[PATCH v6 0/6] support subsets of Float-Point in Integer Registers extensions

Weiwei Li posted 6 patches 2 years, 2 months ago
Test checkpatch passed
Patches applied successfully (tree, apply log)
git fetch https://github.com/patchew-project/qemu tags/patchew/20220211043920.28981-1-liweiwei@iscas.ac.cn
Maintainers: Palmer Dabbelt <palmer@dabbelt.com>, Alistair Francis <alistair.francis@wdc.com>, Bin Meng <bin.meng@windriver.com>
target/riscv/cpu.c                        |  17 ++
target/riscv/cpu.h                        |   4 +
target/riscv/cpu_helper.c                 |   6 +-
target/riscv/csr.c                        |  25 +-
target/riscv/fpu_helper.c                 | 178 ++++++------
target/riscv/helper.h                     |   4 +-
target/riscv/insn_trans/trans_rvd.c.inc   | 285 ++++++++++++++-----
target/riscv/insn_trans/trans_rvf.c.inc   | 314 +++++++++++++-------
target/riscv/insn_trans/trans_rvzfh.c.inc | 332 +++++++++++++++-------
target/riscv/internals.h                  |  32 ++-
target/riscv/translate.c                  | 149 +++++++++-
11 files changed, 974 insertions(+), 372 deletions(-)
[PATCH v6 0/6] support subsets of Float-Point in Integer Registers extensions
Posted by Weiwei Li 2 years, 2 months ago
This patchset implements RISC-V Float-Point in Integer Registers extensions(Version 1.0), which includes Zfinx, Zdinx, Zhinx and Zhinxmin extension. 

Specification:
https://github.com/riscv/riscv-zfinx/blob/main/zfinx-1.0.0.pdf

The port is available here:
https://github.com/plctlab/plct-qemu/tree/plct-zfinx-upstream-v6

To test this implementation, specify cpu argument with 'zfinx =true,zdinx=true,zhinx=true,zhinxmin=true' with 'g=false,f=false,d=false,Zfh=false,Zfhmin=false'
This implementation can pass gcc tests, ci result can be found in https://ci.rvperf.org/job/plct-qemu-zfinx-upstream/.

v6:
* rename flags Z*inx to z*inx
* rebase on apply-to-riscv.next

v5:
* put definition of ftemp and nftemp together, add comments for them
* sperate the declare of variable i from loop 

v4:
* combine register pair check for rv32 zdinx
* clear mstatus.FS when RVF is disabled by write_misa

v3:
* delete unused reset for mstatus.FS
* use positive test for RVF instead of negative test for ZFINX
* replace get_ol with get_xl
* use tcg_gen_concat_tl_i64 to unify tcg_gen_concat_i32_i64 and tcg_gen_deposit_i64

v2:
* hardwire mstatus.FS to zero when enable zfinx
* do register-pair check at the begin of translation
* optimize partial implemention as suggested

Weiwei Li (6):
  target/riscv: add cfg properties for zfinx, zdinx and zhinx{min}
  target/riscv: hardwire mstatus.FS to zero when enable zfinx
  target/riscv: add support for zfinx
  target/riscv: add support for zdinx
  target/riscv: add support for zhinx/zhinxmin
  target/riscv: expose zfinx, zdinx, zhinx{min} properties

 target/riscv/cpu.c                        |  17 ++
 target/riscv/cpu.h                        |   4 +
 target/riscv/cpu_helper.c                 |   6 +-
 target/riscv/csr.c                        |  25 +-
 target/riscv/fpu_helper.c                 | 178 ++++++------
 target/riscv/helper.h                     |   4 +-
 target/riscv/insn_trans/trans_rvd.c.inc   | 285 ++++++++++++++-----
 target/riscv/insn_trans/trans_rvf.c.inc   | 314 +++++++++++++-------
 target/riscv/insn_trans/trans_rvzfh.c.inc | 332 +++++++++++++++-------
 target/riscv/internals.h                  |  32 ++-
 target/riscv/translate.c                  | 149 +++++++++-
 11 files changed, 974 insertions(+), 372 deletions(-)

-- 
2.17.1


Re: [PATCH v6 0/6] support subsets of Float-Point in Integer Registers extensions
Posted by Alistair Francis 2 years, 2 months ago
On Fri, Feb 11, 2022 at 2:49 PM Weiwei Li <liweiwei@iscas.ac.cn> wrote:
>
> This patchset implements RISC-V Float-Point in Integer Registers extensions(Version 1.0), which includes Zfinx, Zdinx, Zhinx and Zhinxmin extension.
>
> Specification:
> https://github.com/riscv/riscv-zfinx/blob/main/zfinx-1.0.0.pdf
>
> The port is available here:
> https://github.com/plctlab/plct-qemu/tree/plct-zfinx-upstream-v6
>
> To test this implementation, specify cpu argument with 'zfinx =true,zdinx=true,zhinx=true,zhinxmin=true' with 'g=false,f=false,d=false,Zfh=false,Zfhmin=false'
> This implementation can pass gcc tests, ci result can be found in https://ci.rvperf.org/job/plct-qemu-zfinx-upstream/.
>
> v6:
> * rename flags Z*inx to z*inx
> * rebase on apply-to-riscv.next
>
> v5:
> * put definition of ftemp and nftemp together, add comments for them
> * sperate the declare of variable i from loop
>
> v4:
> * combine register pair check for rv32 zdinx
> * clear mstatus.FS when RVF is disabled by write_misa
>
> v3:
> * delete unused reset for mstatus.FS
> * use positive test for RVF instead of negative test for ZFINX
> * replace get_ol with get_xl
> * use tcg_gen_concat_tl_i64 to unify tcg_gen_concat_i32_i64 and tcg_gen_deposit_i64
>
> v2:
> * hardwire mstatus.FS to zero when enable zfinx
> * do register-pair check at the begin of translation
> * optimize partial implemention as suggested
>
> Weiwei Li (6):
>   target/riscv: add cfg properties for zfinx, zdinx and zhinx{min}
>   target/riscv: hardwire mstatus.FS to zero when enable zfinx
>   target/riscv: add support for zfinx
>   target/riscv: add support for zdinx
>   target/riscv: add support for zhinx/zhinxmin
>   target/riscv: expose zfinx, zdinx, zhinx{min} properties

Thanks!

Applied to riscv-to-apply.next

Alistair

>
>  target/riscv/cpu.c                        |  17 ++
>  target/riscv/cpu.h                        |   4 +
>  target/riscv/cpu_helper.c                 |   6 +-
>  target/riscv/csr.c                        |  25 +-
>  target/riscv/fpu_helper.c                 | 178 ++++++------
>  target/riscv/helper.h                     |   4 +-
>  target/riscv/insn_trans/trans_rvd.c.inc   | 285 ++++++++++++++-----
>  target/riscv/insn_trans/trans_rvf.c.inc   | 314 +++++++++++++-------
>  target/riscv/insn_trans/trans_rvzfh.c.inc | 332 +++++++++++++++-------
>  target/riscv/internals.h                  |  32 ++-
>  target/riscv/translate.c                  | 149 +++++++++-
>  11 files changed, 974 insertions(+), 372 deletions(-)
>
> --
> 2.17.1
>
>