Re: [PATCH v3 7/9] powerpc64: Add optimized chacha20

[Paul E Murphy <murphyp@linux.ibm.com>]

On 4/19/22 4:28 PM, Adhemerval Zanella via Libc-alpha wrote:
> It adds vectorized ChaCha20 implementation based on libgcrypt
> cipher/chacha20-ppc.c.  It targets POWER8 and it is used on
> default for LE.

> diff --git a/sysdeps/powerpc/powerpc64/chacha20-ppc.c b/sysdeps/powerpc/powerpc64/chacha20-ppc.c
> new file mode 100644
> index 0000000000..e2567c379a
> --- /dev/null
> +++ b/sysdeps/powerpc/powerpc64/chacha20-ppc.c

How difficult is it to keep this synchronized with the upstream version 
in libgcrypt?  Also, this seems like it would be a better placed in the 
power8 subdirectory.

> diff --git a/sysdeps/powerpc/powerpc64/chacha20_arch.h b/sysdeps/powerpc/powerpc64/chacha20_arch.h
> new file mode 100644
> index 0000000000..a18115392f
> --- /dev/null
> +++ b/sysdeps/powerpc/powerpc64/chacha20_arch.h
> @@ -0,0 +1,47 @@
> +/* PowerPC optimization for ChaCha20.
> +   Copyright (C) 2022 Free Software Foundation, Inc.
> +   This file is part of the GNU C Library.
> +
> +   The GNU C Library is free software; you can redistribute it and/or
> +   modify it under the terms of the GNU Lesser General Public
> +   License as published by the Free Software Foundation; either
> +   version 2.1 of the License, or (at your option) any later version.
> +
> +   The GNU C Library is distributed in the hope that it will be useful,
> +   but WITHOUT ANY WARRANTY; without even the implied warranty of
> +   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
> +   Lesser General Public License for more details.
> +
> +   You should have received a copy of the GNU Lesser General Public
> +   License along with the GNU C Library; if not, see
> +   <http://www.gnu.org/licenses/>.  */
> +
> +#include <stdbool.h>
> +#include <ldsodefs.h>
> +
> +unsigned int __chacha20_power8_blocks4 (uint32_t *state, uint8_t *dst,
> +					const uint8_t *src, size_t nblks)
> +     attribute_hidden;
> +
> +static void
> +chacha20_crypt (uint32_t *state, uint8_t *dst,
> +		const uint8_t *src, size_t bytes)
> +{
> +  _Static_assert (CHACHA20_BUFSIZE % 4 == 0,
> +		  "CHACHA20_BUFSIZE not multiple of 4");
> +  _Static_assert (CHACHA20_BUFSIZE >= CHACHA20_BLOCK_SIZE * 4,
> +		  "CHACHA20_BUFSIZE < CHACHA20_BLOCK_SIZE * 4");
> +
> +#ifdef __LITTLE_ENDIAN__
> +  __chacha20_power8_blocks4 (state, dst, src,
> +			     CHACHA20_BUFSIZE / CHACHA20_BLOCK_SIZE);
> +#else
> +  unsigned long int hwcap = GLRO(dl_hwcap);
> +  unsigned long int hwcap2 = GLRO(dl_hwcap2);
> +  if (hwcap2 & PPC_FEATURE2_ARCH_2_07 && hwcap & PPC_FEATURE_HAS_ALTIVEC)
> +    __chacha20_power8_blocks4 (state, dst, src,
> +			       CHACHA20_BUFSIZE / CHACHA20_BLOCK_SIZE);
> +  else
> +    chacha20_crypt_generic (state, dst, src, bytes);
> +#endif

This file doesn't seem to obey the multiarch conventions of other 
powerpc64 specific bits. Is it possible to implement multiarch support 
similar to the libc/libm routines?