Optimize the cache fetch for forward split, pt. 1 (#2187) #2218
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
✅ Deploy Preview for pytorch-fbgemm-docs canceled.
|
|
This pull request was exported from Phabricator. Differential Revision: D51865590 |
q10
force-pushed
the
export-D51865590
branch
from
384e08a to
2b0b81b
Compare
q10
added a commit
to q10/FBGEMM
that referenced
this pull request
Dec 18, 2023
Summary: Rewrite the kernel to use cache_hit_rate enum as template argument. We first check if the cache is empty and pass that value as a template argument. Inside the first kernel, we then determine the cache conflict miss rate, and use this value to as a template parameter when invoking the second kernel, which performs the actual lookup work. We pass in uvm_cache_stats as a run-time argument here instead of passing the cache miss rate as a compile-time argument, because uvm_cache_stats data is only available on the GPU, and incoking a templatized kernel with the cache miss rate as a template argument requires the cache misse information to first be passed back to the host, which is an expensive operation. This is based on the earlier work in stacks D48937380 and D49675672, which have been based on very outdated branches of fbcode. Differential Revision: D51865590
|
This pull request was exported from Phabricator. Differential Revision: D51865590 |
1 similar comment
|
This pull request was exported from Phabricator. Differential Revision: D51865590 |
q10
added a commit
to q10/FBGEMM
that referenced
this pull request
Dec 18, 2023
Summary: Rewrite the kernel to use cache_hit_rate enum as template argument. We first check if the cache is empty and pass that value as a template argument. Inside the first kernel, we then determine the cache conflict miss rate, and use this value to as a template parameter when invoking the second kernel, which performs the actual lookup work. We pass in uvm_cache_stats as a run-time argument here instead of passing the cache miss rate as a compile-time argument, because uvm_cache_stats data is only available on the GPU, and incoking a templatized kernel with the cache miss rate as a template argument requires the cache misse information to first be passed back to the host, which is an expensive operation. This is based on the earlier work in stacks D48937380 and D49675672, which have been based on very outdated branches of fbcode. Differential Revision: D51865590
q10
added a commit
to q10/FBGEMM
that referenced
this pull request
Dec 18, 2023
Summary: Rewrite the kernel to use cache_hit_rate enum as template argument. We first check if the cache is empty and pass that value as a template argument. Inside the first kernel, we then determine the cache conflict miss rate, and use this value to as a template parameter when invoking the second kernel, which performs the actual lookup work. We pass in uvm_cache_stats as a run-time argument here instead of passing the cache miss rate as a compile-time argument, because uvm_cache_stats data is only available on the GPU, and incoking a templatized kernel with the cache miss rate as a template argument requires the cache misse information to first be passed back to the host, which is an expensive operation. This is based on the earlier work in stacks D48937380 and D49675672, which have been based on very outdated branches of fbcode. Differential Revision: D51865590
q10
added a commit
to q10/FBGEMM
that referenced
this pull request
Dec 18, 2023
Summary: Rewrite the kernel to use cache_hit_rate enum as template argument. We first check if the cache is empty and pass that value as a template argument. Inside the first kernel, we then determine the cache conflict miss rate, and use this value to as a template parameter when invoking the second kernel, which performs the actual lookup work. We pass in uvm_cache_stats as a run-time argument here instead of passing the cache miss rate as a compile-time argument, because uvm_cache_stats data is only available on the GPU, and incoking a templatized kernel with the cache miss rate as a template argument requires the cache misse information to first be passed back to the host, which is an expensive operation. This is based on the earlier work in stacks D48937380 and D49675672, which have been based on very outdated branches of fbcode. Differential Revision: D51865590
q10
force-pushed
the
export-D51865590
branch
from
2b0b81b to
ee553e2
Compare
|
This pull request was exported from Phabricator. Differential Revision: D51865590 |
Summary: Rewrite the kernel to use cache_hit_rate enum as template argument. We first check if the cache is empty and pass that value as a template argument. Inside the first kernel, we then determine the cache conflict miss rate, and use this value to as a template parameter when invoking the second kernel, which performs the actual lookup work. We pass in uvm_cache_stats as a run-time argument here instead of passing the cache miss rate as a compile-time argument, because uvm_cache_stats data is only available on the GPU, and incoking a templatized kernel with the cache miss rate as a template argument requires the cache misse information to first be passed back to the host, which is an expensive operation. This is based on the earlier work in stacks D48937380 and D49675672, which have been based on very outdated branches of fbcode. Differential Revision: D51865590
q10
force-pushed
the
export-D51865590
branch
from
ee553e2 to
5e3adb6
Compare
|
This pull request was exported from Phabricator. Differential Revision: D51865590 |
1 similar comment
|
This pull request was exported from Phabricator. Differential Revision: D51865590 |
q10
added a commit
to q10/FBGEMM
that referenced
this pull request
Dec 19, 2023
Summary: Rewrite the kernel to use cache_hit_rate enum as template argument. We first check if the cache is empty and pass that value as a template argument. Inside the first kernel, we then determine the cache conflict miss rate, and use this value to as a template parameter when invoking the second kernel, which performs the actual lookup work. We pass in uvm_cache_stats as a run-time argument here instead of passing the cache miss rate as a compile-time argument, because uvm_cache_stats data is only available on the GPU, and incoking a templatized kernel with the cache miss rate as a template argument requires the cache misse information to first be passed back to the host, which is an expensive operation. This is based on the earlier work in stacks D48937380 and D49675672, which have been based on very outdated branches of fbcode. Differential Revision: D51865590
q10
added a commit
to q10/FBGEMM
that referenced
this pull request
Dec 19, 2023
Summary: Rewrite the kernel to use cache_hit_rate enum as template argument. We first check if the cache is empty and pass that value as a template argument. Inside the first kernel, we then determine the cache conflict miss rate, and use this value to as a template parameter when invoking the second kernel, which performs the actual lookup work. We pass in uvm_cache_stats as a run-time argument here instead of passing the cache miss rate as a compile-time argument, because uvm_cache_stats data is only available on the GPU, and incoking a templatized kernel with the cache miss rate as a template argument requires the cache misse information to first be passed back to the host, which is an expensive operation. This is based on the earlier work in stacks D48937380 and D49675672, which have been based on very outdated branches of fbcode. Differential Revision: D51865590
q10
added a commit
to q10/FBGEMM
that referenced
this pull request
Dec 19, 2023
Summary: Rewrite the kernel to use cache_hit_rate enum as template argument. We first check if the cache is empty and pass that value as a template argument. Inside the first kernel, we then determine the cache conflict miss rate, and use this value to as a template parameter when invoking the second kernel, which performs the actual lookup work. We pass in uvm_cache_stats as a run-time argument here instead of passing the cache miss rate as a compile-time argument, because uvm_cache_stats data is only available on the GPU, and incoking a templatized kernel with the cache miss rate as a template argument requires the cache misse information to first be passed back to the host, which is an expensive operation. This is based on the earlier work in stacks D48937380 and D49675672, which have been based on very outdated branches of fbcode. Differential Revision: D51865590
q10
added a commit
to q10/FBGEMM
that referenced
this pull request
Dec 27, 2023
Summary: Rewrite the kernel to use cache_hit_rate enum as template argument. We first check if the cache is empty and pass that value as a template argument. Inside the first kernel, we then determine the cache conflict miss rate, and use this value to as a template parameter when invoking the second kernel, which performs the actual lookup work. We pass in uvm_cache_stats as a run-time argument here instead of passing the cache miss rate as a compile-time argument, because uvm_cache_stats data is only available on the GPU, and incoking a templatized kernel with the cache miss rate as a template argument requires the cache misse information to first be passed back to the host, which is an expensive operation. This is based on the earlier work in stacks D48937380 and D49675672, which have been based on very outdated branches of fbcode. Reviewed By: spcyppt Differential Revision: D51865590
q10
added a commit
to q10/FBGEMM
that referenced
this pull request
Dec 27, 2023
Summary: Rewrite the kernel to use cache_hit_rate enum as template argument. We first check if the cache is empty and pass that value as a template argument. Inside the first kernel, we then determine the cache conflict miss rate, and use this value to as a template parameter when invoking the second kernel, which performs the actual lookup work. We pass in uvm_cache_stats as a run-time argument here instead of passing the cache miss rate as a compile-time argument, because uvm_cache_stats data is only available on the GPU, and incoking a templatized kernel with the cache miss rate as a template argument requires the cache misse information to first be passed back to the host, which is an expensive operation. This is based on the earlier work in stacks D48937380 and D49675672, which have been based on very outdated branches of fbcode. Reviewed By: spcyppt Differential Revision: D51865590
q10
added a commit
to q10/FBGEMM
that referenced
this pull request
Dec 27, 2023
Summary: Rewrite the kernel to use cache_hit_rate enum as template argument. We first check if the cache is empty and pass that value as a template argument. Inside the first kernel, we then determine the cache conflict miss rate, and use this value to as a template parameter when invoking the second kernel, which performs the actual lookup work. We pass in uvm_cache_stats as a run-time argument here instead of passing the cache miss rate as a compile-time argument, because uvm_cache_stats data is only available on the GPU, and incoking a templatized kernel with the cache miss rate as a template argument requires the cache misse information to first be passed back to the host, which is an expensive operation. This is based on the earlier work in stacks D48937380 and D49675672, which have been based on very outdated branches of fbcode. Reviewed By: spcyppt Differential Revision: D51865590
|
This pull request has been merged in 7dd0c7f. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary:
Rewrite the kernel to use cache_hit_rate enum as template argument. We first check if the cache is empty and pass that value as a template argument. Inside the first kernel, we then determine the cache conflict miss rate, and use this value to as a template parameter when invoking the second kernel, which performs the actual lookup work.
We pass in uvm_cache_stats as a run-time argument here instead of passing the cache miss rate as a compile-time argument, because uvm_cache_stats data is only available on the GPU, and incoking a templatized kernel with the cache miss rate as a template argument requires the cache misse information to first be passed back to the host, which is an expensive operation.
This is based on the earlier work in stacks D48937380 and D49675672, which have been based on very outdated branches of fbcode.
Differential Revision: D51865590