X-Git-Url: http://plrg.eecs.uci.edu/git/?p=oota-llvm.git;a=blobdiff_plain;f=lib%2FTarget%2FSystemZ%2FSystemZInstrInfo.td;h=d5dabc2cd6ab03728ad071ec6205282ecf1c6ab4;hp=31832f70d298d0405b4dcb1027b9f84070047671;hb=8c8ee7cee1b6ed6253ca27666a95a666cd4eea53;hpb=e1b2af731e2a45344a7c502232f66c55cd746da0 diff --git a/lib/Target/SystemZ/SystemZInstrInfo.td b/lib/Target/SystemZ/SystemZInstrInfo.td index 31832f70d29..d5dabc2cd6a 100644 --- a/lib/Target/SystemZ/SystemZInstrInfo.td +++ b/lib/Target/SystemZ/SystemZInstrInfo.td @@ -16,7 +16,7 @@ def ADJCALLSTACKDOWN : Pseudo<(outs), (ins i64imm:$amt), def ADJCALLSTACKUP : Pseudo<(outs), (ins i64imm:$amt1, i64imm:$amt2), [(callseq_end timm:$amt1, timm:$amt2)]>; -let neverHasSideEffects = 1 in { +let hasSideEffects = 0 in { // Takes as input the value of the stack pointer after a dynamic allocation // has been made. Sets the output to the address of the dynamically- // allocated area itself, skipping the outgoing arguments. @@ -32,12 +32,9 @@ let neverHasSideEffects = 1 in { // Control flow instructions //===----------------------------------------------------------------------===// -// A return instruction. R1 is the condition-code mask (all 1s) -// and R2 is the target address, which is always stored in %r14. -let isReturn = 1, isTerminator = 1, isBarrier = 1, hasCtrlDep = 1, - R1 = 15, R2 = 14, isCodeGenOnly = 1 in { - def RET : InstRR<0x07, (outs), (ins), "br\t%r14", [(z_retflag)]>; -} +// A return instruction (br %r14). +let isReturn = 1, isTerminator = 1, isBarrier = 1, hasCtrlDep = 1 in + def Return : Alias<2, (outs), (ins), [(z_retflag)]>; // Unconditional branches. R1 is the condition-code mask (all 1s). let isBranch = 1, isTerminator = 1, isBarrier = 1, R1 = 15 in { @@ -66,10 +63,12 @@ let isBranch = 1, isTerminator = 1, Uses = [CC] in { def BRCL : InstRIL<0xC04, (outs), (ins cond4:$valid, cond4:$R1, brtarget32:$I2), "jg$R1\t$I2", []>; } - def AsmBRC : InstRI<0xA74, (outs), (ins uimm8zx4:$R1, brtarget16:$I2), + def AsmBRC : InstRI<0xA74, (outs), (ins imm32zx4:$R1, brtarget16:$I2), "brc\t$R1, $I2", []>; - def AsmBRCL : InstRIL<0xC04, (outs), (ins uimm8zx4:$R1, brtarget32:$I2), + def AsmBRCL : InstRIL<0xC04, (outs), (ins imm32zx4:$R1, brtarget32:$I2), "brcl\t$R1, $I2", []>; + def AsmBCR : InstRR<0x07, (outs), (ins imm32zx4:$R1, GR64:$R2), + "bcr\t$R1, $R2", []>; } // Fused compare-and-branch instructions. As for normal branches, @@ -94,11 +93,23 @@ multiclass CompareBranches { def GIJ : InstRIEc<0xEC7C, (outs), (ins GR64:$R1, imm64sx8:$I2, ccmask:$M3, brtarget16:$RI4), "cgij"##pos1##"\t$R1, $I2, "##pos2##"$RI4", []>; + def LRJ : InstRIEb<0xEC77, (outs), (ins GR32:$R1, GR32:$R2, ccmask:$M3, + brtarget16:$RI4), + "clrj"##pos1##"\t$R1, $R2, "##pos2##"$RI4", []>; + def LGRJ : InstRIEb<0xEC65, (outs), (ins GR64:$R1, GR64:$R2, ccmask:$M3, + brtarget16:$RI4), + "clgrj"##pos1##"\t$R1, $R2, "##pos2##"$RI4", []>; + def LIJ : InstRIEc<0xEC7F, (outs), (ins GR32:$R1, imm32zx8:$I2, ccmask:$M3, + brtarget16:$RI4), + "clij"##pos1##"\t$R1, $I2, "##pos2##"$RI4", []>; + def LGIJ : InstRIEc<0xEC7D, (outs), (ins GR64:$R1, imm64zx8:$I2, ccmask:$M3, + brtarget16:$RI4), + "clgij"##pos1##"\t$R1, $I2, "##pos2##"$RI4", []>; } } let isCodeGenOnly = 1 in defm C : CompareBranches; -defm AsmC : CompareBranches; +defm AsmC : CompareBranches; // Define AsmParser mnemonics for each general condition-code mask // (integer or floating-point) @@ -108,6 +119,7 @@ multiclass CondExtendedMnemonic ccmask, string name> { "j"##name##"\t$I2", []>; def JG : InstRIL<0xC04, (outs), (ins brtarget32:$I2), "jg"##name##"\t$I2", []>; + def BR : InstRR<0x07, (outs), (ins ADDR64:$R2), "b"##name##"r\t$R2", []>; } def LOCR : FixedCondUnaryRRF<"locr"##name, 0xB9F2, GR32, GR32, ccmask>; def LOCGR : FixedCondUnaryRRF<"locgr"##name, 0xB9E2, GR64, GR64, ccmask>; @@ -152,6 +164,18 @@ multiclass IntCondExtendedMnemonicA ccmask, string name> { def CGI : InstRIEc<0xEC7C, (outs), (ins GR64:$R1, imm64sx8:$I2, brtarget16:$RI4), "cgij"##name##"\t$R1, $I2, $RI4", []>; + def CLR : InstRIEb<0xEC77, (outs), (ins GR32:$R1, GR32:$R2, + brtarget16:$RI4), + "clrj"##name##"\t$R1, $R2, $RI4", []>; + def CLGR : InstRIEb<0xEC65, (outs), (ins GR64:$R1, GR64:$R2, + brtarget16:$RI4), + "clgrj"##name##"\t$R1, $R2, $RI4", []>; + def CLI : InstRIEc<0xEC7F, (outs), (ins GR32:$R1, imm32zx8:$I2, + brtarget16:$RI4), + "clij"##name##"\t$R1, $I2, $RI4", []>; + def CLGI : InstRIEc<0xEC7D, (outs), (ins GR64:$R1, imm64zx8:$I2, + brtarget16:$RI4), + "clgij"##name##"\t$R1, $I2, $RI4", []>; } } multiclass IntCondExtendedMnemonic ccmask, string name1, string name2> @@ -177,22 +201,31 @@ let Defs = [CC] in { // Select instructions //===----------------------------------------------------------------------===// -def Select32 : SelectWrapper; -def Select64 : SelectWrapper; - -defm CondStore8_32 : CondStores, Requires<[FeatureHighWord]>; +def Select32 : SelectWrapper; +def Select64 : SelectWrapper; + +// We don't define 32-bit Mux stores because the low-only STOC should +// always be used if possible. +defm CondStore8Mux : CondStores, + Requires<[FeatureHighWord]>; +defm CondStore16Mux : CondStores, + Requires<[FeatureHighWord]>; +defm CondStore8 : CondStores; -defm CondStore16_32 : CondStores; -defm CondStore32_32 : CondStores; -defm CondStore8 : CondStores; -defm CondStore16 : CondStores; -defm CondStore32 : CondStores; +defm : CondStores64; +defm : CondStores64; +defm : CondStores64; defm CondStore64 : CondStores; @@ -200,33 +233,50 @@ defm CondStore64 : CondStores; - def BRASL : InstRIL<0xC05, (outs), (ins pcrel32call:$I2, variable_ops), - "brasl\t%r14, $I2", [(z_call pcrel32call:$I2)]>; - def BASR : InstRR<0x0D, (outs), (ins ADDR64:$R2, variable_ops), - "basr\t%r14, $R2", [(z_call ADDR64:$R2)]>; +let isCall = 1, Defs = [R14D, CC] in { + def CallBRASL : Alias<6, (outs), (ins pcrel32:$I2, variable_ops), + [(z_call pcrel32:$I2)]>; + def CallBASR : Alias<2, (outs), (ins ADDR64:$R2, variable_ops), + [(z_call ADDR64:$R2)]>; +} + +// Sibling calls. Indirect sibling calls must be via R1, since R2 upwards +// are argument registers and since branching to R0 is a no-op. +let isCall = 1, isTerminator = 1, isReturn = 1, isBarrier = 1 in { + def CallJG : Alias<6, (outs), (ins pcrel32:$I2), + [(z_sibcall pcrel32:$I2)]>; + let Uses = [R1D] in + def CallBR : Alias<2, (outs), (ins), [(z_sibcall R1D)]>; +} + +// TLS calls. These will be lowered into a call to __tls_get_offset, +// with an extra relocation specifying the TLS symbol. +let isCall = 1, Defs = [R14D, CC] in { + def TLS_GDCALL : Alias<6, (outs), (ins tlssym:$I2, variable_ops), + [(z_tls_gdcall tglobaltlsaddr:$I2)]>; + def TLS_LDCALL : Alias<6, (outs), (ins tlssym:$I2, variable_ops), + [(z_tls_ldcall tglobaltlsaddr:$I2)]>; } // Define the general form of the call instructions for the asm parser. // These instructions don't hard-code %r14 as the return address register. -def AsmBRAS : InstRI<0xA75, (outs), (ins GR64:$R1, brtarget16:$I2), - "bras\t$R1, $I2", []>; -def AsmBRASL : InstRIL<0xC05, (outs), (ins GR64:$R1, brtarget32:$I2), - "brasl\t$R1, $I2", []>; -def AsmBASR : InstRR<0x0D, (outs), (ins GR64:$R1, ADDR64:$R2), - "basr\t$R1, $R2", []>; +// Allow an optional TLS marker symbol to generate TLS call relocations. +def BRAS : InstRI<0xA75, (outs), (ins GR64:$R1, brtarget16tls:$I2), + "bras\t$R1, $I2", []>; +def BRASL : InstRIL<0xC05, (outs), (ins GR64:$R1, brtarget32tls:$I2), + "brasl\t$R1, $I2", []>; +def BASR : InstRR<0x0D, (outs), (ins GR64:$R1, ADDR64:$R2), + "basr\t$R1, $R2", []>; //===----------------------------------------------------------------------===// // Move instructions //===----------------------------------------------------------------------===// // Register moves. -let neverHasSideEffects = 1 in { +let hasSideEffects = 0 in { + // Expands to LR, RISBHG or RISBLG, depending on the choice of registers. + def LRMux : UnaryRRPseudo<"l", null_frag, GRX32, GRX32>, + Requires<[FeatureHighWord]>; def LR : UnaryRR <"l", 0x18, null_frag, GR32, GR32>; def LGR : UnaryRRE<"lg", 0xB904, null_frag, GR64, GR64>; } @@ -246,9 +296,12 @@ let Uses = [CC] in { } // Immediate moves. -let neverHasSideEffects = 1, isAsCheapAsAMove = 1, isMoveImm = 1, +let hasSideEffects = 0, isAsCheapAsAMove = 1, isMoveImm = 1, isReMaterializable = 1 in { - // 16-bit sign-extended immediates. + // 16-bit sign-extended immediates. LHIMux expands to LHI or IIHF, + // deopending on the choice of register. + def LHIMux : UnaryRIPseudo, + Requires<[FeatureHighWord]>; def LHI : UnaryRI<"lhi", 0xA78, bitconvert, GR32, imm32sx16>; def LGHI : UnaryRI<"lghi", 0xA79, bitconvert, GR64, imm64sx16>; @@ -266,7 +319,12 @@ let neverHasSideEffects = 1, isAsCheapAsAMove = 1, isMoveImm = 1, // Register loads. let canFoldAsLoad = 1, SimpleBDXLoad = 1 in { + // Expands to L, LY or LFH, depending on the choice of register. + def LMux : UnaryRXYPseudo<"l", load, GRX32, 4>, + Requires<[FeatureHighWord]>; defm L : UnaryRXPair<"l", 0x58, 0xE358, load, GR32, 4>; + def LFH : UnaryRXY<"lfh", 0xE3CA, load, GRH32, 4>, + Requires<[FeatureHighWord]>; def LG : UnaryRXY<"lg", 0xE304, load, GR64, 8>; // These instructions are split after register allocation, so we don't @@ -298,8 +356,12 @@ let Uses = [CC] in { // Register stores. let SimpleBDXStore = 1 in { - let isCodeGenOnly = 1 in - defm ST32 : StoreRXPair<"st", 0x50, 0xE350, store, GR32, 4>; + // Expands to ST, STY or STFH, depending on the choice of register. + def STMux : StoreRXYPseudo, + Requires<[FeatureHighWord]>; + defm ST : StoreRXPair<"st", 0x50, 0xE350, store, GR32, 4>; + def STFH : StoreRXY<"stfh", 0xE3CB, store, GRH32, 4>, + Requires<[FeatureHighWord]>; def STG : StoreRXY<"stg", 0xE324, store, GR64, 8>; // These instructions are split after register allocation, so we don't @@ -309,15 +371,13 @@ let SimpleBDXStore = 1 in { [(store GR128:$src, bdxaddr20only128:$dst)]>; } } -let isCodeGenOnly = 1 in - def STRL32 : StoreRILPC<"strl", 0xC4F, aligned_store, GR32>; +def STRL : StoreRILPC<"strl", 0xC4F, aligned_store, GR32>; def STGRL : StoreRILPC<"stgrl", 0xC4B, aligned_store, GR64>; // Store on condition. let isCodeGenOnly = 1, Uses = [CC] in { - def STOC32 : CondStoreRSY<"stoc", 0xEBF3, GR32, 4>; - def STOC : CondStoreRSY<"stoc", 0xEBF3, GR64, 4>; - def STOCG : CondStoreRSY<"stocg", 0xEBE3, GR64, 8>; + def STOC : CondStoreRSY<"stoc", 0xEBF3, GR32, 4>; + def STOCG : CondStoreRSY<"stocg", 0xEBE3, GR64, 8>; } let Uses = [CC] in { def AsmSTOC : AsmCondStoreRSY<"stoc", 0xEBF3, GR32, 4>; @@ -334,91 +394,87 @@ def MVGHI : StoreSIL<"mvghi", 0xE548, store, imm64sx16>; // Memory-to-memory moves. let mayLoad = 1, mayStore = 1 in - defm MVC : MemorySS<"mvc", 0xD2, z_mvc>; - -defm LoadStore8_32 : MVCLoadStore; -defm LoadStore16_32 : MVCLoadStore; -defm LoadStore32_32 : MVCLoadStore; - -defm LoadStore8 : MVCLoadStore; -defm LoadStore16 : MVCLoadStore; -defm LoadStore32 : MVCLoadStore; -defm LoadStore64 : MVCLoadStore; + defm MVC : MemorySS<"mvc", 0xD2, z_mvc, z_mvc_loop>; + +// String moves. +let mayLoad = 1, mayStore = 1, Defs = [CC] in + defm MVST : StringRRE<"mvst", 0xB255, z_stpcpy>; //===----------------------------------------------------------------------===// // Sign extensions //===----------------------------------------------------------------------===// +// +// Note that putting these before zero extensions mean that we will prefer +// them for anyextload*. There's not really much to choose between the two +// either way, but signed-extending loads have a short LH and a long LHY, +// while zero-extending loads have only the long LLH. +// +//===----------------------------------------------------------------------===// // 32-bit extensions from registers. -let neverHasSideEffects = 1 in { +let hasSideEffects = 0 in { def LBR : UnaryRRE<"lb", 0xB926, sext8, GR32, GR32>; def LHR : UnaryRRE<"lh", 0xB927, sext16, GR32, GR32>; } // 64-bit extensions from registers. -let neverHasSideEffects = 1 in { +let hasSideEffects = 0 in { def LGBR : UnaryRRE<"lgb", 0xB906, sext8, GR64, GR64>; def LGHR : UnaryRRE<"lgh", 0xB907, sext16, GR64, GR64>; def LGFR : UnaryRRE<"lgf", 0xB914, sext32, GR64, GR32>; } let Defs = [CC], CCValues = 0xE, CompareZeroCCMask = 0xE in - def LTGFR : UnaryRRE<"ltgf", 0xB912, null_frag, GR64, GR64>; + def LTGFR : UnaryRRE<"ltgf", 0xB912, null_frag, GR64, GR32>; // Match 32-to-64-bit sign extensions in which the source is already // in a 64-bit register. def : Pat<(sext_inreg GR64:$src, i32), - (LGFR (EXTRACT_SUBREG GR64:$src, subreg_32bit))>; - -// 32-bit extensions from memory. -def LB : UnaryRXY<"lb", 0xE376, sextloadi8, GR32, 1>; -defm LH : UnaryRXPair<"lh", 0x48, 0xE378, sextloadi16, GR32, 2>; -def LHRL : UnaryRILPC<"lhrl", 0xC45, aligned_sextloadi16, GR32>; + (LGFR (EXTRACT_SUBREG GR64:$src, subreg_l32))>; + +// 32-bit extensions from 8-bit memory. LBMux expands to LB or LBH, +// depending on the choice of register. +def LBMux : UnaryRXYPseudo<"lb", asextloadi8, GRX32, 1>, + Requires<[FeatureHighWord]>; +def LB : UnaryRXY<"lb", 0xE376, asextloadi8, GR32, 1>; +def LBH : UnaryRXY<"lbh", 0xE3C0, asextloadi8, GRH32, 1>, + Requires<[FeatureHighWord]>; + +// 32-bit extensions from 16-bit memory. LHMux expands to LH or LHH, +// depending on the choice of register. +def LHMux : UnaryRXYPseudo<"lh", asextloadi16, GRX32, 2>, + Requires<[FeatureHighWord]>; +defm LH : UnaryRXPair<"lh", 0x48, 0xE378, asextloadi16, GR32, 2>; +def LHH : UnaryRXY<"lhh", 0xE3C4, asextloadi16, GRH32, 2>, + Requires<[FeatureHighWord]>; +def LHRL : UnaryRILPC<"lhrl", 0xC45, aligned_asextloadi16, GR32>; // 64-bit extensions from memory. -def LGB : UnaryRXY<"lgb", 0xE377, sextloadi8, GR64, 1>; -def LGH : UnaryRXY<"lgh", 0xE315, sextloadi16, GR64, 2>; -def LGF : UnaryRXY<"lgf", 0xE314, sextloadi32, GR64, 4>; -def LGHRL : UnaryRILPC<"lghrl", 0xC44, aligned_sextloadi16, GR64>; -def LGFRL : UnaryRILPC<"lgfrl", 0xC4C, aligned_sextloadi32, GR64>; +def LGB : UnaryRXY<"lgb", 0xE377, asextloadi8, GR64, 1>; +def LGH : UnaryRXY<"lgh", 0xE315, asextloadi16, GR64, 2>; +def LGF : UnaryRXY<"lgf", 0xE314, asextloadi32, GR64, 4>; +def LGHRL : UnaryRILPC<"lghrl", 0xC44, aligned_asextloadi16, GR64>; +def LGFRL : UnaryRILPC<"lgfrl", 0xC4C, aligned_asextloadi32, GR64>; let Defs = [CC], CCValues = 0xE, CompareZeroCCMask = 0xE in - def LTGF : UnaryRXY<"ltgf", 0xE332, sextloadi32, GR64, 4>; - -// If the sign of a load-extend operation doesn't matter, use the signed ones. -// There's not really much to choose between the sign and zero extensions, -// but LH is more compact than LLH for small offsets. -def : Pat<(i32 (extloadi8 bdxaddr20only:$src)), (LB bdxaddr20only:$src)>; -def : Pat<(i32 (extloadi16 bdxaddr12pair:$src)), (LH bdxaddr12pair:$src)>; -def : Pat<(i32 (extloadi16 bdxaddr20pair:$src)), (LHY bdxaddr20pair:$src)>; - -def : Pat<(i64 (extloadi8 bdxaddr20only:$src)), (LGB bdxaddr20only:$src)>; -def : Pat<(i64 (extloadi16 bdxaddr20only:$src)), (LGH bdxaddr20only:$src)>; -def : Pat<(i64 (extloadi32 bdxaddr20only:$src)), (LGF bdxaddr20only:$src)>; - -// We want PC-relative addresses to be tried ahead of BD and BDX addresses. -// However, BDXs have two extra operands and are therefore 6 units more -// complex. -let AddedComplexity = 7 in { - def : Pat<(i32 (extloadi16 pcrel32:$src)), (LHRL pcrel32:$src)>; - def : Pat<(i64 (extloadi16 pcrel32:$src)), (LGHRL pcrel32:$src)>; -} + def LTGF : UnaryRXY<"ltgf", 0xE332, asextloadi32, GR64, 4>; //===----------------------------------------------------------------------===// // Zero extensions //===----------------------------------------------------------------------===// // 32-bit extensions from registers. -let neverHasSideEffects = 1 in { - def LLCR : UnaryRRE<"llc", 0xB994, zext8, GR32, GR32>; - def LLHR : UnaryRRE<"llh", 0xB995, zext16, GR32, GR32>; +let hasSideEffects = 0 in { + // Expands to LLCR or RISB[LH]G, depending on the choice of registers. + def LLCRMux : UnaryRRPseudo<"llc", zext8, GRX32, GRX32>, + Requires<[FeatureHighWord]>; + def LLCR : UnaryRRE<"llc", 0xB994, zext8, GR32, GR32>; + // Expands to LLHR or RISB[LH]G, depending on the choice of registers. + def LLHRMux : UnaryRRPseudo<"llh", zext16, GRX32, GRX32>, + Requires<[FeatureHighWord]>; + def LLHR : UnaryRRE<"llh", 0xB995, zext16, GR32, GR32>; } // 64-bit extensions from registers. -let neverHasSideEffects = 1 in { +let hasSideEffects = 0 in { def LLGCR : UnaryRRE<"llgc", 0xB984, zext8, GR64, GR64>; def LLGHR : UnaryRRE<"llgh", 0xB985, zext16, GR64, GR64>; def LLGFR : UnaryRRE<"llgf", 0xB916, zext32, GR64, GR32>; @@ -427,19 +483,31 @@ let neverHasSideEffects = 1 in { // Match 32-to-64-bit zero extensions in which the source is already // in a 64-bit register. def : Pat<(and GR64:$src, 0xffffffff), - (LLGFR (EXTRACT_SUBREG GR64:$src, subreg_32bit))>; + (LLGFR (EXTRACT_SUBREG GR64:$src, subreg_l32))>; -// 32-bit extensions from memory. -def LLC : UnaryRXY<"llc", 0xE394, zextloadi8, GR32, 1>; -def LLH : UnaryRXY<"llh", 0xE395, zextloadi16, GR32, 2>; -def LLHRL : UnaryRILPC<"llhrl", 0xC42, aligned_zextloadi16, GR32>; +// 32-bit extensions from 8-bit memory. LLCMux expands to LLC or LLCH, +// depending on the choice of register. +def LLCMux : UnaryRXYPseudo<"llc", azextloadi8, GRX32, 1>, + Requires<[FeatureHighWord]>; +def LLC : UnaryRXY<"llc", 0xE394, azextloadi8, GR32, 1>; +def LLCH : UnaryRXY<"llch", 0xE3C2, azextloadi8, GRH32, 1>, + Requires<[FeatureHighWord]>; + +// 32-bit extensions from 16-bit memory. LLHMux expands to LLH or LLHH, +// depending on the choice of register. +def LLHMux : UnaryRXYPseudo<"llh", azextloadi16, GRX32, 2>, + Requires<[FeatureHighWord]>; +def LLH : UnaryRXY<"llh", 0xE395, azextloadi16, GR32, 2>; +def LLHH : UnaryRXY<"llhh", 0xE3C6, azextloadi16, GRH32, 2>, + Requires<[FeatureHighWord]>; +def LLHRL : UnaryRILPC<"llhrl", 0xC42, aligned_azextloadi16, GR32>; // 64-bit extensions from memory. -def LLGC : UnaryRXY<"llgc", 0xE390, zextloadi8, GR64, 1>; -def LLGH : UnaryRXY<"llgh", 0xE391, zextloadi16, GR64, 2>; -def LLGF : UnaryRXY<"llgf", 0xE316, zextloadi32, GR64, 4>; -def LLGHRL : UnaryRILPC<"llghrl", 0xC46, aligned_zextloadi16, GR64>; -def LLGFRL : UnaryRILPC<"llgfrl", 0xC4E, aligned_zextloadi32, GR64>; +def LLGC : UnaryRXY<"llgc", 0xE390, azextloadi8, GR64, 1>; +def LLGH : UnaryRXY<"llgh", 0xE391, azextloadi16, GR64, 2>; +def LLGF : UnaryRXY<"llgf", 0xE316, azextloadi32, GR64, 4>; +def LLGHRL : UnaryRILPC<"llghrl", 0xC46, aligned_azextloadi16, GR64>; +def LLGFRL : UnaryRILPC<"llgfrl", 0xC4E, aligned_azextloadi32, GR64>; //===----------------------------------------------------------------------===// // Truncations @@ -447,21 +515,31 @@ def LLGFRL : UnaryRILPC<"llgfrl", 0xC4E, aligned_zextloadi32, GR64>; // Truncations of 64-bit registers to 32-bit registers. def : Pat<(i32 (trunc GR64:$src)), - (EXTRACT_SUBREG GR64:$src, subreg_32bit)>; + (EXTRACT_SUBREG GR64:$src, subreg_l32)>; -// Truncations of 32-bit registers to memory. -let isCodeGenOnly = 1 in { - defm STC32 : StoreRXPair<"stc", 0x42, 0xE372, truncstorei8, GR32, 1>; - defm STH32 : StoreRXPair<"sth", 0x40, 0xE370, truncstorei16, GR32, 2>; - def STHRL32 : StoreRILPC<"sthrl", 0xC47, aligned_truncstorei16, GR32>; -} +// Truncations of 32-bit registers to 8-bit memory. STCMux expands to +// STC, STCY or STCH, depending on the choice of register. +def STCMux : StoreRXYPseudo, + Requires<[FeatureHighWord]>; +defm STC : StoreRXPair<"stc", 0x42, 0xE372, truncstorei8, GR32, 1>; +def STCH : StoreRXY<"stch", 0xE3C3, truncstorei8, GRH32, 1>, + Requires<[FeatureHighWord]>; + +// Truncations of 32-bit registers to 16-bit memory. STHMux expands to +// STH, STHY or STHH, depending on the choice of register. +def STHMux : StoreRXYPseudo, + Requires<[FeatureHighWord]>; +defm STH : StoreRXPair<"sth", 0x40, 0xE370, truncstorei16, GR32, 2>; +def STHH : StoreRXY<"sthh", 0xE3C7, truncstorei16, GRH32, 2>, + Requires<[FeatureHighWord]>; +def STHRL : StoreRILPC<"sthrl", 0xC47, aligned_truncstorei16, GR32>; // Truncations of 64-bit registers to memory. -defm STC : StoreRXPair<"stc", 0x42, 0xE372, truncstorei8, GR64, 1>; -defm STH : StoreRXPair<"sth", 0x40, 0xE370, truncstorei16, GR64, 2>; -def STHRL : StoreRILPC<"sthrl", 0xC47, aligned_truncstorei16, GR64>; -defm ST : StoreRXPair<"st", 0x50, 0xE350, truncstorei32, GR64, 4>; -def STRL : StoreRILPC<"strl", 0xC4F, aligned_truncstorei32, GR64>; +defm : StoreGR64Pair; +defm : StoreGR64Pair; +def : StoreGR64PC; +defm : StoreGR64Pair; +def : StoreGR64PC; //===----------------------------------------------------------------------===// // Multi-register moves @@ -478,7 +556,7 @@ def STMG : StoreMultipleRSY<"stmg", 0xEB24, GR64>; //===----------------------------------------------------------------------===// // Byte-swapping register moves. -let neverHasSideEffects = 1 in { +let hasSideEffects = 0 in { def LRVR : UnaryRRE<"lrv", 0xB91F, bswap, GR32, GR32>; def LRVGR : UnaryRRE<"lrvg", 0xB90F, bswap, GR64, GR64>; } @@ -498,7 +576,7 @@ def STRVG : StoreRXY<"strvg", 0xE32F, storeu, //===----------------------------------------------------------------------===// // Load BDX-style addresses. -let neverHasSideEffects = 1, isAsCheapAsAMove = 1, isReMaterializable = 1, +let hasSideEffects = 0, isAsCheapAsAMove = 1, isReMaterializable = 1, DispKey = "la" in { let DispSize = "12" in def LA : InstRX<0x41, (outs GR64:$R1), (ins laaddr12pair:$XBD2), @@ -512,17 +590,49 @@ let neverHasSideEffects = 1, isAsCheapAsAMove = 1, isReMaterializable = 1, // Load a PC-relative address. There's no version of this instruction // with a 16-bit offset, so there's no relaxation. -let neverHasSideEffects = 1, isAsCheapAsAMove = 1, isMoveImm = 1, +let hasSideEffects = 0, isAsCheapAsAMove = 1, isMoveImm = 1, isReMaterializable = 1 in { def LARL : InstRIL<0xC00, (outs GR64:$R1), (ins pcrel32:$I2), "larl\t$R1, $I2", [(set GR64:$R1, pcrel32:$I2)]>; } +// Load the Global Offset Table address. This will be lowered into a +// larl $R1, _GLOBAL_OFFSET_TABLE_ +// instruction. +def GOT : Alias<6, (outs GR64:$R1), (ins), + [(set GR64:$R1, (global_offset_table))]>; + //===----------------------------------------------------------------------===// -// Negation +// Absolute and Negation //===----------------------------------------------------------------------===// +let Defs = [CC] in { + let CCValues = 0xF, CompareZeroCCMask = 0x8 in { + def LPR : UnaryRR <"lp", 0x10, z_iabs, GR32, GR32>; + def LPGR : UnaryRRE<"lpg", 0xB900, z_iabs, GR64, GR64>; + } + let CCValues = 0xE, CompareZeroCCMask = 0xE in + def LPGFR : UnaryRRE<"lpgf", 0xB910, null_frag, GR64, GR32>; +} +def : Pat<(z_iabs32 GR32:$src), (LPR GR32:$src)>; +def : Pat<(z_iabs64 GR64:$src), (LPGR GR64:$src)>; +defm : SXU; +defm : SXU; + +let Defs = [CC] in { + let CCValues = 0xF, CompareZeroCCMask = 0x8 in { + def LNR : UnaryRR <"ln", 0x11, z_inegabs, GR32, GR32>; + def LNGR : UnaryRRE<"lng", 0xB901, z_inegabs, GR64, GR64>; + } + let CCValues = 0xE, CompareZeroCCMask = 0xE in + def LNGFR : UnaryRRE<"lngf", 0xB911, null_frag, GR64, GR32>; +} +def : Pat<(z_inegabs32 GR32:$src), (LNR GR32:$src)>; +def : Pat<(z_inegabs64 GR64:$src), (LNGR GR64:$src)>; +defm : SXU; +defm : SXU; + let Defs = [CC] in { let CCValues = 0xF, CompareZeroCCMask = 0x8 in { def LCR : UnaryRR <"lc", 0x13, ineg, GR32, GR32>; @@ -538,43 +648,51 @@ defm : SXU; //===----------------------------------------------------------------------===// let isCodeGenOnly = 1 in - defm IC32 : BinaryRXPair<"ic", 0x43, 0xE373, inserti8, GR32, zextloadi8, 1>; -defm IC : BinaryRXPair<"ic", 0x43, 0xE373, inserti8, GR64, zextloadi8, 1>; + defm IC32 : BinaryRXPair<"ic", 0x43, 0xE373, inserti8, GR32, azextloadi8, 1>; +defm IC : BinaryRXPair<"ic", 0x43, 0xE373, inserti8, GR64, azextloadi8, 1>; -defm : InsertMem<"inserti8", IC32, GR32, zextloadi8, bdxaddr12pair>; -defm : InsertMem<"inserti8", IC32Y, GR32, zextloadi8, bdxaddr20pair>; +defm : InsertMem<"inserti8", IC32, GR32, azextloadi8, bdxaddr12pair>; +defm : InsertMem<"inserti8", IC32Y, GR32, azextloadi8, bdxaddr20pair>; -defm : InsertMem<"inserti8", IC, GR64, zextloadi8, bdxaddr12pair>; -defm : InsertMem<"inserti8", ICY, GR64, zextloadi8, bdxaddr20pair>; +defm : InsertMem<"inserti8", IC, GR64, azextloadi8, bdxaddr12pair>; +defm : InsertMem<"inserti8", ICY, GR64, azextloadi8, bdxaddr20pair>; // Insertions of a 16-bit immediate, leaving other bits unaffected. // We don't have or_as_insert equivalents of these operations because // OI is available instead. -let isCodeGenOnly = 1 in { - def IILL32 : BinaryRI<"iill", 0xA53, insertll, GR32, imm32ll16>; - def IILH32 : BinaryRI<"iilh", 0xA52, insertlh, GR32, imm32lh16>; -} -def IILL : BinaryRI<"iill", 0xA53, insertll, GR64, imm64ll16>; -def IILH : BinaryRI<"iilh", 0xA52, insertlh, GR64, imm64lh16>; -def IIHL : BinaryRI<"iihl", 0xA51, inserthl, GR64, imm64hl16>; -def IIHH : BinaryRI<"iihh", 0xA50, inserthh, GR64, imm64hh16>; +// +// IIxMux expands to II[LH]x, depending on the choice of register. +def IILMux : BinaryRIPseudo, + Requires<[FeatureHighWord]>; +def IIHMux : BinaryRIPseudo, + Requires<[FeatureHighWord]>; +def IILL : BinaryRI<"iill", 0xA53, insertll, GR32, imm32ll16>; +def IILH : BinaryRI<"iilh", 0xA52, insertlh, GR32, imm32lh16>; +def IIHL : BinaryRI<"iihl", 0xA51, insertll, GRH32, imm32ll16>; +def IIHH : BinaryRI<"iihh", 0xA50, insertlh, GRH32, imm32lh16>; +def IILL64 : BinaryAliasRI; +def IILH64 : BinaryAliasRI; +def IIHL64 : BinaryAliasRI; +def IIHH64 : BinaryAliasRI; // ...likewise for 32-bit immediates. For GR32s this is a general // full-width move. (We use IILF rather than something like LLILF // for 32-bit moves because IILF leaves the upper 32 bits of the // GR64 unchanged.) -let isCodeGenOnly = 1, isAsCheapAsAMove = 1, isMoveImm = 1, - isReMaterializable = 1 in { - def IILF32 : UnaryRIL<"iilf", 0xC09, bitconvert, GR32, uimm32>; +let isAsCheapAsAMove = 1, isMoveImm = 1, isReMaterializable = 1 in { + def IIFMux : UnaryRIPseudo, + Requires<[FeatureHighWord]>; + def IILF : UnaryRIL<"iilf", 0xC09, bitconvert, GR32, uimm32>; + def IIHF : UnaryRIL<"iihf", 0xC08, bitconvert, GRH32, uimm32>; } -def IILF : BinaryRIL<"iilf", 0xC09, insertlf, GR64, imm64lf32>; -def IIHF : BinaryRIL<"iihf", 0xC08, inserthf, GR64, imm64hf32>; +def IILF64 : BinaryAliasRIL; +def IIHF64 : BinaryAliasRIL; // An alternative model of inserthf, with the first operand being // a zero-extended value. def : Pat<(or (zext32 GR32:$src), imm64hf32:$imm), - (IIHF (INSERT_SUBREG (i64 (IMPLICIT_DEF)), GR32:$src, subreg_32bit), - imm64hf32:$imm)>; + (IIHF64 (INSERT_SUBREG (i64 (IMPLICIT_DEF)), GR32:$src, subreg_l32), + imm64hf32:$imm)>; //===----------------------------------------------------------------------===// // Addition @@ -590,17 +708,22 @@ let Defs = [CC], CCValues = 0xF, CompareZeroCCMask = 0x8 in { def AGFR : BinaryRRE<"agf", 0xB918, null_frag, GR64, GR32>; // Addition of signed 16-bit immediates. + defm AHIMux : BinaryRIAndKPseudo<"ahimux", add, GRX32, imm32sx16>; defm AHI : BinaryRIAndK<"ahi", 0xA7A, 0xECD8, add, GR32, imm32sx16>; defm AGHI : BinaryRIAndK<"aghi", 0xA7B, 0xECD9, add, GR64, imm64sx16>; // Addition of signed 32-bit immediates. + def AFIMux : BinaryRIPseudo, + Requires<[FeatureHighWord]>; def AFI : BinaryRIL<"afi", 0xC29, add, GR32, simm32>; + def AIH : BinaryRIL<"aih", 0xCC8, add, GRH32, simm32>, + Requires<[FeatureHighWord]>; def AGFI : BinaryRIL<"agfi", 0xC28, add, GR64, imm64sx32>; // Addition of memory. - defm AH : BinaryRXPair<"ah", 0x4A, 0xE37A, add, GR32, sextloadi16, 2>; + defm AH : BinaryRXPair<"ah", 0x4A, 0xE37A, add, GR32, asextloadi16, 2>; defm A : BinaryRXPair<"a", 0x5A, 0xE35A, add, GR32, load, 4>; - def AGF : BinaryRXY<"agf", 0xE318, add, GR64, sextloadi32, 4>; + def AGF : BinaryRXY<"agf", 0xE318, add, GR64, asextloadi32, 4>; def AG : BinaryRXY<"ag", 0xE308, add, GR64, load, 8>; // Addition to memory. @@ -630,7 +753,7 @@ let Defs = [CC] in { // Addition of memory. defm AL : BinaryRXPair<"al", 0x5E, 0xE35E, addc, GR32, load, 4>; - def ALGF : BinaryRXY<"algf", 0xE31A, addc, GR64, zextloadi32, 4>; + def ALGF : BinaryRXY<"algf", 0xE31A, addc, GR64, azextloadi32, 4>; def ALG : BinaryRXY<"alg", 0xE30A, addc, GR64, load, 8>; } defm : ZXB; @@ -650,7 +773,7 @@ let Defs = [CC], Uses = [CC] in { // Subtraction //===----------------------------------------------------------------------===// -// Plain substraction. Although immediate forms exist, we use the +// Plain subtraction. Although immediate forms exist, we use the // add-immediate instruction instead. let Defs = [CC], CCValues = 0xF, CompareZeroCCMask = 0x8 in { // Subtraction of a register. @@ -659,9 +782,9 @@ let Defs = [CC], CCValues = 0xF, CompareZeroCCMask = 0x8 in { defm SGR : BinaryRREAndK<"sg", 0xB909, 0xB9E9, sub, GR64, GR64>; // Subtraction of memory. - defm SH : BinaryRXPair<"sh", 0x4B, 0xE37B, sub, GR32, sextloadi16, 2>; + defm SH : BinaryRXPair<"sh", 0x4B, 0xE37B, sub, GR32, asextloadi16, 2>; defm S : BinaryRXPair<"s", 0x5B, 0xE35B, sub, GR32, load, 4>; - def SGF : BinaryRXY<"sgf", 0xE319, sub, GR64, sextloadi32, 4>; + def SGF : BinaryRXY<"sgf", 0xE319, sub, GR64, asextloadi32, 4>; def SG : BinaryRXY<"sg", 0xE309, sub, GR64, load, 8>; } defm : SXB; @@ -680,7 +803,7 @@ let Defs = [CC] in { // Subtraction of memory. defm SL : BinaryRXPair<"sl", 0x5F, 0xE35F, subc, GR32, load, 4>; - def SLGF : BinaryRXY<"slgf", 0xE31B, subc, GR64, zextloadi32, 4>; + def SLGF : BinaryRXY<"slgf", 0xE31B, subc, GR64, azextloadi32, 4>; def SLG : BinaryRXY<"slg", 0xE30B, subc, GR64, load, 8>; } defm : ZXB; @@ -710,22 +833,33 @@ let Defs = [CC] in { let isConvertibleToThreeAddress = 1 in { // ANDs of a 16-bit immediate, leaving other bits unaffected. // The CC result only reflects the 16-bit field, not the full register. - let isCodeGenOnly = 1 in { - def NILL32 : BinaryRI<"nill", 0xA57, and, GR32, imm32ll16c>; - def NILH32 : BinaryRI<"nilh", 0xA56, and, GR32, imm32lh16c>; - } - def NILL : BinaryRI<"nill", 0xA57, and, GR64, imm64ll16c>; - def NILH : BinaryRI<"nilh", 0xA56, and, GR64, imm64lh16c>; - def NIHL : BinaryRI<"nihl", 0xA55, and, GR64, imm64hl16c>; - def NIHH : BinaryRI<"nihh", 0xA54, and, GR64, imm64hh16c>; + // + // NIxMux expands to NI[LH]x, depending on the choice of register. + def NILMux : BinaryRIPseudo, + Requires<[FeatureHighWord]>; + def NIHMux : BinaryRIPseudo, + Requires<[FeatureHighWord]>; + def NILL : BinaryRI<"nill", 0xA57, and, GR32, imm32ll16c>; + def NILH : BinaryRI<"nilh", 0xA56, and, GR32, imm32lh16c>; + def NIHL : BinaryRI<"nihl", 0xA55, and, GRH32, imm32ll16c>; + def NIHH : BinaryRI<"nihh", 0xA54, and, GRH32, imm32lh16c>; + def NILL64 : BinaryAliasRI; + def NILH64 : BinaryAliasRI; + def NIHL64 : BinaryAliasRI; + def NIHH64 : BinaryAliasRI; // ANDs of a 32-bit immediate, leaving other bits unaffected. // The CC result only reflects the 32-bit field, which means we can // use it as a zero indicator for i32 operations but not otherwise. - let isCodeGenOnly = 1, CCValues = 0xC, CompareZeroCCMask = 0x8 in - def NILF32 : BinaryRIL<"nilf", 0xC0B, and, GR32, uimm32>; - def NILF : BinaryRIL<"nilf", 0xC0B, and, GR64, imm64lf32c>; - def NIHF : BinaryRIL<"nihf", 0xC0A, and, GR64, imm64hf32c>; + let CCValues = 0xC, CompareZeroCCMask = 0x8 in { + // Expands to NILF or NIHF, depending on the choice of register. + def NIFMux : BinaryRIPseudo, + Requires<[FeatureHighWord]>; + def NILF : BinaryRIL<"nilf", 0xC0B, and, GR32, uimm32>; + def NIHF : BinaryRIL<"nihf", 0xC0A, and, GRH32, uimm32>; + } + def NILF64 : BinaryAliasRIL; + def NIHF64 : BinaryAliasRIL; } // ANDs of memory. @@ -735,7 +869,11 @@ let Defs = [CC] in { } // AND to memory - defm NI : BinarySIPair<"ni", 0x94, 0xEB54, null_frag, uimm8>; + defm NI : BinarySIPair<"ni", 0x94, 0xEB54, null_frag, imm32zx8>; + + // Block AND. + let mayLoad = 1, mayStore = 1 in + defm NC : MemorySS<"nc", 0xD4, z_nc, z_nc_loop>; } defm : RMWIByte; defm : RMWIByte; @@ -753,22 +891,33 @@ let Defs = [CC] in { // ORs of a 16-bit immediate, leaving other bits unaffected. // The CC result only reflects the 16-bit field, not the full register. - let isCodeGenOnly = 1 in { - def OILL32 : BinaryRI<"oill", 0xA5B, or, GR32, imm32ll16>; - def OILH32 : BinaryRI<"oilh", 0xA5A, or, GR32, imm32lh16>; - } - def OILL : BinaryRI<"oill", 0xA5B, or, GR64, imm64ll16>; - def OILH : BinaryRI<"oilh", 0xA5A, or, GR64, imm64lh16>; - def OIHL : BinaryRI<"oihl", 0xA59, or, GR64, imm64hl16>; - def OIHH : BinaryRI<"oihh", 0xA58, or, GR64, imm64hh16>; + // + // OIxMux expands to OI[LH]x, depending on the choice of register. + def OILMux : BinaryRIPseudo, + Requires<[FeatureHighWord]>; + def OIHMux : BinaryRIPseudo, + Requires<[FeatureHighWord]>; + def OILL : BinaryRI<"oill", 0xA5B, or, GR32, imm32ll16>; + def OILH : BinaryRI<"oilh", 0xA5A, or, GR32, imm32lh16>; + def OIHL : BinaryRI<"oihl", 0xA59, or, GRH32, imm32ll16>; + def OIHH : BinaryRI<"oihh", 0xA58, or, GRH32, imm32lh16>; + def OILL64 : BinaryAliasRI; + def OILH64 : BinaryAliasRI; + def OIHL64 : BinaryAliasRI; + def OIHH64 : BinaryAliasRI; // ORs of a 32-bit immediate, leaving other bits unaffected. // The CC result only reflects the 32-bit field, which means we can // use it as a zero indicator for i32 operations but not otherwise. - let isCodeGenOnly = 1, CCValues = 0xC, CompareZeroCCMask = 0x8 in - def OILF32 : BinaryRIL<"oilf", 0xC0D, or, GR32, uimm32>; - def OILF : BinaryRIL<"oilf", 0xC0D, or, GR64, imm64lf32>; - def OIHF : BinaryRIL<"oihf", 0xC0C, or, GR64, imm64hf32>; + let CCValues = 0xC, CompareZeroCCMask = 0x8 in { + // Expands to OILF or OIHF, depending on the choice of register. + def OIFMux : BinaryRIPseudo, + Requires<[FeatureHighWord]>; + def OILF : BinaryRIL<"oilf", 0xC0D, or, GR32, uimm32>; + def OIHF : BinaryRIL<"oihf", 0xC0C, or, GRH32, uimm32>; + } + def OILF64 : BinaryAliasRIL; + def OIHF64 : BinaryAliasRIL; // ORs of memory. let CCValues = 0xC, CompareZeroCCMask = 0x8 in { @@ -777,7 +926,11 @@ let Defs = [CC] in { } // OR to memory - defm OI : BinarySIPair<"oi", 0x96, 0xEB56, null_frag, uimm8>; + defm OI : BinarySIPair<"oi", 0x96, 0xEB56, null_frag, imm32zx8>; + + // Block OR. + let mayLoad = 1, mayStore = 1 in + defm OC : MemorySS<"oc", 0xD6, z_oc, z_oc_loop>; } defm : RMWIByte; defm : RMWIByte; @@ -796,10 +949,15 @@ let Defs = [CC] in { // XORs of a 32-bit immediate, leaving other bits unaffected. // The CC result only reflects the 32-bit field, which means we can // use it as a zero indicator for i32 operations but not otherwise. - let isCodeGenOnly = 1, CCValues = 0xC, CompareZeroCCMask = 0x8 in - def XILF32 : BinaryRIL<"xilf", 0xC07, xor, GR32, uimm32>; - def XILF : BinaryRIL<"xilf", 0xC07, xor, GR64, imm64lf32>; - def XIHF : BinaryRIL<"xihf", 0xC06, xor, GR64, imm64hf32>; + let CCValues = 0xC, CompareZeroCCMask = 0x8 in { + // Expands to XILF or XIHF, depending on the choice of register. + def XIFMux : BinaryRIPseudo, + Requires<[FeatureHighWord]>; + def XILF : BinaryRIL<"xilf", 0xC07, xor, GR32, uimm32>; + def XIHF : BinaryRIL<"xihf", 0xC06, xor, GRH32, uimm32>; + } + def XILF64 : BinaryAliasRIL; + def XIHF64 : BinaryAliasRIL; // XORs of memory. let CCValues = 0xC, CompareZeroCCMask = 0x8 in { @@ -808,7 +966,11 @@ let Defs = [CC] in { } // XOR to memory - defm XI : BinarySIPair<"xi", 0x97, 0xEB57, null_frag, uimm8>; + defm XI : BinarySIPair<"xi", 0x97, 0xEB57, null_frag, imm32zx8>; + + // Block XOR. + let mayLoad = 1, mayStore = 1 in + defm XC : MemorySS<"xc", 0xD7, z_xc, z_xc_loop>; } defm : RMWIByte; defm : RMWIByte; @@ -834,9 +996,9 @@ def MSFI : BinaryRIL<"msfi", 0xC21, mul, GR32, simm32>; def MSGFI : BinaryRIL<"msgfi", 0xC20, mul, GR64, imm64sx32>; // Multiplication of memory. -defm MH : BinaryRXPair<"mh", 0x4C, 0xE37C, mul, GR32, sextloadi16, 2>; +defm MH : BinaryRXPair<"mh", 0x4C, 0xE37C, mul, GR32, asextloadi16, 2>; defm MS : BinaryRXPair<"ms", 0x71, 0xE351, mul, GR32, load, 4>; -def MSGF : BinaryRXY<"msgf", 0xE31C, mul, GR64, sextloadi32, 4>; +def MSGF : BinaryRXY<"msgf", 0xE31C, mul, GR64, asextloadi32, 4>; def MSG : BinaryRXY<"msg", 0xE30C, mul, GR64, load, 8>; // Multiplication of a register, producing two results. @@ -866,27 +1028,27 @@ def DLG : BinaryRXY<"dlg", 0xE387, z_udivrem64, GR128, load, 8>; //===----------------------------------------------------------------------===// // Shift left. -let neverHasSideEffects = 1 in { - defm SLL : ShiftRSAndK<"sll", 0x89, 0xEBDF, shl, GR32>; - def SLLG : ShiftRSY<"sllg", 0xEB0D, shl, GR64>; +let hasSideEffects = 0 in { + defm SLL : BinaryRSAndK<"sll", 0x89, 0xEBDF, shl, GR32>; + def SLLG : BinaryRSY<"sllg", 0xEB0D, shl, GR64>; } // Logical shift right. -let neverHasSideEffects = 1 in { - defm SRL : ShiftRSAndK<"srl", 0x88, 0xEBDE, srl, GR32>; - def SRLG : ShiftRSY<"srlg", 0xEB0C, srl, GR64>; +let hasSideEffects = 0 in { + defm SRL : BinaryRSAndK<"srl", 0x88, 0xEBDE, srl, GR32>; + def SRLG : BinaryRSY<"srlg", 0xEB0C, srl, GR64>; } // Arithmetic shift right. let Defs = [CC], CCValues = 0xE, CompareZeroCCMask = 0xE in { - defm SRA : ShiftRSAndK<"sra", 0x8A, 0xEBDC, sra, GR32>; - def SRAG : ShiftRSY<"srag", 0xEB0A, sra, GR64>; + defm SRA : BinaryRSAndK<"sra", 0x8A, 0xEBDC, sra, GR32>; + def SRAG : BinaryRSY<"srag", 0xEB0A, sra, GR64>; } // Rotate left. -let neverHasSideEffects = 1 in { - def RLL : ShiftRSY<"rll", 0xEB1D, rotl, GR32>; - def RLLG : ShiftRSY<"rllg", 0xEB1C, rotl, GR64>; +let hasSideEffects = 0 in { + def RLL : BinaryRSY<"rll", 0xEB1D, rotl, GR32>; + def RLLG : BinaryRSY<"rllg", 0xEB1C, rotl, GR64>; } // Rotate second operand left and inserted selected bits into first operand. @@ -899,15 +1061,21 @@ let Defs = [CC] in { def RISBG : RotateSelectRIEf<"risbg", 0xEC55, GR64, GR64>; } +// On zEC12 we have a variant of RISBG that does not set CC. +let Predicates = [FeatureMiscellaneousExtensions] in + def RISBGN : RotateSelectRIEf<"risbgn", 0xEC59, GR64, GR64>; + // Forms of RISBG that only affect one word of the destination register. // They do not set CC. -let isCodeGenOnly = 1 in - def RISBLG32 : RotateSelectRIEf<"risblg", 0xEC51, GR32, GR32>, - Requires<[FeatureHighWord]>; -def RISBHG : RotateSelectRIEf<"risbhg", 0xEC5D, GR64, GR64>, - Requires<[FeatureHighWord]>; -def RISBLG : RotateSelectRIEf<"risblg", 0xEC51, GR64, GR64>, - Requires<[FeatureHighWord]>; +let Predicates = [FeatureHighWord] in { + def RISBMux : RotateSelectRIEfPseudo; + def RISBLL : RotateSelectAliasRIEf; + def RISBLH : RotateSelectAliasRIEf; + def RISBHL : RotateSelectAliasRIEf; + def RISBHH : RotateSelectAliasRIEf; + def RISBLG : RotateSelectRIEf<"risblg", 0xEC51, GR32, GR64>; + def RISBHG : RotateSelectRIEf<"risbhg", 0xEC5D, GRH32, GR64>; +} // Rotate second operand left and perform a logical operation with selected // bits of the first operand. The CC result only describes the selected bits, @@ -922,39 +1090,50 @@ let Defs = [CC] in { // Comparison //===----------------------------------------------------------------------===// -// Signed comparisons. +// Signed comparisons. We put these before the unsigned comparisons because +// some of the signed forms have COMPARE AND BRANCH equivalents whereas none +// of the unsigned forms do. let Defs = [CC], CCValues = 0xE in { // Comparison with a register. - def CR : CompareRR <"c", 0x19, z_cmp, GR32, GR32>; + def CR : CompareRR <"c", 0x19, z_scmp, GR32, GR32>; def CGFR : CompareRRE<"cgf", 0xB930, null_frag, GR64, GR32>; - def CGR : CompareRRE<"cg", 0xB920, z_cmp, GR64, GR64>; + def CGR : CompareRRE<"cg", 0xB920, z_scmp, GR64, GR64>; // Comparison with a signed 16-bit immediate. - def CHI : CompareRI<"chi", 0xA7E, z_cmp, GR32, imm32sx16>; - def CGHI : CompareRI<"cghi", 0xA7F, z_cmp, GR64, imm64sx16>; - - // Comparison with a signed 32-bit immediate. - def CFI : CompareRIL<"cfi", 0xC2D, z_cmp, GR32, simm32>; - def CGFI : CompareRIL<"cgfi", 0xC2C, z_cmp, GR64, imm64sx32>; + def CHI : CompareRI<"chi", 0xA7E, z_scmp, GR32, imm32sx16>; + def CGHI : CompareRI<"cghi", 0xA7F, z_scmp, GR64, imm64sx16>; + + // Comparison with a signed 32-bit immediate. CFIMux expands to CFI or CIH, + // depending on the choice of register. + def CFIMux : CompareRIPseudo, + Requires<[FeatureHighWord]>; + def CFI : CompareRIL<"cfi", 0xC2D, z_scmp, GR32, simm32>; + def CIH : CompareRIL<"cih", 0xCCD, z_scmp, GRH32, simm32>, + Requires<[FeatureHighWord]>; + def CGFI : CompareRIL<"cgfi", 0xC2C, z_scmp, GR64, imm64sx32>; // Comparison with memory. - defm CH : CompareRXPair<"ch", 0x49, 0xE379, z_cmp, GR32, sextloadi16, 2>; - defm C : CompareRXPair<"c", 0x59, 0xE359, z_cmp, GR32, load, 4>; - def CGH : CompareRXY<"cgh", 0xE334, z_cmp, GR64, sextloadi16, 2>; - def CGF : CompareRXY<"cgf", 0xE330, z_cmp, GR64, sextloadi32, 4>; - def CG : CompareRXY<"cg", 0xE320, z_cmp, GR64, load, 8>; - def CHRL : CompareRILPC<"chrl", 0xC65, z_cmp, GR32, aligned_sextloadi16>; - def CRL : CompareRILPC<"crl", 0xC6D, z_cmp, GR32, aligned_load>; - def CGHRL : CompareRILPC<"cghrl", 0xC64, z_cmp, GR64, aligned_sextloadi16>; - def CGFRL : CompareRILPC<"cgfrl", 0xC6C, z_cmp, GR64, aligned_sextloadi32>; - def CGRL : CompareRILPC<"cgrl", 0xC68, z_cmp, GR64, aligned_load>; + defm CH : CompareRXPair<"ch", 0x49, 0xE379, z_scmp, GR32, asextloadi16, 2>; + def CMux : CompareRXYPseudo, + Requires<[FeatureHighWord]>; + defm C : CompareRXPair<"c", 0x59, 0xE359, z_scmp, GR32, load, 4>; + def CHF : CompareRXY<"chf", 0xE3CD, z_scmp, GRH32, load, 4>, + Requires<[FeatureHighWord]>; + def CGH : CompareRXY<"cgh", 0xE334, z_scmp, GR64, asextloadi16, 2>; + def CGF : CompareRXY<"cgf", 0xE330, z_scmp, GR64, asextloadi32, 4>; + def CG : CompareRXY<"cg", 0xE320, z_scmp, GR64, load, 8>; + def CHRL : CompareRILPC<"chrl", 0xC65, z_scmp, GR32, aligned_asextloadi16>; + def CRL : CompareRILPC<"crl", 0xC6D, z_scmp, GR32, aligned_load>; + def CGHRL : CompareRILPC<"cghrl", 0xC64, z_scmp, GR64, aligned_asextloadi16>; + def CGFRL : CompareRILPC<"cgfrl", 0xC6C, z_scmp, GR64, aligned_asextloadi32>; + def CGRL : CompareRILPC<"cgrl", 0xC68, z_scmp, GR64, aligned_load>; // Comparison between memory and a signed 16-bit immediate. - def CHHSI : CompareSIL<"chhsi", 0xE554, z_cmp, sextloadi16, imm32sx16>; - def CHSI : CompareSIL<"chsi", 0xE55C, z_cmp, load, imm32sx16>; - def CGHSI : CompareSIL<"cghsi", 0xE558, z_cmp, load, imm64sx16>; + def CHHSI : CompareSIL<"chhsi", 0xE554, z_scmp, asextloadi16, imm32sx16>; + def CHSI : CompareSIL<"chsi", 0xE55C, z_scmp, load, imm32sx16>; + def CGHSI : CompareSIL<"cghsi", 0xE558, z_scmp, load, imm64sx16>; } -defm : SXB; +defm : SXB; // Unsigned comparisons. let Defs = [CC], CCValues = 0xE, IsLogical = 1 in { @@ -963,121 +1142,191 @@ let Defs = [CC], CCValues = 0xE, IsLogical = 1 in { def CLGFR : CompareRRE<"clgf", 0xB931, null_frag, GR64, GR32>; def CLGR : CompareRRE<"clg", 0xB921, z_ucmp, GR64, GR64>; - // Comparison with a signed 32-bit immediate. + // Comparison with an unsigned 32-bit immediate. CLFIMux expands to CLFI + // or CLIH, depending on the choice of register. + def CLFIMux : CompareRIPseudo, + Requires<[FeatureHighWord]>; def CLFI : CompareRIL<"clfi", 0xC2F, z_ucmp, GR32, uimm32>; + def CLIH : CompareRIL<"clih", 0xCCF, z_ucmp, GRH32, uimm32>, + Requires<[FeatureHighWord]>; def CLGFI : CompareRIL<"clgfi", 0xC2E, z_ucmp, GR64, imm64zx32>; // Comparison with memory. + def CLMux : CompareRXYPseudo, + Requires<[FeatureHighWord]>; defm CL : CompareRXPair<"cl", 0x55, 0xE355, z_ucmp, GR32, load, 4>; - def CLGF : CompareRXY<"clgf", 0xE331, z_ucmp, GR64, zextloadi32, 4>; + def CLHF : CompareRXY<"clhf", 0xE3CF, z_ucmp, GRH32, load, 4>, + Requires<[FeatureHighWord]>; + def CLGF : CompareRXY<"clgf", 0xE331, z_ucmp, GR64, azextloadi32, 4>; def CLG : CompareRXY<"clg", 0xE321, z_ucmp, GR64, load, 8>; def CLHRL : CompareRILPC<"clhrl", 0xC67, z_ucmp, GR32, - aligned_zextloadi16>; + aligned_azextloadi16>; def CLRL : CompareRILPC<"clrl", 0xC6F, z_ucmp, GR32, aligned_load>; def CLGHRL : CompareRILPC<"clghrl", 0xC66, z_ucmp, GR64, - aligned_zextloadi16>; + aligned_azextloadi16>; def CLGFRL : CompareRILPC<"clgfrl", 0xC6E, z_ucmp, GR64, - aligned_zextloadi32>; + aligned_azextloadi32>; def CLGRL : CompareRILPC<"clgrl", 0xC6A, z_ucmp, GR64, aligned_load>; // Comparison between memory and an unsigned 8-bit immediate. - defm CLI : CompareSIPair<"cli", 0x95, 0xEB55, z_ucmp, zextloadi8, imm32zx8>; + defm CLI : CompareSIPair<"cli", 0x95, 0xEB55, z_ucmp, azextloadi8, imm32zx8>; // Comparison between memory and an unsigned 16-bit immediate. - def CLHHSI : CompareSIL<"clhhsi", 0xE555, z_ucmp, zextloadi16, imm32zx16>; - def CLFHSI : CompareSIL<"clfhsi", 0xE55D, z_ucmp, load, imm32zx16>; - def CLGHSI : CompareSIL<"clghsi", 0xE559, z_ucmp, load, imm64zx16>; + def CLHHSI : CompareSIL<"clhhsi", 0xE555, z_ucmp, azextloadi16, imm32zx16>; + def CLFHSI : CompareSIL<"clfhsi", 0xE55D, z_ucmp, load, imm32zx16>; + def CLGHSI : CompareSIL<"clghsi", 0xE559, z_ucmp, load, imm64zx16>; } defm : ZXB; // Memory-to-memory comparison. let mayLoad = 1, Defs = [CC] in - defm CLC : MemorySS<"clc", 0xD5, z_clc>; + defm CLC : MemorySS<"clc", 0xD5, z_clc, z_clc_loop>; // String comparison. -let mayLoad = 1, Defs = [CC], Uses = [R0W] in +let mayLoad = 1, Defs = [CC] in defm CLST : StringRRE<"clst", 0xB25D, z_strcmp>; +// Test under mask. +let Defs = [CC] in { + // TMxMux expands to TM[LH]x, depending on the choice of register. + def TMLMux : CompareRIPseudo, + Requires<[FeatureHighWord]>; + def TMHMux : CompareRIPseudo, + Requires<[FeatureHighWord]>; + def TMLL : CompareRI<"tmll", 0xA71, z_tm_reg, GR32, imm32ll16>; + def TMLH : CompareRI<"tmlh", 0xA70, z_tm_reg, GR32, imm32lh16>; + def TMHL : CompareRI<"tmhl", 0xA73, z_tm_reg, GRH32, imm32ll16>; + def TMHH : CompareRI<"tmhh", 0xA72, z_tm_reg, GRH32, imm32lh16>; + + def TMLL64 : CompareAliasRI; + def TMLH64 : CompareAliasRI; + def TMHL64 : CompareAliasRI; + def TMHH64 : CompareAliasRI; + + defm TM : CompareSIPair<"tm", 0x91, 0xEB51, z_tm_mem, anyextloadi8, imm32zx8>; +} + +//===----------------------------------------------------------------------===// +// Prefetch +//===----------------------------------------------------------------------===// + +def PFD : PrefetchRXY<"pfd", 0xE336, z_prefetch>; +def PFDRL : PrefetchRILPC<"pfdrl", 0xC62, z_prefetch>; + //===----------------------------------------------------------------------===// // Atomic operations //===----------------------------------------------------------------------===// -def ATOMIC_SWAPW : AtomicLoadWBinaryReg; -def ATOMIC_SWAP_32 : AtomicLoadBinaryReg32; -def ATOMIC_SWAP_64 : AtomicLoadBinaryReg64; - -def ATOMIC_LOADW_AR : AtomicLoadWBinaryReg; -def ATOMIC_LOADW_AFI : AtomicLoadWBinaryImm; -def ATOMIC_LOAD_AR : AtomicLoadBinaryReg32; -def ATOMIC_LOAD_AHI : AtomicLoadBinaryImm32; -def ATOMIC_LOAD_AFI : AtomicLoadBinaryImm32; -def ATOMIC_LOAD_AGR : AtomicLoadBinaryReg64; -def ATOMIC_LOAD_AGHI : AtomicLoadBinaryImm64; -def ATOMIC_LOAD_AGFI : AtomicLoadBinaryImm64; - -def ATOMIC_LOADW_SR : AtomicLoadWBinaryReg; -def ATOMIC_LOAD_SR : AtomicLoadBinaryReg32; -def ATOMIC_LOAD_SGR : AtomicLoadBinaryReg64; - -def ATOMIC_LOADW_NR : AtomicLoadWBinaryReg; -def ATOMIC_LOADW_NILH : AtomicLoadWBinaryImm; -def ATOMIC_LOAD_NR : AtomicLoadBinaryReg32; -def ATOMIC_LOAD_NILL32 : AtomicLoadBinaryImm32; -def ATOMIC_LOAD_NILH32 : AtomicLoadBinaryImm32; -def ATOMIC_LOAD_NILF32 : AtomicLoadBinaryImm32; -def ATOMIC_LOAD_NGR : AtomicLoadBinaryReg64; -def ATOMIC_LOAD_NILL : AtomicLoadBinaryImm64; -def ATOMIC_LOAD_NILH : AtomicLoadBinaryImm64; -def ATOMIC_LOAD_NIHL : AtomicLoadBinaryImm64; -def ATOMIC_LOAD_NIHH : AtomicLoadBinaryImm64; -def ATOMIC_LOAD_NILF : AtomicLoadBinaryImm64; -def ATOMIC_LOAD_NIHF : AtomicLoadBinaryImm64; +// A serialization instruction that acts as a barrier for all memory +// accesses, which expands to "bcr 14, 0". +let hasSideEffects = 1 in +def Serialize : Alias<2, (outs), (ins), [(z_serialize)]>; + +let Predicates = [FeatureInterlockedAccess1], Defs = [CC] in { + def LAA : LoadAndOpRSY<"laa", 0xEBF8, atomic_load_add_32, GR32>; + def LAAG : LoadAndOpRSY<"laag", 0xEBE8, atomic_load_add_64, GR64>; + def LAAL : LoadAndOpRSY<"laal", 0xEBFA, null_frag, GR32>; + def LAALG : LoadAndOpRSY<"laalg", 0xEBEA, null_frag, GR64>; + def LAN : LoadAndOpRSY<"lan", 0xEBF4, atomic_load_and_32, GR32>; + def LANG : LoadAndOpRSY<"lang", 0xEBE4, atomic_load_and_64, GR64>; + def LAO : LoadAndOpRSY<"lao", 0xEBF6, atomic_load_or_32, GR32>; + def LAOG : LoadAndOpRSY<"laog", 0xEBE6, atomic_load_or_64, GR64>; + def LAX : LoadAndOpRSY<"lax", 0xEBF7, atomic_load_xor_32, GR32>; + def LAXG : LoadAndOpRSY<"laxg", 0xEBE7, atomic_load_xor_64, GR64>; +} + +def ATOMIC_SWAPW : AtomicLoadWBinaryReg; +def ATOMIC_SWAP_32 : AtomicLoadBinaryReg32; +def ATOMIC_SWAP_64 : AtomicLoadBinaryReg64; + +def ATOMIC_LOADW_AR : AtomicLoadWBinaryReg; +def ATOMIC_LOADW_AFI : AtomicLoadWBinaryImm; +let Predicates = [FeatureNoInterlockedAccess1] in { + def ATOMIC_LOAD_AR : AtomicLoadBinaryReg32; + def ATOMIC_LOAD_AHI : AtomicLoadBinaryImm32; + def ATOMIC_LOAD_AFI : AtomicLoadBinaryImm32; + def ATOMIC_LOAD_AGR : AtomicLoadBinaryReg64; + def ATOMIC_LOAD_AGHI : AtomicLoadBinaryImm64; + def ATOMIC_LOAD_AGFI : AtomicLoadBinaryImm64; +} + +def ATOMIC_LOADW_SR : AtomicLoadWBinaryReg; +def ATOMIC_LOAD_SR : AtomicLoadBinaryReg32; +def ATOMIC_LOAD_SGR : AtomicLoadBinaryReg64; + +def ATOMIC_LOADW_NR : AtomicLoadWBinaryReg; +def ATOMIC_LOADW_NILH : AtomicLoadWBinaryImm; +let Predicates = [FeatureNoInterlockedAccess1] in { + def ATOMIC_LOAD_NR : AtomicLoadBinaryReg32; + def ATOMIC_LOAD_NILL : AtomicLoadBinaryImm32; + def ATOMIC_LOAD_NILH : AtomicLoadBinaryImm32; + def ATOMIC_LOAD_NILF : AtomicLoadBinaryImm32; + def ATOMIC_LOAD_NGR : AtomicLoadBinaryReg64; + def ATOMIC_LOAD_NILL64 : AtomicLoadBinaryImm64; + def ATOMIC_LOAD_NILH64 : AtomicLoadBinaryImm64; + def ATOMIC_LOAD_NIHL64 : AtomicLoadBinaryImm64; + def ATOMIC_LOAD_NIHH64 : AtomicLoadBinaryImm64; + def ATOMIC_LOAD_NILF64 : AtomicLoadBinaryImm64; + def ATOMIC_LOAD_NIHF64 : AtomicLoadBinaryImm64; +} def ATOMIC_LOADW_OR : AtomicLoadWBinaryReg; def ATOMIC_LOADW_OILH : AtomicLoadWBinaryImm; -def ATOMIC_LOAD_OR : AtomicLoadBinaryReg32; -def ATOMIC_LOAD_OILL32 : AtomicLoadBinaryImm32; -def ATOMIC_LOAD_OILH32 : AtomicLoadBinaryImm32; -def ATOMIC_LOAD_OILF32 : AtomicLoadBinaryImm32; -def ATOMIC_LOAD_OGR : AtomicLoadBinaryReg64; -def ATOMIC_LOAD_OILL : AtomicLoadBinaryImm64; -def ATOMIC_LOAD_OILH : AtomicLoadBinaryImm64; -def ATOMIC_LOAD_OIHL : AtomicLoadBinaryImm64; -def ATOMIC_LOAD_OIHH : AtomicLoadBinaryImm64; -def ATOMIC_LOAD_OILF : AtomicLoadBinaryImm64; -def ATOMIC_LOAD_OIHF : AtomicLoadBinaryImm64; +let Predicates = [FeatureNoInterlockedAccess1] in { + def ATOMIC_LOAD_OR : AtomicLoadBinaryReg32; + def ATOMIC_LOAD_OILL : AtomicLoadBinaryImm32; + def ATOMIC_LOAD_OILH : AtomicLoadBinaryImm32; + def ATOMIC_LOAD_OILF : AtomicLoadBinaryImm32; + def ATOMIC_LOAD_OGR : AtomicLoadBinaryReg64; + def ATOMIC_LOAD_OILL64 : AtomicLoadBinaryImm64; + def ATOMIC_LOAD_OILH64 : AtomicLoadBinaryImm64; + def ATOMIC_LOAD_OIHL64 : AtomicLoadBinaryImm64; + def ATOMIC_LOAD_OIHH64 : AtomicLoadBinaryImm64; + def ATOMIC_LOAD_OILF64 : AtomicLoadBinaryImm64; + def ATOMIC_LOAD_OIHF64 : AtomicLoadBinaryImm64; +} def ATOMIC_LOADW_XR : AtomicLoadWBinaryReg; def ATOMIC_LOADW_XILF : AtomicLoadWBinaryImm; -def ATOMIC_LOAD_XR : AtomicLoadBinaryReg32; -def ATOMIC_LOAD_XILF32 : AtomicLoadBinaryImm32; -def ATOMIC_LOAD_XGR : AtomicLoadBinaryReg64; -def ATOMIC_LOAD_XILF : AtomicLoadBinaryImm64; -def ATOMIC_LOAD_XIHF : AtomicLoadBinaryImm64; +let Predicates = [FeatureNoInterlockedAccess1] in { + def ATOMIC_LOAD_XR : AtomicLoadBinaryReg32; + def ATOMIC_LOAD_XILF : AtomicLoadBinaryImm32; + def ATOMIC_LOAD_XGR : AtomicLoadBinaryReg64; + def ATOMIC_LOAD_XILF64 : AtomicLoadBinaryImm64; + def ATOMIC_LOAD_XIHF64 : AtomicLoadBinaryImm64; +} def ATOMIC_LOADW_NRi : AtomicLoadWBinaryReg; def ATOMIC_LOADW_NILHi : AtomicLoadWBinaryImm; def ATOMIC_LOAD_NRi : AtomicLoadBinaryReg32; -def ATOMIC_LOAD_NILL32i : AtomicLoadBinaryImm32; -def ATOMIC_LOAD_NILH32i : AtomicLoadBinaryImm32; -def ATOMIC_LOAD_NILF32i : AtomicLoadBinaryImm32; +def ATOMIC_LOAD_NILFi : AtomicLoadBinaryImm32; def ATOMIC_LOAD_NGRi : AtomicLoadBinaryReg64; -def ATOMIC_LOAD_NILLi : AtomicLoadBinaryImm64; -def ATOMIC_LOAD_NILHi : AtomicLoadBinaryImm64; -def ATOMIC_LOAD_NIHLi : AtomicLoadBinaryImm64; -def ATOMIC_LOAD_NIHHi : AtomicLoadBinaryImm64; -def ATOMIC_LOAD_NILFi : AtomicLoadBinaryImm64; -def ATOMIC_LOAD_NIHFi : AtomicLoadBinaryImm64; def ATOMIC_LOADW_MIN : AtomicLoadWBinaryReg; @@ -1115,6 +1364,60 @@ let Defs = [CC] in { def CSG : CmpSwapRSY<"csg", 0xEB30, atomic_cmp_swap_64, GR64>; } +//===----------------------------------------------------------------------===// +// Transactional execution +//===----------------------------------------------------------------------===// + +let Predicates = [FeatureTransactionalExecution] in { + // Transaction Begin + let hasSideEffects = 1, mayStore = 1, + usesCustomInserter = 1, Defs = [CC] in { + def TBEGIN : InstSIL<0xE560, + (outs), (ins bdaddr12only:$BD1, imm32zx16:$I2), + "tbegin\t$BD1, $I2", + [(z_tbegin bdaddr12only:$BD1, imm32zx16:$I2)]>; + def TBEGIN_nofloat : Pseudo<(outs), (ins bdaddr12only:$BD1, imm32zx16:$I2), + [(z_tbegin_nofloat bdaddr12only:$BD1, + imm32zx16:$I2)]>; + def TBEGINC : InstSIL<0xE561, + (outs), (ins bdaddr12only:$BD1, imm32zx16:$I2), + "tbeginc\t$BD1, $I2", + [(int_s390_tbeginc bdaddr12only:$BD1, + imm32zx16:$I2)]>; + } + + // Transaction End + let hasSideEffects = 1, Defs = [CC], BD2 = 0 in + def TEND : InstS<0xB2F8, (outs), (ins), "tend", [(z_tend)]>; + + // Transaction Abort + let hasSideEffects = 1, isTerminator = 1, isBarrier = 1 in + def TABORT : InstS<0xB2FC, (outs), (ins bdaddr12only:$BD2), + "tabort\t$BD2", + [(int_s390_tabort bdaddr12only:$BD2)]>; + + // Nontransactional Store + let hasSideEffects = 1 in + def NTSTG : StoreRXY<"ntstg", 0xE325, int_s390_ntstg, GR64, 8>; + + // Extract Transaction Nesting Depth + let hasSideEffects = 1 in + def ETND : InherentRRE<"etnd", 0xB2EC, GR32, (int_s390_etnd)>; +} + +//===----------------------------------------------------------------------===// +// Processor assist +//===----------------------------------------------------------------------===// + +let Predicates = [FeatureProcessorAssist] in { + let hasSideEffects = 1, R4 = 0 in + def PPA : InstRRF<0xB2E8, (outs), (ins GR64:$R1, GR64:$R2, imm32zx4:$R3), + "ppa\t$R1, $R2, $R3", []>; + def : Pat<(int_s390_ppa_txassist GR32:$src), + (PPA (INSERT_SUBREG (i64 (IMPLICIT_DEF)), GR32:$src, subreg_l32), + 0, 1)>; +} + //===----------------------------------------------------------------------===// // Miscellaneous Instructions. //===----------------------------------------------------------------------===// @@ -1138,19 +1441,18 @@ let Defs = [CC] in { def FLOGR : UnaryRRE<"flog", 0xB983, null_frag, GR128, GR64>; } def : Pat<(ctlz GR64:$src), - (EXTRACT_SUBREG (FLOGR GR64:$src), subreg_high)>; + (EXTRACT_SUBREG (FLOGR GR64:$src), subreg_h64)>; + +// Population count. Counts bits set per byte. +let Predicates = [FeaturePopulationCount], Defs = [CC] in { + def POPCNT : InstRRE<0xB9E1, (outs GR64:$R1), (ins GR64:$R2), + "popcnt\t$R1, $R2", + [(set GR64:$R1, (z_popcnt GR64:$R2))]>; +} // Use subregs to populate the "don't care" bits in a 32-bit to 64-bit anyext. def : Pat<(i64 (anyext GR32:$src)), - (INSERT_SUBREG (i64 (IMPLICIT_DEF)), GR32:$src, subreg_32bit)>; - -// There are no 32-bit equivalents of LLILL and LLILH, so use a full -// 64-bit move followed by a subreg. This preserves the invariant that -// all GR32 operations only modify the low 32 bits. -def : Pat<(i32 imm32ll16:$src), - (EXTRACT_SUBREG (LLILL (LL16 imm:$src)), subreg_32bit)>; -def : Pat<(i32 imm32lh16:$src), - (EXTRACT_SUBREG (LLILH (LH16 imm:$src)), subreg_32bit)>; + (INSERT_SUBREG (i64 (IMPLICIT_DEF)), GR32:$src, subreg_l32)>; // Extend GR32s and GR64s to GR128s. let usesCustomInserter = 1 in { @@ -1159,6 +1461,30 @@ let usesCustomInserter = 1 in { def ZEXT128_64 : Pseudo<(outs GR128:$dst), (ins GR64:$src), []>; } +// Search a block of memory for a character. +let mayLoad = 1, Defs = [CC] in + defm SRST : StringRRE<"srst", 0xb25e, z_search_string>; + +// Other instructions for inline assembly +let hasSideEffects = 1, Defs = [CC], mayStore = 1 in + def STCK : InstS<0xB205, (outs), (ins bdaddr12only:$BD2), + "stck\t$BD2", + []>; +let hasSideEffects = 1, Defs = [CC], mayStore = 1 in + def STCKF : InstS<0xB27C, (outs), (ins bdaddr12only:$BD2), + "stckf\t$BD2", + []>; +let hasSideEffects = 1, Defs = [CC], mayStore = 1 in + def STCKE : InstS<0xB278, (outs), (ins bdaddr12only:$BD2), + "stcke\t$BD2", + []>; +let hasSideEffects = 1, Defs = [CC], mayStore = 1 in + def STFLE : InstS<0xB2B0, (outs), (ins bdaddr12only:$BD2), + "stfle\t$BD2", + []>; + + + //===----------------------------------------------------------------------===// // Peepholes. //===----------------------------------------------------------------------===// @@ -1167,24 +1493,40 @@ let usesCustomInserter = 1 in { defm : ZXB; def : Pat<(add GR64:$src1, imm64zx32:$src2), (ALGFI GR64:$src1, imm64zx32:$src2)>; -def : Pat<(add GR64:$src1, (zextloadi32 bdxaddr20only:$addr)), +def : Pat<(add GR64:$src1, (azextloadi32 bdxaddr20only:$addr)), (ALGF GR64:$src1, bdxaddr20only:$addr)>; // Use SL* for GR64 subtractions of unsigned 32-bit values. defm : ZXB; def : Pat<(add GR64:$src1, imm64zx32n:$src2), (SLGFI GR64:$src1, imm64zx32n:$src2)>; -def : Pat<(sub GR64:$src1, (zextloadi32 bdxaddr20only:$addr)), +def : Pat<(sub GR64:$src1, (azextloadi32 bdxaddr20only:$addr)), (SLGF GR64:$src1, bdxaddr20only:$addr)>; // Optimize sign-extended 1/0 selects to -1/0 selects. This is important // for vector legalization. -def : Pat<(sra (shl (i32 (z_select_ccmask 1, 0, uimm8zx4:$valid, uimm8zx4:$cc)), +def : Pat<(sra (shl (i32 (z_select_ccmask 1, 0, imm32zx4:$valid, imm32zx4:$cc)), (i32 31)), (i32 31)), - (Select32 (LHI -1), (LHI 0), uimm8zx4:$valid, uimm8zx4:$cc)>; -def : Pat<(sra (shl (i64 (anyext (i32 (z_select_ccmask 1, 0, uimm8zx4:$valid, - uimm8zx4:$cc)))), + (Select32 (LHI -1), (LHI 0), imm32zx4:$valid, imm32zx4:$cc)>; +def : Pat<(sra (shl (i64 (anyext (i32 (z_select_ccmask 1, 0, imm32zx4:$valid, + imm32zx4:$cc)))), (i32 63)), (i32 63)), - (Select64 (LGHI -1), (LGHI 0), uimm8zx4:$valid, uimm8zx4:$cc)>; + (Select64 (LGHI -1), (LGHI 0), imm32zx4:$valid, imm32zx4:$cc)>; + +// Peepholes for turning scalar operations into block operations. +defm : BlockLoadStore; +defm : BlockLoadStore; +defm : BlockLoadStore; +defm : BlockLoadStore; +defm : BlockLoadStore; +defm : BlockLoadStore; +defm : BlockLoadStore;