Benchmark: matmul ukernel vs direct codegen #415
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
|
|
||
| // We will time the run, and print it. | ||
| auto start = std::chrono::high_resolution_clock::now(); | ||
| auto run = kernel(bo_instr, instr_v.size(), bo_a, bo_b, bo_c); |
There was a problem hiding this comment.
was this kernel generated with latest iree-amd-aie, if the answer is yes then this would be the wrong way to run it. You can copy the changes done in the tests by Xilinx/mlir-aie#1517
newling
force-pushed
the
matmul_benchmark
branch
from
9bc089d
to
ef6fb1a
Compare
newling
force-pushed
the
matmul_benchmark
branch
from
0f6d244
to
7b7b593
Compare
jsetoain
reviewed
Jul 10, 2024
Comment on lines
+143
to
+147
| BASE_COMPILATION_FLAGS="-iree-hal-target-backends=amd-aie \ | ||
| -iree-amd-aie-peano-install-dir=${PEANO} \ | ||
| -iree-amd-aie-mlir-aie-install-dir=${MLIR_AIE_INSTALL} \ | ||
| -iree-amd-aie-vitis-install-dir=${VITIS} \ | ||
| -iree-amd-aie-show-invoked-commands" |
There was a problem hiding this comment.
This is missing the path to IREE:
-iree-amd-aie-install-dir=$IREE_INSTALL_DIRThere was a problem hiding this comment.
Ok thanks. This is a recent flag addition afaik. Maybe I should ask/push for this PR to be landed so it doesn't go stale..
There was a problem hiding this comment.
Yes, very recent (it just started failing for me). Having this test in CI wouldn't be entirely out of the question, I don't think. It's a good reference for people trying to do some performance analysis (like myself).
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Outdated suggestions cannot be applied.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.
End to end script. Running locally (nuc50)
Direct codegen
Using ukernel
So the ukernel approach is currently 3x faster. This is a lower bound though (i.e. core ukernel probably more than 3x faster). Consider:
Where
other-timeis the same in the 2 experiments, as only the instruction memory is different (identical DMA data movement). We observed thatso that
as
other-timeis the same in both approaches (date movement between DDR <-> memtile <-> core is identical)I think on this phoenix machine, theoretical max is
4 tops/second. So ukernel approach is 50% of theoretical max.Two extremes:
So performance of ukernel is between 3x and 5x better than dcg.