mirror of
https://sourceware.org/git/glibc.git
synced 2025-01-13 20:50:08 +00:00
x86_64: Fix svml_d_exp24_core_avx2.S code formatting
This commit contains following formatting changes 1. Instructions proceeded by a tab. 2. Instruction less than 8 characters in length have a tab between it and the first operand. 3. Instruction greater than 7 characters in length have a space between it and the first operand. 4. Tabs after `#define`d names and their value. 5. 8 space at the beginning of line replaced by tab. 6. Indent comments with code. 7. Remove redundent .text section. 8. 1 space between line content and line comment. 9. Space after all commas. Reviewed-by: Noah Goldstein <goldstein.w.n@gmail.com>
This commit is contained in:
Sunil K Pandey
parent
0ea799391e
commit
2b7494c4f8
@ -22,7 +22,7 @@
|
||||
* exp2(x) = 2^n * T[j] * (1 + P(y))
|
||||
* where
|
||||
* x = m*(1/K) + y, y in [-1/K..1/K]
|
||||
* m = n*K + j, m,n,j - signed integer, j in [-K/2..K/2]
|
||||
* m = n*K + j, m, n,j - signed integer, j in [-K/2..K/2]
|
||||
*
|
||||
* values of 2^j/K are tabulated
|
||||
*
|
||||
@ -43,299 +43,297 @@
|
||||
|
||||
/* Offsets for data table __svml_dexp2_data_internal
|
||||
*/
|
||||
#define _dbT 0
|
||||
#define _dbShifter 1024
|
||||
#define _dPC1 1056
|
||||
#define _dPC2 1088
|
||||
#define _dPC3 1120
|
||||
#define _dPC4 1152
|
||||
#define _lIndexMask 1184
|
||||
#define _iAbsMask 1216
|
||||
#define _iDomainRange 1248
|
||||
#define _dbT 0
|
||||
#define _dbShifter 1024
|
||||
#define _dPC1 1056
|
||||
#define _dPC2 1088
|
||||
#define _dPC3 1120
|
||||
#define _dPC4 1152
|
||||
#define _lIndexMask 1184
|
||||
#define _iAbsMask 1216
|
||||
#define _iDomainRange 1248
|
||||
|
||||
#include <sysdep.h>
|
||||
|
||||
.text
|
||||
.section .text.avx2,"ax",@progbits
|
||||
.section .text.avx2, "ax", @progbits
|
||||
ENTRY(_ZGVdN4v_exp2_avx2)
|
||||
pushq %rbp
|
||||
cfi_def_cfa_offset(16)
|
||||
movq %rsp, %rbp
|
||||
cfi_def_cfa(6, 16)
|
||||
cfi_offset(6, -16)
|
||||
andq $-32, %rsp
|
||||
subq $96, %rsp
|
||||
pushq %rbp
|
||||
cfi_def_cfa_offset(16)
|
||||
movq %rsp, %rbp
|
||||
cfi_def_cfa(6, 16)
|
||||
cfi_offset(6, -16)
|
||||
andq $-32, %rsp
|
||||
subq $96, %rsp
|
||||
|
||||
/* out, basePtr, iIndex, iBaseOfs, iSize, iGran, iOfs */
|
||||
lea __svml_dexp2_data_internal(%rip), %r8
|
||||
vmovupd _dbShifter+__svml_dexp2_data_internal(%rip), %ymm4
|
||||
vmovupd _lIndexMask+__svml_dexp2_data_internal(%rip), %ymm3
|
||||
vmovapd %ymm0, %ymm1
|
||||
/* out, basePtr, iIndex, iBaseOfs, iSize, iGran, iOfs */
|
||||
lea __svml_dexp2_data_internal(%rip), %r8
|
||||
vmovupd _dbShifter+__svml_dexp2_data_internal(%rip), %ymm4
|
||||
vmovupd _lIndexMask+__svml_dexp2_data_internal(%rip), %ymm3
|
||||
vmovapd %ymm0, %ymm1
|
||||
|
||||
/* Load arument */
|
||||
vaddpd %ymm4, %ymm1, %ymm2
|
||||
vsubpd %ymm4, %ymm2, %ymm0
|
||||
/* Load arument */
|
||||
vaddpd %ymm4, %ymm1, %ymm2
|
||||
vsubpd %ymm4, %ymm2, %ymm0
|
||||
|
||||
/* Index and lookup */
|
||||
vandps %ymm3, %ymm2, %ymm9
|
||||
vpandn %ymm2, %ymm3, %ymm2
|
||||
/* Index and lookup */
|
||||
vandps %ymm3, %ymm2, %ymm9
|
||||
vpandn %ymm2, %ymm3, %ymm2
|
||||
|
||||
/* 2^N */
|
||||
vpsllq $45, %ymm2, %ymm3
|
||||
/* 2^N */
|
||||
vpsllq $45, %ymm2, %ymm3
|
||||
|
||||
/* R */
|
||||
vsubpd %ymm0, %ymm1, %ymm15
|
||||
/* R */
|
||||
vsubpd %ymm0, %ymm1, %ymm15
|
||||
|
||||
/* Check for overflow\underflow */
|
||||
vextractf128 $1, %ymm1, %xmm5
|
||||
/* Check for overflow\underflow */
|
||||
vextractf128 $1, %ymm1, %xmm5
|
||||
|
||||
/*
|
||||
* Polynomial
|
||||
* poly(dN) = a1*dR+...+a4*dR^4
|
||||
*/
|
||||
vmovupd _dPC4+__svml_dexp2_data_internal(%rip), %ymm0
|
||||
vshufps $221, %xmm5, %xmm1, %xmm6
|
||||
vandps _iAbsMask+__svml_dexp2_data_internal(%rip), %xmm6, %xmm7
|
||||
vpcmpgtd _iDomainRange+__svml_dexp2_data_internal(%rip), %xmm7, %xmm8
|
||||
vfmadd213pd _dPC3+__svml_dexp2_data_internal(%rip), %ymm15, %ymm0
|
||||
vmovmskps %xmm8, %eax
|
||||
vfmadd213pd _dPC2+__svml_dexp2_data_internal(%rip), %ymm15, %ymm0
|
||||
/*
|
||||
* Polynomial
|
||||
* poly(dN) = a1*dR+...+a4*dR^4
|
||||
*/
|
||||
vmovupd _dPC4+__svml_dexp2_data_internal(%rip), %ymm0
|
||||
vshufps $221, %xmm5, %xmm1, %xmm6
|
||||
vandps _iAbsMask+__svml_dexp2_data_internal(%rip), %xmm6, %xmm7
|
||||
vpcmpgtd _iDomainRange+__svml_dexp2_data_internal(%rip), %xmm7, %xmm8
|
||||
vfmadd213pd _dPC3+__svml_dexp2_data_internal(%rip), %ymm15, %ymm0
|
||||
vmovmskps %xmm8, %eax
|
||||
vfmadd213pd _dPC2+__svml_dexp2_data_internal(%rip), %ymm15, %ymm0
|
||||
|
||||
/* a1+...+a4*dR^3 ! */
|
||||
vfmadd213pd _dPC1+__svml_dexp2_data_internal(%rip), %ymm15, %ymm0
|
||||
vextractf128 $1, %ymm9, %xmm12
|
||||
vmovd %xmm9, %edx
|
||||
vmovd %xmm12, %esi
|
||||
shll $3, %edx
|
||||
vpextrd $2, %xmm9, %ecx
|
||||
shll $3, %esi
|
||||
vpextrd $2, %xmm12, %edi
|
||||
shll $3, %ecx
|
||||
vmovq (%r8,%rdx), %xmm10
|
||||
shll $3, %edi
|
||||
vmovq (%r8,%rsi), %xmm13
|
||||
vmovhpd (%r8,%rcx), %xmm10, %xmm11
|
||||
vmovhpd (%r8,%rdi), %xmm13, %xmm14
|
||||
vinsertf128 $1, %xmm14, %ymm11, %ymm4
|
||||
/* a1+...+a4*dR^3 ! */
|
||||
vfmadd213pd _dPC1+__svml_dexp2_data_internal(%rip), %ymm15, %ymm0
|
||||
vextractf128 $1, %ymm9, %xmm12
|
||||
vmovd %xmm9, %edx
|
||||
vmovd %xmm12, %esi
|
||||
shll $3, %edx
|
||||
vpextrd $2, %xmm9, %ecx
|
||||
shll $3, %esi
|
||||
vpextrd $2, %xmm12, %edi
|
||||
shll $3, %ecx
|
||||
vmovq (%r8, %rdx), %xmm10
|
||||
shll $3, %edi
|
||||
vmovq (%r8, %rsi), %xmm13
|
||||
vmovhpd (%r8, %rcx), %xmm10, %xmm11
|
||||
vmovhpd (%r8, %rdi), %xmm13, %xmm14
|
||||
vinsertf128 $1, %xmm14, %ymm11, %ymm4
|
||||
|
||||
/* dR=dR*dT */
|
||||
vmulpd %ymm15, %ymm4, %ymm15
|
||||
/* dR=dR*dT */
|
||||
vmulpd %ymm15, %ymm4, %ymm15
|
||||
|
||||
/*
|
||||
* Reconstruction
|
||||
* exp2 = {2^N later}*(Tj+Tj*poly)
|
||||
* dN = dT+dT*dR*(a1+...+a4*dR^3)
|
||||
*/
|
||||
vfmadd213pd %ymm4, %ymm15, %ymm0
|
||||
/*
|
||||
* Reconstruction
|
||||
* exp2 = {2^N later}*(Tj+Tj*poly)
|
||||
* dN = dT+dT*dR*(a1+...+a4*dR^3)
|
||||
*/
|
||||
vfmadd213pd %ymm4, %ymm15, %ymm0
|
||||
|
||||
/* quick 2^N */
|
||||
vpaddq %ymm3, %ymm0, %ymm0
|
||||
/* quick 2^N */
|
||||
vpaddq %ymm3, %ymm0, %ymm0
|
||||
|
||||
/* Finish */
|
||||
testl %eax, %eax
|
||||
/* Finish */
|
||||
testl %eax, %eax
|
||||
|
||||
/* Go to special inputs processing branch */
|
||||
jne L(SPECIAL_VALUES_BRANCH)
|
||||
# LOE rbx r12 r13 r14 r15 eax ymm0 ymm1
|
||||
/* Go to special inputs processing branch */
|
||||
jne L(SPECIAL_VALUES_BRANCH)
|
||||
# LOE rbx r12 r13 r14 r15 eax ymm0 ymm1
|
||||
|
||||
/* Restore registers
|
||||
* and exit the function
|
||||
*/
|
||||
/* Restore registers
|
||||
* and exit the function
|
||||
*/
|
||||
|
||||
L(EXIT):
|
||||
movq %rbp, %rsp
|
||||
popq %rbp
|
||||
cfi_def_cfa(7, 8)
|
||||
cfi_restore(6)
|
||||
ret
|
||||
cfi_def_cfa(6, 16)
|
||||
cfi_offset(6, -16)
|
||||
movq %rbp, %rsp
|
||||
popq %rbp
|
||||
cfi_def_cfa(7, 8)
|
||||
cfi_restore(6)
|
||||
ret
|
||||
cfi_def_cfa(6, 16)
|
||||
cfi_offset(6, -16)
|
||||
|
||||
/* Branch to process
|
||||
* special inputs
|
||||
*/
|
||||
/* Branch to process
|
||||
* special inputs
|
||||
*/
|
||||
|
||||
L(SPECIAL_VALUES_BRANCH):
|
||||
vmovupd %ymm1, 32(%rsp)
|
||||
vmovupd %ymm0, 64(%rsp)
|
||||
# LOE rbx r12 r13 r14 r15 eax ymm0
|
||||
vmovupd %ymm1, 32(%rsp)
|
||||
vmovupd %ymm0, 64(%rsp)
|
||||
# LOE rbx r12 r13 r14 r15 eax ymm0
|
||||
|
||||
xorl %edx, %edx
|
||||
# LOE rbx r12 r13 r14 r15 eax edx
|
||||
xorl %edx, %edx
|
||||
# LOE rbx r12 r13 r14 r15 eax edx
|
||||
|
||||
vzeroupper
|
||||
movq %r12, 16(%rsp)
|
||||
/* DW_CFA_expression: r12 (r12) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -80; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xb0, 0xff, 0xff, 0xff, 0x22
|
||||
movl %edx, %r12d
|
||||
movq %r13, 8(%rsp)
|
||||
/* DW_CFA_expression: r13 (r13) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -88; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xa8, 0xff, 0xff, 0xff, 0x22
|
||||
movl %eax, %r13d
|
||||
movq %r14, (%rsp)
|
||||
/* DW_CFA_expression: r14 (r14) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -96; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xa0, 0xff, 0xff, 0xff, 0x22
|
||||
# LOE rbx r15 r12d r13d
|
||||
vzeroupper
|
||||
movq %r12, 16(%rsp)
|
||||
/* DW_CFA_expression: r12 (r12) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -80; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xb0, 0xff, 0xff, 0xff, 0x22
|
||||
movl %edx, %r12d
|
||||
movq %r13, 8(%rsp)
|
||||
/* DW_CFA_expression: r13 (r13) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -88; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xa8, 0xff, 0xff, 0xff, 0x22
|
||||
movl %eax, %r13d
|
||||
movq %r14, (%rsp)
|
||||
/* DW_CFA_expression: r14 (r14) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -96; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xa0, 0xff, 0xff, 0xff, 0x22
|
||||
# LOE rbx r15 r12d r13d
|
||||
|
||||
/* Range mask
|
||||
* bits check
|
||||
*/
|
||||
/* Range mask
|
||||
* bits check
|
||||
*/
|
||||
|
||||
L(RANGEMASK_CHECK):
|
||||
btl %r12d, %r13d
|
||||
btl %r12d, %r13d
|
||||
|
||||
/* Call scalar math function */
|
||||
jc L(SCALAR_MATH_CALL)
|
||||
# LOE rbx r15 r12d r13d
|
||||
/* Call scalar math function */
|
||||
jc L(SCALAR_MATH_CALL)
|
||||
# LOE rbx r15 r12d r13d
|
||||
|
||||
/* Special inputs
|
||||
* processing loop
|
||||
*/
|
||||
/* Special inputs
|
||||
* processing loop
|
||||
*/
|
||||
|
||||
L(SPECIAL_VALUES_LOOP):
|
||||
incl %r12d
|
||||
cmpl $4, %r12d
|
||||
incl %r12d
|
||||
cmpl $4, %r12d
|
||||
|
||||
/* Check bits in range mask */
|
||||
jl L(RANGEMASK_CHECK)
|
||||
# LOE rbx r15 r12d r13d
|
||||
/* Check bits in range mask */
|
||||
jl L(RANGEMASK_CHECK)
|
||||
# LOE rbx r15 r12d r13d
|
||||
|
||||
movq 16(%rsp), %r12
|
||||
cfi_restore(12)
|
||||
movq 8(%rsp), %r13
|
||||
cfi_restore(13)
|
||||
movq (%rsp), %r14
|
||||
cfi_restore(14)
|
||||
vmovupd 64(%rsp), %ymm0
|
||||
movq 16(%rsp), %r12
|
||||
cfi_restore(12)
|
||||
movq 8(%rsp), %r13
|
||||
cfi_restore(13)
|
||||
movq (%rsp), %r14
|
||||
cfi_restore(14)
|
||||
vmovupd 64(%rsp), %ymm0
|
||||
|
||||
/* Go to exit */
|
||||
jmp L(EXIT)
|
||||
/* DW_CFA_expression: r12 (r12) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -80; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xb0, 0xff, 0xff, 0xff, 0x22
|
||||
/* DW_CFA_expression: r13 (r13) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -88; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xa8, 0xff, 0xff, 0xff, 0x22
|
||||
/* DW_CFA_expression: r14 (r14) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -96; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xa0, 0xff, 0xff, 0xff, 0x22
|
||||
# LOE rbx r12 r13 r14 r15 ymm0
|
||||
/* Go to exit */
|
||||
jmp L(EXIT)
|
||||
/* DW_CFA_expression: r12 (r12) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -80; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xb0, 0xff, 0xff, 0xff, 0x22
|
||||
/* DW_CFA_expression: r13 (r13) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -88; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xa8, 0xff, 0xff, 0xff, 0x22
|
||||
/* DW_CFA_expression: r14 (r14) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -96; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xa0, 0xff, 0xff, 0xff, 0x22
|
||||
# LOE rbx r12 r13 r14 r15 ymm0
|
||||
|
||||
/* Scalar math fucntion call
|
||||
* to process special input
|
||||
*/
|
||||
/* Scalar math fucntion call
|
||||
* to process special input
|
||||
*/
|
||||
|
||||
L(SCALAR_MATH_CALL):
|
||||
movl %r12d, %r14d
|
||||
movsd 32(%rsp,%r14,8), %xmm0
|
||||
call exp2@PLT
|
||||
# LOE rbx r14 r15 r12d r13d xmm0
|
||||
movl %r12d, %r14d
|
||||
movsd 32(%rsp, %r14, 8), %xmm0
|
||||
call exp2@PLT
|
||||
# LOE rbx r14 r15 r12d r13d xmm0
|
||||
|
||||
movsd %xmm0, 64(%rsp,%r14,8)
|
||||
movsd %xmm0, 64(%rsp, %r14, 8)
|
||||
|
||||
/* Process special inputs in loop */
|
||||
jmp L(SPECIAL_VALUES_LOOP)
|
||||
# LOE rbx r15 r12d r13d
|
||||
/* Process special inputs in loop */
|
||||
jmp L(SPECIAL_VALUES_LOOP)
|
||||
# LOE rbx r15 r12d r13d
|
||||
END(_ZGVdN4v_exp2_avx2)
|
||||
|
||||
.section .rodata, "a"
|
||||
.align 32
|
||||
.section .rodata, "a"
|
||||
.align 32
|
||||
|
||||
#ifdef __svml_dexp2_data_internal_typedef
|
||||
typedef unsigned int VUINT32;
|
||||
typedef struct
|
||||
{
|
||||
__declspec(align(32)) VUINT32 _dbT[(1<<7)][2];
|
||||
__declspec(align(32)) VUINT32 _dbShifter[4][2];
|
||||
__declspec(align(32)) VUINT32 _dPC1[4][2];
|
||||
__declspec(align(32)) VUINT32 _dPC2[4][2];
|
||||
__declspec(align(32)) VUINT32 _dPC3[4][2];
|
||||
__declspec(align(32)) VUINT32 _dPC4[4][2];
|
||||
__declspec(align(32)) VUINT32 _lIndexMask[4][2];
|
||||
__declspec(align(32)) VUINT32 _iAbsMask[8][1];
|
||||
__declspec(align(32)) VUINT32 _iDomainRange[8][1];
|
||||
typedef struct {
|
||||
__declspec(align(32)) VUINT32 _dbT[(1<<7)][2];
|
||||
__declspec(align(32)) VUINT32 _dbShifter[4][2];
|
||||
__declspec(align(32)) VUINT32 _dPC1[4][2];
|
||||
__declspec(align(32)) VUINT32 _dPC2[4][2];
|
||||
__declspec(align(32)) VUINT32 _dPC3[4][2];
|
||||
__declspec(align(32)) VUINT32 _dPC4[4][2];
|
||||
__declspec(align(32)) VUINT32 _lIndexMask[4][2];
|
||||
__declspec(align(32)) VUINT32 _iAbsMask[8][1];
|
||||
__declspec(align(32)) VUINT32 _iDomainRange[8][1];
|
||||
} __svml_dexp2_data_internal;
|
||||
#endif
|
||||
__svml_dexp2_data_internal:
|
||||
/*== _dbT ==*/
|
||||
.quad 0x3ff0000000000000, 0x3ff0163da9fb3335 /*2^( 0 /128),2^( 1 /128)*/
|
||||
.quad 0x3ff02c9a3e778061, 0x3ff04315e86e7f85 /*2^( 2 /128),2^( 3 /128)*/
|
||||
.quad 0x3ff059b0d3158574, 0x3ff0706b29ddf6de /*2^( 4 /128),2^( 5 /128)*/
|
||||
.quad 0x3ff0874518759bc8, 0x3ff09e3ecac6f383 /*2^( 6 /128),2^( 7 /128)*/
|
||||
.quad 0x3ff0b5586cf9890f, 0x3ff0cc922b7247f7 /*2^( 8 /128),2^( 9 /128)*/
|
||||
.quad 0x3ff0e3ec32d3d1a2, 0x3ff0fb66affed31b /*2^( 10 /128),2^( 11 /128)*/
|
||||
.quad 0x3ff11301d0125b51, 0x3ff12abdc06c31cc /*2^( 12 /128),2^( 13 /128)*/
|
||||
.quad 0x3ff1429aaea92de0, 0x3ff15a98c8a58e51 /*2^( 14 /128),2^( 15 /128)*/
|
||||
.quad 0x3ff172b83c7d517b, 0x3ff18af9388c8dea /*2^( 16 /128),2^( 17 /128)*/
|
||||
.quad 0x3ff1a35beb6fcb75, 0x3ff1bbe084045cd4 /*2^( 18 /128),2^( 19 /128)*/
|
||||
.quad 0x3ff1d4873168b9aa, 0x3ff1ed5022fcd91d /*2^( 20 /128),2^( 21 /128)*/
|
||||
.quad 0x3ff2063b88628cd6, 0x3ff21f49917ddc96 /*2^( 22 /128),2^( 23 /128)*/
|
||||
.quad 0x3ff2387a6e756238, 0x3ff251ce4fb2a63f /*2^( 24 /128),2^( 25 /128)*/
|
||||
.quad 0x3ff26b4565e27cdd, 0x3ff284dfe1f56381 /*2^( 26 /128),2^( 27 /128)*/
|
||||
.quad 0x3ff29e9df51fdee1, 0x3ff2b87fd0dad990 /*2^( 28 /128),2^( 29 /128)*/
|
||||
.quad 0x3ff2d285a6e4030b, 0x3ff2ecafa93e2f56 /*2^( 30 /128),2^( 31 /128)*/
|
||||
.quad 0x3ff306fe0a31b715, 0x3ff32170fc4cd831 /*2^( 32 /128),2^( 33 /128)*/
|
||||
.quad 0x3ff33c08b26416ff, 0x3ff356c55f929ff1 /*2^( 34 /128),2^( 35 /128)*/
|
||||
.quad 0x3ff371a7373aa9cb, 0x3ff38cae6d05d866 /*2^( 36 /128),2^( 37 /128)*/
|
||||
.quad 0x3ff3a7db34e59ff7, 0x3ff3c32dc313a8e5 /*2^( 38 /128),2^( 39 /128)*/
|
||||
.quad 0x3ff3dea64c123422, 0x3ff3fa4504ac801c /*2^( 40 /128),2^( 41 /128)*/
|
||||
.quad 0x3ff4160a21f72e2a, 0x3ff431f5d950a897 /*2^( 42 /128),2^( 43 /128)*/
|
||||
.quad 0x3ff44e086061892d, 0x3ff46a41ed1d0057 /*2^( 44 /128),2^( 45 /128)*/
|
||||
.quad 0x3ff486a2b5c13cd0, 0x3ff4a32af0d7d3de /*2^( 46 /128),2^( 47 /128)*/
|
||||
.quad 0x3ff4bfdad5362a27, 0x3ff4dcb299fddd0d /*2^( 48 /128),2^( 49 /128)*/
|
||||
.quad 0x3ff4f9b2769d2ca7, 0x3ff516daa2cf6642 /*2^( 50 /128),2^( 51 /128)*/
|
||||
.quad 0x3ff5342b569d4f82, 0x3ff551a4ca5d920f /*2^( 52 /128),2^( 53 /128)*/
|
||||
.quad 0x3ff56f4736b527da, 0x3ff58d12d497c7fd /*2^( 54 /128),2^( 55 /128)*/
|
||||
.quad 0x3ff5ab07dd485429, 0x3ff5c9268a5946b7 /*2^( 56 /128),2^( 57 /128)*/
|
||||
.quad 0x3ff5e76f15ad2148, 0x3ff605e1b976dc09 /*2^( 58 /128),2^( 59 /128)*/
|
||||
.quad 0x3ff6247eb03a5585, 0x3ff6434634ccc320 /*2^( 60 /128),2^( 61 /128)*/
|
||||
.quad 0x3ff6623882552225, 0x3ff68155d44ca973 /*2^( 62 /128),2^( 63 /128)*/
|
||||
.quad 0x3ff6a09e667f3bcd, 0x3ff6c012750bdabf /*2^( 64 /128),2^( 65 /128)*/
|
||||
.quad 0x3ff6dfb23c651a2f, 0x3ff6ff7df9519484 /*2^( 66 /128),2^( 67 /128)*/
|
||||
.quad 0x3ff71f75e8ec5f74, 0x3ff73f9a48a58174 /*2^( 68 /128),2^( 69 /128)*/
|
||||
.quad 0x3ff75feb564267c9, 0x3ff780694fde5d3f /*2^( 70 /128),2^( 71 /128)*/
|
||||
.quad 0x3ff7a11473eb0187, 0x3ff7c1ed0130c132 /*2^( 72 /128),2^( 73 /128)*/
|
||||
.quad 0x3ff7e2f336cf4e62, 0x3ff80427543e1a12 /*2^( 74 /128),2^( 75 /128)*/
|
||||
.quad 0x3ff82589994cce13, 0x3ff8471a4623c7ad /*2^( 76 /128),2^( 77 /128)*/
|
||||
.quad 0x3ff868d99b4492ed, 0x3ff88ac7d98a6699 /*2^( 78 /128),2^( 79 /128)*/
|
||||
.quad 0x3ff8ace5422aa0db, 0x3ff8cf3216b5448c /*2^( 80 /128),2^( 81 /128)*/
|
||||
.quad 0x3ff8f1ae99157736, 0x3ff9145b0b91ffc6 /*2^( 82 /128),2^( 83 /128)*/
|
||||
.quad 0x3ff93737b0cdc5e5, 0x3ff95a44cbc8520f /*2^( 84 /128),2^( 85 /128)*/
|
||||
.quad 0x3ff97d829fde4e50, 0x3ff9a0f170ca07ba /*2^( 86 /128),2^( 87 /128)*/
|
||||
.quad 0x3ff9c49182a3f090, 0x3ff9e86319e32323 /*2^( 88 /128),2^( 89 /128)*/
|
||||
.quad 0x3ffa0c667b5de565, 0x3ffa309bec4a2d33 /*2^( 90 /128),2^( 91 /128)*/
|
||||
.quad 0x3ffa5503b23e255d, 0x3ffa799e1330b358 /*2^( 92 /128),2^( 93 /128)*/
|
||||
.quad 0x3ffa9e6b5579fdbf, 0x3ffac36bbfd3f37a /*2^( 94 /128),2^( 95 /128)*/
|
||||
.quad 0x3ffae89f995ad3ad, 0x3ffb0e07298db666 /*2^( 96 /128),2^( 97 /128)*/
|
||||
.quad 0x3ffb33a2b84f15fb, 0x3ffb59728de5593a /*2^( 98 /128),2^( 99 /128)*/
|
||||
.quad 0x3ffb7f76f2fb5e47, 0x3ffba5b030a1064a /*2^( 100 /128),2^( 101 /128)*/
|
||||
.quad 0x3ffbcc1e904bc1d2, 0x3ffbf2c25bd71e09 /*2^( 102 /128),2^( 103 /128)*/
|
||||
.quad 0x3ffc199bdd85529c, 0x3ffc40ab5fffd07a /*2^( 104 /128),2^( 105 /128)*/
|
||||
.quad 0x3ffc67f12e57d14b, 0x3ffc8f6d9406e7b5 /*2^( 106 /128),2^( 107 /128)*/
|
||||
.quad 0x3ffcb720dcef9069, 0x3ffcdf0b555dc3fa /*2^( 108 /128),2^( 109 /128)*/
|
||||
.quad 0x3ffd072d4a07897c, 0x3ffd2f87080d89f2 /*2^( 110 /128),2^( 111 /128)*/
|
||||
.quad 0x3ffd5818dcfba487, 0x3ffd80e316c98398 /*2^( 112 /128),2^( 113 /128)*/
|
||||
.quad 0x3ffda9e603db3285, 0x3ffdd321f301b460 /*2^( 114 /128),2^( 115 /128)*/
|
||||
.quad 0x3ffdfc97337b9b5f, 0x3ffe264614f5a129 /*2^( 116 /128),2^( 117 /128)*/
|
||||
.quad 0x3ffe502ee78b3ff6, 0x3ffe7a51fbc74c83 /*2^( 118 /128),2^( 119 /128)*/
|
||||
.quad 0x3ffea4afa2a490da, 0x3ffecf482d8e67f1 /*2^( 120 /128),2^( 121 /128)*/
|
||||
.quad 0x3ffefa1bee615a27, 0x3fff252b376bba97 /*2^( 122 /128),2^( 123 /128)*/
|
||||
.quad 0x3fff50765b6e4540, 0x3fff7bfdad9cbe14 /*2^( 124 /128),2^( 125 /128)*/
|
||||
.quad 0x3fffa7c1819e90d8, 0x3fffd3c22b8f71f1 /*2^( 126 /128),2^( 127 /128)*/
|
||||
.align 32
|
||||
.quad 0x42c8000000000000, 0x42c8000000000000, 0x42c8000000000000, 0x42c8000000000000 /* _dbShifter - 0x433-7=0x42c shifted right on K!*/
|
||||
//log2(relerr) = -53.547756365162
|
||||
.align 32
|
||||
.quad 0x3fe62e42fefa3685, 0x3fe62e42fefa3685, 0x3fe62e42fefa3685, 0x3fe62e42fefa3685 /* _dPC1 */
|
||||
.align 32
|
||||
.quad 0x3fcebfbdff82ca48, 0x3fcebfbdff82ca48, 0x3fcebfbdff82ca48, 0x3fcebfbdff82ca48 /* _dPC2 */
|
||||
.align 32
|
||||
.quad 0x3fac6b09b180f045, 0x3fac6b09b180f045, 0x3fac6b09b180f045, 0x3fac6b09b180f045 /* _dPC3 */
|
||||
.align 32
|
||||
.quad 0x3f83b2ab5bb1268f, 0x3f83b2ab5bb1268f, 0x3f83b2ab5bb1268f, 0x3f83b2ab5bb1268f /* _dPC4 */
|
||||
.align 32
|
||||
.quad 0x000000000000007f, 0x000000000000007f, 0x000000000000007f, 0x000000000000007f /* _lIndexMask =(2^K-1)*/
|
||||
.align 32
|
||||
.long 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff /* _iAbsMask */
|
||||
.align 32
|
||||
.long 0x408fefff, 0x408fefff, 0x408fefff, 0x408fefff, 0x408fefff, 0x408fefff, 0x408fefff, 0x408fefff /* _iDomainRange */
|
||||
.align 32
|
||||
.type __svml_dexp2_data_internal,@object
|
||||
.size __svml_dexp2_data_internal,.-__svml_dexp2_data_internal
|
||||
/* _dbT */
|
||||
.quad 0x3ff0000000000000, 0x3ff0163da9fb3335 /* 2^( 0 /128), 2^( 1 /128) */
|
||||
.quad 0x3ff02c9a3e778061, 0x3ff04315e86e7f85 /* 2^( 2 /128), 2^( 3 /128) */
|
||||
.quad 0x3ff059b0d3158574, 0x3ff0706b29ddf6de /* 2^( 4 /128), 2^( 5 /128) */
|
||||
.quad 0x3ff0874518759bc8, 0x3ff09e3ecac6f383 /* 2^( 6 /128), 2^( 7 /128) */
|
||||
.quad 0x3ff0b5586cf9890f, 0x3ff0cc922b7247f7 /* 2^( 8 /128), 2^( 9 /128) */
|
||||
.quad 0x3ff0e3ec32d3d1a2, 0x3ff0fb66affed31b /* 2^( 10 /128), 2^( 11 /128) */
|
||||
.quad 0x3ff11301d0125b51, 0x3ff12abdc06c31cc /* 2^( 12 /128), 2^( 13 /128) */
|
||||
.quad 0x3ff1429aaea92de0, 0x3ff15a98c8a58e51 /* 2^( 14 /128), 2^( 15 /128) */
|
||||
.quad 0x3ff172b83c7d517b, 0x3ff18af9388c8dea /* 2^( 16 /128), 2^( 17 /128) */
|
||||
.quad 0x3ff1a35beb6fcb75, 0x3ff1bbe084045cd4 /* 2^( 18 /128), 2^( 19 /128) */
|
||||
.quad 0x3ff1d4873168b9aa, 0x3ff1ed5022fcd91d /* 2^( 20 /128), 2^( 21 /128) */
|
||||
.quad 0x3ff2063b88628cd6, 0x3ff21f49917ddc96 /* 2^( 22 /128), 2^( 23 /128) */
|
||||
.quad 0x3ff2387a6e756238, 0x3ff251ce4fb2a63f /* 2^( 24 /128), 2^( 25 /128) */
|
||||
.quad 0x3ff26b4565e27cdd, 0x3ff284dfe1f56381 /* 2^( 26 /128), 2^( 27 /128) */
|
||||
.quad 0x3ff29e9df51fdee1, 0x3ff2b87fd0dad990 /* 2^( 28 /128), 2^( 29 /128) */
|
||||
.quad 0x3ff2d285a6e4030b, 0x3ff2ecafa93e2f56 /* 2^( 30 /128), 2^( 31 /128) */
|
||||
.quad 0x3ff306fe0a31b715, 0x3ff32170fc4cd831 /* 2^( 32 /128), 2^( 33 /128) */
|
||||
.quad 0x3ff33c08b26416ff, 0x3ff356c55f929ff1 /* 2^( 34 /128), 2^( 35 /128) */
|
||||
.quad 0x3ff371a7373aa9cb, 0x3ff38cae6d05d866 /* 2^( 36 /128), 2^( 37 /128) */
|
||||
.quad 0x3ff3a7db34e59ff7, 0x3ff3c32dc313a8e5 /* 2^( 38 /128), 2^( 39 /128) */
|
||||
.quad 0x3ff3dea64c123422, 0x3ff3fa4504ac801c /* 2^( 40 /128), 2^( 41 /128) */
|
||||
.quad 0x3ff4160a21f72e2a, 0x3ff431f5d950a897 /* 2^( 42 /128), 2^( 43 /128) */
|
||||
.quad 0x3ff44e086061892d, 0x3ff46a41ed1d0057 /* 2^( 44 /128), 2^( 45 /128) */
|
||||
.quad 0x3ff486a2b5c13cd0, 0x3ff4a32af0d7d3de /* 2^( 46 /128), 2^( 47 /128) */
|
||||
.quad 0x3ff4bfdad5362a27, 0x3ff4dcb299fddd0d /* 2^( 48 /128), 2^( 49 /128) */
|
||||
.quad 0x3ff4f9b2769d2ca7, 0x3ff516daa2cf6642 /* 2^( 50 /128), 2^( 51 /128) */
|
||||
.quad 0x3ff5342b569d4f82, 0x3ff551a4ca5d920f /* 2^( 52 /128), 2^( 53 /128) */
|
||||
.quad 0x3ff56f4736b527da, 0x3ff58d12d497c7fd /* 2^( 54 /128), 2^( 55 /128) */
|
||||
.quad 0x3ff5ab07dd485429, 0x3ff5c9268a5946b7 /* 2^( 56 /128), 2^( 57 /128) */
|
||||
.quad 0x3ff5e76f15ad2148, 0x3ff605e1b976dc09 /* 2^( 58 /128), 2^( 59 /128) */
|
||||
.quad 0x3ff6247eb03a5585, 0x3ff6434634ccc320 /* 2^( 60 /128), 2^( 61 /128) */
|
||||
.quad 0x3ff6623882552225, 0x3ff68155d44ca973 /* 2^( 62 /128), 2^( 63 /128) */
|
||||
.quad 0x3ff6a09e667f3bcd, 0x3ff6c012750bdabf /* 2^( 64 /128), 2^( 65 /128) */
|
||||
.quad 0x3ff6dfb23c651a2f, 0x3ff6ff7df9519484 /* 2^( 66 /128), 2^( 67 /128) */
|
||||
.quad 0x3ff71f75e8ec5f74, 0x3ff73f9a48a58174 /* 2^( 68 /128), 2^( 69 /128) */
|
||||
.quad 0x3ff75feb564267c9, 0x3ff780694fde5d3f /* 2^( 70 /128), 2^( 71 /128) */
|
||||
.quad 0x3ff7a11473eb0187, 0x3ff7c1ed0130c132 /* 2^( 72 /128), 2^( 73 /128) */
|
||||
.quad 0x3ff7e2f336cf4e62, 0x3ff80427543e1a12 /* 2^( 74 /128), 2^( 75 /128) */
|
||||
.quad 0x3ff82589994cce13, 0x3ff8471a4623c7ad /* 2^( 76 /128), 2^( 77 /128) */
|
||||
.quad 0x3ff868d99b4492ed, 0x3ff88ac7d98a6699 /* 2^( 78 /128), 2^( 79 /128) */
|
||||
.quad 0x3ff8ace5422aa0db, 0x3ff8cf3216b5448c /* 2^( 80 /128), 2^( 81 /128) */
|
||||
.quad 0x3ff8f1ae99157736, 0x3ff9145b0b91ffc6 /* 2^( 82 /128), 2^( 83 /128) */
|
||||
.quad 0x3ff93737b0cdc5e5, 0x3ff95a44cbc8520f /* 2^( 84 /128), 2^( 85 /128) */
|
||||
.quad 0x3ff97d829fde4e50, 0x3ff9a0f170ca07ba /* 2^( 86 /128), 2^( 87 /128) */
|
||||
.quad 0x3ff9c49182a3f090, 0x3ff9e86319e32323 /* 2^( 88 /128), 2^( 89 /128) */
|
||||
.quad 0x3ffa0c667b5de565, 0x3ffa309bec4a2d33 /* 2^( 90 /128), 2^( 91 /128) */
|
||||
.quad 0x3ffa5503b23e255d, 0x3ffa799e1330b358 /* 2^( 92 /128), 2^( 93 /128) */
|
||||
.quad 0x3ffa9e6b5579fdbf, 0x3ffac36bbfd3f37a /* 2^( 94 /128), 2^( 95 /128) */
|
||||
.quad 0x3ffae89f995ad3ad, 0x3ffb0e07298db666 /* 2^( 96 /128), 2^( 97 /128) */
|
||||
.quad 0x3ffb33a2b84f15fb, 0x3ffb59728de5593a /* 2^( 98 /128), 2^( 99 /128) */
|
||||
.quad 0x3ffb7f76f2fb5e47, 0x3ffba5b030a1064a /* 2^( 100 /128), 2^( 101 /128) */
|
||||
.quad 0x3ffbcc1e904bc1d2, 0x3ffbf2c25bd71e09 /* 2^( 102 /128), 2^( 103 /128) */
|
||||
.quad 0x3ffc199bdd85529c, 0x3ffc40ab5fffd07a /* 2^( 104 /128), 2^( 105 /128) */
|
||||
.quad 0x3ffc67f12e57d14b, 0x3ffc8f6d9406e7b5 /* 2^( 106 /128), 2^( 107 /128) */
|
||||
.quad 0x3ffcb720dcef9069, 0x3ffcdf0b555dc3fa /* 2^( 108 /128), 2^( 109 /128) */
|
||||
.quad 0x3ffd072d4a07897c, 0x3ffd2f87080d89f2 /* 2^( 110 /128), 2^( 111 /128) */
|
||||
.quad 0x3ffd5818dcfba487, 0x3ffd80e316c98398 /* 2^( 112 /128), 2^( 113 /128) */
|
||||
.quad 0x3ffda9e603db3285, 0x3ffdd321f301b460 /* 2^( 114 /128), 2^( 115 /128) */
|
||||
.quad 0x3ffdfc97337b9b5f, 0x3ffe264614f5a129 /* 2^( 116 /128), 2^( 117 /128) */
|
||||
.quad 0x3ffe502ee78b3ff6, 0x3ffe7a51fbc74c83 /* 2^( 118 /128), 2^( 119 /128) */
|
||||
.quad 0x3ffea4afa2a490da, 0x3ffecf482d8e67f1 /* 2^( 120 /128), 2^( 121 /128) */
|
||||
.quad 0x3ffefa1bee615a27, 0x3fff252b376bba97 /* 2^( 122 /128), 2^( 123 /128) */
|
||||
.quad 0x3fff50765b6e4540, 0x3fff7bfdad9cbe14 /* 2^( 124 /128), 2^( 125 /128) */
|
||||
.quad 0x3fffa7c1819e90d8, 0x3fffd3c22b8f71f1 /* 2^( 126 /128), 2^( 127 /128) */
|
||||
.align 32
|
||||
.quad 0x42c8000000000000, 0x42c8000000000000, 0x42c8000000000000, 0x42c8000000000000 /* _dbShifter - 0x433-7=0x42c shifted right on K! */
|
||||
//log2(relerr) = -53.547756365162
|
||||
.align 32
|
||||
.quad 0x3fe62e42fefa3685, 0x3fe62e42fefa3685, 0x3fe62e42fefa3685, 0x3fe62e42fefa3685 /* _dPC1 */
|
||||
.align 32
|
||||
.quad 0x3fcebfbdff82ca48, 0x3fcebfbdff82ca48, 0x3fcebfbdff82ca48, 0x3fcebfbdff82ca48 /* _dPC2 */
|
||||
.align 32
|
||||
.quad 0x3fac6b09b180f045, 0x3fac6b09b180f045, 0x3fac6b09b180f045, 0x3fac6b09b180f045 /* _dPC3 */
|
||||
.align 32
|
||||
.quad 0x3f83b2ab5bb1268f, 0x3f83b2ab5bb1268f, 0x3f83b2ab5bb1268f, 0x3f83b2ab5bb1268f /* _dPC4 */
|
||||
.align 32
|
||||
.quad 0x000000000000007f, 0x000000000000007f, 0x000000000000007f, 0x000000000000007f /* _lIndexMask =(2^K-1) */
|
||||
.align 32
|
||||
.long 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff /* _iAbsMask */
|
||||
.align 32
|
||||
.long 0x408fefff, 0x408fefff, 0x408fefff, 0x408fefff, 0x408fefff, 0x408fefff, 0x408fefff, 0x408fefff /* _iDomainRange */
|
||||
.align 32
|
||||
.type __svml_dexp2_data_internal, @object
|
||||
.size __svml_dexp2_data_internal, .-__svml_dexp2_data_internal
|
||||
|
||||
Loading…
Reference in New Issue
Block a user