From: Charlie Jenkins <charlie@rivosinc.com>
To: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Conor Dooley <conor@kernel.org>,
Paul Walmsley <paul.walmsley@sifive.com>,
aou@eecs.berkeley.edu, linux-riscv@lists.infradead.org,
linux-kernel@vger.kernel.org
Subject: Re: [PATCH] riscv: lib: Support csum on GCC <11
Date: Thu, 18 Jan 2024 14:38:46 -0800 [thread overview]
Message-ID: <ZamodtvgqscAqW8N@ghost> (raw)
In-Reply-To: <mhng-6127df88-9768-4f9d-8d78-a1aeb4213451@palmer-ri-x1c9>
On Thu, Jan 18, 2024 at 02:20:42PM -0800, Palmer Dabbelt wrote:
> On Thu, 18 Jan 2024 14:05:44 PST (-0800), Conor Dooley wrote:
> > On Thu, Jan 18, 2024 at 01:53:59PM -0800, Charlie Jenkins wrote:
> > > The OutputOperands field for asm goto statements is only supported
> > > starting from GCC 11. Split the asm goto to remove the use of this
> > > feature.
> > >
> > > Signed-off-by: Charlie Jenkins <charlie@rivosinc.com>
> > > Fixes: a04c192eabfb ("riscv: Add checksum library")
> > > ---
> > > The OutputOperands field for asm goto statements is only supported
> > > starting from GCC 11. Split the asm goto to remove the use of this
> > > feature.
> >
> > Maybe this is a super naive question, but is it possible to just not
> > use the custom csum code for gcc older than 11?
>
> Charlie and I were talking, these old GCC versions also don't support ZBB.
> So I think we can get away with something like
I sent an updated version that uses CC_HAS_ASM_GOTO_TIED_OUTPUT since
there may exist a compiler that supports ZBB but doesn't support output
in asm goto.
- Charlie
>
> diff --git a/arch/riscv/lib/csum.c b/arch/riscv/lib/csum.c
> index 06ce8e7250d9..17f883b612c9 100644
> --- a/arch/riscv/lib/csum.c
> +++ b/arch/riscv/lib/csum.c
> @@ -158,10 +158,16 @@ do_csum_with_alignment(const unsigned char *buff, int len)
>
> /*
> * Zbb support saves 6 instructions, so not worth checking without
> - * alternatives if supported
> + * alternatives if supported.
> + *
> + * Note that we pull the check for ZBB into the preprocessor proper
> + * here, as otherwise GCC will attempt to compile the code inside the
> + * "if (IS_ENABLED(ZBB)" block which fails because GCC10 doesn't
> + * support ASM goto output operands. GCC 10 also doesn't support ZBB,
> + * so we're safe with that check here.
> */
> - if (IS_ENABLED(CONFIG_RISCV_ISA_ZBB) &&
> - IS_ENABLED(CONFIG_RISCV_ALTERNATIVE)) {
> +#if defined(CONFIG_RISCV_ISA_ZBB)
> + if (IS_ENABLED(CONFIG_RISCV_ALTERNATIVE)) {
> unsigned long fold_temp;
>
> /*
> @@ -213,6 +219,7 @@ do_csum_with_alignment(const unsigned char *buff, int len)
> end:
> return csum >> 16;
> }
> +#endif /*CONFIG_RISCV_ISA_ZBB*/
> no_zbb:
> #ifndef CONFIG_32BIT
> csum += ror64(csum, 32);
> @@ -244,10 +251,11 @@ do_csum_no_alignment(const unsigned char *buff, int len)
>
> /*
> * Zbb support saves 6 instructions, so not worth checking without
> - * alternatives if supported
> + * alternatives if supported. See above for the ZBB preprocessor
> + * check.
> */
> - if (IS_ENABLED(CONFIG_RISCV_ISA_ZBB) &&
> - IS_ENABLED(CONFIG_RISCV_ALTERNATIVE)) {
> +#if defined(CONFIG_RISCV_ISA_ZBB)
> + if (IS_ENABLED(CONFIG_RISCV_ALTERNATIVE)) {
> unsigned long fold_temp;
>
> /*
> @@ -287,6 +295,7 @@ do_csum_no_alignment(const unsigned char *buff, int len)
> #endif /* !CONFIG_32BIT */
> return csum >> 16;
> }
> +#endif /*CONFIG_RISCV_ISA_ZBB*/
> no_zbb:
> #ifndef CONFIG_32BIT
> csum += ror64(csum, 32);
>
> which is building for me on GCC-10/defconfig.
>
> >
> > > ---
> > > arch/riscv/lib/csum.c | 42 ++++++++++++++++++++++++++++++------------
> > > 1 file changed, 30 insertions(+), 12 deletions(-)
> > >
> > > diff --git a/arch/riscv/lib/csum.c b/arch/riscv/lib/csum.c
> > > index 06ce8e7250d9..23be289f52b6 100644
> > > --- a/arch/riscv/lib/csum.c
> > > +++ b/arch/riscv/lib/csum.c
> > > @@ -177,22 +177,35 @@ do_csum_with_alignment(const unsigned char *buff, int len)
> > > : no_zbb);
> > > #ifdef CONFIG_32BIT
> > > - asm_volatile_goto(".option push \n\
> > > + /*
> > > + * OutputOperands in asm goto is not supported until GCC 11, so
> > > + * this asm has to be split to be compatible.
> > > + */
> > > + asm (".option push \n\
> > > .option arch,+zbb \n\
> > > rori %[fold_temp], %[csum], 16 \n\
> > > andi %[offset], %[offset], 1 \n\
> > > add %[csum], %[fold_temp], %[csum] \n\
> > > - beq %[offset], zero, %l[end] \n\
> > > - rev8 %[csum], %[csum] \n\
> > > .option pop"
> > > : [csum] "+r" (csum), [fold_temp] "=&r" (fold_temp)
> > > - : [offset] "r" (offset)
> > > - :
> > > - : end);
> > > + : [offset] "r" (offset));
> > > +
> > > + if (offset == 0)
> > > + goto end;
> > > +
> > > + asm (".option push \n\
> > > + .option arch, +zbb \n\
> > > + rev8 %[csum], %[csum] \n\
> > > + .option pop"
> > > + : [csum] "+r" (csum));
> > > return (unsigned short)csum;
> > > #else /* !CONFIG_32BIT */
> > > - asm_volatile_goto(".option push \n\
> > > + /*
> > > + * OutputOperands in asm goto is not supported until GCC 11, so
> > > + * this asm has to be split to be compatible.
> > > + */
> > > + asm (".option push \n\
> > > .option arch,+zbb \n\
> > > rori %[fold_temp], %[csum], 32 \n\
> > > add %[csum], %[fold_temp], %[csum] \n\
> > > @@ -200,13 +213,18 @@ do_csum_with_alignment(const unsigned char *buff, int len)
> > > roriw %[fold_temp], %[csum], 16 \n\
> > > addw %[csum], %[fold_temp], %[csum] \n\
> > > andi %[offset], %[offset], 1 \n\
> > > - beq %[offset], zero, %l[end] \n\
> > > - rev8 %[csum], %[csum] \n\
> > > .option pop"
> > > : [csum] "+r" (csum), [fold_temp] "=&r" (fold_temp)
> > > - : [offset] "r" (offset)
> > > - :
> > > - : end);
> > > + : [offset] "r" (offset));
> > > +
> > > + if (offset == 0)
> > > + goto end;
> > > +
> > > + asm (".option push \n\
> > > + .option arch, +zbb \n\
> > > + rev8 %[csum], %[csum] \n\
> > > + .option pop"
> > > + : [csum] "+r" (csum));
> > > return (csum << 16) >> 48;
> > > #endif /* !CONFIG_32BIT */
> > >
> > > ---
> > > base-commit: 080c4324fa5e81ff3780206a138223abfb57a68e
> > > change-id: 20240118-csum_remove_output_operands_asm_goto-49922c141ce7
> > > --
> > > - Charlie
> > >
> > >
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
WARNING: multiple messages have this Message-ID (diff)
From: Charlie Jenkins <charlie@rivosinc.com>
To: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Conor Dooley <conor@kernel.org>,
Paul Walmsley <paul.walmsley@sifive.com>,
aou@eecs.berkeley.edu, linux-riscv@lists.infradead.org,
linux-kernel@vger.kernel.org
Subject: Re: [PATCH] riscv: lib: Support csum on GCC <11
Date: Thu, 18 Jan 2024 14:38:46 -0800 [thread overview]
Message-ID: <ZamodtvgqscAqW8N@ghost> (raw)
In-Reply-To: <mhng-6127df88-9768-4f9d-8d78-a1aeb4213451@palmer-ri-x1c9>
On Thu, Jan 18, 2024 at 02:20:42PM -0800, Palmer Dabbelt wrote:
> On Thu, 18 Jan 2024 14:05:44 PST (-0800), Conor Dooley wrote:
> > On Thu, Jan 18, 2024 at 01:53:59PM -0800, Charlie Jenkins wrote:
> > > The OutputOperands field for asm goto statements is only supported
> > > starting from GCC 11. Split the asm goto to remove the use of this
> > > feature.
> > >
> > > Signed-off-by: Charlie Jenkins <charlie@rivosinc.com>
> > > Fixes: a04c192eabfb ("riscv: Add checksum library")
> > > ---
> > > The OutputOperands field for asm goto statements is only supported
> > > starting from GCC 11. Split the asm goto to remove the use of this
> > > feature.
> >
> > Maybe this is a super naive question, but is it possible to just not
> > use the custom csum code for gcc older than 11?
>
> Charlie and I were talking, these old GCC versions also don't support ZBB.
> So I think we can get away with something like
I sent an updated version that uses CC_HAS_ASM_GOTO_TIED_OUTPUT since
there may exist a compiler that supports ZBB but doesn't support output
in asm goto.
- Charlie
>
> diff --git a/arch/riscv/lib/csum.c b/arch/riscv/lib/csum.c
> index 06ce8e7250d9..17f883b612c9 100644
> --- a/arch/riscv/lib/csum.c
> +++ b/arch/riscv/lib/csum.c
> @@ -158,10 +158,16 @@ do_csum_with_alignment(const unsigned char *buff, int len)
>
> /*
> * Zbb support saves 6 instructions, so not worth checking without
> - * alternatives if supported
> + * alternatives if supported.
> + *
> + * Note that we pull the check for ZBB into the preprocessor proper
> + * here, as otherwise GCC will attempt to compile the code inside the
> + * "if (IS_ENABLED(ZBB)" block which fails because GCC10 doesn't
> + * support ASM goto output operands. GCC 10 also doesn't support ZBB,
> + * so we're safe with that check here.
> */
> - if (IS_ENABLED(CONFIG_RISCV_ISA_ZBB) &&
> - IS_ENABLED(CONFIG_RISCV_ALTERNATIVE)) {
> +#if defined(CONFIG_RISCV_ISA_ZBB)
> + if (IS_ENABLED(CONFIG_RISCV_ALTERNATIVE)) {
> unsigned long fold_temp;
>
> /*
> @@ -213,6 +219,7 @@ do_csum_with_alignment(const unsigned char *buff, int len)
> end:
> return csum >> 16;
> }
> +#endif /*CONFIG_RISCV_ISA_ZBB*/
> no_zbb:
> #ifndef CONFIG_32BIT
> csum += ror64(csum, 32);
> @@ -244,10 +251,11 @@ do_csum_no_alignment(const unsigned char *buff, int len)
>
> /*
> * Zbb support saves 6 instructions, so not worth checking without
> - * alternatives if supported
> + * alternatives if supported. See above for the ZBB preprocessor
> + * check.
> */
> - if (IS_ENABLED(CONFIG_RISCV_ISA_ZBB) &&
> - IS_ENABLED(CONFIG_RISCV_ALTERNATIVE)) {
> +#if defined(CONFIG_RISCV_ISA_ZBB)
> + if (IS_ENABLED(CONFIG_RISCV_ALTERNATIVE)) {
> unsigned long fold_temp;
>
> /*
> @@ -287,6 +295,7 @@ do_csum_no_alignment(const unsigned char *buff, int len)
> #endif /* !CONFIG_32BIT */
> return csum >> 16;
> }
> +#endif /*CONFIG_RISCV_ISA_ZBB*/
> no_zbb:
> #ifndef CONFIG_32BIT
> csum += ror64(csum, 32);
>
> which is building for me on GCC-10/defconfig.
>
> >
> > > ---
> > > arch/riscv/lib/csum.c | 42 ++++++++++++++++++++++++++++++------------
> > > 1 file changed, 30 insertions(+), 12 deletions(-)
> > >
> > > diff --git a/arch/riscv/lib/csum.c b/arch/riscv/lib/csum.c
> > > index 06ce8e7250d9..23be289f52b6 100644
> > > --- a/arch/riscv/lib/csum.c
> > > +++ b/arch/riscv/lib/csum.c
> > > @@ -177,22 +177,35 @@ do_csum_with_alignment(const unsigned char *buff, int len)
> > > : no_zbb);
> > > #ifdef CONFIG_32BIT
> > > - asm_volatile_goto(".option push \n\
> > > + /*
> > > + * OutputOperands in asm goto is not supported until GCC 11, so
> > > + * this asm has to be split to be compatible.
> > > + */
> > > + asm (".option push \n\
> > > .option arch,+zbb \n\
> > > rori %[fold_temp], %[csum], 16 \n\
> > > andi %[offset], %[offset], 1 \n\
> > > add %[csum], %[fold_temp], %[csum] \n\
> > > - beq %[offset], zero, %l[end] \n\
> > > - rev8 %[csum], %[csum] \n\
> > > .option pop"
> > > : [csum] "+r" (csum), [fold_temp] "=&r" (fold_temp)
> > > - : [offset] "r" (offset)
> > > - :
> > > - : end);
> > > + : [offset] "r" (offset));
> > > +
> > > + if (offset == 0)
> > > + goto end;
> > > +
> > > + asm (".option push \n\
> > > + .option arch, +zbb \n\
> > > + rev8 %[csum], %[csum] \n\
> > > + .option pop"
> > > + : [csum] "+r" (csum));
> > > return (unsigned short)csum;
> > > #else /* !CONFIG_32BIT */
> > > - asm_volatile_goto(".option push \n\
> > > + /*
> > > + * OutputOperands in asm goto is not supported until GCC 11, so
> > > + * this asm has to be split to be compatible.
> > > + */
> > > + asm (".option push \n\
> > > .option arch,+zbb \n\
> > > rori %[fold_temp], %[csum], 32 \n\
> > > add %[csum], %[fold_temp], %[csum] \n\
> > > @@ -200,13 +213,18 @@ do_csum_with_alignment(const unsigned char *buff, int len)
> > > roriw %[fold_temp], %[csum], 16 \n\
> > > addw %[csum], %[fold_temp], %[csum] \n\
> > > andi %[offset], %[offset], 1 \n\
> > > - beq %[offset], zero, %l[end] \n\
> > > - rev8 %[csum], %[csum] \n\
> > > .option pop"
> > > : [csum] "+r" (csum), [fold_temp] "=&r" (fold_temp)
> > > - : [offset] "r" (offset)
> > > - :
> > > - : end);
> > > + : [offset] "r" (offset));
> > > +
> > > + if (offset == 0)
> > > + goto end;
> > > +
> > > + asm (".option push \n\
> > > + .option arch, +zbb \n\
> > > + rev8 %[csum], %[csum] \n\
> > > + .option pop"
> > > + : [csum] "+r" (csum));
> > > return (csum << 16) >> 48;
> > > #endif /* !CONFIG_32BIT */
> > >
> > > ---
> > > base-commit: 080c4324fa5e81ff3780206a138223abfb57a68e
> > > change-id: 20240118-csum_remove_output_operands_asm_goto-49922c141ce7
> > > --
> > > - Charlie
> > >
> > >
next prev parent reply other threads:[~2024-01-18 22:39 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-01-18 21:53 [PATCH] riscv: lib: Support csum on GCC <11 Charlie Jenkins
2024-01-18 21:53 ` Charlie Jenkins
2024-01-18 22:05 ` Conor Dooley
2024-01-18 22:05 ` Conor Dooley
2024-01-18 22:20 ` Palmer Dabbelt
2024-01-18 22:20 ` Palmer Dabbelt
2024-01-18 22:38 ` Charlie Jenkins [this message]
2024-01-18 22:38 ` Charlie Jenkins
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=ZamodtvgqscAqW8N@ghost \
--to=charlie@rivosinc.com \
--cc=aou@eecs.berkeley.edu \
--cc=conor@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-riscv@lists.infradead.org \
--cc=palmer@dabbelt.com \
--cc=paul.walmsley@sifive.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.