| /llvm-project-15.0.7/llvm/test/CodeGen/X86/ |
| H A D | scheduler-backtracking.ll | 13 ; ILP-LABEL: test1: 14 ; ILP: # %bb.0: 15 ; ILP-NEXT: pushq %r14 16 ; ILP-NEXT: pushq %rbx 57 ; ILP-NEXT: retq 250 ; ILP-LABEL: test2: 288 ; ILP-NEXT: retq 456 ; ILP-LABEL: test3: 500 ; ILP-NEXT: retq 691 ; ILP-LABEL: test4: [all …]
|
| H A D | early-ifcvt-remarks.ll | 46 ; CHECK-SAME: which cannot be hidden by available ILP.
|
| /llvm-project-15.0.7/llvm/test/tools/llvm-mca/X86/Barcelona/ |
| H A D | dependency-breaking-sbb-1.s | 5 # implicit dependency on EFLAGS which limits the ILP. So, the hardware backend
|
| H A D | partial-reg-update-3.s | 4 # The ILP is limited by the false dependency on %dx. So, the mov cannot execute
|
| /llvm-project-15.0.7/llvm/test/tools/llvm-mca/X86/BtVer2/ |
| H A D | dependency-breaking-sbb-1.s | 7 # implicit dependency on EFLAGS which limits the ILP. So, the hardware backend
|
| H A D | partial-reg-update-3.s | 6 # The ILP is limited by the false dependency on %dx. So, the mov cannot execute
|
| /llvm-project-15.0.7/llvm/test/tools/llvm-mca/X86/BdVer2/ |
| H A D | dependency-breaking-sbb-1.s | 7 # implicit dependency on EFLAGS which limits the ILP. So, the hardware backend
|
| H A D | partial-reg-update-3.s | 6 # The ILP is limited by the false dependency on %dx. So, the mov cannot execute
|
| /llvm-project-15.0.7/llvm/test/tools/llvm-mca/X86/Znver2/ |
| H A D | partial-reg-update-3.s | 4 # The ILP is limited by the false dependency on %dx. So, the mov cannot execute
|
| /llvm-project-15.0.7/llvm/test/tools/llvm-mca/X86/Znver1/ |
| H A D | partial-reg-update-3.s | 4 # The ILP is limited by the false dependency on %dx. So, the mov cannot execute
|
| /llvm-project-15.0.7/llvm/test/CodeGen/AArch64/ |
| H A D | arm64-neon-v1i1-setcc.ll | 55 ; scalar setcc. If anything, it exposes more ILP.
|
| H A D | bitfield-insert.ll | 526 ; 'and' with a 'movk', which would decrease ILP while using the same number of
|
| /llvm-project-15.0.7/llvm/test/tools/llvm-mca/X86/Znver3/ |
| H A D | partial-reg-update-3.s | 4 # The ILP is limited by the false dependency on %dx. So, the mov cannot execute
|
| /llvm-project-15.0.7/llvm/docs/HistoricalNotes/ |
| H A D | 2001-02-09-AdveComments.txt | 103 while providing easy ILP for single-threaded onces. But it is probably
|
| /llvm-project-15.0.7/llvm/test/CodeGen/AMDGPU/ |
| H A D | schedule-regpressure-limit3.ll | 2 …mdgcn -mcpu=tonga -misched=gcn-ilp -verify-machineinstrs < %s | FileCheck --check-prefix=GCN-ILP %s 4 …he scheduler when only one wave is requested. The result should be high register usage and max ILP. 8 ; GCN-ILP: NumVgprs: {{[0-9][0-9][0-9]$}} 10 ; FIXME: The machine scheduler is doing a poor job at maximizing ILP here.
|
| /llvm-project-15.0.7/llvm/lib/CodeGen/SelectionDAG/ |
| H A D | ScheduleDAGRRList.cpp | 2496 bool LStall = (!checkPref || left->SchedulingPref == Sched::ILP) && in BUCompareLatency() 2498 bool RStall = (!checkPref || right->SchedulingPref == Sched::ILP) && in BUCompareLatency() 2514 if (!checkPref || (left->SchedulingPref == Sched::ILP || in BUCompareLatency() 2515 right->SchedulingPref == Sched::ILP)) { in BUCompareLatency()
|
| H A D | SelectionDAGISel.cpp | 274 assert(TLI->getSchedulingPreference() == Sched::ILP && in createDefaultScheduler()
|
| /llvm-project-15.0.7/llvm/docs/ |
| H A D | Vectorizers.rst | 353 The Loop Vectorizer increases the instruction level parallelism (ILP) by
|
| /llvm-project-15.0.7/llvm/docs/CommandGuide/ |
| H A D | llvm-mca.rst | 595 vhaddps. Long data dependencies negatively impact the ILP (Instruction Level 605 which may limit the ILP. Last row, ``<total>``, shows a global average over all
|
| /llvm-project-15.0.7/llvm/lib/CodeGen/ |
| H A D | TargetLoweringBase.cpp | 719 SchedPreferenceInfo = Sched::ILP; in TargetLoweringBase()
|
| /llvm-project-15.0.7/llvm/include/llvm/CodeGen/ |
| H A D | TargetLowering.h | 102 ILP, // Scheduling for ILP in low register pressure mode. enumerator
|
| /llvm-project-15.0.7/clang/docs/ |
| H A D | LanguageExtensions.rst | 3896 improve instruction-level parallelism (ILP) using advanced hardware features, 3958 opportunities for ILP. Loops can be fully or partially unrolled. Full unrolling
|
| /llvm-project-15.0.7/llvm/lib/Target/ARM/ |
| H A D | ARMISelLowering.cpp | 1949 return Sched::ILP; in getSchedulingPreference() 1964 return Sched::ILP; in getSchedulingPreference()
|
| /llvm-project-15.0.7/llvm/lib/Target/PowerPC/ |
| H A D | PPCISelLowering.cpp | 16831 return Sched::ILP; in getSchedulingPreference()
|
| /llvm-project-15.0.7/llvm/lib/Target/X86/ |
| H A D | X86ISelLowering.cpp | 124 setSchedulingPreference(Sched::ILP); in X86TargetLowering() 126 setSchedulingPreference(Sched::ILP); in X86TargetLowering()
|