Data trace support for compare and swap

[ved-rivos]

The Zacas extension introduces compare-and-swap atomic operations. It also introduces double-word CAS for RV32/64 and quad-word CAS for RV64. The data trace interface may need to be updated to factor these extensions.

[IainCRobertson]

Thanks @ved-rivos I wasn't aware of this. Ideally this would have been handled as part of the process of ratifying Zacas. We need a better way of ensuring that trace implications of new extensions are considered as part of the review and ratification process of that extension.

In terms of how to proceed with this, do you think we should try to fold this in to the RHTI fast-track (which was supposed to be non-normative in nature) or create a separate fast-track to add Zacas support? The latter seems more appropriate to me.

[IainCRobertson]

A compare and swap that succeeds can be treated identically to 'Atomic swap' (dtype = 8). A compare and swap that fails will need additional support however.

Firstly, there is a spare dtype code (7) we can use to indicate "Compare and swap failure".

Secondly however, this will need an update to the E-Trace spec, because a compare and swap that fails produces a load only, not a load-store pair. Tables 28 and 31 need updating to say that the format field can take the value 7, indicating a load resulting from a CAS failure.

[ved-rivos]

Does the double word compare and swap pose any new challenges ?

[IainCRobertson]

No. Both the bus widths and transaction sizes are abstracted/parameterized to be as large as is required. The transactions size is conveyed via the dsize input and this is independent of the load store type (dtype). Iain From: Ved Shanbhogue ***@***.***> Sent: 26 September 2025 16:04 To: riscv-non-isa/hart-trace-interface ***@***.***> Cc: Robertson, Iain (DI SW EDA DDCP TST RD SLE EAH) ***@***.***>; Comment ***@***.***> Subject: Re: [riscv-non-isa/hart-trace-interface] Data trace support for compare and swap (Issue #5) [https://avatars.githubusercontent.com/u/91900059?s=20&v=4]ved-rivos left a comment (riscv-non-isa/hart-trace-interface#5)<#5 (comment)> Does the double word compare and swap pose any new challenges ? - Reply to this email directly, view it on GitHub<#5 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ALQOPSSU2VKB7EOUTV6OXED3UVIXZAVCNFSM6AAAAACBHGC6E6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTGMZZGA4DMNZTHE>. You are receiving this because you commented.Message ID: ***@***.******@***.***>>

[IainCRobertson]

Updated included with other Atomic clarifications: #33