LX 1.00 Benchmark File
==============================================================================

See bottom of file for benchmarking setup.


A. Test 1

View all archives on DH1:
(29 -lh5- archives, totalling 8 MB)

   Utility  Version  Time (sec)  Command line
   -------  -------  ----------  ------------
1. LV       1.00     12.07       LV >nil: -q dh1:#?
2. LhA      1.38     12.07       LhA >nil: -q v dh1:#?
3. LX/020   1.00     12.10       LX020 >nil: -q v dh1:#?
4. LX       1.00     12.14       LX >nil: -q v dh1:#?
5. LZ       1.92     13.38       LZ >nil: -N v dh1:#?
6. Haddock  1.20     20.30       HD >nil: -rQsrxE*pi-1e.09 V dh1:#?

Comments: As you can see, LV, LhA, LX and LX/020 operate at almost exactly the
          same speed for the viewing operation.


B. Test 2

Test all files in archives on DH1:
(29 -lh5- archives, totalling 8 MB)

   Utility  Version  Time (sec)  Options
   -------  -------  ----------  -------
1. LX/020   1.00     92.45       LX020 >nil: -q t
2. LX       1.00     95.23       LX >nil: -q t
3. LhA      1.38     103.72      LhA >nil: -q t
4. LZ       1.92     115.67      LZ >nil: -N t

Comments: LX and LX/020 show their speed, a combination of optimised assembly
          language, and asynchronous disk i/o.


C. Test 3

Test one large file in RAM:
(410K archived, 1500K decompressed, -lh5-)

   Utility  Version  Time (sec)  Options
   -------  -------  ----------  -------
1. LX/020   1.00     4.93        LX020 >nil: -q t
2. LX       1.00     5.35        LX >nil: -q t
3. LhA      1.38     5.42        LhA >nil: -q t
4. LZ       1.92     10.98       LZ >nil: -N t

Comments: Testing entirely in RAM, the asynchronous disk i/o advantage of LX
          is nullified, and, in fact, becomes a deficiency, although it
          remains faster than LhA.


D. Test 4

Test one large file in RAM:
(410K archived, 1500K decompressed, -lh1-)

   Utility  Version  Time (sec)  Options
   -------  -------  ----------  -------
1. LX/020   1.00     12.25       LX020 >nil: -q t
2. LX       1.00     12.75       LX >nil: -q t
3. LhA      1.38     15.34       LhA >nil: -q t
4. LZ       1.92     22.35       LZ >nil: -N t

Comments: LX and LX/020 are far faster at decompressing -lh1- archives.


E. Test 5

Test all files in archives on DF0:
(1 -lh5- archive, total 817K)

   Utility  Version  Time (sec)  Options
   -------  -------  ----------  -------
1. LX       1.00     55.22       LX >nil: -q t
2. LX/020   1.00     55.30       LX020 >nil: -q t
3. LhA      1.38     80.68       LhA >nil: -q t
4. LZ       1.92     96.27       LZ >nil: -N t

Comments: Demonstrates the advantage of LX's asynchronous disk i/o.

As a sidenote, here are the stats for the 'copy' command
(copy the archive from DF0: to RAM:)

   Utility  Version  Time (sec)
   -------  -------  ----------
   Copy     38.1     39.21

==============================================================================

The above benchmarks were performed using:

Amiga 3000 (25 MHz 68030)
6 MB Fast RAM, 2 MB Chip RAM
52 MB Quantum LPS

All archivers and the timing program (RTimer) were in RAM:

All console output was suppressed in all tests.

The computer was re-booted between each test to avoid memory fragmentation.

The Data and Instruction caches of the 68030 were enabled, as was BURST mode
(for both Data and Instructions).

These benchmarks provide a very general idea of the speed improvement of LX
over other archives.  Your mileage will vary.

==============================================================================