Small suggested clarifications

[rsnikhil]

The word "E-Trace", used on pages 9, 10 and 79, presumably stands for "Efficient Trace", but AFAIK is nowhere so defined.

Suggest adding an item in Section 1.1 Terminology for this.

[rsnikhil]

‘ucause’ is mentioned in Section 3.2.3; Section 4.1 second bullet; table 4.
‘utval’ is mentioned in Section 4.1 second bullet; Table 4; Table 17; 7.3.2.

AFAIK these CSRs were removed from the Privileged ISA spec?

[rsnikhil]

(Note: a new and separate Issue has been opened for this: #142 (comment))

In Section 4.1, we have:

The following information is mandatory:
...
* The instruction_type of retired instructions for:
  * Jumps with a target that cannot be inferred from the source code;
...

and

The following information is optional:
...
* The instruction_type of instructions for:
  ...
  * Jumps with a target that cannot be inferred from the source code;

This seems contradictory (both mandatory and optional)?

[IainCRobertson]

Rishiyur, That's a good observation, and does indeed appear contradictory. It came about as a result of an earlier clarification attempt. The list of optional types in section 4.1 corresponds to the more detailed classifications in section 4.11. Originally, this was broken down into Calls, Tail-calls, Returns, Co-routine swap and Other. However, there is no formal definition of tail-call in the RISC-V ISA, and so references to 'tail call' in both section 4.1.1 and 4.1 were replaced with 'jump'. With hindsight, this is too general, and probably needs to be something like "jump with discarded link address" or "call with discarded link address" Iain From: Rishiyur S. Nikhil ***@***.***> Sent: 03 September 2024 18:32 To: riscv-non-isa/riscv-trace-spec ***@***.***> Cc: Subscribed ***@***.***> Subject: Re: [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) In Section 4.1, we have: The following information is mandatory: ... * The instruction_type of retired instructions for: * Jumps with a target that cannot be inferred from the source code; ... and The following information is optional: ... * The instruction_type of instructions for: ... * Jumps with a target that cannot be inferred from the source code; This seems contradictory (both mandatory and optional)? - Reply to this email directly, view it on GitHub<#138 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ALQOPSV4KFRG7M7M4S5F2K3ZUXW7VAVCNFSM6AAAAABNQZL5VGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGMRXGA3TAMJUHE>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.******@***.***>>

[IainCRobertson]

FWIW, when CTR went through ARC review, ARC requested the following definitions for the various control transfer types: [image: image.png]

…

On Wed, Sep 4, 2024 at 1:29 AM Mr Iain Robertson via lists.riscv.org ***@***.***> wrote: Rishiyur, That’s a good observation, and does indeed appear contradictory. It came about as a result of an earlier clarification attempt. The list of optional types in section 4.1 corresponds to the more detailed classifications in section 4.11. Originally, this was broken down into Calls, Tail-calls, Returns, Co-routine swap and Other. However, there is no formal definition of tail-call in the RISC-V ISA, and so references to ‘tail call’ in both section 4.1.1 and 4.1 were replaced with ‘jump’. With hindsight, this is too general, and probably needs to be something like “jump with discarded link address” or “call with discarded link address” Please open an issue against the spec ( https://github.com/riscv-non-isa/riscv-trace-spec/issues) and I’ll address this in the next update. Iain *From:* Rishiyur S. Nikhil ***@***.***> *Sent:* 03 September 2024 18:32 *To:* riscv-non-isa/riscv-trace-spec ***@***.***> *Cc:* Subscribed ***@***.***> *Subject:* Re: [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) In Section 4.1, we have: The following information is mandatory: ... * The instruction_type of retired instructions for: * Jumps with a target that cannot be inferred from the source code; ... and The following information is optional: ... * The instruction_type of instructions for: ... * Jumps with a target that cannot be inferred from the source code; This seems contradictory (both mandatory and optional)? — Reply to this email directly, view it on GitHub <#138 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALQOPSV4KFRG7M7M4S5F2K3ZUXW7VAVCNFSM6AAAAABNQZL5VGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGMRXGA3TAMJUHE> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***> _._,_._,_ ------------------------------ Links: You receive all messages sent to this group. View/Reply Online (#598) <https://lists.riscv.org/g/sig-debug-trace-perf-mon/message/598> | Reply To Group ***@***.***?subject=Re:%20Re%3A%20%5BRISC-V%5D%20%5Bsig-debug-trace-perf-mon%5D%20%5Briscv-non-isa%2Friscv-trace-spec%5D%20Small%20suggested%20clarifications%20%28Issue%20%23138%29> | Reply To Sender ***@***.***?subject=Private:%20Re:%20Re%3A%20%5BRISC-V%5D%20%5Bsig-debug-trace-perf-mon%5D%20%5Briscv-non-isa%2Friscv-trace-spec%5D%20Small%20suggested%20clarifications%20%28Issue%20%23138%29> | Mute This Topic <https://lists.riscv.org/mt/108261477/6477655> | New Topic <https://lists.riscv.org/g/sig-debug-trace-perf-mon/post> Your Subscription <https://lists.riscv.org/g/sig-debug-trace-perf-mon/editsub/6477655> | Contact Group Owner ***@***.***> | Unsubscribe <https://lists.riscv.org/g/sig-debug-trace-perf-mon/unsub> [ ***@***.*** _._,_._,_

[IainCRobertson]

Beeman, Thanks – that looks like the terminology we should adopt. Unfortunately your screen-grab doesn’t show up in the GIT issue – can you paste plain text or include a reference that I can follow later please? Iain From: Beeman Strong ***@***.***> Sent: 04 September 2024 15:45 To: ***@***.***; Robertson, Iain (DI SW ICS DDCP TST RD EAH) ***@***.***> Cc: riscv-non-isa/riscv-trace-spec ***@***.***>; riscv-non-isa/riscv-trace-spec ***@***.***>; Subscribed ***@***.***> Subject: Re: [RISC-V] [sig-debug-trace-perf-mon] [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) FWIW, when CTR went through ARC review, ARC requested the following definitions for the various control transfer types: ***@***.*** On Wed, Sep 4, 2024 at 1:29 AM Mr Iain Robertson via lists.riscv.org<http://lists.riscv.org/> ***@***.******@***.***>> wrote: Rishiyur, That’s a good observation, and does indeed appear contradictory. It came about as a result of an earlier clarification attempt. The list of optional types in section 4.1 corresponds to the more detailed classifications in section 4.11. Originally, this was broken down into Calls, Tail-calls, Returns, Co-routine swap and Other. However, there is no formal definition of tail-call in the RISC-V ISA, and so references to ‘tail call’ in both section 4.1.1 and 4.1 were replaced with ‘jump’. With hindsight, this is too general, and probably needs to be something like “jump with discarded link address” or “call with discarded link address” Please open an issue against the spec (https://github.com/riscv-non-isa/riscv-trace-spec/issues) and I’ll address this in the next update. Iain From: Rishiyur S. Nikhil ***@***.******@***.***>> Sent: 03 September 2024 18:32 To: riscv-non-isa/riscv-trace-spec ***@***.******@***.***>> Cc: Subscribed ***@***.******@***.***>> Subject: Re: [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) In Section 4.1, we have: The following information is mandatory: ... * The instruction_type of retired instructions for: * Jumps with a target that cannot be inferred from the source code; ... and The following information is optional: ... * The instruction_type of instructions for: ... * Jumps with a target that cannot be inferred from the source code; This seems contradictory (both mandatory and optional)? — Reply to this email directly, view it on GitHub<#138 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ALQOPSV4KFRG7M7M4S5F2K3ZUXW7VAVCNFSM6AAAAABNQZL5VGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGMRXGA3TAMJUHE>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.******@***.***>> _._,_._,_

…

________________________________ Links: You receive all messages sent to this group. View/Reply Online (#598)<https://lists.riscv.org/g/sig-debug-trace-perf-mon/message/598> | Reply To ***@***.***?subject=Re:%20Re%3A%20%5BRISC-V%5D%20%5Bsig-debug-trace-perf-mon%5D%20%5Briscv-non-isa%2Friscv-trace-spec%5D%20Small%20suggested%20clarifications%20%28Issue%20%23138%29> | Reply To ***@***.***?subject=Private:%20Re:%20Re%3A%20%5BRISC-V%5D%20%5Bsig-debug-trace-perf-mon%5D%20%5Briscv-non-isa%2Friscv-trace-spec%5D%20Small%20suggested%20clarifications%20%28Issue%20%23138%29> | Mute This Topic<https://lists.riscv.org/mt/108261477/6477655> | New Topic<https://lists.riscv.org/g/sig-debug-trace-perf-mon/post> Your Subscription<https://lists.riscv.org/g/sig-debug-trace-perf-mon/editsub/6477655> | Contact Group ***@***.***> | Unsubscribe<https://lists.riscv.org/g/sig-debug-trace-perf-mon/unsub> ***@***.******@***.***>] _._,_._,_

[IainCRobertson]

Sure, I'll paste the adoc text. On Wed, Sep 4, 2024 at 8:40 AM Robertson, Iain ***@***.***> wrote:

…

Beeman, Thanks – that looks like the terminology we should adopt. Unfortunately your screen-grab doesn’t show up in the GIT issue – can you paste plain text or include a reference that I can follow later please? Iain *From:* Beeman Strong ***@***.***> *Sent:* 04 September 2024 15:45 *To:* ***@***.***; Robertson, Iain (DI SW ICS DDCP TST RD EAH) ***@***.***> *Cc:* riscv-non-isa/riscv-trace-spec < ***@***.***>; riscv-non-isa/riscv-trace-spec ***@***.***>; Subscribed ***@***.***> *Subject:* Re: [RISC-V] [sig-debug-trace-perf-mon] [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) FWIW, when CTR went through ARC review, ARC requested the following definitions for the various control transfer types: On Wed, Sep 4, 2024 at 1:29 AM Mr Iain Robertson via lists.riscv.org ***@***.***> wrote: Rishiyur, That’s a good observation, and does indeed appear contradictory. It came about as a result of an earlier clarification attempt. The list of optional types in section 4.1 corresponds to the more detailed classifications in section 4.11. Originally, this was broken down into Calls, Tail-calls, Returns, Co-routine swap and Other. However, there is no formal definition of tail-call in the RISC-V ISA, and so references to ‘tail call’ in both section 4.1.1 and 4.1 were replaced with ‘jump’. With hindsight, this is too general, and probably needs to be something like “jump with discarded link address” or “call with discarded link address” Please open an issue against the spec ( https://github.com/riscv-non-isa/riscv-trace-spec/issues) and I’ll address this in the next update. Iain *From:* Rishiyur S. Nikhil ***@***.***> *Sent:* 03 September 2024 18:32 *To:* riscv-non-isa/riscv-trace-spec ***@***.***> *Cc:* Subscribed ***@***.***> *Subject:* Re: [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) In Section 4.1, we have: The following information is mandatory: ... * The instruction_type of retired instructions for: * Jumps with a target that cannot be inferred from the source code; ... and The following information is optional: ... * The instruction_type of instructions for: ... * Jumps with a target that cannot be inferred from the source code; This seems contradictory (both mandatory and optional)? — Reply to this email directly, view it on GitHub <#138 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALQOPSV4KFRG7M7M4S5F2K3ZUXW7VAVCNFSM6AAAAABNQZL5VGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGMRXGA3TAMJUHE> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***> _._,_._,_ ------------------------------ Links: You receive all messages sent to this group. View/Reply Online (#598) <https://lists.riscv.org/g/sig-debug-trace-perf-mon/message/598> | Reply To Group ***@***.***?subject=Re:%20Re%3A%20%5BRISC-V%5D%20%5Bsig-debug-trace-perf-mon%5D%20%5Briscv-non-isa%2Friscv-trace-spec%5D%20Small%20suggested%20clarifications%20%28Issue%20%23138%29> | Reply To Sender ***@***.***?subject=Private:%20Re:%20Re%3A%20%5BRISC-V%5D%20%5Bsig-debug-trace-perf-mon%5D%20%5Briscv-non-isa%2Friscv-trace-spec%5D%20Small%20suggested%20clarifications%20%28Issue%20%23138%29> | Mute This Topic <https://lists.riscv.org/mt/108261477/6477655> | New Topic <https://lists.riscv.org/g/sig-debug-trace-perf-mon/post> Your Subscription <https://lists.riscv.org/g/sig-debug-trace-perf-mon/editsub/6477655> | Contact Group Owner ***@***.***> | Unsubscribe <https://lists.riscv.org/g/sig-debug-trace-perf-mon/unsub> [ ***@***.*** _._,_._,_

[IainCRobertson]

It was also improved in N-Trace. See this: https://github.com/riscv-non-isa/tg-nexus-trace/blob/main/docs/RISC-V-N-Trace.adoc#2-trace-ingress-port IMPORTANT: I am against that PR mentioned below. We agreed to EXTRACT this into SINGLE source (as fast-track). Thanks, Robert

…

________________________________ From: ***@***.*** ***@***.***> on behalf of Mr Iain Robertson via lists.riscv.org ***@***.***> Sent: Wednesday, September 4, 2024 8:40 AM To: Beeman Strong ***@***.***>; ***@***.*** ***@***.***> Cc: riscv-non-isa/riscv-trace-spec ***@***.***>; riscv-non-isa/riscv-trace-spec ***@***.***>; Subscribed ***@***.***> Subject: [EXTERNAL]Re: [RISC-V] [sig-debug-trace-perf-mon] [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) Beeman, Thanks – that looks like the terminology we should adopt. Unfortunately your screen-grab doesn’t show up in the GIT issue – can you paste plain text or include a reference that I can follow later please? Iain From: Beeman Strong ***@***.***> Sent: 04 September 2024 15:45 To: ***@***.***; Robertson, Iain (DI SW ICS DDCP TST RD EAH) ***@***.***> Cc: riscv-non-isa/riscv-trace-spec ***@***.***>; riscv-non-isa/riscv-trace-spec ***@***.***>; Subscribed ***@***.***> Subject: Re: [RISC-V] [sig-debug-trace-perf-mon] [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) FWIW, when CTR went through ARC review, ARC requested the following definitions for the various control transfer types: ***@***.*** On Wed, Sep 4, 2024 at 1:29 AM Mr Iain Robertson via lists.riscv.org<http://lists.riscv.org/> ***@***.******@***.***>> wrote: Rishiyur, That’s a good observation, and does indeed appear contradictory. It came about as a result of an earlier clarification attempt. The list of optional types in section 4.1 corresponds to the more detailed classifications in section 4.11. Originally, this was broken down into Calls, Tail-calls, Returns, Co-routine swap and Other. However, there is no formal definition of tail-call in the RISC-V ISA, and so references to ‘tail call’ in both section 4.1.1 and 4.1 were replaced with ‘jump’. With hindsight, this is too general, and probably needs to be something like “jump with discarded link address” or “call with discarded link address” Please open an issue against the spec (https://github.com/riscv-non-isa/riscv-trace-spec/issues) and I’ll address this in the next update. Iain From: Rishiyur S. Nikhil ***@***.******@***.***>> Sent: 03 September 2024 18:32 To: riscv-non-isa/riscv-trace-spec ***@***.******@***.***>> Cc: Subscribed ***@***.******@***.***>> Subject: Re: [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) In Section 4.1, we have: The following information is mandatory: ... * The instruction_type of retired instructions for: * Jumps with a target that cannot be inferred from the source code; ... and The following information is optional: ... * The instruction_type of instructions for: ... * Jumps with a target that cannot be inferred from the source code; This seems contradictory (both mandatory and optional)? — Reply to this email directly, view it on GitHub<#138 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ALQOPSV4KFRG7M7M4S5F2K3ZUXW7VAVCNFSM6AAAAABNQZL5VGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGMRXGA3TAMJUHE>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.******@***.***>> _._,_._,_

________________________________ Links: You receive all messages sent to this group. View/Reply Online (#610)<https://lists.riscv.org/g/sig-debug-trace-perf-mon/message/610> | Reply To ***@***.***?subject=Re:%20Re%3A%20%5BRISC-V%5D%20%5Bsig-debug-trace-perf-mon%5D%20%5Briscv-non-isa%2Friscv-trace-spec%5D%20Small%20suggested%20clarifications%20%28Issue%20%23138%29> | Reply To ***@***.***?subject=Private:%20Re:%20Re%3A%20%5BRISC-V%5D%20%5Bsig-debug-trace-perf-mon%5D%20%5Briscv-non-isa%2Friscv-trace-spec%5D%20Small%20suggested%20clarifications%20%28Issue%20%23138%29> | Mute This Topic<https://lists.riscv.org/mt/108261477/1782486> | New Topic<https://lists.riscv.org/g/sig-debug-trace-perf-mon/post> Your Subscription<https://lists.riscv.org/g/sig-debug-trace-perf-mon/editsub/1782486> | Contact Group ***@***.***> | Unsubscribe<https://lists.riscv.org/g/sig-debug-trace-perf-mon/unsub> ***@***.*** _._,_._,_

[IainCRobertson]

I see the issue hasn't been filed yet, so pasting the text here: |INDCALLINH |Inhibit recording of indirect calls. See <<Transfer Type Filtering>>. |DIRCALLINH |Inhibit recording of direct calls. See <<Transfer Type Filtering>>. |INDJMPINH |Inhibit recording of indirect jumps (without linkage). See <<Transfer Type Filtering>>. |DIRJMPINH |Inhibit recording of direct jumps (without linkage). See <<Transfer Type Filtering>>. |CORSWAPINH |Inhibit recording of co-routine swaps. See <<Transfer Type Filtering>>. |RETINH |Inhibit recording of function returns. See <<Transfer Type Filtering>>. |INDLJMPINH |Inhibit recording of other indirect jumps (with linkage). See <<Transfer Type Filtering>>. |DIRLJMPINH |Inhibit recording of other direct jumps (with linkage). See <<Transfer Type Filtering>>.

…

On Wed, Sep 4, 2024 at 8:58 AM Beeman Strong ***@***.***> wrote: Sure, I'll paste the adoc text. On Wed, Sep 4, 2024 at 8:40 AM Robertson, Iain ***@***.***> wrote: > Beeman, > > Thanks – that looks like the terminology we should > adopt. Unfortunately your screen-grab doesn’t show up in the GIT issue – > can you paste plain text or include a reference that I can follow later > please? > > > > Iain > > > > *From:* Beeman Strong ***@***.***> > *Sent:* 04 September 2024 15:45 > *To:* ***@***.***; Robertson, Iain (DI SW > ICS DDCP TST RD EAH) ***@***.***> > *Cc:* riscv-non-isa/riscv-trace-spec < > ***@***.***>; > riscv-non-isa/riscv-trace-spec ***@***.***>; > Subscribed ***@***.***> > *Subject:* Re: [RISC-V] [sig-debug-trace-perf-mon] > [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) > > > > FWIW, when CTR went through ARC review, ARC requested the following > definitions for the various control transfer types: > > > > > > > > On Wed, Sep 4, 2024 at 1:29 AM Mr Iain Robertson via lists.riscv.org > ***@***.***> wrote: > > Rishiyur, > > That’s a good observation, and does indeed appear > contradictory. It came about as a result of an earlier clarification > attempt. The list of optional types in section 4.1 corresponds to the more > detailed classifications in section 4.11. Originally, this was broken down > into Calls, Tail-calls, Returns, Co-routine swap and Other. However, there > is no formal definition of tail-call in the RISC-V ISA, and so references > to ‘tail call’ in both section 4.1.1 and 4.1 were replaced with ‘jump’. > With hindsight, this is too general, and probably needs to be something > like “jump with discarded link address” or “call with discarded link > address” > > > > Please open an issue against the spec ( > https://github.com/riscv-non-isa/riscv-trace-spec/issues) and I’ll > address this in the next update. > > > > Iain > > > > *From:* Rishiyur S. Nikhil ***@***.***> > *Sent:* 03 September 2024 18:32 > *To:* riscv-non-isa/riscv-trace-spec ***@***.*** > > > *Cc:* Subscribed ***@***.***> > *Subject:* Re: [riscv-non-isa/riscv-trace-spec] Small suggested > clarifications (Issue #138) > > > > In Section 4.1, we have: > > The following information is mandatory: > > ... > > * The instruction_type of retired instructions for: > > * Jumps with a target that cannot be inferred from the source code; > > ... > > and > > The following information is optional: > > ... > > * The instruction_type of instructions for: > > ... > > * Jumps with a target that cannot be inferred from the source code; > > This seems contradictory (both mandatory and optional)? > > — > Reply to this email directly, view it on GitHub > <#138 (comment)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/ALQOPSV4KFRG7M7M4S5F2K3ZUXW7VAVCNFSM6AAAAABNQZL5VGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGMRXGA3TAMJUHE> > . > You are receiving this because you are subscribed to this thread.Message > ID: ***@***.***> > > _._,_._,_ > ------------------------------ > > Links: > > You receive all messages sent to this group. > > View/Reply Online (#598) > <https://lists.riscv.org/g/sig-debug-trace-perf-mon/message/598> | Reply > To Group > ***@***.***?subject=Re:%20Re%3A%20%5BRISC-V%5D%20%5Bsig-debug-trace-perf-mon%5D%20%5Briscv-non-isa%2Friscv-trace-spec%5D%20Small%20suggested%20clarifications%20%28Issue%20%23138%29> > | Reply To Sender > ***@***.***?subject=Private:%20Re:%20Re%3A%20%5BRISC-V%5D%20%5Bsig-debug-trace-perf-mon%5D%20%5Briscv-non-isa%2Friscv-trace-spec%5D%20Small%20suggested%20clarifications%20%28Issue%20%23138%29> > | Mute This Topic <https://lists.riscv.org/mt/108261477/6477655> | New > Topic <https://lists.riscv.org/g/sig-debug-trace-perf-mon/post> > Your Subscription > <https://lists.riscv.org/g/sig-debug-trace-perf-mon/editsub/6477655> | Contact > Group Owner ***@***.***> | > Unsubscribe <https://lists.riscv.org/g/sig-debug-trace-perf-mon/unsub> [ > ***@***.*** > > _._,_._,_ > >

[IainCRobertson]

Robert, The issue is for a request to clarify chapter 4. Whether that is done as part of an incremental update to the E-Trace spec (along with some other clarifications and corrections) before we pull out chapter 4, or whether chapter 4 gets pulled out first and the clarifications are handled as part of that makes no difference – the important thing is that we have a record of the need for clarification. Iain From: Robert Chyla ***@***.***> Sent: 04 September 2024 17:32 To: Beeman Strong ***@***.***>; ***@***.***; Robertson, Iain (DI SW ICS DDCP TST RD EAH) ***@***.***> Cc: riscv-non-isa/riscv-trace-spec ***@***.***>; riscv-non-isa/riscv-trace-spec ***@***.***>; Subscribed ***@***.***> Subject: Re: [EXTERNAL]Re: [RISC-V] [sig-debug-trace-perf-mon] [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) It was also improved in N-Trace. See this: https://github.com/riscv-non-isa/tg-nexus-trace/blob/main/docs/RISC-V-N-Trace.adoc#2-trace-ingress-port IMPORTANT: I am against that PR mentioned below. We agreed to EXTRACT this into SINGLE source (as fast-track). Thanks, Robert

…

________________________________ From: ***@***.******@***.***> ***@***.******@***.***>> on behalf of Mr Iain Robertson via lists.riscv.org ***@***.******@***.***>> Sent: Wednesday, September 4, 2024 8:40 AM To: Beeman Strong ***@***.******@***.***>>; ***@***.******@***.***> ***@***.******@***.***>> Cc: riscv-non-isa/riscv-trace-spec ***@***.******@***.***>>; riscv-non-isa/riscv-trace-spec ***@***.******@***.***>>; Subscribed ***@***.******@***.***>> Subject: [EXTERNAL]Re: [RISC-V] [sig-debug-trace-perf-mon] [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) Beeman, Thanks – that looks like the terminology we should adopt. Unfortunately your screen-grab doesn’t show up in the GIT issue – can you paste plain text or include a reference that I can follow later please? Iain From: Beeman Strong ***@***.******@***.***>> Sent: 04 September 2024 15:45 To: ***@***.******@***.***>; Robertson, Iain (DI SW ICS DDCP TST RD EAH) ***@***.******@***.***>> Cc: riscv-non-isa/riscv-trace-spec ***@***.******@***.***>>; riscv-non-isa/riscv-trace-spec ***@***.******@***.***>>; Subscribed ***@***.******@***.***>> Subject: Re: [RISC-V] [sig-debug-trace-perf-mon] [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) FWIW, when CTR went through ARC review, ARC requested the following definitions for the various control transfer types: ***@***.*** On Wed, Sep 4, 2024 at 1:29 AM Mr Iain Robertson via lists.riscv.org<http://lists.riscv.org/> ***@***.******@***.***>> wrote: Rishiyur, That’s a good observation, and does indeed appear contradictory. It came about as a result of an earlier clarification attempt. The list of optional types in section 4.1 corresponds to the more detailed classifications in section 4.11. Originally, this was broken down into Calls, Tail-calls, Returns, Co-routine swap and Other. However, there is no formal definition of tail-call in the RISC-V ISA, and so references to ‘tail call’ in both section 4.1.1 and 4.1 were replaced with ‘jump’. With hindsight, this is too general, and probably needs to be something like “jump with discarded link address” or “call with discarded link address” Please open an issue against the spec (https://github.com/riscv-non-isa/riscv-trace-spec/issues) and I’ll address this in the next update. Iain From: Rishiyur S. Nikhil ***@***.******@***.***>> Sent: 03 September 2024 18:32 To: riscv-non-isa/riscv-trace-spec ***@***.******@***.***>> Cc: Subscribed ***@***.******@***.***>> Subject: Re: [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) In Section 4.1, we have: The following information is mandatory: ... * The instruction_type of retired instructions for: * Jumps with a target that cannot be inferred from the source code; ... and The following information is optional: ... * The instruction_type of instructions for: ... * Jumps with a target that cannot be inferred from the source code; This seems contradictory (both mandatory and optional)? — Reply to this email directly, view it on GitHub<#138 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ALQOPSV4KFRG7M7M4S5F2K3ZUXW7VAVCNFSM6AAAAABNQZL5VGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGMRXGA3TAMJUHE>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.******@***.***>> _._,_._,_

________________________________ Links: You receive all messages sent to this group. View/Reply Online (#610)<https://lists.riscv.org/g/sig-debug-trace-perf-mon/message/610> | Reply To ***@***.***?subject=Re:%20Re%3A%20%5BRISC-V%5D%20%5Bsig-debug-trace-perf-mon%5D%20%5Briscv-non-isa%2Friscv-trace-spec%5D%20Small%20suggested%20clarifications%20%28Issue%20%23138%29> | Reply To ***@***.***?subject=Private:%20Re:%20Re%3A%20%5BRISC-V%5D%20%5Bsig-debug-trace-perf-mon%5D%20%5Briscv-non-isa%2Friscv-trace-spec%5D%20Small%20suggested%20clarifications%20%28Issue%20%23138%29> | Mute This Topic<https://lists.riscv.org/mt/108261477/1782486> | New Topic<https://lists.riscv.org/g/sig-debug-trace-perf-mon/post> Your Subscription<https://lists.riscv.org/g/sig-debug-trace-perf-mon/editsub/1782486> | Contact Group ***@***.***> | Unsubscribe<https://lists.riscv.org/g/sig-debug-trace-perf-mon/unsub> ***@***.*** _._,_._,_

[IainCRobertson]

N-Trace has 'correct' (ARC-blessed ...) names. We (me, Jay, Ved=ARC) spent a lot of time (several meetings with ARC) making 'itype' names correctly named and correctly explained. I assume 'extracted spec' will inherit this. Thanks, Robert

…

________________________________ From: Robertson, Iain ***@***.***> Sent: Wednesday, September 4, 2024 9:50 AM To: Robert Chyla ***@***.***>; Beeman Strong ***@***.***>; ***@***.*** ***@***.***> Cc: riscv-non-isa/riscv-trace-spec ***@***.***>; riscv-non-isa/riscv-trace-spec ***@***.***>; Subscribed ***@***.***> Subject: RE: [EXTERNAL]Re: [RISC-V] [sig-debug-trace-perf-mon] [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) Robert, The issue is for a request to clarify chapter 4. Whether that is done as part of an incremental update to the E-Trace spec (along with some other clarifications and corrections) before we pull out chapter 4, or whether chapter 4 gets pulled out first and the clarifications are handled as part of that makes no difference – the important thing is that we have a record of the need for clarification. Iain From: Robert Chyla ***@***.***> Sent: 04 September 2024 17:32 To: Beeman Strong ***@***.***>; ***@***.***; Robertson, Iain (DI SW ICS DDCP TST RD EAH) ***@***.***> Cc: riscv-non-isa/riscv-trace-spec ***@***.***>; riscv-non-isa/riscv-trace-spec ***@***.***>; Subscribed ***@***.***> Subject: Re: [EXTERNAL]Re: [RISC-V] [sig-debug-trace-perf-mon] [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) It was also improved in N-Trace. See this: https://github.com/riscv-non-isa/tg-nexus-trace/blob/main/docs/RISC-V-N-Trace.adoc#2-trace-ingress-port IMPORTANT: I am against that PR mentioned below. We agreed to EXTRACT this into SINGLE source (as fast-track). Thanks, Robert

________________________________ From: ***@***.******@***.***> ***@***.******@***.***>> on behalf of Mr Iain Robertson via lists.riscv.org ***@***.******@***.***>> Sent: Wednesday, September 4, 2024 8:40 AM To: Beeman Strong ***@***.******@***.***>>; ***@***.******@***.***> ***@***.******@***.***>> Cc: riscv-non-isa/riscv-trace-spec ***@***.******@***.***>>; riscv-non-isa/riscv-trace-spec ***@***.******@***.***>>; Subscribed ***@***.******@***.***>> Subject: [EXTERNAL]Re: [RISC-V] [sig-debug-trace-perf-mon] [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) Beeman, Thanks – that looks like the terminology we should adopt. Unfortunately your screen-grab doesn’t show up in the GIT issue – can you paste plain text or include a reference that I can follow later please? Iain From: Beeman Strong ***@***.******@***.***>> Sent: 04 September 2024 15:45 To: ***@***.******@***.***>; Robertson, Iain (DI SW ICS DDCP TST RD EAH) ***@***.******@***.***>> Cc: riscv-non-isa/riscv-trace-spec ***@***.******@***.***>>; riscv-non-isa/riscv-trace-spec ***@***.******@***.***>>; Subscribed ***@***.******@***.***>> Subject: Re: [RISC-V] [sig-debug-trace-perf-mon] [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) FWIW, when CTR went through ARC review, ARC requested the following definitions for the various control transfer types: ***@***.*** On Wed, Sep 4, 2024 at 1:29 AM Mr Iain Robertson via lists.riscv.org<http://lists.riscv.org/> ***@***.******@***.***>> wrote: Rishiyur, That’s a good observation, and does indeed appear contradictory. It came about as a result of an earlier clarification attempt. The list of optional types in section 4.1 corresponds to the more detailed classifications in section 4.11. Originally, this was broken down into Calls, Tail-calls, Returns, Co-routine swap and Other. However, there is no formal definition of tail-call in the RISC-V ISA, and so references to ‘tail call’ in both section 4.1.1 and 4.1 were replaced with ‘jump’. With hindsight, this is too general, and probably needs to be something like “jump with discarded link address” or “call with discarded link address” Please open an issue against the spec (https://github.com/riscv-non-isa/riscv-trace-spec/issues) and I’ll address this in the next update. Iain From: Rishiyur S. Nikhil ***@***.******@***.***>> Sent: 03 September 2024 18:32 To: riscv-non-isa/riscv-trace-spec ***@***.******@***.***>> Cc: Subscribed ***@***.******@***.***>> Subject: Re: [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) In Section 4.1, we have: The following information is mandatory: ... * The instruction_type of retired instructions for: * Jumps with a target that cannot be inferred from the source code; ... and The following information is optional: ... * The instruction_type of instructions for: ... * Jumps with a target that cannot be inferred from the source code; This seems contradictory (both mandatory and optional)? — Reply to this email directly, view it on GitHub<#138 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ALQOPSV4KFRG7M7M4S5F2K3ZUXW7VAVCNFSM6AAAAABNQZL5VGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGMRXGA3TAMJUHE>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.******@***.***>> _._,_._,_

________________________________ Links: You receive all messages sent to this group. View/Reply Online (#610)<https://lists.riscv.org/g/sig-debug-trace-perf-mon/message/610> | Reply To ***@***.***?subject=Re:%20Re%3A%20%5BRISC-V%5D%20%5Bsig-debug-trace-perf-mon%5D%20%5Briscv-non-isa%2Friscv-trace-spec%5D%20Small%20suggested%20clarifications%20%28Issue%20%23138%29> | Reply To ***@***.***?subject=Private:%20Re:%20Re%3A%20%5BRISC-V%5D%20%5Bsig-debug-trace-perf-mon%5D%20%5Briscv-non-isa%2Friscv-trace-spec%5D%20Small%20suggested%20clarifications%20%28Issue%20%23138%29> | Mute This Topic<https://lists.riscv.org/mt/108261477/1782486> | New Topic<https://lists.riscv.org/g/sig-debug-trace-perf-mon/post> Your Subscription<https://lists.riscv.org/g/sig-debug-trace-perf-mon/editsub/1782486> | Contact Group ***@***.***> | Unsubscribe<https://lists.riscv.org/g/sig-debug-trace-perf-mon/unsub> ***@***.*** _._,_._,_

[IainCRobertson]

Robert, Yes, and those are the same names Beeman put forward. Again, no reason not to update the E-Trace spec if it happens that an update is done before chapter 4 is pulled out. Iain From: Robert Chyla ***@***.***> Sent: 04 September 2024 17:54 To: Robertson, Iain (DI SW ICS DDCP TST RD EAH) ***@***.***>; Beeman Strong ***@***.***>; ***@***.*** Cc: riscv-non-isa/riscv-trace-spec ***@***.***>; riscv-non-isa/riscv-trace-spec ***@***.***>; Subscribed ***@***.***> Subject: Re: [EXTERNAL]Re: [RISC-V] [sig-debug-trace-perf-mon] [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) N-Trace has 'correct' (ARC-blessed ...) names. We (me, Jay, Ved=ARC) spent a lot of time (several meetings with ARC) making 'itype' names correctly named and correctly explained. I assume 'extracted spec' will inherit this. Thanks, Robert

…

________________________________ From: Robertson, Iain ***@***.******@***.***>> Sent: Wednesday, September 4, 2024 9:50 AM To: Robert Chyla ***@***.******@***.***>>; Beeman Strong ***@***.******@***.***>>; ***@***.******@***.***> ***@***.******@***.***>> Cc: riscv-non-isa/riscv-trace-spec ***@***.******@***.***>>; riscv-non-isa/riscv-trace-spec ***@***.******@***.***>>; Subscribed ***@***.******@***.***>> Subject: RE: [EXTERNAL]Re: [RISC-V] [sig-debug-trace-perf-mon] [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) Robert, The issue is for a request to clarify chapter 4. Whether that is done as part of an incremental update to the E-Trace spec (along with some other clarifications and corrections) before we pull out chapter 4, or whether chapter 4 gets pulled out first and the clarifications are handled as part of that makes no difference – the important thing is that we have a record of the need for clarification. Iain From: Robert Chyla ***@***.******@***.***>> Sent: 04 September 2024 17:32 To: Beeman Strong ***@***.******@***.***>>; ***@***.******@***.***>; Robertson, Iain (DI SW ICS DDCP TST RD EAH) ***@***.******@***.***>> Cc: riscv-non-isa/riscv-trace-spec ***@***.******@***.***>>; riscv-non-isa/riscv-trace-spec ***@***.******@***.***>>; Subscribed ***@***.******@***.***>> Subject: Re: [EXTERNAL]Re: [RISC-V] [sig-debug-trace-perf-mon] [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) It was also improved in N-Trace. See this: https://github.com/riscv-non-isa/tg-nexus-trace/blob/main/docs/RISC-V-N-Trace.adoc#2-trace-ingress-port IMPORTANT: I am against that PR mentioned below. We agreed to EXTRACT this into SINGLE source (as fast-track). Thanks, Robert

________________________________ From: ***@***.******@***.***> ***@***.******@***.***>> on behalf of Mr Iain Robertson via lists.riscv.org ***@***.******@***.***>> Sent: Wednesday, September 4, 2024 8:40 AM To: Beeman Strong ***@***.******@***.***>>; ***@***.******@***.***> ***@***.******@***.***>> Cc: riscv-non-isa/riscv-trace-spec ***@***.******@***.***>>; riscv-non-isa/riscv-trace-spec ***@***.******@***.***>>; Subscribed ***@***.******@***.***>> Subject: [EXTERNAL]Re: [RISC-V] [sig-debug-trace-perf-mon] [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) Beeman, Thanks – that looks like the terminology we should adopt. Unfortunately your screen-grab doesn’t show up in the GIT issue – can you paste plain text or include a reference that I can follow later please? Iain From: Beeman Strong ***@***.******@***.***>> Sent: 04 September 2024 15:45 To: ***@***.******@***.***>; Robertson, Iain (DI SW ICS DDCP TST RD EAH) ***@***.******@***.***>> Cc: riscv-non-isa/riscv-trace-spec ***@***.******@***.***>>; riscv-non-isa/riscv-trace-spec ***@***.******@***.***>>; Subscribed ***@***.******@***.***>> Subject: Re: [RISC-V] [sig-debug-trace-perf-mon] [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) FWIW, when CTR went through ARC review, ARC requested the following definitions for the various control transfer types: ***@***.*** On Wed, Sep 4, 2024 at 1:29 AM Mr Iain Robertson via lists.riscv.org<http://lists.riscv.org/> ***@***.******@***.***>> wrote: Rishiyur, That’s a good observation, and does indeed appear contradictory. It came about as a result of an earlier clarification attempt. The list of optional types in section 4.1 corresponds to the more detailed classifications in section 4.11. Originally, this was broken down into Calls, Tail-calls, Returns, Co-routine swap and Other. However, there is no formal definition of tail-call in the RISC-V ISA, and so references to ‘tail call’ in both section 4.1.1 and 4.1 were replaced with ‘jump’. With hindsight, this is too general, and probably needs to be something like “jump with discarded link address” or “call with discarded link address” Please open an issue against the spec (https://github.com/riscv-non-isa/riscv-trace-spec/issues) and I’ll address this in the next update. Iain From: Rishiyur S. Nikhil ***@***.******@***.***>> Sent: 03 September 2024 18:32 To: riscv-non-isa/riscv-trace-spec ***@***.******@***.***>> Cc: Subscribed ***@***.******@***.***>> Subject: Re: [riscv-non-isa/riscv-trace-spec] Small suggested clarifications (Issue #138) In Section 4.1, we have: The following information is mandatory: ... * The instruction_type of retired instructions for: * Jumps with a target that cannot be inferred from the source code; ... and The following information is optional: ... * The instruction_type of instructions for: ... * Jumps with a target that cannot be inferred from the source code; This seems contradictory (both mandatory and optional)? — Reply to this email directly, view it on GitHub<#138 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ALQOPSV4KFRG7M7M4S5F2K3ZUXW7VAVCNFSM6AAAAABNQZL5VGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGMRXGA3TAMJUHE>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.******@***.***>> _._,_._,_

________________________________ Links: You receive all messages sent to this group. View/Reply Online (#610)<https://lists.riscv.org/g/sig-debug-trace-perf-mon/message/610> | Reply To ***@***.***?subject=Re:%20Re%3A%20%5BRISC-V%5D%20%5Bsig-debug-trace-perf-mon%5D%20%5Briscv-non-isa%2Friscv-trace-spec%5D%20Small%20suggested%20clarifications%20%28Issue%20%23138%29> | Reply To ***@***.***?subject=Private:%20Re:%20Re%3A%20%5BRISC-V%5D%20%5Bsig-debug-trace-perf-mon%5D%20%5Briscv-non-isa%2Friscv-trace-spec%5D%20Small%20suggested%20clarifications%20%28Issue%20%23138%29> | Mute This Topic<https://lists.riscv.org/mt/108261477/1782486> | New Topic<https://lists.riscv.org/g/sig-debug-trace-perf-mon/post> Your Subscription<https://lists.riscv.org/g/sig-debug-trace-perf-mon/editsub/1782486> | Contact Group ***@***.***> | Unsubscribe<https://lists.riscv.org/g/sig-debug-trace-perf-mon/unsub> ***@***.*** _._,_._,_

[IainCRobertson]

@rsnikhil please review #161 - does this address all of your concerns?
(Note that some parts of this are addressed by Issue #142 , for which there is a separate pull request)