summary refs log tree commit diff
path: root/sysdeps/ieee754/flt-32/math_config.h
Commit message (Collapse)AuthorAgeFilesLines
* aarch64: Add vector implementations of cos routinesJoe Ramsay2023-06-301-2/+0
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | Replace the loop-over-scalar placeholder routines with optimised implementations from Arm Optimized Routines (AOR). Also add some headers containing utilities for aarch64 libmvec routines, and update libm-test-ulps. Data tables for new routines are used via a pointer with a barrier on it, in order to prevent overly aggressive constant inlining in GCC. This allows a single adrp, combined with offset loads, to be used for every constant in the table. Special-case handlers are marked NOINLINE in order to confine the save/restore overhead of switching from vector to normal calling standard. This way we only incur the extra memory access in the exceptional cases. NOINLINE definitions have been moved to math_private.h in order to reduce duplication. AOR exposes a config option, WANT_SIMD_EXCEPT, to enable selective masking (and later fixing up) of invalid lanes, in order to trigger fp exceptions correctly (AdvSIMD only). This is tested and maintained in AOR, however it is configured off at source level here for performance reasons. We keep the WANT_SIMD_EXCEPT blocks in routine sources to greatly simplify the upstreaming process from AOR to glibc. Reviewed-by: Szabolcs Nagy <szabolcs.nagy@arm.com>
* math: Remove the error handling wrapper from fmod and fmodfAdhemerval Zanella Netto2023-04-031-0/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The error handling is moved to sysdeps/ieee754 version with no SVID support. The compatibility symbol versions still use the wrapper with SVID error handling around the new code. There is no new symbol version nor compatibility code on !LIBM_SVID_COMPAT targets (e.g. riscv). The ia64 is unchanged, since it still uses the arch specific __libm_error_region on its implementation. For both i686 and m68k, which provive arch specific implementation, wrappers are added so no new symbol are added (which would require to change the implementations). It shows an small improvement, the results for fmod: Architecture | Input | master | patch -----------------|-----------------|----------|-------- x86_64 (Ryzen 9) | subnormals | 12.5049 | 9.40992 x86_64 (Ryzen 9) | normal | 296.939 | 296.738 x86_64 (Ryzen 9) | close-exponents | 16.0244 | 13.119 aarch64 (N1) | subnormal | 6.81778 | 4.33313 aarch64 (N1) | normal | 155.620 | 152.915 aarch64 (N1) | close-exponents | 8.21306 | 5.76138 armhf (N1) | subnormal | 15.1083 | 14.5746 armhf (N1) | normal | 244.833 | 241.738 armhf (N1) | close-exponents | 21.8182 | 22.457 Checked on x86_64-linux-gnu, i686-linux-gnu, and aarch64-linux-gnu. Reviewed-by: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
* math: Improve fmodfAdhemerval Zanella Netto2023-04-031-0/+41
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This uses a new algorithm similar to already proposed earlier [1]. With x = mx * 2^ex and y = my * 2^ey (mx, my, ex, ey being integers), the simplest implementation is: mx * 2^ex == 2 * mx * 2^(ex - 1) while (ex > ey) { mx *= 2; --ex; mx %= my; } With mx/my being mantissa of double floating pointer, on each step the argument reduction can be improved 8 (which is sizeof of uint32_t minus MANTISSA_WIDTH plus the signal bit): while (ex > ey) { mx << 8; ex -= 8; mx %= my; } */ The implementation uses builtin clz and ctz, along with shifts to convert hx/hy back to doubles. Different than the original patch, this path assume modulo/divide operation is slow, so use multiplication with invert values. I see the following performance improvements using fmod benchtests (result only show the 'mean' result): Architecture | Input | master | patch -----------------|-----------------|----------|-------- x86_64 (Ryzen 9) | subnormals | 17.2549 | 12.0318 x86_64 (Ryzen 9) | normal | 85.4096 | 49.9641 x86_64 (Ryzen 9) | close-exponents | 19.1072 | 15.8224 aarch64 (N1) | subnormal | 10.2182 | 6.81778 aarch64 (N1) | normal | 60.0616 | 20.3667 aarch64 (N1) | close-exponents | 11.5256 | 8.39685 I also see similar improvements on arm-linux-gnueabihf when running on the N1 aarch64 chips, where it a lot of soft-fp implementation (for modulo, and multiplication): Architecture | Input | master | patch -----------------|-----------------|----------|-------- armhf (N1) | subnormal | 11.6662 | 10.8955 armhf (N1) | normal | 69.2759 | 34.1524 armhf (N1) | close-exponents | 13.6472 | 18.2131 Instead of using the math_private.h definitions, I used the math_config.h instead which is used on newer math implementations. Co-authored-by: kirill <kirill.okhotnikov@gmail.com> [1] https://sourceware.org/pipermail/libc-alpha/2020-November/119794.html Reviewed-by: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
* Update copyright dates with scripts/update-copyrightsJoseph Myers2023-01-061-1/+1
|
* Update copyright dates with scripts/update-copyrightsPaul Eggert2022-01-011-1/+1
| | | | | | | | | | | | | | | | | | | | | | | I used these shell commands: ../glibc/scripts/update-copyrights $PWD/../gnulib/build-aux/update-copyright (cd ../glibc && git commit -am"[this commit message]") and then ignored the output, which consisted lines saying "FOO: warning: copyright statement not found" for each of 7061 files FOO. I then removed trailing white space from math/tgmath.h, support/tst-support-open-dev-null-range.c, and sysdeps/x86_64/multiarch/strlen-vec.S, to work around the following obscure pre-commit check failure diagnostics from Savannah. I don't know why I run into these diagnostics whereas others evidently do not. remote: *** 912-#endif remote: *** 913: remote: *** 914- remote: *** error: lines with trailing whitespace found ... remote: *** error: sysdeps/unix/sysv/linux/statx_cp.c: trailing lines
* math: Simplify hypotf implementationAdhemerval Zanella2021-12-131-0/+9
| | | | | | | | | | | | Use a more optimized comparison for check for NaN and infinite and add an inlined issignaling implementation for float. With gcc it results in 2 FP comparisons. The file Copyright is also changed to use GPL, the implementation was completely changed by 7c10fd3515f to use double precision instead of scaling and this change removes all the GET_FLOAT_WORD usage. Checked on x86_64-linux-gnu.
* Update copyright dates with scripts/update-copyrightsPaul Eggert2021-01-021-1/+1
| | | | | | | | | | | | | | | | I used these shell commands: ../glibc/scripts/update-copyrights $PWD/../gnulib/build-aux/update-copyright (cd ../glibc && git commit -am"[this commit message]") and then ignored the output, which consisted lines saying "FOO: warning: copyright statement not found" for each of 6694 files FOO. I then removed trailing white space from benchtests/bench-pthread-locks.c and iconvdata/tst-iconv-big5-hkscs-to-2ucs4.c, to work around this diagnostic from Savannah: remote: *** pre-commit check failed ... remote: *** error: lines with trailing whitespace found remote: error: hook declined to update refs/heads/master
* math: Optimized generic exp10f with wrappersPaul Zimmermann2020-06-191-1/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | It is inspired by expf and reuses its tables and internal functions. The error checks are inlined and errno setting is in separate tail called functions, but the wrappers are kept in this patch to handle the _LIB_VERSION==_SVID_ case. Double precision arithmetics is used which is expected to be faster on most targets (including soft-float) than using single precision and it is easier to get good precision result with it. Result for x86_64 (i7-4790K CPU @ 4.00GHz) are: Before new code: "exp10f": { "workload-spec2017.wrf (adapted)": { "duration": 4.0414e+09, "iterations": 1.00128e+08, "reciprocal-throughput": 26.6818, "latency": 54.043, "max-throughput": 3.74787e+07, "min-throughput": 1.85038e+07 } With new code: "exp10f": { "workload-spec2017.wrf (adapted)": { "duration": 4.11951e+09, "iterations": 1.23968e+08, "reciprocal-throughput": 21.0581, "latency": 45.4028, "max-throughput": 4.74876e+07, "min-throughput": 2.20251e+07 } Result for aarch64 (A72 @ 2GHz) are: Before new code: "exp10f": { "workload-spec2017.wrf (adapted)": { "duration": 4.62362e+09, "iterations": 3.3376e+07, "reciprocal-throughput": 127.698, "latency": 149.365, "max-throughput": 7.831e+06, "min-throughput": 6.69501e+06 } With new code: "exp10f": { "workload-spec2017.wrf (adapted)": { "duration": 4.29108e+09, "iterations": 6.6752e+07, "reciprocal-throughput": 51.2111, "latency": 77.3568, "max-throughput": 1.9527e+07, "min-throughput": 1.29271e+07 } Checked on x86_64-linux-gnu, powerpc64le-linux-gnu, aarch64-linux-gnu, and sparc64-linux-gnu.
* Update copyright dates with scripts/update-copyrights.Joseph Myers2020-01-011-1/+1
|
* Remove duplicate inline implementation of issignalingfPaul A. Clarke2019-11-221-9/+0
| | | | | | | | | | | | | | | | | | | | Very recent commit 854e91bf6b4221f424ffa13b9ef50f35623b7b74 enabled inline of issignalingf() in general (__issignalingf in include/math.h). There is another implementation for an inline use of issignalingf (issignalingf_inline in sysdeps/ieee754/flt-32/math_config.h) which could instead make use of the new enablement. Replace the use of issignalingf_inline with __issignaling. Using issignaling (instead of __issignalingf) will allow future enhancements to the type-generic implementation, issignaling, to be automatically adopted. The implementations are slightly different, and compile to slightly different code, but I measured no significant performance difference. The second implementation was brought to my attention by: Suggested-by: Joseph Myers <joseph@codesourcery.com> Reviewed-by: Joseph Myers <joseph@codesourcery.com>
* Prefer https to http for gnu.org and fsf.org URLsPaul Eggert2019-09-071-1/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Also, change sources.redhat.com to sourceware.org. This patch was automatically generated by running the following shell script, which uses GNU sed, and which avoids modifying files imported from upstream: sed -ri ' s,(http|ftp)(://(.*\.)?(gnu|fsf|sourceware)\.org($|[^.]|\.[^a-z])),https\2,g s,(http|ftp)(://(.*\.)?)sources\.redhat\.com($|[^.]|\.[^a-z]),https\2sourceware.org\4,g ' \ $(find $(git ls-files) -prune -type f \ ! -name '*.po' \ ! -name 'ChangeLog*' \ ! -path COPYING ! -path COPYING.LIB \ ! -path manual/fdl-1.3.texi ! -path manual/lgpl-2.1.texi \ ! -path manual/texinfo.tex ! -path scripts/config.guess \ ! -path scripts/config.sub ! -path scripts/install-sh \ ! -path scripts/mkinstalldirs ! -path scripts/move-if-change \ ! -path INSTALL ! -path locale/programs/charmap-kw.h \ ! -path po/libc.pot ! -path sysdeps/gnu/errlist.c \ ! '(' -name configure \ -execdir test -f configure.ac -o -f configure.in ';' ')' \ ! '(' -name preconfigure \ -execdir test -f preconfigure.ac ';' ')' \ -print) and then by running 'make dist-prepare' to regenerate files built from the altered files, and then executing the following to cleanup: chmod a+x sysdeps/unix/sysv/linux/riscv/configure # Omit irrelevant whitespace and comment-only changes, # perhaps from a slightly-different Autoconf version. git checkout -f \ sysdeps/csky/configure \ sysdeps/hppa/configure \ sysdeps/riscv/configure \ sysdeps/unix/sysv/linux/csky/configure # Omit changes that caused a pre-commit check to fail like this: # remote: *** error: sysdeps/powerpc/powerpc64/ppc-mcount.S: trailing lines git checkout -f \ sysdeps/powerpc/powerpc64/ppc-mcount.S \ sysdeps/unix/sysv/linux/s390/s390-64/syscall.S # Omit change that caused a pre-commit check to fail like this: # remote: *** error: sysdeps/sparc/sparc64/multiarch/memcpy-ultra3.S: last line does not end in newline git checkout -f sysdeps/sparc/sparc64/multiarch/memcpy-ultra3.S
* Update copyright dates with scripts/update-copyrights.Joseph Myers2019-01-011-1/+1
| | | | | | | * All files with FSF copyright notices: Update copyright dates using scripts/update-copyrights. * locale/programs/charmap-kw.h: Regenerated. * locale/programs/locfile-kw.h: Likewise.
* Clean up converttoint handling and document the semanticsSzabolcs Nagy2018-08-101-5/+15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This patch currently only affects aarch64. The roundtoint and converttoint internal functions are only called with small values, so 32 bit result is enough for converttoint and it is a signed int conversion so the return type is changed to int32_t. The original idea was to help the compiler keeping the result in uint64_t, then it's clear that no sign extension is needed and there is no accidental undefined or implementation defined signed int arithmetics. But it turns out gcc does a good job with inlining so changing the type has no overhead and the semantics of the conversion is less surprising this way. Since we want to allow the asuint64 (x + 0x1.8p52) style conversion, the top bits were never usable and the existing code ensures that only the bottom 32 bits of the conversion result are used. On aarch64 the neon intrinsics (which round ties to even) are changed to round and lround (which round ties away from zero) this does not affect the results in a significant way, but more portable (relies on round and lround being inlined which works with -fno-math-errno). The TOINT_SHIFT and TOINT_RINT macros were removed, only keep separate code paths for TOINT_INTRINSICS and !TOINT_INTRINSICS. * sysdeps/aarch64/fpu/math_private.h (roundtoint): Use round. (converttoint): Use lround. * sysdeps/ieee754/flt-32/math_config.h (roundtoint): Declare and document the semantics when TOINT_INTRINSICS is set. (converttoint): Likewise. (TOINT_RINT): Remove. (TOINT_SHIFT): Remove. * sysdeps/ieee754/flt-32/e_expf.c (__expf): Remove the TOINT_RINT code path.
* Use uint32_t sign in single precision math error handling functionsSzabolcs Nagy2018-07-021-4/+4
| | | | | | | | | | | | | | | | | | | | | Ideally sign should be bool, but sometimes (e.g. in powf) it's more efficient to pass a non-zero value than 1 to indicate that the sign should be set. The fixed size int is less ambigous than unsigned long. * sysdeps/ieee754/flt-32/e_powf.c (__powf): Use uint32_t. (exp2f_inline): Likewise. * sysdeps/ieee754/flt-32/math_config.h (__math_oflowf): Likewise. (__math_uflowf): Likewise. (__math_may_uflowf): Likewise. (__math_divzerof): Likewise. (__math_invalidf): Likewise. * sysdeps/ieee754/flt-32/math_errf.c (xflowf): Likewise. (__math_oflowf): Likewise. (__math_uflowf): Likewise. (__math_may_uflowf): Likewise. (__math_divzerof): Likewise. (__math_invalidf): Likewise.
* Update copyright dates with scripts/update-copyrights.Joseph Myers2018-01-011-1/+1
| | | | | | | * All files with FSF copyright notices: Update copyright dates using scripts/update-copyrights. * locale/programs/charmap-kw.h: Regenerated. * locale/programs/locfile-kw.h: Likewise.
* New generic powfSzabolcs Nagy2017-09-291-0/+27
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | without wrapper on aarch64: powf reciprocal-throughput: 4.2x faster powf latency: 2.6x faster old worst-case error: 1.11 ulp new worst-case error: 0.82 ulp aarch64 .text size: -780 bytes aarch64 .rodata size: +144 bytes powf(x,y) is implemented as exp2(y*log2(x)) with the same algorithms that are used in exp2f and log2f, except that the log2f polynomial is larger for extra precision and its output (and exp2f input) may be scaled by a power of 2 (POWF_SCALE) to simplify the argument reduction step of exp2 (possible when efficient round and convert toint operation is available). The special case handling tries to minimize the checks in the hot path. When the input of exp2_inline is checked, int arithmetics is used as that was faster on the tested aarch64 cores. * math/Makefile (type-float-routines): Add e_powf_log2_data. * sysdeps/ieee754/flt-32/e_powf.c: New implementation. * sysdeps/ieee754/flt-32/e_powf_log2_data.c: New file. * sysdeps/ieee754/flt-32/math_config.h (__powf_log2_data): Define. (issignalingf_inline): Likewise. (POWF_LOG2_TABLE_BITS): Likewise. (POWF_LOG2_POLY_ORDER): Likewise. (POWF_SCALE_BITS): Likewise. (POWF_SCALE): Likewise. * sysdeps/i386/fpu/e_powf_log2_data.c: New file. * sysdeps/ia64/fpu/e_powf_log2_data.c: New file. * sysdeps/m68k/m680x0/fpu/e_powf_log2_data.c: New file.
* New generic log2fSzabolcs Nagy2017-09-291-0/+11
| | | | | | | | | | | | | | | | | | | | | | | Similar to the new logf: double precision arithmetics and a small lookup table is used. The argument reduction step is the same as in the new logf. without wrapper on aarch64: log2f reciprocal-throughput: 2.3x faster log2f latency: 2.1x faster old worst case error: 1.72 ulp new worst case error: 0.75 ulp aarch64 .text size: -252 bytes aarch64 .rodata size: +244 bytes * math/Makefile (type-float-routines): Add e_log2f_data. * sysdeps/ieee754/flt-32/e_log2f.c: New implementation. * sysdeps/ieee754/flt-32/e_log2f_data.c: New file. * sysdeps/ieee754/flt-32/math_config.h (__log2f_data): Define. (LOG2F_TABLE_BITS, LOG2F_POLY_ORDER): Define. * sysdeps/i386/fpu/e_log2f_data.c: New file. * sysdeps/ia64/fpu/e_log2f_data.c: New file. * sysdeps/m68k/m680x0/fpu/e_log2f_data.c: New file.
* New generic logfSzabolcs Nagy2017-09-291-0/+12
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | without wrapper on aarch64: logf reciprocal-throughput: 2.2x faster logf latency: 1.9x faster old worst case error: 0.89 ulp new worst case error: 0.82 ulp aarch64 .text size: -356 bytes aarch64 .rodata size: +240 bytes Uses double precision arithmetics and a lookup table to allow smaller polynomial and avoid the use of division. Data is in a separate translation unit with fixed layout to prevent the compiler generating suboptimal literal access. Errors are handled inline according to POSIX rules, but this patch keeps the wrapper with SVID compatible error handling. Needs libm-test-ulps adjustment for clogf in non-nearest rounding mode. * math/Makefile (type-float-routines): Add e_logf_data. * sysdeps/ieee754/flt-32/e_logf.c: New implementation. * sysdeps/ieee754/flt-32/e_logf_data.c: New file. * sysdeps/ieee754/flt-32/math_config.h (__logf_data): Define. (LOGF_TABLE_BITS, LOGF_POLY_ORDER): Define. * sysdeps/i386/fpu/e_logf_data.c: New file. * sysdeps/ia64/fpu/e_logf_data.c: New file. * sysdeps/m68k/m680x0/fpu/e_logf_data.c: New file.
* Optimized generic expf and exp2f with wrappersSzabolcs Nagy2017-09-251-0/+114
Based on new expf and exp2f code from https://github.com/ARM-software/optimized-routines/ with wrapper on aarch64: expf reciprocal-throughput: 2.3x faster expf latency: 1.7x faster without wrapper on aarch64: expf reciprocal-throughput: 3.3x faster expf latency: 1.7x faster without wrapper on aarch64: exp2f reciprocal-throughput: 2.8x faster exp2f latency: 1.3x faster libm.so size on aarch64: .text size: -152 bytes .rodata size: -1740 bytes expf/exp2f worst case nearest rounding error: 0.502 ulp worst case non-nearest rounding error: 1 ulp Error checks are inline and errno setting is in separate tail called functions, but the wrappers are kept in this patch to handle the _LIB_VERSION==_SVID_ case. (So e.g. errno is set twice for expf calls and once for __expf_finite calls on targets where the new code is used.) Double precision arithmetics is used which is expected to be faster on most targets (including soft-float) than using single precision and it is easier to get good precision result with it. Const data is kept in a separate translation unit which complicates maintenance a bit, but is expected to give good code for literal loads on most targets and allows sharing data across expf, exp2f and powf. (This data is disabled on i386, m68k and ia64 which have their own expf, exp2f and powf code.) Some details may need target specific tweaks: - best convert and round to int operation in the arg reduction may be different across targets. - code was optimized on fma target, optimal polynomial eval may be different without fma. - gcc does not always generate good code for fp bit representation access via unions or it may be inherently slow on some targets. The libm-test-ulps will need adjustment because.. - The argument reduction ideally uses nearest rounded rint, but that is not efficient on most targets, so the polynomial can get evaluated on a wider interval in non-nearest rounding mode making 1 ulp errors common in that case. - The polynomial is evaluated such that it may have 1 ulp error on negative tiny inputs with upward rounding. * math/Makefile (type-float-routines): Add math_errf and e_exp2f_data. * sysdeps/aarch64/fpu/math_private.h (TOINT_INTRINSICS): Define. (roundtoint, converttoint): Likewise. * sysdeps/ieee754/flt-32/e_expf.c: New implementation. * sysdeps/ieee754/flt-32/e_exp2f.c: New implementation. * sysdeps/ieee754/flt-32/e_exp2f_data.c: New file. * sysdeps/ieee754/flt-32/math_config.h: New file. * sysdeps/ieee754/flt-32/math_errf.c: New file. * sysdeps/ieee754/flt-32/t_exp2f.h: Remove. * sysdeps/i386/fpu/e_exp2f_data.c: New file. * sysdeps/i386/fpu/math_errf.c: New file. * sysdeps/ia64/fpu/e_exp2f_data.c: New file. * sysdeps/ia64/fpu/math_errf.c: New file. * sysdeps/m68k/m680x0/fpu/e_exp2f_data.c: New file. * sysdeps/m68k/m680x0/fpu/math_errf.c: New file.
generated by cgit-pink 1.4.1 (git 2.36.1) at 2024-08-15 17:45:09 +0000