mirrors/ollama-for-amd
1
0
Fork 0
mirror of https://github.com/likelovewant/ollama-for-amd.git synced 2025-12-21 14:26:30 +00:00
Code Issues Packages Projects Releases Wiki Activity

feat: llama.cpp bump (17f7f4) for SSM performance improvements (#13408)

Browse Source 
* feat: Bump llama.cpp to the latest master (17f7f4b)

This brings in significant improvements to prefill performance for all
models using the SSM_CONV and SSM_SCAN ops (granite4, jamba, falcon-h,
nemotron-h, Qwen3 Next) on Apple Metal.

See https://github.com/ggml-org/llama.cpp/pull/17876

Branch: LlamaCPPMetalSSMImprovements

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* feat: Update patches 1-4

Branch: LlamaCPPMetalSSMImprovements

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix: Update patches 5-12

Branch: LlamaCPPMetalSSMImprovements

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* feat: Update patches 13-18

Branch: LlamaCPPMetalSSMImprovements

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* feat: Update patch 20

Branch: LlamaCPPMetalSSMImprovements

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* feat: Update patches 21-31

Branch: LlamaCPPMetalSSMImprovements

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* feat: Sync vendored code

The two files I'm not sure about here are the swap from gemma3-iswa.cpp to
gemma3.cpp (I chose to include this because I think it's required), and the
inclusion of `ggml-zendnn.h` which I chose to omit.

Branch: LlamaCPPMetalSSMImprovements

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

---------

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
This commit is contained in:
Gabe Goodhart
2025-12-10 13:59:27 -07:00
committed by GitHub
parent c34fc64688
commit b95693056c
115 changed files with 5176 additions and 2585 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
llama/patches/0028-Add-memory-detection-using-DXGI-PDH.patch
View File

@@ -12,10 +12,10 @@ Subject: [PATCH] Add memory detection using DXGI + PDH
create mode 100644 ggml/src/mem_dxgi_pdh.cpp
diff --git a/ggml/src/CMakeLists.txt b/ggml/src/CMakeLists.txt
index ac8f38464..faa1beed2 100644
index 99ae293cc..9a134b7af 100644
--- a/ggml/src/CMakeLists.txt
+++ b/ggml/src/CMakeLists.txt
@@ -211,6 +211,7 @@ add_library(ggml-base
@@ -207,6 +207,7 @@ add_library(ggml-base
ggml-quants.h
mem_hip.cpp
mem_nvml.cpp
@@ -38,7 +38,7 @@ index 1c07e767a..0da3e065b 100644
#ifdef __cplusplus
}
diff --git a/ggml/src/ggml-vulkan/ggml-vulkan.cpp b/ggml/src/ggml-vulkan/ggml-vulkan.cpp
index a234eda2e..c98f98c73 100644
index d9f4d34f5..8a83427fb 100644
--- a/ggml/src/ggml-vulkan/ggml-vulkan.cpp
+++ b/ggml/src/ggml-vulkan/ggml-vulkan.cpp
@@ -74,6 +74,7 @@ DispatchLoaderDynamic & ggml_vk_default_dispatcher();
@@ -49,7 +49,7 @@ index a234eda2e..c98f98c73 100644
typedef struct VkPhysicalDeviceShaderBfloat16FeaturesKHR {
VkStructureType sType;
@@ -13655,6 +13656,7 @@ struct ggml_backend_vk_device_context {
@@ -13576,6 +13577,7 @@ struct ggml_backend_vk_device_context {
std::string pci_id;
std::string id;
std::string uuid;
@@ -57,7 +57,7 @@ index a234eda2e..c98f98c73 100644
int major;
int minor;
int driver_major;
@@ -13673,6 +13675,20 @@ void ggml_backend_vk_get_device_memory(ggml_backend_vk_device_context *ctx, size
@@ -13594,6 +13596,20 @@ void ggml_backend_vk_get_device_memory(ggml_backend_vk_device_context *ctx, size
vk::PhysicalDeviceProperties2 props2;
vkdev.getProperties2(&props2);
@@ -78,7 +78,7 @@ index a234eda2e..c98f98c73 100644
if (!is_integrated_gpu)
{
@@ -13704,7 +13720,6 @@ void ggml_backend_vk_get_device_memory(ggml_backend_vk_device_context *ctx, size
@@ -13625,7 +13641,6 @@ void ggml_backend_vk_get_device_memory(ggml_backend_vk_device_context *ctx, size
}
// else fallback to memory budget if supported
@@ -86,7 +86,7 @@ index a234eda2e..c98f98c73 100644
if (membudget_supported) {
memprops.pNext = &budgetprops;
}
@@ -14440,7 +14455,6 @@ static ggml_backend_dev_t ggml_backend_vk_reg_get_device(ggml_backend_reg_t reg,
@@ -14357,7 +14372,6 @@ static ggml_backend_dev_t ggml_backend_vk_reg_get_device(ggml_backend_reg_t reg,
/* .reg = */ reg,
/* .context = */ ctx,
});
@@ -94,7 +94,7 @@ index a234eda2e..c98f98c73 100644
// Gather additional information about the device
int dev_idx = vk_instance.device_indices[i];
vk::PhysicalDeviceProperties props1;
@@ -14463,6 +14477,14 @@ static ggml_backend_dev_t ggml_backend_vk_reg_get_device(ggml_backend_reg_t reg,
@@ -14380,6 +14394,14 @@ static ggml_backend_dev_t ggml_backend_vk_reg_get_device(ggml_backend_reg_t reg,
}
}
ctx->uuid = oss.str();