Commit 301db6337 for llama.cpp

commit 301db633702f1d2c107d4feaa4268f07c0b508c5
Author: Ruben Ortlam <rortlam@redhat.com>
Date:   Mon Nov 16 14:22:57 2026 +0100

    vulkan: fix Nvidia fp32 FA shmem overflow, disable shmem staging if it takes too much memory (#32244)

diff --git a/ggml/src/ggml-vulkan/ggml-vulkan.cpp b/ggml/src/ggml-vulkan/ggml-vulkan.cpp
index 1c0775307..f07135351 100644
--- a/ggml/src/ggml-vulkan/ggml-vulkan.cpp
+++ b/ggml/src/ggml-vulkan/ggml-vulkan.cpp
@@ -1258,10 +1258,14 @@ static vk_fa_tuning_params get_fa_tuning_params_scalar(const vk_device& device,

     result.shmem_staging = (device->vendor_id == VK_VENDOR_ID_NVIDIA && hsk < 256 && hsv < 256) ? 1 : 0;

-    if (!reduce_block_rows && !ggml_vk_flash_attn_scalar_shmem_support(device, result, hsk, hsv, f32acc, k_type, v_type)) {
+    if (!reduce_block_rows && result.block_rows > 1 && !ggml_vk_flash_attn_scalar_shmem_support(device, result, hsk, hsv, f32acc, k_type, v_type)) {
         result.block_rows /= 2;
     }

+    if (result.shmem_staging && !ggml_vk_flash_attn_scalar_shmem_support(device, result, hsk, hsv, f32acc, k_type, v_type)) {
+        result.shmem_staging = 0;
+    }
+
     // With head sizes above 128, 16 rows per workgroup spill registers on Qualcomm and run about
     // 40x slower; at large KV a single dispatch can then exceed the Windows GPU timeout.
     if (device->vendor_id == VK_VENDOR_ID_QUALCOMM && (hsk > 128 || hsv > 128)) {
@@ -8528,7 +8532,9 @@ bool ggml_vk_flash_attn_scalar_shmem_support(const vk_device& device, const vk_f
         kblocksh_size = 0;
     }

-    const uint32_t total_size = tmpsh + tmpshv4 + masksh + iq_shmem + Qf + kvsh + kblocksh_size;
+    const uint32_t occupancy_limiter = std::max(params.limit_occupancy_shmem, 1u) * 4 * sizeof(float);
+
+    const uint32_t total_size = tmpsh + tmpshv4 + masksh + iq_shmem + Qf + kvsh + kblocksh_size + occupancy_limiter;
     const bool supported = total_size <= device->properties.limits.maxComputeSharedMemorySize;

     VK_LOG_DEBUG("ggml_vk_flash_attn_scalar_shmem_support(HSK=" << hsk << ", HSV=" << hsv << ", mmq=" << mmq << ", total_size=" << total_size << ", supported=" << supported);