Follow-up to "lz4frame: avoid signed overflow", per the #1824 review:
saturate at LZ4_ACCELERATION_MAX rather than INT_MAX,
since higher values make no difference.
LZ4_ACCELERATION_DEFAULT and LZ4_ACCELERATION_MAX move from lz4.c
to the LZ4_STATIC_LINKING_ONLY section of lz4.h, which lz4.c, lz4frame.c,
bench.c and frametest.c already include, so their values derive from it.
The LZ4_compress_fast() documentation already referred to both constants.
bench.c computed the acceleration the same way (-cLevel + 1),
so `lz4 -b --fast=2147483647` hit the same signed overflow:
apply the same cap there.
test-lz4-basic.sh: --fast=2147483647 must produce the same frame
as --fast=65536 (fails on dev, where acceleration wraps to the default),
and its benchmark runs (caught by the UBSan CI job before this fix).
lz4.1: --fast values above 65536 make no difference.
as cpus get faster, it's now possible for some of them to reach speeds > 10000 MB/s on single core.
Ensure that the display remains properly aligned in this case.
benchmark dictionary mode, implemented in #808, is incorrect.
It would desynchronize compression and decompression as soon as 2+ files are dictionary compressed.
Also : slightly improved traces system, to also include __LINE__ number
Support the -D command line option for running benchmarks. The
benchmark code was slightly restructured to factor out the calls
that need to be different for each benchmark scenario. Since there
are now 4 scenarios (all combinations of fast/HC and with/without
dictionary), the logic was getting somewhat convoluted otherwise.
This was done by extending the compressionParameters struct that
previously contained just a single function pointer. It now
contains 4 function pointers for init/reset/compress/cleanup,
with the related state. The functions get a pointer to the
structure as their first argument (inspired by C++), so that they
can access the state values in the struct.
it was a fairly complex scenario,
involving source files > 64K
and some extraordinary conditions related to specific layout of ranges of zeroes.
and only on level 9.
Previous method would produce too many time() invocations,
becoming a significant fraction of workload measured.
The new strategy is to use time() only once per batch,
and dynamically resize batch size so that each round lasts approximately 1 second.
This only matters for small inputs.
Measurement for large files (such as silesia.tar) are much less impacted
(though decoding speed is so fast that even medium-size files will notice an improvement).
The timer used was only accurate up to 0.01 seconds. This timer is accurate up to 1 ns.
It is a monotonic timer that measures the real time difference, not on CPU time.
Copied the benchmark code from 6ab4d5e904
