The ISA-L EC code has been written using RVV vector instructions and the minimum multiplication table,
resulting in a performance improvement of over 10 times compared to the existing implementation.
Signed-off-by: Shuo Lv <lv.shuo@sanechips.com.cn>
Added new RAID performance application which consolidates the
existing XOR and P+Q gen performance applications.
This application accepts buffer sizes to benchmark,
as a single value, list or range, and the RAID function
to test and the number of sources.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
To benchmark a cold cache scenario, the option `--cold`
has been added as a parameter of the CRC benchmark application,
where the addresses of the input buffers are randomize
within a 1GB preallocated memory buffer.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Added new CRC performance application which consolidates the
existing CRC performance applications (CRC16, CRC32 and CRC64).
This application accepts buffer sizes to benchmark,
as a single value, list or range, and the CRC function
to test (or all of them).
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
This makes it easier to compare the constants used for crc/crc64_*_by8.asm,
crc/crc64_*_by16_10.asm, and crc/aarch64/crc64_*_pmull.h
Note that this revealed some discrepancies:
ecma_refl: br_high != rk8 (92d8af2baf0e1e85 vs 92d8af2baf0e1e84)
iso_refl: br_high != rk8 (b000000000000001 vs b000000000000000)
jones_refl: br_high != rk8 (2b5926535897936b vs 2b5926535897936a)
but they should be innocuous.
Signed-off-by: Tim Burke <tim.burke@gmail.com>
As already announced in issue #296, we are removing 32-bit x86 support,
which was not being validated anyway.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
NASM version 2.14.01 supports all x86 ISA in this library.
Since this version has been out since 2018, it is safe to
only permit the library to be compiled with this minimum version,
as announced in issue #297.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
The riscv64 dispatcher code uses the same PROVIDER_INFO macro as the
aarch64 dispatcher and have the same kind of warnings during compilation:
igzip/riscv64/igzip_multibinary_riscv64_dispatcher.c:39:24: warning: type of 'adler32_base' does not match original declaration [-Wlto-type-mismatch]
39 | return PROVIDER_BASIC(adler32);
| ^
igzip/adler32_base.c:34:1: note: return value type mismatch
34 | adler32_base(uint32_t adler32, uint8_t *start, uint64_t length)
| ^
igzip/adler32_base.c:34:1: note: type 'uint32_t' should match type 'void'
igzip/adler32_base.c:34:1: note: 'adler32_base' was previously declared here
This commit introduces the same correction for riscv64.
Signed-off-by: Mattias Ellert <mattias.ellert@physics.uu.se>
The PROVIDER_INFO macro used in the aarch64 code declares all
functions with the signature:
extern void function(void);
The actual return type and parameter list of the functions are however
different. The declarations provided by the PROVIDER_INFO macro
therfore conflicts with the actual declarations of the functions
elsewhere in the code, causing compiler warnings.
This commit drops the PROVIDER_INFO macro and provides proper function
declarations, eiter by including a header file or by providing a
forward declaration. This corresponds to how the code for the other
architectures are handlinging this issue.
Signed-off-by: Mattias Ellert <mattias.ellert@physics.uu.se>
When using RISC-V GCC 14, `gcc -O0` passes the test, but `gcc -O2` fails.
The log shows that it enters the branch `if (c_dut != c_ref) {`
even though `c_dut` and `c_ref` have the same value.
Adding `volatile` allows the test to pass.
Signed-off-by: sunyuechi <sunyuechi@iscas.ac.cn>
When a file contains a number of repeated '0x00' or '0xff'
bytes, the block header is copied from a precomputed header,
which only worked for Little-Endian systems.
Fixes#311.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
igzip/igzip_icf_body.c:7:1: warning: type of 'gen_icf_map_lh1' does not match original declaration [-Wlto-type-mismatch]
7 | gen_icf_map_lh1(struct isal_zstream *, struct deflate_icf *, uint32_t);
| ^
igzip/igzip_base_aliases.c:177:1: note: return value type mismatch
177 | gen_icf_map_lh1(struct isal_zstream *stream, struct deflate_icf *matches_icf_lookup,
| ^
igzip/igzip_base_aliases.c:177:1: note: type 'void' should match type 'uint64_t'
igzip/igzip_base_aliases.c:177:1: note: 'gen_icf_map_lh1' was previously declared here
igzip/igzip_base_aliases.c:177:1: note: code may be misoptimized unless '-fno-strict-aliasing' is used
igzip/igzip_icf_body.c:9:1: warning: type of 'set_long_icf_fg' does not match original declaration [-Wlto-type-mismatch]
9 | set_long_icf_fg(uint8_t *, uint64_t, uint64_t, struct deflate_icf *);
| ^
igzip/igzip_base_aliases.c:170:1: note: type mismatch in parameter 2
170 | set_long_icf_fg(uint8_t *next_in, uint8_t *end_in, struct deflate_icf *match_lookup,
| ^
igzip/igzip_base_aliases.c:170:1: note: 'set_long_icf_fg' was previously declared here
igzip/igzip_base_aliases.c:170:1: note: code may be misoptimized unless '-fno-strict-aliasing' is used
igzip/igzip_base_aliases.c:62:1: warning: type of 'set_long_icf_fg_base' does not match original declaration [-Wlto-type-mismatch]
62 | set_long_icf_fg_base(uint8_t *next_in, uint8_t *end_in, struct deflate_icf *match_lookup,
| ^
igzip/igzip_icf_body.c:34:1: note: type mismatch in parameter 2
34 | set_long_icf_fg_base(uint8_t *next_in, uint64_t processed, uint64_t input_size,
| ^
igzip/igzip_icf_body.c:34:1: note: 'set_long_icf_fg_base' was previously declared here
igzip/igzip_icf_body.c:34:1: note: code may be misoptimized unless '-fno-strict-aliasing' is used
igzip/igzip_base_aliases.c:54:1: warning: type of 'adler32_base' does not match original declaration [-Wlto-type-mismatch]
54 | adler32_base(uint32_t init, const unsigned char *buf, uint64_t len);
| ^
igzip/adler32_base.c:34:1: note: type mismatch in parameter 3
34 | adler32_base(uint32_t adler32, uint8_t *start, uint32_t length)
| ^
igzip/adler32_base.c:34:1: note: type 'uint32_t' should match type 'uint64_t'
igzip/adler32_base.c:34:1: note: 'adler32_base' was previously declared here
igzip/adler32_base.c:34:1: note: code may be misoptimized unless '-fno-strict-aliasing' is used
Signed-off-by: Mattias Ellert <mattias.ellert@physics.uu.se>
The soname is equal to current minus age.
In version 2.31.0 current is 2 and age is set to 0.
In version 2.31.1 current is 2 and age is set to 1.
This means the soname goes backwards from 2 to 1.
The full library version changes from 2.0.31 to 1.1.31
The soname should not go backwards, so this soname change looks like a
mistake that should be reverted.
The current, revision, age for a library should change in one of three ways:
1) increase current by one, reset revision and age to 0.
2) increase current by one, reset revision to 0 and increase age by 1.
3) increase revision by 1, retain the values of current and age.
1) is for non-backward compatible changes to the library (changes or
removals to the old ABI). Soname changes and applications using the
library must be recompiled.
2) is for when there are ABI additions to the library, but no ABI
changes or removals. Application compiled against the old version of
the library don't need to be recompiled, and the soname (current minus
age) does not change.
3) is for minor updates with no ABI additions, changes or removals.
The major, minor, patch version of the software project should not be
used as current, revision, age for the library. Especially true for
using the patch version as age, because that means the soname goes
backwards for patch releases as happened here.
Signed-off-by: Mattias Ellert <mattias.ellert@physics.uu.se>
