From: Andrew Carlotti <andrew.carlotti@arm.com>
To: Srinath Parvathaneni <srinath.parvathaneni@arm.com>
Cc: binutils@sourceware.org, richard.earnshaw@arm.com, nickc@redhat.com
Subject: Re: [PATCH v3 0/5] aarch64: Add support for sme2 and sve2 BFloat16 feature.
Date: Thu, 9 Jan 2025 12:11:18 +0000 [thread overview]
Message-ID: <35beba38-3479-190d-50f1-445817c221c3@e124511.cambridge.arm.com> (raw)
In-Reply-To: <20250107113415.2836224-1-srinath.parvathaneni@arm.com>
On Tue, Jan 07, 2025 at 11:34:10AM +0000, Srinath Parvathaneni wrote:
> Changes v2-v3:
>
> In 1/5:
> * Removed FEAT_SVE_SVE2_B16B16 entry from virtual feature bits and
> aarch64_virtual_dependency_table aarch64_dependencies.
> * Retained the same feature list macro (B16B16_SVE2) for FEAT_SVE_B16B16
> to avoid unnecessary changes in opcode tables.
> * Replaced OP_SVE_SMSS with OP_SVE_HMHH in B16B16_SVE2_INSNC expansions.
> * Renamed test filenames, FEAT_SVE_B16B16 (+sve2) tests are renamed with
> prefix "sve-b16b16-sve2", eg sve-b16b16-sve2-1.s.
> * Added ` ?` to the vector lists in the tests with z vectors
> e.g. `{z0\.h-z3\.h}` -> `{ ?z0\.h-z3\.h ?}`.
> * Documented +sve-b16b16 option.
>
> In 2/5
> * Removed FEAT_SVE_SME2_B16B16 entry from virtual feature bits and
> aarch64_virtual_dependency_table aarch64_dependencies.
> * Macro using the feature list (&aarch64_feature_sve_b16b16_sme2) is
> renamed to SVE_B16B16_SME.
> * F_OD(2) and F_OD(4) are dropped from SVE_B16B16_SME_INSN "bfclamp"
> instructions.
> * Renamed test filenames, FEAT_SVE_B16B16 (+sme2) tests are renamed with
> prefix "sve-b16b16-sme2", eg sve-b16b16-sme2-1.s.
> * Added ` ?` to the vector lists in the tests with z vectors
> e.g. `{z0\.h-z3\.h}` -> `{ ?z0\.h-z3\.h ?}`.
>
> In 4/5:
> * Fixed the gas/NEWS entry for feature name.
> * Documented +sme-b16b16 option.
> * Renamed test filenames, FEAT_SME_B16B16 tests are renamed with prefix
> "sme-b16b16", eg sme-b16b16-1.s.
> * For some the instructions, added tests with different index value in
> the last operands.
> * Corrected the feature list and dependencies for the option "sme-b16b16".
>
> Changes v1 -> v2:
>
> * Previously in version v1 this patch series was posted on top of master + FEAT_SME_F16F16 [3],
> but this failed to apply by the Linaro CI due to dependency on FEAT_SME_F16F16. I have
> rebased the patch series on later master (without FEAT_SME_F16F16) in version v2.
>
> version v1:
> In the current assembler, SVE2 Bfloat16 instructions are implemented with tick
> FEAT_B16B16 and command line flag "+b16b16" and this feature was suspended
> due to incomplete support.
>
> In the new spec available here(SVE[1], SME[2]), FEAT_B16B16 is replaced with
> FEAT_SVE_B16B16 and FEAT_SME_B16B16 and command line flag "+b16b16" is replaced
> with "+sve-b16b16"(+sme2 or +sve2) and "+sme-b16b16".
>
> More details about the Bfloat16 are provided below:
> The Bfloat16 feature in sve2 and sme2 is divided into 3 combinations.
> * SVE Z-targeting non-widening BFloat16 instructions under FEAT_SVE_B16B16
> implemented with command line flag "+sve-b16b16+sve2".
> * SME Z-targeting multi-vector non-widening BFloat16 instructions under
> FEAT_SVE_B16B16 implemented with command line flag "+sve-b16b16+sme2".
> * SME ZA-targeting non-widening BFloat16 instructions under FEAT_SME_B16B16
> implemented with command line flag "+sme-b16b16".
> * Added ` ?` to the vector lists in the tests with z vectors
> e.g. `{z0\.h-z3\.h}` -> `{ ?z0\.h-z3\.h ?}`.
>
> This following 5 patch series add support for above combinations and instructions:
> Srinath Parvathaneni (5):
> aarch64: Add support for FEAT_SVE_B16B16 feature.
> aarch64: Add support for FEAT_SVE_B16B16 min and max instructions.
> aarch64: Add support for FEAT_SVE_B16B16 min and max instructions (autogenerated files).
> aarch64: Add support for FEAT_SME_B16B16 feature.
> aarch64: Add support for FEAT_SME_B16B16 feature (autogenerated files).
>
> SVE[1]: https://developer.arm.com/documentation/ddi0602/2024-06/SVE-Instructions?lang=en
> SME[2]: https://developer.arm.com/documentation/ddi0602/2024-06/SME-Instructions?lang=en
> FEAT_SME_F16F16: https://sourceware.org/pipermail/binutils/2025-January/138318.html
>
> Regression testing for aarch64-none-elf target and found no regressions.
>
> Ok for binutils-master?
I think this is nearly there, thanks - just a few small issues remaining in 2/5
and 4/5.
>
> Regards,
> Srinath.
>
> Srinath Parvathaneni (5):
> aarch64: Add support for FEAT_SVE_B16B16 feature.
> aarch64: Add support for FEAT_SVE_B16B16 min and max instructions.
> aarch64: Add support for FEAT_SVE_B16B16 min and max instructions
> (autogenerated files).
> aarch64: Add support for FEAT_SME_B16B16 feature.
> aarch64: Add support for FEAT_SME_B16B16 feature (autogenerated
> files).
>
> gas/NEWS | 7 +-
> gas/config/tc-aarch64.c | 6 +-
> gas/doc/c-aarch64.texi | 6 +-
> .../gas/aarch64/bfloat16-1-invalid.d | 4 -
> .../gas/aarch64/bfloat16-2-invalid.d | 5 -
> gas/testsuite/gas/aarch64/bfloat16-bad.d | 4 -
> gas/testsuite/gas/aarch64/bfloat16-bad.l | 100 --
> gas/testsuite/gas/aarch64/sme-b16b16-1.d | 126 +++
> gas/testsuite/gas/aarch64/sme-b16b16-1.s | 143 +++
> gas/testsuite/gas/aarch64/sme-b16b16-bad-1.d | 3 +
> gas/testsuite/gas/aarch64/sme-b16b16-bad-1.l | 193 ++++
> gas/testsuite/gas/aarch64/sme-b16b16-bad-1.s | 173 ++++
> .../gas/aarch64/sve-b16b16-sme2-bad-1.d | 3 +
> .../gas/aarch64/sve-b16b16-sme2-bad-1.l | 159 ++++
> .../gas/aarch64/sve-b16b16-sme2-bad-1.s | 136 +++
> gas/testsuite/gas/aarch64/sve-b16b16-sme2.d | 122 +++
> gas/testsuite/gas/aarch64/sve-b16b16-sme2.s | 138 +++
> .../gas/aarch64/sve-b16b16-sve2-bad-1.d | 3 +
> ...16-1-invalid.l => sve-b16b16-sve2-bad-1.l} | 0
> ...16-1-invalid.s => sve-b16b16-sve2-bad-1.s} | 0
> .../gas/aarch64/sve-b16b16-sve2-bad-2.d | 3 +
> ...16-2-invalid.l => sve-b16b16-sve2-bad-2.l} | 24 +-
> ...16-2-invalid.s => sve-b16b16-sve2-bad-2.s} | 0
> .../{bfloat16-1.d => sve-b16b16-sve2.d} | 5 +-
> .../{bfloat16-1.s => sve-b16b16-sve2.s} | 0
> include/opcode/aarch64.h | 6 +-
> opcodes/aarch64-dis-2.c | 886 +++++++++++-------
> opcodes/aarch64-tbl.h | 87 +-
> 28 files changed, 1866 insertions(+), 476 deletions(-)
> delete mode 100644 gas/testsuite/gas/aarch64/bfloat16-1-invalid.d
> delete mode 100644 gas/testsuite/gas/aarch64/bfloat16-2-invalid.d
> delete mode 100644 gas/testsuite/gas/aarch64/bfloat16-bad.d
> delete mode 100644 gas/testsuite/gas/aarch64/bfloat16-bad.l
> create mode 100644 gas/testsuite/gas/aarch64/sme-b16b16-1.d
> create mode 100644 gas/testsuite/gas/aarch64/sme-b16b16-1.s
> create mode 100644 gas/testsuite/gas/aarch64/sme-b16b16-bad-1.d
> create mode 100644 gas/testsuite/gas/aarch64/sme-b16b16-bad-1.l
> create mode 100644 gas/testsuite/gas/aarch64/sme-b16b16-bad-1.s
> create mode 100644 gas/testsuite/gas/aarch64/sve-b16b16-sme2-bad-1.d
> create mode 100644 gas/testsuite/gas/aarch64/sve-b16b16-sme2-bad-1.l
> create mode 100644 gas/testsuite/gas/aarch64/sve-b16b16-sme2-bad-1.s
> create mode 100644 gas/testsuite/gas/aarch64/sve-b16b16-sme2.d
> create mode 100644 gas/testsuite/gas/aarch64/sve-b16b16-sme2.s
> create mode 100644 gas/testsuite/gas/aarch64/sve-b16b16-sve2-bad-1.d
> rename gas/testsuite/gas/aarch64/{bfloat16-1-invalid.l => sve-b16b16-sve2-bad-1.l} (100%)
> rename gas/testsuite/gas/aarch64/{bfloat16-1-invalid.s => sve-b16b16-sve2-bad-1.s} (100%)
> create mode 100644 gas/testsuite/gas/aarch64/sve-b16b16-sve2-bad-2.d
> rename gas/testsuite/gas/aarch64/{bfloat16-2-invalid.l => sve-b16b16-sve2-bad-2.l} (93%)
> rename gas/testsuite/gas/aarch64/{bfloat16-2-invalid.s => sve-b16b16-sve2-bad-2.s} (100%)
> rename gas/testsuite/gas/aarch64/{bfloat16-1.d => sve-b16b16-sve2.d} (97%)
> rename gas/testsuite/gas/aarch64/{bfloat16-1.s => sve-b16b16-sve2.s} (100%)
>
> --
> 2.25.1
>
prev parent reply other threads:[~2025-01-09 12:11 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-01-07 11:34 Srinath Parvathaneni
2025-01-07 11:34 ` [PATCH v3 1/5] aarch64: Add support for FEAT_SVE_B16B16 feature Srinath Parvathaneni
2025-01-07 11:34 ` [PATCH v3 2/5] aarch64: Add support for FEAT_SVE_B16B16 min and max instructions Srinath Parvathaneni
2025-01-09 11:59 ` Andrew Carlotti
2025-01-07 11:34 ` [PATCH v3 3/5] aarch64: Add support for FEAT_SVE_B16B16 min and max instructions (autogenerated files) Srinath Parvathaneni
2025-01-07 11:34 ` [PATCH v3 4/5] aarch64: Add support for FEAT_SME_B16B16 feature Srinath Parvathaneni
2025-01-09 12:09 ` Andrew Carlotti
2025-01-07 11:34 ` [PATCH v3 5/5] aarch64: Add support for FEAT_SME_B16B16 feature (autogenerated files) Srinath Parvathaneni
2025-01-09 12:11 ` Andrew Carlotti [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=35beba38-3479-190d-50f1-445817c221c3@e124511.cambridge.arm.com \
--to=andrew.carlotti@arm.com \
--cc=binutils@sourceware.org \
--cc=nickc@redhat.com \
--cc=richard.earnshaw@arm.com \
--cc=srinath.parvathaneni@arm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).