From: Johannes Gäßler <redacted>
Date: Thu, 27 Jun 2024 14:26:05 +0000 (+0200)
Subject: CUDA: fix MMQ stream-k for --split-mode row (#8167)
X-Git-Tag: upstream/0.0.4488~1238
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=85a267daaa1c6f8fd69160445bcb88717031d10c;p=pkg%2Fggml%2Fsources%2Fllama.cpp

CUDA: fix MMQ stream-k for --split-mode row (#8167)
---

diff --git a/ggml/src/ggml-cuda/mmq.cuh b/ggml/src/ggml-cuda/mmq.cuh
index 31fcbf13..1396e7a7 100644
--- a/ggml/src/ggml-cuda/mmq.cuh
+++ b/ggml/src/ggml-cuda/mmq.cuh
@@ -2475,7 +2475,7 @@ static void launch_mul_mat_q(ggml_backend_cuda_context & ctx, const mmq_args & a
 
     const dim3 block_nums_mmq(nsm, 1, 1);
 
-    ggml_cuda_pool & pool = ctx.pool();
+    ggml_cuda_pool & pool = ctx.pool(id);
     ggml_cuda_pool_alloc<float> tmp_fixup(pool, block_nums_mmq.x * mmq_x*mmq_y);
 
     if (args.ne01 % mmq_y == 0) {