diff options
Diffstat (limited to 'llvm/lib')
-rw-r--r-- | llvm/lib/Target/X86/X86Instr3DNow.td | 3 | ||||
-rw-r--r-- | llvm/lib/Target/X86/X86InstrMMX.td | 3 | ||||
-rwxr-xr-x | llvm/lib/Target/X86/X86SchedBroadwell.td | 8 | ||||
-rw-r--r-- | llvm/lib/Target/X86/X86SchedHaswell.td | 8 | ||||
-rw-r--r-- | llvm/lib/Target/X86/X86SchedSandyBridge.td | 1 | ||||
-rw-r--r-- | llvm/lib/Target/X86/X86SchedSkylakeClient.td | 8 | ||||
-rwxr-xr-x | llvm/lib/Target/X86/X86SchedSkylakeServer.td | 8 | ||||
-rw-r--r-- | llvm/lib/Target/X86/X86Schedule.td | 3 | ||||
-rw-r--r-- | llvm/lib/Target/X86/X86ScheduleAtom.td | 4 | ||||
-rw-r--r-- | llvm/lib/Target/X86/X86ScheduleBtVer2.td | 1 | ||||
-rw-r--r-- | llvm/lib/Target/X86/X86ScheduleSLM.td | 1 | ||||
-rw-r--r-- | llvm/lib/Target/X86/X86ScheduleZnver1.td | 1 |
12 files changed, 15 insertions, 34 deletions
diff --git a/llvm/lib/Target/X86/X86Instr3DNow.td b/llvm/lib/Target/X86/X86Instr3DNow.td index 0c834260a5b..c64c6be3cf0 100644 --- a/llvm/lib/Target/X86/X86Instr3DNow.td +++ b/llvm/lib/Target/X86/X86Instr3DNow.td @@ -74,8 +74,7 @@ defm PFSUBR : I3DNow_binop_rm_int<0xAA, "pfsubr", WriteFAdd, 1>; defm PI2FD : I3DNow_conv_rm_int<0x0D, "pi2fd", WriteCvtI2F>; defm PMULHRW : I3DNow_binop_rm_int<0xB7, "pmulhrw", SchedWriteVecIMul.MMX, 1>; -// FIXME: Is there a better scheduler class for EMMS/FEMMS? -let SchedRW = [WriteMicrocoded] in +let SchedRW = [WriteEMMS] in def FEMMS : I3DNow<0x0E, RawFrm, (outs), (ins), "femms", [(int_x86_mmx_femms)]>, TB; diff --git a/llvm/lib/Target/X86/X86InstrMMX.td b/llvm/lib/Target/X86/X86InstrMMX.td index c410ddb4c60..d41641ad158 100644 --- a/llvm/lib/Target/X86/X86InstrMMX.td +++ b/llvm/lib/Target/X86/X86InstrMMX.td @@ -153,8 +153,7 @@ multiclass sse12_cvt_pint_3addr<bits<8> opc, RegisterClass SrcRC, // MMX EMMS Instruction //===----------------------------------------------------------------------===// -// FIXME: Is there a better scheduler class for EMMS/FEMMS? -let SchedRW = [WriteMicrocoded] in +let SchedRW = [WriteEMMS] in def MMX_EMMS : MMXI<0x77, RawFrm, (outs), (ins), "emms", [(int_x86_mmx_emms)]>; //===----------------------------------------------------------------------===// diff --git a/llvm/lib/Target/X86/X86SchedBroadwell.td b/llvm/lib/Target/X86/X86SchedBroadwell.td index 21d0c8a629d..edd81bed65f 100755 --- a/llvm/lib/Target/X86/X86SchedBroadwell.td +++ b/llvm/lib/Target/X86/X86SchedBroadwell.td @@ -205,6 +205,7 @@ def : WriteRes<WriteCvtF2FSt, [BWPort1,BWPort4,BWPort237]> { def : WriteRes<WriteVecLoad, [BWPort23]> { let Latency = 5; } def : WriteRes<WriteVecStore, [BWPort237, BWPort4]>; def : WriteRes<WriteVecMove, [BWPort015]>; +defm : X86WriteRes<WriteEMMS, [BWPort01,BWPort15,BWPort015,BWPort0156], 31, [8,1,21,1], 31>; defm : BWWriteResPair<WriteVecALU, [BWPort15], 1, [1], 1, 5>; // Vector integer ALU op, no logicals. defm : BWWriteResPair<WriteVecALUY, [BWPort15], 1, [1], 1, 6>; // Vector integer ALU op, no logicals (YMM/ZMM). @@ -1779,13 +1780,6 @@ def BWWriteResGroup186 : SchedWriteRes<[BWPort4,BWPort6,BWPort23,BWPort237,BWPor } def: InstRW<[BWWriteResGroup186], (instregex "^XSAVE$", "XSAVEC", "XSAVES", "XSAVEOPT")>; -def BWWriteResGroup187 : SchedWriteRes<[BWPort01,BWPort15,BWPort015,BWPort0156]> { - let Latency = 31; - let NumMicroOps = 31; - let ResourceCycles = [8,1,21,1]; -} -def: InstRW<[BWWriteResGroup187], (instregex "MMX_EMMS")>; - def BWWriteResGroup189 : SchedWriteRes<[BWPort0,BWPort015,BWFPDivider]> { let Latency = 29; let NumMicroOps = 3; diff --git a/llvm/lib/Target/X86/X86SchedHaswell.td b/llvm/lib/Target/X86/X86SchedHaswell.td index cacf24f1e34..79a9e7a847e 100644 --- a/llvm/lib/Target/X86/X86SchedHaswell.td +++ b/llvm/lib/Target/X86/X86SchedHaswell.td @@ -147,6 +147,7 @@ defm : HWWriteResPair<WriteIDiv, [HWPort0, HWDivider], 25, [1,10], 1, 4>; def : WriteRes<WriteFStore, [HWPort237, HWPort4]>; def : WriteRes<WriteFLoad, [HWPort23]> { let Latency = 5; } def : WriteRes<WriteFMove, [HWPort5]>; +defm : X86WriteRes<WriteEMMS, [HWPort01,HWPort15,HWPort015,HWPort0156], 31, [8,1,21,1], 31>; defm : HWWriteResPair<WriteFAdd, [HWPort1], 3, [1], 1, 6>; defm : HWWriteResPair<WriteFAddY, [HWPort1], 3, [1], 1, 7>; @@ -2105,13 +2106,6 @@ def HWWriteResGroup171 : SchedWriteRes<[HWPort5,HWPort6,HWPort23,HWPort237,HWPor def: InstRW<[HWWriteResGroup171], (instregex "OUT(8|16|32)ir", "OUT(8|16|32)rr")>; -def HWWriteResGroup172 : SchedWriteRes<[HWPort01,HWPort15,HWPort015,HWPort0156]> { - let Latency = 31; - let NumMicroOps = 31; - let ResourceCycles = [8,1,21,1]; -} -def: InstRW<[HWWriteResGroup172], (instregex "MMX_EMMS")>; - def HWWriteResGroup173 : SchedWriteRes<[HWPort0,HWPort15,HWFPDivider]> { let Latency = 35; let NumMicroOps = 3; diff --git a/llvm/lib/Target/X86/X86SchedSandyBridge.td b/llvm/lib/Target/X86/X86SchedSandyBridge.td index 9c424c06536..a39e5b2bf28 100644 --- a/llvm/lib/Target/X86/X86SchedSandyBridge.td +++ b/llvm/lib/Target/X86/X86SchedSandyBridge.td @@ -137,6 +137,7 @@ defm : SBWriteResPair<WriteBZHI, [SBPort1], 1>; def : WriteRes<WriteFStore, [SBPort23, SBPort4]>; def : WriteRes<WriteFLoad, [SBPort23]> { let Latency = 6; } def : WriteRes<WriteFMove, [SBPort5]>; +defm : X86WriteRes<WriteEMMS, [SBPort015], 31, [31], 31>; defm : SBWriteResPair<WriteFAdd, [SBPort1], 3, [1], 1, 6>; defm : SBWriteResPair<WriteFAddY, [SBPort1], 3, [1], 1, 7>; diff --git a/llvm/lib/Target/X86/X86SchedSkylakeClient.td b/llvm/lib/Target/X86/X86SchedSkylakeClient.td index 1e34b840915..8cb6b14239d 100644 --- a/llvm/lib/Target/X86/X86SchedSkylakeClient.td +++ b/llvm/lib/Target/X86/X86SchedSkylakeClient.td @@ -150,6 +150,7 @@ defm : SKLWriteResPair<WriteJump, [SKLPort06], 1>; def : WriteRes<WriteFLoad, [SKLPort23]> { let Latency = 6; } def : WriteRes<WriteFStore, [SKLPort237, SKLPort4]>; def : WriteRes<WriteFMove, [SKLPort015]>; +defm : X86WriteRes<WriteEMMS, [SKLPort05,SKLPort0156], 10, [9,1], 10>; defm : SKLWriteResPair<WriteFAdd, [SKLPort01], 4, [1], 1, 6>; // Floating point add/sub. defm : SKLWriteResPair<WriteFAddY, [SKLPort01], 4, [1], 1, 7>; // Floating point add/sub (YMM/ZMM). @@ -1600,13 +1601,6 @@ def SKLWriteResGroup143 : SchedWriteRes<[SKLPort4,SKLPort6,SKLPort23,SKLPort237, } def: InstRW<[SKLWriteResGroup143], (instregex "XCHG(8|16|32|64)rm")>; -def SKLWriteResGroup144 : SchedWriteRes<[SKLPort05,SKLPort0156]> { - let Latency = 10; - let NumMicroOps = 10; - let ResourceCycles = [9,1]; -} -def: InstRW<[SKLWriteResGroup144], (instregex "MMX_EMMS")>; - def SKLWriteResGroup145 : SchedWriteRes<[SKLPort0,SKLFPDivider]> { let Latency = 11; let NumMicroOps = 1; diff --git a/llvm/lib/Target/X86/X86SchedSkylakeServer.td b/llvm/lib/Target/X86/X86SchedSkylakeServer.td index c22c864425f..c945fae34d5 100755 --- a/llvm/lib/Target/X86/X86SchedSkylakeServer.td +++ b/llvm/lib/Target/X86/X86SchedSkylakeServer.td @@ -150,6 +150,7 @@ defm : SKXWriteResPair<WriteJump, [SKXPort06], 1>; def : WriteRes<WriteFLoad, [SKXPort23]> { let Latency = 5; } def : WriteRes<WriteFStore, [SKXPort237, SKXPort4]>; def : WriteRes<WriteFMove, [SKXPort015]>; +defm : X86WriteRes<WriteEMMS, [SKXPort05,SKXPort0156], 10, [9,1], 10>; defm : SKXWriteResPair<WriteFAdd, [SKXPort015], 4, [1], 1, 6>; // Floating point add/sub. defm : SKXWriteResPair<WriteFAddY,[SKXPort015], 4, [1], 1, 7>; // Floating point add/sub (YMM/ZMM). @@ -2693,13 +2694,6 @@ def SKXWriteResGroup157 : SchedWriteRes<[SKXPort4,SKXPort6,SKXPort23,SKXPort237, } def: InstRW<[SKXWriteResGroup157], (instregex "XCHG(8|16|32|64)rm")>; -def SKXWriteResGroup158 : SchedWriteRes<[SKXPort05,SKXPort0156]> { - let Latency = 10; - let NumMicroOps = 10; - let ResourceCycles = [9,1]; -} -def: InstRW<[SKXWriteResGroup158], (instregex "MMX_EMMS")>; - def SKXWriteResGroup159 : SchedWriteRes<[SKXPort0,SKXFPDivider]> { let Latency = 11; let NumMicroOps = 1; diff --git a/llvm/lib/Target/X86/X86Schedule.td b/llvm/lib/Target/X86/X86Schedule.td index e44aeaf0e17..768e5e15d6a 100644 --- a/llvm/lib/Target/X86/X86Schedule.td +++ b/llvm/lib/Target/X86/X86Schedule.td @@ -216,6 +216,9 @@ defm WriteAESKeyGen : X86SchedWritePair; // Key Generation. // Carry-less multiplication instructions. defm WriteCLMul : X86SchedWritePair; +// EMMS/FEMMS +def WriteEMMS : SchedWrite; + // Load/store MXCSR def WriteLDMXCSR : SchedWrite; def WriteSTMXCSR : SchedWrite; diff --git a/llvm/lib/Target/X86/X86ScheduleAtom.td b/llvm/lib/Target/X86/X86ScheduleAtom.td index 4c64657c70a..1b333718042 100644 --- a/llvm/lib/Target/X86/X86ScheduleAtom.td +++ b/llvm/lib/Target/X86/X86ScheduleAtom.td @@ -201,6 +201,7 @@ def : WriteRes<WriteNop, [AtomPort01]>; def : WriteRes<WriteFLoad, [AtomPort0]>; def : WriteRes<WriteFStore, [AtomPort0]>; def : WriteRes<WriteFMove, [AtomPort01]>; +defm : X86WriteRes<WriteEMMS,[AtomPort01], 5, [5], 1>; defm : AtomWriteResPair<WriteFAdd, [AtomPort0], [AtomPort0], 5, 5, [5], [5]>; defm : AtomWriteResPair<WriteFAddY, [AtomPort0], [AtomPort0], 5, 5, [5], [5]>; @@ -490,8 +491,7 @@ def AtomWrite01_5 : SchedWriteRes<[AtomPort01]> { let Latency = 5; let ResourceCycles = [5]; } -def : InstRW<[AtomWrite01_5], (instrs FLDCW16m, ST_FP80m, - MMX_EMMS)>; +def : InstRW<[AtomWrite01_5], (instrs FLDCW16m, ST_FP80m)>; def : InstRW<[AtomWrite01_5], (instregex "MMX_PH(ADD|SUB)S?Wrr")>; def AtomWrite01_6 : SchedWriteRes<[AtomPort01]> { diff --git a/llvm/lib/Target/X86/X86ScheduleBtVer2.td b/llvm/lib/Target/X86/X86ScheduleBtVer2.td index d9dccf40829..8521ed3881d 100644 --- a/llvm/lib/Target/X86/X86ScheduleBtVer2.td +++ b/llvm/lib/Target/X86/X86ScheduleBtVer2.td @@ -314,6 +314,7 @@ def : WriteRes<WriteNop, [JALU01]> { let Latency = 1; } def : WriteRes<WriteFLoad, [JLAGU, JFPU01, JFPX]> { let Latency = 5; } def : WriteRes<WriteFStore, [JSAGU, JFPU1, JSTC]>; def : WriteRes<WriteFMove, [JFPU01, JFPX]>; +def : WriteRes<WriteEMMS, [JFPU01, JFPX]> { let Latency = 2; } defm : JWriteResFpuPair<WriteFAdd, [JFPU0, JFPA], 3>; defm : JWriteResYMMPair<WriteFAddY, [JFPU0, JFPA], 3, [2,2], 2>; diff --git a/llvm/lib/Target/X86/X86ScheduleSLM.td b/llvm/lib/Target/X86/X86ScheduleSLM.td index 08b3f36728e..dec522ea97d 100644 --- a/llvm/lib/Target/X86/X86ScheduleSLM.td +++ b/llvm/lib/Target/X86/X86ScheduleSLM.td @@ -128,6 +128,7 @@ defm : SLMWriteResPair<WriteIDiv, [SLM_IEC_RSV01, SLMDivider], 25, [1,25], 1, 4> def : WriteRes<WriteFStore, [SLM_FPC_RSV01, SLM_MEC_RSV]>; def : WriteRes<WriteFLoad, [SLM_MEC_RSV]> { let Latency = 3; } def : WriteRes<WriteFMove, [SLM_FPC_RSV01]>; +defm : X86WriteRes<WriteEMMS, [SLM_FPC_RSV01], 10, [10], 9>; defm : SLMWriteResPair<WriteFAdd, [SLM_FPC_RSV1], 3>; defm : SLMWriteResPair<WriteFAddY, [SLM_FPC_RSV1], 3>; diff --git a/llvm/lib/Target/X86/X86ScheduleZnver1.td b/llvm/lib/Target/X86/X86ScheduleZnver1.td index 3e20cb2c75d..cf95ac1fa0b 100644 --- a/llvm/lib/Target/X86/X86ScheduleZnver1.td +++ b/llvm/lib/Target/X86/X86ScheduleZnver1.td @@ -234,6 +234,7 @@ def : WriteRes<WriteCvtF2FSt, [ZnFPU3, ZnAGU]>; def : WriteRes<WriteVecStore, [ZnAGU]>; def : WriteRes<WriteVecMove, [ZnFPU]>; def : WriteRes<WriteVecLoad, [ZnAGU]> { let Latency = 8; } +def : WriteRes<WriteEMMS, [ZnFPU]> { let Latency = 2; } defm : ZnWriteResFpuPair<WriteVecShift, [ZnFPU], 1>; defm : ZnWriteResFpuPair<WriteVecShiftX, [ZnFPU2], 1>; |