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

Jump to bottom

[TOPI] Minor perf improvement for GPU scatter #7233

Merged
merged 16 commits into from
Jan 19, 2021
Merged

[TOPI] Minor perf improvement for GPU scatter #7233

Changes from 1 commit
Commits
Show all changes
16 commits
Select commit Hold shift + click to select a range
4020730
improve scatter 4d init
masahi Jan 1, 2021
cb0e21e
do not launch sorting based scatter for small input
masahi Jan 1, 2021
1b76109
do not use hard coded num threads
masahi Jan 8, 2021
d60acbe
separate sort based implementation
masahi Jan 10, 2021
b6d6541
register scatter as autotvm task
masahi Jan 11, 2021
9c87d8d
add missing import
masahi Jan 11, 2021
9b077db
fix strategy
masahi Jan 11, 2021
3090af8
add dedicated schedule and dummy flop
masahi Jan 11, 2021
28c08dc
add test tuning script
masahi Jan 11, 2021
bf27ba8
try adding dummy knob
Jan 11, 2021
a876443
skip random_fill when a tuning workload is from scatter
masahi Jan 18, 2021
5cfd32e
cleanup memcpy ir
masahi Jan 18, 2021
528f755
remove scatter tuning script
masahi Jan 18, 2021
4f844f5
make sure zero init arguments
masahi Jan 18, 2021
6e3acff
add comment on why skip random init for scatter
masahi Jan 18, 2021
7b1a6ea
restore ctx sync
masahi Jan 19, 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
Prev Previous commit
Next Next commit
do not use hard coded num threads
  • Loading branch information
@masahi
masahi committed Jan 18, 2021
commit 1b7610932cc25d1aa00c7d40033911229221d4b4
3 changes: 2 additions & 1 deletion python/tvm/topi/cuda/scatter.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -314,8 +314,9 @@ def gen_ir_4d(data, indices, updates, axis, out, update_func):
data_ptr = ib.buffer_ptr(data)
with ib.new_scope():
fused = n * c * h * w
num_thread = 1024
num_thread = int(tvm.target.Target.current(allow_none=False).max_num_threads)
num_blocks = ceil_div(fused, num_thread)

bx = te.thread_axis("blockIdx.x")
ib.scope_attr(bx, "thread_extent", num_blocks)
tx = te.thread_axis("threadIdx.x")
Expand Down