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.

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

DPU embedding table implementation with DPU multiColumn strategy #21

Open
wants to merge 98 commits into
base: master
Choose a base branch
from

Conversation

dgerinmem
Copy link

No description provided.

SylvanBrocard and others added 21 commits July 6, 2022 16:09
…rin/pipeline

# Conflicts:
#	upmem/Makefile
#	upmem/src/embedding.c
#	upmem/src/synthetic_dataset_embedding.c
# Conflicts:
#	upmem/Makefile
#	upmem/src/dpu/dpu_embedding.c
#	upmem/src/synthetic_dataset_embedding.c
…ipeline

# Conflicts:
#	upmem/Makefile
#	upmem/src/embedding.c
#	upmem/src/synthetic_dataset_embedding.c
…ults(callback)) (#1)

* preparation for async

* imports cleanup

* unused parameter

* fixed time counter

* added perf counters + experiment

* BUG & PERF fix : separate alloc_dpu_results

* preparation for async

* refacto : dont use precompiler constants

* 32 bits int in kernel

* formatting

* now supports >512 batches

* refacto : separates indices, offsets and input info allocation

* param : switch to original

* fixed

* put counters in ifdef blocks

* updated clangd

* imports cleanup

* unused parameter

* fixed time counter

* added perf counters + experiment

* 32 bits int in kernel

* formatting

* now supports >512 batches

* put counters in ifdef blocks

* updated clangd

* measuring verification time

* pipeline stage 1 [build_synthetic_input_data -> synthetic_inference]

* few merge fix

* bug fix

Co-authored-by: SylvanBrocard <[email protected]>
bugfix time measurement
use 10 re-run for DPU
# Conflicts:
#	upmem/Makefile
#	upmem/include/emb_types.h
#	upmem/include/embedding.h
#	upmem/src/embedding.c
#	upmem/src/synthetic_dataset_embedding.c
feature : RAND_INPUT_SIZE=1 -> synthetic input indices generation
                               with fixed batch size, random Pooling Factor (num indices per lookup)
add PERF.md for perf log
        separate testbench and embedding lib
        clean names
        update README
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants