Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Raise exception on too much shared memory requested #160

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Raise exception on too much shared memory requested #160

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 17 additions & 0 deletions jax_triton/triton_lib.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,6 +15,7 @@
"""Module for calling Triton kernels from JAX."""
import functools
import os
import sys
import types
from typing import Any, Callable, Dict, Optional, Protocol, Sequence, Tuple, Union
import weakref
Expand Down Expand Up @@ -134,6 +135,20 @@ def ptx_get_kernel_name(module) -> str:
return tc.get_kernel_name(module, pattern='// .globl')


# From https://en.wikipedia.org/wiki/CUDA#Technical_Specification
# "Amount of shared memory per multiprocessor". In bytes.
_SHARED_MEMORY_PER_SM = {
70: 98304,
72: 98304,
75: 65536,
80: 167936,
86: 102400,
87: 167936,
89: 102400,
90: 233472,
}


def compile_ttir_inplace(
ttir,
device: int = 0,
Expand Down Expand Up @@ -163,6 +178,8 @@ def compile_ttir_inplace(
ttgir.dump()
raise ValueError("TTGIR->LLIR pass failed!") from e
shared_mem = _triton.get_shared_memory_size(ttgir)
if shared_mem > _SHARED_MEMORY_PER_SM.get(compute_capability, sys.maxsize):
raise RuntimeError("Shared memory requested exceeds device resources.")
if dump:
print(llir)
ptx = tc.llir_to_ptx(llir, compute_capability)
Expand Down