1
0
mirror of https://github.com/golang/go synced 2024-11-19 18:54:41 -07:00
go/src/math/asinh_s390x.s
Bill O'Farrell 88672de7af math: use SIMD to accelerate additional scalar math functions on s390x
As necessary, math functions were structured to use stubs, so that they can
be accelerated with assembly on any platform.

Technique used was minimax polynomial approximation using tables of
polynomial coefficients, with argument range reduction.

Benchmark         New     Old     Speedup
BenchmarkAcos     12.2    47.5    3.89
BenchmarkAcosh    18.5    56.2    3.04
BenchmarkAsin     13.1    40.6    3.10
BenchmarkAsinh    19.4    62.8    3.24
BenchmarkAtan     10.1    23      2.28
BenchmarkAtanh    19.1    53.2    2.79
BenchmarkAtan2    16.5    33.9    2.05
BenchmarkCbrt     14.8    58      3.92
BenchmarkErf      10.8    20.1    1.86
BenchmarkErfc     11.2    23.5    2.10
BenchmarkExp      8.77    53.8    6.13
BenchmarkExpm1    10.1    38.3    3.79
BenchmarkLog      13.1    40.1    3.06
BenchmarkLog1p    12.7    38.3    3.02
BenchmarkPowInt   31.7    40.5    1.28
BenchmarkPowFrac  33.1    141     4.26
BenchmarkTan      11.5    30      2.61

Accuracy was tested against a high precision
reference function to determine maximum error.
Note: ulperr is error in "units in the last place"

       max
      ulperr
Acos  1.15
Acosh 1.07
Asin  2.22
Asinh 1.72
Atan  1.41
Atanh 3.00
Atan2 1.45
Cbrt  1.18
Erf   1.29
Erfc  4.82
Exp   1.00
Expm1 2.26
Log   0.94
Log1p 2.39
Tan   3.14

Pow will have 99.99% correctly rounded results with reasonable inputs
producing numeric (non Inf or NaN) results

Change-Id: I850e8cf7b70426e8b54ec49d74acd4cddc8c6cb2
Reviewed-on: https://go-review.googlesource.com/38585
Reviewed-by: Michael Munday <munday@ca.ibm.com>
Run-TryBot: Michael Munday <munday@ca.ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-05-08 19:52:30 +00:00

224 lines
6.1 KiB
ArmAsm

// Copyright 2017 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
#include "textflag.h"
// Minimax polynomial coefficients and other constants
DATA ·asinhrodataL18<> + 0(SB)/8, $0.749999999977387502E-01
DATA ·asinhrodataL18<> + 8(SB)/8, $-.166666666666657082E+00
DATA ·asinhrodataL18<> + 16(SB)/8, $0.303819368237360639E-01
DATA ·asinhrodataL18<> + 24(SB)/8, $-.446428569571752982E-01
DATA ·asinhrodataL18<> + 32(SB)/8, $0.173500047922695924E-01
DATA ·asinhrodataL18<> + 40(SB)/8, $-.223719767210027185E-01
DATA ·asinhrodataL18<> + 48(SB)/8, $0.113655037946822130E-01
DATA ·asinhrodataL18<> + 56(SB)/8, $0.579747490622448943E-02
DATA ·asinhrodataL18<> + 64(SB)/8, $-.139372433914359122E-01
DATA ·asinhrodataL18<> + 72(SB)/8, $-.218674325255800840E-02
DATA ·asinhrodataL18<> + 80(SB)/8, $-.891074277756961157E-02
DATA ·asinhrodataL18<> + 88(SB)/8, $.41375273347623353626
DATA ·asinhrodataL18<> + 96(SB)/8, $.51487302528619766235E+04
DATA ·asinhrodataL18<> + 104(SB)/8, $-1.67526912689208984375
DATA ·asinhrodataL18<> + 112(SB)/8, $0.181818181818181826E+00
DATA ·asinhrodataL18<> + 120(SB)/8, $-.165289256198351540E-01
DATA ·asinhrodataL18<> + 128(SB)/8, $0.200350613573012186E-02
DATA ·asinhrodataL18<> + 136(SB)/8, $-.273205381970859341E-03
DATA ·asinhrodataL18<> + 144(SB)/8, $0.397389654305194527E-04
DATA ·asinhrodataL18<> + 152(SB)/8, $0.938370938292558173E-06
DATA ·asinhrodataL18<> + 160(SB)/8, $0.212881813645679599E-07
DATA ·asinhrodataL18<> + 168(SB)/8, $-.602107458843052029E-05
DATA ·asinhrodataL18<> + 176(SB)/8, $-.148682720127920854E-06
DATA ·asinhrodataL18<> + 184(SB)/8, $-5.5
DATA ·asinhrodataL18<> + 192(SB)/8, $1.0
DATA ·asinhrodataL18<> + 200(SB)/8, $1.0E-20
GLOBL ·asinhrodataL18<> + 0(SB), RODATA, $208
// Table of log correction terms
DATA ·asinhtab2080<> + 0(SB)/8, $0.585235384085551248E-01
DATA ·asinhtab2080<> + 8(SB)/8, $0.412206153771168640E-01
DATA ·asinhtab2080<> + 16(SB)/8, $0.273839003221648339E-01
DATA ·asinhtab2080<> + 24(SB)/8, $0.166383778368856480E-01
DATA ·asinhtab2080<> + 32(SB)/8, $0.866678223433169637E-02
DATA ·asinhtab2080<> + 40(SB)/8, $0.319831684989627514E-02
DATA ·asinhtab2080<> + 48(SB)/8, $0.0
DATA ·asinhtab2080<> + 56(SB)/8, $-.113006378583725549E-02
DATA ·asinhtab2080<> + 64(SB)/8, $-.367979419636602491E-03
DATA ·asinhtab2080<> + 72(SB)/8, $0.213172484510484979E-02
DATA ·asinhtab2080<> + 80(SB)/8, $0.623271047682013536E-02
DATA ·asinhtab2080<> + 88(SB)/8, $0.118140812789696885E-01
DATA ·asinhtab2080<> + 96(SB)/8, $0.187681358930914206E-01
DATA ·asinhtab2080<> + 104(SB)/8, $0.269985148668178992E-01
DATA ·asinhtab2080<> + 112(SB)/8, $0.364186619761331328E-01
DATA ·asinhtab2080<> + 120(SB)/8, $0.469505379381388441E-01
GLOBL ·asinhtab2080<> + 0(SB), RODATA, $128
// Asinh returns the inverse hyperbolic sine of the argument.
//
// Special cases are:
// Asinh(±0) = ±0
// Asinh(±Inf) = ±Inf
// Asinh(NaN) = NaN
// The algorithm used is minimax polynomial approximation
// with coefficients determined with a Remez exchange algorithm.
TEXT ·asinhAsm(SB), NOSPLIT, $0-16
FMOVD x+0(FP), F0
MOVD $·asinhrodataL18<>+0(SB), R9
WORD $0xB3CD00C0 //lgdr %r12, %f0
WORD $0xC0293FDF //iilf %r2,1071644671
BYTE $0xFF
BYTE $0xFF
SRAD $32, R12
WORD $0xB917001C //llgtr %r1,%r12
MOVW R1, R6
MOVW R2, R7
CMPBLE R6, R7, L2
WORD $0xC0295FEF //iilf %r2,1609564159
BYTE $0xFF
BYTE $0xFF
MOVW R2, R7
CMPBLE R6, R7, L14
L3:
WORD $0xC0297FEF //iilf %r2,2146435071
BYTE $0xFF
BYTE $0xFF
CMPW R1, R2
BGT L1
WORD $0xB3120000 //ltdbr %f0,%f0
FMOVD F0, F10
BLTU L15
L9:
FMOVD $0, F0
WFADB V0, V10, V0
WORD $0xC0398006 //iilf %r3,2147909631
BYTE $0x7F
BYTE $0xFF
WORD $0xB3CD0050 //lgdr %r5, %f0
SRAD $32, R5
MOVH $0x0, R2
SUBW R5, R3
FMOVD $0, F8
WORD $0xEC4320AF //risbg %r4,%r3,32,128+47,0
BYTE $0x00
BYTE $0x55
BYTE $0x18 //lr %r1,%r4
BYTE $0x14
WORD $0xEC24001F //risbgn %r2,%r4,64-64+0,64-64+0+32-1,64-0-32
BYTE $0x20
BYTE $0x59
SUBW $0x100000, R1
SRAW $8, R1, R1
ORW $0x45000000, R1
BR L6
L2:
MOVD $0x30000000, R2
CMPW R1, R2
BGT L16
FMOVD 200(R9), F2
FMADD F2, F0, F0
L1:
FMOVD F0, ret+8(FP)
RET
L14:
WORD $0xB3120000 //ltdbr %f0,%f0
BLTU L17
FMOVD F0, F10
L4:
FMOVD 192(R9), F2
WFMADB V0, V0, V2, V0
WORD $0xB3120000 //ltdbr %f0,%f0
FSQRT F0, F8
L5:
WFADB V8, V10, V0
WORD $0xC0398006 //iilf %r3,2147909631
BYTE $0x7F
BYTE $0xFF
WORD $0xB3CD0050 //lgdr %r5, %f0
SRAD $32, R5
MOVH $0x0, R2
SUBW R5, R3
WORD $0xEC4320AF //risbg %r4,%r3,32,128+47,0
BYTE $0x00
BYTE $0x55
SRAW $8, R4, R1
WORD $0xEC24001F //risbgn %r2,%r4,64-64+0,64-64+0+32-1,64-0-32
BYTE $0x20
BYTE $0x59
ORW $0x45000000, R1
L6:
WORD $0xB3C10022 //ldgr %f2,%r2
FMOVD 184(R9), F0
WFMADB V8, V2, V0, V8
FMOVD 176(R9), F4
WFMADB V10, V2, V8, V2
FMOVD 168(R9), F0
FMOVD 160(R9), F6
FMOVD 152(R9), F1
WFMADB V2, V6, V4, V6
WFMADB V2, V1, V0, V1
WFMDB V2, V2, V4
FMOVD 144(R9), F0
WFMADB V6, V4, V1, V6
FMOVD 136(R9), F1
WORD $0xEC3339BC //risbg %r3,%r3,57,128+60,64-13
BYTE $0x33
BYTE $0x55
WFMADB V2, V0, V1, V0
FMOVD 128(R9), F1
WFMADB V4, V6, V0, V6
FMOVD 120(R9), F0
WFMADB V2, V1, V0, V1
VLVGF $0, R1, V0
WFMADB V4, V6, V1, V4
LDEBR F0, F0
FMOVD 112(R9), F6
WFMADB V2, V4, V6, V4
MOVD $·asinhtab2080<>+0(SB), R1
FMOVD 104(R9), F1
WORD $0x68331000 //ld %f3,0(%r3,%r1)
FMOVD 96(R9), F6
WFMADB V2, V4, V3, V2
WFMADB V0, V1, V6, V0
FMOVD 88(R9), F4
WFMADB V0, V4, V2, V0
MOVD R12, R6
CMPBGT R6, $0, L1
WORD $0xB3130000 //lcdbr %f0,%f0
FMOVD F0, ret+8(FP)
RET
L16:
WFMDB V0, V0, V1
FMOVD 80(R9), F6
WFMDB V1, V1, V4
FMOVD 72(R9), F2
WFMADB V4, V2, V6, V2
FMOVD 64(R9), F3
FMOVD 56(R9), F6
WFMADB V4, V2, V3, V2
FMOVD 48(R9), F3
WFMADB V4, V6, V3, V6
FMOVD 40(R9), F5
FMOVD 32(R9), F3
WFMADB V4, V2, V5, V2
WFMADB V4, V6, V3, V6
FMOVD 24(R9), F5
FMOVD 16(R9), F3
WFMADB V4, V2, V5, V2
WFMADB V4, V6, V3, V6
FMOVD 8(R9), F5
FMOVD 0(R9), F3
WFMADB V4, V2, V5, V2
WFMADB V4, V6, V3, V4
WFMDB V0, V1, V6
WFMADB V1, V4, V2, V4
FMADD F4, F6, F0
FMOVD F0, ret+8(FP)
RET
L17:
WORD $0xB31300A0 //lcdbr %f10,%f0
BR L4
L15:
WORD $0xB31300A0 //lcdbr %f10,%f0
BR L9