This patch fixes a bug in floating point operands parsing, when instruction alias...

[oota-llvm.git] / lib / Target / X86 / X86Schedule.td

diff --git a/lib/Target/X86/X86Schedule.td b/lib/Target/X86/X86Schedule.td
index 782e771fd0403a529468dde2f0567cc29f6ed1eb..0556437b839b966a71e6713d15973aa371aa36d9 100644 (file)
--- a/lib/Target/X86/X86Schedule.td
+++ b/lib/Target/X86/X86Schedule.td
@@ -42,6 +42,7 @@ multiclass X86SchedWritePair {
 // Arithmetic.
 defm WriteALU  : X86SchedWritePair; // Simple integer ALU op.
 defm WriteIMul : X86SchedWritePair; // Integer multiplication.
+def  WriteIMulH : SchedWrite;       // Integer multiplication, high part.
 defm WriteIDiv : X86SchedWritePair; // Integer division.
 def  WriteLEA  : SchedWrite;        // LEA instructions can't fold loads.
 
@@ -67,6 +68,10 @@ defm WriteFMul  : X86SchedWritePair; // Floating point multiplication.
 defm WriteFDiv  : X86SchedWritePair; // Floating point division.
 defm WriteFSqrt : X86SchedWritePair; // Floating point square root.
 defm WriteFRcp  : X86SchedWritePair; // Floating point reciprocal.
+defm WriteFMA   : X86SchedWritePair; // Fused Multiply Add.
+
+// FMA Scheduling helper class.
+class FMASC { X86FoldableSchedWrite Sched = WriteFAdd; }
 
 // Vector integer operations.
 defm WriteVecALU   : X86SchedWritePair; // Vector integer ALU op, no logicals.
@@ -136,9 +141,12 @@ def IIC_IDIV64      : InstrItinClass;
 // neg/not/inc/dec
 def IIC_UNARY_REG   : InstrItinClass;
 def IIC_UNARY_MEM   : InstrItinClass;
-// add/sub/and/or/xor/adc/sbc/cmp/test
+// add/sub/and/or/xor/sbc/cmp/test
 def IIC_BIN_MEM     : InstrItinClass;
 def IIC_BIN_NONMEM  : InstrItinClass;
+// adc/sbc
+def IIC_BIN_CARRY_MEM     : InstrItinClass;
+def IIC_BIN_CARRY_NONMEM  : InstrItinClass;
 // shift/rotate
 def IIC_SR          : InstrItinClass;
 // shift double
@@ -245,11 +253,11 @@ def IIC_SSE_INTSH_P_RR : InstrItinClass;
 def IIC_SSE_INTSH_P_RM : InstrItinClass;
 def IIC_SSE_INTSH_P_RI : InstrItinClass;
 
-def IIC_SSE_CMPP_RR : InstrItinClass;
-def IIC_SSE_CMPP_RM : InstrItinClass;
+def IIC_SSE_INTSHDQ_P_RI : InstrItinClass;
 
 def IIC_SSE_SHUFP : InstrItinClass;
-def IIC_SSE_PSHUF : InstrItinClass;
+def IIC_SSE_PSHUF_RI : InstrItinClass;
+def IIC_SSE_PSHUF_MI : InstrItinClass;
 
 def IIC_SSE_UNPCK : InstrItinClass;
 
@@ -262,10 +270,14 @@ def IIC_SSE_PINSRW : InstrItinClass;
 def IIC_SSE_PABS_RR : InstrItinClass;
 def IIC_SSE_PABS_RM : InstrItinClass;
 
-def IIC_SSE_SQRTP_RR : InstrItinClass;
-def IIC_SSE_SQRTP_RM : InstrItinClass;
-def IIC_SSE_SQRTS_RR : InstrItinClass;
-def IIC_SSE_SQRTS_RM : InstrItinClass;
+def IIC_SSE_SQRTPS_RR : InstrItinClass;
+def IIC_SSE_SQRTPS_RM : InstrItinClass;
+def IIC_SSE_SQRTSS_RR : InstrItinClass;
+def IIC_SSE_SQRTSS_RM : InstrItinClass;
+def IIC_SSE_SQRTPD_RR : InstrItinClass;
+def IIC_SSE_SQRTPD_RM : InstrItinClass;
+def IIC_SSE_SQRTSD_RR : InstrItinClass;
+def IIC_SSE_SQRTSD_RM : InstrItinClass;
 
 def IIC_SSE_RCPP_RR : InstrItinClass;
 def IIC_SSE_RCPP_RM : InstrItinClass;
@@ -307,7 +319,8 @@ def IIC_SSE_PSIGN_RM : InstrItinClass;
 
 def IIC_SSE_PMADD : InstrItinClass;
 def IIC_SSE_PMULHRSW : InstrItinClass;
-def IIC_SSE_PALIGNR : InstrItinClass;
+def IIC_SSE_PALIGNRR : InstrItinClass;
+def IIC_SSE_PALIGNRM : InstrItinClass;
 def IIC_SSE_MWAIT : InstrItinClass;
 def IIC_SSE_MONITOR : InstrItinClass;
 
@@ -483,8 +496,8 @@ def IIC_PUSH_REG : InstrItinClass;
 def IIC_PUSH_F : InstrItinClass;
 def IIC_PUSH_A : InstrItinClass;
 def IIC_BSWAP : InstrItinClass;
-def IIC_BSF : InstrItinClass;
-def IIC_BSR : InstrItinClass;
+def IIC_BIT_SCAN_MEM : InstrItinClass;
+def IIC_BIT_SCAN_REG : InstrItinClass;
 def IIC_MOVS : InstrItinClass;
 def IIC_STOS : InstrItinClass;
 def IIC_SCAS : InstrItinClass;
@@ -531,6 +544,33 @@ def IIC_BOUND : InstrItinClass;
 def IIC_ARPL_REG : InstrItinClass;
 def IIC_ARPL_MEM : InstrItinClass;
 def IIC_MOVBE : InstrItinClass;
+def IIC_AES   : InstrItinClass;
+def IIC_BLEND_MEM : InstrItinClass;
+def IIC_BLEND_NOMEM : InstrItinClass;
+def IIC_CBW   : InstrItinClass;
+def IIC_CRC32_REG : InstrItinClass;
+def IIC_CRC32_MEM : InstrItinClass;
+def IIC_SSE_DPPD_RR : InstrItinClass;
+def IIC_SSE_DPPD_RM : InstrItinClass;
+def IIC_SSE_DPPS_RR : InstrItinClass;
+def IIC_SSE_DPPS_RM : InstrItinClass;
+def IIC_MMX_EMMS : InstrItinClass;
+def IIC_SSE_EXTRACTPS_RR : InstrItinClass;
+def IIC_SSE_EXTRACTPS_RM : InstrItinClass;
+def IIC_SSE_INSERTPS_RR : InstrItinClass;
+def IIC_SSE_INSERTPS_RM : InstrItinClass;
+def IIC_SSE_MPSADBW_RR : InstrItinClass;
+def IIC_SSE_MPSADBW_RM : InstrItinClass;
+def IIC_SSE_PMULLD_RR : InstrItinClass;
+def IIC_SSE_PMULLD_RM : InstrItinClass;
+def IIC_SSE_ROUNDPS_REG : InstrItinClass;
+def IIC_SSE_ROUNDPS_MEM : InstrItinClass;
+def IIC_SSE_ROUNDPD_REG : InstrItinClass;
+def IIC_SSE_ROUNDPD_MEM : InstrItinClass;
+def IIC_SSE_POPCNT_RR : InstrItinClass;
+def IIC_SSE_POPCNT_RM : InstrItinClass;
+def IIC_SSE_PCLMULQDQ_RR : InstrItinClass;
+def IIC_SSE_PCLMULQDQ_RM : InstrItinClass;
 
 def IIC_NOP : InstrItinClass;
 
@@ -542,8 +582,9 @@ def IIC_NOP : InstrItinClass;
 // Resources beyond the decoder operate on micro-ops and are bufferred
 // so adjacent micro-ops don't directly compete.
 //
-// MinLatency=0 indicates that RAW dependencies can be decoded in the
-// same cycle.
+// MicroOpBufferSize > 1 indicates that RAW dependencies can be
+// decoded in the same cycle. The value 32 is a reasonably arbitrary
+// number of in-flight instructions.
 //
 // HighLatency=10 is optimistic. X86InstrInfo::isHighLatencyDef
 // indicates high latency opcodes. Alternatively, InstrItinData
@@ -551,18 +592,15 @@ def IIC_NOP : InstrItinClass;
 // latencies. Since these latencies are not used for pipeline hazards,
 // they do not need to be exact.
 //
-// ILPWindow=10 is an arbitrary threshold that approximates cycles of
-// latency hidden by instruction buffers. The actual value is not very
-// important but should be zero for inorder and nonzero for OOO processors.
-//
-// The GenericModel contains no instruciton itineraries.
+// The GenericModel contains no instruction itineraries.
 def GenericModel : SchedMachineModel {
   let IssueWidth = 4;
-  let MinLatency = 0;
+  let MicroOpBufferSize = 32;
   let LoadLatency = 4;
   let HighLatency = 10;
-  let ILPWindow = 10;
 }
 
 include "X86ScheduleAtom.td"
 include "X86SchedSandyBridge.td"
+include "X86SchedHaswell.td"
+include "X86ScheduleSLM.td"