read-only-zero case when CSR disabled - misa.yaml file #918

jevinjojo · 2025-07-22T12:27:58Z

Issue

Fixes MISA can be read-only-0 #576

Proposed Changes

All CSR fields always return 0 when misa CSR = false
writes Ignored when CSR disabled

jevinjojo · 2025-07-22T12:29:06Z

@ThinkOpenly @dhower-qc Please take a look

dhower-qc

Thanks for the contribution! Just a few small nits

dhower-qc · 2025-07-22T12:53:03Z

spec/std/isa/csr/misa.yaml

    type: RO
    reset_value(): |
-      return (MXLEN == 32) ? 2'b01 : 2'b10;
+      return (!MISA_CSR_IMPLEMENTED) ? 0 : (MXLEN == 32) ? 2'b01 : 2'b10;


I would suggest either adding parenthesis or converting this to if (MISA_CSR_IMPLEMENTED) { .. } else { ... } for clarity.

(applies many times in this PR)

should we also convert those existing ternary conditions to if-else blocks.

should we also convert those existing ternary conditions to if-else blocks.

leave the existing ternaries, please.

dhower-qc · 2025-07-22T12:54:32Z

spec/std/isa/csr/misa.yaml

      .. Turn off each present bit invidually and try affected behaviors
      . Check
-      .. Fail unless turning off bit disables extension as expected
+      .. Fail unless turning off bit disables extension as expected


You'll need a newline at the end for regressions to pass

dhower-qc · 2025-07-22T12:56:02Z

Partially addresses #576 ; still leaves checks of CSR[misa].extension without checking for read-only-0 status.

codecov · 2025-07-22T13:02:25Z

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 43.30%. Comparing base (161b320) to head (f857538).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #918   +/-   ##
=======================================
  Coverage   43.30%   43.30%           
=======================================
  Files          10       10           
  Lines        4787     4787           
  Branches     1298     1298           
=======================================
  Hits         2073     2073           
  Misses       2714     2714

Flag	Coverage Δ
idlc	`43.30% <ø> (ø)`

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

jevinjojo · 2025-07-23T16:07:37Z

Partially addresses #576 ; still leaves checks of CSR[misa].extension without checking for read-only-0 status.

@dhower-qc @ThinkOpenly Shall I go ahead with misa field update, so this can complete the remaining work

ThinkOpenly · 2025-07-24T03:27:55Z

Partially addresses #576 ; still leaves checks of CSR[misa].extension without checking for read-only-0 status.

@dhower-qc @ThinkOpenly Shall I go ahead with misa field update, so this can complete the remaining work

$ grep 'CSR\[misa\]' $(find spec/std/isa -type f) | wc -l
558

@jevinjojo, you mean you don't want to change all 558 occurences where misa is read? 😉

@dhower-qc, I presume the tedious fix would be to change all of them from:

CSR[misa].X == value

to something like:

(!MISA_CSR_IMPLEMENTED) || (CSR[misa].X == value)

Basically, let everything "pass" if misa is read-only-0, right?

Do we want a more maintainable implementation? I can't think of a great one offhand.

jevinjojo · 2025-07-24T08:20:51Z

@ThinkOpenly @dhower-qc No 😄, I was referring the phase 3 from our issue breakdown, updating the CSR files

ThinkOpenly · 2025-07-27T02:09:29Z

No 😄, I was referring the phase 3 from our issue breakdown, updating the CSR files

Phase 3 is updating the 558 references. :-)

I suggest you address the existing comments in the content you've already submitted first, then decide if you wish to proceed with phase 3 or just open an issue to have that addressed later.

(If you do wish to proceed, we should discuss the best approach first.)

jevinjojo · 2025-07-28T16:06:43Z

@ThinkOpenly @dhower-qc added the guards to CSR[misa] reference in the core specs and CSR yaml files, also updated the ruby code-gen template. take a look

ThinkOpenly

I think this is very close to complete. Thank you for your efforts.

I suggest a few ways to make the changes simpler.

ThinkOpenly · 2025-07-30T22:46:11Z

spec/std/isa/csr/cycle.yaml

    }
  } else if (mode() == PrivilegeMode::U) {
-    if (CSR[misa].S == 1'b1) {
+    if ((!MISA_CSR_IMPLEMENTED) || ((!MISA_CSR_IMPLEMENTED) || (CSR[misa].S == 1'b1))) {


This seems redundant. Can it be simplified to:

Suggested change

if ((!MISA_CSR_IMPLEMENTED) || ((!MISA_CSR_IMPLEMENTED) || (CSR[misa].S == 1'b1))) {

if (!MISA_CSR_IMPLEMENTED || CSR[misa].S == 1'b1) {

Here and elsewhere, of course.

Almost. We can't just give a 'pass' when misa is read-only-0. In that case, we have to check 'implemented?(ExtensionName::NAME)'. When misa is not read-only-0, having the bit set implies the extension is implemented (and has not been dynamically disabled). When misa is read-only-0, the misa bit does not give any information on the extension.

Ah. Understood. If misa is not implemented, don't just avoid it, grab the setting from the configuration. So, for this instance:

Suggested change

if ((!MISA_CSR_IMPLEMENTED) || ((!MISA_CSR_IMPLEMENTED) || (CSR[misa].S == 1'b1))) {

if ((!MISA_CSR_IMPLEMENTED && implemented?(ExtensionName::S) || ((!MISA_CSR_IMPLEMENTED) || (CSR[misa].S == 1'b1))) {

ThinkOpenly · 2025-07-30T22:52:23Z

spec/std/isa/csr/misa.yaml

+      if (MISA_CSR_IMPLEMENTED) {
+        return (implemented?(ExtensionName::A) && MUTABLE_MISA_A) ? CsrFieldType::RW : CsrFieldType::RO;
+      } else {
+        return CsrFieldType::RO;
+      }


Could you use the slightly simpler implementation below, here and elsewhere?

Suggested change

if (MISA_CSR_IMPLEMENTED) {

return (implemented?(ExtensionName::A) && MUTABLE_MISA_A) ? CsrFieldType::RW : CsrFieldType::RO;

} else {

return CsrFieldType::RO;

}

if (MISA_CSR_IMPLEMENTED && implemented?(ExtensionName::A) && MUTABLE_MISA_A) {

return CsrFieldType::RW;

} else {

return CsrFieldType::RO;

}

ThinkOpenly · 2025-07-30T22:55:40Z

spec/std/isa/csr/misa.yaml

+      if (MISA_CSR_IMPLEMENTED) {
+        return 1;
+      } else {
+        return 0;
+      }


This is reasonably replaced with a ternary.

Suggested change

if (MISA_CSR_IMPLEMENTED) {

return 1;

} else {

return 0;

}

return (MISA_CSR_IMPLEMENTED) ? 1 : 0;

ThinkOpenly · 2025-07-30T22:58:28Z

spec/std/isa/csr/misa.yaml

+  if (MISA_CSR_IMPLEMENTED) {
+    return (
+      (CSR[misa].MXL << (xlen() - 2)) |
+      (CSR[misa].V << 21) |
+      (CSR[misa].U << 20) |
+      (CSR[misa].S << 18) |
+      (CSR[misa].M << 12) |
+      (CSR[misa].I << 8) |
+      (CSR[misa].H << 7) |
+      ((CSR[misa].A & CSR[misa].M & CSR[misa].F & CSR[misa].D) << 6) | # 'G'
+      (CSR[misa].F << 5) |
+      (CSR[misa].D << 3) |
+      (CSR[misa].C << 2) |
+      (CSR[misa].B << 1) |
+      CSR[misa].A);
+  } else {
+    return 0;
+  }


There's no need to change the indentation here if you insert a test/return at the beginning:

Suggested change

if (MISA_CSR_IMPLEMENTED) {

return (

(CSR[misa].MXL << (xlen() - 2)) |

(CSR[misa].V << 21) |

(CSR[misa].U << 20) |

(CSR[misa].S << 18) |

(CSR[misa].M << 12) |

(CSR[misa].I << 8) |

(CSR[misa].H << 7) |

((CSR[misa].A & CSR[misa].M & CSR[misa].F & CSR[misa].D) << 6) | # 'G'

(CSR[misa].F << 5) |

(CSR[misa].D << 3) |

(CSR[misa].C << 2) |

(CSR[misa].B << 1) |

CSR[misa].A);

} else {

return 0;

}

if (!MISA_CSR_IMPLEMENTED) {

return 0;

}

return (

(CSR[misa].MXL << (xlen() - 2)) |

(CSR[misa].V << 21) |

(CSR[misa].U << 20) |

(CSR[misa].S << 18) |

(CSR[misa].M << 12) |

(CSR[misa].I << 8) |

(CSR[misa].H << 7) |

((CSR[misa].A & CSR[misa].M & CSR[misa].F & CSR[misa].D) << 6) | # 'G'

(CSR[misa].F << 5) |

(CSR[misa].D << 3) |

(CSR[misa].C << 2) |

(CSR[misa].B << 1) |

CSR[misa].A);

ThinkOpenly · 2025-07-30T23:02:32Z

spec/std/isa/ext/Sm.yaml

+  MISA_READ_ONLY_ZERO:
+    description: |
+      When true, indicates the `misa` CSR is hardwired to read-only zero.
+      When false, the `misa` CSR behaves as a standard WARL register.
+      Only applicable when MISA_CSR_IMPLEMENTED == false.
+    schema:
+      type: boolean
+      default: false


I don't think this is needed, as it is redundant with MISA_CSR_IMPLEMENTED.

ThinkOpenly · 2025-07-30T23:06:03Z

tools/ruby-gems/udb/lib/udb/obj/csr.rb

    case @data["length"]
    when "MXLEN"
-      "CSR[misa].MXL == 0"
+      "(!MISA_CSR_IMPLEMENTED) || (CSR[misa].MXL == 0)"


If misa is not implemented, it is read-only-0. Since this is just testing against zero, you need not test to see if it's implemented.

Same comment applies anywhere the test is with zero (equal or not equal).

jevinjojo · 2025-08-02T07:13:54Z

@ThinkOpenly @dhower-qc I think it's almost done :)

ThinkOpenly · 2025-08-02T12:01:22Z

spec/std/isa/csr/misa.yaml

-      if (csr_value.F == 0 && csr_value.D == 1) {
-        return UNDEFINED_LEGAL_DETERMINISTIC;
-      }
+      if (MISA_CSR_IMPLEMENTED) {


@dhower-qc, is this test needed? In the !MISA_CSR_IMPLEMENTED case, all fields are marked read-only, so that means we never call the sw_write() method, is that correct?

I think we need a little flow chart as to how "writable" is checked. I suspect something like:

"writable" attribute for the CSR

"type"/"type()" attribute/method for the field

"legal?()" method

"sw_write()" method logic

If that's reasonable accurate, then we won't get to the "sw_write()" method in the read-only-0 case, because the write is prevented by the "type()" method. So, no special code is needed here.

@dhower-qc, is this test needed? In the !MISA_CSR_IMPLEMENTED case, all fields are marked read-only, so that means we never call the sw_write() method, is that correct?

correct

I think we need a little flow chart as to how "writable" is checked. I suspect something like:

"writable" attribute for the CSR

"type"/"type()" attribute/method for the field

"legal?()" method

"sw_write()" method logic

If that's reasonable accurate, then we won't get to the "sw_write()" method in the read-only-0 case, because the write is prevented by the "type()" method. So, no special code is needed here.

isn't actually used anywhere. we should remove it. Its purpose is handled through sw_write/sw_read, which together try to define what happens if the write is illegal.

Otherwise, the order is correct (1-2-4)

ThinkOpenly · 2025-08-02T12:04:25Z

spec/std/isa/csr/misa.yaml

+      }
    legal?(csr_value): |
-      return !(csr_value.Q == 1 && csr_value.D == 0);
+      if (MISA_CSR_IMPLEMENTED) {


Similar question to above, is this test needed in the !MISA_CSR_IMPLEMENTED case? (Do we ever need to test for legal values of a CSR if that CSR is read-only?)

When MISA_CSR_IMPLEMENTED is false, the CSR is read-only-0, so no writes can actually occur. any write value is legal?

I want to assert that in the read-only-0 case, we don't need to test for legal values because no values will ever be accepted/written. I don't think you need any new code here.

As I said above, let's just remove the legal? field

ThinkOpenly · 2025-08-02T12:11:23Z

spec/std/isa/csr/misa.yaml

      Indicates support for the `V` (vector) extension.

-      [when,"MUTABLE_MISA_V == true"]
+      [when,"MUTABLE_MISA_V == true && MISA_CSR_IMPLEMENTED == true"]


I imagine we don't need this, since if any bits of the CSR are mutable, the CSR must be implemented. I'm not sure if we enforce that, though, which we should. To do that, we probably need some "EXTRA_VALIDATION" in Sm.yaml somewhere to check if any MUTABLE parameters are true when IMPLEMENTED is false. Is that something you want to include here, or shall we open a new issue?

in the Sm.yaml file an extra_validation check under MISA_CSR_IMPLEMENTED and validation verifies that if any MISA bit (A-Z) is also need to marked as mutable. do we need to add error message if the condition violated

in the Sm.yaml file an extra_validation check under MISA_CSR_IMPLEMENTED and validation verifies that if any MISA bit (A-Z) is also need to marked as mutable.

I'm leaning toward adding the extra_validation in the extensions where the mutability parameter is defined. For example, in spec/std/isa/ext/A.yaml, under MUTABLE_MISA_A add something like:

extra_validation: assert MISA_CSR_IMPLEMENTED == true

(untested). In that way, "everybody" only needs to know that one "implemented" parameter, rather than Sm.yaml having to know every other "mutability" parameter.

do we need to add error message if the condition violated

No, that will come automatically when the architecture is resolved.

ThinkOpenly · 2025-08-02T13:14:35Z

spec/std/isa/ext/Sm.yaml

-  MTVEC_MODES:
+      default: true
+MTVEC_MODES:


Is this related? I think this should be removed from this PR.

spec/std/isa/ext/Zhinx.yaml

ThinkOpenly · 2025-08-02T13:19:46Z

spec/std/isa/isa/fp.idl

  }
  body {
-    if (MUTABLE_MISA_F && CSR[misa].F == 0) {
+    if (MUTABLE_MISA_F && ((!MISA_CSR_IMPLEMENTED && !implemented?(ExtensionName::F)) || (CSR[misa].F == 0))) {


This change is unnecessary. If the F bit is mutable, then the CSR is implemented.

ThinkOpenly · 2025-08-02T13:23:00Z

spec/std/isa/isa/globals.isa

      CSR[mcause].INT = 1'b0;
      CSR[mcause].CODE = $bits(exception_code);
-      if (CSR[misa].H == 1) {
+      if ((!MISA_CSR_IMPLEMENTED) || (CSR[misa].H == 1)) {


This needs the longer test:

Suggested change

if ((!MISA_CSR_IMPLEMENTED) || (CSR[misa].H == 1)) {

if ((!MISA_CSR_IMPLEMENTED && implemented?(ExtensionName::H)) || (CSR[misa].H == 1)) {

ThinkOpenly · 2025-08-02T13:23:41Z

spec/std/isa/isa/globals.isa

      }
      CSR[mstatus].MPP = $bits(from_mode);
-    } else if (CSR[misa].S == 1 && (handling_mode == PrivilegeMode::S)) {
+    } else if ((!MISA_CSR_IMPLEMENTED) || (CSR[misa].S == 1 )&& (handling_mode == PrivilegeMode::S)) {


This needs the longer test, too.

ThinkOpenly · 2025-08-02T13:23:56Z

spec/std/isa/isa/globals.isa

      CSR[scause].CODE = $bits(exception_code);
      CSR[mstatus].SPP = $bits(from_mode)[0];
-      if (CSR[misa].H == 1) {
+      if ((!MISA_CSR_IMPLEMENTED) || (CSR[misa].H == 1)) {


This needs the longer test, too.

ThinkOpenly · 2025-08-02T13:24:37Z

spec/std/isa/isa/globals.isa

        }
      }
-    } else if (CSR[misa].H == 1 && (handling_mode == PrivilegeMode::VS)) {
+    } else if ((!MISA_CSR_IMPLEMENTED) || (CSR[misa].H == 1 )&& (handling_mode == PrivilegeMode::VS)) {


longer test

ThinkOpenly · 2025-08-02T13:36:35Z

spec/std/isa/isa/globals.isa

      return 32;
    } else {
      if (mode() == PrivilegeMode::M) {
-        if (CSR[misa].MXL == $bits(XRegWidth::XLEN32)) {
+        if ((!MISA_CSR_IMPLEMENTED) || (CSR[misa].MXL == $bits(XRegWidth::XLEN32))) {
          return 32;
-        } else if (CSR[misa].MXL == $bits(XRegWidth::XLEN64)) {
+        } else if ((!MISA_CSR_IMPLEMENTED) || (CSR[misa].MXL == $bits(XRegWidth::XLEN64))) {
          return 64;
        }


I'm not quite sure here, but pretty sure this needs to change. This always returns 32 if !MISA_CSR_IMPLEMENTED. @dhower-qc ?

yea, this can just check that the MXLEN parameter is 32/64. even when misa is implemented, MXL is read-only (this was a backward-looking, incompatible change made to the spec)

ThinkOpenly · 2025-08-02T13:38:58Z

spec/std/isa/isa/globals.isa

    # with the use of mstatus.MPRV
    if (mode() == PrivilegeMode::M) {
-      if (CSR[misa].U == 1 && CSR[mstatus].MPRV == 1) {
+      if ((!MISA_CSR_IMPLEMENTED) || (CSR[misa].U == 1 )&& CSR[mstatus].MPRV == 1) {


this needs a longer test, and be careful with how the logic is grouped. Using A || B && C could be confusing (and seems to be incorrect here).

ThinkOpenly · 2025-08-02T13:43:19Z

spec/std/isa/isa/globals.isa

+      if ((!MISA_CSR_IMPLEMENTED) || (CSR[misa].U == 1 )&& CSR[mstatus].MPRV == 1) {
        if (CSR[mstatus].MPP == 0b00) {
-          if (CSR[misa].H == 1 && mpv() == 0b1) {
+          if ((!MISA_CSR_IMPLEMENTED) || (CSR[misa].H == 1 )&& mpv() == 0b1) {


ThinkOpenly · 2025-08-02T13:43:38Z

spec/std/isa/isa/globals.isa

+        } else if ((!MISA_CSR_IMPLEMENTED) || (CSR[misa].S == 1 )&& CSR[mstatus].MPP == 0b01) {
+          if ((!MISA_CSR_IMPLEMENTED) || (CSR[misa].H == 1 )&& mpv() == 0b1) {


same... there are more cases below where you are just checking for MISA_CSR_IMPLEMENTED, but you also need to check implemented?..., OK?

ThinkOpenly

I submitted a bunch of "single comments", above, rather than a single review. Hopefully it all makes sense.

dhower-qc · 2025-08-08T15:37:41Z

Keep up the good work!

jevinjojo · 2025-08-10T08:50:12Z

@ThinkOpenly @dhower-qc did the changes, can you verify if I missed anything.

ThinkOpenly

Some minor, but important changes.

spec/std/isa/ext/A.yaml

spec/std/isa/ext/B.yaml

spec/std/isa/ext/C.yaml

spec/std/isa/ext/Sm.yaml

spec/std/isa/ext/Zhinx.yaml

spec/std/isa/isa/globals.isa

ThinkOpenly · 2025-08-12T15:36:34Z

spec/std/isa/isa/interrupts.idl

-          CSR[mstatush].MPV = $bits(mode())[2];
-        }
+      if ((!MISA_CSR_IMPLEMENTED && implemented?(ExtensionName::H)) || (CSR[misa].H == 1'b1)) {
+        CSR[mstatus].MPV = $bits(mode())[2];


The original code can't be simplified here. Notice that the "then" block is using mstatus and the else block is using mstatush, so you'll need to revert that part of this change, but keep the "MISA_CSR_IMPLEMENTED" change.

Signed-off-by: Jevin Jojo <[email protected]>

jevinjojo · 2025-08-17T15:42:43Z

@ThinkOpenly @dhower-qc Any changes required.

dhower-qc

Great work. Still a bit more to do :) Check for references in spec/**/inst/*.yaml for misa; there should be a lot.

dhower-qc · 2025-08-20T21:21:28Z

spec/std/isa/csr/misa.yaml

+  Reports the XLEN and "major" extensions supported by the ISA.
+  [when,"MISA_CSR_IMPLEMENTED == false"] 
+  This CSR is read-only-0 (all bits hardwired to zero) and cannot be written.


We've update how this is handled:

Suggested change

Reports the XLEN and "major" extensions supported by the ISA.

[when,"MISA_CSR_IMPLEMENTED == false"]

This CSR is read-only-0 (all bits hardwired to zero) and cannot be written.

- id: csr-misa-purpose

text: |

Reports the XLEN and "major" extensions supported by the ISA.

when:

param:

name: MISA_CSR_IMPLEMENTED

value: true

- id: csr-misa-unimplemented

text: |

This CSR is read-only-0 (all bits hardwired to zero) and cannot be written.

when:

param:

name: MISA_CSR_IMPLEMENTED

value: false

Note that the schema will reject this until #891 ships

dhower-qc · 2025-08-20T21:26:24Z

spec/std/isa/ext/A.yaml

      the extension can be disabled in the `misa.A` bit.
    schema:
      type: boolean
+    extra_validation: assert MISA_CSR_IMPLEMENTED == true


MISA_CSR_IMPLEMENTED only needs to be true if MUTABLE_MISA_A is true:

Suggested change

extra_validation: assert MISA_CSR_IMPLEMENTED == true

extra_validation: assert MISA_CSR_IMPLEMENTED == true if MUTABLE_MISA_A

dhower-qc · 2025-08-20T21:27:29Z

spec/std/isa/ext/Zhinx.yaml

still here?

I checked with git diff to see if there line ending/permission changes, also tried resetting it. but didn't work. I'm not sure can you tell me what might be causing this?

dhower-qc · 2025-08-20T21:28:22Z

spec/std/isa/isa/globals.isa

-    } else {
-      if (mode() == PrivilegeMode::M) {
-        if (CSR[misa].MXL == $bits(XRegWidth::XLEN32)) {
-          return 32;
-        } else if (CSR[misa].MXL == $bits(XRegWidth::XLEN64)) {
-          return 64;
-        }
-      } else if (implemented?(ExtensionName::S) && mode() == PrivilegeMode::S) {
-        if (CSR[mstatus].SXL == $bits(XRegWidth::XLEN32)) {
-          return 32;
-        } else if (CSR[mstatus].SXL == $bits(XRegWidth::XLEN64)) {
-          return 64;
-        }
-      } else if (implemented?(ExtensionName::U) && mode() == PrivilegeMode::U) {
-        if (CSR[mstatus].UXL == $bits(XRegWidth::XLEN32)) {
-          return 32;
-        } else if (CSR[mstatus].UXL == $bits(XRegWidth::XLEN64)) {
-          return 64;
-        }
-      } else if (implemented?(ExtensionName::H) && mode() == PrivilegeMode::VS) {
-        if (CSR[hstatus].VSXL == $bits(XRegWidth::XLEN32)) {
-          return 32;
-        } else if (CSR[hstatus].VSXL == $bits(XRegWidth::XLEN64)) {
-          return 64;
-        }
-      } else if (implemented?(ExtensionName::H) && mode() == PrivilegeMode::VU) {
-        if (CSR[vsstatus].UXL == $bits(XRegWidth::XLEN32)) {
-          return 32;
-        } else if (CSR[vsstatus].UXL == $bits(XRegWidth::XLEN64)) {
-          return 64;
-        }
+    } else if (mode() == PrivilegeMode::M) {
+      return MXLEN;
+    } else if (implemented?(ExtensionName::S) && mode() == PrivilegeMode::S) {
+      if (CSR[mstatus].SXL == $bits(XRegWidth::XLEN32)) {
+        return 32;
+      } else if (CSR[mstatus].SXL == $bits(XRegWidth::XLEN64)) {
+        return 64;
+      }
+    } else if (implemented?(ExtensionName::U) && mode() == PrivilegeMode::U) {
+      if (CSR[mstatus].UXL == $bits(XRegWidth::XLEN32)) {
+        return 32;
+      } else if (CSR[mstatus].UXL == $bits(XRegWidth::XLEN64)) {
+        return 64;
+      }
+    } else if (implemented?(ExtensionName::H) && mode() == PrivilegeMode::VS) {
+      if (CSR[hstatus].VSXL == $bits(XRegWidth::XLEN32)) {
+        return 32;
+      } else if (CSR[hstatus].VSXL == $bits(XRegWidth::XLEN64)) {
+        return 64;
+      }
+    } else if (implemented?(ExtensionName::H) && mode() == PrivilegeMode::VU) {
+      if (CSR[vsstatus].UXL == $bits(XRegWidth::XLEN32)) {
+        return 32;
+      } else if (CSR[vsstatus].UXL == $bits(XRegWidth::XLEN64)) {
+        return 64;


Why is this diff here?

dhower-qc · 2025-08-20T21:28:32Z

spec/std/isa/isa/globals.isa

+        return 64;
      }
    }
+


dhower-qc · 2025-08-20T21:30:41Z

tools/ruby-gems/udb/lib/udb/obj/csr.rb

    case @data["length"]
    when "MXLEN"
-      "CSR[misa].MXL == 1"
+      "(!MISA_CSR_IMPLEMENTED) || (CSR[misa].MXL == 1)"


Suggested change

"(!MISA_CSR_IMPLEMENTED) || (CSR[misa].MXL == 1)"

"MXLEN == 64"

What you wrote isn't quite right, since it doesn't tell you what MXLEN is when misa isn't implemented. MXLEN can never change, so we don't need to consult misa at all here

jevinjojo · 2025-09-04T02:13:10Z

@dhower-qc @ThinkOpenly Will do the changes in a few days.

Signed-off-by: Jevin Jojo <[email protected]>

ThinkOpenly · 2025-09-17T21:01:51Z

Should we expect additional changes here?

jevinjojo added 2 commits July 22, 2025 12:07

read-only-zero when MISA_CSR_IMPLEMENTED=false

b1a2a4d

read-only-zero when MISA_CSR_IMPLEMENTED=false

f857538

jevinjojo requested review from ThinkOpenly and dhower-qc as code owners July 22, 2025 12:27

dhower-qc requested changes Jul 22, 2025

View reviewed changes

guard CSR[misa] checks

a176914

jevinjojo force-pushed the misa-readonly-zero branch from e57c2d6 to a176914 Compare July 28, 2025 16:00

ThinkOpenly requested changes Jul 30, 2025

View reviewed changes

latest

7ae01d6

jevinjojo force-pushed the misa-readonly-zero branch from aa4ab1f to 7ae01d6 Compare August 2, 2025 07:11

ThinkOpenly reviewed Aug 2, 2025

View reviewed changes

spec/std/isa/ext/Zhinx.yaml Show resolved Hide resolved

ThinkOpenly reviewed Aug 2, 2025

View reviewed changes

ThinkOpenly requested changes Aug 2, 2025

View reviewed changes

cleanup

15c0c7a

jevinjojo force-pushed the misa-readonly-zero branch from fa3c6e5 to 15c0c7a Compare August 10, 2025 08:46

ThinkOpenly requested changes Aug 12, 2025

View reviewed changes

indentations

1fbc3e7

jevinjojo force-pushed the misa-readonly-zero branch from 9fed2ef to 1fbc3e7 Compare August 16, 2025 08:54

Merge branch 'main' into misa-readonly-zero

37cf99a

Signed-off-by: Jevin Jojo <[email protected]>

jevinjojo requested review from ThinkOpenly and dhower-qc August 16, 2025 08:57

Merge branch 'main' into misa-readonly-zero

4ede842

dhower-qc requested changes Aug 20, 2025

View reviewed changes

Merge branch 'main' into misa-readonly-zero

0bf0534

Signed-off-by: Jevin Jojo <[email protected]>

	if ((!MISA_CSR_IMPLEMENTED) \|\| ((!MISA_CSR_IMPLEMENTED) \|\| (CSR[misa].S == 1'b1))) {
	if (!MISA_CSR_IMPLEMENTED \|\| CSR[misa].S == 1'b1) {

	if ((!MISA_CSR_IMPLEMENTED) \|\| ((!MISA_CSR_IMPLEMENTED) \|\| (CSR[misa].S == 1'b1))) {
	if ((!MISA_CSR_IMPLEMENTED && implemented?(ExtensionName::S) \|\| ((!MISA_CSR_IMPLEMENTED) \|\| (CSR[misa].S == 1'b1))) {

	if ((!MISA_CSR_IMPLEMENTED) \|\| (CSR[misa].H == 1)) {
	if ((!MISA_CSR_IMPLEMENTED && implemented?(ExtensionName::H)) \|\| (CSR[misa].H == 1)) {

		} else if ((!MISA_CSR_IMPLEMENTED) \|\| (CSR[misa].S == 1 )&& CSR[mstatus].MPP == 0b01) {
		if ((!MISA_CSR_IMPLEMENTED) \|\| (CSR[misa].H == 1 )&& mpv() == 0b1) {

-  Reports the XLEN and "major" extensions supported by the ISA.
-  [when,"MISA_CSR_IMPLEMENTED == false"]
-  This CSR is read-only-0 (all bits hardwired to zero) and cannot be written.
+  - id: csr-misa-purpose
+    text: |
+      Reports the XLEN and "major" extensions supported by the ISA.
+    when:
+      param:
+        name: MISA_CSR_IMPLEMENTED
+        value: true
+  - id: csr-misa-unimplemented
+    text: |
+      This CSR is read-only-0 (all bits hardwired to zero) and cannot be written.
+    when:
+      param:
+        name: MISA_CSR_IMPLEMENTED
+        value: false

	extra_validation: assert MISA_CSR_IMPLEMENTED == true
	extra_validation: assert MISA_CSR_IMPLEMENTED == true if MUTABLE_MISA_A

	"(!MISA_CSR_IMPLEMENTED) \|\| (CSR[misa].MXL == 1)"
	"MXLEN == 64"

read-only-zero case when CSR disabled - misa.yaml file #918

Are you sure you want to change the base?

read-only-zero case when CSR disabled - misa.yaml file #918

Uh oh!

Conversation

jevinjojo commented Jul 22, 2025

Issue

Proposed Changes

Uh oh!

jevinjojo commented Jul 22, 2025

Uh oh!

dhower-qc left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

dhower-qc commented Jul 22, 2025

Uh oh!

codecov bot commented Jul 22, 2025

Codecov Report

Uh oh!

jevinjojo commented Jul 23, 2025

Uh oh!

ThinkOpenly commented Jul 24, 2025

Uh oh!

jevinjojo commented Jul 24, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ThinkOpenly commented Jul 27, 2025

Uh oh!

jevinjojo commented Jul 28, 2025

Uh oh!

ThinkOpenly left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jevinjojo commented Aug 2, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jevinjojo Aug 4, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

jevinjojo commented Jul 24, 2025 •

edited

Loading

jevinjojo Aug 4, 2025 •

edited

Loading

ThinkOpenly Aug 2, 2025 •

edited

Loading