Abbreviations used in this document are listed here.
| Abbreviation | Description |
|---|---|
| APERF | Actual Performance Frequency Clock Counter |
| ASLR | Address Space Layout Randomization |
| CCD | Core Complex Die that can contain one or more CCX(s) and GMI2 Fabric port(s) connecting to IOD |
| CLI | Command Line Interface |
| CPI | Cycles Per Instruction |
| CSV | Comma Separated Values format |
| DC | Data Cache |
| DIMM | Dual In-line Memory Module |
| DRAM | Dynamic Random Access Memory |
| DTLB | Data Translation Lookaside Buffer |
| EBP | Event Based Profiling, uses Core PMC events |
| GUI | Graphical User Interface |
| IBS | Instruction Based Sampling |
| IC | Instruction Cache |
| IOD | IO Die |
| IPC | Instructions Per Cycle |
| ITLB | Instruction Translation Lookaside Buffer |
| MPERF | Maximum Performance Frequency Clock Counter |
| MSR | Model Specific Register |
| NB | Northbridge |
| OS | Operating System |
| P0Freq | P0State Frequency |
| PMC | Performance Monitoring Counter |
| PTI | Per Thousand Instructions |
| RAPL | Running Average Power Limit |
| SMU | System Management Unit |
| TBP | Time Based Profiling |
| TSC | Time Stamp Counter |
| UMC |
Unified Memory Controllers Upto 8 UMCs, each supporting one DRAM channel per socket; each channel can have up to 2 DIMMs |
Terminology
The terms used in this document are listed here.
| Term | Description |
|---|---|
| AMD uProf | The product name uProf. |
| AMDuProfGUI | The name of the graphical user interface tool. |
| AMDuProfCLI | The name of the command line interface tool. |
| AMDuProfPcm | The name of the command line interface tool for System Analysis. |
| AMDuProfSys | The name of the python based command line interface tool for System Analysis. |
| Client | Instance of AMD uProf or AMDuProfCLI running on a host system. |
| Core | The logical core number, a core can contain one or two CPU(s) depending on the SMT configuration. |
| Core Complex (CCX) | Consists of one or many cores and a cache system. |
| CPU | Logical CPU numbers as considered by the operating system. |
| Host system | System in which the AMD uProf client process runs. |
| L1D, L1I Cache | CPU exclusive data and instruction cache. |
| L2 Cache | Shared by all the CPUs within the core. |
| L3 Cache | Shared by all the CPUs within CCX. |
| Node | Logical NUMA node. |
| Performance Profiling (or) CPU Profiling | Identify and analyze the performance bottlenecks. Performance Profiling and CPU Profiling denotes the same. |
| Socket | The logical socket number, a socket can contain multiple nodes. |
| System Analysis | Refers to AMDuProfPcm or AMDuProfSys tools. |
| Target system | System in which the profile data is collected. |