diff options
| author | Szabolcs Nagy <szabolcs.nagy@arm.com> | 2017-09-04 17:55:33 +0100 |
|---|---|---|
| committer | Szabolcs Nagy <szabolcs.nagy@arm.com> | 2017-09-29 11:51:12 +0100 |
| commit | e260a9f8a9279c231405593c449e1f5bd39b3fd1 (patch) | |
| tree | 7b80b0dc7fab5bec38a78ebbf13b940386873294 /sysdeps/ieee754/flt-32/math_config.h | |
| parent | eb38f71b3504ac132b5c65ba23a15ddb9d02a435 (diff) | |
| download | glibc-e260a9f8a9279c231405593c449e1f5bd39b3fd1.tar.gz glibc-e260a9f8a9279c231405593c449e1f5bd39b3fd1.tar.xz glibc-e260a9f8a9279c231405593c449e1f5bd39b3fd1.zip | |
New generic powf
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. 2017-09-19 Szabolcs Nagy <szabolcs.nagy@arm.com> * 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_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.
Diffstat (limited to 'sysdeps/ieee754/flt-32/math_config.h')
| -rw-r--r-- | sysdeps/ieee754/flt-32/math_config.h | 27 |
1 files changed, 27 insertions, 0 deletions
diff --git a/sysdeps/ieee754/flt-32/math_config.h b/sysdeps/ieee754/flt-32/math_config.h index f869fbc66c..7e78cb0c96 100644 --- a/sysdeps/ieee754/flt-32/math_config.h +++ b/sysdeps/ieee754/flt-32/math_config.h @@ -21,6 +21,7 @@ #include <math.h> #include <math_private.h> +#include <nan-high-order-bit.h> #include <stdint.h> #ifndef WANT_ROUNDING @@ -90,6 +91,15 @@ asdouble (uint64_t i) return u.f; } +static inline int +issignalingf_inline (float x) +{ + uint32_t ix = asuint (x); + if (HIGH_ORDER_BIT_IS_SET_FOR_SNAN) + return (ix & 0x7fc00000) == 0x7fc00000; + return 2 * (ix ^ 0x00400000) > 2u * 0x7fc00000; +} + #define NOINLINE __attribute__ ((noinline)) attribute_hidden float __math_oflowf (unsigned long); @@ -134,4 +144,21 @@ extern const struct log2f_data double poly[LOG2F_POLY_ORDER]; } __log2f_data attribute_hidden; +#define POWF_LOG2_TABLE_BITS 4 +#define POWF_LOG2_POLY_ORDER 5 +#if TOINT_INTRINSICS +#define POWF_SCALE_BITS EXP2F_TABLE_BITS +#else +#define POWF_SCALE_BITS 0 +#endif +#define POWF_SCALE ((double) (1 << POWF_SCALE_BITS)) +extern const struct powf_log2_data +{ + struct + { + double invc, logc; + } tab[1 << POWF_LOG2_TABLE_BITS]; + double poly[POWF_LOG2_POLY_ORDER]; +} __powf_log2_data attribute_hidden; + #endif |