SPRUI30H November 2015 – May 2024 DRA745 , DRA746 , DRA750 , DRA756
The L3_MAIN interconnect supports a built-in performance monitoring feature by implementing a statistics collector (NOC_SC) component, which computes traffic statistics within a user-defined window and periodically reports to the user through the CT_STM interface. Ten NOC_SC instances are instantiated in the device:
Statistics collectors (SC_SDRAM and SC_LAT) can report:
The performance metrics are interleaved with software instrumentation data at the L3_MAIN interconnect level.
The performance monitoring probes implement three main functions:
The probes can be configured to detect the NTTP and OCP link events summarized in Table 33-17.
Link Event | NTTP | OCP | Definition |
---|---|---|---|
NONE | ✓ | ✓ | No event selected |
ANY | ✓ | ✓ | Any clock cycles |
TRANSFER | ✓ | ✓ | Word has been accepted by the receiver. |
WAIT | ✓ | – | Transfer has been initiated but the transmitter currently has no data to send. |
BUSY | ✓ | ✓ | Receiver applies flow control |
PKT | ✓ | ✓ | Transfer of a new packet header |
DATA | ✓ | ✓ | Transfer of a payload word |
IDLES | ✓ | ✓ | No communication over the link |
LATENCY | ✓ | – | Debug bit detection |
The probes can be configured to filter the traffic based on the criteria summarized in Table 33-18.
Filters | Comment |
---|---|
Master address | Mask and match |
Slave address(2) | |
UserInfo | |
Read | Opcode is a load |
Write | Opcode is a store |
Error | Mask and match |
OCP address(1) |
Table 33-19 specifies the master address mapping (all statistics collectors).
Master-ID | Initiator |
---|---|
0x0 | MPU |
0x10 | CS_DAP |
0x14 | IEEE1500_2_OCP |
0x20 | DSP1 MDMA |
0x24 | DSP1 CFG |
0x28 | DSP1 DMA |
0x2C | DSP2 DMA |
0x30 | DSP2 CFG |
0x34 | DSP2 MDMA |
0x3A | IVA ICONT1 |
0x42 | EVE1 P1 |
0x46 | EVE2 P1 |
0x4A | Reserved |
0x4E | Reserved |
0x50 | Reserved |
0x54 | Reserved |
0x58 | Reserved |
0x5C | Reserved |
0x60 | IPU1 |
0x64 | IPU2 |
0x68 | DMA_SYSTEM RD |
0x6A | DMA_SYSTEM WR |
0x6C | Reserved |
0x6E | Reserved |
0x70 | EDMA_TC1 WR |
0x72 | EDMA_TC1 RD |
0x74 | EDMA_TC2 WR |
0x76 | EDMA_TC2 RD |
0x80 | DSS |
0x84 | MLB |
0x86 | MMU1 |
0x88 | PCIe_SS1 |
0x8C | PCIe_SS2 |
0x8E | MMU2 |
0x90 | VIP1 P1 |
0x92 | VIP1 P2 |
0x94 | VIP2 P1 |
0x96 | VIP2 P2 |
0x98 | VIP3 P1 |
0x9A | VIP3 P2 |
0x9C | VPE P1 |
0x9E | VPE P2 |
0xA0 | MMC1 |
0xA2 | GPU P1 |
0xA4 | MMC2 |
0xA6 | GPU P2 |
0xA8 | BB2D P1 |
0xAA | BB2D P2 |
0xAC | GMAC_SW |
0xAE | Reserved |
0xB0 | USB4 |
0xB4 | USB1 |
0xB8 | USB2 |
0xBC | USB3 |
0xC0 | Reserved |
0xC4 | Reserved |
0xCC | SATA |
0xD2 | EVE1 P2 |
0xD6 | EVE2 P2 |
0xDA | Reserved |
0xDE | Reserved |
Table 33-20 specifies the slave address mapping (SC_LAT only).
Slave | Address (hex) |
---|---|
Reserved | 0x0 |
Reserved | 0x1 |
DMM P1 | 0x2 |
DMM P2 | 0x3 |
DSP1 SDMA | 0x4 |
DSP2 SDMA | 0x5 |
DSS | 0x6 |
EVE1 | 0x7 |
EVE2 | 0x8 |
Reserved | 0x9 |
Reserved | 0xA |
BB2D | 0xB |
GPMC | 0xC |
GPU | 0xD |
HOST_CLK1_1 | 0xE |
HOST_CLK1_2 | 0xF |
IPU1 | 0x10 |
IPU2 | 0x11 |
IVA CONFIG | 0x12 |
IVA SL2IF | 0x13 |
L4_CFG | 0x14 |
L4_PER1 P1 | 0x15 |
L4_PER1 P2 | 0x16 |
L4_PER1 P3 | 0x17 |
L4_PER2 P1 | 0x18 |
L3_INSTR | 0x19 |
L4_PER2 P3 | 0x1A |
L4_PER3 P1 | 0x1B |
L4_PER3 P2 | 0x1C |
L4_PER3 P3 | 0x1D |
L4_WKUP | 0x1E |
McASP1 | 0x1F |
McASP2 | 0x20 |
McASP3 | 0x21 |
MMU1 | 0x22 |
MMU2 | 0x23 |
OCMC_RAM1 | 0x24 |
OCMC_RAM2 | 0x25 |
OCMC_RAM3 | 0x26 |
Reserved | 0x27 |
PCIe_SS1 | 0x28 |
PCIe_SS2 | 0x29 |
Reserved | 0x2A |
Reserved | 0x2B |
Reserved | 0x2C |
Reserved | 0x2D |
Reserved | 0x2E |
Reserved | 0x2F |
EDMA_TPCC | 0x30 |
EDMA_TC1 | 0x31 |
EDMA_TC2 | 0x32 |
Reserved | 0x35 |
VCP1 | 0x36 |
VCP2 | 0x37 |
QSPI | 0x39 |
HOST_CLK2_1 | 0x40 |
DEBUGSS CT_TBR | 0x41 |
L4_PER2 P2 | 0x42 |
The probes implement a user-defined set of counters that aggregate the events sampled by the detector and filtered according to the user setup.
Statistics collectors counter values are accessible by application software.
Table 33-21 summarizes the performance probe aggregation modes.
Aggregation Mode | Description |
---|---|
FILTER_HIT | The counter increments by 1 when the filter hits. |
MIN_MAX_HIT | The counter increments by 1 when the filter hits and the selected event information is within range. |
– Payload length (bytes) | |
– Pressure value | |
– Request/response latency (clock cycles) | |
EVT_INFO | The selected event information is added to the counter value when the filter hits. |
Payload length (bytes) | |
Pressure value | |
Request/response latency (clock cycles) | |
AND_FILTER | The counter increments by 1 when all unit filters hit. |
OR_FILTER | The counter increments by 1 when at least one unit filter hits. |
SUM_REQ_EVT | The counter sums the events from any request port. |
SUM_RSP_EVT | The counter sums the events from any response port. |
SUM_ALL_EVT | The counter sums the events from any port. |
EXT_EVT | The counter increments by 1 when selected external event input signal is sampled high. |