pali112/ollama
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge remote-tracking branch 'upstream/main' into vulkanV3

Browse Source
This commit is contained in:
Inforithmics
2025-08-09 21:59:30 +02:00
parent 45dbd14645 79f6376f5b
commit f8ed1541ed
471 changed files with 209409 additions and 46508 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
discover/amd_linux.go
View File

@@ -58,7 +58,7 @@ func AMDGetGPUInfo() ([]RocmGPUInfo, error) {
driverMajor, driverMinor, err := AMDDriverVersion()
if err != nil {
// TODO - if we see users crash and burn with the upstreamed kernel this can be adjusted to hard-fail rocm support and fallback to CPU
slog.Warn("ollama recommends running the https://www.amd.com/en/support/linux-drivers", "error", err)
slog.Warn("ollama recommends running the https://www.amd.com/en/support/download/linux-drivers.html", "error", err)
}
// Determine if the user has already pre-selected which GPUs to look at, then ignore the others

2
discover/cpu_common.go
View File

@@ -12,7 +12,7 @@ func IsNUMA() bool {
// numa support in llama.cpp is linux only
return false
}
ids := map[string]interface{}{}
ids := map[string]any{}
packageIds, _ := filepath.Glob("/sys/devices/system/cpu/cpu*/topology/physical_package_id")
for _, packageId := range packageIds {
id, err := os.ReadFile(packageId)

4
discover/cuda_common.go
View File

@@ -3,6 +3,7 @@
package discover
import (
"fmt"
"log/slog"
"os"
"regexp"
@@ -55,10 +56,13 @@ func cudaVariant(gpuInfo CudaGPUInfo) string {
}
}
}
return "sbsa"
}
// driver 12.0 has problems with the cuda v12 library, so run v11 on those older drivers
if gpuInfo.DriverMajor < 12 || (gpuInfo.DriverMajor == 12 && gpuInfo.DriverMinor == 0) {
// The detected driver is older than Feb 2023
slog.Warn("old CUDA driver detected - please upgrade to a newer driver", "version", fmt.Sprintf("%d.%d", gpuInfo.DriverMajor, gpuInfo.DriverMinor))
return "v11"
}
return "v12"

2
discover/gpu.go
View File

@@ -785,7 +785,7 @@ func LoadVulkanMgmt(vulkanLibPaths []string, capLibPaths []string) (int, *C.vk_h
}
func getVerboseState() C.uint16_t {
if envconfig.Debug() {
if envconfig.LogLevel() < slog.LevelInfo {
return C.uint16_t(1)
}
return C.uint16_t(0)

2
discover/gpu_info.h
View File

@@ -27,12 +27,14 @@
#endif
#ifndef LOG
#define LOG(verbose, ...) \
do { \
if (verbose) { \
fprintf(stderr, __VA_ARGS__); \
} \
} while (0)
#endif
#ifdef __cplusplus
extern "C" {

11
discover/gpu_info_cudart.c
View File

@@ -1,6 +1,7 @@
#ifndef __APPLE__ // TODO - maybe consider nvidia support on intel macs?
#include <string.h>
#include <inttypes.h>
#include "gpu_info_cudart.h"
void cudart_init(char *cudart_lib_path, cudart_init_resp_t *resp) {
@@ -58,7 +59,7 @@ void cudart_init(char *cudart_lib_path, cudart_init_resp_t *resp) {
LOG(resp->ch.verbose, "cudaSetDevice err: %d\n", ret);
UNLOAD_LIBRARY(resp->ch.handle);
resp->ch.handle = NULL;
if (ret == CUDA_ERROR_INSUFFICIENT_DRIVER) {
if (ret == CUDART_ERROR_INSUFFICIENT_DRIVER) {
resp->err = strdup("your nvidia driver is too old or missing. If you have a CUDA GPU please upgrade to run ollama");
return;
}
@@ -168,9 +169,9 @@ void cudart_bootstrap(cudart_handle_t h, int i, mem_info_t *resp) {
resp->free = memInfo.free;
resp->used = memInfo.used;
LOG(h.verbose, "[%s] CUDA totalMem %lu\n", resp->gpu_id, resp->total);
LOG(h.verbose, "[%s] CUDA freeMem %lu\n", resp->gpu_id, resp->free);
LOG(h.verbose, "[%s] CUDA usedMem %lu\n", resp->gpu_id, resp->used);
LOG(h.verbose, "[%s] CUDA totalMem %" PRId64 "\n", resp->gpu_id, resp->total);
LOG(h.verbose, "[%s] CUDA freeMem %" PRId64 "\n", resp->gpu_id, resp->free);
LOG(h.verbose, "[%s] CUDA usedMem %" PRId64 "\n", resp->gpu_id, resp->used);
LOG(h.verbose, "[%s] Compute Capability %d.%d\n", resp->gpu_id, resp->major, resp->minor);
}
@@ -180,4 +181,4 @@ void cudart_release(cudart_handle_t h) {
h.handle = NULL;
}
#endif // __APPLE__
#endif // __APPLE__

7
discover/gpu_info_nvcuda.c
View File

@@ -1,6 +1,7 @@
#ifndef __APPLE__ // TODO - maybe consider nvidia support on intel macs?
#include <string.h>
#include <inttypes.h>
#include "gpu_info_nvcuda.h"
void nvcuda_init(char *nvcuda_lib_path, nvcuda_init_resp_t *resp) {
@@ -193,8 +194,8 @@ void nvcuda_bootstrap(nvcuda_handle_t h, int i, mem_info_t *resp) {
resp->total = memInfo.total;
resp->free = memInfo.free;
LOG(h.verbose, "[%s] CUDA totalMem %lu mb\n", resp->gpu_id, resp->total / 1024 / 1024);
LOG(h.verbose, "[%s] CUDA freeMem %lu mb\n", resp->gpu_id, resp->free / 1024 / 1024);
LOG(h.verbose, "[%s] CUDA totalMem %" PRId64 "mb\n", resp->gpu_id, resp->total / 1024 / 1024);
LOG(h.verbose, "[%s] CUDA freeMem %" PRId64 "mb\n", resp->gpu_id, resp->free / 1024 / 1024);
LOG(h.verbose, "[%s] Compute Capability %d.%d\n", resp->gpu_id, resp->major, resp->minor);
@@ -247,4 +248,4 @@ void nvcuda_release(nvcuda_handle_t h) {
h.handle = NULL;
}
#endif // __APPLE__
#endif // __APPLE__

5
discover/gpu_linux.go
View File

@@ -129,6 +129,7 @@ func GetCPUDetails() ([]CPU, error) {
if err != nil {
return nil, err
}
defer file.Close()
return linuxCPUDetails(file)
}
@@ -186,13 +187,11 @@ func linuxCPUDetails(file io.Reader) ([]CPU, error) {
for id, s := range socketByID {
s.CoreCount = len(coreBySocket[id])
s.ThreadCount = 0
for _, tc := range threadsByCoreBySocket[id] {
s.ThreadCount += tc
}
// This only works if HT is enabled, consider a more reliable model, maybe cache size comparisons?
efficiencyCoreCount := 0
for _, threads := range threadsByCoreBySocket[id] {
s.ThreadCount += threads
if threads == 1 {
efficiencyCoreCount++
}

2
discover/path.go
View File

@@ -12,7 +12,7 @@ import (
// '../lib/ollama' on Linux and the executable's directory on macOS
// note: distribution builds, additional GPU-specific libraries are
// found in subdirectories of the returned path, such as
// 'cuda_v11', 'cuda_v12', 'rocm', etc.
// 'cuda_v12', 'rocm', etc.
var LibOllamaPath string = func() string {
exe, err := os.Executable()
if err != nil {