0028-CUDA-Changing-the-CUDA-scheduling-strategy-to-spin-1.patch 1.64 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
From: Julius Tischbein <ju.tischbein@gmail.com>
Date: Wed, 15 Oct 2025 13:54:15 +0200
Subject: [PATCH] CUDA: Changing the CUDA scheduling strategy to spin (#16585)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* CUDA set scheduling strategy to spinning for cc121

* Using prop.major and prop.minor, include HIP and MUSA

* Exclude HIP and MUSA

* Remove trailing whitespace

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>

* Remove empty line

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>

---------

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
---
 ggml/src/ggml-cuda/ggml-cuda.cu | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/ggml/src/ggml-cuda/ggml-cuda.cu b/ggml/src/ggml-cuda/ggml-cuda.cu
31
index b075a18be..d62f412d6 100644
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
--- a/ggml/src/ggml-cuda/ggml-cuda.cu
+++ b/ggml/src/ggml-cuda/ggml-cuda.cu
@@ -340,6 +340,15 @@ static ggml_cuda_device_info ggml_cuda_init() {
         } else if (device_name.substr(0, 21) == "NVIDIA GeForce GTX 16") {
             turing_devices_without_mma.push_back({ id, device_name });
         }
+
+        // Temporary performance fix:
+        // Setting device scheduling strategy for iGPUs with cc121 to "spinning" to avoid delays in cuda synchronize calls.
+        // TODO: Check for future drivers the default scheduling strategy and
+        // remove this call again when cudaDeviceScheduleSpin is default.
+        if (prop.major == 12 && prop.minor == 1) {
+            CUDA_CHECK(cudaSetDeviceFlags(cudaDeviceScheduleSpin));
+        }
+
 #endif  // defined(GGML_USE_HIP)
     }