Sch: Restore per-processor pressure in task cost estimation

[jpsamaroo]

estimate_task_costs! dropped the est_business term (each candidate processor's currently-reserved compute pressure) in b16afad, leaving only a fixed per-task cross-worker transfer cost in the comparison. That fixed cost always makes the scheduler's own worker look cheapest, so under the default scope every task piled onto worker 1 regardless of how busy it became (#663).

Restore the pressure term, reading it from worker_time_pressure's current LockedObject/atomic-counter representation, so busier processors are considered costlier and work spreads across workers again.

Fixes #663

Written by Claude Sonnet

[github-actions]

Dagger benchmarks: dirty vs master

Median time

	master	dirty	master / dirty
array/dagger/N=1024 (block 512)/add (X + X)	5 ± 0.27 ms	4.56 ± 1.2 ms	1.1 ± 0.3
array/dagger/N=1024 (block 512)/alloc (rand)	1.67 ± 0.018 ms	2.3 ± 0.55 ms	0.728 ± 0.18
array/dagger/N=1024 (block 512)/broadcast (X .+ 1)	2.29 ± 0.61 ms	1.9 ± 0.65 ms	1.2 ± 0.52
array/dagger/N=1024 (block 512)/map (sin.(X))	5.61 ± 0.51 ms	5.12 ± 1.5 ms	1.1 ± 0.33
array/dagger/N=1024 (block 512)/norm	1.19 ± 0.1 ms	1.15 ± 0.024 ms	1.03 ± 0.091
array/dagger/N=1024 (block 512)/reduce (sum)	2.46 ± 1.4 ms	2.38 ± 0.97 ms	1.03 ± 0.72
array/dagger/N=1024 (block 512)/transpose (permutedims)	4.69 ± 0.62 ms	4.62 ± 0.59 ms	1.02 ± 0.19
array/dagger/N=256 (block 256)/add (X + X)	0.813 ± 0.032 ms	0.773 ± 2 ms	1.05 ± 2.7
array/dagger/N=256 (block 256)/alloc (rand)	0.717 ± 0.099 ms	0.657 ± 0.045 ms	1.09 ± 0.17
array/dagger/N=256 (block 256)/broadcast (X .+ 1)	0.455 ± 0.079 ms	0.442 ± 0.019 ms	1.03 ± 0.18
array/dagger/N=256 (block 256)/map (sin.(X))	1 ± 0.14 ms	0.946 ± 0.0094 ms	1.06 ± 0.15
array/dagger/N=256 (block 256)/norm	0.477 ± 0.0061 ms	0.419 ± 0.11 ms	1.14 ± 0.29
array/dagger/N=256 (block 256)/reduce (sum)	0.846 ± 0.13 ms	0.847 ± 0.036 ms	0.998 ± 0.16
array/dagger/N=256 (block 256)/transpose (permutedims)	0.792 ± 0.012 ms	0.698 ± 0.054 ms	1.14 ± 0.09
linalg/dagger/N=1024 (block 512)/cholesky	22.5 ± 4.5 ms	19.9 ± 3.2 ms	1.13 ± 0.29
linalg/dagger/N=1024 (block 512)/lu	0.0427 ± 0.00062 s	0.0407 ± 0.0045 s	1.05 ± 0.12
linalg/dagger/N=1024 (block 512)/matmul (A*A)	0.0447 ± 0.026 s	0.052 ± 0.0056 s	0.861 ± 0.51
linalg/dagger/N=1024 (block 512)/matvec (A*x)	3.56 ± 2.4 ms	3.12 ± 1.5 ms	1.14 ± 0.95
linalg/dagger/N=1024 (block 512)/qr	0.112 ± 0.0087 s	0.112 ± 0.0055 s	1 ± 0.093
linalg/dagger/N=1024 (block 512)/solve (A\b via lu)	0.0509 ± 0.0028 s	0.0504 ± 0.0021 s	1.01 ± 0.071
linalg/dagger/N=1024 (block 512)/svd	0.0374 h	0.037 h	1.01
linalg/dagger/N=1024 (block 512)/syrk (A'*A)	0.0325 ± 0.00057 s	0.0332 ± 0.0046 s	0.979 ± 0.14
linalg/dagger/N=256 (block 256)/cholesky	6.16 ± 0.81 ms	4.59 ± 2.5 ms	1.34 ± 0.76
linalg/dagger/N=256 (block 256)/lu	6.6 ± 3.9 ms	4.53 ± 0.5 ms	1.46 ± 0.89
linalg/dagger/N=256 (block 256)/matmul (A*A)	2.33 ± 0.83 ms	2.12 ± 0.74 ms	1.1 ± 0.55
linalg/dagger/N=256 (block 256)/matvec (A*x)	1.98 ± 2.4 ms	1.07 ± 2.9 ms	1.85 ± 5.5
linalg/dagger/N=256 (block 256)/qr	4.97 ± 0.69 ms	5.25 ± 0.97 ms	0.947 ± 0.22
linalg/dagger/N=256 (block 256)/solve (A\b via lu)	10.2 ± 1.2 ms	10.1 ± 10 ms	1.01 ± 1
linalg/dagger/N=256 (block 256)/svd	0.608 ± 0.0021 s	0.571 ± 0.04 s	1.07 ± 0.075
linalg/dagger/N=256 (block 256)/syrk (A'*A)	3.53 ± 0.91 ms	3.76 ± 1.4 ms	0.939 ± 0.43
stencil/dagger/N=1024 (block 512)/alloc (neighbors Wrap)	7.69 ± 1.3 ms	6.49 ± 0.68 ms	1.19 ± 0.24
stencil/dagger/N=1024 (block 512)/assign (const)	1.84 ± 1.6 ms	0.893 ± 0.029 ms	2.06 ± 1.8
stencil/dagger/N=1024 (block 512)/multi-expr	4.76 ± 2.5 ms	2.61 ± 2.3 ms	1.83 ± 1.9
stencil/dagger/N=1024 (block 512)/neighbors (Clamp)	6.58 ± 0.65 ms	5.1 ± 1.3 ms	1.29 ± 0.35
stencil/dagger/N=1024 (block 512)/neighbors (Pad)	5.85 ± 0.37 ms	5.54 ± 0.14 ms	1.06 ± 0.071
stencil/dagger/N=1024 (block 512)/neighbors (Reflect)	6 ± 0.71 ms	5.69 ± 0.67 ms	1.05 ± 0.18
stencil/dagger/N=1024 (block 512)/neighbors (Wrap)	6.47 ± 0.57 ms	5.58 ± 0.3 ms	1.16 ± 0.12
stencil/dagger/N=1024 (block 512)/update (+)	1.42 ± 1.7 ms	1.9 ± 1.4 ms	0.751 ± 1.1
stencil/dagger/N=256 (block 256)/alloc (neighbors Wrap)	1.82 ± 0.075 ms	1.63 ± 0.047 ms	1.11 ± 0.056
stencil/dagger/N=256 (block 256)/assign (const)	0.481 ± 0.04 ms	0.551 ± 0.23 ms	0.872 ± 0.38
stencil/dagger/N=256 (block 256)/multi-expr	1.96 ± 1.3 ms	1.43 ± 1.2 ms	1.37 ± 1.5
stencil/dagger/N=256 (block 256)/neighbors (Clamp)	1.46 ± 0.031 ms	1.45 ± 0.54 ms	1.01 ± 0.37
stencil/dagger/N=256 (block 256)/neighbors (Pad)	1.34 ± 0.1 ms	2.45 ± 1.5 ms	0.546 ± 0.34
stencil/dagger/N=256 (block 256)/neighbors (Reflect)	1.36 ± 0.14 ms	1.17 ± 0.12 ms	1.16 ± 0.17
stencil/dagger/N=256 (block 256)/neighbors (Wrap)	1.48 ± 0.3 ms	1.28 ± 0.23 ms	1.16 ± 0.31
stencil/dagger/N=256 (block 256)/update (+)	1.1 ± 0.76 ms	0.71 ± 0.23 ms	1.55 ± 1.2
time_to_load	1.06 ± 0.0073 s	1.05 ± 0.0058 s	1 ± 0.0089

Allocations / memory

	master	dirty	master / dirty
array/dagger/N=1024 (block 512)/add (X + X)	5.47 k allocs: 8.21 MB	5.52 k allocs: 8.22 MB	1
array/dagger/N=1024 (block 512)/alloc (rand)	2.74 k allocs: 8.09 MB	2.69 k allocs: 8.09 MB	1
array/dagger/N=1024 (block 512)/broadcast (X .+ 1)	2.39 k allocs: 8.08 MB	2.4 k allocs: 8.08 MB	1
array/dagger/N=1024 (block 512)/map (sin.(X))	2.07 k allocs: 8.07 MB	2.08 k allocs: 8.07 MB	1
array/dagger/N=1024 (block 512)/norm	2.71 k allocs: 0.0906 MB	2.68 k allocs: 0.0891 MB	1.02
array/dagger/N=1024 (block 512)/reduce (sum)	4.11 k allocs: 0.141 MB	3.9 k allocs: 0.133 MB	1.06
array/dagger/N=1024 (block 512)/transpose (permutedims)	3.32 k allocs: 8.14 MB	3.28 k allocs: 8.13 MB	1
array/dagger/N=256 (block 256)/add (X + X)	1.71 k allocs: 0.574 MB	1.7 k allocs: 0.574 MB	1
array/dagger/N=256 (block 256)/alloc (rand)	0.749 k allocs: 0.526 MB	0.749 k allocs: 0.526 MB	1
array/dagger/N=256 (block 256)/broadcast (X .+ 1)	0.646 k allocs: 0.522 MB	0.635 k allocs: 0.522 MB	1
array/dagger/N=256 (block 256)/map (sin.(X))	0.569 k allocs: 0.519 MB	0.569 k allocs: 0.519 MB	1
array/dagger/N=256 (block 256)/norm	0.71 k allocs: 24.5 kB	0.702 k allocs: 24.2 kB	1.01
array/dagger/N=256 (block 256)/reduce (sum)	1.29 k allocs: 0.0464 MB	1.3 k allocs: 0.0466 MB	0.995
array/dagger/N=256 (block 256)/transpose (permutedims)	1.1 k allocs: 0.551 MB	1.1 k allocs: 0.551 MB	1
linalg/dagger/N=1024 (block 512)/cholesky	7.81 k allocs: 10.3 MB	7.77 k allocs: 10.3 MB	1
linalg/dagger/N=1024 (block 512)/lu	15.6 k allocs: 14.6 MB	15.5 k allocs: 14.6 MB	1
linalg/dagger/N=1024 (block 512)/matmul (A*A)	7.57 k allocs: 8.28 MB	7.63 k allocs: 8.28 MB	1
linalg/dagger/N=1024 (block 512)/matvec (A*x)	6.14 k allocs: 0.246 MB	5.95 k allocs: 0.238 MB	1.03
linalg/dagger/N=1024 (block 512)/qr	11.4 k allocs: 9.59 MB	11 k allocs: 9.57 MB	1
linalg/dagger/N=1024 (block 512)/solve (A\b via lu)	29.3 k allocs: 15.2 MB	29.6 k allocs: 15.2 MB	1
linalg/dagger/N=1024 (block 512)/svd	0.0488 M allocs: 0.197 GB	0.0485 M allocs: 0.197 GB	1
linalg/dagger/N=1024 (block 512)/syrk (A'*A)	8.87 k allocs: 24.4 MB	8.81 k allocs: 24.4 MB	1
linalg/dagger/N=256 (block 256)/cholesky	3.03 k allocs: 0.633 MB	3.02 k allocs: 0.633 MB	1
linalg/dagger/N=256 (block 256)/lu	5.42 k allocs: 1.24 MB	5.35 k allocs: 1.24 MB	1
linalg/dagger/N=256 (block 256)/matmul (A*A)	1.92 k allocs: 0.581 MB	1.91 k allocs: 0.581 MB	1
linalg/dagger/N=256 (block 256)/matvec (A*x)	2.34 k allocs: 0.099 MB	2.33 k allocs: 0.0986 MB	1
linalg/dagger/N=256 (block 256)/qr	3.44 k allocs: 0.779 MB	3.43 k allocs: 0.778 MB	1
linalg/dagger/N=256 (block 256)/solve (A\b via lu)	12 k allocs: 1.53 MB	11.8 k allocs: 1.52 MB	1
linalg/dagger/N=256 (block 256)/svd	15.4 k allocs: 6.71 MB	15.4 k allocs: 6.71 MB	1
linalg/dagger/N=256 (block 256)/syrk (A'*A)	2.97 k allocs: 2.63 MB	2.97 k allocs: 2.63 MB	1
stencil/dagger/N=1024 (block 512)/alloc (neighbors Wrap)	4.7 k allocs: 8.2 MB	4.72 k allocs: 8.2 MB	1
stencil/dagger/N=1024 (block 512)/assign (const)	2.1 k allocs: 0.0926 MB	2.15 k allocs: 0.0943 MB	0.982
stencil/dagger/N=1024 (block 512)/multi-expr	4.78 k allocs: 0.211 MB	4.76 k allocs: 0.21 MB	1
stencil/dagger/N=1024 (block 512)/neighbors (Clamp)	3.73 k allocs: 0.231 MB	3.75 k allocs: 0.231 MB	0.997
stencil/dagger/N=1024 (block 512)/neighbors (Pad)	3.6 k allocs: 0.227 MB	3.6 k allocs: 0.227 MB	1
stencil/dagger/N=1024 (block 512)/neighbors (Reflect)	3.71 k allocs: 0.262 MB	3.71 k allocs: 0.262 MB	1
stencil/dagger/N=1024 (block 512)/neighbors (Wrap)	3.56 k allocs: 0.166 MB	3.55 k allocs: 0.166 MB	1
stencil/dagger/N=1024 (block 512)/update (+)	2.59 k allocs: 0.115 MB	2.73 k allocs: 2.12 MB	0.0544
stencil/dagger/N=256 (block 256)/alloc (neighbors Wrap)	1.54 k allocs: 0.574 MB	1.53 k allocs: 0.574 MB	1
stencil/dagger/N=256 (block 256)/assign (const)	0.795 k allocs: 0.0403 MB	0.806 k allocs: 0.0406 MB	0.991
stencil/dagger/N=256 (block 256)/multi-expr	1.72 k allocs: 0.0865 MB	1.71 k allocs: 0.0863 MB	1
stencil/dagger/N=256 (block 256)/neighbors (Clamp)	1.23 k allocs: 0.0708 MB	1.3 k allocs: 0.0735 MB	0.964
stencil/dagger/N=256 (block 256)/neighbors (Pad)	1.2 k allocs: 0.0702 MB	1.19 k allocs: 0.0694 MB	1.01
stencil/dagger/N=256 (block 256)/neighbors (Reflect)	1.25 k allocs: 0.0797 MB	1.31 k allocs: 0.0818 MB	0.975
stencil/dagger/N=256 (block 256)/neighbors (Wrap)	1.17 k allocs: 0.0619 MB	1.2 k allocs: 0.0632 MB	0.979
stencil/dagger/N=256 (block 256)/update (+)	0.917 k allocs: 0.0459 MB	0.917 k allocs: 0.0459 MB	1
time_to_load	0.147 k allocs: 10.8 kB	0.147 k allocs: 10.8 kB	1

Plots

⚠️ Regressions (time > 25.0% and outside the reported ±spread; allocs/memory > 25.0%)

• stencil/dagger/N=1024 (block 512)/update (+) (memory): +1739.3%
• array/dagger/N=1024 (block 512)/alloc (rand) (time): +37.4%

Within noise (9 metric(s) past threshold but inside the ±spread; not counted)

• stencil/dagger/N=256 (block 256)/neighbors (Pad) (time): 83.2%
• stencil/dagger/N=1024 (block 512)/update (+) (time): 33.1%
• linalg/dagger/N=256 (block 256)/cholesky (time): -25.4%
• stencil/dagger/N=256 (block 256)/multi-expr (time): -26.8%
• linalg/dagger/N=256 (block 256)/lu (time): -31.3%
• stencil/dagger/N=256 (block 256)/update (+) (time): -35.4%
• stencil/dagger/N=1024 (block 512)/multi-expr (time): -45.3%
• linalg/dagger/N=256 (block 256)/matvec (A*x) (time): -45.8%
• stencil/dagger/N=1024 (block 512)/assign (const) (time): -51.4%

Full results and plots (download the benchmark-results artifact).