tinygrad/tinygrad
2
0
Fork 3
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-08-29 10:56:08 +00:00
Code Issues Packages Projects Releases Wiki Activity

am: disable page migration (#10608)

Browse Source 
* am: disable page migration

* fixed

* enable

* fxi

* typ

* fix check
This commit is contained in:
nimlgen
2025-06-03 18:51:28 +03:00
committed by GitHub
parent 1c1f578490
commit 4381b54543
1 changed files with 9 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tinygrad/runtime/ops_amd.py
+9 -4
View File
@@ -476,7 +476,7 @@ class AMDAllocator(HCQAllocator['AMDDevice']):
def map(self, buf:HCQBuffer): self.dev.dev_iface.map(buf._base if buf._base is not None else buf)
MAP_FIXED, MAP_NORESERVE, MAP_LOCKED = 0x10, 0x400, 0 if OSX else 0x2000
MAP_FIXED, MAP_NORESERVE, MAP_LOCKED, MAP_POPULATE = 0x10, 0x400, 0 if OSX else 0x2000, getattr(mmap, "MAP_POPULATE", 0 if OSX else 0x008000)
@dataclass(frozen=True)
class ProfileSQTTEvent(ProfileEvent): device:str; se:int; blob:bytes; itrace:bool # noqa: E702
@@ -505,7 +505,7 @@ class AMDQueueDesc:
if CPUProgram.atomic_lib is not None: CPUProgram.atomic_lib.atomic_thread_fence(__ATOMIC_SEQ_CST:=5)
# Flush hdp if queue is in dev mem.
if dev.is_am() and not dev.is_usb() and getenv("AMD_ALLOC_QUEUE_DEV_MEM", 1): dev.dev_iface.adev.gmc.flush_hdp()
if dev.is_am() and not dev.is_usb(): dev.dev_iface.adev.gmc.flush_hdp()
for doorbell in self.doorbells: doorbell[0] = self.put_value
class KFDIface:
@@ -665,6 +665,11 @@ class PCIIface:
try: FileIOInterface(f"/sys/bus/pci/devices/{self.pcibus}/resource0_resize", os.O_RDWR).write(str(supported_sizes.bit_length() - 1))
except OSError as e: raise RuntimeError(f"Cannot resize BAR: {e}. Ensure the resizable BAR option is enabled on your system.") from e
# Disable migartion of locked pages.
if first_dev and FileIOInterface(reloc_sysfs:="/proc/sys/vm/compact_unevictable_allowed", os.O_RDONLY).read()[0] != "0":
os.system(cmd:=f"sudo sh -c 'echo 0 > {reloc_sysfs}'")
assert FileIOInterface(reloc_sysfs, os.O_RDONLY).read()[0] == "0", f"Failed to disable migration of locked pages. Please run {cmd} manually."
# Try to init vfio. Use it if success.
if getenv("VFIO", 0):
try:
@@ -736,10 +741,10 @@ class PCIIface:
def alloc(self, size:int, host=False, uncached=False, cpu_access=False):
if host or (not getenv("AMD_ALLOC_QUEUE_DEV_MEM", 1) and uncached and cpu_access): # host or gtt-like memory.
vaddr = self.adev.mm.alloc_vaddr(size:=round_up(size, mmap.PAGESIZE), align=mmap.PAGESIZE)
va = FileIOInterface.anon_mmap(vaddr, size, mmap.PROT_READ | mmap.PROT_WRITE, mmap.MAP_SHARED | mmap.MAP_ANONYMOUS | MAP_LOCKED | MAP_FIXED, 0)
FileIOInterface.anon_mmap(vaddr, size, mmap.PROT_READ|mmap.PROT_WRITE, mmap.MAP_SHARED|mmap.MAP_ANONYMOUS|MAP_POPULATE|MAP_LOCKED|MAP_FIXED, 0)
# Read pagemap to get the physical address of each page. The pages are locked.
self.pagemap.seek(va // mmap.PAGESIZE * 8)
self.pagemap.seek(vaddr // mmap.PAGESIZE * 8)
paddrs = [((x & ((1<<55) - 1)) * mmap.PAGESIZE, mmap.PAGESIZE) for x in array.array('Q', self.pagemap.read(size//mmap.PAGESIZE*8, binary=True))]
am_mapping = self.adev.mm.map_range(vaddr, size, paddrs, system=True, snooped=True, uncached=True)
return HCQBuffer(vaddr, size, meta=AMAllocationMeta(self.dev, [self.dev], am_mapping, has_cpu_mapping=True),