Intel® Stratix® 10 SX Device Errata

Download PDF
ID 683399
Date 7/21/2022
Public
Document Table of Contents
Document Table of Contents x
1. About this Document 2. Intel® -Specific Errata for the Intel® Stratix® 10 SX Devices 3. Arm* Cortex*-A53 MPCore* and CoreSight* Errata 4. Document Revision History for Intel® Stratix® 10 SX Device Errata
2. Intel® -Specific Errata for the Intel® Stratix® 10 SX Devices x
2.1. FPGA 2.2. Device Power Supply, Core, Configuration and I/O 2.3. Hard Processor System
2.2. Device Power Supply, Core, Configuration and I/O x
2.2.1. M20K Simple Quad-Port Mode Support
2.3. Hard Processor System x
2.3.1. HPS Stops on the First Read Request to SDRAM 2.3.2. Write Data Can Appear at an AXI Interface before the Write Address, which can Cause a Deadlock Condition
3. Arm* Cortex*-A53 MPCore* and CoreSight* Errata x
3.1. 843819: Memory Locations May be Accessed Speculatively Due to Instruction Fetches When HCR.VM is Set 3.2. 845719: A Load May Read Incorrect Data 3.3. 855871: ETM Does Not Report IDLE State When Disabled Using OSLOCK 3.4. 855872: A Store-Exclusive Instruction May Pass When it Should Fail 3.5. 711668: Configuration Extension Register Has Wrong Value Status 3.6. 720107: Periodic Synchronization Can Be Delayed and Cause Overflow 3.7. 855873: An Eviction Might Overtake a Cache Clean Operation 3.8. 853172: ETM May Assert AFREADY Before All Data Has Been Output 3.9. 836870: Non-Allocating Reads May Prevent a Store Exclusive From Passing 3.10. 836919: Write of the JMCR in EL0 Does Not Generate an UNDEFINED Exception 3.11. 845819: Instruction Sequences Containing AES Instructions May Produce Incorrect Results 3.12. 851672: ETM May Trace an Incorrect Exception Address 3.13. 851871: ETM May Lose Counter Events While Entering WFx Mode 3.14. 852071: Direct Branch Instructions Executed Before a Trace Flush May be Output in an Atom Packet After Flush Acknowledgment 3.15. 852521: A64 Unconditional Branch May Jump to Incorrect Address 3.16. 855827: PMU Counter Values May Be Inaccurate When Monitoring Certain Events 3.17. 855829: Reads of PMEVCNTR<n> are not Masked by HDCR.HPMN 3.18. 855830: Loads of Mismatched Size May not be Single-Copy Atomic
1. About this Document
2. Intel® -Specific Errata for the Intel® Stratix® 10 SX Devices
3. Arm* Cortex*-A53 MPCore* and CoreSight* Errata
4. Document Revision History for Intel® Stratix® 10 SX Device Errata

3.6. 720107: Periodic Synchronization Can Be Delayed and Cause Overflow

Description

The Program Trace Macrocell (PTM) is required to insert synchronization information into the trace stream at intervals in order to allow a partial trace dump to be decompressed. The synchronization period can be controlled either external to the PTM or by a counter that by default, requests synchronization every 1024 bytes of trace.

Synchronization does not need to be inserted precisely at a regular interval, so allowance is made to delay synchronization if the PTM is required to generate other trace packets, or if there is not sufficient space in the FIFO. To guarantee that some synchronization packets are always inserted regardless of the conditions, a 'forced overflow' mechanism shuts off trace if a new synchronization request occurs before the previous request was satisfied. This forced overflow mechanism is minimally intrusive to the trace stream, but ensures that synchronization is inserted after no more than 2x the requested interval.

Due to this erratum, some specific sequences of instructions prevent the PTM from being able to insert any synchronization into the trace stream while that instruction sequence continues. Typically, there is just a short delay which may not be noticed and the synchronization is inserted once the particular pattern of waypoints changes. It is possible that overflows are generated in the trace regardless of the utilization of the FIFO in the PTM. In this scenario, typically only a single byte of trace (up to 5 waypoints) is lost.

The scenario does not correspond to sustained high rates of trace generation which could genuinely cause the FIFO to become full.

Scenarios that cause this erratum are rare, and are limited to code iterating around a loop many times. The loop would contain several branches and be dependent on memory accesses completing.

This erratum can occur under the following conditions:

  1. Tracing is enabled.
  2. Branch broadcasting is disabled.
  3. Cycle accuracy is disabled.
  4. The processor executes tight loops of repeating code, lasting longer than the configured synchronization period.

Impact

This erratum reduces the frequency of periodic synchronization and potentially causes trace overflows where some trace is lost. In the case of an overflow, trace following the overflow can be correctly decompressed. This erratum is more noticeable if the periodic synchronization requests are more frequent.

Workaround

The most appropriate workaround depends on the use-case for the trace:

  • If the trace buffer is large enough, consider increasing the synchronization period by setting the ETMSYNCFR to a higher value.
  • If no trace must be lost (and periodic synchronization does not have to be guaranteed), set bit[0] of ETMAUXCR to disable the forced overflow function. Synchronization is inserted at the earliest opportunity, dependent on the executed instruction stream.

Category

Category 3

Level Two Title