Commit 4bc272fd7 for llama.cpp

commit 4bc272fd729bd094c0422e4b8353da8d2fec91f8
Author: Jeff Bolz <jbolz@nvidia.com>
Date:   Wed Sep 16 18:40:07 2026 -0500

    vulkan: work around NV bug with argsort_large.comp (#28975)

diff --git a/ggml/src/ggml-vulkan/vulkan-shaders/argsort_large.comp b/ggml/src/ggml-vulkan/vulkan-shaders/argsort_large.comp
index b2df44137..f6a29be29 100644
--- a/ggml/src/ggml-vulkan/vulkan-shaders/argsort_large.comp
+++ b/ggml/src/ggml-vulkan/vulkan-shaders/argsort_large.comp
@@ -27,6 +27,8 @@ layout (push_constant) uniform parameter {
     uint inner_end;
 } p;

+shared int s;
+
 void argsort(bool needs_bounds_check, const uint row) {
     // bitonic sort
     int col = int(gl_GlobalInvocationID.x);
@@ -35,6 +37,12 @@ void argsort(bool needs_bounds_check, const uint row) {
     const uint row_offset = row * p.ncols;
     uint idx_offset = row * p.ncols_padded;

+    // workaround for NV driver/compiler bug - dummy use of shared memory
+    if (gl_LocalInvocationIndex == 0) {
+        s = 0;
+        col += s;
+    }
+
     bool need_barrier = false;

     // initialize indices