Reduce function calls in hot path

[oneonestar]

Description

Reduce function calls in hot path when reading TextFile.
For some reason JVM didn't inline Slice.getByte() and that caused a performance regression (compare Trino 388 and Trino 428).

Additional context and related issues

$ java -version
openjdk version "17.0.4.1" 2022-08-12
OpenJDK Runtime Environment Temurin-17.0.4.1+1 (build 17.0.4.1+1)
OpenJDK 64-Bit Server VM Temurin-17.0.4.1+1 (build 17.0.4.1+1, mixed mode, sharing)

 CREATE TABLE test (
    ...
    foo varchar,
    err_code integer,
    log_date varchar,
    log_hour varchar,
    log_min varchar
 )
 WITH (
    format = 'TEXTFILE',
    partitioned_by = ARRAY['log_date','log_hour','log_min'],
    textfile_field_separator = U&'\0009'
 )

explain analyze verbose
select
     count(*) as request_count
 from test
where log_date between  '20231101' and '20231130' and foo != 'bar' and err_code = 0;

Before:

    └─ ScanFilterProject[table = ...., filterPredicate = ((""foo"" <> VARCHAR 'bar') AND (""err_code"" = 0))]
           Layout: []
           Estimates: {rows: ? (0B), cpu: ?, memory: 0B, network: 0B}/{rows: ? (0B), cpu: ?, memory: 0B, network: 0B}/{rows: ? (0B), cpu: 0, memory: 0B, network: 0B}
           CPU: 3.87h (99.88%), Scheduled: 3.93h (99.88%), Blocked: 0.00ns (0.00%), Output: 2561824485 rows (0B)
           metrics:
             'CPU time distribution (s)' = {count=720, p01=2.13, p05=2.36, p10=3.12, p25=7.32, p50=18.91, p75=26.43, p90=39.03, p95=43.56, p99=53.73, min=1.86, max=63.28}
             'Filter CPU time' = {duration=1.21m}
             'Input rows distribution' = {count=720, p01=643102.00, p05=723008.49, p10=959696.92, p25=2202411.75, p50=5724157.26, p75=7800053.53, p90=11579449.88, p95=12846610.81, p99=15835723.00, min=551069.00, max=18246492.00}
             'Scheduled time distribution (s)' = {count=720, p01=2.17, p05=2.42, p10=3.21, p25=7.46, p50=19.18, p75=26.88, p90=39.37, p95=44.30, p99=54.87, min=1.93, max=63.85}
           Input avg.: 5762501.43 rows, Input std.dev.: 66.94%
           foo := foo:string:REGULAR
           err_code := err_code:int:REGULAR
           log_date:string:PARTITION_KEY
               :: [[20231101], [20231102], [20231103], [20231104], [20231105], [20231106], [20231107], [20231108], [20231109], [20231110], [20231111], [20231112], [20231113], [20231114], [20231115], [20231116], [20231117], [20231118], [20231119], [20231120], [20231121], [20231122], [20231123], [20231124], [20231125], [20231126], [20231127], [20231128], [20231129], [20231130]]
           log_min:string:PARTITION_KEY
               :: [[0]]
           log_hour:string:PARTITION_KEY
               :: [[0], [1], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [2], [20], [21], [22], [23], [3], [4], [5], [6], [7], [8], [9]]
           Input: 4149001027 rows (70.55GB), Filtered: 38.25%, Physical input: 317.99GB, Physical input time: 6658800.00ms

After:

    └─ ScanFilterProject[table = ... filterPredicate = ((""foo"" <> VARCHAR 'bar') AND (""err_code"" = 0))]
           Layout: []
           Estimates: {rows: ? (0B), cpu: ?, memory: 0B, network: 0B}/{rows: ? (0B), cpu: ?, memory: 0B, network: 0B}/{rows: ? (0B), cpu: 0, memory: 0B, network: 0B}
           CPU: 3.09h (99.99%), Scheduled: 3.17h (99.99%), Blocked: 0.00ns (0.00%), Output: 2561824485 rows (0B)
           metrics:
             'CPU time distribution (s)' = {count=720, p01=1.72, p05=2.28, p10=2.88, p25=6.53, p50=15.68, p75=20.91, p90=30.21, p95=33.17, p99=39.95, min=1.34, max=45.44}
             'Filter CPU time' = {duration=59.80s}
             'Input rows distribution' = {count=720, p01=643102.00, p05=722484.69, p10=962167.67, p25=2202245.63, p50=5737402.26, p75=7801883.69, p90=11601371.25, p95=12858997.28, p99=15835723.00, min=551069.00, max=18246492.00}
             'Scheduled time distribution (s)' = {count=720, p01=1.82, p05=2.37, p10=2.96, p25=6.69, p50=16.00, p75=21.37, p90=31.16, p95=34.23, p99=42.22, min=1.38, max=46.59}
           Input avg.: 5762501.43 rows, Input std.dev.: 66.94%
           foo := foo:string:REGULAR
           err_code := err_code:int:REGULAR
           log_date:string:PARTITION_KEY
               :: [[20231101], [20231102], [20231103], [20231104], [20231105], [20231106], [20231107], [20231108], [20231109], [20231110], [20231111], [20231112], [20231113], [20231114], [20231115], [20231116], [20231117], [20231118], [20231119], [20231120], [20231121], [20231122], [20231123], [20231124], [20231125], [20231126], [20231127], [20231128], [20231129], [20231130]]
           log_hour:string:PARTITION_KEY
               :: [[0], [1], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [2], [20], [21], [22], [23], [3], [4], [5], [6], [7], [8], [9]]
           log_min:string:PARTITION_KEY
               :: [[0]]
           Input: 4149001027 rows (70.55GB), Filtered: 38.25%, Physical input: 317.99GB, Physical input time: 6814200.00ms

Trino 388 (For reference)

└─ ScanFilterProject[table = ..., filterPredicate = ((""foo"" <> VARCHAR 'bar') AND (""err_code"" = 0))]
         Layout: []
         Estimates: {rows: ? (0B), cpu: ?, memory: 0B, network: 0B}/{rows: ? (0B), cpu: ?, memory: 0B, network: 0B}/{rows: ? (0B), cpu: ?, memory: 0B, network: 0B}
         CPU: 3.58h (99.98%), Scheduled: 3.64h (99.98%), Blocked: 0.00ns (0.00%), Output: 2561824485 rows (0B)
         metrics:
           'Input distribution' = {count=721.00, p01=632889.00, p05=721412.80, p10=956607.82, p25=2181684.40, p50=5721734.28, p75=7801610.53, p90=11581353.33, p95=12852007.22, p99=15835723.00, min=0.00, max=18246492.00}
         Input avg.: 5754509.05 rows, Input std.dev.: 67.09%
         foo := foo:string:REGULAR
         err_code := err_code:int:REGULAR
         log_min:string:PARTITION_KEY
             :: [[0]]
         log_hour:string:PARTITION_KEY
             :: [[0], [1], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [2], [20], [21], [22], [23], [3], [4], [5], [6], [7], [8], [9]]
         log_date:string:PARTITION_KEY
             :: [[20231101], [20231102], [20231103], [20231104], [20231105], [20231106], [20231107], [20231108], [20231109], [20231110], [20231111], [20231112], [20231113], [20231114], [20231115], [20231116], [20231117], [20231118], [20231119], [20231120], [20231121], [20231122], [20231123], [20231124], [20231125], [20231126], [20231127], [20231128], [20231129], [20231130]]
         Input: 4149001027 rows (317.99GB), Filtered: 38.25%

---

CPU frame graph where:

• Top: before patch
• Middle: inline getByte()
• Bottom: also inline isEscapeByte()

---

Compare Trino 388 and Trino 428

Release notes

( ) This is not user-visible or is docs only, and no release notes are required.
( ) Release notes are required. Please propose a release note for me.
(X) Release notes are required, with the following suggested text:

# Section
* Improve performance when querying TextFile table ({issue}`issuenumber`)

[github-actions]

This pull request has gone a while without any activity. Tagging the Trino developer relations team: @bitsondatadev @colebow @mosabua

[hashhar]

ping @sopel39 / @Dith3r

[mosabua]

Related .. does this still apply and fix the issue for Java 21 .. since that is now the required runtime.

[sopel39]

merged, thanks!