Skip to content

Commit

Permalink
Update charter.adoc
Browse files Browse the repository at this point in the history
Incorporate feedback from the first Performance Event Sampling TG meeting

Signed-off-by: Beeman Strong <[email protected]>
  • Loading branch information
bcstrongx authored Jun 11, 2024
1 parent 58f5066 commit 7cec7d8
Showing 1 changed file with 7 additions and 7 deletions.
14 changes: 7 additions & 7 deletions charter.adoc
Original file line number Diff line number Diff line change
@@ -1,12 +1,12 @@
= Preliminary Performance Sampling TG Charter
= Preliminary Performance Event Sampling TG Charter

RISC-V hardware performance monitoring counters (Zihpm) provide support for counting performance events, and, with Sscofpmf, support for basic, interrupt-based performance sampling. However, on most implementations sampling interrupts will skid, such that the resulting trap is taken some number of cycles and/or instructions after the instruction that caused the overflow retires. As a result the PC collected by the profiler will rarely match that of the causal instruction, since the PC will typically advance during the skid period. Other state that a profiler may want to collect (registers, call-stack, counter values, etc) is likely to be overwritten or modified as well.
RISC-V hardware performance monitoring counters (Zihpm) provide support for counting performance events, and, with Sscofpmf, support for basic, interrupt-based performance event sampling. However, on most implementations sampling interrupts will skid, such that the resulting trap is taken some number of cycles and/or instructions after the instruction that caused the overflow retires. As a result the PC collected by the profiler will rarely match that of the causal instruction, since the PC will typically advance during the skid period. Other state that a profiler may want to collect (registers, call-stack, counter values, etc) is likely to be overwritten or modified as well.

The Performance Sampling TG aims to address these limitations by defining two new ISA extensions:
The Performance Event Sampling TG aims to address these limitations by defining two new ISA extensions:

* An extension that enables precise attribution of samples based on select (non-speculative) events to the instruction that caused the counter overflow, despite implementations where the associated sampling interrupt may skid. This will provide more directly actionable information to the user, by precisely identifying the instructions that are most often experiencing performance events.
* An extension that enables sampling of instructions and/or uops, selected at dispatch, with collection of runtime event occurrences and latencies incurred by the instruction/uop. Such samples can be filtered based on instruction/uop type, events incurred, or latencies observed, allowing the user to focus on samples of interest. Further, associated sampling interrupts can be skidless, allowing the user to collect additional sample state (call-stack, register values) reliably.
* An extension that enables precise attribution of samples based on select events (e.g., instruction/uop retirement events) to the instruction that caused the counter overflow, despite implementations where the associated sampling interrupt may skid. This will provide more directly actionable information to the user, by precisely identifying the instructions that are most often experiencing performance events.
* An extension that enables sampling of instructions and/or uops, with collection of runtime event occurrences and latencies incurred by the instruction/uop. Such samples can be filtered based on instruction/uop type, events incurred, or latencies observed, allowing the user to focus on samples of interest. Further, associated sampling interrupts can be skidless, allowing the user to collect additional sample state (call-stack, register values) reliably.
Each extension will be crafted to be implementation-friendly even for high-performance, out-of-order microarchitectures, aiming to require no additional performance overhead beyond that resulting from the handling of sampling interrupts.
Each extension will be crafted to be implementation-friendly even for high-performance, out-of-order microarchitectures, aiming to require no additional performance overhead beyond that resulting from the handling of sampling interrupts. The extensions will be compatible with the H extension, and supoprt RISC-V security objectives.

The TG will prototype the new extensions in Spike or Qemu, along with Linux perf support to demonstrate the end-to-end solution.
The TG will prototype support for the new extensions in Qemu and Linux perf, to demonstrate the usability of the ISA for kernels and tools.

0 comments on commit 7cec7d8

Please sign in to comment.