All lookdict* functions are implemented like pseudo language:

```
lookup()
if not collision:
    return result

while True:
    perturb_shift()
    lookup()
    if not collision:
        return result
```

This patch changes it as:

```
while True:
    lookup()
    if not collision:
        return result
    perturb_shift()
```

It removes 100 lines of code. Good. But how about performance?

When this patch is applied to 4a534c45bbf6:

```
$ ../python.patched -m perf compare_to default.json patched2.json -G --min-speed=2
Slower (4):
- xml_etree_generate: 271 ms +- 6 ms -> 283 ms +- 9 ms: 1.04x slower (+4%)
- nqueens: 263 ms +- 4 ms -> 272 ms +- 3 ms: 1.04x slower (+4%)
- scimark_monte_carlo: 272 ms +- 10 ms -> 280 ms +- 14 ms: 1.03x slower (+3%)
- scimark_lu: 435 ms +- 23 ms -> 446 ms +- 32 ms: 1.03x slower (+3%)

Faster (7):
- call_method: 15.2 ms +- 0.2 ms -> 14.7 ms +- 0.4 ms: 1.04x faster (-4%)
- call_simple: 14.4 ms +- 0.2 ms -> 13.9 ms +- 0.3 ms: 1.04x faster (-4%)
- xml_etree_iterparse: 227 ms +- 9 ms -> 219 ms +- 7 ms: 1.04x faster (-3%)
- scimark_sor: 527 ms +- 10 ms -> 510 ms +- 11 ms: 1.03x faster (-3%)
- call_method_slots: 14.7 ms +- 0.5 ms -> 14.3 ms +- 0.2 ms: 1.03x faster (-3%)
- genshi_text: 90.2 ms +- 1.1 ms -> 87.8 ms +- 1.1 ms: 1.03x faster (-3%)
- django_template: 403 ms +- 5 ms -> 394 ms +- 4 ms: 1.02x faster (-2%)

Benchmark hidden because not significant (53): 2to3, ...
```

And when this patch applied to 1a97b10cb420 :

```
$ ../python.patched -m perf compare_to default.json patched.json -G --min-speed=2                                                                                                               
Slower (6):
- call_simple: 13.5 ms +- 0.5 ms -> 14.4 ms +- 0.4 ms: 1.07x slower (+7%)
- xml_etree_generate: 270 ms +- 6 ms -> 287 ms +- 5 ms: 1.06x slower (+6%)
- xml_etree_process: 240 ms +- 6 ms -> 247 ms +- 4 ms: 1.03x slower (+3%)
- regex_compile: 429 ms +- 3 ms -> 440 ms +- 5 ms: 1.03x slower (+3%)
- call_method_unknown: 16.1 ms +- 0.2 ms -> 16.5 ms +- 0.3 ms: 1.02x slower (+2%)
- logging_simple: 31.2 us +- 0.4 us -> 32.0 us +- 0.3 us: 1.02x slower (+2%)

Faster (8):
- genshi_text: 90.6 ms +- 1.4 ms -> 87.6 ms +- 1.2 ms: 1.03x faster (-3%)
- scimark_sor: 513 ms +- 11 ms -> 497 ms +- 12 ms: 1.03x faster (-3%)
- genshi_xml: 200 ms +- 2 ms -> 194 ms +- 2 ms: 1.03x faster (-3%)
- unpickle_pure_python: 857 us +- 21 us -> 835 us +- 13 us: 1.03x faster (-3%)
- python_startup_no_site: 9.95 ms +- 0.02 ms -> 9.74 ms +- 0.02 ms: 1.02x faster (-2%)
- json_dumps: 29.7 ms +- 0.4 ms -> 29.1 ms +- 0.4 ms: 1.02x faster (-2%)
- xml_etree_iterparse: 225 ms +- 9 ms -> 220 ms +- 5 ms: 1.02x faster (-2%)
- chameleon: 31.1 ms +- 0.3 ms -> 30.5 ms +- 0.5 ms: 1.02x faster (-2%)

Benchmark hidden because not significant (50): 2to3, ...
```

I can't see any stable and significant performance regression.
I'll try to create some micro benchmarks.