Re: [PATCH v8] eal/x86: improve rte_memcpy const size 16 performance

2024-07-09 Thread Patrick Robb
On Tue, Jul 9, 2024 at 8:48 AM David Marchand wrote: > > On Tue, Jul 9, 2024 at 2:43 PM Morten Brørup > wrote: > > PS: The CI should catch this stuff. > > Restoring OVS tests in CI has been requested and I think it was being worked > on. > Not sure where we are atm, Patrick? > OvS and SPDK com

RE: [PATCH v8] eal/x86: improve rte_memcpy const size 16 performance

2024-07-09 Thread Morten Brørup
> From: David Marchand [mailto:david.march...@redhat.com] > Sent: Tuesday, 9 July 2024 14.48 > > On Tue, Jul 9, 2024 at 2:43 PM Morten Brørup > wrote: > > PS: The CI should catch this stuff. Working on fixing it now, some old variants of rte_mov16() do the extra cast, and some don't. It could b

Re: [PATCH v8] eal/x86: improve rte_memcpy const size 16 performance

2024-07-09 Thread David Marchand
On Tue, Jul 9, 2024 at 2:43 PM Morten Brørup wrote: > PS: The CI should catch this stuff. Restoring OVS tests in CI has been requested and I think it was being worked on. Not sure where we are atm, Patrick? -- David Marchand

RE: [PATCH v8] eal/x86: improve rte_memcpy const size 16 performance

2024-07-09 Thread Morten Brørup
> From: David Marchand [mailto:david.march...@redhat.com] > Sent: Tuesday, 9 July 2024 13.43 > > Hello, > > On Tue, Jul 9, 2024 at 11:24 AM David Marchand > wrote: > > > > On Mon, Jun 10, 2024 at 3:40 PM Konstantin Ananyev > > wrote: > > > > When the rte_memcpy() size is 16, the same 16 bytes a

Re: [PATCH v8] eal/x86: improve rte_memcpy const size 16 performance

2024-07-09 Thread David Marchand
Hello, On Tue, Jul 9, 2024 at 11:24 AM David Marchand wrote: > > On Mon, Jun 10, 2024 at 3:40 PM Konstantin Ananyev > wrote: > > > When the rte_memcpy() size is 16, the same 16 bytes are copied twice. > > > In the case where the size is known to be 16 at build time, omit the > > > duplicate copy

Re: [PATCH v8] eal/x86: improve rte_memcpy const size 16 performance

2024-07-09 Thread David Marchand
On Mon, Jun 10, 2024 at 3:40 PM Konstantin Ananyev wrote: > > When the rte_memcpy() size is 16, the same 16 bytes are copied twice. > > In the case where the size is known to be 16 at build time, omit the > > duplicate copy. > > > > Reduced the amount of effectively copy-pasted code by using #ifde

RE: [PATCH v8] eal/x86: improve rte_memcpy const size 16 performance

2024-06-10 Thread Morten Brørup
> From: Konstantin Ananyev [mailto:konstantin.anan...@huawei.com] > Sent: Monday, 10 June 2024 15.40 > > > When the rte_memcpy() size is 16, the same 16 bytes are copied twice. > > In the case where the size is known to be 16 at build time, omit the > > duplicate copy. > > > > Reduced the amount o

RE: [PATCH v8] eal/x86: improve rte_memcpy const size 16 performance

2024-06-10 Thread Konstantin Ananyev
> When the rte_memcpy() size is 16, the same 16 bytes are copied twice. > In the case where the size is known to be 16 at build time, omit the > duplicate copy. > > Reduced the amount of effectively copy-pasted code by using #ifdef > inside functions instead of outside functions. > > Suggested-

RE: [PATCH v8] eal/x86: improve rte_memcpy const size 16 performance

2024-06-10 Thread Morten Brørup
PING for review. The CI failures can be ignored: Most of the CI doesn't support the Depends-on tag, and this patch uses __rte_constant(), provided by Tyler's patch series [1]. [1]: https://inbox.dpdk.org/dev/1710970416-27841-1-git-send-email-roret...@linux.microsoft.com/ -Morten > From: Morte

[PATCH v8] eal/x86: improve rte_memcpy const size 16 performance

2024-05-30 Thread Morten Brørup
When the rte_memcpy() size is 16, the same 16 bytes are copied twice. In the case where the size is known to be 16 at build time, omit the duplicate copy. Reduced the amount of effectively copy-pasted code by using #ifdef inside functions instead of outside functions. Suggested-by: Stephen Hemmin