Intel-XE Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: "K V P, Satyanarayana" <satyanarayana.k.v.p@intel.com>
To: <tilak.tirumalesh.tangudu@intel.com>,
	<piotr.piorkowski@intel.com>, <jesus.narvaez@intel.com>,
	<daniele.ceraolospurio@intel.com>, <stuart.summers@intel.com>,
	<intel-xe@lists.freedesktop.org>
Subject: Re: [PATCH 3/4] drm/xe: Use MI_SEMAPHORE_WAIT_64 for fake watchdog where supported
Date: Fri, 20 Nov 2026 13:20:30 +0530	[thread overview]
Message-ID: <f530bf9a-7427-4b97-a30c-8924c4e448e8@intel.com> (raw)
In-Reply-To: <20261113155734.4164230-4-tilak.tirumalesh.tangudu@intel.com>


On 13-Nov-26 9:27 PM, tilak.tirumalesh.tangudu@intel.com wrote:
> From: Tangudu Tilak Tirumalesh <tilak.tirumalesh.tangudu@intel.com>
>
> Use the MI_SEMAPHORE_WAIT_64 instruction in emit_fake_watchdog() on
> platforms whose GT supports it, per gt->info.has_sem64 (Wa_16029281913).
> Platforms without support keep using the existing 32-bit
> MI_SEMAPHORE_WAIT path.
>
> Assisted-by: LLM
> Signed-off-by: Tangudu Tilak Tirumalesh <tilak.tirumalesh.tangudu@intel.com>
> ---
>   drivers/gpu/drm/xe/xe_ring_ops.c | 24 +++++++++++++++++++-----
>   1 file changed, 19 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/gpu/drm/xe/xe_ring_ops.c b/drivers/gpu/drm/xe/xe_ring_ops.c
> index ad63181938e3..fbff2973a204 100644
> --- a/drivers/gpu/drm/xe/xe_ring_ops.c
> +++ b/drivers/gpu/drm/xe/xe_ring_ops.c
> @@ -316,11 +316,25 @@ static int emit_fake_watchdog(struct xe_lrc *lrc, u32 *dw, int i)
>   	dw[i++] = PR_CTR_CTRL(0).addr;
>   	dw[i++] = CTR_LOGIC_OP(START);
>   
> -	dw[i++] = MI_SEMAPHORE_WAIT | MI_SEMW_GGTT | MI_SEMW_POLL | MI_SEMW_COMPARE(SAD_EQ_SDD);
> -	dw[i++] = 0xdead; /* this should never be seen */
> -	dw[i++] = lower_32_bits(xe_lrc_ggtt_addr(lrc));
> -	dw[i++] = upper_32_bits(xe_lrc_ggtt_addr(lrc));
> -	dw[i++] = 0; /* unused token */
> +	/* Wa_16029281913: legacy semaphore has an issue on some Xe3p platforms, prefer sem64 */

No need to mention this WA. As per below condition, if has_sem64 is 
enabled, sem64 will be used.

If we need to use this only for this WA, we may need to derive this in 
some different way.

> +	if (lrc->gt->info.has_sem64) {
> +		dw[i++] = MI_SEMAPHORE_WAIT_64 | MI_SEMW_GGTT | MI_SEMW64_32B_COMPARE |
> +			  MI_SEMW64_INHIBIT_CTX_SWITCH | MI_SEMW64_INTERRUPT |
> +			  MI_SEMW_COMPARE(SAD_EQ_SDD);
> +		dw[i++] = 0xdead; /* this should never be seen */
> +		dw[i++] = 0;
Mention why is it zero here.
> +		dw[i++] = lower_32_bits(xe_lrc_ggtt_addr(lrc));
> +		dw[i++] = upper_32_bits(xe_lrc_ggtt_addr(lrc));
> +		dw[i++] = 0; /* token, must be 0 */
> +		dw[i++] = 0;
> +	} else {
> +		dw[i++] = MI_SEMAPHORE_WAIT | MI_SEMW_GGTT
> +			  | MI_SEMW_POLL | MI_SEMW_COMPARE(SAD_EQ_SDD);
> +		dw[i++] = 0xdead; /* this should never be seen */
> +		dw[i++] = lower_32_bits(xe_lrc_ggtt_addr(lrc));
> +		dw[i++] = upper_32_bits(xe_lrc_ggtt_addr(lrc));
> +		dw[i++] = 0; /* unused token */
> +	}
>   
>   	dw[i++] = MI_LOAD_REGISTER_IMM | MI_LRI_NUM_REGS(1) | MI_LRI_LRM_CS_MMIO;
>   	dw[i++] = PR_CTR_CTRL(0).addr;

  reply	other threads:[~2026-11-20  7:51 UTC|newest]

Thread overview: 17+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-13 15:57 [PATCH 0/4] drm/xe: Use MI_SEMAPHORE_WAIT_64 (Wa_16029281913) tilak.tirumalesh.tangudu
2026-11-13 15:57 ` [PATCH 1/4] drm/xe: Add has_sem64 bit to graphics/media IP descriptors tilak.tirumalesh.tangudu
2026-11-13 16:20   ` sashiko-bot
2026-11-13 15:57 ` [PATCH 2/4] drm/xe: Add MI_SEMAPHORE_WAIT_64 instruction definition tilak.tirumalesh.tangudu
2026-11-20  7:48   ` K V P, Satyanarayana
2026-11-20 11:32     ` Tangudu, Tilak Tirumalesh
2026-11-20 11:40       ` Tangudu, Tilak Tirumalesh
2026-11-20 12:36     ` Tangudu, Tilak Tirumalesh
2026-11-13 15:57 ` [PATCH 3/4] drm/xe: Use MI_SEMAPHORE_WAIT_64 for fake watchdog where supported tilak.tirumalesh.tangudu
2026-11-20  7:50   ` K V P, Satyanarayana [this message]
2026-11-13 15:57 ` [PATCH 4/4] drm/xe: Use MI_SEMAPHORE_WAIT_64 in ULLS postamble " tilak.tirumalesh.tangudu
2026-11-13 16:34 ` ✗ CI.checkpatch: warning for drm/xe: Use MI_SEMAPHORE_WAIT_64 (Wa_16029281913) (rev3) Patchwork
2026-11-13 16:36 ` ✓ CI.KUnit: success " Patchwork
2026-11-13 17:09 ` ✓ Xe.CI.BAT: " Patchwork
2026-11-14  3:12 ` ✓ Xe.CI.FULL: " Patchwork
  -- strict thread matches above, loose matches on Subject: below --
2026-11-20 12:02 [PATCH 0/4] drm/xe: Use MI_SEMAPHORE_WAIT_64 (Wa_16029281913) tilak.tirumalesh.tangudu
2026-11-20 12:02 ` [PATCH 3/4] drm/xe: Use MI_SEMAPHORE_WAIT_64 for fake watchdog where supported tilak.tirumalesh.tangudu
2026-11-20 12:18   ` K V P, Satyanarayana

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=f530bf9a-7427-4b97-a30c-8924c4e448e8@intel.com \
    --to=satyanarayana.k.v.p@intel.com \
    --cc=daniele.ceraolospurio@intel.com \
    --cc=intel-xe@lists.freedesktop.org \
    --cc=jesus.narvaez@intel.com \
    --cc=piotr.piorkowski@intel.com \
    --cc=stuart.summers@intel.com \
    --cc=tilak.tirumalesh.tangudu@intel.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox