380 Commits

Author	SHA1	Message	Date
Mika Lindqvist	2c98ece180	[configure] Add support for RISC-V ZBC extension	2025-05-28 13:32:14 +02:00
yintong	830995ff78	riscv: add bash configure script and related ci support for riscv	2025-05-01 22:59:25 +02:00
Adam Stylinski	46fc33f39d	SSE4.1 optimized chorba ... This is ~25-30% faster than the SSE2 variant on a core2 quad. The main reason for this has to do with the fact that, while incurring far fewer shifts, an entirely separate stack buffer has to be managed that is the size of the L1 cache on most CPUs. This was one of the main reasons the 32k specialized function was slower for the scalar counterpart, despite auto vectorizing. The auto vectorized loop was setting up the stack buffer at unaligned offsets, which is detrimental to performance pre-nehalem. Additionally, we were losing a fair bit of time to the zero initialization, which we are now doing more selectively. There are a ton of loads and stores happening, and for sure we are bound on the fill buffer + store forwarding. An SSE2 version of this code is probably possible by simply replacing the shifts with unpacks with zero and the palignr's with shufpd's. I'm just not sure it'll be all that worth it, though. We are gating against SSE4.1 not because we are using specifically a 4.1 instruction but because that marks when Wolfdale came out and palignr became a lot faster.	2025-04-15 14:11:12 +02:00
Detlef Riekenberg	5a232688e1	port: Use __cpuid only, when available. ... Add a fallback, when __cpuid is not available	2025-04-15 14:08:46 +02:00
Hans Kristian Rosbach	00a3168d5d	Add AVX512 version of compare256 ... Improve the speed of sub-16 byte matches by first using a 128-bit intrinsic, after that use only 512-bit intrinsics. This requires us to overlap on the last run, but this is cheaper than processing the tail using a 256-bit and then a 128-bit run. Change benchmark steps to avoid it hitting chunk boundaries of one or the other function as much, this gives more fair benchmarks.	2025-04-14 23:28:38 +02:00
Cristi Vîjdea	b2a0ffb5e9	Pass POSIX_C_SOURCE for std::alligned_alloc try_compile checks ... On FreeBSD 11, definining POSIX_C_SOURCE to a lower level has the efect of inhibiting the language level (__ISO_C_VISIBLE ) to be lower than C11, even in the presence of -std=c11 Since the check_symbol_exists runs without setting POSIX_C_SOURCE, this means that we will spuriously define HAVE_ALIGNED_ALLOC, while in the actual build it is not going to be defined ref: https://github.com/freebsd/freebsd-src/blob/stable/11/sys/sys/cdefs.h#L738	2025-04-07 23:36:27 +02:00
Adam Stylinski	724dc0cfb4	Explicit SSE2 vectorization of Chorba CRC method ... The version that's currently in the generic implementation for 32768 byte buffers leverages the stack. It manages to autovectorize but unfortunately the trips to the stack hurt its performance for CPUs which need this the most. This version is explicitly SIMD vectorized and doesn't use trips to the stack. In my testing it's ~10% faster than the "small" variant, and about 42% faster than the "32768" variant.	2025-03-28 20:43:59 +01:00
Detlef Riekenberg	89f90d3e9c	ports: Use memalign or _aligned_malloc, when available. Fallback to malloc ... Using "_WIN32" to decide, if the MSVC extensions _aligned_malloc / _aligned_free are available is a bug that breaks other Compiler on Windows. (OpenWatcom as Example) Regards ... Detlef	2025-03-17 16:36:12 +01:00
Vladislav Shchapov	85223403d9	Change flags to "-Werror=unguarded-availability", "-Werror=unguarded-availability-new" and add it to maybe affected symbol checking ... Signed-off-by: Vladislav Shchapov <vladislav@shchapov.ru>	2025-03-01 16:45:04 +01:00
Vladislav Shchapov	c3bd3d7645	Restore support macOS prior 10.15 ... Signed-off-by: Vladislav Shchapov <vladislav@shchapov.ru>	2025-03-01 16:45:04 +01:00
Andrew Murray	e8d8049382	Added --installnamedir	2025-02-16 18:15:43 +01:00
Cameron Cawley	aecfe160c5	Provide --without-acle/-DWITH_ACLE options for backward compatibility	2025-02-12 13:54:30 +01:00
Cameron Cawley	231c4b3a64	Use -Wa,-march with older ARM toolchains	2025-02-12 13:54:30 +01:00
Cameron Cawley	7ea78f12c8	Provide an inline asm fallback for the ARMv8 intrinsics	2025-02-12 13:54:30 +01:00
Cameron Cawley	721c488aff	Rename most ACLE references to ARMv8	2025-02-12 13:54:30 +01:00
Vladislav Shchapov	05305edee8	Remove unused include directories ... Signed-off-by: Vladislav Shchapov <vladislav@shchapov.ru>	2025-01-27 12:38:30 +01:00
Hans Kristian Rosbach	509f6b5818	Since we long ago make unaligned reads safe (by using memcpy or intrinsics), ... it is time to replace the UNALIGNED_OK checks that have since really only been used to select the optimal comparison sizes for the arch instead.	2024-12-21 00:46:48 +01:00
Hans Kristian Rosbach	037ab0fd35	Revert "Since we long ago make unaligned reads safe (by using memcpy or intrinsics)," ... This reverts commit 80fffd72f3. It was mistakenly pushed to develop instead of going through a PR and the appropriate reviews.	2024-12-17 23:09:31 +01:00
Hans Kristian Rosbach	80fffd72f3	Since we long ago make unaligned reads safe (by using memcpy or intrinsics), ... it is time to replace the UNALIGNED_OK checks that have since really only been used to select the optimal comparison sizes for the arch instead.	2024-12-17 23:02:32 +01:00
Adam Stylinski	7020cb3f74	Enable AVX2 functions to be built with BMI2 instructions ... While these are technically different instructions, no such CPU exists that has AVX2 that doesn't have BMI2. Enabling BMI2 allows us to eliminate several flag stalls by having flagless versions of shifts, and allows us to not clobber and move around GPRs so much in scalar code. There's usually a sizeable benefit for enabling it. Since we're building with BMI2 for AVX2 functions, let's also just make sure the CPU claims to support it (just to cover our bases).	2024-12-07 22:32:29 +01:00
Adam Stylinski	0ed5ac8289	Make an AVX512 inflate fast with low cost masked writes ... This takes advantage of the fact that on AVX512 architectures, masked moves are incredibly cheap. There are many places where we have to fallback to the safe C implementation of chunkcopy_safe because of the assumed overwriting that occurs. We're to sidestep most of the branching needed here by simply controlling the bounds of our writes with a mask.	2024-11-20 22:14:44 +01:00
Mika Lindqvist	1ec47b7245	configure: add --mandir to override $mandir on command line.	2024-10-10 13:21:32 +02:00
Mika Lindqvist	22a4cbbc5a	configure: Fix linker flags for Haiku.	2024-10-10 13:21:20 +02:00
Alexander Smorkalov	4549279dbf	Fixed false positive HAVE_ARMV6_INTRIN value on old ARM platforms.	2024-09-11 12:40:39 +02:00
Harmen Stoppels	5732b39b1c	don't use zlib-ng's -Wl,--version-script in tests (#1750) ... lld 18 errors when a version script assigns a version to a symbol that is not defined in the object files. Therefore configure scripts should not use zlib-ng's version script -- all tests will fail. Also test whether the linker supports the flag instead of assuming.	2024-06-28 13:21:33 +02:00
Un1q32	c5b4b35106	Improved ACLE check (#1727) ... Co-authored-by: Cameron Cawley <ccawley2011@gmail.com>	2024-06-13 13:23:29 +02:00
Hans Kristian Rosbach	5b208676f8	Clean up memory allocation functions that are no longer used, and its tests. ... Co-authored-by: Ilya Leoshkevich <iii@linux.ibm.com>	2024-05-28 16:35:13 +02:00
Hans Kristian Rosbach	130055e8d1	Rewrite deflate memory allocation. ... Deflate used to call allocate 5 times during init. - 5 calls to external alloc function now becomes 1 - Handling alignment of allocated buffers is simplified - Efforts to align the allocated buffer now needs to happen only once. - Individual buffers are ordered so that they have natural sequential alignment. - Due to reduced losses to alignment, we allocate less memory in total. - While doing alloc(), we now store pointer to corresponding free(), avoiding crashes with applications that incorrectly set alloc/free pointers after running init function. - Removed need for extra padding after window, chunked reads can now go beyond the window buffer without causing a segfault. Co-authored-by: Ilya Leoshkevich <iii@linux.ibm.com>	2024-05-28 16:35:13 +02:00
Vladislav Shchapov	d27ff46e8a	Fix -std option in configure script ... Signed-off-by: Vladislav Shchapov <vladislav@shchapov.ru>	2024-05-06 22:50:52 +02:00
Un1q32	ee618f6f19	make darwin cross compilation possible ... https://github.com/madler/zlib/pull/789	2024-05-06 10:03:57 +02:00
Hans Kristian Rosbach	9953f12e21	Move update_hash(), insert_string() and quick_insert_string() out of functable ... and remove SSE4.2 and ACLE optimizations. The functable overhead is higher than the benefit from using optimized functions.	2024-02-23 13:34:10 +01:00
Mika T. Lindqvist	9d945f0d71	Fix xsave intrinsic test for clang, and gcc 8.2 or later, and icc.	2024-02-18 10:10:45 +01:00
Mark Adler	13a775881a	Remove -w compile option in configure test. ... Not all C compilers have a -w option. madler/zlib#88ec24670e9c484219392def5f2f4b5e3f839db3	2024-02-07 19:15:56 +01:00
Nathan Moinvaziri	8ef6098a65	Enable LIT_MEM by default expect when WITH_REDUCED_MEM is ON.	2024-02-07 19:15:56 +01:00
Vladislav Shchapov	44e6bfcc5b	Remove unused macro X86_MASK_INTRIN. ... Signed-off-by: Vladislav Shchapov <vladislav@shchapov.ru>	2024-01-25 10:21:49 +01:00
Vladislav Shchapov	408ad95795	Reorder in which intrinsics support is checked. ... Signed-off-by: Vladislav Shchapov <vladislav@shchapov.ru>	2024-01-25 10:21:49 +01:00
Hans Kristian Rosbach	4e132cc0ec	Move adler32 C fallbacks to arch/generic	2024-01-19 15:22:34 +01:00
Hans Kristian Rosbach	6f38b4c5fc	Simplify includes	2024-01-19 15:22:34 +01:00
Hans Kristian Rosbach	0993a7e96c	Don't attempt ARMv6 detection on AARCH64.	2023-12-29 15:51:27 +01:00
Vladislav Shchapov	dfced56c57	Remove unused Z_TLS and HAVE_THREAD_LOCAL ... Signed-off-by: Vladislav Shchapov <vladislav@shchapov.ru>	2023-12-25 20:47:24 +01:00
Mika T. Lindqvist	6592accb25	[configure] Propagate -DWITH_GZFILEOP to pkgconfig.	2023-11-08 11:34:53 +01:00
Hajin Jang	f9228d8475	Support llvm-mingw toolchain ... zlib-ng requires some patches to make it compilable on LLVM-mingw. 1. Add -Wno-pedantic-ms-format only if a toolchain is MinGW GCC. - llvm-mingw does not support it, causing build to break. 2. Include arm_neon.h instead of arm64_neon.h (aarch64 only). - arm64_neon.h is MSVC only. - GCC, Clang does not have arm64_neon.h but arm_neon.h on aarch64. - Also applied to configure and detect-instrinsics.cmake	2023-09-28 00:15:12 +02:00
Cameron Cawley	16fe1f885e	Add ARMv6 version of slide_hash	2023-09-16 11:11:18 +02:00
Cameron Cawley	1c1e728637	Use GCC cpuid intrinsics with MinGW	2023-09-16 11:08:25 +02:00
Nathan Moinvaziri	4838f0e967	Merge 32-bit arm hardware flag detection sections in configure and cmake	2023-09-13 11:55:01 +02:00
Nathan Moinvaziri	881aa3e60e	Merge both without_optimizations sections in configure for arm.	2023-09-13 11:55:01 +02:00
Nathan Moinvaziri	f0ffa3d57a	Clean up detection on ARM variants and allow ACLE on all ARM archs.	2023-09-13 11:55:01 +02:00
Harmen Stoppels	ca2d4e5adc	cast _xgetbv to int to silence conversion warning	2023-09-13 11:54:42 +02:00
Harmen Stoppels	cf029967c2	another	2023-09-13 11:54:42 +02:00
Harmen Stoppels	b7ac847935	configure: ensure instructions are not optimized out before assembler is run	2023-09-13 11:54:42 +02:00