-
Notifications
You must be signed in to change notification settings - Fork 3.5k
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.
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
[MetaSchedule] Update scripts for subgraph tuning #10501
[MetaSchedule] Update scripts for subgraph tuning #10501
Conversation
#define TVM_REGISTER_CUDA_TAG(Name, Arch, SharedMem, RegPerBlock) \ | ||
TVM_REGISTER_TARGET_TAG(Name).set_config({ \ | ||
{"kind", String("cuda")}, \ | ||
{"arch", String(Arch)}, \ | ||
{"shared_memory_per_block", Integer(SharedMem)}, \ | ||
{"registers_per_block", Integer(RegPerBlock)}, \ | ||
{"max_shared_memory_per_block", Integer(SharedMem)}, \ |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Per discussion with @masahi, we updated shared_memory_per_block
to max_shared_memory_per_block
to be consistent with Vulkan settings. I'm not fully convinced but would love to follow the convention as Masa suggested.
@@ -105,7 +105,6 @@ class VerifyGPUCodeNode : public PostprocNode { | |||
Target target = context->target.value(); | |||
this->target_constraints_ = Map<String, PrimExpr>{ | |||
{"max_shared_memory_per_block", Extract(target, "shared_memory_per_block")}, | |||
{"max_local_memory_per_block", Extract(target, "registers_per_block")}, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
ab6743d
to
df45fe7
Compare
df45fe7
to
30af4b9
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
No description provided.