Remove lkball indexing

[codecov[bot]]

Codecov Report

Merging #210 into master will increase coverage by 0.01%. The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #210      +/-   ##
==========================================
+ Coverage   93.96%   93.97%   +0.01%     
==========================================
  Files         237      237              
  Lines       11114    11112       -2     
==========================================
  Hits        10443    10443              
+ Misses        671      669       -2

Flag	Coverage Δ
#_shrug_Testing_CPU	91% <ø> (+0.01%)	:arrow_up:
#_shrug_Testing_CUDA	93.96% <ø> (+0.01%)	:arrow_up:

Impacted Files	Coverage Δ
tmol/score/lk_ball/script_modules.py	90.69% <ø> (+4.03%)	:arrow_up:

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 28744be...e503468. Read the comment docs.

[codecov[bot]]

Codecov Report

Merging #210 into master will increase coverage by 0.01%. The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #210      +/-   ##
==========================================
+ Coverage   93.96%   93.97%   +0.01%     
==========================================
  Files         237      237              
  Lines       11114    11112       -2     
==========================================
  Hits        10443    10443              
+ Misses        671      669       -2

Flag	Coverage Δ
#_shrug_Testing_CPU	91% <ø> (+0.01%)	:arrow_up:
#_shrug_Testing_CUDA	93.96% <ø> (+0.01%)	:arrow_up:

Impacted Files	Coverage Δ
tmol/score/lk_ball/script_modules.py	90.69% <ø> (+4.03%)	:arrow_up:

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 28744be...e503468. Read the comment docs.

[codecov[bot]]

Codecov Report

Merging #210 into master will increase coverage by 0.01%. The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #210      +/-   ##
==========================================
+ Coverage   93.96%   93.97%   +0.01%     
==========================================
  Files         237      237              
  Lines       11114    11112       -2     
==========================================
  Hits        10443    10443              
+ Misses        671      669       -2

Flag	Coverage Δ
#_shrug_Testing_CPU	91% <ø> (+0.01%)	:arrow_up:
#_shrug_Testing_CUDA	93.96% <ø> (+0.01%)	:arrow_up:

Impacted Files	Coverage Δ
tmol/score/lk_ball/script_modules.py	90.69% <ø> (+4.03%)	:arrow_up:

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 28744be...e503468. Read the comment docs.

[codecov[bot]]

Codecov Report

Merging #210 into master will increase coverage by 0.02%. The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #210      +/-   ##
==========================================
+ Coverage   93.96%   93.98%   +0.02%     
==========================================
  Files         237      237              
  Lines       11114    11124      +10     
==========================================
+ Hits        10443    10455      +12     
+ Misses        671      669       -2

Flag	Coverage Δ
#_shrug_Testing_CPU	91.01% <100%> (+0.02%)	:arrow_up:
#_shrug_Testing_CUDA	93.96% <100%> (+0.02%)	:arrow_up:

Impacted Files	Coverage Δ
tmol/score/lk_ball/script_modules.py	89.74% <ø> (+3.07%)	:arrow_up:
tmol/tests/score/lk_ball/test_script_modules.py	100% <100%> (ø)	:arrow_up:
tmol/score/lk_ball/score_graph.py	100% <100%> (ø)	:arrow_up:

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update ab6d903...5787493. Read the comment docs.

[asford]

🤔

It looks like removing the python-level operation without a subsequent update to the kernel to sparse-ify the calculation is causing a slowdown. I'll have to review a perf trace off the branch to understand what's going on. Maybe increased cost of eval, or perhaps the cost of folding derivatives back in through the water locations? Plot of the comparative performance:

[fdimaio]

It looks like removing the python-level operation without a subsequent update to the kernel to sparse-ify the calculation is causing a slowdown.

Agreed. I saw this last night. All of the added runtime is (as expected) in the second kernel. My first guess is to use similar logic to hbond to sparsify the calculation. Will try this out today.

[asford]

Awesome! Thanks.

On Wed, Jul 10, 2019, 09:19 Frank DiMaio notifications@github.com wrote:

It looks like removing the python-level operation without a subsequent update to the kernel to sparse-ify the calculation is causing a slowdown.

Agreed. I saw this last night. All of the added runtime is (as expected) in the second kernel. My first guess is to use similar logic to hbond to sparsify the calculation. Will try this out today.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/uw-ipd/tmol/pull/210?email_source=notifications&email_token=AACFBKEZRE24FU6CHSRWTMDP6YDWXA5CNFSM4H7KI3YKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODZT7UCI#issuecomment-510130697, or mute the thread https://github.com/notifications/unsubscribe-auth/AACFBKFXQJ4ZBASKNYFD6S3P6YDWXANCNFSM4H7KI3YA .

[fdimaio]

@asford I refactored using indexed dispatch and now we see a decent speedup.

Also thanks for the graph checkins, it is very useful for evaluating code changes.