From: "Zbigniew Kempczyński" <zbigniew.kempczynski@intel.com>
To: igt-dev@lists.freedesktop.org
Cc: "Zbigniew Kempczyński" <zbigniew.kempczynski@intel.com>,
"Francois Dugast" <francois.dugast@intel.com>,
"Priyanka Dandamudi" <priyanka.dandamudi@intel.com>
Subject: [PATCH i-g-t v2 2/5] lib/intel_compute_square_kernels: use stoppable loop for LNL/BMG
Date: Fri, 4 Apr 2025 14:31:37 +0200 [thread overview]
Message-ID: <20250404123140.260143-3-zbigniew.kempczynski@intel.com> (raw)
In-Reply-To: <20250404123140.260143-1-zbigniew.kempczynski@intel.com>
Instead of tweaked loop start using loop in which we may stop
it via simple cpu write to memory. Currently this is possible
for LNL and BMG platforms.
Signed-off-by: Zbigniew Kempczyński <zbigniew.kempczynski@intel.com>
Cc: Francois Dugast <francois.dugast@intel.com>
Cc: Priyanka Dandamudi <priyanka.dandamudi@intel.com>
---
lib/intel_compute_square_kernels.c | 41 ++++++++++++++++++++++++++++++
1 file changed, 41 insertions(+)
diff --git a/lib/intel_compute_square_kernels.c b/lib/intel_compute_square_kernels.c
index 76c48c4511..626dbc4cec 100644
--- a/lib/intel_compute_square_kernels.c
+++ b/lib/intel_compute_square_kernels.c
@@ -3844,6 +3844,43 @@ static const unsigned char xe2lpg_kernel_inc_bin[] = {
0x00, 0x00, 0x00, 0x00
};
+/*
+ * Opencl code is in opencl/loop.cl
+ *
+ * To work properly it requires to use uncached reads, so ocloc has to
+ * be called with: -options " -igc_opts 'LscLoadCacheControlOverride=1' arg
+*/
+
+static const unsigned char xe2lpg_kernel_loop_bin[] = {
+ 0x65, 0x00, 0x00, 0x80, 0x20, 0x82, 0x05, 0x7f, 0x04, 0x00, 0x00, 0x02,
+ 0xc0, 0xff, 0xff, 0xff, 0x40, 0x19, 0x00, 0x80, 0x20, 0x82, 0x05, 0x7f,
+ 0x04, 0x7f, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x31, 0x20, 0x01, 0x80,
+ 0x00, 0x00, 0x0c, 0x02, 0x8f, 0x7f, 0x00, 0xfa, 0x03, 0x00, 0x70, 0xf6,
+ 0x61, 0x00, 0x10, 0x2c, 0x01, 0x00, 0x10, 0x00, 0x66, 0x09, 0x00, 0x80,
+ 0x20, 0x82, 0x01, 0x80, 0x00, 0x80, 0x00, 0x02, 0xc0, 0x04, 0x00, 0x40,
+ 0x01, 0x09, 0x8c, 0x3c, 0x00, 0x00, 0x10, 0x00, 0x61, 0x80, 0x84, 0xa4,
+ 0x04, 0x02, 0x10, 0x00, 0x31, 0x21, 0x01, 0x80, 0x00, 0x00, 0x0c, 0x03,
+ 0x0c, 0x04, 0x00, 0xfb, 0x00, 0x00, 0xa0, 0x00, 0x70, 0x81, 0x14, 0x80,
+ 0x60, 0x86, 0x01, 0x00, 0x04, 0x03, 0x00, 0x16, 0x34, 0x12, 0x34, 0x12,
+ 0x20, 0x00, 0x00, 0x94, 0x00, 0x40, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+ 0xd8, 0xff, 0xff, 0xff, 0x61, 0x00, 0x10, 0x28, 0x7f, 0x01, 0x10, 0x00,
+ 0x31, 0x22, 0x02, 0x80, 0x04, 0x00, 0x00, 0x00, 0x0c, 0x7f, 0x20, 0x30,
+ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00
+};
+
unsigned char xelpg_kernel_square_bin[] = {
0x61, 0x00, 0x03, 0x80, 0x20, 0x42, 0x05, 0x7f, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x65, 0x00, 0x00, 0x80, 0x20, 0x82, 0x45, 0x7f,
@@ -6629,6 +6666,8 @@ const struct intel_compute_kernels intel_compute_square_kernels[] = {
.long_kernel_size = sizeof(xe2lpg_kernel_inc_bin),
.sip_kernel = xe2lpg_kernel_sip_bin,
.sip_kernel_size = sizeof(xe2lpg_kernel_sip_bin),
+ .loop_kernel = xe2lpg_kernel_loop_bin,
+ .loop_kernel_size = sizeof(xe2lpg_kernel_loop_bin),
},
{
.ip_ver = IP_VER(20, 04),
@@ -6638,6 +6677,8 @@ const struct intel_compute_kernels intel_compute_square_kernels[] = {
.long_kernel_size = sizeof(xe2lpg_kernel_inc_bin),
.sip_kernel = xe2lpg_kernel_sip_bin,
.sip_kernel_size = sizeof(xe2lpg_kernel_sip_bin),
+ .loop_kernel = xe2lpg_kernel_loop_bin,
+ .loop_kernel_size = sizeof(xe2lpg_kernel_loop_bin),
},
{
.ip_ver = IP_VER(30, 00),
--
2.34.1
next prev parent reply other threads:[~2025-04-04 12:31 UTC|newest]
Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-04-04 12:31 [PATCH i-g-t v2 0/5] Replace loop in WMTP tests Zbigniew Kempczyński
2025-04-04 12:31 ` [PATCH i-g-t v2 1/5] lib/intel_compute: add support for stoppable loop Zbigniew Kempczyński
2025-04-04 13:18 ` Francois Dugast
2025-04-04 12:31 ` Zbigniew Kempczyński [this message]
2025-04-04 12:31 ` [PATCH i-g-t v2 3/5] lib/intel_compute_square_kernel: add loop shader binary for PTL Zbigniew Kempczyński
2025-04-04 12:31 ` [PATCH i-g-t v2 4/5] tests/xe_compute_preempt: adjust number of children according to ram size Zbigniew Kempczyński
2025-04-07 8:49 ` Dandamudi, Priyanka
2025-04-07 19:01 ` Zbigniew Kempczyński
2025-04-08 12:12 ` Kamil Konieczny
2025-04-09 8:47 ` Dandamudi, Priyanka
2025-04-04 12:31 ` [PATCH i-g-t v2 5/5] tests/xe_compute_preempt: consume all ram for wmtp Zbigniew Kempczyński
2025-04-04 14:34 ` ✓ Xe.CI.BAT: success for Replace loop in WMTP tests (rev2) Patchwork
2025-04-04 14:55 ` ✓ i915.CI.BAT: " Patchwork
2025-04-04 17:35 ` ✗ i915.CI.Full: failure " 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=20250404123140.260143-3-zbigniew.kempczynski@intel.com \
--to=zbigniew.kempczynski@intel.com \
--cc=francois.dugast@intel.com \
--cc=igt-dev@lists.freedesktop.org \
--cc=priyanka.dandamudi@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