UUID result
Update UUID, so that we return raw bytes instead of object wrapper.
Removes CqlTypeClass, to reduce the overhead from it's type. This means, that array will be treated as ambiguous types in result wrapping.
This provides about 25% time improvement time elapsed:
Before
sudo perf stat node ./benchmark/logic/concurrent_select.js scylladb-javascript-driver 300000
Performance counter stats for 'node ./benchmark/logic/concurrent_select.js scylladb-javascript-driver 300000':
33,278.45 msec task-clock # 1.618 CPUs utilized
448,180 context-switches # 13.468 K/sec
160,090 cpu-migrations # 4.811 K/sec
67,942 page-faults # 2.042 K/sec
124,413,648,358 cycles # 3.739 GHz
48,582,787,534 stalled-cycles-frontend # 39.05% frontend cycles idle
109,354,731,390 instructions # 0.88 insn per cycle
# 0.44 stalled cycles per insn
20,753,228,584 branches # 623.624 M/sec
585,833,793 branch-misses # 2.82% of all branches
20.562829716 seconds time elapsed
24.754476000 seconds user
8.881299000 seconds sys
After:
sudo perf stat node ./benchmark/logic/concurrent_select.js scylladb-javascript-driver 300000
Performance counter stats for 'node ./benchmark/logic/concurrent_select.js scylladb-javascript-driver 300000':
28,658.52 msec task-clock # 1.718 CPUs utilized
384,096 context-switches # 13.403 K/sec
126,869 cpu-migrations # 4.427 K/sec
41,419 page-faults # 1.445 K/sec
105,013,076,940 cycles # 3.664 GHz
40,753,024,207 stalled-cycles-frontend # 38.81% frontend cycles idle
95,126,229,441 instructions # 0.91 insn per cycle
# 0.43 stalled cycles per insn
18,313,277,308 branches # 639.017 M/sec
529,455,973 branch-misses # 2.89% of all branches
16.684932131 seconds time elapsed
20.733184000 seconds user
8.202016000 seconds sys