/* Function sincosf vectorized with AVX-512. KNL and SKX versions.
Copyright (C) 2014-2015 Free Software Foundation, Inc.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, see
. */
#include
#include "svml_s_trig_data.h"
#include "svml_s_wrapper_impl.h"
/*
ALGORITHM DESCRIPTION:
1) Range reduction to [-Pi/4; +Pi/4] interval
a) Grab sign from source argument and save it.
b) Remove sign using AND operation
c) Getting octant Y by 2/Pi multiplication
d) Add "Right Shifter" value
e) Treat obtained value as integer S for destination sign setting.
SS = ((S-S&1)&2)<<30; For sin part
SC = ((S+S&1)&2)<<30; For cos part
f) Change destination sign if source sign is negative
using XOR operation.
g) Subtract "Right Shifter" (0x4B000000) value
h) Subtract Y*(PI/2) from X argument, where PI/2 divided to 4 parts:
X = X - Y*PI1 - Y*PI2 - Y*PI3 - Y*PI4;
2) Polynomial (minimax for sin within [-Pi/4; +Pi/4] interval)
a) Calculate X^2 = X * X
b) Calculate 2 polynomials for sin and cos:
RS = X * ( A0 + X^2 * (A1 + x^2 * (A2 + x^2 * (A3))));
RC = B0 + X^2 * (B1 + x^2 * (B2 + x^2 * (B3 + x^2 * (B4))));
c) Swap RS & RC if if first bit of obtained value after
Right Shifting is set to 1. Using And, Andnot & Or operations.
3) Destination sign setting
a) Set shifted destination sign using XOR operation:
R1 = XOR( RS, SS );
R2 = XOR( RC, SC ). */
.text
ENTRY (_ZGVeN16vvv_sincosf_knl)
#ifndef HAVE_AVX512_ASM_SUPPORT
WRAPPER_IMPL_AVX512_fFF _ZGVdN8vvv_sincosf
#else
pushq %rbp
cfi_adjust_cfa_offset (8)
cfi_rel_offset (%rbp, 0)
movq %rsp, %rbp
cfi_def_cfa_register (%rbp)
andq $-64, %rsp
subq $1344, %rsp
movq __svml_s_trig_data@GOTPCREL(%rip), %rax
vmovaps %zmm0, %zmm2
movl $-1, %edx
vmovups __sAbsMask(%rax), %zmm0
vmovups __sInvPI(%rax), %zmm3
/* Absolute argument computation */
vpandd %zmm0, %zmm2, %zmm1
vmovups __sPI1_FMA(%rax), %zmm5
vmovups __sSignMask(%rax), %zmm9
vpandnd %zmm2, %zmm0, %zmm0
/* h) Subtract Y*(PI/2) from X argument, where PI/2 divided to 3 parts:
X = X - Y*PI1 - Y*PI2 - Y*PI3 */
vmovaps %zmm1, %zmm6
vmovaps %zmm1, %zmm8
/* c) Getting octant Y by 2/Pi multiplication
d) Add "Right Shifter" value */
vfmadd213ps __sRShifter(%rax), %zmm1, %zmm3
vmovups __sPI3_FMA(%rax), %zmm7
/* g) Subtract "Right Shifter" (0x4B000000) value */
vsubps __sRShifter(%rax), %zmm3, %zmm12
/* e) Treat obtained value as integer S for destination sign setting */
vpslld $31, %zmm3, %zmm13
vmovups __sA7_FMA(%rax), %zmm14
vfnmadd231ps %zmm12, %zmm5, %zmm6
/* 2) Polynomial (minimax for sin within [-Pi/4; +Pi/4] interval)
a) Calculate X^2 = X * X
b) Calculate 2 polynomials for sin and cos:
RS = X * ( A0 + X^2 * (A1 + x^2 * (A2 + x^2 * (A3))));
RC = B0 + X^2 * (B1 + x^2 * (B2 + x^2 * (B3 + x^2 * (B4)))) */
vmovaps %zmm14, %zmm15
vmovups __sA9_FMA(%rax), %zmm3
vcmpps $22, __sRangeReductionVal(%rax), %zmm1, %k1
vpbroadcastd %edx, %zmm1{%k1}{z}
vfnmadd231ps __sPI2_FMA(%rax), %zmm12, %zmm6
vptestmd %zmm1, %zmm1, %k0
vpandd %zmm6, %zmm9, %zmm11
kmovw %k0, %ecx
vpxord __sOneHalf(%rax), %zmm11, %zmm4
/* Result sign calculations */
vpternlogd $150, %zmm13, %zmm9, %zmm11
/* Add correction term 0.5 for cos() part */
vaddps %zmm4, %zmm12, %zmm10
vfnmadd213ps %zmm6, %zmm7, %zmm12
vfnmadd231ps %zmm10, %zmm5, %zmm8
vpxord %zmm13, %zmm12, %zmm13
vmulps %zmm13, %zmm13, %zmm12
vfnmadd231ps __sPI2_FMA(%rax), %zmm10, %zmm8
vfmadd231ps __sA9_FMA(%rax), %zmm12, %zmm15
vfnmadd213ps %zmm8, %zmm7, %zmm10
vfmadd213ps __sA5_FMA(%rax), %zmm12, %zmm15
vpxord %zmm11, %zmm10, %zmm5
vmulps %zmm5, %zmm5, %zmm4
vfmadd213ps __sA3(%rax), %zmm12, %zmm15
vfmadd213ps %zmm14, %zmm4, %zmm3
vmulps %zmm12, %zmm15, %zmm14
vfmadd213ps __sA5_FMA(%rax), %zmm4, %zmm3
vfmadd213ps %zmm13, %zmm13, %zmm14
vfmadd213ps __sA3(%rax), %zmm4, %zmm3
vpxord %zmm0, %zmm14, %zmm0
vmulps %zmm4, %zmm3, %zmm3
vfmadd213ps %zmm5, %zmm5, %zmm3
testl %ecx, %ecx
jne .LBL_1_3
.LBL_1_2:
cfi_remember_state
vmovups %zmm0, (%rdi)
vmovups %zmm3, (%rsi)
movq %rbp, %rsp
cfi_def_cfa_register (%rsp)
popq %rbp
cfi_adjust_cfa_offset (-8)
cfi_restore (%rbp)
ret
.LBL_1_3:
cfi_restore_state
vmovups %zmm2, 1152(%rsp)
vmovups %zmm0, 1216(%rsp)
vmovups %zmm3, 1280(%rsp)
je .LBL_1_2
xorb %dl, %dl
kmovw %k4, 1048(%rsp)
xorl %eax, %eax
kmovw %k5, 1040(%rsp)
kmovw %k6, 1032(%rsp)
kmovw %k7, 1024(%rsp)
vmovups %zmm16, 960(%rsp)
vmovups %zmm17, 896(%rsp)
vmovups %zmm18, 832(%rsp)
vmovups %zmm19, 768(%rsp)
vmovups %zmm20, 704(%rsp)
vmovups %zmm21, 640(%rsp)
vmovups %zmm22, 576(%rsp)
vmovups %zmm23, 512(%rsp)
vmovups %zmm24, 448(%rsp)
vmovups %zmm25, 384(%rsp)
vmovups %zmm26, 320(%rsp)
vmovups %zmm27, 256(%rsp)
vmovups %zmm28, 192(%rsp)
vmovups %zmm29, 128(%rsp)
vmovups %zmm30, 64(%rsp)
vmovups %zmm31, (%rsp)
movq %rsi, 1056(%rsp)
movq %r12, 1096(%rsp)
cfi_offset_rel_rsp (12, 1096)
movb %dl, %r12b
movq %r13, 1088(%rsp)
cfi_offset_rel_rsp (13, 1088)
movl %eax, %r13d
movq %r14, 1080(%rsp)
cfi_offset_rel_rsp (14, 1080)
movl %ecx, %r14d
movq %r15, 1072(%rsp)
cfi_offset_rel_rsp (15, 1072)
movq %rbx, 1064(%rsp)
movq %rdi, %rbx
cfi_remember_state
.LBL_1_6:
btl %r13d, %r14d
jc .LBL_1_13
.LBL_1_7:
lea 1(%r13), %esi
btl %esi, %r14d
jc .LBL_1_10
.LBL_1_8:
addb $1, %r12b
addl $2, %r13d
cmpb $16, %r12b
jb .LBL_1_6
movq %rbx, %rdi
kmovw 1048(%rsp), %k4
movq 1056(%rsp), %rsi
kmovw 1040(%rsp), %k5
movq 1096(%rsp), %r12
cfi_restore (%r12)
kmovw 1032(%rsp), %k6
movq 1088(%rsp), %r13
cfi_restore (%r13)
kmovw 1024(%rsp), %k7
vmovups 960(%rsp), %zmm16
vmovups 896(%rsp), %zmm17
vmovups 832(%rsp), %zmm18
vmovups 768(%rsp), %zmm19
vmovups 704(%rsp), %zmm20
vmovups 640(%rsp), %zmm21
vmovups 576(%rsp), %zmm22
vmovups 512(%rsp), %zmm23
vmovups 448(%rsp), %zmm24
vmovups 384(%rsp), %zmm25
vmovups 320(%rsp), %zmm26
vmovups 256(%rsp), %zmm27
vmovups 192(%rsp), %zmm28
vmovups 128(%rsp), %zmm29
vmovups 64(%rsp), %zmm30
vmovups (%rsp), %zmm31
movq 1080(%rsp), %r14
cfi_restore (%r14)
movq 1072(%rsp), %r15
cfi_restore (%r15)
movq 1064(%rsp), %rbx
vmovups 1216(%rsp), %zmm0
vmovups 1280(%rsp), %zmm3
jmp .LBL_1_2
.LBL_1_10:
cfi_restore_state
movzbl %r12b, %r15d
vmovss 1156(%rsp,%r15,8), %xmm0
call sinf@PLT
vmovss %xmm0, 1220(%rsp,%r15,8)
vmovss 1156(%rsp,%r15,8), %xmm0
call cosf@PLT
vmovss %xmm0, 1284(%rsp,%r15,8)
jmp .LBL_1_8
.LBL_1_13:
movzbl %r12b, %r15d
vmovss 1152(%rsp,%r15,8), %xmm0
call sinf@PLT
vmovss %xmm0, 1216(%rsp,%r15,8)
vmovss 1152(%rsp,%r15,8), %xmm0
call cosf@PLT
vmovss %xmm0, 1280(%rsp,%r15,8)
jmp .LBL_1_7
#endif
END (_ZGVeN16vvv_sincosf_knl)
ENTRY (_ZGVeN16vvv_sincosf_skx)
#ifndef HAVE_AVX512_ASM_SUPPORT
WRAPPER_IMPL_AVX512_fFF _ZGVdN8vvv_sincosf
#else
pushq %rbp
cfi_adjust_cfa_offset (8)
cfi_rel_offset (%rbp, 0)
movq %rsp, %rbp
cfi_def_cfa_register (%rbp)
andq $-64, %rsp
subq $1344, %rsp
movq __svml_s_trig_data@GOTPCREL(%rip), %rax
vmovaps %zmm0, %zmm4
vmovups __sAbsMask(%rax), %zmm3
vmovups __sInvPI(%rax), %zmm5
vmovups __sRShifter(%rax), %zmm6
vmovups __sPI1_FMA(%rax), %zmm9
vmovups __sPI2_FMA(%rax), %zmm10
vmovups __sSignMask(%rax), %zmm14
vmovups __sOneHalf(%rax), %zmm7
vmovups __sPI3_FMA(%rax), %zmm12
/* Absolute argument computation */
vandps %zmm3, %zmm4, %zmm2
/* c) Getting octant Y by 2/Pi multiplication
d) Add "Right Shifter" value */
vfmadd213ps %zmm6, %zmm2, %zmm5
vcmpps $18, __sRangeReductionVal(%rax), %zmm2, %k1
/* e) Treat obtained value as integer S for destination sign setting */
vpslld $31, %zmm5, %zmm0
/* g) Subtract "Right Shifter" (0x4B000000) value */
vsubps %zmm6, %zmm5, %zmm5
vmovups __sA3(%rax), %zmm6
/* h) Subtract Y*(PI/2) from X argument, where PI/2 divided to 3 parts:
X = X - Y*PI1 - Y*PI2 - Y*PI3 */
vmovaps %zmm2, %zmm11
vfnmadd231ps %zmm5, %zmm9, %zmm11
vfnmadd231ps %zmm5, %zmm10, %zmm11
vandps %zmm11, %zmm14, %zmm1
vxorps %zmm1, %zmm7, %zmm8
/* Result sign calculations */
vpternlogd $150, %zmm0, %zmm14, %zmm1
vmovups .L_2il0floatpacket.13(%rip), %zmm14
/* Add correction term 0.5 for cos() part */
vaddps %zmm8, %zmm5, %zmm15
vfnmadd213ps %zmm11, %zmm12, %zmm5
vandnps %zmm4, %zmm3, %zmm11
vmovups __sA7_FMA(%rax), %zmm3
vmovaps %zmm2, %zmm13
vfnmadd231ps %zmm15, %zmm9, %zmm13
vxorps %zmm0, %zmm5, %zmm9
vmovups __sA5_FMA(%rax), %zmm0
vfnmadd231ps %zmm15, %zmm10, %zmm13
vmulps %zmm9, %zmm9, %zmm8
vfnmadd213ps %zmm13, %zmm12, %zmm15
vmovups __sA9_FMA(%rax), %zmm12
vxorps %zmm1, %zmm15, %zmm1
vmulps %zmm1, %zmm1, %zmm13
/* 2) Polynomial (minimax for sin within [-Pi/4; +Pi/4] interval)
a) Calculate X^2 = X * X
b) Calculate 2 polynomials for sin and cos:
RS = X * ( A0 + X^2 * (A1 + x^2 * (A2 + x^2 * (A3))));
RC = B0 + X^2 * (B1 + x^2 * (B2 + x^2 * (B3 + x^2 * (B4)))) */
vmovaps %zmm12, %zmm7
vfmadd213ps %zmm3, %zmm8, %zmm7
vfmadd213ps %zmm3, %zmm13, %zmm12
vfmadd213ps %zmm0, %zmm8, %zmm7
vfmadd213ps %zmm0, %zmm13, %zmm12
vfmadd213ps %zmm6, %zmm8, %zmm7
vfmadd213ps %zmm6, %zmm13, %zmm12
vmulps %zmm8, %zmm7, %zmm10
vmulps %zmm13, %zmm12, %zmm3
vfmadd213ps %zmm9, %zmm9, %zmm10
vfmadd213ps %zmm1, %zmm1, %zmm3
vxorps %zmm11, %zmm10, %zmm0
vpandnd %zmm2, %zmm2, %zmm14{%k1}
vptestmd %zmm14, %zmm14, %k0
kmovw %k0, %ecx
testl %ecx, %ecx
jne .LBL_2_3
.LBL_2_2:
cfi_remember_state
vmovups %zmm0, (%rdi)
vmovups %zmm3, (%rsi)
movq %rbp, %rsp
cfi_def_cfa_register (%rsp)
popq %rbp
cfi_adjust_cfa_offset (-8)
cfi_restore (%rbp)
ret
.LBL_2_3:
cfi_restore_state
vmovups %zmm4, 1152(%rsp)
vmovups %zmm0, 1216(%rsp)
vmovups %zmm3, 1280(%rsp)
je .LBL_2_2
xorb %dl, %dl
xorl %eax, %eax
kmovw %k4, 1048(%rsp)
kmovw %k5, 1040(%rsp)
kmovw %k6, 1032(%rsp)
kmovw %k7, 1024(%rsp)
vmovups %zmm16, 960(%rsp)
vmovups %zmm17, 896(%rsp)
vmovups %zmm18, 832(%rsp)
vmovups %zmm19, 768(%rsp)
vmovups %zmm20, 704(%rsp)
vmovups %zmm21, 640(%rsp)
vmovups %zmm22, 576(%rsp)
vmovups %zmm23, 512(%rsp)
vmovups %zmm24, 448(%rsp)
vmovups %zmm25, 384(%rsp)
vmovups %zmm26, 320(%rsp)
vmovups %zmm27, 256(%rsp)
vmovups %zmm28, 192(%rsp)
vmovups %zmm29, 128(%rsp)
vmovups %zmm30, 64(%rsp)
vmovups %zmm31, (%rsp)
movq %rsi, 1056(%rsp)
movq %r12, 1096(%rsp)
cfi_offset_rel_rsp (12, 1096)
movb %dl, %r12b
movq %r13, 1088(%rsp)
cfi_offset_rel_rsp (13, 1088)
movl %eax, %r13d
movq %r14, 1080(%rsp)
cfi_offset_rel_rsp (14, 1080)
movl %ecx, %r14d
movq %r15, 1072(%rsp)
cfi_offset_rel_rsp (15, 1072)
movq %rbx, 1064(%rsp)
movq %rdi, %rbx
cfi_remember_state
.LBL_2_6:
btl %r13d, %r14d
jc .LBL_2_13
.LBL_2_7:
lea 1(%r13), %esi
btl %esi, %r14d
jc .LBL_2_10
.LBL_2_8:
incb %r12b
addl $2, %r13d
cmpb $16, %r12b
jb .LBL_2_6
kmovw 1048(%rsp), %k4
movq %rbx, %rdi
kmovw 1040(%rsp), %k5
kmovw 1032(%rsp), %k6
kmovw 1024(%rsp), %k7
vmovups 960(%rsp), %zmm16
vmovups 896(%rsp), %zmm17
vmovups 832(%rsp), %zmm18
vmovups 768(%rsp), %zmm19
vmovups 704(%rsp), %zmm20
vmovups 640(%rsp), %zmm21
vmovups 576(%rsp), %zmm22
vmovups 512(%rsp), %zmm23
vmovups 448(%rsp), %zmm24
vmovups 384(%rsp), %zmm25
vmovups 320(%rsp), %zmm26
vmovups 256(%rsp), %zmm27
vmovups 192(%rsp), %zmm28
vmovups 128(%rsp), %zmm29
vmovups 64(%rsp), %zmm30
vmovups (%rsp), %zmm31
vmovups 1216(%rsp), %zmm0
vmovups 1280(%rsp), %zmm3
movq 1056(%rsp), %rsi
movq 1096(%rsp), %r12
cfi_restore (%r12)
movq 1088(%rsp), %r13
cfi_restore (%r13)
movq 1080(%rsp), %r14
cfi_restore (%r14)
movq 1072(%rsp), %r15
cfi_restore (%r15)
movq 1064(%rsp), %rbx
jmp .LBL_2_2
.LBL_2_10:
cfi_restore_state
movzbl %r12b, %r15d
vmovss 1156(%rsp,%r15,8), %xmm0
vzeroupper
vmovss 1156(%rsp,%r15,8), %xmm0
call sinf@PLT
vmovss %xmm0, 1220(%rsp,%r15,8)
vmovss 1156(%rsp,%r15,8), %xmm0
call cosf@PLT
vmovss %xmm0, 1284(%rsp,%r15,8)
jmp .LBL_2_8
.LBL_2_13:
movzbl %r12b, %r15d
vmovss 1152(%rsp,%r15,8), %xmm0
vzeroupper
vmovss 1152(%rsp,%r15,8), %xmm0
call sinf@PLT
vmovss %xmm0, 1216(%rsp,%r15,8)
vmovss 1152(%rsp,%r15,8), %xmm0
call cosf@PLT
vmovss %xmm0, 1280(%rsp,%r15,8)
jmp .LBL_2_7
#endif
END (_ZGVeN16vvv_sincosf_skx)
.section .rodata, "a"
.L_2il0floatpacket.13:
.long 0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff
.type .L_2il0floatpacket.13,@object