Skip to content

[WIP] [Codegen] [CUDA] enable fast math flags #9252

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

Closed
AndrewZhaoLuo wants to merge 3 commits into from
Closed

[WIP] [Codegen] [CUDA] enable fast math flags #9252

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
6ff80f9
results
AndrewZhaoLuo Oct 11, 2021
b0d6408
syntax err
AndrewZhaoLuo Oct 11, 2021
7eef0f5
use fmad ptxas compiler option
AndrewZhaoLuo Oct 13, 2021
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
8 changes: 6 additions & 2 deletions python/tvm/contrib/nvcc.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,15 +18,15 @@
"""Utility to invoke nvcc compiler in the system"""
from __future__ import absolute_import as _abs

import subprocess
import os
import subprocess
import warnings

import tvm._ffi
from tvm.runtime import ndarray as nd

from . import utils
from .._ffi.base import py_str
from . import utils


def compile_cuda(code, target="ptx", arch=None, options=None, path_target=None):
Expand Down Expand Up @@ -78,6 +78,10 @@ def compile_cuda(code, target="ptx", arch=None, options=None, path_target=None):
else:
cmd += ["-arch", arch]

# Fast math, pass in through options later
cmd += ["--use_fast_math"]
cmd += ["--ptxas-options=--fmad"]

if options:
if isinstance(options, str):
cmd += [options]
Expand Down
2 changes: 2 additions & 0 deletions src/target/opt/build_cuda_on.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -102,6 +102,8 @@ std::string NVRTCCompile(const std::string& code, bool include_path = false) {
compile_params.push_back(include_option);
}

compile_params.push_back("--use_fast_math");

for (const auto& string : compile_params) {
param_cstrings.push_back(string.c_str());
}
Expand Down