github/tinygrad
1
1
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-01-24 06:18:01 -05:00
Code Issues Packages Projects Releases Wiki Activity

amd better allocation error messages (#5462)

Browse Source 
* amd better allocation error messages

* a bit better
This commit is contained in:
nimlgen
2024-07-13 22:55:09 +03:00
committed by GitHub
parent 2427f149a3
commit 67f70cef02
1 changed files with 10 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
tinygrad/runtime/ops_amd.py
View File

@@ -351,12 +351,8 @@ class AMDAllocator(HCQCompatAllocator):
def __init__(self, device:AMDDevice): super().__init__(device, batch_size=SDMA_MAX_COPY_SIZE)
def _alloc(self, size:int, options:BufferOptions) -> HCQCompatAllocRes:
try:
if options.host: return self.device._gpu_alloc(size, kfd.KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, public=True)
return self.device._gpu_alloc(size, kfd.KFD_IOC_ALLOC_MEM_FLAGS_VRAM, public=options.cpu_access)
except OSError as e:
if e.errno == errno.ENOMEM: raise MemoryError("Cannot allocate memory") from e
raise
if options.host: return self.device._gpu_alloc(size, kfd.KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, public=True)
return self.device._gpu_alloc(size, kfd.KFD_IOC_ALLOC_MEM_FLAGS_VRAM, public=options.cpu_access)
def _free(self, opaque, options:BufferOptions): self.device._gpu_free(opaque)
@@ -394,7 +390,14 @@ class AMDDevice(HCQCompatCompiled):
else:
buf, addr = 0, libc.mmap(0, size, 0, mmap.MAP_PRIVATE|mmap.MAP_ANONYMOUS|MAP_NORESERVE, -1, 0)
assert addr != 0xffffffffffffffff
mem = kio.alloc_memory_of_gpu(self.kfd, va_addr=addr, size=size, base=addr, length=size, gpu_id=self.gpu_id, flags=flags, mmap_offset=buf)
try: mem = kio.alloc_memory_of_gpu(self.kfd, va_addr=addr, size=size, base=addr, length=size, gpu_id=self.gpu_id, flags=flags, mmap_offset=buf)
except OSError as e:
if e.errno == errno.EINVAL and (flags & kfd.KFD_IOC_ALLOC_MEM_FLAGS_VRAM) and public:
raise MemoryError("Cannot allocate host-visible VRAM. Ensure the resizable BAR option is enabled on your system.") from e
if e.errno == errno.ENOMEM: raise MemoryError("Cannot allocate memory: no memory is available.") from e
raise
if not (flags & kfd.KFD_IOC_ALLOC_MEM_FLAGS_USERPTR):
buf = libc.mmap(mem.va_addr, mem.size, mmap.PROT_READ|mmap.PROT_WRITE, mmap.MAP_SHARED|MAP_FIXED, self.drm_fd, mem.mmap_offset)
assert addr == buf == mem.va_addr