mirror of
https://github.com/dogkeeper886/ollama37.git
synced 2025-12-10 15:57:04 +00:00
ef14fb5b26
This commit represents a complete rework after pulling the latest changes from official ollama/ollama repository and re-applying Tesla K80 compatibility patches. ## Key Changes ### CUDA Compute Capability 3.7 Support (Tesla K80) - Added sm_37 (compute 3.7) to CMAKE_CUDA_ARCHITECTURES in CMakeLists.txt - Updated CMakePresets.json to include compute 3.7 in "CUDA 11" preset - Using 37-virtual (PTX with JIT compilation) for maximum compatibility ### Legacy Toolchain Compatibility - **NVIDIA Driver**: 470.256.02 (last version supporting Kepler/K80) - **CUDA Version**: 11.4.4 (last CUDA 11.x supporting compute 3.7) - **GCC Version**: 10.5.0 (required by CUDA 11.4 host_config.h) ### CPU Architecture Trade-offs Due to GCC 10.5 limitation, sacrificed newer CPU optimizations: - Alderlake CPU variant enabled WITHOUT AVX_VNNI (requires GCC 11+) - Still supports: SSE4.2, AVX, F16C, AVX2, BMI2, FMA - Performance impact: ~3-7% on newer CPUs (acceptable for K80 compatibility) ### Build System Updates - Modified ml/backend/ggml/ggml/src/ggml-cuda/CMakeLists.txt for compute 3.7 - Added -Wno-deprecated-gpu-targets flag to suppress warnings - Updated ml/backend/ggml/ggml/src/CMakeLists.txt for Alderlake without AVX_VNNI ### Upstream Sync Merged latest llama.cpp changes including: - Enhanced KV cache management with ISWA and hybrid memory support - Improved multi-modal support (mtmd framework) - New model architectures (Gemma3, Llama4, Qwen3, etc.) - GPU backend improvements for CUDA, Metal, and ROCm - Updated quantization support and GGUF format handling ### Documentation - Updated CLAUDE.md with comprehensive build instructions - Documented toolchain constraints and CPU architecture trade-offs - Removed outdated CI/CD workflows (tesla-k80-*.yml) - Cleaned up temporary development artifacts ## Rationale This fork maintains Tesla K80 GPU support (compute 3.7) which was dropped in official Ollama due to legacy driver/CUDA requirements. The toolchain constraint creates a deadlock: - K80 → Driver 470 → CUDA 11.4 → GCC 10 → No AVX_VNNI We accept the loss of cutting-edge CPU optimizations to enable running modern LLMs on legacy but still capable Tesla K80 hardware (12GB VRAM per GPU). 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
86 lines
3.2 KiB
Diff
86 lines
3.2 KiB
Diff
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
|
|
From: jmorganca <jmorganca@gmail.com>
|
|
Date: Tue, 8 Apr 2025 20:31:38 -0700
|
|
Subject: [PATCH] sort devices by score
|
|
|
|
in the ggml backend loading code, devices
|
|
are now sorted by score, ensuring the device
|
|
with the fastest acceleration is loaded
|
|
---
|
|
ggml/src/ggml-backend-reg.cpp | 21 +++++++++++++--------
|
|
1 file changed, 13 insertions(+), 8 deletions(-)
|
|
|
|
diff --git a/ggml/src/ggml-backend-reg.cpp b/ggml/src/ggml-backend-reg.cpp
|
|
index 136afec7..f794d9cf 100644
|
|
--- a/ggml/src/ggml-backend-reg.cpp
|
|
+++ b/ggml/src/ggml-backend-reg.cpp
|
|
@@ -175,7 +175,7 @@ struct ggml_backend_reg_entry {
|
|
|
|
struct ggml_backend_registry {
|
|
std::vector<ggml_backend_reg_entry> backends;
|
|
- std::vector<ggml_backend_dev_t> devices;
|
|
+ std::vector<std::pair<ggml_backend_dev_t, int>> devices;
|
|
|
|
ggml_backend_registry() {
|
|
#ifdef GGML_USE_CUDA
|
|
@@ -223,7 +223,7 @@ struct ggml_backend_registry {
|
|
}
|
|
}
|
|
|
|
- void register_backend(ggml_backend_reg_t reg, dl_handle_ptr handle = nullptr) {
|
|
+ void register_backend(ggml_backend_reg_t reg, int score = -1, dl_handle_ptr handle = nullptr) {
|
|
if (!reg) {
|
|
return;
|
|
}
|
|
@@ -234,15 +234,20 @@ struct ggml_backend_registry {
|
|
#endif
|
|
backends.push_back({ reg, std::move(handle) });
|
|
for (size_t i = 0; i < ggml_backend_reg_dev_count(reg); i++) {
|
|
- register_device(ggml_backend_reg_dev_get(reg, i));
|
|
+ register_device(ggml_backend_reg_dev_get(reg, i), score);
|
|
}
|
|
}
|
|
|
|
- void register_device(ggml_backend_dev_t device) {
|
|
+ void register_device(ggml_backend_dev_t device, int score = -1) {
|
|
#ifndef NDEBUG
|
|
GGML_LOG_DEBUG("%s: registered device %s (%s)\n", __func__, ggml_backend_dev_name(device), ggml_backend_dev_description(device));
|
|
#endif
|
|
- devices.push_back(device);
|
|
+ devices.push_back({device, score});
|
|
+ std::stable_sort(devices.begin(), devices.end(),
|
|
+ [](const auto & a, const auto & b) {
|
|
+ return a.second > b.second;
|
|
+ }
|
|
+ );
|
|
}
|
|
|
|
ggml_backend_reg_t load_backend(const fs::path & path, bool silent) {
|
|
@@ -286,7 +291,7 @@ struct ggml_backend_registry {
|
|
|
|
GGML_LOG_INFO("%s: loaded %s backend from %s\n", __func__, ggml_backend_reg_name(reg), path_str(path).c_str());
|
|
|
|
- register_backend(reg, std::move(handle));
|
|
+ register_backend(reg, score_fn ? score_fn() : -1, std::move(handle));
|
|
|
|
return reg;
|
|
}
|
|
@@ -309,7 +314,7 @@ struct ggml_backend_registry {
|
|
// remove devices
|
|
devices.erase(
|
|
std::remove_if(devices.begin(), devices.end(),
|
|
- [reg](ggml_backend_dev_t dev) { return ggml_backend_dev_backend_reg(dev) == reg; }),
|
|
+ [reg](std::pair<ggml_backend_dev_t, int> dev) { return ggml_backend_dev_backend_reg(dev.first) == reg; }),
|
|
devices.end());
|
|
|
|
// remove backend
|
|
@@ -367,7 +372,7 @@ size_t ggml_backend_dev_count() {
|
|
|
|
ggml_backend_dev_t ggml_backend_dev_get(size_t index) {
|
|
GGML_ASSERT(index < ggml_backend_dev_count());
|
|
- return get_reg().devices[index];
|
|
+ return get_reg().devices[index].first;
|
|
}
|
|
|
|
ggml_backend_dev_t ggml_backend_dev_by_name(const char * name) {
|