From: Jeff Bolz <redacted>
Date: Thu, 23 Jan 2025 07:01:17 +0000 (-0600)
Subject: vulkan: fix diag_mask_inf (llama/11323)
X-Git-Tag: upstream/1.7.4+95~42
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=373670613961d0426d05626ff5ff8fff619a648d;p=pkg%2Fggml%2Fsources%2Fwhisper.cpp

vulkan: fix diag_mask_inf (llama/11323)

With robustbufferaccess disabled, this shader was showing OOB stores. There
is a bounds check in the code, but the workgrouop dimensions were reversed vs
CUDA and it was running the wrong number of threads. So fix the workgroup
dimensions and disable robustness for this pipeline.
---

diff --git a/ggml/src/ggml-vulkan/ggml-vulkan.cpp b/ggml/src/ggml-vulkan/ggml-vulkan.cpp
index 78c2f5c4..eae4f073 100644
--- a/ggml/src/ggml-vulkan/ggml-vulkan.cpp
+++ b/ggml/src/ggml-vulkan/ggml-vulkan.cpp
@@ -2012,7 +2012,7 @@ static void ggml_vk_load_shaders(vk_device& device) {
     ggml_vk_create_pipeline(device, device->pipeline_leaky_relu_f32, "leaky_relu_f32", leaky_relu_f32_len, leaky_relu_f32_data, "main", 2, sizeof(vk_op_push_constants), {512, 1, 1}, {}, 1);
     ggml_vk_create_pipeline(device, device->pipeline_tanh_f32, "tanh_f32", tanh_f32_len, tanh_f32_data, "main", 2, sizeof(vk_op_push_constants), {512, 1, 1}, {}, 1);
 
-    ggml_vk_create_pipeline(device, device->pipeline_diag_mask_inf_f32, "diag_mask_inf_f32", diag_mask_inf_f32_len, diag_mask_inf_f32_data, "main", 2, sizeof(vk_op_diag_mask_push_constants), {512, 1, 1}, {}, 1);
+    ggml_vk_create_pipeline(device, device->pipeline_diag_mask_inf_f32, "diag_mask_inf_f32", diag_mask_inf_f32_len, diag_mask_inf_f32_data, "main", 2, sizeof(vk_op_diag_mask_push_constants), {1, 512, 1}, {}, 1, true);
 
     ggml_vk_create_pipeline(device, device->pipeline_soft_max_f32, "soft_max_f32", soft_max_f32_len, soft_max_f32_data, "main", 3, sizeof(vk_op_soft_max_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
     ggml_vk_create_pipeline(device, device->pipeline_soft_max_f32_wg512, "soft_max_f32_wg512", soft_max_f32_len, soft_max_f32_data, "main", 3, sizeof(vk_op_soft_max_push_constants), {1, 1, 1}, { 512 }, 1);
diff --git a/ggml/src/ggml-vulkan/vulkan-shaders/diag_mask_inf.comp b/ggml/src/ggml-vulkan/vulkan-shaders/diag_mask_inf.comp
index 4e68742b..26d8bc22 100644
--- a/ggml/src/ggml-vulkan/vulkan-shaders/diag_mask_inf.comp
+++ b/ggml/src/ggml-vulkan/vulkan-shaders/diag_mask_inf.comp
@@ -12,7 +12,7 @@ layout (push_constant) uniform parameter
 
 #include "types.comp"
 
-layout(local_size_x = 512, local_size_y = 1, local_size_z = 1) in;
+layout(local_size_x = 1, local_size_y = 512, local_size_z = 1) in;
 
 layout (binding = 0) readonly buffer X {A_TYPE data_a[];};
 layout (binding = 1) writeonly buffer D {D_TYPE data_d[];};