Add vector-times-matrix-transposed benchmark (V2) #40
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
This adds benchmarks for `vmt`, with very similar supporting structure to the existing `mmt` benchmark, but with different strategies tuned for matvec. This add three strategies: 1) Treat it like a reduction with one workgroup per row, relying on cache to get reuse of the vector. 2) Copy the vector to shared memory using all threads in the workgroup and then process N0 rows per workgroup, with WG_Y | N0 threadgroups. 3) Use a fixed number of workgroups and each workgroup strides the full problem space. This should limit the overhead of setting up the vector in shared memory, as well as improves scheduling overhead. Currently, the best configuration for each of the above three strategies are in the same performance ballpark (~20us for a 4096 * 4096x4096 matvec on an AMD 7900xtx).
* Drop alternative implementations * Pick a SIMD-contigous access pattern * Add more problem size configurations * Measure performance in GB/s * Add comments
kuhar
commented
Oct 26, 2023
|
I added code to prefetch LHS and RHS in hope to hide latency. I'm seeing better numbers now: |
Load 128-bits at the time
|
New numbers with increased load type: |
|
@antiagainst @qedawkins I'm pretty happy with this implementation. Should we merge? |
Works for me, can I give it a pass tomorrow first? |
qedawkins
approved these changes
Nov 2, 2023
|
Hi, ~/code/uVkCompute/build/benchmarks/vmt |
|
@oscarbg noting as of today, you can see the GSL compile target here: 3049af9#diff-62da6f62b4091626b341c9d8333d332aee35c053ff57cacebbb57792b987702aR30 This is more to communicate that it has been tuned and tested on rdna3, and in the future we may add more target-specific options to GLSL. |
@oscarbg also this indicates that one of the assumptions made in the GLSL does not hold on this target. |
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.
Based on #38 by @qedawkins, and earlier mmt by @kuhar.
Add benchmarks for
vmt, with very similar supporting structure to the existingmmtbenchmark.Changes compared to #38:
The performance depends heavily on the problem size. On 7900XTX, I'm seeing numbers up to 945 GB/s on 8k problem size.