Improve CUDA performance

[flaub]

No description provided.

[github-actions]

Benchmark for Linux-cuda c05f8ee

Click to hide benchmark

Test	Base	PR	%
fib/100/execute	5.2±0.12ms	5.2±0.10ms	0.00%
fib/100/prove	1261.3±8.11ms	1062.5±3.27ms	-15.76%
fib/100/total	1264.0±11.24ms	1066.4±6.42ms	-15.63%
fib/1000/execute	5.7±0.10ms	5.7±0.13ms	0.00%
fib/1000/prove	1301.0±7.55ms	1098.0±7.15ms	-15.60%
fib/1000/total	1302.8±9.03ms	1102.8±3.99ms	-15.35%
fib/10000/execute	10.3±0.09ms	10.2±0.12ms	-0.97%
fib/10000/prove	5.1±0.04s	4.1±0.05s	-19.61%
fib/10000/total	5.2±0.04s	4.1±0.04s	-21.15%

Benchmark for Linux-default

    <details open>
      <summary>Click to hide benchmark</summary>
      Benchmarks have changed between the two branches, unable to diff.
    </details>

Benchmark for macOS-default c05f8ee

Click to hide benchmark

Test	Base	PR	%
fib/100/execute	2.8±0.19ms	2.7±0.13ms	-3.57%
fib/100/prove	3.6±0.05s	3.6±0.04s	0.00%
fib/100/total	3.6±0.04s	3.6±0.03s	0.00%
fib/1000/execute	2.9±0.06ms	2.9±0.08ms	0.00%
fib/1000/prove	3.6±0.04s	3.6±0.09s	0.00%
fib/1000/total	3.6±0.07s	3.6±0.02s	0.00%
fib/10000/execute	5.2±0.14ms	5.0±0.06ms	-3.85%
fib/10000/prove	14.6±0.30s	14.6±0.12s	0.00%
fib/10000/total	14.7±0.32s	14.6±0.13s	-0.68%

Benchmark for macOS-metal c05f8ee

Click to hide benchmark

Test	Base	PR	%
fib/100/execute	2.8±0.07ms	2.8±0.05ms	0.00%
fib/100/prove	849.2±4.39ms	848.4±5.39ms	-0.09%
fib/100/total	875.0±5.88ms	872.6±7.61ms	-0.27%
fib/1000/execute	3.0±0.11ms	2.9±0.04ms	-3.33%
fib/1000/prove	869.7±4.82ms	866.4±3.52ms	-0.38%
fib/1000/total	898.0±4.98ms	896.8±7.91ms	-0.13%
fib/10000/execute	5.1±0.07ms	5.0±0.15ms	-1.96%
fib/10000/prove	3.4±0.01s	3.3±0.01s	-2.94%
fib/10000/total	3.4±0.01s	3.3±0.01s	-2.94%

[jbruestle]

Looks good minus a the fixed constant and hidden dependence on another constant.

[github-actions]

Benchmark for Linux-cuda db050b2

Click to hide benchmark

Test	Base	PR	%
fib/100/execute	5.2±0.09ms	5.2±0.09ms	0.00%
fib/100/prove	1288.5±5.52ms	1103.4±3.48ms	-14.37%
fib/100/total	1300.1±8.07ms	1101.0±3.32ms	-15.31%
fib/1000/execute	5.7±0.11ms	5.7±0.12ms	0.00%
fib/1000/prove	1331.8±6.04ms	1146.9±5.59ms	-13.88%
fib/1000/total	1318.3±7.91ms	1133.5±3.71ms	-14.02%
fib/10000/execute	10.3±0.14ms	10.2±0.14ms	-0.97%
fib/10000/prove	5.2±0.04s	4.3±0.04s	-17.31%
fib/10000/total	5.2±0.05s	4.3±0.04s	-17.31%

Benchmark for Linux-default

    <details open>
      <summary>Click to hide benchmark</summary>
      Benchmarks have changed between the two branches, unable to diff.
    </details>

Benchmark for macOS-default db050b2

Click to hide benchmark

Test	Base	PR	%
fib/100/execute	2.7±0.13ms	2.7±0.10ms	0.00%
fib/100/prove	3.7±0.06s	3.7±0.03s	0.00%
fib/100/total	3.7±0.05s	3.7±0.04s	0.00%
fib/1000/execute	2.9±0.06ms	2.9±0.07ms	0.00%
fib/1000/prove	3.7±0.07s	3.7±0.05s	0.00%
fib/1000/total	3.7±0.05s	3.7±0.06s	0.00%
fib/10000/execute	5.2±0.12ms	5.1±0.11ms	-1.92%
fib/10000/prove	15.3±0.12s	15.2±0.14s	-0.65%
fib/10000/total	15.3±0.14s	15.2±0.13s	-0.65%

Benchmark for macOS-metal

    <details open>
      <summary>Click to hide benchmark</summary>
      Benchmarks have changed between the two branches, unable to diff.
    </details>