diff options
Diffstat (limited to 'sysdeps/x86_64/fpu/multiarch/svml_d_exp2_core_sse4.S')
-rw-r--r-- | sysdeps/x86_64/fpu/multiarch/svml_d_exp2_core_sse4.S | 225 |
1 files changed, 225 insertions, 0 deletions
diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_exp2_core_sse4.S b/sysdeps/x86_64/fpu/multiarch/svml_d_exp2_core_sse4.S new file mode 100644 index 0000000000..1f5445924a --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_exp2_core_sse4.S @@ -0,0 +1,225 @@ +/* Function exp vectorized with SSE4. + 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 + <http://www.gnu.org/licenses/>. */ + +#include <sysdep.h> +#include "svml_d_exp_data.h" + + .text +ENTRY (_ZGVbN2v_exp_sse4) +/* + ALGORITHM DESCRIPTION: + + Argument representation: + N = rint(X*2^k/ln2) = 2^k*M+j + X = N*ln2/2^k + r = M*ln2 + ln2*(j/2^k) + r + then -ln2/2^(k+1) < r < ln2/2^(k+1) + Alternatively: + N = trunc(X*2^k/ln2) + then 0 < r < ln2/2^k + + Result calculation: + exp(X) = exp(M*ln2 + ln2*(j/2^k) + r) + = 2^M * 2^(j/2^k) * exp(r) + 2^M is calculated by bit manipulation + 2^(j/2^k) is stored in table + exp(r) is approximated by polynomial. + + The table lookup is skipped if k = 0. */ + + pushq %rbp + cfi_adjust_cfa_offset (8) + cfi_rel_offset (%rbp, 0) + movq %rsp, %rbp + cfi_def_cfa_register (%rbp) + andq $-64, %rsp + subq $320, %rsp + movaps %xmm0, %xmm3 + movq __svml_dexp_data@GOTPCREL(%rip), %r8 + +/* iAbsX = (int)(lX>>32), lX = *(longlong*)&X */ + pshufd $221, %xmm3, %xmm7 + movups __dbInvLn2(%r8), %xmm0 + +/* dK = X*dbInvLn2 */ + mulpd %xmm3, %xmm0 + movq __iAbsMask(%r8), %xmm5 + movq __iDomainRange(%r8), %xmm6 + +/* iAbsX = iAbsX&iAbsMask */ + pand %xmm5, %xmm7 + +/* iRangeMask = (iAbsX>iDomainRange) */ + pcmpgtd %xmm6, %xmm7 + +/* Mask = iRangeMask?1:0, set mask for overflow/underflow */ + movmskps %xmm7, %eax + +/* dN = rint(X*2^k/Ln2) */ + xorps %xmm7, %xmm7 + movups __dbLn2hi(%r8), %xmm5 + movups __dbLn2lo(%r8), %xmm6 + roundpd $0, %xmm0, %xmm7 + +/* dR = X - dN*dbLn2hi, dbLn2hi is 52-8-k hi bits of ln2/2^k */ + mulpd %xmm7, %xmm5 + +/* dR = dR - dN*dbLn2lo, dbLn2lo is 40..94 bits of lo part of ln2/2^k */ + mulpd %xmm6, %xmm7 + movups __dbShifter(%r8), %xmm4 + +/* dM = X*dbInvLn2+dbShifter */ + addpd %xmm0, %xmm4 + movaps %xmm3, %xmm0 + subpd %xmm5, %xmm0 + subpd %xmm7, %xmm0 + movups __dPC2(%r8), %xmm5 + +/* exp(r) = b0+r*(b0+r*(b1+r*b2)) */ + mulpd %xmm0, %xmm5 + addpd __dPC1(%r8), %xmm5 + mulpd %xmm0, %xmm5 + movups __dPC0(%r8), %xmm6 + addpd %xmm6, %xmm5 + mulpd %xmm5, %xmm0 + movdqu __lIndexMask(%r8), %xmm2 + +/* lIndex = (*(longlong*)&dM)&lIndexMask, lIndex is the lower K bits of lM */ + movdqa %xmm2, %xmm1 + +/* lM = (*(longlong*)&dM)&(~lIndexMask) */ + pandn %xmm4, %xmm2 + pand %xmm4, %xmm1 + +/* lM = lM<<(52-K), 2^M */ + psllq $42, %xmm2 + +/* table lookup for dT[j] = 2^(j/2^k) */ + movd %xmm1, %edx + pextrw $4, %xmm1, %ecx + addpd %xmm0, %xmm6 + shll $3, %edx + shll $3, %ecx + movq (%r8,%rdx), %xmm0 + andl $3, %eax + movhpd (%r8,%rcx), %xmm0 + +/* 2^(j/2^k) * exp(r) */ + mulpd %xmm6, %xmm0 + +/* multiply by 2^M through integer add */ + paddq %xmm2, %xmm0 + jne .LBL_1_3 + +.LBL_1_2: + cfi_remember_state + 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 + movups %xmm3, 192(%rsp) + movups %xmm0, 256(%rsp) + je .LBL_1_2 + + xorb %cl, %cl + xorl %edx, %edx + movups %xmm8, 112(%rsp) + movups %xmm9, 96(%rsp) + movups %xmm10, 80(%rsp) + movups %xmm11, 64(%rsp) + movups %xmm12, 48(%rsp) + movups %xmm13, 32(%rsp) + movups %xmm14, 16(%rsp) + movups %xmm15, (%rsp) + movq %rsi, 136(%rsp) + movq %rdi, 128(%rsp) + movq %r12, 168(%rsp) + cfi_offset_rel_rsp (12, 168) + movb %cl, %r12b + movq %r13, 160(%rsp) + cfi_offset_rel_rsp (13, 160) + movl %eax, %r13d + movq %r14, 152(%rsp) + cfi_offset_rel_rsp (14, 152) + movl %edx, %r14d + movq %r15, 144(%rsp) + cfi_offset_rel_rsp (15, 144) + cfi_remember_state + +.LBL_1_6: + btl %r14d, %r13d + jc .LBL_1_12 + +.LBL_1_7: + lea 1(%r14), %esi + btl %esi, %r13d + jc .LBL_1_10 + +.LBL_1_8: + incb %r12b + addl $2, %r14d + cmpb $16, %r12b + jb .LBL_1_6 + + movups 112(%rsp), %xmm8 + movups 96(%rsp), %xmm9 + movups 80(%rsp), %xmm10 + movups 64(%rsp), %xmm11 + movups 48(%rsp), %xmm12 + movups 32(%rsp), %xmm13 + movups 16(%rsp), %xmm14 + movups (%rsp), %xmm15 + movq 136(%rsp), %rsi + movq 128(%rsp), %rdi + movq 168(%rsp), %r12 + cfi_restore (%r12) + movq 160(%rsp), %r13 + cfi_restore (%r13) + movq 152(%rsp), %r14 + cfi_restore (%r14) + movq 144(%rsp), %r15 + cfi_restore (%r15) + movups 256(%rsp), %xmm0 + jmp .LBL_1_2 + +.LBL_1_10: + cfi_restore_state + movzbl %r12b, %r15d + shlq $4, %r15 + movsd 200(%rsp,%r15), %xmm0 + + call exp@PLT + + movsd %xmm0, 264(%rsp,%r15) + jmp .LBL_1_8 + +.LBL_1_12: + movzbl %r12b, %r15d + shlq $4, %r15 + movsd 192(%rsp,%r15), %xmm0 + + call exp@PLT + + movsd %xmm0, 256(%rsp,%r15) + jmp .LBL_1_7 + +END (_ZGVbN2v_exp_sse4) |