[PATCH v3] gdb/record: Support fense.tso, zicond, bitmanip, sinval and zihintntl in RISC-V

Fri Nov 7 01:01:01 GMT 2025

We haven't used riscv-opcodes to generate these macros for a long time in
binutils, probably because of the license issue which I forgot...  However,
if I remember correctly, these codes in opcodes/riscv-opc.c are used to
simplify the number of macros defined, so I would like to keep the existing
code.  If you are adding new instructions with new defined macros, then
it's all Okay.

Thanks
Nelson

On Thu, Nov 6, 2025 at 1:36 AM Timur Golubovich <timurgol007@gmail.com>
wrote:

> > so is it possible to apply these defined stuffs into the missing
> DECLARE_INSN macros, rather than define a bunch of extra new ones?
>
> Well, I can do this, but won't it be better to change
> MATCH_UNSHFLI|MATCH_SHAMT_ZIP_32 to MATCH_UNZIP in
> opcodes/riscv-opc.c? Because the changes that I made in this patch
> were generated by riscv-opcodes, and I absolutely don't understand why
> it was necessary to build this MATCH using |.
>
> On Tue, Nov 4, 2025 at 6:14 AM Nelson Chu <nelson@rivosinc.com> wrote:
> >
> >
> >
> > On Tue, Nov 4, 2025 at 5:37 AM Timur Golubovich <timurgol007@gmail.com>
> wrote:
> >>
> >> ---------- Forwarded message ---------
> >> From: timurgol007 <timurgol007@gmail.com>
> >> Date: Sat, Nov 1, 2025 at 2:28 PM
> >> Subject: [PATCH v3] gdb/record: Support fense.tso, zicond, bitmanip,
> >> sinval and zihintntl in RISC-V
> >> To: <tom@tromey.com>, <guinevere@redhat.com>, <aburgess@redhat.com>,
> >> <gdb-patches@sourceware.org>
> >> Cc: timurgol007 <timurgol007@gmail.com>
> >>
> >>
> >> Added record-full support for these extensions. As some opcodes were
> >> missed in riscv-opc.h, added them there using riscv-opcodes repo.
> >
> >
> > I think I can just review the riscv-opc.h parts.
> >
> >>
> >> ---
> >>  gdb/riscv-tdep.c           | 83 ++++++++++++++++++++++++++++++++++++--
> >>  include/opcode/riscv-opc.h | 43 ++++++++++++++++++++
> >>  2 files changed, 123 insertions(+), 3 deletions(-)
> >>
> >> diff --git a/include/opcode/riscv-opc.h b/include/opcode/riscv-opc.h
> >> index 1c649628390..6a5d54b1615 100644
> >> --- a/include/opcode/riscv-opc.h
> >> +++ b/include/opcode/riscv-opc.h
> >> @@ -579,6 +579,32 @@
> >>  #define MASK_RORW  0xfe00707f
> >>  #define MATCH_RORIW 0x6000501b
> >>  #define MASK_RORIW  0xfe00707f
> >> +#define MATCH_BCLRI_RV32 0x48001013
> >> +#define MASK_BCLRI_RV32 0xfe00707f
> >> +#define MATCH_BEXTI_RV32 0x48005013
> >> +#define MASK_BEXTI_RV32 0xfe00707f
> >> +#define MATCH_BINVI_RV32 0x68001013
> >> +#define MASK_BINVI_RV32 0xfe00707f
> >> +#define MATCH_BREV8 0x68705013
> >> +#define MASK_BREV8 0xfff0707f
> >> +#define MATCH_BSETI_RV32 0x28001013
> >> +#define MASK_BSETI_RV32 0xfe00707f
> >> +#define MATCH_ORC_B 0x28705013
> >> +#define MASK_ORC_B 0xfff0707f
> >> +#define MATCH_REV8 0x6b805013
> >> +#define MASK_REV8 0xfff0707f
> >> +#define MATCH_REV8_RV32 0x69805013
> >> +#define MASK_REV8_RV32 0xfff0707f
> >> +#define MATCH_RORI_RV32 0x60005013
> >> +#define MASK_RORI_RV32 0xfe00707f
> >> +#define MATCH_UNZIP 0x8f05013
> >> +#define MASK_UNZIP 0xfff0707f
> >> +#define MATCH_ZEXT_H 0x800403b
> >> +#define MASK_ZEXT_H 0xfff0707f
> >> +#define MATCH_ZEXT_H_RV32 0x8004033
> >> +#define MASK_ZEXT_H_RV32 0xfff0707f
> >> +#define MATCH_ZIP 0x8f01013
> >> +#define MASK_ZIP 0xfff0707f
> >>  #define MATCH_SH1ADD 0x20002033
> >>  #define MASK_SH1ADD  0xfe00707f
> >>  #define MATCH_SH2ADD 0x20004033
> >> @@ -2269,6 +2295,8 @@
> >>  #define MASK_C_SEXT_H 0xfc7f
> >>  #define MATCH_C_ZEXT_W 0x9c71
> >>  #define MASK_C_ZEXT_W 0xfc7f
> >> +#define MATCH_C_SEXT_W 0x2001
> >> +#define MASK_C_SEXT_W 0xf07f
> >>  #define MATCH_C_NOT 0x9c75
> >>  #define MASK_C_NOT 0xfc7f
> >>  #define MATCH_C_MUL 0x9c41
> >> @@ -4370,6 +4398,7 @@ DECLARE_INSN(sd, MATCH_SD, MASK_SD)
> >>  DECLARE_INSN(pause, MATCH_PAUSE, MASK_PAUSE)
> >>  DECLARE_INSN(fence, MATCH_FENCE, MASK_FENCE)
> >>  DECLARE_INSN(fence_i, MATCH_FENCE_I, MASK_FENCE_I)
> >> +DECLARE_INSN(fence_tso, MATCH_FENCE_TSO, MASK_FENCE_TSO)
> >>  DECLARE_INSN(mul, MATCH_MUL, MASK_MUL)
> >>  DECLARE_INSN(mulh, MATCH_MULH, MASK_MULH)
> >>  DECLARE_INSN(mulhsu, MATCH_MULHSU, MASK_MULHSU)
> >> @@ -4570,6 +4599,19 @@ DECLARE_INSN(cpopw, MATCH_CPOPW, MASK_CPOPW)
> >>  DECLARE_INSN(rolw, MATCH_ROLW, MASK_ROLW)
> >>  DECLARE_INSN(rorw, MATCH_RORW, MASK_RORW)
> >>  DECLARE_INSN(roriw, MATCH_RORIW, MASK_RORIW)
> >> +DECLARE_INSN(bclri_rv32, MATCH_BCLRI_RV32, MASK_BCLRI_RV32)
> >> +DECLARE_INSN(bexti_rv32, MATCH_BEXTI_RV32, MASK_BEXTI_RV32)
> >> +DECLARE_INSN(binvi_rv32, MATCH_BINVI_RV32, MASK_BINVI_RV32)
> >> +DECLARE_INSN(brev8, MATCH_BREV8, MASK_BREV8)
> >> +DECLARE_INSN(bseti_rv32, MATCH_BSETI_RV32, MASK_BSETI_RV32)
> >> +DECLARE_INSN(orc_b, MATCH_ORC_B, MASK_ORC_B)
> >> +DECLARE_INSN(rev8, MATCH_REV8, MASK_REV8)
> >> +DECLARE_INSN(rev8_rv32, MATCH_REV8_RV32, MASK_REV8_RV32)
> >> +DECLARE_INSN(rori_rv32, MATCH_RORI_RV32, MASK_RORI_RV32)
> >> +DECLARE_INSN(unzip, MATCH_UNZIP, MASK_UNZIP)
> >> +DECLARE_INSN(zext_h, MATCH_ZEXT_H, MASK_ZEXT_H)
> >> +DECLARE_INSN(zext_h_rv32, MATCH_ZEXT_H_RV32, MASK_ZEXT_H_RV32)
> >> +DECLARE_INSN(zip, MATCH_ZIP, MASK_ZIP)
> >>  DECLARE_INSN(sh1add, MATCH_SH1ADD, MASK_SH1ADD)
> >>  DECLARE_INSN(sh2add, MATCH_SH2ADD, MASK_SH2ADD)
> >>  DECLARE_INSN(sh3add, MATCH_SH3ADD, MASK_SH3ADD)
> >> @@ -4803,6 +4845,7 @@ DECLARE_INSN(vsm3me_vv, MATCH_VSM3ME_VV,
> MASK_VSM3ME_VV)
> >>  /* Zcb instructions.  */
> >>  DECLARE_INSN(c_sext_b, MATCH_C_SEXT_B, MASK_C_SEXT_B)
> >>  DECLARE_INSN(c_sext_h, MATCH_C_SEXT_H, MASK_C_SEXT_H)
> >> +DECLARE_INSN(c_sext_w, MATCH_C_SEXT_W, MASK_C_SEXT_W)
> >>  DECLARE_INSN(c_zext_b, MATCH_C_ZEXT_B, MASK_C_ZEXT_B)
> >>  DECLARE_INSN(c_zext_h, MATCH_C_ZEXT_H, MASK_C_ZEXT_H)
> >>  DECLARE_INSN(c_zext_w, MATCH_C_ZEXT_W, MASK_C_ZEXT_W)
> >
> >
> > I haven't check the whole instruction, but at least for something like
> unzip, according to the opcode table, it's match and mask functions are,
> > MATCH_UNSHFLI|MATCH_SHAMT_ZIP_32 and MASK_UNSHFLI|MASK_SHAMT, so is it
> possible to apply these defined stuffs into the missing DECLARE_INSN
> macros, rather than define a bunch of extra new ones?
> >
> > Thanks
> > Nelson
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://sourceware.org/pipermail/binutils/attachments/20251107/937bc987/attachment-0001.htm>