CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-27 08:21:30 +00:00
Code Issues Packages Projects Releases Wiki Activity

HIP: remove the use of __HIP_PLATFORM_AMD__, explicitly support only AMD targets (#14945)

Browse Source
This commit is contained in:
uvos
2025-07-29 20:23:04 +02:00
committed by GitHub
parent b77d11179d
commit aa79524c51
10 changed files with 67 additions and 65 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
ggml/src/ggml-cuda/ggml-cuda.cu
View File

@@ -128,7 +128,7 @@ static cudaError_t ggml_cuda_device_malloc(void ** ptr, size_t size, int device)
return err;
}
#if defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__)
#if defined(GGML_USE_HIP)
static int ggml_cuda_parse_id(char devName[]) {
// A list of possible Target IDs can be found under the rocclr/clr repo in device.cpp
// these values are not stable so this is susceptible to breakage
@@ -175,10 +175,10 @@ static int ggml_cuda_parse_id(char devName[]) {
archNum += archMinor;
return archNum;
}
#endif // defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__)
#endif // defined(GGML_USE_HIP)
static ggml_cuda_device_info ggml_cuda_init() {
#ifdef __HIP_PLATFORM_AMD__
#if defined(GGML_USE_HIP)
// Workaround for a rocBLAS bug when using multiple graphics cards:
// https://github.com/ROCmSoftwarePlatform/rocBLAS/issues/1346
{
@@ -251,7 +251,7 @@ static ggml_cuda_device_info ggml_cuda_init() {
info.devices[id].nsm = prop.multiProcessorCount;
info.devices[id].smpb = prop.sharedMemPerBlock;
info.devices[id].warp_size = prop.warpSize;
#if defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__)
#if defined(GGML_USE_HIP)
info.devices[id].smpbo = prop.sharedMemPerBlock;
info.devices[id].cc = ggml_cuda_parse_id(prop.gcnArchName);
@@ -281,7 +281,7 @@ static ggml_cuda_device_info ggml_cuda_init() {
info.devices[id].cc = 100*prop.major + 10*prop.minor;
GGML_LOG_INFO(" Device %d: %s, compute capability %d.%d, VMM: %s\n",
id, prop.name, prop.major, prop.minor, device_vmm ? "yes" : "no");
#endif // defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__)
#endif // defined(GGML_USE_HIP)
}
for (int id = 0; id < info.device_count; ++id) {