Date: Thu, 8 Nov 2018 15:12:37 +0000 (UTC) From: Mateusz Guzik <mjg@FreeBSD.org> To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: svn commit: r340252 - head/sys/amd64/amd64 Message-ID: <201811081512.wA8FCbqG086293@repo.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: mjg Date: Thu Nov 8 15:12:36 2018 New Revision: 340252 URL: https://svnweb.freebsd.org/changeset/base/340252 Log: amd64: align memset buffers to 16 bytes before using rep stos Both Intel manual and Agner Fog's docs suggest aligning to 16. See the review for benchmark results. Reviewed by: kib (previous version) Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D17661 Modified: head/sys/amd64/amd64/support.S Modified: head/sys/amd64/amd64/support.S ============================================================================== --- head/sys/amd64/amd64/support.S Thu Nov 8 14:46:21 2018 (r340251) +++ head/sys/amd64/amd64/support.S Thu Nov 8 15:12:36 2018 (r340252) @@ -515,24 +515,38 @@ END(memcpy_erms) 1256: movq %rdi,%r9 movq %r10,%rax + testl $15,%edi + jnz 3f +1: .if \erms == 1 rep stosb movq %r9,%rax .else + movq %rcx,%rdx shrq $3,%rcx rep stosq movq %r9,%rax andl $7,%edx - jnz 1f + jnz 2f POP_FRAME_POINTER ret -1: +2: movq %r10,-8(%rdi,%rdx) .endif POP_FRAME_POINTER ret + ALIGN_TEXT +3: + movq %r10,(%rdi) + movq %r10,8(%rdi) + movq %rdi,%r8 + andq $15,%r8 + leaq -16(%rcx,%r8),%rcx + neg %r8 + leaq 16(%rdi,%r8),%rdi + jmp 1b .endm ENTRY(memset_std)
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201811081512.wA8FCbqG086293>