<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=Windows-1252">
<style type="text/css" style="display:none;"> P {margin-top:0;margin-bottom:0;} </style>
</head>
<body dir="ltr">
<div class="elementToProof" style="margin-top: 1em; margin-bottom: 1em; font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);">
I’d like to provide additional context regarding the necessity of these patches.</div>
<div class="elementToProof" style="font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);">
The shift from dynamic load balancing mode to fixed mode, with CCS usage restricted to a single unit, has led to a notable performance regression, with workloads experiencing an approximately 10% FPS drop.</div>
<div class="elementToProof" style="font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);">
<br>
</div>
<div class="elementToProof" style="font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);">
For example, on DG2, the ResNet-50 inference benchmark previously achieved ~10,500 FPS in dynamic load balancing mode. However, after limiting CCS to 1 in fixed mode, performance dropped to ~9,200 FPS. With these patches, enabling all 4 CCS units via sysfs
(in fixed mode) restores performance back to nearly 10,500 FPS, effectively matching the previous dynamic mode results.</div>
<div class="elementToProof" style="font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);">
<br>
</div>
<div class="elementToProof" style="font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);">
Given customer expectations to maintain prior performance levels, these patches are essential to ensuring workloads utilizing multiple CCS units do not experience unnecessary degradation. The proposed sysfs interface provides configurability, allowing controlled
re-enablement of all 4 CCS units while keeping fixed mode intact. Since fixed mode is now in use, having a configurable approach ensures flexibility to address different scenarios that may arise.</div>
<div class="elementToProof" style="font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);">
<br>
</div>
<div class="elementToProof" style="font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);">
Let me know if you need further details.</div>
<div class="elementToProof" style="font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);">
<br>
</div>
<div class="elementToProof" style="font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);">
Regards,</div>
<div class="elementToProof" style="font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);">
Arshad</div>
<div id="appendonsend"></div>
<hr style="display:inline-block;width:98%" tabindex="-1">
<div id="divRplyFwdMsg" dir="ltr"><font face="Calibri, sans-serif" style="font-size:11pt" color="#000000"><b>From:</b> Joonas Lahtinen <joonas.lahtinen@linux.intel.com><br>
<b>Sent:</b> Thursday, March 27, 2025 2:49 PM<br>
<b>To:</b> Andi Shyti <andi.shyti@linux.intel.com><br>
<b>Cc:</b> Andi Shyti <andi.shyti@linux.intel.com>; dri-devel <dri-devel@lists.freedesktop.org>; intel-gfx <intel-gfx@lists.freedesktop.org>; Tvrtko Ursulin <tursulin@ursulin.net>; Chris Wilson <chris.p.wilson@linux.intel.com>; Simona Vetter <simona.vetter@ffwll.ch>;
Mehmood, Arshad <arshad.mehmood@intel.com>; Mrozek, Michal <michal.mrozek@intel.com>; Andi Shyti <andi.shyti@kernel.org>; Ayyalasomayajula, Usharani <usharani.ayyalasomayajula@intel.com><br>
<b>Subject:</b> Re: [PATCH v4 00/15] CCS static load balance</font>
<div> </div>
</div>
<div class="BodyFragment"><font size="2"><span style="font-size:11pt;">
<div class="PlainText">Quoting Andi Shyti (2025-03-25 12:52:58)<br>
> On Tue, Mar 25, 2025 at 10:24:42AM +0200, Joonas Lahtinen wrote:<br>
<br>
<SNIP><br>
<br>
> > Do you have a reference to some GitLab issues or maybe some external<br>
> > project issues where regressions around here are discussed?<br>
> <br>
> AFAIK, there's no GitLab issue for this because we're not fixing<br>
> a bug here; we're adding a new sysfs interface.<br>
<br>
This sysfs interface was exactly designed to address performance<br>
regressions coming from limiting the number of CCS to 1.<br>
<br>
So unless we have a specific workload and end-user reporting a <br>
regression on it, there's no incentive to spend any further time here.<br>
<br>
<SNIP><br>
<br>
> Arshad and Usha can definitely help if there are any technical<br>
> questions about how the application uses the interface.<br>
<br>
I don't have any technical questions as I specified the interface<br>
initially :)<br>
<br>
This is not about technical opens about how the interface works.<br>
To recap, when we initially implemented the 1CCS mode, we got active<br>
feedback on the community on regressions.<br>
<br>
We were careful to verify that all userspace would cleanly fall back to<br>
using 1CCS mode after it was implemented. And indeed, nobody has been<br>
asking for the 4CCS mode back after the 1CCS mode bugs were fixed.<br>
<br>
So as far as I see it, there are no users for this interface in<br>
upstream, and thus we should not spend the time on it.<br>
<br>
Regards, Joonas<br>
</div>
</span></font></div>
</body>
</html>