diff options
Diffstat (limited to 'llvm/lib/Target')
| -rw-r--r-- | llvm/lib/Target/ARM/ARMInstrFormats.td | 1 | ||||
| -rw-r--r-- | llvm/lib/Target/ARM/ARMInstrInfo.td | 18 | ||||
| -rw-r--r-- | llvm/lib/Target/ARM/ARMInstrThumb2.td | 3 | ||||
| -rw-r--r-- | llvm/lib/Target/ARM/ARMInstrVFP.td | 11 | ||||
| -rw-r--r-- | llvm/lib/Target/ARM/ARMScheduleM4.td | 24 |
5 files changed, 40 insertions, 17 deletions
diff --git a/llvm/lib/Target/ARM/ARMInstrFormats.td b/llvm/lib/Target/ARM/ARMInstrFormats.td index 53af05cf5f4..dd14ebda707 100644 --- a/llvm/lib/Target/ARM/ARMInstrFormats.td +++ b/llvm/lib/Target/ARM/ARMInstrFormats.td @@ -465,6 +465,7 @@ class AsmPseudoInst<string asm, dag iops, dag oops = (outs)> let isCodeGenOnly = 0; // So we get asm matcher for it. let AsmString = asm; let isPseudo = 1; + let hasNoSchedulingInfo = 1; } class ARMAsmPseudo<string asm, dag iops, dag oops = (outs)> diff --git a/llvm/lib/Target/ARM/ARMInstrInfo.td b/llvm/lib/Target/ARM/ARMInstrInfo.td index 766c87ed8c5..e25260d8b47 100644 --- a/llvm/lib/Target/ARM/ARMInstrInfo.td +++ b/llvm/lib/Target/ARM/ARMInstrInfo.td @@ -1967,7 +1967,7 @@ multiclass AI_str1nopc<bit isByte, string opc, InstrItinClass iii, /// the function. The first operand is the ID# for this instruction, the second /// is the index into the MachineConstantPool that this is, the third is the /// size in bytes of this constant pool entry. -let hasSideEffects = 0, isNotDuplicable = 1 in +let hasSideEffects = 0, isNotDuplicable = 1, hasNoSchedulingInfo = 1 in def CONSTPOOL_ENTRY : PseudoInst<(outs), (ins cpinst_operand:$instid, cpinst_operand:$cpidx, i32imm:$size), NoItinerary, []>; @@ -4895,14 +4895,13 @@ def SB : AInoP<(outs), (ins), MiscFrm, NoItinerary, "sb", "", []>, let hasSideEffects = 1; } -let usesCustomInserter = 1, Defs = [CPSR] in { - -// Pseudo instruction that combines movs + predicated rsbmi -// to implement integer ABS +let usesCustomInserter = 1, Defs = [CPSR], hasNoSchedulingInfo = 1 in { + // Pseudo instruction that combines movs + predicated rsbmi + // to implement integer ABS def ABS : ARMPseudoInst<(outs GPR:$dst), (ins GPR:$src), 8, NoItinerary, []>; } -let usesCustomInserter = 1, Defs = [CPSR] in { +let usesCustomInserter = 1, Defs = [CPSR], hasNoSchedulingInfo = 1 in { def COPY_STRUCT_BYVAL_I32 : PseudoInst< (outs), (ins GPR:$dst, GPR:$src, i32imm:$size, i32imm:$alignment), NoItinerary, @@ -5604,12 +5603,12 @@ def MSRbanked : ABI<0b0001, (outs), (ins banked_reg:$banked, GPRnopc:$Rn), def win__chkstk : SDNode<"ARMISD::WIN__CHKSTK", SDTNone, [SDNPHasChain, SDNPSideEffect]>; -let usesCustomInserter = 1, Uses = [R4], Defs = [R4, SP] in +let usesCustomInserter = 1, Uses = [R4], Defs = [R4, SP], hasNoSchedulingInfo = 1 in def WIN__CHKSTK : PseudoInst<(outs), (ins), NoItinerary, [(win__chkstk)]>; def win__dbzchk : SDNode<"ARMISD::WIN__DBZCHK", SDT_WIN__DBZCHK, [SDNPHasChain, SDNPSideEffect, SDNPOutGlue]>; -let usesCustomInserter = 1, Defs = [CPSR] in +let usesCustomInserter = 1, Defs = [CPSR], hasNoSchedulingInfo = 1 in def WIN__DBZCHK : PseudoInst<(outs), (ins tGPR:$divisor), NoItinerary, [(win__dbzchk tGPR:$divisor)]>; @@ -6156,7 +6155,7 @@ def : InstAlias<"umull${s}${p} $RdLo, $RdHi, $Rn, $Rm", def ITasm : ARMAsmPseudo<"it$mask $cc", (ins it_pred:$cc, it_mask:$mask)>, ComplexDeprecationPredicate<"IT">; -let mayLoad = 1, mayStore =1, hasSideEffects = 1 in +let mayLoad = 1, mayStore =1, hasSideEffects = 1, hasNoSchedulingInfo = 1 in def SPACE : PseudoInst<(outs GPR:$Rd), (ins i32imm:$size, GPR:$Rn), NoItinerary, [(set GPR:$Rd, (int_arm_space timm:$size, GPR:$Rn))]>; @@ -6199,4 +6198,5 @@ def CompilerBarrier : PseudoInst<(outs), (ins i32imm:$ordering), NoItinerary, let hasSideEffects = 1; let Size = 0; let AsmString = "@ COMPILER BARRIER"; + let hasNoSchedulingInfo = 1; } diff --git a/llvm/lib/Target/ARM/ARMInstrThumb2.td b/llvm/lib/Target/ARM/ARMInstrThumb2.td index b165b85a433..ef5d0908b77 100644 --- a/llvm/lib/Target/ARM/ARMInstrThumb2.td +++ b/llvm/lib/Target/ARM/ARMInstrThumb2.td @@ -4086,7 +4086,7 @@ def t2LDRpci_pic : PseudoInst<(outs rGPR:$dst), (ins i32imm:$addr, pclabel:$cp), // Pseudo isntruction that combines movs + predicated rsbmi // to implement integer ABS -let usesCustomInserter = 1, Defs = [CPSR] in { +let usesCustomInserter = 1, Defs = [CPSR], hasNoSchedulingInfo = 1 in { def t2ABS : PseudoInst<(outs rGPR:$dst), (ins rGPR:$src), NoItinerary, []>, Requires<[IsThumb2]>; } @@ -5088,6 +5088,7 @@ def t2BF_LabelPseudo : t2PseudoInst<(outs ), (ins pclabel:$cp), 0, NoItinerary, []> { let isTerminator = 1; let Predicates = [IsThumb2, HasV8_1MMainline, HasLOB]; + let hasNoSchedulingInfo = 1; } def t2BFi : t2BF<(ins bflabel_u4:$b_label, bflabel_s16:$label, pred:$p), diff --git a/llvm/lib/Target/ARM/ARMInstrVFP.td b/llvm/lib/Target/ARM/ARMInstrVFP.td index 4a95630ab7e..d3380ab7cef 100644 --- a/llvm/lib/Target/ARM/ARMInstrVFP.td +++ b/llvm/lib/Target/ARM/ARMInstrVFP.td @@ -324,7 +324,7 @@ defm : VFPDTAnyInstAlias<"vpop${p}", "$r", // However, there is no UAL syntax for them, so we keep them around for // (dis)assembly only. multiclass vfp_ldstx_mult<string asm, bit L_bit> { - let Predicates = [HasFPRegs] in { + let Predicates = [HasFPRegs], hasNoSchedulingInfo = 1 in { // Unknown precision def XIA : AXXI4<(outs), (ins GPR:$Rn, pred:$p, dpr_reglist:$regs, variable_ops), @@ -1732,7 +1732,8 @@ def VTOSHS : AVConv1XInsS_Encode<0b11101, 0b11, 0b1110, 0b1010, 0, def VTOUHS : AVConv1XInsS_Encode<0b11101, 0b11, 0b1111, 0b1010, 0, (outs SPR:$dst), (ins SPR:$a, fbits16:$fbits), - IIC_fpCVTSI, "vcvt", ".u16.f32\t$dst, $a, $fbits", []> { + IIC_fpCVTSI, "vcvt", ".u16.f32\t$dst, $a, $fbits", []>, + Sched<[WriteFPCVT]> { // Some single precision VFP instructions may be executed on both NEON and // VFP pipelines on A8. let D = VFPNeonA8Domain; @@ -1740,7 +1741,8 @@ def VTOUHS : AVConv1XInsS_Encode<0b11101, 0b11, 0b1111, 0b1010, 0, def VTOSLS : AVConv1XInsS_Encode<0b11101, 0b11, 0b1110, 0b1010, 1, (outs SPR:$dst), (ins SPR:$a, fbits32:$fbits), - IIC_fpCVTSI, "vcvt", ".s32.f32\t$dst, $a, $fbits", []> { + IIC_fpCVTSI, "vcvt", ".s32.f32\t$dst, $a, $fbits", []>, + Sched<[WriteFPCVT]> { // Some single precision VFP instructions may be executed on both NEON and // VFP pipelines on A8. let D = VFPNeonA8Domain; @@ -1748,7 +1750,8 @@ def VTOSLS : AVConv1XInsS_Encode<0b11101, 0b11, 0b1110, 0b1010, 1, def VTOULS : AVConv1XInsS_Encode<0b11101, 0b11, 0b1111, 0b1010, 1, (outs SPR:$dst), (ins SPR:$a, fbits32:$fbits), - IIC_fpCVTSI, "vcvt", ".u32.f32\t$dst, $a, $fbits", []> { + IIC_fpCVTSI, "vcvt", ".u32.f32\t$dst, $a, $fbits", []>, + Sched<[WriteFPCVT]> { // Some single precision VFP instructions may be executed on both NEON and // VFP pipelines on A8. let D = VFPNeonA8Domain; diff --git a/llvm/lib/Target/ARM/ARMScheduleM4.td b/llvm/lib/Target/ARM/ARMScheduleM4.td index 38c8ea2b4f3..bfa5fc0d713 100644 --- a/llvm/lib/Target/ARM/ARMScheduleM4.td +++ b/llvm/lib/Target/ARM/ARMScheduleM4.td @@ -18,6 +18,9 @@ def CortexM4Model : SchedMachineModel { let PostRAScheduler = 1; let CompleteModel = 0; + let UnsupportedFeatures = [IsARM, HasNEON, HasDotProd, HasZCZ, HasMVEInt, + IsNotMClass, HasDPVFP, HasFPARMv8, HasFullFP16, Has8MSecExt, HasV8, + HasV8_3a, HasTrustZone, HasDFB, IsWindows]; } @@ -50,6 +53,7 @@ def : M4UnitL2<WriteMAC16>; def : M4UnitL2<WriteDIV>; def : M4UnitL2I<(instregex "(t|t2)LDM")>; +def : M4UnitL2I<(instregex "(t|t2)LDR")>; // Stores we use a latency of 1 as they have no outputs @@ -78,9 +82,20 @@ def : M4UnitL1<WriteNoop>; def : M4UnitL1<WritePreLd>; def : M4UnitL1I<(instregex "(t|t2)MOV")>; def : M4UnitL1I<(instrs COPY)>; -def : M4UnitL1I<(instregex "t2IT")>; -def : M4UnitL1I<(instregex "t2SEL", "t2USAD8", - "t2(S|Q|SH|U|UQ|UH)(ADD16|ASX|SAX|SUB16|ADD8|SUB8)", "t2USADA8", "(t|t2)REV")>; +def : M4UnitL1I<(instregex "t2IT", "t2MSR", "t2MRS")>; +def : M4UnitL1I<(instregex "t2CLREX")>; +def : M4UnitL1I<(instregex "t2SEL", "t2USAD8", "t2SML[AS]", + "t2(S|Q|SH|U|UQ|UH|QD)(ADD|ASX|SAX|SUB)", "t2USADA8", "(t|t2)REV")>; + +// These instructions are not of much interest to scheduling as they will not +// be generated or it is not very useful to schedule them. They are here to make +// the model more complete. +def : M4UnitL1I<(instregex "t2CDP", "t2LDC", "t2MCR", "t2MRC", "t2MRRC", "t2STC")>; +def : M4UnitL1I<(instregex "tCPS", "t2ISB", "t2DSB", "t2DMB", "t2?HINT$")>; +def : M4UnitL1I<(instregex "t2?UDF$", "tBKPT", "t2DBG")>; +def : M4UnitL1I<(instregex "t?2?Int_eh_sjlj_", "tADDframe", "t?ADJCALL")>; +def : M4UnitL1I<(instregex "CMP_SWAP", "JUMPTABLE", "MEMCPY")>; +def : M4UnitL1I<(instregex "VSETLNi32", "VGETLNi32")>; def : ReadAdvance<ReadALU, 0>; def : ReadAdvance<ReadALUsr, 0>; @@ -112,6 +127,9 @@ def : M4UnitL1<WriteVST1>; def : M4UnitL1<WriteVST2>; def : M4UnitL1<WriteVST3>; def : M4UnitL1<WriteVST4>; +def : M4UnitL1I<(instregex "VMOVS", "FCONSTS", "VCMP", "VNEG", "VABS")>; +def : M4UnitL2I<(instregex "VMOVD")>; +def : M4UnitL1I<(instregex "VMRS", "VMSR", "FMSTAT")>; def : ReadAdvance<ReadFPMUL, 0>; def : ReadAdvance<ReadFPMAC, 0>; |
