xref: /freebsd/contrib/llvm-project/llvm/lib/Target/SystemZ/SystemZInstrInfo.td (revision 924226fba12cc9a228c73b956e1b7fa24c60b055)
1//===-- SystemZInstrInfo.td - General SystemZ instructions ----*- tblgen-*-===//
2//
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
6//
7//===----------------------------------------------------------------------===//
8
9def IsTargetXPLINK64      : Predicate<"Subtarget->isTargetXPLINK64()">;
10def IsTargetELF           : Predicate<"Subtarget->isTargetELF()">;
11
12//===----------------------------------------------------------------------===//
13// Stack allocation
14//===----------------------------------------------------------------------===//
15
16// The callseq_start node requires the hasSideEffects flag, even though these
17// instructions are noops on SystemZ.
18let 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)]>;
23}
24
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.
28//
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.
32def ADJDYNALLOC : Pseudo<(outs GR64:$dst), (ins dynalloc12only:$src),
33                         [(set GR64:$dst, dynalloc12only:$src)]>;
34
35let 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))]>;
40
41let Defs = [R1D, R15D, CC], Uses = [R15D], hasNoSchedulingInfo = 1,
42    hasSideEffects = 1 in
43  def PROBED_STACKALLOC : Pseudo<(outs), (ins i64imm:$stacksize), []>;
44
45//===----------------------------------------------------------------------===//
46// Branch instructions
47//===----------------------------------------------------------------------===//
48
49// Conditional branches.
50let 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]>;
65    }
66  }
67
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]>;
77  }
78
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]>;
90    }
91  }
92}
93
94// Unconditional branches.  These are in fact simply variants of the
95// conditional branches with the condition mask set to "always".
96let 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]>;
104  }
105}
106
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.
110def NOP  : InstAlias<"nop\t$XBD", (BCAsm 0, bdxaddr12only:$XBD), 0>;
111let isAsmParserOnly = 1, hasNoSchedulingInfo = 1, M1 = 0, XBD2 = 0 in
112  def NOP_bare  : InstRXb<0x47,(outs), (ins), "nop", []>;
113def NOPR : InstAlias<"nopr\t$R", (BCRAsm 0, GR64:$R), 0>;
114def NOPR_bare : InstAlias<"nopr", (BCRAsm 0, R0D), 0>;
115
116// An alias of BRC 0, label
117def JNOP : InstAlias<"jnop\t$RI2", (BRCAsm 0, brtarget16:$RI2), 0>;
118
119// An alias of BRCL 0, label
120// jgnop on att ; jlnop on hlasm
121def JGNOP : InstAlias<"{jgnop|jlnop}\t$RI2", (BRCLAsm 0, brtarget32:$RI2), 0>;
122
123// Fused compare-and-branch instructions.
124//
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.
129let 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.
133  let Defs = [CC] in {
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>;
142  }
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>;
152  }
153
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 {
157    let Defs = [CC] 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,
161                                          imm32sx8>;
162      def CGIJAsm#V  : FixedCmpBranchRIEc<ICV<V>, "cgij",  0xEC7C, GR64,
163                                          imm64sx8>;
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,
167                                          imm32zx8>;
168      def CLGIJAsm#V : FixedCmpBranchRIEc<ICV<V>, "clgij", 0xEC7D, GR64,
169                                          imm64zx8>;
170    }
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,
175                                         imm32sx8>;
176      def CGIBAsm#V  : FixedCmpBranchRIS<ICV<V>, "cgib",  0xECFC, GR64,
177                                         imm64sx8>;
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,
181                                         imm32zx8>;
182      def CLGIBAsm#V : FixedCmpBranchRIS<ICV<V>, "clgib", 0xECFD, GR64,
183                                         imm64zx8>;
184    }
185  }
186}
187
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.
190let isBranch = 1, isTerminator = 1 in {
191  let Defs = [CC] in {
192    def BRCT  : BranchUnaryRI<"brct",  0xA76, GR32>;
193    def BRCTG : BranchUnaryRI<"brctg", 0xA77, GR64>;
194  }
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]>;
198
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>;
203}
204
205let isBranch = 1, isTerminator = 1 in {
206  let Defs = [CC] 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>;
211  }
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>;
216}
217
218//===----------------------------------------------------------------------===//
219// Trap instructions
220//===----------------------------------------------------------------------===//
221
222// Unconditional trap.
223let hasCtrlDep = 1, hasSideEffects = 1 in
224  def Trap : Alias<4, (outs), (ins), [(trap)]>;
225
226// Conditional trap.
227let hasCtrlDep = 1, Uses = [CC], hasSideEffects = 1 in
228  def CondTrap : Alias<4, (outs), (ins cond4:$valid, cond4:$R1), []>;
229
230// Fused compare-and-trap instructions.
231let 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>;
244  }
245
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,
253                                         imm32sx16>;
254    def CGITAsm#V  : FixedCmpBranchRIEa<ICV<V>, "cgit",  0xEC70, GR64,
255                                         imm64sx16>;
256    def CLFITAsm#V : FixedCmpBranchRIEa<ICV<V>, "clfit", 0xEC73, GR32,
257                                         imm32zx16>;
258    def CLGITAsm#V : FixedCmpBranchRIEa<ICV<V>, "clgit", 0xEC71, GR64,
259                                         imm64zx16>;
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>;
263    }
264  }
265}
266
267//===----------------------------------------------------------------------===//
268// Call and return instructions
269//===----------------------------------------------------------------------===//
270
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.
273let 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>;
278}
279
280// z/OS XPLINK
281let 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:$R2)]>;
287  }
288}
289
290// Regular calls.
291// z/Linux ELF
292let Predicates = [IsTargetELF] in {
293  let isCall = 1, Defs = [R14D, CC], Uses = [FPC] in {
294    def CallBRASL : Alias<6, (outs), (ins pcrel32:$I2, variable_ops),
295                          [(z_call pcrel32:$I2)]>;
296    def CallBASR  : Alias<2, (outs), (ins ADDR64:$R2, variable_ops),
297                          [(z_call ADDR64:$R2)]>;
298  }
299
300  // TLS calls.  These will be lowered into a call to __tls_get_offset,
301  // with an extra relocation specifying the TLS symbol.
302  let isCall = 1, Defs = [R14D, CC] in {
303    def TLS_GDCALL : Alias<6, (outs), (ins tlssym:$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:$I2)]>;
307  }
308}
309
310// Sibling calls. Indirect sibling calls must be via R6 for XPLink,
311// R1 used for ELF
312let isCall = 1, isTerminator = 1, isReturn = 1, isBarrier = 1 in {
313  def CallJG : Alias<6, (outs), (ins pcrel32:$I2),
314                     [(z_sibcall pcrel32:$I2)]>;
315  def CallBR : Alias<2, (outs), (ins ADDR64:$R2),
316                     [(z_sibcall ADDR64:$R2)]>;
317}
318
319// Conditional sibling calls.
320let CCMaskFirst = 1, isCall = 1, isTerminator = 1, isReturn = 1 in {
321  def CallBRCL : Alias<6, (outs), (ins cond4:$valid, cond4:$R1,
322                                   pcrel32:$I2), []>;
323  def CallBCR : Alias<2, (outs), (ins cond4:$valid, cond4:$R1,
324                                  ADDR64:$R2), []>;
325}
326
327// Fused compare and conditional sibling calls.
328let isCall = 1, isTerminator = 1, isReturn = 1 in {
329  def CRBCall : Alias<6, (outs), (ins GR32:$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:$R4), []>;
337}
338
339// A return instruction (br %r14) for ELF and (b 2 %r7) for XPLink.
340let isReturn = 1, isTerminator = 1, isBarrier = 1, hasCtrlDep = 1 in
341  def Return : Alias<2, (outs), (ins), [(z_retflag)]>;
342
343// A conditional return instruction (bcr <cond>, %r14).
344let isReturn = 1, isTerminator = 1, hasCtrlDep = 1, CCMaskFirst = 1, Uses = [CC] in
345  def CondReturn : Alias<2, (outs), (ins cond4:$valid, cond4:$R1), []>;
346
347// Fused compare and conditional returns.
348let isReturn = 1, isTerminator = 1, hasCtrlDep = 1 in {
349  def CRBReturn : Alias<6, (outs), (ins GR32:$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), []>;
357}
358
359//===----------------------------------------------------------------------===//
360// Select instructions
361//===----------------------------------------------------------------------===//
362
363def Select32    : SelectWrapper<i32, GR32>,
364                  Requires<[FeatureNoLoadStoreOnCond]>;
365def Select64    : SelectWrapper<i64, GR64>,
366                  Requires<[FeatureNoLoadStoreOnCond]>;
367
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.
370defm CondStore8Mux  : CondStores<GRX32, nonvolatile_truncstorei8,
371                                 nonvolatile_anyextloadi8, bdxaddr20only>,
372                      Requires<[FeatureHighWord]>;
373defm CondStore16Mux : CondStores<GRX32, nonvolatile_truncstorei16,
374                                 nonvolatile_anyextloadi16, bdxaddr20only>,
375                      Requires<[FeatureHighWord]>;
376defm CondStore32Mux : CondStores<GRX32, simple_store,
377                                 simple_load, bdxaddr20only>,
378                      Requires<[FeatureLoadStoreOnCond2]>;
379defm CondStore8     : CondStores<GR32, nonvolatile_truncstorei8,
380                                 nonvolatile_anyextloadi8, bdxaddr20only>;
381defm CondStore16    : CondStores<GR32, nonvolatile_truncstorei16,
382                                 nonvolatile_anyextloadi16, bdxaddr20only>;
383defm CondStore32    : CondStores<GR32, simple_store,
384                                 simple_load, bdxaddr20only>;
385
386defm : CondStores64<CondStore8, CondStore8Inv, nonvolatile_truncstorei8,
387                    nonvolatile_anyextloadi8, bdxaddr20only>;
388defm : CondStores64<CondStore16, CondStore16Inv, nonvolatile_truncstorei16,
389                    nonvolatile_anyextloadi16, bdxaddr20only>;
390defm : CondStores64<CondStore32, CondStore32Inv, nonvolatile_truncstorei32,
391                    nonvolatile_anyextloadi32, bdxaddr20only>;
392defm CondStore64 : CondStores<GR64, simple_store,
393                              simple_load, bdxaddr20only>;
394
395//===----------------------------------------------------------------------===//
396// Move instructions
397//===----------------------------------------------------------------------===//
398
399// Register moves.
400def LR  : UnaryRR <"lr",  0x18,   null_frag, GR32, GR32>;
401def LGR : UnaryRRE<"lgr", 0xB904, null_frag, GR64, GR64>;
402
403let 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>;
406}
407
408let usesCustomInserter = 1, hasNoSchedulingInfo = 1 in
409  def PAIR128 : Pseudo<(outs GR128:$dst), (ins GR64:$hi, GR64:$lo), []>;
410
411// Immediate moves.
412let 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>;
419
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>;
425
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>;
430}
431
432// Register loads.
433let 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>;
441
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))]>;
447  }
448}
449let 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>;
452}
453
454let canFoldAsLoad = 1 in {
455  def LRL  : UnaryRILPC<"lrl",  0xC4D, aligned_load, GR32>;
456  def LGRL : UnaryRILPC<"lgrl", 0xC48, aligned_load, GR64>;
457}
458
459// Load and zero rightmost byte.
460let 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)>;
467}
468
469// Load and trap.
470let 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>;
474}
475
476// Register stores.
477let 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>;
485
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)]>;
491  }
492}
493def STRL  : StoreRILPC<"strl", 0xC4F, aligned_store, GR32>;
494def STGRL : StoreRILPC<"stgrl", 0xC4B, aligned_store, GR64>;
495
496// 8-bit immediate stores to 8-bit fields.
497defm MVI : StoreSIPair<"mvi", 0x92, 0xEB52, truncstorei8, imm32zx8trunc>;
498
499// 16-bit immediate stores to 16-, 32- or 64-bit fields.
500def MVHHI : StoreSIL<"mvhhi", 0xE544, truncstorei16, imm32sx16trunc>;
501def MVHI  : StoreSIL<"mvhi",  0xE54C, store,         imm32sx16>;
502def MVGHI : StoreSIL<"mvghi", 0xE548, store,         imm64sx16>;
503
504// Memory-to-memory moves.
505let mayLoad = 1, mayStore = 1 in
506  defm MVC : MemorySS<"mvc", 0xD2, z_mvc>;
507let 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>;
511}
512
513// Memset[Length][Byte] pseudos.
514def MemsetImmImm : MemsetPseudo<imm64, imm32zx8trunc>;
515def MemsetImmReg : MemsetPseudo<imm64, GR32>;
516def MemsetRegImm : MemsetPseudo<ADDR64, imm32zx8trunc>;
517def MemsetRegReg : MemsetPseudo<ADDR64, GR32>;
518
519// Move right.
520let Predicates = [FeatureMiscellaneousExtensions3],
521    mayLoad = 1, mayStore = 1, Uses = [R0L] in
522  def MVCRL : SideEffectBinarySSE<"mvcrl", 0xE50A>;
523
524// String moves.
525let mayLoad = 1, mayStore = 1, Defs = [CC] in
526  defm MVST : StringRRE<"mvst", 0xB255, z_stpcpy>;
527
528//===----------------------------------------------------------------------===//
529// Conditional move instructions
530//===----------------------------------------------------------------------===//
531
532let Predicates = [FeatureMiscellaneousExtensions3], Uses = [CC] in {
533  // Select.
534  let isCommutable = 1 in {
535    // Expands to SELR or SELFHR or a branch-and-move sequence,
536    // depending on the choice of registers.
537    def  SELRMux : CondBinaryRRFaPseudo<"MUXselr", GRX32, GRX32, GRX32>;
538    defm SELFHR  : CondBinaryRRFaPair<"selfhr", 0xB9C0, GRH32, GRH32, GRH32>;
539    defm SELR    : CondBinaryRRFaPair<"selr",   0xB9F0, GR32, GR32, GR32>;
540    defm SELGR   : CondBinaryRRFaPair<"selgr",  0xB9E3, GR64, GR64, GR64>;
541  }
542
543  // Define AsmParser extended mnemonics for each general condition-code mask.
544  foreach V = [ "E", "NE", "H", "NH", "L", "NL", "HE", "NHE", "LE", "NLE",
545                "Z", "NZ", "P", "NP", "M", "NM", "LH", "NLH", "O", "NO" ] in {
546    def SELRAsm#V   : FixedCondBinaryRRFa<CV<V>, "selr",   0xB9F0,
547                                          GR32, GR32, GR32>;
548    def SELFHRAsm#V : FixedCondBinaryRRFa<CV<V>, "selfhr", 0xB9C0,
549                                          GRH32, GRH32, GRH32>;
550    def SELGRAsm#V  : FixedCondBinaryRRFa<CV<V>, "selgr",  0xB9E3,
551                                          GR64, GR64, GR64>;
552  }
553}
554
555let Predicates = [FeatureLoadStoreOnCond2], Uses = [CC] in {
556  // Load immediate on condition.  Matched via DAG pattern and created
557  // by the PeepholeOptimizer via FoldImmediate.
558
559  // Expands to LOCHI or LOCHHI, depending on the choice of register.
560  def LOCHIMux : CondBinaryRIEPseudo<GRX32, imm32sx16>;
561  defm LOCHHI  : CondBinaryRIEPair<"lochhi", 0xEC4E, GRH32, imm32sx16>;
562  defm LOCHI   : CondBinaryRIEPair<"lochi",  0xEC42, GR32, imm32sx16>;
563  defm LOCGHI  : CondBinaryRIEPair<"locghi", 0xEC46, GR64, imm64sx16>;
564
565  // Move register on condition.  Matched via DAG pattern and
566  // created by early if-conversion.
567  let isCommutable = 1 in {
568    // Expands to LOCR or LOCFHR or a branch-and-move sequence,
569    // depending on the choice of registers.
570    def LOCRMux : CondBinaryRRFPseudo<"MUXlocr", GRX32, GRX32>;
571    defm LOCFHR : CondBinaryRRFPair<"locfhr", 0xB9E0, GRH32, GRH32>;
572  }
573
574  // Load on condition.  Matched via DAG pattern.
575  // Expands to LOC or LOCFH, depending on the choice of register.
576  defm LOCMux : CondUnaryRSYPseudoAndMemFold<"MUXloc", simple_load, GRX32, 4>;
577  defm LOCFH : CondUnaryRSYPair<"locfh", 0xEBE0, simple_load, GRH32, 4>;
578
579  // Store on condition.  Expanded from CondStore* pseudos.
580  // Expands to STOC or STOCFH, depending on the choice of register.
581  def STOCMux : CondStoreRSYPseudo<GRX32, 4>;
582  defm STOCFH : CondStoreRSYPair<"stocfh", 0xEBE1, GRH32, 4>;
583
584  // Define AsmParser extended mnemonics for each general condition-code mask.
585  foreach V = [ "E", "NE", "H", "NH", "L", "NL", "HE", "NHE", "LE", "NLE",
586                "Z", "NZ", "P", "NP", "M", "NM", "LH", "NLH", "O", "NO" ] in {
587    def LOCHIAsm#V  : FixedCondBinaryRIE<CV<V>, "lochi",  0xEC42, GR32,
588                                         imm32sx16>;
589    def LOCGHIAsm#V : FixedCondBinaryRIE<CV<V>, "locghi", 0xEC46, GR64,
590                                         imm64sx16>;
591    def LOCHHIAsm#V : FixedCondBinaryRIE<CV<V>, "lochhi", 0xEC4E, GRH32,
592                                         imm32sx16>;
593    def LOCFHRAsm#V : FixedCondBinaryRRF<CV<V>, "locfhr", 0xB9E0, GRH32, GRH32>;
594    def LOCFHAsm#V  : FixedCondUnaryRSY<CV<V>, "locfh",  0xEBE0, GRH32, 4>;
595    def STOCFHAsm#V : FixedCondStoreRSY<CV<V>, "stocfh", 0xEBE1, GRH32, 4>;
596  }
597}
598
599let Predicates = [FeatureLoadStoreOnCond], Uses = [CC] in {
600  // Move register on condition.  Matched via DAG pattern and
601  // created by early if-conversion.
602  let isCommutable = 1 in {
603    defm LOCR  : CondBinaryRRFPair<"locr",  0xB9F2, GR32, GR32>;
604    defm LOCGR : CondBinaryRRFPair<"locgr", 0xB9E2, GR64, GR64>;
605  }
606
607  // Load on condition.  Matched via DAG pattern.
608  defm LOC  : CondUnaryRSYPair<"loc",  0xEBF2, simple_load, GR32, 4>;
609  defm LOCG : CondUnaryRSYPairAndMemFold<"locg", 0xEBE2, simple_load, GR64, 8>;
610
611  // Store on condition.  Expanded from CondStore* pseudos.
612  defm STOC  : CondStoreRSYPair<"stoc",  0xEBF3, GR32, 4>;
613  defm STOCG : CondStoreRSYPair<"stocg", 0xEBE3, GR64, 8>;
614
615  // Define AsmParser extended mnemonics for each general condition-code mask.
616  foreach V = [ "E", "NE", "H", "NH", "L", "NL", "HE", "NHE", "LE", "NLE",
617                "Z", "NZ", "P", "NP", "M", "NM", "LH", "NLH", "O", "NO" ] in {
618    def LOCRAsm#V   : FixedCondBinaryRRF<CV<V>, "locr",  0xB9F2, GR32, GR32>;
619    def LOCGRAsm#V  : FixedCondBinaryRRF<CV<V>, "locgr", 0xB9E2, GR64, GR64>;
620    def LOCAsm#V    : FixedCondUnaryRSY<CV<V>, "loc",   0xEBF2, GR32, 4>;
621    def LOCGAsm#V   : FixedCondUnaryRSY<CV<V>, "locg",  0xEBE2, GR64, 8>;
622    def STOCAsm#V   : FixedCondStoreRSY<CV<V>, "stoc",  0xEBF3, GR32, 4>;
623    def STOCGAsm#V  : FixedCondStoreRSY<CV<V>, "stocg", 0xEBE3, GR64, 8>;
624  }
625}
626//===----------------------------------------------------------------------===//
627// Sign extensions
628//===----------------------------------------------------------------------===//
629//
630// Note that putting these before zero extensions mean that we will prefer
631// them for anyextload*.  There's not really much to choose between the two
632// either way, but signed-extending loads have a short LH and a long LHY,
633// while zero-extending loads have only the long LLH.
634//
635//===----------------------------------------------------------------------===//
636
637// 32-bit extensions from registers.
638def LBR : UnaryRRE<"lbr", 0xB926, sext8,  GR32, GR32>;
639def LHR : UnaryRRE<"lhr", 0xB927, sext16, GR32, GR32>;
640
641// 64-bit extensions from registers.
642def LGBR : UnaryRRE<"lgbr", 0xB906, sext8,  GR64, GR64>;
643def LGHR : UnaryRRE<"lghr", 0xB907, sext16, GR64, GR64>;
644def LGFR : UnaryRRE<"lgfr", 0xB914, sext32, GR64, GR32>;
645
646let Defs = [CC], CCValues = 0xE, CompareZeroCCMask = 0xE in
647  def LTGFR : UnaryRRE<"ltgfr", 0xB912, null_frag, GR64, GR32>;
648
649// Match 32-to-64-bit sign extensions in which the source is already
650// in a 64-bit register.
651def : Pat<(sext_inreg GR64:$src, i32),
652          (LGFR (EXTRACT_SUBREG GR64:$src, subreg_l32))>;
653
654// 32-bit extensions from 8-bit memory.  LBMux expands to LB or LBH,
655// depending on the choice of register.
656def LBMux : UnaryRXYPseudo<"lb", asextloadi8, GRX32, 1>,
657            Requires<[FeatureHighWord]>;
658def LB  : UnaryRXY<"lb", 0xE376, asextloadi8, GR32, 1>;
659def LBH : UnaryRXY<"lbh", 0xE3C0, asextloadi8, GRH32, 1>,
660          Requires<[FeatureHighWord]>;
661
662// 32-bit extensions from 16-bit memory.  LHMux expands to LH or LHH,
663// depending on the choice of register.
664def LHMux : UnaryRXYPseudo<"lh", asextloadi16, GRX32, 2>,
665            Requires<[FeatureHighWord]>;
666defm LH   : UnaryRXPair<"lh", 0x48, 0xE378, asextloadi16, GR32, 2>;
667def  LHH  : UnaryRXY<"lhh", 0xE3C4, asextloadi16, GRH32, 2>,
668            Requires<[FeatureHighWord]>;
669def  LHRL : UnaryRILPC<"lhrl", 0xC45, aligned_asextloadi16, GR32>;
670
671// 64-bit extensions from memory.
672def LGB   : UnaryRXY<"lgb", 0xE377, asextloadi8,  GR64, 1>;
673def LGH   : UnaryRXY<"lgh", 0xE315, asextloadi16, GR64, 2>;
674def LGF   : UnaryRXY<"lgf", 0xE314, asextloadi32, GR64, 4>;
675def LGHRL : UnaryRILPC<"lghrl", 0xC44, aligned_asextloadi16, GR64>;
676def LGFRL : UnaryRILPC<"lgfrl", 0xC4C, aligned_asextloadi32, GR64>;
677let Defs = [CC], CCValues = 0xE, CompareZeroCCMask = 0xE in
678  def LTGF : UnaryRXY<"ltgf", 0xE332, asextloadi32, GR64, 4>;
679
680//===----------------------------------------------------------------------===//
681// Zero extensions
682//===----------------------------------------------------------------------===//
683
684// 32-bit extensions from registers.
685
686// Expands to LLCR or RISB[LH]G, depending on the choice of registers.
687def LLCRMux : UnaryRRPseudo<"llcr", zext8, GRX32, GRX32>,
688              Requires<[FeatureHighWord]>;
689def LLCR    : UnaryRRE<"llcr", 0xB994, zext8,  GR32, GR32>;
690// Expands to LLHR or RISB[LH]G, depending on the choice of registers.
691def LLHRMux : UnaryRRPseudo<"llhr", zext16, GRX32, GRX32>,
692              Requires<[FeatureHighWord]>;
693def LLHR    : UnaryRRE<"llhr", 0xB995, zext16, GR32, GR32>;
694
695// 64-bit extensions from registers.
696def LLGCR : UnaryRRE<"llgcr", 0xB984, zext8,  GR64, GR64>;
697def LLGHR : UnaryRRE<"llghr", 0xB985, zext16, GR64, GR64>;
698def LLGFR : UnaryRRE<"llgfr", 0xB916, zext32, GR64, GR32>;
699
700// Match 32-to-64-bit zero extensions in which the source is already
701// in a 64-bit register.
702def : Pat<(and GR64:$src, 0xffffffff),
703          (LLGFR (EXTRACT_SUBREG GR64:$src, subreg_l32))>;
704
705// 32-bit extensions from 8-bit memory.  LLCMux expands to LLC or LLCH,
706// depending on the choice of register.
707def LLCMux : UnaryRXYPseudo<"llc", azextloadi8, GRX32, 1>,
708             Requires<[FeatureHighWord]>;
709def LLC  : UnaryRXY<"llc", 0xE394, azextloadi8, GR32, 1>;
710def LLCH : UnaryRXY<"llch", 0xE3C2, azextloadi8, GRH32, 1>,
711           Requires<[FeatureHighWord]>;
712
713// 32-bit extensions from 16-bit memory.  LLHMux expands to LLH or LLHH,
714// depending on the choice of register.
715def LLHMux : UnaryRXYPseudo<"llh", azextloadi16, GRX32, 2>,
716             Requires<[FeatureHighWord]>;
717def LLH   : UnaryRXY<"llh", 0xE395, azextloadi16, GR32, 2>;
718def LLHH  : UnaryRXY<"llhh", 0xE3C6, azextloadi16, GRH32, 2>,
719            Requires<[FeatureHighWord]>;
720def LLHRL : UnaryRILPC<"llhrl", 0xC42, aligned_azextloadi16, GR32>;
721
722// 64-bit extensions from memory.
723def LLGC   : UnaryRXY<"llgc", 0xE390, azextloadi8,  GR64, 1>;
724def LLGH   : UnaryRXY<"llgh", 0xE391, azextloadi16, GR64, 2>;
725def LLGF   : UnaryRXY<"llgf", 0xE316, azextloadi32, GR64, 4>;
726def LLGHRL : UnaryRILPC<"llghrl", 0xC46, aligned_azextloadi16, GR64>;
727def LLGFRL : UnaryRILPC<"llgfrl", 0xC4E, aligned_azextloadi32, GR64>;
728
729// 31-to-64-bit zero extensions.
730def LLGTR : UnaryRRE<"llgtr", 0xB917, null_frag, GR64, GR64>;
731def LLGT  : UnaryRXY<"llgt",  0xE317, null_frag, GR64, 4>;
732def : Pat<(and GR64:$src, 0x7fffffff),
733          (LLGTR GR64:$src)>;
734def : Pat<(and (i64 (azextloadi32 bdxaddr20only:$src)), 0x7fffffff),
735          (LLGT bdxaddr20only:$src)>;
736
737// Load and zero rightmost byte.
738let Predicates = [FeatureLoadAndZeroRightmostByte] in {
739  def LLZRGF : UnaryRXY<"llzrgf", 0xE33A, null_frag, GR64, 4>;
740  def : Pat<(and (i64 (azextloadi32 bdxaddr20only:$src)), 0xffffff00),
741            (LLZRGF bdxaddr20only:$src)>;
742}
743
744// Load and trap.
745let Predicates = [FeatureLoadAndTrap], hasSideEffects = 1 in {
746  def LLGFAT : UnaryRXY<"llgfat", 0xE39D, null_frag, GR64, 4>;
747  def LLGTAT : UnaryRXY<"llgtat", 0xE39C, null_frag, GR64, 4>;
748}
749
750// Extend GR64s to GR128s.
751let usesCustomInserter = 1, hasNoSchedulingInfo = 1 in
752  def ZEXT128 : Pseudo<(outs GR128:$dst), (ins GR64:$src), []>;
753
754//===----------------------------------------------------------------------===//
755// "Any" extensions
756//===----------------------------------------------------------------------===//
757
758// Use subregs to populate the "don't care" bits in a 32-bit to 64-bit anyext.
759def : Pat<(i64 (anyext GR32:$src)),
760          (INSERT_SUBREG (i64 (IMPLICIT_DEF)), GR32:$src, subreg_l32)>;
761
762// Extend GR64s to GR128s.
763let usesCustomInserter = 1, hasNoSchedulingInfo = 1 in
764  def AEXT128 : Pseudo<(outs GR128:$dst), (ins GR64:$src), []>;
765
766//===----------------------------------------------------------------------===//
767// Truncations
768//===----------------------------------------------------------------------===//
769
770// Truncations of 64-bit registers to 32-bit registers.
771def : Pat<(i32 (trunc GR64:$src)),
772          (EXTRACT_SUBREG GR64:$src, subreg_l32)>;
773
774// Truncations of 32-bit registers to 8-bit memory.  STCMux expands to
775// STC, STCY or STCH, depending on the choice of register.
776def STCMux : StoreRXYPseudo<truncstorei8, GRX32, 1>,
777             Requires<[FeatureHighWord]>;
778defm STC : StoreRXPair<"stc", 0x42, 0xE372, truncstorei8, GR32, 1>;
779def STCH : StoreRXY<"stch", 0xE3C3, truncstorei8, GRH32, 1>,
780           Requires<[FeatureHighWord]>;
781
782// Truncations of 32-bit registers to 16-bit memory.  STHMux expands to
783// STH, STHY or STHH, depending on the choice of register.
784def STHMux : StoreRXYPseudo<truncstorei16, GRX32, 1>,
785             Requires<[FeatureHighWord]>;
786defm STH : StoreRXPair<"sth", 0x40, 0xE370, truncstorei16, GR32, 2>;
787def STHH : StoreRXY<"sthh", 0xE3C7, truncstorei16, GRH32, 2>,
788           Requires<[FeatureHighWord]>;
789def STHRL : StoreRILPC<"sthrl", 0xC47, aligned_truncstorei16, GR32>;
790
791// Truncations of 64-bit registers to memory.
792defm : StoreGR64Pair<STC, STCY, truncstorei8>;
793defm : StoreGR64Pair<STH, STHY, truncstorei16>;
794def  : StoreGR64PC<STHRL, aligned_truncstorei16>;
795defm : StoreGR64Pair<ST, STY, truncstorei32>;
796def  : StoreGR64PC<STRL, aligned_truncstorei32>;
797
798// Store characters under mask -- not (yet) used for codegen.
799defm STCM : StoreBinaryRSPair<"stcm", 0xBE, 0xEB2D, GR32, 0>;
800def STCMH : StoreBinaryRSY<"stcmh", 0xEB2C, GRH32, 0>;
801
802//===----------------------------------------------------------------------===//
803// Multi-register moves
804//===----------------------------------------------------------------------===//
805
806// Multi-register loads.
807defm LM : LoadMultipleRSPair<"lm", 0x98, 0xEB98, GR32>;
808def LMG : LoadMultipleRSY<"lmg", 0xEB04, GR64>;
809def LMH : LoadMultipleRSY<"lmh", 0xEB96, GRH32>;
810def LMD : LoadMultipleSSe<"lmd", 0xEF, GR64>;
811
812// Multi-register stores.
813defm STM : StoreMultipleRSPair<"stm", 0x90, 0xEB90, GR32>;
814def STMG : StoreMultipleRSY<"stmg", 0xEB24, GR64>;
815def STMH : StoreMultipleRSY<"stmh", 0xEB26, GRH32>;
816
817//===----------------------------------------------------------------------===//
818// Byte swaps
819//===----------------------------------------------------------------------===//
820
821// Byte-swapping register moves.
822def LRVR  : UnaryRRE<"lrvr",  0xB91F, bswap, GR32, GR32>;
823def LRVGR : UnaryRRE<"lrvgr", 0xB90F, bswap, GR64, GR64>;
824
825// Byte-swapping loads.
826def LRVH : UnaryRXY<"lrvh", 0xE31F, z_loadbswap16, GR32, 2>;
827def LRV  : UnaryRXY<"lrv",  0xE31E, z_loadbswap32, GR32, 4>;
828def LRVG : UnaryRXY<"lrvg", 0xE30F, z_loadbswap64, GR64, 8>;
829
830// Byte-swapping stores.
831def STRVH : StoreRXY<"strvh", 0xE33F, z_storebswap16, GR32, 2>;
832def STRV  : StoreRXY<"strv",  0xE33E, z_storebswap32, GR32, 4>;
833def STRVG : StoreRXY<"strvg", 0xE32F, z_storebswap64, GR64, 8>;
834
835// Byte-swapping memory-to-memory moves.
836let mayLoad = 1, mayStore = 1 in
837  def MVCIN : SideEffectBinarySSa<"mvcin", 0xE8>;
838
839//===----------------------------------------------------------------------===//
840// Load address instructions
841//===----------------------------------------------------------------------===//
842
843// Load BDX-style addresses.
844let isAsCheapAsAMove = 1, isReMaterializable = 1 in
845  defm LA : LoadAddressRXPair<"la", 0x41, 0xE371, bitconvert>;
846
847// Load a PC-relative address.  There's no version of this instruction
848// with a 16-bit offset, so there's no relaxation.
849let isAsCheapAsAMove = 1, isMoveImm = 1, isReMaterializable = 1 in
850  def LARL : LoadAddressRIL<"larl", 0xC00, bitconvert>;
851
852// Load the Global Offset Table address.  This will be lowered into a
853//     larl $R1, _GLOBAL_OFFSET_TABLE_
854// instruction.
855def GOT : Alias<6, (outs GR64:$R1), (ins),
856                [(set GR64:$R1, (global_offset_table))]>;
857
858//===----------------------------------------------------------------------===//
859// Absolute and Negation
860//===----------------------------------------------------------------------===//
861
862let Defs = [CC] in {
863  let CCValues = 0xF, CompareZeroCCMask = 0x8 in {
864    def LPR  : UnaryRR <"lpr",  0x10,   abs, GR32, GR32>;
865    def LPGR : UnaryRRE<"lpgr", 0xB900, abs, GR64, GR64>;
866  }
867  let CCValues = 0xE, CompareZeroCCMask = 0xE in
868    def LPGFR : UnaryRRE<"lpgfr", 0xB910, null_frag, GR64, GR32>;
869}
870defm : SXU<abs, LPGFR>;
871
872let Defs = [CC] in {
873  let CCValues = 0xF, CompareZeroCCMask = 0x8 in {
874    def LNR  : UnaryRR <"lnr",  0x11,   z_inegabs, GR32, GR32>;
875    def LNGR : UnaryRRE<"lngr", 0xB901, z_inegabs, GR64, GR64>;
876  }
877  let CCValues = 0xE, CompareZeroCCMask = 0xE in
878    def LNGFR : UnaryRRE<"lngfr", 0xB911, null_frag, GR64, GR32>;
879}
880defm : SXU<z_inegabs, LNGFR>;
881
882let Defs = [CC] in {
883  let CCValues = 0xF, CompareZeroCCMask = 0x8 in {
884    def LCR  : UnaryRR <"lcr",  0x13,   ineg, GR32, GR32>;
885    def LCGR : UnaryRRE<"lcgr", 0xB903, ineg, GR64, GR64>;
886  }
887  let CCValues = 0xE, CompareZeroCCMask = 0xE in
888    def LCGFR : UnaryRRE<"lcgfr", 0xB913, null_frag, GR64, GR32>;
889}
890defm : SXU<ineg, LCGFR>;
891
892//===----------------------------------------------------------------------===//
893// Insertion
894//===----------------------------------------------------------------------===//
895
896let isCodeGenOnly = 1 in
897  defm IC32 : BinaryRXPair<"ic", 0x43, 0xE373, inserti8, GR32, azextloadi8, 1>;
898defm IC : BinaryRXPair<"ic", 0x43, 0xE373, inserti8, GR64, azextloadi8, 1>;
899
900defm : InsertMem<"inserti8", IC32,  GR32, azextloadi8, bdxaddr12pair>;
901defm : InsertMem<"inserti8", IC32Y, GR32, azextloadi8, bdxaddr20pair>;
902
903defm : InsertMem<"inserti8", IC,  GR64, azextloadi8, bdxaddr12pair>;
904defm : InsertMem<"inserti8", ICY, GR64, azextloadi8, bdxaddr20pair>;
905
906// Insert characters under mask -- not (yet) used for codegen.
907let Defs = [CC] in {
908  defm ICM : TernaryRSPair<"icm", 0xBF, 0xEB81, GR32, 0>;
909  def ICMH : TernaryRSY<"icmh", 0xEB80, GRH32, 0>;
910}
911
912// Insertions of a 16-bit immediate, leaving other bits unaffected.
913// We don't have or_as_insert equivalents of these operations because
914// OI is available instead.
915//
916// IIxMux expands to II[LH]x, depending on the choice of register.
917def IILMux : BinaryRIPseudo<insertll, GRX32, imm32ll16>,
918             Requires<[FeatureHighWord]>;
919def IIHMux : BinaryRIPseudo<insertlh, GRX32, imm32lh16>,
920             Requires<[FeatureHighWord]>;
921def IILL : BinaryRI<"iill", 0xA53, insertll, GR32, imm32ll16>;
922def IILH : BinaryRI<"iilh", 0xA52, insertlh, GR32, imm32lh16>;
923def IIHL : BinaryRI<"iihl", 0xA51, insertll, GRH32, imm32ll16>;
924def IIHH : BinaryRI<"iihh", 0xA50, insertlh, GRH32, imm32lh16>;
925def IILL64 : BinaryAliasRI<insertll, GR64, imm64ll16>;
926def IILH64 : BinaryAliasRI<insertlh, GR64, imm64lh16>;
927def IIHL64 : BinaryAliasRI<inserthl, GR64, imm64hl16>;
928def IIHH64 : BinaryAliasRI<inserthh, GR64, imm64hh16>;
929
930// ...likewise for 32-bit immediates.  For GR32s this is a general
931// full-width move.  (We use IILF rather than something like LLILF
932// for 32-bit moves because IILF leaves the upper 32 bits of the
933// GR64 unchanged.)
934let isAsCheapAsAMove = 1, isMoveImm = 1, isReMaterializable = 1 in {
935  def IIFMux : UnaryRIPseudo<bitconvert, GRX32, uimm32>,
936               Requires<[FeatureHighWord]>;
937  def IILF : UnaryRIL<"iilf", 0xC09, bitconvert, GR32, uimm32>;
938  def IIHF : UnaryRIL<"iihf", 0xC08, bitconvert, GRH32, uimm32>;
939}
940def IILF64 : BinaryAliasRIL<insertlf, GR64, imm64lf32>;
941def IIHF64 : BinaryAliasRIL<inserthf, GR64, imm64hf32>;
942
943// An alternative model of inserthf, with the first operand being
944// a zero-extended value.
945def : Pat<(or (zext32 GR32:$src), imm64hf32:$imm),
946          (IIHF64 (INSERT_SUBREG (i64 (IMPLICIT_DEF)), GR32:$src, subreg_l32),
947                  imm64hf32:$imm)>;
948
949//===----------------------------------------------------------------------===//
950// Addition
951//===----------------------------------------------------------------------===//
952
953// Addition producing a signed overflow flag.
954let Defs = [CC], CCValues = 0xF, CCIfNoSignedWrap = 1 in {
955  // Addition of a register.
956  let isCommutable = 1 in {
957    defm AR : BinaryRRAndK<"ar", 0x1A, 0xB9F8, z_sadd, GR32, GR32>;
958    defm AGR : BinaryRREAndK<"agr", 0xB908, 0xB9E8, z_sadd, GR64, GR64>;
959  }
960  def AGFR : BinaryRRE<"agfr", 0xB918, null_frag, GR64, GR32>;
961
962  // Addition to a high register.
963  def AHHHR : BinaryRRFa<"ahhhr", 0xB9C8, null_frag, GRH32, GRH32, GRH32>,
964              Requires<[FeatureHighWord]>;
965  def AHHLR : BinaryRRFa<"ahhlr", 0xB9D8, null_frag, GRH32, GRH32, GR32>,
966              Requires<[FeatureHighWord]>;
967
968  // Addition of signed 16-bit immediates.
969  defm AHIMux : BinaryRIAndKPseudo<"ahimux", z_sadd, GRX32, imm32sx16>;
970  defm AHI  : BinaryRIAndK<"ahi",  0xA7A, 0xECD8, z_sadd, GR32, imm32sx16>;
971  defm AGHI : BinaryRIAndK<"aghi", 0xA7B, 0xECD9, z_sadd, GR64, imm64sx16>;
972
973  // Addition of signed 32-bit immediates.
974  def AFIMux : BinaryRIPseudo<z_sadd, GRX32, simm32>,
975               Requires<[FeatureHighWord]>;
976  def AFI  : BinaryRIL<"afi",  0xC29, z_sadd, GR32, simm32>;
977  def AIH  : BinaryRIL<"aih",  0xCC8, z_sadd, GRH32, simm32>,
978             Requires<[FeatureHighWord]>;
979  def AGFI : BinaryRIL<"agfi", 0xC28, z_sadd, GR64, imm64sx32>;
980
981  // Addition of memory.
982  defm AH  : BinaryRXPair<"ah", 0x4A, 0xE37A, z_sadd, GR32, asextloadi16, 2>;
983  defm A   : BinaryRXPairAndPseudo<"a",  0x5A, 0xE35A, z_sadd, GR32, load, 4>;
984  def  AGH : BinaryRXY<"agh", 0xE338, z_sadd, GR64, asextloadi16, 2>,
985             Requires<[FeatureMiscellaneousExtensions2]>;
986  def  AGF : BinaryRXY<"agf", 0xE318, z_sadd, GR64, asextloadi32, 4>;
987  defm AG  : BinaryRXYAndPseudo<"ag",  0xE308, z_sadd, GR64, load, 8>;
988
989  // Addition to memory.
990  def ASI  : BinarySIY<"asi",  0xEB6A, add, imm32sx8>;
991  def AGSI : BinarySIY<"agsi", 0xEB7A, add, imm64sx8>;
992}
993defm : SXB<z_sadd, GR64, AGFR>;
994
995// Addition producing a carry.
996let Defs = [CC], CCValues = 0xF, IsLogical = 1 in {
997  // Addition of a register.
998  let isCommutable = 1 in {
999    defm ALR : BinaryRRAndK<"alr", 0x1E, 0xB9FA, z_uadd, GR32, GR32>;
1000    defm ALGR : BinaryRREAndK<"algr", 0xB90A, 0xB9EA, z_uadd, GR64, GR64>;
1001  }
1002  def ALGFR : BinaryRRE<"algfr", 0xB91A, null_frag, GR64, GR32>;
1003
1004  // Addition to a high register.
1005  def ALHHHR : BinaryRRFa<"alhhhr", 0xB9CA, null_frag, GRH32, GRH32, GRH32>,
1006               Requires<[FeatureHighWord]>;
1007  def ALHHLR : BinaryRRFa<"alhhlr", 0xB9DA, null_frag, GRH32, GRH32, GR32>,
1008               Requires<[FeatureHighWord]>;
1009
1010  // Addition of signed 16-bit immediates.
1011  def ALHSIK  : BinaryRIE<"alhsik",  0xECDA, z_uadd, GR32, imm32sx16>,
1012                Requires<[FeatureDistinctOps]>;
1013  def ALGHSIK : BinaryRIE<"alghsik", 0xECDB, z_uadd, GR64, imm64sx16>,
1014                Requires<[FeatureDistinctOps]>;
1015
1016  // Addition of unsigned 32-bit immediates.
1017  def ALFI  : BinaryRIL<"alfi",  0xC2B, z_uadd, GR32, uimm32>;
1018  def ALGFI : BinaryRIL<"algfi", 0xC2A, z_uadd, GR64, imm64zx32>;
1019
1020  // Addition of signed 32-bit immediates.
1021  def ALSIH : BinaryRIL<"alsih", 0xCCA, null_frag, GRH32, simm32>,
1022              Requires<[FeatureHighWord]>;
1023
1024  // Addition of memory.
1025  defm AL   : BinaryRXPairAndPseudo<"al", 0x5E, 0xE35E, z_uadd, GR32, load, 4>;
1026  def  ALGF : BinaryRXY<"algf", 0xE31A, z_uadd, GR64, azextloadi32, 4>;
1027  defm ALG  : BinaryRXYAndPseudo<"alg",  0xE30A, z_uadd, GR64, load, 8>;
1028
1029  // Addition to memory.
1030  def ALSI  : BinarySIY<"alsi",  0xEB6E, null_frag, imm32sx8>;
1031  def ALGSI : BinarySIY<"algsi", 0xEB7E, null_frag, imm64sx8>;
1032}
1033defm : ZXB<z_uadd, GR64, ALGFR>;
1034
1035// Addition producing and using a carry.
1036let Defs = [CC], Uses = [CC], CCValues = 0xF, IsLogical = 1 in {
1037  // Addition of a register.
1038  def ALCR  : BinaryRRE<"alcr",  0xB998, z_addcarry, GR32, GR32>;
1039  def ALCGR : BinaryRRE<"alcgr", 0xB988, z_addcarry, GR64, GR64>;
1040
1041  // Addition of memory.
1042  def ALC  : BinaryRXY<"alc",  0xE398, z_addcarry, GR32, load, 4>;
1043  def ALCG : BinaryRXY<"alcg", 0xE388, z_addcarry, GR64, load, 8>;
1044}
1045
1046// Addition that does not modify the condition code.
1047def ALSIHN : BinaryRIL<"alsihn", 0xCCB, null_frag, GRH32, simm32>,
1048             Requires<[FeatureHighWord]>;
1049
1050
1051//===----------------------------------------------------------------------===//
1052// Subtraction
1053//===----------------------------------------------------------------------===//
1054
1055// Subtraction producing a signed overflow flag.
1056let Defs = [CC], CCValues = 0xF, CompareZeroCCMask = 0x8,
1057    CCIfNoSignedWrap = 1 in {
1058  // Subtraction of a register.
1059  defm SR : BinaryRRAndK<"sr", 0x1B, 0xB9F9, z_ssub, GR32, GR32>;
1060  def SGFR : BinaryRRE<"sgfr", 0xB919, null_frag, GR64, GR32>;
1061  defm SGR : BinaryRREAndK<"sgr", 0xB909, 0xB9E9, z_ssub, GR64, GR64>;
1062
1063  // Subtraction from a high register.
1064  def SHHHR : BinaryRRFa<"shhhr", 0xB9C9, null_frag, GRH32, GRH32, GRH32>,
1065              Requires<[FeatureHighWord]>;
1066  def SHHLR : BinaryRRFa<"shhlr", 0xB9D9, null_frag, GRH32, GRH32, GR32>,
1067              Requires<[FeatureHighWord]>;
1068
1069  // Subtraction of memory.
1070  defm SH  : BinaryRXPair<"sh", 0x4B, 0xE37B, z_ssub, GR32, asextloadi16, 2>;
1071  defm S   : BinaryRXPairAndPseudo<"s", 0x5B, 0xE35B, z_ssub, GR32, load, 4>;
1072  def  SGH : BinaryRXY<"sgh", 0xE339, z_ssub, GR64, asextloadi16, 2>,
1073             Requires<[FeatureMiscellaneousExtensions2]>;
1074  def  SGF : BinaryRXY<"sgf", 0xE319, z_ssub, GR64, asextloadi32, 4>;
1075  defm SG  : BinaryRXYAndPseudo<"sg",  0xE309, z_ssub, GR64, load, 8>;
1076}
1077defm : SXB<z_ssub, GR64, SGFR>;
1078
1079// Subtracting an immediate is the same as adding the negated immediate.
1080let AddedComplexity = 1 in {
1081  def : Pat<(z_ssub GR32:$src1, imm32sx16n:$src2),
1082            (AHIMux GR32:$src1, imm32sx16n:$src2)>,
1083        Requires<[FeatureHighWord]>;
1084  def : Pat<(z_ssub GR32:$src1, simm32n:$src2),
1085            (AFIMux GR32:$src1, simm32n:$src2)>,
1086        Requires<[FeatureHighWord]>;
1087  def : Pat<(z_ssub GR32:$src1, imm32sx16n:$src2),
1088            (AHI GR32:$src1, imm32sx16n:$src2)>;
1089  def : Pat<(z_ssub GR32:$src1, simm32n:$src2),
1090            (AFI GR32:$src1, simm32n:$src2)>;
1091  def : Pat<(z_ssub GR64:$src1, imm64sx16n:$src2),
1092            (AGHI GR64:$src1, imm64sx16n:$src2)>;
1093  def : Pat<(z_ssub GR64:$src1, imm64sx32n:$src2),
1094            (AGFI GR64:$src1, imm64sx32n:$src2)>;
1095}
1096
1097// And vice versa in one special case, where we need to load a
1098// constant into a register in any case, but the negated constant
1099// requires fewer instructions to load.
1100def : Pat<(z_saddo GR64:$src1, imm64lh16n:$src2),
1101          (SGR GR64:$src1, (LLILH imm64lh16n:$src2))>;
1102def : Pat<(z_saddo GR64:$src1, imm64lf32n:$src2),
1103          (SGR GR64:$src1, (LLILF imm64lf32n:$src2))>;
1104
1105// Subtraction producing a carry.
1106let Defs = [CC], CCValues = 0x7, IsLogical = 1 in {
1107  // Subtraction of a register.
1108  defm SLR : BinaryRRAndK<"slr", 0x1F, 0xB9FB, z_usub, GR32, GR32>;
1109  def SLGFR : BinaryRRE<"slgfr", 0xB91B, null_frag, GR64, GR32>;
1110  defm SLGR : BinaryRREAndK<"slgr", 0xB90B, 0xB9EB, z_usub, GR64, GR64>;
1111
1112  // Subtraction from a high register.
1113  def SLHHHR : BinaryRRFa<"slhhhr", 0xB9CB, null_frag, GRH32, GRH32, GRH32>,
1114               Requires<[FeatureHighWord]>;
1115  def SLHHLR : BinaryRRFa<"slhhlr", 0xB9DB, null_frag, GRH32, GRH32, GR32>,
1116               Requires<[FeatureHighWord]>;
1117
1118  // Subtraction of unsigned 32-bit immediates.
1119  def SLFI  : BinaryRIL<"slfi",  0xC25, z_usub, GR32, uimm32>;
1120  def SLGFI : BinaryRIL<"slgfi", 0xC24, z_usub, GR64, imm64zx32>;
1121
1122  // Subtraction of memory.
1123  defm SL   : BinaryRXPairAndPseudo<"sl", 0x5F, 0xE35F, z_usub, GR32, load, 4>;
1124  def  SLGF : BinaryRXY<"slgf", 0xE31B, z_usub, GR64, azextloadi32, 4>;
1125  defm SLG  : BinaryRXYAndPseudo<"slg",  0xE30B, z_usub, GR64, load, 8>;
1126}
1127defm : ZXB<z_usub, GR64, SLGFR>;
1128
1129// Subtracting an immediate is the same as adding the negated immediate.
1130let AddedComplexity = 1 in {
1131  def : Pat<(z_usub GR32:$src1, imm32sx16n:$src2),
1132            (ALHSIK GR32:$src1, imm32sx16n:$src2)>,
1133        Requires<[FeatureDistinctOps]>;
1134  def : Pat<(z_usub GR64:$src1, imm64sx16n:$src2),
1135            (ALGHSIK GR64:$src1, imm64sx16n:$src2)>,
1136        Requires<[FeatureDistinctOps]>;
1137}
1138
1139// And vice versa in one special case (but we prefer addition).
1140def : Pat<(add GR64:$src1, imm64zx32n:$src2),
1141          (SLGFI GR64:$src1, imm64zx32n:$src2)>;
1142
1143// Subtraction producing and using a carry.
1144let Defs = [CC], Uses = [CC], CCValues = 0xF, IsLogical = 1 in {
1145  // Subtraction of a register.
1146  def SLBR  : BinaryRRE<"slbr",  0xB999, z_subcarry, GR32, GR32>;
1147  def SLBGR : BinaryRRE<"slbgr", 0xB989, z_subcarry, GR64, GR64>;
1148
1149  // Subtraction of memory.
1150  def SLB  : BinaryRXY<"slb",  0xE399, z_subcarry, GR32, load, 4>;
1151  def SLBG : BinaryRXY<"slbg", 0xE389, z_subcarry, GR64, load, 8>;
1152}
1153
1154
1155//===----------------------------------------------------------------------===//
1156// AND
1157//===----------------------------------------------------------------------===//
1158
1159let Defs = [CC] in {
1160  // ANDs of a register.
1161  let isCommutable = 1, CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1162    defm NR : BinaryRRAndK<"nr", 0x14, 0xB9F4, and, GR32, GR32>;
1163    defm NGR : BinaryRREAndK<"ngr", 0xB980, 0xB9E4, and, GR64, GR64>;
1164  }
1165
1166  let isConvertibleToThreeAddress = 1 in {
1167    // ANDs of a 16-bit immediate, leaving other bits unaffected.
1168    // The CC result only reflects the 16-bit field, not the full register.
1169    //
1170    // NIxMux expands to NI[LH]x, depending on the choice of register.
1171    def NILMux : BinaryRIPseudo<and, GRX32, imm32ll16c>,
1172                 Requires<[FeatureHighWord]>;
1173    def NIHMux : BinaryRIPseudo<and, GRX32, imm32lh16c>,
1174                 Requires<[FeatureHighWord]>;
1175    def NILL : BinaryRI<"nill", 0xA57, and, GR32, imm32ll16c>;
1176    def NILH : BinaryRI<"nilh", 0xA56, and, GR32, imm32lh16c>;
1177    def NIHL : BinaryRI<"nihl", 0xA55, and, GRH32, imm32ll16c>;
1178    def NIHH : BinaryRI<"nihh", 0xA54, and, GRH32, imm32lh16c>;
1179    def NILL64 : BinaryAliasRI<and, GR64, imm64ll16c>;
1180    def NILH64 : BinaryAliasRI<and, GR64, imm64lh16c>;
1181    def NIHL64 : BinaryAliasRI<and, GR64, imm64hl16c>;
1182    def NIHH64 : BinaryAliasRI<and, GR64, imm64hh16c>;
1183
1184    // ANDs of a 32-bit immediate, leaving other bits unaffected.
1185    // The CC result only reflects the 32-bit field, which means we can
1186    // use it as a zero indicator for i32 operations but not otherwise.
1187    let CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1188      // Expands to NILF or NIHF, depending on the choice of register.
1189      def NIFMux : BinaryRIPseudo<and, GRX32, uimm32>,
1190                   Requires<[FeatureHighWord]>;
1191      def NILF : BinaryRIL<"nilf", 0xC0B, and, GR32, uimm32>;
1192      def NIHF : BinaryRIL<"nihf", 0xC0A, and, GRH32, uimm32>;
1193    }
1194    def NILF64 : BinaryAliasRIL<and, GR64, imm64lf32c>;
1195    def NIHF64 : BinaryAliasRIL<and, GR64, imm64hf32c>;
1196  }
1197
1198  // ANDs of memory.
1199  let CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1200    defm N  : BinaryRXPairAndPseudo<"n", 0x54, 0xE354, and, GR32, load, 4>;
1201    defm NG : BinaryRXYAndPseudo<"ng", 0xE380, and, GR64, load, 8>;
1202  }
1203
1204  // AND to memory
1205  defm NI : BinarySIPair<"ni", 0x94, 0xEB54, null_frag, imm32zx8>;
1206
1207  // Block AND.
1208  let mayLoad = 1, mayStore = 1 in
1209    defm NC : MemorySS<"nc", 0xD4, z_nc>;
1210}
1211defm : RMWIByte<and, bdaddr12pair, NI>;
1212defm : RMWIByte<and, bdaddr20pair, NIY>;
1213
1214//===----------------------------------------------------------------------===//
1215// OR
1216//===----------------------------------------------------------------------===//
1217
1218let Defs = [CC] in {
1219  // ORs of a register.
1220  let isCommutable = 1, CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1221    defm OR : BinaryRRAndK<"or", 0x16, 0xB9F6, or, GR32, GR32>;
1222    defm OGR : BinaryRREAndK<"ogr", 0xB981, 0xB9E6, or, GR64, GR64>;
1223  }
1224
1225  // ORs of a 16-bit immediate, leaving other bits unaffected.
1226  // The CC result only reflects the 16-bit field, not the full register.
1227  //
1228  // OIxMux expands to OI[LH]x, depending on the choice of register.
1229  def OILMux : BinaryRIPseudo<or, GRX32, imm32ll16>,
1230               Requires<[FeatureHighWord]>;
1231  def OIHMux : BinaryRIPseudo<or, GRX32, imm32lh16>,
1232               Requires<[FeatureHighWord]>;
1233  def OILL : BinaryRI<"oill", 0xA5B, or, GR32, imm32ll16>;
1234  def OILH : BinaryRI<"oilh", 0xA5A, or, GR32, imm32lh16>;
1235  def OIHL : BinaryRI<"oihl", 0xA59, or, GRH32, imm32ll16>;
1236  def OIHH : BinaryRI<"oihh", 0xA58, or, GRH32, imm32lh16>;
1237  def OILL64 : BinaryAliasRI<or, GR64, imm64ll16>;
1238  def OILH64 : BinaryAliasRI<or, GR64, imm64lh16>;
1239  def OIHL64 : BinaryAliasRI<or, GR64, imm64hl16>;
1240  def OIHH64 : BinaryAliasRI<or, GR64, imm64hh16>;
1241
1242  // ORs of a 32-bit immediate, leaving other bits unaffected.
1243  // The CC result only reflects the 32-bit field, which means we can
1244  // use it as a zero indicator for i32 operations but not otherwise.
1245  let CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1246    // Expands to OILF or OIHF, depending on the choice of register.
1247    def OIFMux : BinaryRIPseudo<or, GRX32, uimm32>,
1248                 Requires<[FeatureHighWord]>;
1249    def OILF : BinaryRIL<"oilf", 0xC0D, or, GR32, uimm32>;
1250    def OIHF : BinaryRIL<"oihf", 0xC0C, or, GRH32, uimm32>;
1251  }
1252  def OILF64 : BinaryAliasRIL<or, GR64, imm64lf32>;
1253  def OIHF64 : BinaryAliasRIL<or, GR64, imm64hf32>;
1254
1255  // ORs of memory.
1256  let CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1257    defm O  : BinaryRXPairAndPseudo<"o", 0x56, 0xE356, or, GR32, load, 4>;
1258    defm OG : BinaryRXYAndPseudo<"og", 0xE381, or, GR64, load, 8>;
1259  }
1260
1261  // OR to memory
1262  defm OI : BinarySIPair<"oi", 0x96, 0xEB56, null_frag, imm32zx8>;
1263
1264  // Block OR.
1265  let mayLoad = 1, mayStore = 1 in
1266    defm OC : MemorySS<"oc", 0xD6, z_oc>;
1267}
1268defm : RMWIByte<or, bdaddr12pair, OI>;
1269defm : RMWIByte<or, bdaddr20pair, OIY>;
1270
1271//===----------------------------------------------------------------------===//
1272// XOR
1273//===----------------------------------------------------------------------===//
1274
1275let Defs = [CC] in {
1276  // XORs of a register.
1277  let isCommutable = 1, CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1278    defm XR : BinaryRRAndK<"xr", 0x17, 0xB9F7, xor, GR32, GR32>;
1279    defm XGR : BinaryRREAndK<"xgr", 0xB982, 0xB9E7, xor, GR64, GR64>;
1280  }
1281
1282  // XORs of a 32-bit immediate, leaving other bits unaffected.
1283  // The CC result only reflects the 32-bit field, which means we can
1284  // use it as a zero indicator for i32 operations but not otherwise.
1285  let CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1286    // Expands to XILF or XIHF, depending on the choice of register.
1287    def XIFMux : BinaryRIPseudo<xor, GRX32, uimm32>,
1288                 Requires<[FeatureHighWord]>;
1289    def XILF : BinaryRIL<"xilf", 0xC07, xor, GR32, uimm32>;
1290    def XIHF : BinaryRIL<"xihf", 0xC06, xor, GRH32, uimm32>;
1291  }
1292  def XILF64 : BinaryAliasRIL<xor, GR64, imm64lf32>;
1293  def XIHF64 : BinaryAliasRIL<xor, GR64, imm64hf32>;
1294
1295  // XORs of memory.
1296  let CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1297    defm X  : BinaryRXPairAndPseudo<"x",0x57, 0xE357, xor, GR32, load, 4>;
1298    defm XG : BinaryRXYAndPseudo<"xg", 0xE382, xor, GR64, load, 8>;
1299  }
1300
1301  // XOR to memory
1302  defm XI : BinarySIPair<"xi", 0x97, 0xEB57, null_frag, imm32zx8>;
1303
1304  // Block XOR.
1305  let mayLoad = 1, mayStore = 1 in
1306    defm XC : MemorySS<"xc", 0xD7, z_xc>;
1307}
1308defm : RMWIByte<xor, bdaddr12pair, XI>;
1309defm : RMWIByte<xor, bdaddr20pair, XIY>;
1310
1311//===----------------------------------------------------------------------===//
1312// Combined logical operations
1313//===----------------------------------------------------------------------===//
1314
1315let Predicates = [FeatureMiscellaneousExtensions3],
1316    Defs = [CC] in {
1317  // AND with complement.
1318  let CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1319    def NCRK : BinaryRRFa<"ncrk", 0xB9F5, andc, GR32, GR32, GR32>;
1320    def NCGRK : BinaryRRFa<"ncgrk", 0xB9E5, andc, GR64, GR64, GR64>;
1321  }
1322
1323  // OR with complement.
1324  let CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1325    def OCRK : BinaryRRFa<"ocrk", 0xB975, orc, GR32, GR32, GR32>;
1326    def OCGRK : BinaryRRFa<"ocgrk", 0xB965, orc, GR64, GR64, GR64>;
1327  }
1328
1329  // NAND.
1330  let isCommutable = 1, CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1331    def NNRK : BinaryRRFa<"nnrk", 0xB974, nand, GR32, GR32, GR32>;
1332    def NNGRK : BinaryRRFa<"nngrk", 0xB964, nand, GR64, GR64, GR64>;
1333  }
1334
1335  // NOR.
1336  let isCommutable = 1, CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1337    def NORK : BinaryRRFa<"nork", 0xB976, nor, GR32, GR32, GR32>;
1338    def NOGRK : BinaryRRFa<"nogrk", 0xB966, nor, GR64, GR64, GR64>;
1339  }
1340
1341  // NXOR.
1342  let isCommutable = 1, CCValues = 0xC, CompareZeroCCMask = 0x8 in {
1343    def NXRK : BinaryRRFa<"nxrk", 0xB977, nxor, GR32, GR32, GR32>;
1344    def NXGRK : BinaryRRFa<"nxgrk", 0xB967, nxor, GR64, GR64, GR64>;
1345  }
1346}
1347
1348//===----------------------------------------------------------------------===//
1349// Multiplication
1350//===----------------------------------------------------------------------===//
1351
1352// Multiplication of a register, setting the condition code.  We prefer these
1353// over MS(G)R if available, even though we cannot use the condition code,
1354// since they are three-operand instructions.
1355let Predicates = [FeatureMiscellaneousExtensions2],
1356    Defs = [CC], isCommutable = 1 in {
1357  def MSRKC  : BinaryRRFa<"msrkc",  0xB9FD, mul, GR32, GR32, GR32>;
1358  def MSGRKC : BinaryRRFa<"msgrkc", 0xB9ED, mul, GR64, GR64, GR64>;
1359}
1360
1361// Multiplication of a register.
1362let isCommutable = 1 in {
1363  def MSR  : BinaryRRE<"msr",  0xB252, mul, GR32, GR32>;
1364  def MSGR : BinaryRRE<"msgr", 0xB90C, mul, GR64, GR64>;
1365}
1366def MSGFR : BinaryRRE<"msgfr", 0xB91C, null_frag, GR64, GR32>;
1367defm : SXB<mul, GR64, MSGFR>;
1368
1369// Multiplication of a signed 16-bit immediate.
1370def MHI  : BinaryRI<"mhi",  0xA7C, mul, GR32, imm32sx16>;
1371def MGHI : BinaryRI<"mghi", 0xA7D, mul, GR64, imm64sx16>;
1372
1373// Multiplication of a signed 32-bit immediate.
1374def MSFI  : BinaryRIL<"msfi",  0xC21, mul, GR32, simm32>;
1375def MSGFI : BinaryRIL<"msgfi", 0xC20, mul, GR64, imm64sx32>;
1376
1377// Multiplication of memory.
1378defm MH   : BinaryRXPair<"mh", 0x4C, 0xE37C, mul, GR32, asextloadi16, 2>;
1379defm MS   : BinaryRXPair<"ms", 0x71, 0xE351, mul, GR32, load, 4>;
1380def  MGH  : BinaryRXY<"mgh", 0xE33C, mul, GR64, asextloadi16, 2>,
1381            Requires<[FeatureMiscellaneousExtensions2]>;
1382def  MSGF : BinaryRXY<"msgf", 0xE31C, mul, GR64, asextloadi32, 4>;
1383def  MSG  : BinaryRXY<"msg",  0xE30C, mul, GR64, load, 8>;
1384
1385// Multiplication of memory, setting the condition code.
1386let Predicates = [FeatureMiscellaneousExtensions2], Defs = [CC] in {
1387  defm MSC  : BinaryRXYAndPseudo<"msc",  0xE353, null_frag, GR32, load, 4>;
1388  defm MSGC : BinaryRXYAndPseudo<"msgc", 0xE383, null_frag, GR64, load, 8>;
1389}
1390
1391// Multiplication of a register, producing two results.
1392def MR   : BinaryRR <"mr",    0x1C,   null_frag, GR128, GR32>;
1393def MGRK : BinaryRRFa<"mgrk", 0xB9EC, null_frag, GR128, GR64, GR64>,
1394           Requires<[FeatureMiscellaneousExtensions2]>;
1395def MLR  : BinaryRRE<"mlr",  0xB996, null_frag, GR128, GR32>;
1396def MLGR : BinaryRRE<"mlgr", 0xB986, null_frag, GR128, GR64>;
1397
1398def : Pat<(z_smul_lohi GR64:$src1, GR64:$src2),
1399          (MGRK GR64:$src1, GR64:$src2)>;
1400def : Pat<(z_umul_lohi GR64:$src1, GR64:$src2),
1401          (MLGR (AEXT128 GR64:$src1), GR64:$src2)>;
1402
1403// Multiplication of memory, producing two results.
1404def M   : BinaryRX <"m",   0x5C,   null_frag, GR128, load, 4>;
1405def MFY : BinaryRXY<"mfy", 0xE35C, null_frag, GR128, load, 4>;
1406def MG  : BinaryRXY<"mg",  0xE384, null_frag, GR128, load, 8>,
1407          Requires<[FeatureMiscellaneousExtensions2]>;
1408def ML  : BinaryRXY<"ml",  0xE396, null_frag, GR128, load, 4>;
1409def MLG : BinaryRXY<"mlg", 0xE386, null_frag, GR128, load, 8>;
1410
1411def : Pat<(z_smul_lohi GR64:$src1, (i64 (load bdxaddr20only:$src2))),
1412          (MG (AEXT128 GR64:$src1), bdxaddr20only:$src2)>;
1413def : Pat<(z_umul_lohi GR64:$src1, (i64 (load bdxaddr20only:$src2))),
1414          (MLG (AEXT128 GR64:$src1), bdxaddr20only:$src2)>;
1415
1416//===----------------------------------------------------------------------===//
1417// Division and remainder
1418//===----------------------------------------------------------------------===//
1419
1420let hasSideEffects = 1 in {  // Do not speculatively execute.
1421  // Division and remainder, from registers.
1422  def DR    : BinaryRR <"dr",    0x1D,   null_frag, GR128, GR32>;
1423  def DSGFR : BinaryRRE<"dsgfr", 0xB91D, null_frag, GR128, GR32>;
1424  def DSGR  : BinaryRRE<"dsgr",  0xB90D, null_frag, GR128, GR64>;
1425  def DLR   : BinaryRRE<"dlr",   0xB997, null_frag, GR128, GR32>;
1426  def DLGR  : BinaryRRE<"dlgr",  0xB987, null_frag, GR128, GR64>;
1427
1428  // Division and remainder, from memory.
1429  def D    : BinaryRX <"d",    0x5D,   null_frag, GR128, load, 4>;
1430  def DSGF : BinaryRXY<"dsgf", 0xE31D, null_frag, GR128, load, 4>;
1431  def DSG  : BinaryRXY<"dsg",  0xE30D, null_frag, GR128, load, 8>;
1432  def DL   : BinaryRXY<"dl",   0xE397, null_frag, GR128, load, 4>;
1433  def DLG  : BinaryRXY<"dlg",  0xE387, null_frag, GR128, load, 8>;
1434}
1435def : Pat<(z_sdivrem GR64:$src1, GR32:$src2),
1436          (DSGFR (AEXT128 GR64:$src1), GR32:$src2)>;
1437def : Pat<(z_sdivrem GR64:$src1, (i32 (load bdxaddr20only:$src2))),
1438          (DSGF (AEXT128 GR64:$src1), bdxaddr20only:$src2)>;
1439def : Pat<(z_sdivrem GR64:$src1, GR64:$src2),
1440          (DSGR (AEXT128 GR64:$src1), GR64:$src2)>;
1441def : Pat<(z_sdivrem GR64:$src1, (i64 (load bdxaddr20only:$src2))),
1442          (DSG (AEXT128 GR64:$src1), bdxaddr20only:$src2)>;
1443
1444def : Pat<(z_udivrem GR32:$src1, GR32:$src2),
1445          (DLR (ZEXT128 (INSERT_SUBREG (i64 (IMPLICIT_DEF)), GR32:$src1,
1446                                       subreg_l32)), GR32:$src2)>;
1447def : Pat<(z_udivrem GR32:$src1, (i32 (load bdxaddr20only:$src2))),
1448          (DL (ZEXT128 (INSERT_SUBREG (i64 (IMPLICIT_DEF)), GR32:$src1,
1449                                      subreg_l32)), bdxaddr20only:$src2)>;
1450def : Pat<(z_udivrem GR64:$src1, GR64:$src2),
1451          (DLGR (ZEXT128 GR64:$src1), GR64:$src2)>;
1452def : Pat<(z_udivrem GR64:$src1, (i64 (load bdxaddr20only:$src2))),
1453          (DLG (ZEXT128 GR64:$src1), bdxaddr20only:$src2)>;
1454
1455//===----------------------------------------------------------------------===//
1456// Shifts
1457//===----------------------------------------------------------------------===//
1458
1459// Logical shift left.
1460defm SLL : BinaryRSAndK<"sll", 0x89, 0xEBDF, shiftop<shl>, GR32>;
1461def SLLG : BinaryRSY<"sllg", 0xEB0D, shiftop<shl>, GR64>;
1462def SLDL : BinaryRS<"sldl", 0x8D, null_frag, GR128>;
1463
1464// Arithmetic shift left.
1465let Defs = [CC] in {
1466  defm SLA : BinaryRSAndK<"sla", 0x8B, 0xEBDD, null_frag, GR32>;
1467  def SLAG : BinaryRSY<"slag", 0xEB0B, null_frag, GR64>;
1468  def SLDA : BinaryRS<"slda", 0x8F, null_frag, GR128>;
1469}
1470
1471// Logical shift right.
1472defm SRL : BinaryRSAndK<"srl", 0x88, 0xEBDE, shiftop<srl>, GR32>;
1473def SRLG : BinaryRSY<"srlg", 0xEB0C, shiftop<srl>, GR64>;
1474def SRDL : BinaryRS<"srdl", 0x8C, null_frag, GR128>;
1475
1476// Arithmetic shift right.
1477let Defs = [CC], CCValues = 0xE, CompareZeroCCMask = 0xE in {
1478  defm SRA : BinaryRSAndK<"sra", 0x8A, 0xEBDC, shiftop<sra>, GR32>;
1479  def SRAG : BinaryRSY<"srag", 0xEB0A, shiftop<sra>, GR64>;
1480  def SRDA : BinaryRS<"srda", 0x8E, null_frag, GR128>;
1481}
1482
1483// Rotate left.
1484def RLL  : BinaryRSY<"rll",  0xEB1D, shiftop<rotl>, GR32>;
1485def RLLG : BinaryRSY<"rllg", 0xEB1C, shiftop<rotl>, GR64>;
1486
1487// Rotate second operand left and inserted selected bits into first operand.
1488// These can act like 32-bit operands provided that the constant start and
1489// end bits (operands 2 and 3) are in the range [32, 64).
1490let Defs = [CC] in {
1491  let isCodeGenOnly = 1 in
1492    def RISBG32 : RotateSelectRIEf<"risbg", 0xEC55, GR32, GR32>;
1493  let CCValues = 0xE, CompareZeroCCMask = 0xE in
1494    def RISBG : RotateSelectRIEf<"risbg", 0xEC55, GR64, GR64>;
1495}
1496
1497// On zEC12 we have a variant of RISBG that does not set CC.
1498let Predicates = [FeatureMiscellaneousExtensions] in
1499  def RISBGN : RotateSelectRIEf<"risbgn", 0xEC59, GR64, GR64>;
1500
1501// Forms of RISBG that only affect one word of the destination register.
1502// They do not set CC.
1503let Predicates = [FeatureHighWord] in {
1504  def RISBMux : RotateSelectRIEfPseudo<GRX32, GRX32>;
1505  def RISBLL  : RotateSelectAliasRIEf<GR32,  GR32>;
1506  def RISBLH  : RotateSelectAliasRIEf<GR32,  GRH32>;
1507  def RISBHL  : RotateSelectAliasRIEf<GRH32, GR32>;
1508  def RISBHH  : RotateSelectAliasRIEf<GRH32, GRH32>;
1509  def RISBLG  : RotateSelectRIEf<"risblg", 0xEC51, GR32, GR64>;
1510  def RISBHG  : RotateSelectRIEf<"risbhg", 0xEC5D, GRH32, GR64>;
1511}
1512
1513// Rotate second operand left and perform a logical operation with selected
1514// bits of the first operand.  The CC result only describes the selected bits,
1515// so isn't useful for a full comparison against zero.
1516let Defs = [CC] in {
1517  def RNSBG : RotateSelectRIEf<"rnsbg", 0xEC54, GR64, GR64>;
1518  def ROSBG : RotateSelectRIEf<"rosbg", 0xEC56, GR64, GR64>;
1519  def RXSBG : RotateSelectRIEf<"rxsbg", 0xEC57, GR64, GR64>;
1520}
1521
1522//===----------------------------------------------------------------------===//
1523// Comparison
1524//===----------------------------------------------------------------------===//
1525
1526// Signed comparisons.  We put these before the unsigned comparisons because
1527// some of the signed forms have COMPARE AND BRANCH equivalents whereas none
1528// of the unsigned forms do.
1529let Defs = [CC], CCValues = 0xE in {
1530  // Comparison with a register.
1531  def CR   : CompareRR <"cr",   0x19,   z_scmp,    GR32, GR32>;
1532  def CGFR : CompareRRE<"cgfr", 0xB930, null_frag, GR64, GR32>;
1533  def CGR  : CompareRRE<"cgr",  0xB920, z_scmp,    GR64, GR64>;
1534
1535  // Comparison with a high register.
1536  def CHHR : CompareRRE<"chhr", 0xB9CD, null_frag, GRH32, GRH32>,
1537             Requires<[FeatureHighWord]>;
1538  def CHLR : CompareRRE<"chlr", 0xB9DD, null_frag, GRH32, GR32>,
1539             Requires<[FeatureHighWord]>;
1540
1541  // Comparison with a signed 16-bit immediate.  CHIMux expands to CHI or CIH,
1542  // depending on the choice of register.
1543  def CHIMux : CompareRIPseudo<z_scmp, GRX32, imm32sx16>,
1544               Requires<[FeatureHighWord]>;
1545  def CHI  : CompareRI<"chi",  0xA7E, z_scmp, GR32, imm32sx16>;
1546  def CGHI : CompareRI<"cghi", 0xA7F, z_scmp, GR64, imm64sx16>;
1547
1548  // Comparison with a signed 32-bit immediate.  CFIMux expands to CFI or CIH,
1549  // depending on the choice of register.
1550  def CFIMux : CompareRIPseudo<z_scmp, GRX32, simm32>,
1551               Requires<[FeatureHighWord]>;
1552  def CFI  : CompareRIL<"cfi",  0xC2D, z_scmp, GR32, simm32>;
1553  def CIH  : CompareRIL<"cih",  0xCCD, z_scmp, GRH32, simm32>,
1554             Requires<[FeatureHighWord]>;
1555  def CGFI : CompareRIL<"cgfi", 0xC2C, z_scmp, GR64, imm64sx32>;
1556
1557  // Comparison with memory.
1558  defm CH    : CompareRXPair<"ch", 0x49, 0xE379, z_scmp, GR32, asextloadi16, 2>;
1559  def  CMux  : CompareRXYPseudo<z_scmp, GRX32, load, 4>,
1560               Requires<[FeatureHighWord]>;
1561  defm C     : CompareRXPair<"c",  0x59, 0xE359, z_scmp, GR32, load, 4>;
1562  def  CHF   : CompareRXY<"chf", 0xE3CD, z_scmp, GRH32, load, 4>,
1563               Requires<[FeatureHighWord]>;
1564  def  CGH   : CompareRXY<"cgh", 0xE334, z_scmp, GR64, asextloadi16, 2>;
1565  def  CGF   : CompareRXY<"cgf", 0xE330, z_scmp, GR64, asextloadi32, 4>;
1566  def  CG    : CompareRXY<"cg",  0xE320, z_scmp, GR64, load, 8>;
1567  def  CHRL  : CompareRILPC<"chrl",  0xC65, z_scmp, GR32, aligned_asextloadi16>;
1568  def  CRL   : CompareRILPC<"crl",   0xC6D, z_scmp, GR32, aligned_load>;
1569  def  CGHRL : CompareRILPC<"cghrl", 0xC64, z_scmp, GR64, aligned_asextloadi16>;
1570  def  CGFRL : CompareRILPC<"cgfrl", 0xC6C, z_scmp, GR64, aligned_asextloadi32>;
1571  def  CGRL  : CompareRILPC<"cgrl",  0xC68, z_scmp, GR64, aligned_load>;
1572
1573  // Comparison between memory and a signed 16-bit immediate.
1574  def CHHSI : CompareSIL<"chhsi", 0xE554, z_scmp, asextloadi16, imm32sx16>;
1575  def CHSI  : CompareSIL<"chsi",  0xE55C, z_scmp, load, imm32sx16>;
1576  def CGHSI : CompareSIL<"cghsi", 0xE558, z_scmp, load, imm64sx16>;
1577}
1578defm : SXB<z_scmp, GR64, CGFR>;
1579
1580// Unsigned comparisons.
1581let Defs = [CC], CCValues = 0xE, IsLogical = 1 in {
1582  // Comparison with a register.
1583  def CLR   : CompareRR <"clr",   0x15,   z_ucmp,    GR32, GR32>;
1584  def CLGFR : CompareRRE<"clgfr", 0xB931, null_frag, GR64, GR32>;
1585  def CLGR  : CompareRRE<"clgr",  0xB921, z_ucmp,    GR64, GR64>;
1586
1587  // Comparison with a high register.
1588  def CLHHR : CompareRRE<"clhhr", 0xB9CF, null_frag, GRH32, GRH32>,
1589              Requires<[FeatureHighWord]>;
1590  def CLHLR : CompareRRE<"clhlr", 0xB9DF, null_frag, GRH32, GR32>,
1591              Requires<[FeatureHighWord]>;
1592
1593  // Comparison with an unsigned 32-bit immediate.  CLFIMux expands to CLFI
1594  // or CLIH, depending on the choice of register.
1595  def CLFIMux : CompareRIPseudo<z_ucmp, GRX32, uimm32>,
1596                Requires<[FeatureHighWord]>;
1597  def CLFI  : CompareRIL<"clfi",  0xC2F, z_ucmp, GR32, uimm32>;
1598  def CLIH  : CompareRIL<"clih",  0xCCF, z_ucmp, GRH32, uimm32>,
1599              Requires<[FeatureHighWord]>;
1600  def CLGFI : CompareRIL<"clgfi", 0xC2E, z_ucmp, GR64, imm64zx32>;
1601
1602  // Comparison with memory.
1603  def  CLMux  : CompareRXYPseudo<z_ucmp, GRX32, load, 4>,
1604                Requires<[FeatureHighWord]>;
1605  defm CL     : CompareRXPair<"cl", 0x55, 0xE355, z_ucmp, GR32, load, 4>;
1606  def  CLHF   : CompareRXY<"clhf", 0xE3CF, z_ucmp, GRH32, load, 4>,
1607                Requires<[FeatureHighWord]>;
1608  def  CLGF   : CompareRXY<"clgf", 0xE331, z_ucmp, GR64, azextloadi32, 4>;
1609  def  CLG    : CompareRXY<"clg",  0xE321, z_ucmp, GR64, load, 8>;
1610  def  CLHRL  : CompareRILPC<"clhrl",  0xC67, z_ucmp, GR32,
1611                             aligned_azextloadi16>;
1612  def  CLRL   : CompareRILPC<"clrl",   0xC6F, z_ucmp, GR32,
1613                             aligned_load>;
1614  def  CLGHRL : CompareRILPC<"clghrl", 0xC66, z_ucmp, GR64,
1615                             aligned_azextloadi16>;
1616  def  CLGFRL : CompareRILPC<"clgfrl", 0xC6E, z_ucmp, GR64,
1617                             aligned_azextloadi32>;
1618  def  CLGRL  : CompareRILPC<"clgrl",  0xC6A, z_ucmp, GR64,
1619                             aligned_load>;
1620
1621  // Comparison between memory and an unsigned 8-bit immediate.
1622  defm CLI : CompareSIPair<"cli", 0x95, 0xEB55, z_ucmp, azextloadi8, imm32zx8>;
1623
1624  // Comparison between memory and an unsigned 16-bit immediate.
1625  def CLHHSI : CompareSIL<"clhhsi", 0xE555, z_ucmp, azextloadi16, imm32zx16>;
1626  def CLFHSI : CompareSIL<"clfhsi", 0xE55D, z_ucmp, load, imm32zx16>;
1627  def CLGHSI : CompareSIL<"clghsi", 0xE559, z_ucmp, load, imm64zx16>;
1628}
1629defm : ZXB<z_ucmp, GR64, CLGFR>;
1630
1631// Memory-to-memory comparison.
1632let mayLoad = 1, Defs = [CC] in {
1633  defm CLC : CompareMemorySS<"clc", 0xD5, z_clc>;
1634  def CLCL  : SideEffectBinaryMemMemRR<"clcl", 0x0F, GR128, GR128>;
1635  def CLCLE : SideEffectTernaryMemMemRS<"clcle", 0xA9, GR128, GR128>;
1636  def CLCLU : SideEffectTernaryMemMemRSY<"clclu", 0xEB8F, GR128, GR128>;
1637}
1638
1639// String comparison.
1640let mayLoad = 1, Defs = [CC] in
1641  defm CLST : StringRRE<"clst", 0xB25D, z_strcmp>;
1642
1643// Test under mask.
1644let Defs = [CC] in {
1645  // TMxMux expands to TM[LH]x, depending on the choice of register.
1646  def TMLMux : CompareRIPseudo<z_tm_reg, GRX32, imm32ll16>,
1647               Requires<[FeatureHighWord]>;
1648  def TMHMux : CompareRIPseudo<z_tm_reg, GRX32, imm32lh16>,
1649               Requires<[FeatureHighWord]>;
1650  def TMLL : CompareRI<"tmll", 0xA71, z_tm_reg, GR32, imm32ll16>;
1651  def TMLH : CompareRI<"tmlh", 0xA70, z_tm_reg, GR32, imm32lh16>;
1652  def TMHL : CompareRI<"tmhl", 0xA73, z_tm_reg, GRH32, imm32ll16>;
1653  def TMHH : CompareRI<"tmhh", 0xA72, z_tm_reg, GRH32, imm32lh16>;
1654
1655  def TMLL64 : CompareAliasRI<z_tm_reg, GR64, imm64ll16>;
1656  def TMLH64 : CompareAliasRI<z_tm_reg, GR64, imm64lh16>;
1657  def TMHL64 : CompareAliasRI<z_tm_reg, GR64, imm64hl16>;
1658  def TMHH64 : CompareAliasRI<z_tm_reg, GR64, imm64hh16>;
1659
1660  defm TM : CompareSIPair<"tm", 0x91, 0xEB51, z_tm_mem, anyextloadi8, imm32zx8>;
1661}
1662
1663def TML : InstAlias<"tml\t$R, $I", (TMLL GR32:$R, imm32ll16:$I), 0>;
1664def TMH : InstAlias<"tmh\t$R, $I", (TMLH GR32:$R, imm32lh16:$I), 0>;
1665
1666// Compare logical characters under mask -- not (yet) used for codegen.
1667let Defs = [CC] in {
1668  defm CLM : CompareRSPair<"clm", 0xBD, 0xEB21, GR32, 0>;
1669  def CLMH : CompareRSY<"clmh", 0xEB20, GRH32, 0>;
1670}
1671
1672//===----------------------------------------------------------------------===//
1673// Prefetch and execution hint
1674//===----------------------------------------------------------------------===//
1675
1676let mayLoad = 1, mayStore = 1 in {
1677  def PFD : PrefetchRXY<"pfd", 0xE336, z_prefetch>;
1678  def PFDRL : PrefetchRILPC<"pfdrl", 0xC62, z_prefetch>;
1679}
1680
1681let Predicates = [FeatureExecutionHint], hasSideEffects = 1 in {
1682  // Branch Prediction Preload
1683  def BPP : BranchPreloadSMI<"bpp", 0xC7>;
1684  def BPRP : BranchPreloadMII<"bprp", 0xC5>;
1685
1686  // Next Instruction Access Intent
1687  def NIAI : SideEffectBinaryIE<"niai", 0xB2FA, imm32zx4, imm32zx4>;
1688}
1689
1690//===----------------------------------------------------------------------===//
1691// Atomic operations
1692//===----------------------------------------------------------------------===//
1693
1694// A serialization instruction that acts as a barrier for all memory
1695// accesses, which expands to "bcr 14, 0".
1696let hasSideEffects = 1 in
1697def Serialize : Alias<2, (outs), (ins), []>;
1698
1699// A pseudo instruction that serves as a compiler barrier.
1700let hasSideEffects = 1, hasNoSchedulingInfo = 1 in
1701def MemBarrier : Pseudo<(outs), (ins), [(z_membarrier)]>;
1702
1703let Predicates = [FeatureInterlockedAccess1], Defs = [CC] in {
1704  def LAA   : LoadAndOpRSY<"laa",   0xEBF8, atomic_load_add_32, GR32>;
1705  def LAAG  : LoadAndOpRSY<"laag",  0xEBE8, atomic_load_add_64, GR64>;
1706  def LAAL  : LoadAndOpRSY<"laal",  0xEBFA, null_frag, GR32>;
1707  def LAALG : LoadAndOpRSY<"laalg", 0xEBEA, null_frag, GR64>;
1708  def LAN   : LoadAndOpRSY<"lan",   0xEBF4, atomic_load_and_32, GR32>;
1709  def LANG  : LoadAndOpRSY<"lang",  0xEBE4, atomic_load_and_64, GR64>;
1710  def LAO   : LoadAndOpRSY<"lao",   0xEBF6, atomic_load_or_32, GR32>;
1711  def LAOG  : LoadAndOpRSY<"laog",  0xEBE6, atomic_load_or_64, GR64>;
1712  def LAX   : LoadAndOpRSY<"lax",   0xEBF7, atomic_load_xor_32, GR32>;
1713  def LAXG  : LoadAndOpRSY<"laxg",  0xEBE7, atomic_load_xor_64, GR64>;
1714}
1715
1716def ATOMIC_SWAPW   : AtomicLoadWBinaryReg<z_atomic_swapw>;
1717def ATOMIC_SWAP_32 : AtomicLoadBinaryReg32<atomic_swap_32>;
1718def ATOMIC_SWAP_64 : AtomicLoadBinaryReg64<atomic_swap_64>;
1719
1720def ATOMIC_LOADW_AR  : AtomicLoadWBinaryReg<z_atomic_loadw_add>;
1721def ATOMIC_LOADW_AFI : AtomicLoadWBinaryImm<z_atomic_loadw_add, simm32>;
1722let Predicates = [FeatureNoInterlockedAccess1] in {
1723  def ATOMIC_LOAD_AR   : AtomicLoadBinaryReg32<atomic_load_add_32>;
1724  def ATOMIC_LOAD_AHI  : AtomicLoadBinaryImm32<atomic_load_add_32, imm32sx16>;
1725  def ATOMIC_LOAD_AFI  : AtomicLoadBinaryImm32<atomic_load_add_32, simm32>;
1726  def ATOMIC_LOAD_AGR  : AtomicLoadBinaryReg64<atomic_load_add_64>;
1727  def ATOMIC_LOAD_AGHI : AtomicLoadBinaryImm64<atomic_load_add_64, imm64sx16>;
1728  def ATOMIC_LOAD_AGFI : AtomicLoadBinaryImm64<atomic_load_add_64, imm64sx32>;
1729}
1730
1731def ATOMIC_LOADW_SR : AtomicLoadWBinaryReg<z_atomic_loadw_sub>;
1732def ATOMIC_LOAD_SR  : AtomicLoadBinaryReg32<atomic_load_sub_32>;
1733def ATOMIC_LOAD_SGR : AtomicLoadBinaryReg64<atomic_load_sub_64>;
1734
1735def ATOMIC_LOADW_NR   : AtomicLoadWBinaryReg<z_atomic_loadw_and>;
1736def ATOMIC_LOADW_NILH : AtomicLoadWBinaryImm<z_atomic_loadw_and, imm32lh16c>;
1737let Predicates = [FeatureNoInterlockedAccess1] in {
1738  def ATOMIC_LOAD_NR     : AtomicLoadBinaryReg32<atomic_load_and_32>;
1739  def ATOMIC_LOAD_NILL   : AtomicLoadBinaryImm32<atomic_load_and_32,
1740                                                 imm32ll16c>;
1741  def ATOMIC_LOAD_NILH   : AtomicLoadBinaryImm32<atomic_load_and_32,
1742                                                 imm32lh16c>;
1743  def ATOMIC_LOAD_NILF   : AtomicLoadBinaryImm32<atomic_load_and_32, uimm32>;
1744  def ATOMIC_LOAD_NGR    : AtomicLoadBinaryReg64<atomic_load_and_64>;
1745  def ATOMIC_LOAD_NILL64 : AtomicLoadBinaryImm64<atomic_load_and_64,
1746                                                 imm64ll16c>;
1747  def ATOMIC_LOAD_NILH64 : AtomicLoadBinaryImm64<atomic_load_and_64,
1748                                                 imm64lh16c>;
1749  def ATOMIC_LOAD_NIHL64 : AtomicLoadBinaryImm64<atomic_load_and_64,
1750                                                 imm64hl16c>;
1751  def ATOMIC_LOAD_NIHH64 : AtomicLoadBinaryImm64<atomic_load_and_64,
1752                                                 imm64hh16c>;
1753  def ATOMIC_LOAD_NILF64 : AtomicLoadBinaryImm64<atomic_load_and_64,
1754                                                 imm64lf32c>;
1755  def ATOMIC_LOAD_NIHF64 : AtomicLoadBinaryImm64<atomic_load_and_64,
1756                                                 imm64hf32c>;
1757}
1758
1759def ATOMIC_LOADW_OR     : AtomicLoadWBinaryReg<z_atomic_loadw_or>;
1760def ATOMIC_LOADW_OILH   : AtomicLoadWBinaryImm<z_atomic_loadw_or, imm32lh16>;
1761let Predicates = [FeatureNoInterlockedAccess1] in {
1762  def ATOMIC_LOAD_OR     : AtomicLoadBinaryReg32<atomic_load_or_32>;
1763  def ATOMIC_LOAD_OILL   : AtomicLoadBinaryImm32<atomic_load_or_32, imm32ll16>;
1764  def ATOMIC_LOAD_OILH   : AtomicLoadBinaryImm32<atomic_load_or_32, imm32lh16>;
1765  def ATOMIC_LOAD_OILF   : AtomicLoadBinaryImm32<atomic_load_or_32, uimm32>;
1766  def ATOMIC_LOAD_OGR    : AtomicLoadBinaryReg64<atomic_load_or_64>;
1767  def ATOMIC_LOAD_OILL64 : AtomicLoadBinaryImm64<atomic_load_or_64, imm64ll16>;
1768  def ATOMIC_LOAD_OILH64 : AtomicLoadBinaryImm64<atomic_load_or_64, imm64lh16>;
1769  def ATOMIC_LOAD_OIHL64 : AtomicLoadBinaryImm64<atomic_load_or_64, imm64hl16>;
1770  def ATOMIC_LOAD_OIHH64 : AtomicLoadBinaryImm64<atomic_load_or_64, imm64hh16>;
1771  def ATOMIC_LOAD_OILF64 : AtomicLoadBinaryImm64<atomic_load_or_64, imm64lf32>;
1772  def ATOMIC_LOAD_OIHF64 : AtomicLoadBinaryImm64<atomic_load_or_64, imm64hf32>;
1773}
1774
1775def ATOMIC_LOADW_XR     : AtomicLoadWBinaryReg<z_atomic_loadw_xor>;
1776def ATOMIC_LOADW_XILF   : AtomicLoadWBinaryImm<z_atomic_loadw_xor, uimm32>;
1777let Predicates = [FeatureNoInterlockedAccess1] in {
1778  def ATOMIC_LOAD_XR     : AtomicLoadBinaryReg32<atomic_load_xor_32>;
1779  def ATOMIC_LOAD_XILF   : AtomicLoadBinaryImm32<atomic_load_xor_32, uimm32>;
1780  def ATOMIC_LOAD_XGR    : AtomicLoadBinaryReg64<atomic_load_xor_64>;
1781  def ATOMIC_LOAD_XILF64 : AtomicLoadBinaryImm64<atomic_load_xor_64, imm64lf32>;
1782  def ATOMIC_LOAD_XIHF64 : AtomicLoadBinaryImm64<atomic_load_xor_64, imm64hf32>;
1783}
1784
1785def ATOMIC_LOADW_NRi    : AtomicLoadWBinaryReg<z_atomic_loadw_nand>;
1786def ATOMIC_LOADW_NILHi  : AtomicLoadWBinaryImm<z_atomic_loadw_nand,
1787                                               imm32lh16c>;
1788def ATOMIC_LOAD_NRi     : AtomicLoadBinaryReg32<atomic_load_nand_32>;
1789def ATOMIC_LOAD_NILLi   : AtomicLoadBinaryImm32<atomic_load_nand_32,
1790                                                imm32ll16c>;
1791def ATOMIC_LOAD_NILHi   : AtomicLoadBinaryImm32<atomic_load_nand_32,
1792                                                imm32lh16c>;
1793def ATOMIC_LOAD_NILFi   : AtomicLoadBinaryImm32<atomic_load_nand_32, uimm32>;
1794def ATOMIC_LOAD_NGRi    : AtomicLoadBinaryReg64<atomic_load_nand_64>;
1795def ATOMIC_LOAD_NILL64i : AtomicLoadBinaryImm64<atomic_load_nand_64,
1796                                                imm64ll16c>;
1797def ATOMIC_LOAD_NILH64i : AtomicLoadBinaryImm64<atomic_load_nand_64,
1798                                                imm64lh16c>;
1799def ATOMIC_LOAD_NIHL64i : AtomicLoadBinaryImm64<atomic_load_nand_64,
1800                                                imm64hl16c>;
1801def ATOMIC_LOAD_NIHH64i : AtomicLoadBinaryImm64<atomic_load_nand_64,
1802                                                imm64hh16c>;
1803def ATOMIC_LOAD_NILF64i : AtomicLoadBinaryImm64<atomic_load_nand_64,
1804                                                imm64lf32c>;
1805def ATOMIC_LOAD_NIHF64i : AtomicLoadBinaryImm64<atomic_load_nand_64,
1806                                                imm64hf32c>;
1807
1808def ATOMIC_LOADW_MIN    : AtomicLoadWBinaryReg<z_atomic_loadw_min>;
1809def ATOMIC_LOAD_MIN_32  : AtomicLoadBinaryReg32<atomic_load_min_32>;
1810def ATOMIC_LOAD_MIN_64  : AtomicLoadBinaryReg64<atomic_load_min_64>;
1811
1812def ATOMIC_LOADW_MAX    : AtomicLoadWBinaryReg<z_atomic_loadw_max>;
1813def ATOMIC_LOAD_MAX_32  : AtomicLoadBinaryReg32<atomic_load_max_32>;
1814def ATOMIC_LOAD_MAX_64  : AtomicLoadBinaryReg64<atomic_load_max_64>;
1815
1816def ATOMIC_LOADW_UMIN   : AtomicLoadWBinaryReg<z_atomic_loadw_umin>;
1817def ATOMIC_LOAD_UMIN_32 : AtomicLoadBinaryReg32<atomic_load_umin_32>;
1818def ATOMIC_LOAD_UMIN_64 : AtomicLoadBinaryReg64<atomic_load_umin_64>;
1819
1820def ATOMIC_LOADW_UMAX   : AtomicLoadWBinaryReg<z_atomic_loadw_umax>;
1821def ATOMIC_LOAD_UMAX_32 : AtomicLoadBinaryReg32<atomic_load_umax_32>;
1822def ATOMIC_LOAD_UMAX_64 : AtomicLoadBinaryReg64<atomic_load_umax_64>;
1823
1824def ATOMIC_CMP_SWAPW
1825  : Pseudo<(outs GR32:$dst), (ins bdaddr20only:$addr, GR32:$cmp, GR32:$swap,
1826                                  ADDR32:$bitshift, ADDR32:$negbitshift,
1827                                  uimm32:$bitsize),
1828           [(set GR32:$dst,
1829                 (z_atomic_cmp_swapw bdaddr20only:$addr, GR32:$cmp, GR32:$swap,
1830                                     ADDR32:$bitshift, ADDR32:$negbitshift,
1831                                     uimm32:$bitsize))]> {
1832  let Defs = [CC];
1833  let mayLoad = 1;
1834  let mayStore = 1;
1835  let usesCustomInserter = 1;
1836  let hasNoSchedulingInfo = 1;
1837}
1838
1839// Test and set.
1840let mayLoad = 1, Defs = [CC] in
1841  def TS : StoreInherentS<"ts", 0x9300, null_frag, 1>;
1842
1843// Compare and swap.
1844let Defs = [CC] in {
1845  defm CS  : CmpSwapRSPair<"cs", 0xBA, 0xEB14, z_atomic_cmp_swap, GR32>;
1846  def  CSG : CmpSwapRSY<"csg", 0xEB30, z_atomic_cmp_swap, GR64>;
1847}
1848
1849// Compare double and swap.
1850let Defs = [CC] in {
1851  defm CDS  : CmpSwapRSPair<"cds", 0xBB, 0xEB31, null_frag, GR128>;
1852  def  CDSG : CmpSwapRSY<"cdsg", 0xEB3E, z_atomic_cmp_swap_128, GR128>;
1853}
1854
1855// Compare and swap and store.
1856let Uses = [R0L, R1D], Defs = [CC], mayStore = 1, mayLoad = 1 in
1857  def CSST : SideEffectTernarySSF<"csst", 0xC82, GR64>;
1858
1859// Perform locked operation.
1860let Uses = [R0L, R1D], Defs = [CC], mayStore = 1, mayLoad =1 in
1861  def PLO : SideEffectQuaternarySSe<"plo", 0xEE, GR64>;
1862
1863// Load/store pair from/to quadword.
1864def LPQ  : UnaryRXY<"lpq", 0xE38F, z_atomic_load_128, GR128, 16>;
1865def STPQ : StoreRXY<"stpq", 0xE38E, z_atomic_store_128, GR128, 16>;
1866
1867// Load pair disjoint.
1868let Predicates = [FeatureInterlockedAccess1], Defs = [CC] in {
1869  def LPD  : BinarySSF<"lpd", 0xC84, GR128>;
1870  def LPDG : BinarySSF<"lpdg", 0xC85, GR128>;
1871}
1872
1873//===----------------------------------------------------------------------===//
1874// Translate and convert
1875//===----------------------------------------------------------------------===//
1876
1877let mayLoad = 1, mayStore = 1 in
1878  def TR : SideEffectBinarySSa<"tr", 0xDC>;
1879
1880let mayLoad = 1, Defs = [CC, R0L, R1D] in {
1881  def TRT  : SideEffectBinarySSa<"trt", 0xDD>;
1882  def TRTR : SideEffectBinarySSa<"trtr", 0xD0>;
1883}
1884
1885let mayLoad = 1, mayStore = 1, Uses = [R0L] in
1886  def TRE : SideEffectBinaryMemMemRRE<"tre", 0xB2A5, GR128, GR64>;
1887
1888let mayLoad = 1, Uses = [R1D], Defs = [CC] in {
1889  defm TRTE  : BinaryMemRRFcOpt<"trte",  0xB9BF, GR128, GR64>;
1890  defm TRTRE : BinaryMemRRFcOpt<"trtre", 0xB9BD, GR128, GR64>;
1891}
1892
1893let mayLoad = 1, mayStore = 1, Uses = [R0L, R1D], Defs = [CC] in {
1894  defm TROO : SideEffectTernaryMemMemRRFcOpt<"troo", 0xB993, GR128, GR64>;
1895  defm TROT : SideEffectTernaryMemMemRRFcOpt<"trot", 0xB992, GR128, GR64>;
1896  defm TRTO : SideEffectTernaryMemMemRRFcOpt<"trto", 0xB991, GR128, GR64>;
1897  defm TRTT : SideEffectTernaryMemMemRRFcOpt<"trtt", 0xB990, GR128, GR64>;
1898}
1899
1900let mayLoad = 1, mayStore = 1, Defs = [CC] in {
1901  defm CU12 : SideEffectTernaryMemMemRRFcOpt<"cu12", 0xB2A7, GR128, GR128>;
1902  defm CU14 : SideEffectTernaryMemMemRRFcOpt<"cu14", 0xB9B0, GR128, GR128>;
1903  defm CU21 : SideEffectTernaryMemMemRRFcOpt<"cu21", 0xB2A6, GR128, GR128>;
1904  defm CU24 : SideEffectTernaryMemMemRRFcOpt<"cu24", 0xB9B1, GR128, GR128>;
1905  def  CU41 : SideEffectBinaryMemMemRRE<"cu41", 0xB9B2, GR128, GR128>;
1906  def  CU42 : SideEffectBinaryMemMemRRE<"cu42", 0xB9B3, GR128, GR128>;
1907
1908  let isAsmParserOnly = 1 in {
1909    defm CUUTF : SideEffectTernaryMemMemRRFcOpt<"cuutf", 0xB2A6, GR128, GR128>;
1910    defm CUTFU : SideEffectTernaryMemMemRRFcOpt<"cutfu", 0xB2A7, GR128, GR128>;
1911  }
1912}
1913
1914//===----------------------------------------------------------------------===//
1915// Message-security assist
1916//===----------------------------------------------------------------------===//
1917
1918let mayLoad = 1, mayStore = 1, Uses = [R0L, R1D], Defs = [CC] in {
1919  def KM  : SideEffectBinaryMemMemRRE<"km",  0xB92E, GR128, GR128>;
1920  def KMC : SideEffectBinaryMemMemRRE<"kmc", 0xB92F, GR128, GR128>;
1921
1922  def KIMD : SideEffectBinaryMemRRE<"kimd", 0xB93E, GR64, GR128>;
1923  def KLMD : SideEffectBinaryMemRRE<"klmd", 0xB93F, GR64, GR128>;
1924  def KMAC : SideEffectBinaryMemRRE<"kmac", 0xB91E, GR64, GR128>;
1925
1926  let Predicates = [FeatureMessageSecurityAssist4] in {
1927    def KMF   : SideEffectBinaryMemMemRRE<"kmf", 0xB92A, GR128, GR128>;
1928    def KMO   : SideEffectBinaryMemMemRRE<"kmo", 0xB92B, GR128, GR128>;
1929    def KMCTR : SideEffectTernaryMemMemMemRRFb<"kmctr", 0xB92D,
1930                                               GR128, GR128, GR128>;
1931    def PCC   : SideEffectInherentRRE<"pcc", 0xB92C>;
1932  }
1933
1934  let Predicates = [FeatureMessageSecurityAssist5] in
1935    def PPNO : SideEffectBinaryMemMemRRE<"ppno", 0xB93C, GR128, GR128>;
1936  let Predicates = [FeatureMessageSecurityAssist7], isAsmParserOnly = 1 in
1937    def PRNO : SideEffectBinaryMemMemRRE<"prno", 0xB93C, GR128, GR128>;
1938
1939  let Predicates = [FeatureMessageSecurityAssist8] in
1940    def KMA : SideEffectTernaryMemMemMemRRFb<"kma", 0xB929,
1941                                              GR128, GR128, GR128>;
1942
1943  let Predicates = [FeatureMessageSecurityAssist9] in
1944    def KDSA : SideEffectBinaryMemRRE<"kdsa", 0xB93A, GR64, GR128>;
1945}
1946
1947//===----------------------------------------------------------------------===//
1948// Guarded storage
1949//===----------------------------------------------------------------------===//
1950
1951// These instructions use and/or modify the guarded storage control
1952// registers, which we do not otherwise model, so they should have
1953// hasSideEffects.
1954let Predicates = [FeatureGuardedStorage], hasSideEffects = 1 in {
1955  def LGG : UnaryRXY<"lgg", 0xE34C, null_frag, GR64, 8>;
1956  def LLGFSG : UnaryRXY<"llgfsg", 0xE348, null_frag, GR64, 4>;
1957
1958  let mayLoad = 1 in
1959    def LGSC : SideEffectBinaryRXY<"lgsc", 0xE34D, GR64>;
1960  let mayStore = 1 in
1961    def STGSC : SideEffectBinaryRXY<"stgsc", 0xE349, GR64>;
1962}
1963
1964//===----------------------------------------------------------------------===//
1965// Decimal arithmetic
1966//===----------------------------------------------------------------------===//
1967
1968defm CVB  : BinaryRXPair<"cvb",0x4F, 0xE306, null_frag, GR32, load, 4>;
1969def  CVBG : BinaryRXY<"cvbg", 0xE30E, null_frag, GR64, load, 8>;
1970
1971defm CVD  : StoreRXPair<"cvd", 0x4E, 0xE326, null_frag, GR32, 4>;
1972def  CVDG : StoreRXY<"cvdg", 0xE32E, null_frag, GR64, 8>;
1973
1974let mayLoad = 1, mayStore = 1 in {
1975  def MVN : SideEffectBinarySSa<"mvn", 0xD1>;
1976  def MVZ : SideEffectBinarySSa<"mvz", 0xD3>;
1977  def MVO : SideEffectBinarySSb<"mvo", 0xF1>;
1978
1979  def PACK : SideEffectBinarySSb<"pack", 0xF2>;
1980  def PKA  : SideEffectBinarySSf<"pka", 0xE9>;
1981  def PKU  : SideEffectBinarySSf<"pku", 0xE1>;
1982  def UNPK : SideEffectBinarySSb<"unpk", 0xF3>;
1983  let Defs = [CC] in {
1984    def UNPKA : SideEffectBinarySSa<"unpka", 0xEA>;
1985    def UNPKU : SideEffectBinarySSa<"unpku", 0xE2>;
1986  }
1987}
1988
1989let mayLoad = 1, mayStore = 1 in {
1990  let Defs = [CC] in {
1991    def AP : SideEffectBinarySSb<"ap", 0xFA>;
1992    def SP : SideEffectBinarySSb<"sp", 0xFB>;
1993    def ZAP : SideEffectBinarySSb<"zap", 0xF8>;
1994    def SRP : SideEffectTernarySSc<"srp", 0xF0>;
1995  }
1996  def MP : SideEffectBinarySSb<"mp", 0xFC>;
1997  def DP : SideEffectBinarySSb<"dp", 0xFD>;
1998  let Defs = [CC] in {
1999    def ED : SideEffectBinarySSa<"ed", 0xDE>;
2000    def EDMK : SideEffectBinarySSa<"edmk", 0xDF>;
2001  }
2002}
2003
2004let Defs = [CC] in {
2005  def CP : CompareSSb<"cp", 0xF9>;
2006  def TP : TestRSL<"tp", 0xEBC0>;
2007}
2008
2009//===----------------------------------------------------------------------===//
2010// Access registers
2011//===----------------------------------------------------------------------===//
2012
2013// Read a 32-bit access register into a GR32.  As with all GR32 operations,
2014// the upper 32 bits of the enclosing GR64 remain unchanged, which is useful
2015// when a 64-bit address is stored in a pair of access registers.
2016def EAR : UnaryRRE<"ear", 0xB24F, null_frag, GR32, AR32>;
2017
2018// Set access register.
2019def SAR : UnaryRRE<"sar", 0xB24E, null_frag, AR32, GR32>;
2020
2021// Copy access register.
2022def CPYA : UnaryRRE<"cpya", 0xB24D, null_frag, AR32, AR32>;
2023
2024// Load address extended.
2025defm LAE : LoadAddressRXPair<"lae", 0x51, 0xE375, null_frag>;
2026
2027// Load access multiple.
2028defm LAM : LoadMultipleRSPair<"lam", 0x9A, 0xEB9A, AR32>;
2029
2030// Store access multiple.
2031defm STAM : StoreMultipleRSPair<"stam", 0x9B, 0xEB9B, AR32>;
2032
2033//===----------------------------------------------------------------------===//
2034// Program mask and addressing mode
2035//===----------------------------------------------------------------------===//
2036
2037// Extract CC and program mask into a register.  CC ends up in bits 29 and 28.
2038let Uses = [CC] in
2039  def IPM : InherentRRE<"ipm", 0xB222, GR32, z_ipm>;
2040
2041// Set CC and program mask from a register.
2042let hasSideEffects = 1, Defs = [CC] in
2043  def SPM : SideEffectUnaryRR<"spm", 0x04, GR32>;
2044
2045// Branch and link - like BAS, but also extracts CC and program mask.
2046let isCall = 1, Uses = [CC], Defs = [CC] in {
2047  def BAL  : CallRX<"bal", 0x45>;
2048  def BALR : CallRR<"balr", 0x05>;
2049}
2050
2051// Test addressing mode.
2052let Defs = [CC] in
2053  def TAM : SideEffectInherentE<"tam", 0x010B>;
2054
2055// Set addressing mode.
2056let hasSideEffects = 1 in {
2057  def SAM24 : SideEffectInherentE<"sam24", 0x010C>;
2058  def SAM31 : SideEffectInherentE<"sam31", 0x010D>;
2059  def SAM64 : SideEffectInherentE<"sam64", 0x010E>;
2060}
2061
2062// Branch and set mode.  Not really a call, but also sets an output register.
2063let isBranch = 1, isTerminator = 1, isBarrier = 1 in
2064  def BSM : CallRR<"bsm", 0x0B>;
2065
2066// Branch and save and set mode.
2067let isCall = 1, Defs = [CC] in
2068  def BASSM : CallRR<"bassm", 0x0C>;
2069
2070//===----------------------------------------------------------------------===//
2071// Transactional execution
2072//===----------------------------------------------------------------------===//
2073
2074let hasSideEffects = 1, Predicates = [FeatureTransactionalExecution] in {
2075  // Transaction Begin
2076  let mayStore = 1, usesCustomInserter = 1, Defs = [CC] in {
2077    def TBEGIN : TestBinarySIL<"tbegin", 0xE560, z_tbegin, imm32zx16>;
2078    let hasNoSchedulingInfo = 1 in
2079     def TBEGIN_nofloat : TestBinarySILPseudo<z_tbegin_nofloat, imm32zx16>;
2080    def TBEGINC : SideEffectBinarySIL<"tbeginc", 0xE561,
2081                                      int_s390_tbeginc, imm32zx16>;
2082  }
2083
2084  // Transaction End
2085  let Defs = [CC] in
2086    def TEND : TestInherentS<"tend", 0xB2F8, z_tend>;
2087
2088  // Transaction Abort
2089  let isTerminator = 1, isBarrier = 1, mayStore = 1,
2090      hasSideEffects = 1 in
2091    def TABORT : SideEffectAddressS<"tabort", 0xB2FC, int_s390_tabort>;
2092
2093  // Nontransactional Store
2094  def NTSTG : StoreRXY<"ntstg", 0xE325, int_s390_ntstg, GR64, 8>;
2095
2096  // Extract Transaction Nesting Depth
2097  def ETND : InherentRRE<"etnd", 0xB2EC, GR32, int_s390_etnd>;
2098}
2099
2100//===----------------------------------------------------------------------===//
2101// Processor assist
2102//===----------------------------------------------------------------------===//
2103
2104let Predicates = [FeatureProcessorAssist] in {
2105  let hasSideEffects = 1 in
2106    def PPA : SideEffectTernaryRRFc<"ppa", 0xB2E8, GR64, GR64, imm32zx4>;
2107  def : Pat<(int_s390_ppa_txassist GR32:$src),
2108            (PPA (INSERT_SUBREG (i64 (IMPLICIT_DEF)), GR32:$src, subreg_l32),
2109                 zero_reg, 1)>;
2110}
2111
2112//===----------------------------------------------------------------------===//
2113// Miscellaneous Instructions.
2114//===----------------------------------------------------------------------===//
2115
2116// Find leftmost one, AKA count leading zeros.  The instruction actually
2117// returns a pair of GR64s, the first giving the number of leading zeros
2118// and the second giving a copy of the source with the leftmost one bit
2119// cleared.  We only use the first result here.
2120let Defs = [CC] in
2121  def FLOGR : UnaryRRE<"flogr", 0xB983, null_frag, GR128, GR64>;
2122def : Pat<(i64 (ctlz GR64:$src)),
2123          (EXTRACT_SUBREG (FLOGR GR64:$src), subreg_h64)>;
2124
2125// Population count.  Counts bits set per byte or doubleword.
2126let Predicates = [FeatureMiscellaneousExtensions3] in {
2127  let Defs = [CC] in
2128    def POPCNTOpt : BinaryRRFc<"popcnt", 0xB9E1, GR64, GR64>;
2129  def : Pat<(ctpop GR64:$src), (POPCNTOpt GR64:$src, 8)>;
2130}
2131let Predicates = [FeaturePopulationCount], Defs = [CC] in
2132  def POPCNT : UnaryRRE<"popcnt", 0xB9E1, z_popcnt, GR64, GR64>;
2133
2134// Search a block of memory for a character.
2135let mayLoad = 1, Defs = [CC] in
2136  defm SRST : StringRRE<"srst", 0xB25E, z_search_string>;
2137let mayLoad = 1, Defs = [CC], Uses = [R0L] in
2138  def SRSTU : SideEffectBinaryMemMemRRE<"srstu", 0xB9BE, GR64, GR64>;
2139
2140// Compare until substring equal.
2141let mayLoad = 1, Defs = [CC], Uses = [R0L, R1L] in
2142  def CUSE : SideEffectBinaryMemMemRRE<"cuse", 0xB257, GR128, GR128>;
2143
2144// Compare and form codeword.
2145let mayLoad = 1, Defs = [CC, R1D, R2D, R3D], Uses = [R1D, R2D, R3D] in
2146  def CFC : SideEffectAddressS<"cfc", 0xB21A, null_frag>;
2147
2148// Update tree.
2149let mayLoad = 1, mayStore = 1, Defs = [CC, R0D, R1D, R2D, R3D, R5D],
2150    Uses = [R0D, R1D, R2D, R3D, R4D, R5D] in
2151  def UPT : SideEffectInherentE<"upt", 0x0102>;
2152
2153// Checksum.
2154let mayLoad = 1, Defs = [CC] in
2155  def CKSM : SideEffectBinaryMemMemRRE<"cksm", 0xB241, GR64, GR128>;
2156
2157// Compression call.
2158let mayLoad = 1, mayStore = 1, Defs = [CC, R1D], Uses = [R0L, R1D] in
2159  def CMPSC : SideEffectBinaryMemMemRRE<"cmpsc", 0xB263, GR128, GR128>;
2160
2161// Sort lists.
2162let Predicates = [FeatureEnhancedSort],
2163    mayLoad = 1, mayStore = 1, Defs = [CC], Uses = [R0L, R1D] in
2164  def SORTL : SideEffectBinaryMemMemRRE<"sortl", 0xB938, GR128, GR128>;
2165
2166// Deflate conversion call.
2167let Predicates = [FeatureDeflateConversion],
2168    mayLoad = 1, mayStore = 1, Defs = [CC], Uses = [R0L, R1D] in
2169  def DFLTCC : SideEffectTernaryMemMemRRFa<"dfltcc", 0xB939,
2170                                           GR128, GR128, GR64>;
2171
2172// NNPA.
2173let Predicates = [FeatureNNPAssist],
2174    mayLoad = 1, mayStore = 1, Defs = [R0D, CC], Uses = [R0D, R1D] in
2175  def NNPA : SideEffectInherentRRE<"nnpa", 0xB93B>;
2176
2177// Execute.
2178let hasSideEffects = 1 in {
2179  def EX   : SideEffectBinaryRX<"ex", 0x44, ADDR64>;
2180  def EXRL : SideEffectBinaryRILPC<"exrl", 0xC60, ADDR64>;
2181  let hasNoSchedulingInfo = 1 in
2182    def EXRL_Pseudo : Alias<6, (outs), (ins i64imm:$TargetOpc, ADDR64:$lenMinus1,
2183                                          bdaddr12only:$bdl1, bdaddr12only:$bd2),
2184                                          []>;
2185}
2186
2187//===----------------------------------------------------------------------===//
2188// .insn directive instructions
2189//===----------------------------------------------------------------------===//
2190
2191let isCodeGenOnly = 1, hasSideEffects = 1 in {
2192  def InsnE   : DirectiveInsnE<(outs), (ins imm64zx16:$enc), ".insn e,$enc", []>;
2193  def InsnRI  : DirectiveInsnRI<(outs), (ins imm64zx32:$enc, AnyReg:$R1,
2194                                             imm32sx16:$I2),
2195                                ".insn ri,$enc,$R1,$I2", []>;
2196  def InsnRIE : DirectiveInsnRIE<(outs), (ins imm64zx48:$enc, AnyReg:$R1,
2197                                              AnyReg:$R3, brtarget16:$I2),
2198                                 ".insn rie,$enc,$R1,$R3,$I2", []>;
2199  def InsnRIL : DirectiveInsnRIL<(outs), (ins imm64zx48:$enc, AnyReg:$R1,
2200                                              brtarget32:$I2),
2201                                 ".insn ril,$enc,$R1,$I2", []>;
2202  def InsnRILU : DirectiveInsnRIL<(outs), (ins imm64zx48:$enc, AnyReg:$R1,
2203                                               uimm32:$I2),
2204                                  ".insn rilu,$enc,$R1,$I2", []>;
2205  def InsnRIS : DirectiveInsnRIS<(outs),
2206                                 (ins imm64zx48:$enc, AnyReg:$R1,
2207                                      imm32sx8:$I2, imm32zx4:$M3,
2208                                      bdaddr12only:$BD4),
2209                                 ".insn ris,$enc,$R1,$I2,$M3,$BD4", []>;
2210  def InsnRR : DirectiveInsnRR<(outs),
2211                               (ins imm64zx16:$enc, AnyReg:$R1, AnyReg:$R2),
2212                               ".insn rr,$enc,$R1,$R2", []>;
2213  def InsnRRE : DirectiveInsnRRE<(outs), (ins imm64zx32:$enc,
2214                                              AnyReg:$R1, AnyReg:$R2),
2215                                 ".insn rre,$enc,$R1,$R2", []>;
2216  def InsnRRF : DirectiveInsnRRF<(outs),
2217                                 (ins imm64zx32:$enc, AnyReg:$R1, AnyReg:$R2,
2218                                      AnyReg:$R3, imm32zx4:$M4),
2219                                 ".insn rrf,$enc,$R1,$R2,$R3,$M4", []>;
2220  def InsnRRS : DirectiveInsnRRS<(outs),
2221                                 (ins imm64zx48:$enc, AnyReg:$R1,
2222                                      AnyReg:$R2, imm32zx4:$M3,
2223                                      bdaddr12only:$BD4),
2224                                 ".insn rrs,$enc,$R1,$R2,$M3,$BD4", []>;
2225  def InsnRS  : DirectiveInsnRS<(outs),
2226                                (ins imm64zx32:$enc, AnyReg:$R1,
2227                                     AnyReg:$R3, bdaddr12only:$BD2),
2228                                ".insn rs,$enc,$R1,$R3,$BD2", []>;
2229  def InsnRSE : DirectiveInsnRSE<(outs),
2230                                 (ins imm64zx48:$enc, AnyReg:$R1,
2231                                      AnyReg:$R3, bdaddr12only:$BD2),
2232                                 ".insn rse,$enc,$R1,$R3,$BD2", []>;
2233  def InsnRSI : DirectiveInsnRSI<(outs),
2234                                 (ins imm64zx48:$enc, AnyReg:$R1,
2235                                      AnyReg:$R3, brtarget16:$RI2),
2236                                 ".insn rsi,$enc,$R1,$R3,$RI2", []>;
2237  def InsnRSY : DirectiveInsnRSY<(outs),
2238                                 (ins imm64zx48:$enc, AnyReg:$R1,
2239                                      AnyReg:$R3, bdaddr20only:$BD2),
2240                                 ".insn rsy,$enc,$R1,$R3,$BD2", []>;
2241  def InsnRX  : DirectiveInsnRX<(outs), (ins imm64zx32:$enc, AnyReg:$R1,
2242                                             bdxaddr12only:$XBD2),
2243                                ".insn rx,$enc,$R1,$XBD2", []>;
2244  def InsnRXE : DirectiveInsnRXE<(outs), (ins imm64zx48:$enc, AnyReg:$R1,
2245                                              bdxaddr12only:$XBD2),
2246                                 ".insn rxe,$enc,$R1,$XBD2", []>;
2247  def InsnRXF : DirectiveInsnRXF<(outs),
2248                                 (ins imm64zx48:$enc, AnyReg:$R1,
2249                                      AnyReg:$R3, bdxaddr12only:$XBD2),
2250                                 ".insn rxf,$enc,$R1,$R3,$XBD2", []>;
2251  def InsnRXY : DirectiveInsnRXY<(outs), (ins imm64zx48:$enc, AnyReg:$R1,
2252                                              bdxaddr20only:$XBD2),
2253                                 ".insn rxy,$enc,$R1,$XBD2", []>;
2254  def InsnS : DirectiveInsnS<(outs),
2255                             (ins imm64zx32:$enc, bdaddr12only:$BD2),
2256                             ".insn s,$enc,$BD2", []>;
2257  def InsnSI : DirectiveInsnSI<(outs),
2258                               (ins imm64zx32:$enc, bdaddr12only:$BD1,
2259                                    imm32sx8:$I2),
2260                               ".insn si,$enc,$BD1,$I2", []>;
2261  def InsnSIY : DirectiveInsnSIY<(outs),
2262                                 (ins imm64zx48:$enc,
2263                                      bdaddr20only:$BD1, imm32zx8:$I2),
2264                                 ".insn siy,$enc,$BD1,$I2", []>;
2265  def InsnSIL : DirectiveInsnSIL<(outs),
2266                                 (ins imm64zx48:$enc, bdaddr12only:$BD1,
2267                                      imm32zx16:$I2),
2268                                 ".insn sil,$enc,$BD1,$I2", []>;
2269  def InsnSS : DirectiveInsnSS<(outs),
2270                               (ins imm64zx48:$enc, bdraddr12only:$RBD1,
2271                                    bdaddr12only:$BD2, AnyReg:$R3),
2272                               ".insn ss,$enc,$RBD1,$BD2,$R3", []>;
2273  def InsnSSE : DirectiveInsnSSE<(outs),
2274                                 (ins imm64zx48:$enc,
2275                                      bdaddr12only:$BD1,bdaddr12only:$BD2),
2276                                 ".insn sse,$enc,$BD1,$BD2", []>;
2277  def InsnSSF : DirectiveInsnSSF<(outs),
2278                                 (ins imm64zx48:$enc, bdaddr12only:$BD1,
2279                                      bdaddr12only:$BD2, AnyReg:$R3),
2280                                 ".insn ssf,$enc,$BD1,$BD2,$R3", []>;
2281  def InsnVRI : DirectiveInsnVRI<(outs),
2282                                 (ins imm64zx48:$enc, VR128:$V1, VR128:$V2,
2283                                  imm32zx12:$I3, imm32zx4:$M4, imm32zx4:$M5),
2284                                 ".insn vri,$enc,$V1,$V2,$I3,$M4,$M5", []>;
2285  def InsnVRR : DirectiveInsnVRR<(outs),
2286                                 (ins imm64zx48:$enc, VR128:$V1, VR128:$V2,
2287                                  VR128:$V3, imm32zx4:$M4, imm32zx4:$M5,
2288                                  imm32zx4:$M6),
2289                                  ".insn vrr,$enc,$V1,$V2,$V3,$M4,$M5,$M6", []>;
2290  def InsnVRS : DirectiveInsnVRS<(outs),
2291                                 (ins imm64zx48:$enc, AnyReg:$R1, VR128:$V3,
2292                                  bdaddr12only:$BD2, imm32zx4:$M4),
2293                                 ".insn vrs,$enc,$BD2,$M4", []>;
2294  def InsnVRV : DirectiveInsnVRV<(outs),
2295                                 (ins imm64zx48:$enc, VR128:$V1,
2296                                      bdvaddr12only:$VBD2, imm32zx4:$M3),
2297                                 ".insn vrv,$enc,$V1,$VBD2,$M3", []>;
2298  def InsnVRX : DirectiveInsnVRX<(outs),
2299                                 (ins imm64zx48:$enc, VR128:$V1,
2300                                  bdxaddr12only:$XBD2, imm32zx4:$M3),
2301                                 ".insn vrx,$enc,$V1,$XBD2,$M3", []>;
2302  def InsnVSI : DirectiveInsnVSI<(outs),
2303                                 (ins imm64zx48:$enc, VR128:$V1,
2304                                  bdaddr12only:$BD2, imm32zx8:$I3),
2305                                  ".insn vsi,$enc,$V1,$BD2,$I3", []>;
2306}
2307
2308//===----------------------------------------------------------------------===//
2309// Peepholes.
2310//===----------------------------------------------------------------------===//
2311
2312// Avoid generating 2 XOR instructions. (xor (and x, y), y) is
2313// equivalent to (and (xor x, -1), y)
2314def : Pat<(and (xor GR64:$x, (i64 -1)), GR64:$y),
2315                          (XGR GR64:$y, (NGR GR64:$y, GR64:$x))>;
2316
2317// Shift/rotate instructions only use the last 6 bits of the second operand
2318// register, so we can safely use NILL (16 fewer bits than NILF) to only AND the
2319// last 16 bits.
2320// Complexity is added so that we match this before we match NILF on the AND
2321// operation alone.
2322let AddedComplexity = 4 in {
2323  def : Pat<(shl GR32:$val, (and GR32:$shift, imm32zx16trunc:$imm)),
2324            (SLL GR32:$val, (NILL GR32:$shift, imm32zx16trunc:$imm), 0)>;
2325
2326  def : Pat<(sra GR32:$val, (and GR32:$shift, imm32zx16trunc:$imm)),
2327            (SRA GR32:$val, (NILL GR32:$shift, imm32zx16trunc:$imm), 0)>;
2328
2329  def : Pat<(srl GR32:$val, (and GR32:$shift, imm32zx16trunc:$imm)),
2330            (SRL GR32:$val, (NILL GR32:$shift, imm32zx16trunc:$imm), 0)>;
2331
2332  def : Pat<(shl GR64:$val, (and GR32:$shift, imm32zx16trunc:$imm)),
2333            (SLLG GR64:$val, (NILL GR32:$shift, imm32zx16trunc:$imm), 0)>;
2334
2335  def : Pat<(sra GR64:$val, (and GR32:$shift, imm32zx16trunc:$imm)),
2336            (SRAG GR64:$val, (NILL GR32:$shift, imm32zx16trunc:$imm), 0)>;
2337
2338  def : Pat<(srl GR64:$val, (and GR32:$shift, imm32zx16trunc:$imm)),
2339            (SRLG GR64:$val, (NILL GR32:$shift, imm32zx16trunc:$imm), 0)>;
2340
2341  def : Pat<(rotl GR32:$val, (and GR32:$shift, imm32zx16trunc:$imm)),
2342            (RLL GR32:$val, (NILL GR32:$shift, imm32zx16trunc:$imm), 0)>;
2343
2344  def : Pat<(rotl GR64:$val, (and GR32:$shift, imm32zx16trunc:$imm)),
2345            (RLLG GR64:$val, (NILL GR32:$shift, imm32zx16trunc:$imm), 0)>;
2346}
2347
2348// Substitute (x*64-s) with (-s), since shift/rotate instructions only
2349// use the last 6 bits of the second operand register (making it modulo 64).
2350let AddedComplexity = 4 in {
2351  def : Pat<(shl GR64:$val, (sub imm32mod64,  GR32:$shift)),
2352            (SLLG GR64:$val, (LCR GR32:$shift), 0)>;
2353
2354  def : Pat<(sra GR64:$val, (sub imm32mod64,  GR32:$shift)),
2355            (SRAG GR64:$val, (LCR GR32:$shift), 0)>;
2356
2357  def : Pat<(srl GR64:$val, (sub imm32mod64,  GR32:$shift)),
2358            (SRLG GR64:$val, (LCR GR32:$shift), 0)>;
2359
2360  def : Pat<(rotl GR64:$val, (sub imm32mod64,  GR32:$shift)),
2361            (RLLG GR64:$val, (LCR GR32:$shift), 0)>;
2362}
2363
2364// Peepholes for turning scalar operations into block operations.  The length
2365// is given as one less for these pseudos.
2366defm : BlockLoadStore<anyextloadi8, i32, MVCImm, NCImm, OCImm, XCImm, 0>;
2367defm : BlockLoadStore<anyextloadi16, i32, MVCImm, NCImm, OCImm, XCImm, 1>;
2368defm : BlockLoadStore<load, i32, MVCImm, NCImm, OCImm, XCImm, 3>;
2369defm : BlockLoadStore<anyextloadi8, i64, MVCImm, NCImm, OCImm, XCImm, 0>;
2370defm : BlockLoadStore<anyextloadi16, i64, MVCImm, NCImm, OCImm, XCImm, 1>;
2371defm : BlockLoadStore<anyextloadi32, i64, MVCImm, NCImm, OCImm, XCImm, 3>;
2372defm : BlockLoadStore<load, i64, MVCImm, NCImm, OCImm, XCImm, 7>;
2373
2374//===----------------------------------------------------------------------===//
2375// Mnemonic Aliases
2376//===----------------------------------------------------------------------===//
2377
2378def JCT   : MnemonicAlias<"jct", "brct">;
2379def JCTG  : MnemonicAlias<"jctg", "brctg">;
2380def JAS   : MnemonicAlias<"jas", "bras">;
2381def JASL  : MnemonicAlias<"jasl", "brasl">;
2382def JXH   : MnemonicAlias<"jxh", "brxh">;
2383def JXLE  : MnemonicAlias<"jxle", "brxle">;
2384def JXHG  : MnemonicAlias<"jxhg", "brxhg">;
2385def JXLEG : MnemonicAlias<"jxleg", "brxlg">;
2386
2387def BRU   : MnemonicAlias<"bru", "j">;
2388def BRUL  : MnemonicAlias<"brul", "jg", "att">;
2389def BRUL_HLASM  : MnemonicAlias<"brul", "jlu", "hlasm">;
2390
2391foreach 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">;
2396}
2397