Intel-XE Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Harish Chegondi <harish.chegondi@intel.com>
To: Matt Roper <matthew.d.roper@intel.com>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>,
	<intel-xe@lists.freedesktop.org>,
	Shekhar Chauhan <shekhar.chauhan@intel.com>,
	Balasubramani Vivekanandan <balasubramani.vivekanandan@intel.com>,
	Tejas Upadhyay <tejas.upadhyay@intel.com>,
	Ashutosh Dixit <ashutosh.dixit@intel.com>
Subject: Re: [PATCH 23/23] drm/xe/xe3p: Add xe3p EU stall data format
Date: Wed, 15 Oct 2025 16:05:58 -0700	[thread overview]
Message-ID: <aPAo1peMIoiiEplw@intel.com> (raw)
In-Reply-To: <20251014191139.GA5409@mdroper-desk1.amr.corp.intel.com>

On Tue, Oct 14, 2025 at 12:11:39PM -0700, Matt Roper wrote:
> On Mon, Oct 13, 2025 at 08:24:55PM -0700, Lucas De Marchi wrote:
> > From: Harish Chegondi <harish.chegondi@intel.com>
> > 
> > Starting with Xe3p, IP address in EU stall data is increases to 61 bits.
> 
> As far as I can see, the EU stall format isn't documented anywhere in
> the bspec (which seems like a major oversight!) but the early
> architecture document related to this change seems to indicate that a
> full 64-bit IP was supposed to be written by hardware now
> 
>         "Bit filed[sic] 0-28 is now 0-63"
> 
> Your structure is only using 61 bits (which definitely makes sense if
> the addresses are aligned since the lowest three bits would always be
> 0's); is there somewhere other than the bspec or architecture document
> that we should be looking at to confirm this?
I agree that EU stall is not properly documented in the B-Spec.
I looked up an HSD for Xe3p to confirm and make this change. Also
checked with the architect to confirm this change.
> 
> If you've been working directly with the hardware architects on this
> feature, please poke them about documenting this clearly in the bspec
> (with proper tagging by IP/platform) so that we can verify this is
> handled correctly and also notice if/when the format changes again on
> future platforms.
Sure, I will follow up with the HW architects to make sure the EU stall
data formats are properly documented in the Bspec and are properly
tagged.
> 
> 
> Matt

Thank You
Harish.
> 
> > 
> > Cc: Ashutosh Dixit <ashutosh.dixit@intel.com>
> > Signed-off-by: Harish Chegondi <harish.chegondi@intel.com>
> > Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
> > ---
> >  drivers/gpu/drm/xe/xe_eu_stall.c | 23 +++++++++++++++++++++++
> >  1 file changed, 23 insertions(+)
> > 
> > diff --git a/drivers/gpu/drm/xe/xe_eu_stall.c b/drivers/gpu/drm/xe/xe_eu_stall.c
> > index f5cfdf29fde34..2bc6b593ff172 100644
> > --- a/drivers/gpu/drm/xe/xe_eu_stall.c
> > +++ b/drivers/gpu/drm/xe/xe_eu_stall.c
> > @@ -124,6 +124,27 @@ struct xe_eu_stall_data_xe2 {
> >  	__u64 unused[6];
> >  } __packed;
> >  
> > +/*
> > + * EU stall data format for Xe3p arch GPUs.
> > + */
> > +struct xe_eu_stall_data_xe3p {
> > +	__u64 ip_addr:61;	  /* Bits 0  to 60  */
> > +	__u64 tdr_count:8;	  /* Bits 61 to 68  */
> > +	__u64 other_count:8;	  /* Bits 69 to 76  */
> > +	__u64 control_count:8;	  /* Bits 77 to 84  */
> > +	__u64 pipestall_count:8;  /* Bits 85 to 92  */
> > +	__u64 send_count:8;	  /* Bits 93 to 100 */
> > +	__u64 dist_acc_count:8;   /* Bits 101 to 108 */
> > +	__u64 sbid_count:8;	  /* Bits 109 to 116 */
> > +	__u64 sync_count:8;	  /* Bits 117 to 124 */
> > +	__u64 inst_fetch_count:8; /* Bits 125 to 132 */
> > +	__u64 active_count:8;	  /* Bits 133 to 140 */
> > +	__u64 ex_id:3;		  /* Bits 141 to 143 */
> > +	__u64 end_flag:1;	  /* Bit  144 */
> > +	__u64 unused_bits:47;
> > +	__u64 unused[5];
> > +} __packed;
> > +
> >  const u64 eu_stall_sampling_rates[] = {251, 251 * 2, 251 * 3, 251 * 4, 251 * 5, 251 * 6, 251 * 7};
> >  
> >  /**
> > @@ -169,6 +190,8 @@ size_t xe_eu_stall_data_record_size(struct xe_device *xe)
> >  
> >  	if (xe->info.platform == XE_PVC)
> >  		record_size = sizeof(struct xe_eu_stall_data_pvc);
> > +	else if (GRAPHICS_VER(xe) >= 35)
> > +		record_size = sizeof(struct xe_eu_stall_data_xe3p);
> >  	else if (GRAPHICS_VER(xe) >= 20)
> >  		record_size = sizeof(struct xe_eu_stall_data_xe2);
> >  
> > 
> > -- 
> > 2.51.0
> > 
> 
> -- 
> Matt Roper
> Graphics Software Engineer
> Linux GPU Platform Enablement
> Intel Corporation

  reply	other threads:[~2025-10-15 23:06 UTC|newest]

Thread overview: 78+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-10-14  3:24 [PATCH 00/23] drm/xe: Add Xe3p support Lucas De Marchi
2025-10-14  3:24 ` [PATCH 01/23] drm/xe/xe3: Add support for graphics IP versions 30.04 & 30.05 Lucas De Marchi
2025-10-14  6:17   ` Shekhar Chauhan
2025-10-14 16:11   ` Matt Roper
2025-10-14  3:24 ` [PATCH 02/23] drm/xe/xe3p_lpm: Add support for media IP versions 35.00 & 35.03 Lucas De Marchi
2025-10-14  6:22   ` Shekhar Chauhan
2025-10-14 16:14   ` Matt Roper
2025-10-14  3:24 ` [PATCH 03/23] drm/xe/xe3p_lpm: Configure MAIN_GAMCTRL_QUEUE_SELECT Lucas De Marchi
2025-10-14 16:34   ` Matt Roper
2025-10-15  2:28     ` Lucas De Marchi
2025-10-15  6:36       ` Vivekanandan, Balasubramani
2025-10-15 14:03         ` Lucas De Marchi
2025-10-16 14:20           ` Vivekanandan, Balasubramani
2025-10-15 14:59       ` Matt Roper
2025-10-14  3:24 ` [PATCH 04/23] drm/xe/xe3p_lpm: Add special check in Media GT for Main GAMCTRL Lucas De Marchi
2025-10-14 16:36   ` Matt Roper
2025-10-14  3:24 ` [PATCH 05/23] drm/xe/xe3p_lpm: Stop reading the CTC_MODE register Lucas De Marchi
2025-10-14 11:58   ` Shekhar Chauhan
2025-10-14 16:52     ` Matt Roper
2025-10-15  3:41       ` Lucas De Marchi
2025-10-15  9:19         ` Vivekanandan, Balasubramani
2025-10-15 15:04           ` Matt Roper
2025-10-14 16:40   ` Matt Roper
2025-10-14 16:53     ` Matt Roper
2025-10-14  3:24 ` [PATCH 06/23] drm/xe/xe3p_lpm: Skip disabling NOA on unsupported IPs Lucas De Marchi
2025-10-14 17:04   ` Matt Roper
2025-10-14  3:24 ` [PATCH 07/23] drm/xe/xe3p_lpm: Handle MCR steering Lucas De Marchi
2025-10-15  9:56   ` Vivekanandan, Balasubramani
2025-10-14  3:24 ` [PATCH 08/23] drm/xe/xe3p: Stop programming RCU_MODE's fixed slice mode setting Lucas De Marchi
2025-10-15 12:24   ` Vivekanandan, Balasubramani
2025-10-14  3:24 ` [PATCH 09/23] drm/xe/xe3p: Determine service copy availability from fuse Lucas De Marchi
2025-10-15 20:14   ` Gustavo Sousa
2025-10-14  3:24 ` [PATCH 10/23] drm/xe/xe3p: Skip TD flush Lucas De Marchi
2025-10-14 19:35   ` Matt Roper
2025-10-14  3:24 ` [PATCH 11/23] drm/xe/xe3p: Enable L2 flush optimization feature Lucas De Marchi
2025-10-14 19:43   ` Matt Roper
2025-10-15  4:02     ` Lucas De Marchi
2025-10-14  3:24 ` [PATCH 12/23] drm/xe/xe3p: Flush userptr/shrinker bo cachelines manually Lucas De Marchi
2025-10-14 12:58   ` Thomas Hellström
2025-10-15 18:42     ` Lucas De Marchi
2025-10-14  3:24 ` [PATCH 13/23] drm/xe: Dump CURRENT_LRCA and CSMQDEBUG registers Lucas De Marchi
2025-10-14 17:24   ` Matt Roper
2025-10-15  4:07     ` Lucas De Marchi
2025-10-14  3:24 ` [PATCH 14/23] drm/xe/nvl: Define NVL-S platform Lucas De Marchi
2025-10-14  7:34   ` Shekhar Chauhan
2025-10-14  3:24 ` [PATCH 15/23] drm/xe/nvls: Define GuC firmware for NVL-S Lucas De Marchi
2025-10-15 18:49   ` Lucas De Marchi
2025-10-14  3:24 ` [PATCH 16/23] drm/xe/nvls: Attach MOCS table " Lucas De Marchi
2025-10-14  7:45   ` Shekhar Chauhan
2025-10-14  3:24 ` [PATCH 17/23] drm/xe/xe3p_xpc: Add Xe3p_XPC IP definition Lucas De Marchi
2025-10-14  8:04   ` Shekhar Chauhan
2025-10-14  8:12     ` Shekhar Chauhan
2025-10-14 17:33     ` Matt Roper
2025-10-15  2:47       ` Shekhar Chauhan
2025-10-14 17:36   ` Matt Roper
2025-10-14  3:24 ` [PATCH 18/23] drm/xe/xe3p_xpc: Add L3 bank mask Lucas De Marchi
2025-10-14 17:46   ` Matt Roper
2025-10-14  3:24 ` [PATCH 19/23] drm/xe/xe3p_xpc: Add MCR steering Lucas De Marchi
2025-10-14  3:24 ` [PATCH 20/23] drm/xe/xe3p_xpc: Add support for compute walker for non-MSIx Lucas De Marchi
2025-10-14 18:07   ` Matt Roper
2025-10-15 17:07     ` Lucas De Marchi
2025-10-15 17:12       ` Matt Roper
2025-10-14  3:24 ` [PATCH 21/23] drm/xe/xe3p_xpc: Skip compression tuning on platforms without flatccs Lucas De Marchi
2025-10-14  8:09   ` Shekhar Chauhan
2025-10-14  8:13     ` Shekhar Chauhan
2025-10-14 18:14     ` Matt Roper
2025-10-15  2:52       ` Shekhar Chauhan
2025-10-15 14:38         ` Lucas De Marchi
2025-10-15 15:34           ` Shekhar Chauhan
2025-10-14  3:24 ` [PATCH 22/23] drm/xe/xe3p_xpc: Setup PAT table Lucas De Marchi
2025-10-16 14:07   ` Vivekanandan, Balasubramani
2025-10-14  3:24 ` [PATCH 23/23] drm/xe/xe3p: Add xe3p EU stall data format Lucas De Marchi
2025-10-14 19:11   ` Matt Roper
2025-10-15 23:05     ` Harish Chegondi [this message]
2025-10-14  3:45 ` ✗ CI.checkpatch: warning for drm/xe: Add Xe3p support Patchwork
2025-10-14  3:47 ` ✓ CI.KUnit: success " Patchwork
2025-10-14  4:22 ` ✓ Xe.CI.BAT: " Patchwork
2025-10-14 12:06 ` ✓ Xe.CI.Full: " Patchwork

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=aPAo1peMIoiiEplw@intel.com \
    --to=harish.chegondi@intel.com \
    --cc=ashutosh.dixit@intel.com \
    --cc=balasubramani.vivekanandan@intel.com \
    --cc=intel-xe@lists.freedesktop.org \
    --cc=lucas.demarchi@intel.com \
    --cc=matthew.d.roper@intel.com \
    --cc=shekhar.chauhan@intel.com \
    --cc=tejas.upadhyay@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