[v5,00/45] target/arm: Scalable Matrix Extension

Changes for v5:
  * Use macros for vertical tile slice addressing.
  * Other misc adjustments per review.

Patches without r-b:
  03-target-arm-Trap-non-streaming-usage-when-Streamin.patch
  07-target-arm-Mark-PMULL-FMMLA-as-non-streaming.patch
  19-target-arm-Implement-SME-MOVA.patch
  20-target-arm-Implement-SME-LD1-ST1.patch
  23-target-arm-Implement-SME-ADDHA-ADDVA.patch
  24-target-arm-Implement-FMOPA-FMOPS-non-widening.patch
  25-target-arm-Implement-BFMOPA-BFMOPS.patch
  26-target-arm-Implement-FMOPA-FMOPS-widening.patch
  35-linux-user-aarch64-Add-SM-bit-to-SVE-signal-conte.patch
  37-linux-user-aarch64-Do-not-allow-duplicate-or-shor.patch


r~


Richard Henderson (45):
  target/arm: Handle SME in aarch64_cpu_dump_state
  target/arm: Add infrastructure for disas_sme
  target/arm: Trap non-streaming usage when Streaming SVE is active
  target/arm: Mark ADR as non-streaming
  target/arm: Mark RDFFR, WRFFR, SETFFR as non-streaming
  target/arm: Mark BDEP, BEXT, BGRP, COMPACT, FEXPA, FTSSEL as
    non-streaming
  target/arm: Mark PMULL, FMMLA as non-streaming
  target/arm: Mark FTSMUL, FTMAD, FADDA as non-streaming
  target/arm: Mark SMMLA, UMMLA, USMMLA as non-streaming
  target/arm: Mark string/histo/crypto as non-streaming
  target/arm: Mark gather/scatter load/store as non-streaming
  target/arm: Mark gather prefetch as non-streaming
  target/arm: Mark LDFF1 and LDNF1 as non-streaming
  target/arm: Mark LD1RO as non-streaming
  target/arm: Add SME enablement checks
  target/arm: Handle SME in sve_access_check
  target/arm: Implement SME RDSVL, ADDSVL, ADDSPL
  target/arm: Implement SME ZERO
  target/arm: Implement SME MOVA
  target/arm: Implement SME LD1, ST1
  target/arm: Export unpredicated ld/st from translate-sve.c
  target/arm: Implement SME LDR, STR
  target/arm: Implement SME ADDHA, ADDVA
  target/arm: Implement FMOPA, FMOPS (non-widening)
  target/arm: Implement BFMOPA, BFMOPS
  target/arm: Implement FMOPA, FMOPS (widening)
  target/arm: Implement SME integer outer product
  target/arm: Implement PSEL
  target/arm: Implement REVD
  target/arm: Implement SCLAMP, UCLAMP
  target/arm: Reset streaming sve state on exception boundaries
  target/arm: Enable SME for -cpu max
  linux-user/aarch64: Clear tpidr2_el0 if CLONE_SETTLS
  linux-user/aarch64: Reset PSTATE.SM on syscalls
  linux-user/aarch64: Add SM bit to SVE signal context
  linux-user/aarch64: Tidy target_restore_sigframe error return
  linux-user/aarch64: Do not allow duplicate or short sve records
  linux-user/aarch64: Verify extra record lock succeeded
  linux-user/aarch64: Move sve record checks into restore
  linux-user/aarch64: Implement SME signal handling
  linux-user: Rename sve prctls
  linux-user/aarch64: Implement PR_SME_GET_VL, PR_SME_SET_VL
  target/arm: Only set ZEN in reset if SVE present
  target/arm: Enable SME for user-only
  linux-user/aarch64: Add SME related hwcap entries

 docs/system/arm/emulation.rst     |    4 +
 linux-user/aarch64/target_cpu.h   |    5 +-
 linux-user/aarch64/target_prctl.h |   56 +-
 target/arm/cpu.h                  |    7 +
 target/arm/helper-sme.h           |  126 ++++
 target/arm/helper-sve.h           |    4 +
 target/arm/helper.h               |   18 +
 target/arm/translate-a64.h        |   45 ++
 target/arm/translate.h            |   16 +
 target/arm/sme-fa64.decode        |   60 ++
 target/arm/sme.decode             |   88 +++
 target/arm/sve.decode             |   41 +-
 linux-user/aarch64/cpu_loop.c     |    9 +
 linux-user/aarch64/signal.c       |  243 ++++++-
 linux-user/elfload.c              |   20 +
 linux-user/syscall.c              |   28 +-
 target/arm/cpu.c                  |   35 +-
 target/arm/cpu64.c                |   11 +
 target/arm/helper.c               |   56 +-
 target/arm/sme_helper.c           | 1124 +++++++++++++++++++++++++++++
 target/arm/sve_helper.c           |   28 +
 target/arm/translate-a64.c        |  103 ++-
 target/arm/translate-sme.c        |  373 ++++++++++
 target/arm/translate-sve.c        |  393 ++++++++--
 target/arm/translate-vfp.c        |   12 +
 target/arm/translate.c            |    2 +
 target/arm/vec_helper.c           |   24 +
 target/arm/meson.build            |    3 +
 28 files changed, 2799 insertions(+), 135 deletions(-)
 create mode 100644 target/arm/sme-fa64.decode
 create mode 100644 target/arm/sme.decode
 create mode 100644 target/arm/translate-sme.c

On Wed, 6 Jul 2022 at 09:40, Richard Henderson
<richard.henderson@linaro.org> wrote:
>
> Changes for v5:
>   * Use macros for vertical tile slice addressing.
>   * Other misc adjustments per review.
>
> Patches without r-b:
>   03-target-arm-Trap-non-streaming-usage-when-Streamin.patch
>   07-target-arm-Mark-PMULL-FMMLA-as-non-streaming.patch
>   19-target-arm-Implement-SME-MOVA.patch
>   20-target-arm-Implement-SME-LD1-ST1.patch
>   23-target-arm-Implement-SME-ADDHA-ADDVA.patch
>   24-target-arm-Implement-FMOPA-FMOPS-non-widening.patch
>   25-target-arm-Implement-BFMOPA-BFMOPS.patch
>   26-target-arm-Implement-FMOPA-FMOPS-widening.patch
>   35-linux-user-aarch64-Add-SM-bit-to-SVE-signal-conte.patch
>   37-linux-user-aarch64-Do-not-allow-duplicate-or-shor.patch

Now all reviewed. There were just enough fixes needed that
I don't quite feel comfortable doing them at my end, so if you
could roll a v6 I think that would be better.

thanks
-- PMM