mirror of
https://github.com/dogkeeper886/ollama37.git
synced 2025-12-09 23:37:06 +00:00
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with comprehensive build instructions - Documented toolchain constraints and CPU architecture trade-offs - Removed outdated CI/CD workflows (tesla-k80-*.yml) - Cleaned up temporary development artifacts ## Rationale This fork maintains Tesla K80 GPU support (compute 3.7) which was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>
74 lines
1.9 KiB
Go
74 lines
1.9 KiB
Go
package discover
|
|
|
|
import (
|
|
"log/slog"
|
|
"os"
|
|
"regexp"
|
|
"runtime"
|
|
"strconv"
|
|
"strings"
|
|
|
|
"github.com/ollama/ollama/ml"
|
|
)
|
|
|
|
// Jetson devices have JETSON_JETPACK="x.y.z" factory set to the Jetpack version installed.
|
|
// Included to drive logic for reducing Ollama-allocated overhead on L4T/Jetson devices.
|
|
var CudaTegra string = os.Getenv("JETSON_JETPACK")
|
|
|
|
// GetSystemInfo returns the last cached state of the GPUs on the system
|
|
func GetSystemInfo() ml.SystemInfo {
|
|
memInfo, err := GetCPUMem()
|
|
if err != nil {
|
|
slog.Warn("error looking up system memory", "error", err)
|
|
}
|
|
var threadCount int
|
|
cpus := GetCPUDetails()
|
|
for _, c := range cpus {
|
|
threadCount += c.CoreCount - c.EfficiencyCoreCount
|
|
}
|
|
|
|
if threadCount == 0 {
|
|
// Fall back to Go's num CPU
|
|
threadCount = runtime.NumCPU()
|
|
}
|
|
|
|
return ml.SystemInfo{
|
|
ThreadCount: threadCount,
|
|
TotalMemory: memInfo.TotalMemory,
|
|
FreeMemory: memInfo.FreeMemory,
|
|
FreeSwap: memInfo.FreeSwap,
|
|
}
|
|
}
|
|
|
|
func cudaJetpack() string {
|
|
if runtime.GOARCH == "arm64" && runtime.GOOS == "linux" {
|
|
if CudaTegra != "" {
|
|
ver := strings.Split(CudaTegra, ".")
|
|
if len(ver) > 0 {
|
|
return "jetpack" + ver[0]
|
|
}
|
|
} else if data, err := os.ReadFile("/etc/nv_tegra_release"); err == nil {
|
|
r := regexp.MustCompile(` R(\d+) `)
|
|
m := r.FindSubmatch(data)
|
|
if len(m) != 2 {
|
|
slog.Info("Unexpected format for /etc/nv_tegra_release. Set JETSON_JETPACK to select version")
|
|
} else {
|
|
if l4t, err := strconv.Atoi(string(m[1])); err == nil {
|
|
// Note: mapping from L4t -> JP is inconsistent (can't just subtract 30)
|
|
// https://developer.nvidia.com/embedded/jetpack-archive
|
|
switch l4t {
|
|
case 35:
|
|
return "jetpack5"
|
|
case 36:
|
|
return "jetpack6"
|
|
default:
|
|
// Newer Jetson systems use the SBSU runtime
|
|
slog.Debug("unrecognized L4T version", "nv_tegra_release", string(data))
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
return ""
|
|
}
|