1 files changed, 123 insertions, 0 deletions
diff --git a/aux-files/stable-diffusion-cpp/vulkan-support.patch b/aux-files/stable-diffusion-cpp/vulkan-support.patch
new file mode 100644
index 0000000..88f5fbf
--- /dev/null
+++ b/aux-files/stable-diffusion-cpp/vulkan-support.patch
@@ -0,0 +1,123 @@
+From 15553084e5d1a7d6abd3aac176bee95b0247c379 Mon Sep 17 00:00:00 2001
+From: sohzm <sohxm7@gmail.com>
+Date: Wed, 19 Jun 2024 04:08:18 +0530
+Subject: [PATCH] wip
+
+---
+ CMakeLists.txt       | 7 +++++++
+ ggml                 | 2 +-
+ ggml_extend.hpp      | 6 +++++-
+ model.cpp            | 4 ++++
+ stable-diffusion.cpp | 6 +++++-
+ upscaler.cpp         | 4 ++++
+ 6 files changed, 26 insertions(+), 3 deletions(-)
+
+diff --git a/CMakeLists.txt b/CMakeLists.txt
+index 28a03fb..6b5679f 100644
+--- a/CMakeLists.txt
++++ b/CMakeLists.txt
+@@ -27,6 +27,7 @@ option(SD_BUILD_EXAMPLES             "sd: build examples" ${SD_STANDALONE})
+ option(SD_CUBLAS                     "sd: cuda backend" OFF)
+ option(SD_HIPBLAS                    "sd: rocm backend" OFF)
+ option(SD_METAL                      "sd: metal backend" OFF)
++option(SD_VULKAN                     "sd: vulkan backend" OFF)
+ option(SD_FLASH_ATTN                 "sd: use flash attention for x4 less memory usage" OFF)
+ option(SD_FAST_SOFTMAX               "sd: x1.5 faster softmax, indeterministic (sometimes, same seed don't generate same image), cuda only" OFF)
+ option(SD_BUILD_SHARED_LIBS          "sd: build shared libs" OFF)
+@@ -44,6 +45,12 @@ if(SD_METAL)
+     add_definitions(-DSD_USE_METAL)
+ endif()
+ 
++if (SD_VULKAN)
++    message("Use Vulkan as backend stable-diffusion")
++    set(GGML_VULKAN ON)
++    add_definitions(-DSD_USE_VULKAN)
++endif ()
++
+ if (SD_HIPBLAS)
+     message("Use HIPBLAS as backend stable-diffusion")
+     set(GGML_HIPBLAS ON)
+diff --git a/ggml b/ggml
+index 9d562d7..5653a19 160000
+--- a/ggml
++++ b/ggml
+@@ -1 +1 @@
+-Subproject commit 9d562d712513c77a4de44ad0428be62bc3f2a9cf
++Subproject commit 5653a195935ea3ac54652644c9daf154dbc1571b
+diff --git a/ggml_extend.hpp b/ggml_extend.hpp
+index dbe9303..1236996 100644
+--- a/ggml_extend.hpp
++++ b/ggml_extend.hpp
+@@ -32,6 +32,10 @@
+ #include "ggml-metal.h"
+ #endif
+ 
++#ifdef SD_USE_VULKAN
++#include "ggml-vulkan.h"
++#endif
++
+ #include "rng.hpp"
+ #include "util.h"
+ 
+@@ -588,7 +592,7 @@ __STATIC_INLINE__ struct ggml_tensor* ggml_nn_attention(struct ggml_context* ctx
+                                                         struct ggml_tensor* k,
+                                                         struct ggml_tensor* v,
+                                                         bool mask = false) {
+-#if defined(SD_USE_FLASH_ATTENTION) && !defined(SD_USE_CUBLAS) && !defined(SD_USE_METAL)
++#if defined(SD_USE_FLASH_ATTENTION) && !defined(SD_USE_CUBLAS) && !defined(SD_USE_METAL) && !defined(SD_USE_VULKAN)
+     struct ggml_tensor* kqv = ggml_flash_attn(ctx, q, k, v, false);  // [N * n_head, n_token, d_head]
+ #else
+     float d_head = (float)q->ne[0];
+diff --git a/model.cpp b/model.cpp
+index c4556a9..db8bae8 100644
+--- a/model.cpp
++++ b/model.cpp
+@@ -21,6 +21,10 @@
+ #include "ggml-metal.h"
+ #endif
+ 
++#ifdef SD_USE_VULKAN
++#include "ggml-vulkan.h"
++#endif
++
+ #define ST_HEADER_SIZE_LEN 8
+ 
+ uint64_t read_u64(uint8_t* buffer) {
+diff --git a/stable-diffusion.cpp b/stable-diffusion.cpp
+index 8e439d2..3521e76 100644
+--- a/stable-diffusion.cpp
++++ b/stable-diffusion.cpp
+@@ -154,13 +154,17 @@ class StableDiffusionGGML {
+         ggml_backend_metal_log_set_callback(ggml_log_callback_default, nullptr);
+         backend = ggml_backend_metal_init();
+ #endif
++#ifdef SD_USE_VULKAN
++        LOG_DEBUG("Using Vulkan backend");
++        backend = ggml_backend_vk_init();
++#endif
+ 
+         if (!backend) {
+             LOG_DEBUG("Using CPU backend");
+             backend = ggml_backend_cpu_init();
+         }
+ #ifdef SD_USE_FLASH_ATTENTION
+-#if defined(SD_USE_CUBLAS) || defined(SD_USE_METAL)
++#if defined(SD_USE_CUBLAS) || defined(SD_USE_METAL) || defined(SD_USE_VULKAN)
+         LOG_WARN("Flash Attention not supported with GPU Backend");
+ #else
+         LOG_INFO("Flash Attention enabled");
+diff --git a/upscaler.cpp b/upscaler.cpp
+index 0e3f95d..7623f9b 100644
+--- a/upscaler.cpp
++++ b/upscaler.cpp
+@@ -24,6 +24,10 @@ struct UpscalerGGML {
+         ggml_backend_metal_log_set_callback(ggml_log_callback_default, nullptr);
+         backend = ggml_backend_metal_init();
+ #endif
++#ifdef SD_USE_VULKAN
++        LOG_DEBUG("Using Vulkan backend");
++        backend = ggml_backend_vk_init(0);
++#endif
+ 
+         if (!backend) {
+             LOG_DEBUG("Using CPU backend");