1, Defs = [CC] in
1835 def TS : StoreInherentS<"ts", 0x9300, null_frag, 1>;
1837 // Compare and swap.
1838 let Defs = [CC] in {
1839 defm CS : CmpSwapRSPair<"cs", 0xBA, 0xEB14, z_atomic_cmp_swap, GR32>;
1840 def CSG : CmpSwapRSY<"csg", 0xEB30, z_atomic_cmp_swap, GR64>;
1843 // Compare double and swap.
1844 let Defs = [CC] in {
1845 defm CDS : CmpSwapRSPair<"cds", 0xBB, 0xEB31, null_frag, GR128>;
1846 def CDSG : CmpSwapRSY<"cdsg", 0xEB3E, z_atomic_cmp_swap_128, GR128>;
1849 // Compare and swap and store.
1850 let Uses = [R0L, R1D], Defs = [CC], mayStore = 1, mayLoad = 1 in
1851 def CSST : SideEffectTernarySSF<"csst", 0xC82, GR64>;
1853 // Perform locked operation.
1854 let Uses = [R0L, R1D], Defs = [CC], mayStore = 1, mayLoad =1 in
1855 def PLO : SideEffectQuaternarySSe<"plo", 0xEE, GR64>;
1857 // Load/store pair from/to quadword.
1858 def LPQ : UnaryRXY<"lpq", 0xE38F, z_atomic_load_128, GR128, 16>;
1859 def STPQ : StoreRXY<"stpq", 0xE38E, z_atomic_store_128, GR128, 16>;
1861 // Load pair disjoint.
1862 let Predicates = [FeatureInterlockedAccess1], Defs = [CC] in {
1863 def LPD : BinarySSF<"lpd", 0xC84, GR128>;
1864 def LPDG : BinarySSF<"lpdg", 0xC85, GR128>;
1867 //===----------------------------------------------------------------------===//
1868 // Translate and convert
1869 //===----------------------------------------------------------------------===//
1871 let mayLoad = 1, mayStore = 1 in
1872 def TR : SideEffectBinarySSa<"tr", 0xDC>;
1874 let mayLoad = 1, Defs = [CC, R0L, R1D] in {
1875 def TRT : SideEffectBinarySSa<"trt", 0xDD>;
1876 def TRTR : SideEffectBinarySSa<"trtr", 0xD0>;
1879 let mayLoad = 1, mayStore = 1, Uses = [R0L] in
1880 def TRE : SideEffectBinaryMemMemRRE<"tre", 0xB2A5, GR128, GR64>;
1882 let mayLoad = 1, Uses = [R1D], Defs = [CC] in {
1883 defm TRTE : BinaryMemRRFcOpt<"trte", 0xB9BF, GR128, GR64>;
1884 defm TRTRE : BinaryMemRRFcOpt<"trtre", 0xB9BD, GR128, GR64>;
1887 let mayLoad = 1, mayStore = 1, Uses = [R0L, R1D], Defs = [CC] in {
1888 defm TROO : SideEffectTernaryMemMemRRFcOpt<"troo", 0xB993, GR128, GR64>;
1889 defm TROT : SideEffectTernaryMemMemRRFcOpt<"trot", 0xB992, GR128, GR64>;
1890 defm TRTO : SideEffectTernaryMemMemRRFcOpt<"trto", 0xB991, GR128, GR64>;
1891 defm TRTT : SideEffectTernaryMemMemRRFcOpt<"trtt", 0xB990, GR128, GR64>;
1894 let mayLoad = 1, mayStore = 1, Defs = [CC] in {
1895 defm CU12 : SideEffectTernaryMemMemRRFcOpt<"cu12", 0xB2A7, GR128, GR128>;
1896 defm CU14 : SideEffectTernaryMemMemRRFcOpt<"cu14", 0xB9B0, GR128, GR128>;
1897 defm CU21 : SideEffectTernaryMemMemRRFcOpt<"cu21", 0xB2A6, GR128, GR128>;
1898 defm CU24 : SideEffectTernaryMemMemRRFcOpt<"cu24", 0xB9B1, GR128, GR128>;
1899 def CU41 : SideEffectBinaryMemMemRRE<"cu41", 0xB9B2, GR128, GR128>;
1900 def CU42 : SideEffectBinaryMemMemRRE<"cu42", 0xB9B3, GR128, GR128>;
1902 let isAsmParserOnly = 1 in {
1903 defm CUUTF : SideEffectTernaryMemMemRRFcOpt<"cuutf", 0xB2A6, GR128, GR128>;
1904 defm CUTFU : SideEffectTernaryMemMemRRFcOpt<"cutfu", 0xB2A7, GR128, GR128>;
1908 //===----------------------------------------------------------------------===//
1909 // Message-security assist
1910 //===----------------------------------------------------------------------===//
1912 let mayLoad = 1, mayStore = 1, Uses = [R0L, R1D], Defs = [CC] in {
1913 def KM : SideEffectBinaryMemMemRRE<"km", 0xB92E, GR128, GR128>;
1914 def KMC : SideEffectBinaryMemMemRRE<"kmc", 0xB92F, GR128, GR128>;
1916 def KIMD : SideEffectBinaryMemRRE<"kimd", 0xB93E, GR64, GR128>;
1917 def KLMD : SideEffectBinaryMemRRE<"klmd", 0xB93F, GR64, GR128>;
1918 def KMAC : SideEffectBinaryMemRRE<"kmac", 0xB91E, GR64, GR128>;
1920 let Predicates = [FeatureMessageSecurityAssist4] in {
1921 def KMF : SideEffectBinaryMemMemRRE<"kmf", 0xB92A, GR128, GR128>;
1922 def KMO : SideEffectBinaryMemMemRRE<"kmo", 0xB92B, GR128, GR128>;
1923 def KMCTR : SideEffectTernaryMemMemMemRRFb<"kmctr", 0xB92D,
1924 GR128, GR128, GR128>;
1925 def PCC : SideEffectInherentRRE<"pcc", 0xB92C>;
1928 let Predicates = [FeatureMessageSecurityAssist5] in
1929 def PPNO : SideEffectBinaryMemMemRRE<"ppno", 0xB93C, GR128, GR128>;
1930 let Predicates = [FeatureMessageSecurityAssist7], isAsmParserOnly = 1 in
1931 def PRNO : SideEffectBinaryMemMemRRE<"prno", 0xB93C, GR128, GR128>;
1933 let Predicates = [FeatureMessageSecurityAssist8] in
1934 def KMA : SideEffectTernaryMemMemMemRRFb<"kma", 0xB929,
1935 GR128, GR128, GR128>;
1937 let Predicates = [FeatureMessageSecurityAssist9] in
1938 def KDSA : SideEffectBinaryMemRRE<"kdsa", 0xB93A, GR64, GR128>;
1941 //===----------------------------------------------------------------------===//
1943 //===----------------------------------------------------------------------===//
1945 // These instructions use and/or modify the guarded storage control
1946 // registers, which we do not otherwise model, so they should have
1948 let Predicates = [FeatureGuardedStorage], hasSideEffects = 1 in {
1949 def LGG : UnaryRXY<"lgg", 0xE34C, null_frag, GR64, 8>;
1950 def LLGFSG : UnaryRXY<"llgfsg", 0xE348, null_frag, GR64, 4>;
1953 def LGSC : SideEffectBinaryRXY<"lgsc", 0xE34D, GR64>;
1955 def STGSC : SideEffectBinaryRXY<"stgsc", 0xE349, GR64>;
1958 //===----------------------------------------------------------------------===//
1959 // Decimal arithmetic
1960 //===----------------------------------------------------------------------===//
1962 defm CVB : BinaryRXPair<"cvb",0x4F, 0xE306, null_frag, GR32, load, 4>;
1963 def CVBG : BinaryRXY<"cvbg", 0xE30E, null_frag, GR64, load, 8>;
1965 defm CVD : StoreRXPair<"cvd", 0x4E, 0xE326, null_frag, GR32, 4>;
1966 def CVDG : StoreRXY<"cvdg", 0xE32E, null_frag, GR64, 8>;
1968 let mayLoad = 1, mayStore = 1 in {
1969 def MVN : SideEffectBinarySSa<"mvn", 0xD1>;
1970 def MVZ : SideEffectBinarySSa<"mvz", 0xD3>;
1971 def MVO : SideEffectBinarySSb<"mvo", 0xF1>;
1973 def PACK : SideEffectBinarySSb<"pack", 0xF2>;
1974 def PKA : SideEffectBinarySSf<"pka", 0xE9>;
1975 def PKU : SideEffectBinarySSf<"pku", 0xE1>;
1976 def UNPK : SideEffectBinarySSb<"unpk", 0xF3>;
1977 let Defs = [CC] in {
1978 def UNPKA : SideEffectBinarySSa<"unpka", 0xEA>;
1979 def UNPKU : SideEffectBinarySSa<"unpku", 0xE2>;
1983 let mayLoad = 1, mayStore = 1 in {
1984 let Defs = [CC] in {
1985 def AP : SideEffectBinarySSb<"ap", 0xFA>;
1986 def SP : SideEffectBinarySSb<"sp", 0xFB>;
1987 def ZAP : SideEffectBinarySSb<"zap", 0xF8>;
1988 def SRP : SideEffectTernarySSc<"srp", 0xF0>;
1990 def MP : SideEffectBinarySSb<"mp", 0xFC>;
1991 def DP : SideEffectBinarySSb<"dp", 0xFD>;
1992 let Defs = [CC] in {
1993 def ED : SideEffectBinarySSa<"ed", 0xDE>;
1994 def EDMK : SideEffectBinarySSa<"edmk", 0xDF>;
1998 let Defs = [CC] in {
1999 def CP : CompareSSb<"cp", 0xF9>;
2000 def TP : TestRSL<"tp", 0xEBC0>;
2003 //===----------------------------------------------------------------------===//
2005 //===----------------------------------------------------------------------===//
2007 // Read a 32-bit access register into a GR32. As with all GR32 operations,
2008 // the upper 32 bits of the enclosing GR64 remain unchanged, which is useful
2009 // when a 64-bit address is stored in a pair of access registers.
2010 def EAR : UnaryRRE<"ear", 0xB24F, null_frag, GR32, AR32>;
2012 // Set access register.
2013 def SAR : UnaryRRE<"sar", 0xB24E, null_frag, AR32, GR32>;
2015 // Copy access register.
2016 def CPYA : UnaryRRE<"cpya", 0xB24D, null_frag, AR32, AR32>;
2018 // Load address extended.
2019 defm LAE : LoadAddressRXPair<"lae", 0x51, 0xE375, null_frag>;
2021 // Load access multiple.
2022 defm LAM : LoadMultipleRSPair<"lam", 0x9A, 0xEB9A, AR32>;
2024 // Store access multiple.
2025 defm STAM : StoreMultipleRSPair<"stam", 0x9B, 0xEB9B, AR32>;
2027 //===----------------------------------------------------------------------===//
2028 // Program mask and addressing mode
2029 //===----------------------------------------------------------------------===//
2031 // Extract CC and program mask into a register. CC ends up in bits 29 and 28.
2033 def IPM : InherentRRE<"ipm", 0xB222, GR32, z_ipm>;
2035 // Set CC and program mask from a register.
2036 let hasSideEffects = 1, Defs = [CC] in
2037 def SPM : SideEffectUnaryRR<"spm", 0x04, GR32>;
2039 // Branch and link - like BAS, but also extracts CC and program mask.
2040 let isCall = 1, Uses = [CC], Defs = [CC] in {
2041 def BAL : CallRX<"bal", 0x45>;
2042 def BALR : CallRR<"balr", 0x05>;
2045 // Test addressing mode.
2047 def TAM : SideEffectInherentE<"tam", 0x010B>;
2049 // Set addressing mode.
2050 let hasSideEffects = 1 in {
2051 def SAM24 : SideEffectInherentE<"sam24", 0x010C>;
2052 def SAM31 : SideEffectInherentE<"sam31", 0x010D>;
2053 def SAM64 : SideEffectInherentE<"sam64", 0x010E>;
2056 // Branch and set mode. Not really a call, but also sets an output register.
2057 let isBranch = 1, isTerminator = 1, isBarrier = 1 in
2058 def BSM : CallRR<"bsm", 0x0B>;
2060 // Branch and save and set mode.
2061 let isCall = 1, Defs = [CC] in
2062 def BASSM : CallRR<"bassm", 0x0C>;
2064 //===----------------------------------------------------------------------===//
2065 // Transactional execution
2066 //===----------------------------------------------------------------------===//
2068 let hasSideEffects = 1, Predicates = [FeatureTransactionalExecution] in {
2069 // Transaction Begin
2070 let mayStore = 1, usesCustomInserter = 1, Defs = [CC] in {
2071 def TBEGIN : TestBinarySIL<"tbegin", 0xE560, z_tbegin, imm32zx16>;
2072 let hasNoSchedulingInfo = 1 in
2073 def TBEGIN_nofloat : TestBinarySILPseudo<z_tbegin_nofloat, imm32zx16>;
2074 def TBEGINC : SideEffectBinarySIL<"tbeginc", 0xE561,
2075 int_s390_tbeginc, imm32zx16>;
2080 def TEND : TestInherentS<"tend", 0xB2F8, z_tend>;
2082 // Transaction Abort
2083 let isTerminator = 1, isBarrier = 1, mayStore = 1,
2084 hasSideEffects = 1 in
2085 def TABORT : SideEffectAddressS<"tabort", 0xB2FC, int_s390_tabort>;
2087 // Nontransactional Store
2088 def NTSTG : StoreRXY<"ntstg", 0xE325, int_s390_ntstg, GR64, 8>;
2090 // Extract Transaction Nesting Depth
2091 def ETND : InherentRRE<"etnd", 0xB2EC, GR32, int_s390_etnd>;
2094 //===----------------------------------------------------------------------===//
2096 //===----------------------------------------------------------------------===//
2098 let Predicates = [FeatureProcessorAssist] in {
2099 let hasSideEffects = 1 in
2100 def PPA : SideEffectTernaryRRFc<"ppa", 0xB2E8, GR64, GR64, imm32zx4>;
2101 def : Pat<(int_s390_ppa_txassist GR32:$src),
2102 (PPA (INSERT_SUBREG (i64 (IMPLICIT_DEF)), GR32:$src, subreg_l32),
2106 //===----------------------------------------------------------------------===//
2107 // Miscellaneous Instructions.
2108 //===----------------------------------------------------------------------===//
2110 // Find leftmost one, AKA count leading zeros. The instruction actually
2111 // returns a pair of GR64s, the first giving the number of leading zeros
2112 // and the second giving a copy of the source with the leftmost one bit
2113 // cleared. We only use the first result here.
2115 def FLOGR : UnaryRRE<"flogr", 0xB983, null_frag, GR128, GR64>;
2116 def : Pat<(i64 (ctlz GR64:$src)),
2117 (EXTRACT_SUBREG (FLOGR GR64:$src), subreg_h64)>;
2119 // Population count. Counts bits set per byte or doubleword.
2120 let Predicates = [FeatureMiscellaneousExtensions3] in {
2122 def POPCNTOpt : BinaryRRFc<"popcnt", 0xB9E1, GR64, GR64>;
2123 def : Pat<(ctpop GR64:$src), (POPCNTOpt GR64:$src, 8)>;
2125 let Predicates = [FeaturePopulationCount], Defs = [CC] in
2126 def POPCNT : UnaryRRE<"popcnt", 0xB9E1, z_popcnt, GR64, GR64>;
2128 // Search a block of memory for a character.
2129 let mayLoad = 1, Defs = [CC] in
2130 defm SRST : StringRRE<"srst", 0xB25E, z_search_string>;
2131 let mayLoad = 1, Defs = [CC], Uses = [R0L] in
2132 def SRSTU : SideEffectBinaryMemMemRRE<"srstu", 0xB9BE, GR64, GR64>;
2134 // Compare until substring equal.
2135 let mayLoad = 1, Defs = [CC], Uses = [R0L, R1L] in
2136 def CUSE : SideEffectBinaryMemMemRRE<"cuse", 0xB257, GR128, GR128>;
2138 // Compare and form codeword.
2139 let mayLoad = 1, Defs = [CC, R1D, R2D, R3D], Uses = [R1D, R2D, R3D] in
2140 def CFC : SideEffectAddressS<"cfc", 0xB21A, null_frag>;
2143 let mayLoad = 1, mayStore = 1, Defs = [CC, R0D, R1D, R2D, R3D, R5D],
2144 Uses = [R0D, R1D, R2D, R3D, R4D, R5D] in
2145 def UPT : SideEffectInherentE<"upt", 0x0102>;
2148 let mayLoad = 1, Defs = [CC] in
2149 def CKSM : SideEffectBinaryMemMemRRE<"cksm", 0xB241, GR64, GR128>;
2151 // Compression call.
2152 let mayLoad = 1, mayStore = 1, Defs = [CC, R1D], Uses = [R0L, R1D] in
2153 def CMPSC : SideEffectBinaryMemMemRRE<"cmpsc", 0xB263, GR128, GR128>;
2156 let Predicates = [FeatureEnhancedSort],
2157 mayLoad = 1, mayStore = 1, Defs = [CC], Uses = [R0L, R1D] in
2158 def SORTL : SideEffectBinaryMemMemRRE<"sortl", 0xB938, GR128, GR128>;
2160 // Deflate conversion call.
2161 let Predicates = [FeatureDeflateConversion],
2162 mayLoad = 1, mayStore = 1, Defs = [CC], Uses = [R0L, R1D] in
2163 def DFLTCC : SideEffectTernaryMemMemRRFa<"dfltcc", 0xB939,
2164 GR128, GR128, GR64>;
2167 let Predicates = [FeatureNNPAssist],
2168 mayLoad = 1, mayStore = 1, Defs = [R0D, CC], Uses = [R0D, R1D] in
2169 def NNPA : SideEffectInherentRRE<"nnpa", 0xB93B>;
2172 let hasSideEffects = 1 in {
2173 def EX : SideEffectBinaryRX<"ex", 0x44, ADDR64>;
2174 def EXRL : SideEffectBinaryRILPC<"exrl", 0xC60, ADDR64>;
2175 let hasNoSchedulingInfo = 1 in
2176 def EXRL_Pseudo : Pseudo<(outs), (ins i64imm:$TargetOpc, ADDR64:$lenMinus1,
2177 bdaddr12only:$bdl1, bdaddr12only:$bd2),
2181 //===----------------------------------------------------------------------===//
2182 // .insn directive instructions
2183 //===----------------------------------------------------------------------===//
2185 let isCodeGenOnly = 1, hasSideEffects = 1 in {
2186 def InsnE : DirectiveInsnE<(outs), (ins imm64zx16:$enc), ".insn e,$enc", []>;
2187 def InsnRI : DirectiveInsnRI<(outs), (ins imm64zx32:$enc, AnyReg:$R1,
2189 ".insn ri,$enc,$R1,$I2", []>;
2190 def InsnRIE : DirectiveInsnRIE<(outs), (ins imm64zx48:$enc, AnyReg:$R1,
2191 AnyReg:$R3, brtarget16:$I2),
2192 ".insn rie,$enc,$R1,$R3,$I2", []>;
2193 def InsnRIL : DirectiveInsnRIL<(outs), (ins imm64zx48:$enc, AnyReg:$R1,
2195 ".insn ril,$enc,$R1,$I2", []>;
2196 def InsnRILU : DirectiveInsnRIL<(outs), (ins imm64zx48:$enc, AnyReg:$R1,
2198 ".insn rilu,$enc,$R1,$I2", []>;
2199 def InsnRIS : DirectiveInsnRIS<(outs),
2200 (ins imm64zx48:$enc, AnyReg:$R1,
2201 imm32sx8:$I2, imm32zx4:$M3,
2203 ".insn ris,$enc,$R1,$I2,$M3,$BD4", []>;
2204 def InsnRR : DirectiveInsnRR<(outs),
2205 (ins imm64zx16:$enc, AnyReg:$R1, AnyReg:$R2),
2206 ".insn rr,$enc,$R1,$R2", []>;
2207 def InsnRRE : DirectiveInsnRRE<(outs), (ins imm64zx32:$enc,
2208 AnyReg:$R1, AnyReg:$R2),
2209 ".insn rre,$enc,$R1,$R2", []>;
2210 def InsnRRF : DirectiveInsnRRF<(outs),
2211 (ins imm64zx32:$enc, AnyReg:$R1, AnyReg:$R2,
2212 AnyReg:$R3, imm32zx4:$M4),
2213 ".insn rrf,$enc,$R1,$R2,$R3,$M4", []>;
2214 def InsnRRS : DirectiveInsnRRS<(outs),
2215 (ins imm64zx48:$enc, AnyReg:$R1,
2216 AnyReg:$R2, imm32zx4:$M3,
2218 ".insn rrs,$enc,$R1,$R2,$M3,$BD4", []>;
2219 def InsnRS : DirectiveInsnRS<(outs),
2220 (ins imm64zx32:$enc, AnyReg:$R1,
2221 AnyReg:$R3, bdaddr12only:$BD2),
2222 ".insn rs,$enc,$R1,$R3,$BD2", []>;
2223 def InsnRSE : DirectiveInsnRSE<(outs),
2224 (ins imm64zx48:$enc, AnyReg:$R1,
2225 AnyReg:$R3, bdaddr12only:$BD2),
2226 ".insn rse,$enc,$R1,$R3,$BD2", []>;
2227 def InsnRSI : DirectiveInsnRSI<(outs),
2228 (ins imm64zx48:$enc, AnyReg:$R1,
2229 AnyReg:$R3, brtarget16:$RI2),
2230 ".insn rsi,$enc,$R1,$R3,$RI2", []>;
2231 def InsnRSY : DirectiveInsnRSY<(outs),
2232 (ins imm64zx48:$enc, AnyReg:$R1,
2233 AnyReg:$R3, bdaddr20only:$BD2),
2234 ".insn rsy,$enc,$R1,$R3,$BD2", []>;
2235 def InsnRX : DirectiveInsnRX<(outs), (ins imm64zx32:$enc, AnyReg:$R1,
2236 bdxaddr12only:$XBD2),
2237 ".insn rx,$enc,$R1,$XBD2", []>;
2238 def InsnRXE : DirectiveInsnRXE<(outs), (ins imm64zx48:$enc, AnyReg:$R1,
2239 bdxaddr12only:$XBD2),
2240 ".insn rxe,$enc,$R1,$XBD2", []>;
2241 def InsnRXF : DirectiveInsnRXF<(outs),
2242 (ins imm64zx48:$enc, AnyReg:$R1,
2243 AnyReg:$R3, bdxaddr12only:$XBD2),
2244 ".insn rxf,$enc,$R1,$R3,$XBD2", []>;
2245 def InsnRXY : DirectiveInsnRXY<(outs), (ins imm64zx48:$enc, AnyReg:$R1,
2246 bdxaddr20only:$XBD2),
2247 ".insn rxy,$enc,$R1,$XBD2", []>;
2248 def InsnS : DirectiveInsnS<(outs),
2249 (ins imm64zx32:$enc, bdaddr12only:$BD2),
2250 ".insn s,$enc,$BD2", []>;
2251 def InsnSI : DirectiveInsnSI<(outs),
2252 (ins imm64zx32:$enc, bdaddr12only:$BD1,
2254 ".insn si,$enc,$BD1,$I2", []>;
2255 def InsnSIY : DirectiveInsnSIY<(outs),
2256 (ins imm64zx48:$enc,
2257 bdaddr20only:$BD1, imm32zx8:$I2),
2258 ".insn siy,$enc,$BD1,$I2", []>;
2259 def InsnSIL : DirectiveInsnSIL<(outs),
2260 (ins imm64zx48:$enc, bdaddr12only:$BD1,
2262 ".insn sil,$enc,$BD1,$I2", []>;
2263 def InsnSS : DirectiveInsnSS<(outs),
2264 (ins imm64zx48:$enc, bdraddr12only:$RBD1,
2265 bdaddr12only:$BD2, AnyReg:$R3),
2266 ".insn ss,$enc,$RBD1,$BD2,$R3", []>;
2267 def InsnSSE : DirectiveInsnSSE<(outs),
2268 (ins imm64zx48:$enc,
2269 bdaddr12only:$BD1,bdaddr12only:$BD2),
2270 ".insn sse,$enc,$BD1,$BD2", []>;
2271 def InsnSSF : DirectiveInsnSSF<(outs),
2272 (ins imm64zx48:$enc, bdaddr12only:$BD1,
2273 bdaddr12only:$BD2, AnyReg:$R3),
2274 ".insn ssf,$enc,$BD1,$BD2,$R3", []>;
2275 def InsnVRI : DirectiveInsnVRI<(outs),
2276 (ins imm64zx48:$enc, VR128:$V1, VR128:$V2,
2277 imm32zx12:$I3, imm32zx4:$M4, imm32zx4:$M5),
2278 ".insn vri,$enc,$V1,$V2,$I3,$M4,$M5", []>;
2279 def InsnVRR : DirectiveInsnVRR<(outs),
2280 (ins imm64zx48:$enc, VR128:$V1, VR128:$V2,
2281 VR128:$V3, imm32zx4:$M4, imm32zx4:$M5,
2283 ".insn vrr,$enc,$V1,$V2,$V3,$M4,$M5,$M6", []>;
2284 def InsnVRS : DirectiveInsnVRS<(outs),
2285 (ins imm64zx48:$enc, AnyReg:$R1, VR128:$V3,
2286 bdaddr12only:$BD2, imm32zx4:$M4),
2287 ".insn vrs,$enc,$BD2,$M4", []>;
2288 def InsnVRV : DirectiveInsnVRV<(outs),
2289 (ins imm64zx48:$enc, VR128:$V1,
2290 bdvaddr12only:$VBD2, imm32zx4:$M3),
2291 ".insn vrv,$enc,$V1,$VBD2,$M3", []>;
2292 def InsnVRX : DirectiveInsnVRX<(outs),
2293 (ins imm64zx48:$enc, VR128:$V1,
2294 bdxaddr12only:$XBD2, imm32zx4:$M3),
2295 ".insn vrx,$enc,$V1,$XBD2,$M3", []>;
2296 def InsnVSI : DirectiveInsnVSI<(outs),
2297 (ins imm64zx48:$enc, VR128:$V1,
2298 bdaddr12only:$BD2, imm32zx8:$I3),
2299 ".insn vsi,$enc,$V1,$BD2,$I3", []>;
2302 //===----------------------------------------------------------------------===//
2304 //===----------------------------------------------------------------------===//
2306 // Avoid generating 2 XOR instructions. (xor (and x, y), y) is
2307 // equivalent to (and (xor x, -1), y)
2308 def : Pat<(and (xor GR64:$x, (i64 -1)), GR64:$y),
2309 (XGR GR64:$y, (NGR GR64:$y, GR64:$x))>;
2311 // Shift/rotate instructions only use the last 6 bits of the second operand
2312 // register, so we can safely use NILL (16 fewer bits than NILF) to only AND the
2314 // Complexity is added so that we match this before we match NILF on the AND
2316 let AddedComplexity = 4 in {
2317 def : Pat<(shl GR32:$val, (and GR32:$shift, imm32zx16trunc:$imm)),
2318 (SLL GR32:$val, (NILL GR32:$shift, imm32zx16trunc:$imm), 0)>;
2320 def : Pat<(sra GR32:$val, (and GR32:$shift, imm32zx16trunc:$imm)),
2321 (SRA GR32:$val, (NILL GR32:$shift, imm32zx16trunc:$imm), 0)>;
2323 def : Pat<(srl GR32:$val, (and GR32:$shift, imm32zx16trunc:$imm)),
2324 (SRL GR32:$val, (NILL GR32:$shift, imm32zx16trunc:$imm), 0)>;
2326 def : Pat<(shl GR64:$val, (and GR32:$shift, imm32zx16trunc:$imm)),
2327 (SLLG GR64:$val, (NILL GR32:$shift, imm32zx16trunc:$imm), 0)>;
2329 def : Pat<(sra GR64:$val, (and GR32:$shift, imm32zx16trunc:$imm)),
2330 (SRAG GR64:$val, (NILL GR32:$shift, imm32zx16trunc:$imm), 0)>;
2332 def : Pat<(srl GR64:$val, (and GR32:$shift, imm32zx16trunc:$imm)),
2333 (SRLG GR64:$val, (NILL GR32:$shift, imm32zx16trunc:$imm), 0)>;
2335 def : Pat<(rotl GR32:$val, (and GR32:$shift, imm32zx16trunc:$imm)),
2336 (RLL GR32:$val, (NILL GR32:$shift, imm32zx16trunc:$imm), 0)>;
2338 def : Pat<(rotl GR64:$val, (and GR32:$shift, imm32zx16trunc:$imm)),
2339 (RLLG GR64:$val, (NILL GR32:$shift, imm32zx16trunc:$imm), 0)>;
2342 // Substitute (x*64-s) with (-s), since shift/rotate instructions only
2343 // use the last 6 bits of the second operand register (making it modulo 64).
2344 let AddedComplexity = 4 in {
2345 def : Pat<(shl GR64:$val, (sub imm32mod64, GR32:$shift)),
2346 (SLLG GR64:$val, (LCR GR32:$shift), 0)>;
2348 def : Pat<(sra GR64:$val, (sub imm32mod64, GR32:$shift)),
2349 (SRAG GR64:$val, (LCR GR32:$shift), 0)>;
2351 def : Pat<(srl GR64:$val, (sub imm32mod64, GR32:$shift)),
2352 (SRLG GR64:$val, (LCR GR32:$shift), 0)>;
2354 def : Pat<(rotl GR64:$val, (sub imm32mod64, GR32:$shift)),
2355 (RLLG GR64:$val, (LCR GR32:$shift), 0)>;
2358 // Peepholes for turning scalar operations into block operations.
2359 defm : BlockLoadStore<anyextloadi8, i32, MVCSequence, NCSequence, OCSequence,
2361 defm : BlockLoadStore<anyextloadi16, i32, MVCSequence, NCSequence, OCSequence,
2363 defm : BlockLoadStore<load, i32, MVCSequence, NCSequence, OCSequence,
2365 defm : BlockLoadStore<anyextloadi8, i64, MVCSequence, NCSequence,
2366 OCSequence, XCSequence, 1>;
2367 defm : BlockLoadStore<anyextloadi16, i64, MVCSequence, NCSequence, OCSequence,
2369 defm : BlockLoadStore<anyextloadi32, i64, MVCSequence, NCSequence, OCSequence,
2371 defm : BlockLoadStore<load, i64, MVCSequence, NCSequence, OCSequence,
2374 //===----------------------------------------------------------------------===//
2376 //===----------------------------------------------------------------------===//
2378 def JCT : MnemonicAlias<"jct", "brct">;
2379 def JCTG : MnemonicAlias<"jctg", "brctg">;
2380 def JAS : MnemonicAlias<"jas", "bras">;
2381 def JASL : MnemonicAlias<"jasl", "brasl">;
2382 def JXH : MnemonicAlias<"jxh", "brxh">;
2383 def JXLE : MnemonicAlias<"jxle", "brxle">;
2384 def JXHG : MnemonicAlias<"jxhg", "brxhg">;
2385 def JXLEG : MnemonicAlias<"jxleg", "brxlg">;
2387 def BRU : MnemonicAlias<"bru", "j">;
2388 def BRUL : MnemonicAlias<"brul", "jg", "att">;
2389 def BRUL_HLASM : MnemonicAlias<"brul", "jlu", "hlasm">;
2391 foreach V = [ "E", "NE", "H", "NH", "L", "NL", "HE", "NHE", "LE", "NLE",
2392 "Z", "NZ", "P", "NP", "M", "NM", "LH", "NLH", "O", "NO" ] in {
2393 defm BRUAsm#V : MnemonicCondBranchAlias <CV<V>, "br#", "j#">;
2394 defm BRULAsm#V : MnemonicCondBranchAlias <CV<V>, "br#l", "jg#", "att">;
2395 defm BRUL_HLASMAsm#V : MnemonicCondBranchAlias <CV<V>, "br#l", "jl#", "hlasm">;