mirror of git://sourceware.org/git/glibc.git
x86_64: Fix svml_d_cosh8_core_avx512.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:
parent
e4d257bdf9
commit
a92606d154
|
|
@ -34,290 +34,288 @@
|
|||
|
||||
/* Offsets for data table __svml_dcosh_data_internal
|
||||
*/
|
||||
#define _dTp_h 0
|
||||
#define _dTn_h 128
|
||||
#define _dbShifter_UISA 256
|
||||
#define _dPC2_UISA 320
|
||||
#define _dPC3_UISA 384
|
||||
#define _dPC4_UISA 448
|
||||
#define _dPC5_UISA 512
|
||||
#define _dPC6_UISA 576
|
||||
#define _dPC7_UISA 640
|
||||
#define _dbInvLn2 704
|
||||
#define _dbLn2hi 768
|
||||
#define _dbLn2lo 832
|
||||
#define _dbShifter 896
|
||||
#define _dPC2 960
|
||||
#define _dPC3 1024
|
||||
#define _dPC4 1088
|
||||
#define _lExpMask 1152
|
||||
#define _dSign 1216
|
||||
#define _iDomainRange 1280
|
||||
#define _dTp_h 0
|
||||
#define _dTn_h 128
|
||||
#define _dbShifter_UISA 256
|
||||
#define _dPC2_UISA 320
|
||||
#define _dPC3_UISA 384
|
||||
#define _dPC4_UISA 448
|
||||
#define _dPC5_UISA 512
|
||||
#define _dPC6_UISA 576
|
||||
#define _dPC7_UISA 640
|
||||
#define _dbInvLn2 704
|
||||
#define _dbLn2hi 768
|
||||
#define _dbLn2lo 832
|
||||
#define _dbShifter 896
|
||||
#define _dPC2 960
|
||||
#define _dPC3 1024
|
||||
#define _dPC4 1088
|
||||
#define _lExpMask 1152
|
||||
#define _dSign 1216
|
||||
#define _iDomainRange 1280
|
||||
|
||||
#include <sysdep.h>
|
||||
|
||||
.text
|
||||
.section .text.evex512,"ax",@progbits
|
||||
.section .text.evex512, "ax", @progbits
|
||||
ENTRY(_ZGVeN8v_cosh_skx)
|
||||
pushq %rbp
|
||||
cfi_def_cfa_offset(16)
|
||||
movq %rsp, %rbp
|
||||
cfi_def_cfa(6, 16)
|
||||
cfi_offset(6, -16)
|
||||
andq $-64, %rsp
|
||||
subq $192, %rsp
|
||||
vmovups _dSign+__svml_dcosh_data_internal(%rip), %zmm11
|
||||
vmovups _dbShifter_UISA+__svml_dcosh_data_internal(%rip), %zmm15
|
||||
pushq %rbp
|
||||
cfi_def_cfa_offset(16)
|
||||
movq %rsp, %rbp
|
||||
cfi_def_cfa(6, 16)
|
||||
cfi_offset(6, -16)
|
||||
andq $-64, %rsp
|
||||
subq $192, %rsp
|
||||
vmovups _dSign+__svml_dcosh_data_internal(%rip), %zmm11
|
||||
vmovups _dbShifter_UISA+__svml_dcosh_data_internal(%rip), %zmm15
|
||||
|
||||
/*
|
||||
* Load argument
|
||||
* dM = x*2^K/log(2) + RShifter
|
||||
*/
|
||||
vmovups _dbInvLn2+__svml_dcosh_data_internal(%rip), %zmm4
|
||||
vmovups _dbLn2hi+__svml_dcosh_data_internal(%rip), %zmm2
|
||||
vmovups _dbLn2lo+__svml_dcosh_data_internal(%rip), %zmm3
|
||||
vmovups _dPC7_UISA+__svml_dcosh_data_internal(%rip), %zmm8
|
||||
vmovups _dPC6_UISA+__svml_dcosh_data_internal(%rip), %zmm9
|
||||
vmovups _dPC2_UISA+__svml_dcosh_data_internal(%rip), %zmm7
|
||||
vmovups _dPC3_UISA+__svml_dcosh_data_internal(%rip), %zmm6
|
||||
vmovaps %zmm0, %zmm10
|
||||
/*
|
||||
* Load argument
|
||||
* dM = x*2^K/log(2) + RShifter
|
||||
*/
|
||||
vmovups _dbInvLn2+__svml_dcosh_data_internal(%rip), %zmm4
|
||||
vmovups _dbLn2hi+__svml_dcosh_data_internal(%rip), %zmm2
|
||||
vmovups _dbLn2lo+__svml_dcosh_data_internal(%rip), %zmm3
|
||||
vmovups _dPC7_UISA+__svml_dcosh_data_internal(%rip), %zmm8
|
||||
vmovups _dPC6_UISA+__svml_dcosh_data_internal(%rip), %zmm9
|
||||
vmovups _dPC2_UISA+__svml_dcosh_data_internal(%rip), %zmm7
|
||||
vmovups _dPC3_UISA+__svml_dcosh_data_internal(%rip), %zmm6
|
||||
vmovaps %zmm0, %zmm10
|
||||
|
||||
/* Abs argument */
|
||||
vandnpd %zmm10, %zmm11, %zmm5
|
||||
/* Abs argument */
|
||||
vandnpd %zmm10, %zmm11, %zmm5
|
||||
|
||||
/* Index and lookup */
|
||||
vmovups __svml_dcosh_data_internal(%rip), %zmm11
|
||||
vmovups _dTn_h+__svml_dcosh_data_internal(%rip), %zmm0
|
||||
vfmadd213pd {rn-sae}, %zmm15, %zmm5, %zmm4
|
||||
/* Index and lookup */
|
||||
vmovups __svml_dcosh_data_internal(%rip), %zmm11
|
||||
vmovups _dTn_h+__svml_dcosh_data_internal(%rip), %zmm0
|
||||
vfmadd213pd {rn-sae}, %zmm15, %zmm5, %zmm4
|
||||
|
||||
/*
|
||||
* Check for overflow\underflow
|
||||
*
|
||||
*/
|
||||
vpsrlq $32, %zmm5, %zmm12
|
||||
/*
|
||||
* Check for overflow\underflow
|
||||
*
|
||||
*/
|
||||
vpsrlq $32, %zmm5, %zmm12
|
||||
|
||||
/* dN = dM - RShifter */
|
||||
vsubpd {rn-sae}, %zmm15, %zmm4, %zmm1
|
||||
vpmovqd %zmm12, %ymm13
|
||||
vpermt2pd _dTn_h+64+__svml_dcosh_data_internal(%rip), %zmm4, %zmm0
|
||||
vpermt2pd _dTp_h+64+__svml_dcosh_data_internal(%rip), %zmm4, %zmm11
|
||||
/* dN = dM - RShifter */
|
||||
vsubpd {rn-sae}, %zmm15, %zmm4, %zmm1
|
||||
vpmovqd %zmm12, %ymm13
|
||||
vpermt2pd _dTn_h+64+__svml_dcosh_data_internal(%rip), %zmm4, %zmm0
|
||||
vpermt2pd _dTp_h+64+__svml_dcosh_data_internal(%rip), %zmm4, %zmm11
|
||||
|
||||
/* dR = dX - dN*Log2_hi/2^K */
|
||||
vfnmadd231pd {rn-sae}, %zmm2, %zmm1, %zmm5
|
||||
/* dR = dX - dN*Log2_hi/2^K */
|
||||
vfnmadd231pd {rn-sae}, %zmm2, %zmm1, %zmm5
|
||||
|
||||
/*
|
||||
* poly(r) = Gmjp(1 + a2*r^2 + a4*r^4) + Gmjn*(r+ a3*r^3 +a5*r^5) =
|
||||
* = Gmjp_h +Gmjp_l+ Gmjp*r^2*(a2 + a4*r^2) + Gmjn*(r+ r^3*(a3 +a5*r^2)
|
||||
*/
|
||||
vmovups _dPC5_UISA+__svml_dcosh_data_internal(%rip), %zmm12
|
||||
vpsllq $48, %zmm4, %zmm2
|
||||
/*
|
||||
* poly(r) = Gmjp(1 + a2*r^2 + a4*r^4) + Gmjn*(r+ a3*r^3 +a5*r^5) =
|
||||
* = Gmjp_h +Gmjp_l+ Gmjp*r^2*(a2 + a4*r^2) + Gmjn*(r+ r^3*(a3 +a5*r^2)
|
||||
*/
|
||||
vmovups _dPC5_UISA+__svml_dcosh_data_internal(%rip), %zmm12
|
||||
vpsllq $48, %zmm4, %zmm2
|
||||
|
||||
/* dR = dX - dN*Log2_hi/2^K */
|
||||
vfnmadd231pd {rn-sae}, %zmm3, %zmm1, %zmm5
|
||||
vmulpd {rn-sae}, %zmm5, %zmm5, %zmm1
|
||||
vfmadd231pd {rn-sae}, %zmm1, %zmm8, %zmm12
|
||||
vmovups _dPC4_UISA+__svml_dcosh_data_internal(%rip), %zmm8
|
||||
vfmadd213pd {rn-sae}, %zmm6, %zmm1, %zmm12
|
||||
vfmadd231pd {rn-sae}, %zmm1, %zmm9, %zmm8
|
||||
vfmadd213pd {rn-sae}, %zmm7, %zmm1, %zmm8
|
||||
vpcmpgtd _iDomainRange+__svml_dcosh_data_internal(%rip), %ymm13, %ymm14
|
||||
vmovmskps %ymm14, %edx
|
||||
/* dR = dX - dN*Log2_hi/2^K */
|
||||
vfnmadd231pd {rn-sae}, %zmm3, %zmm1, %zmm5
|
||||
vmulpd {rn-sae}, %zmm5, %zmm5, %zmm1
|
||||
vfmadd231pd {rn-sae}, %zmm1, %zmm8, %zmm12
|
||||
vmovups _dPC4_UISA+__svml_dcosh_data_internal(%rip), %zmm8
|
||||
vfmadd213pd {rn-sae}, %zmm6, %zmm1, %zmm12
|
||||
vfmadd231pd {rn-sae}, %zmm1, %zmm9, %zmm8
|
||||
vfmadd213pd {rn-sae}, %zmm7, %zmm1, %zmm8
|
||||
vpcmpgtd _iDomainRange+__svml_dcosh_data_internal(%rip), %ymm13, %ymm14
|
||||
vmovmskps %ymm14, %edx
|
||||
|
||||
/* dOut=r^2*(a2 + a4*r^2) */
|
||||
vmulpd {rn-sae}, %zmm1, %zmm8, %zmm6
|
||||
/* dOut=r^2*(a2 + a4*r^2) */
|
||||
vmulpd {rn-sae}, %zmm1, %zmm8, %zmm6
|
||||
|
||||
/* lM now is an EXP(2^N) */
|
||||
vpandq _lExpMask+__svml_dcosh_data_internal(%rip), %zmm2, %zmm3
|
||||
vpaddq %zmm3, %zmm11, %zmm4
|
||||
vpsubq %zmm3, %zmm0, %zmm0
|
||||
vsubpd {rn-sae}, %zmm0, %zmm4, %zmm14
|
||||
vaddpd {rn-sae}, %zmm0, %zmm4, %zmm13
|
||||
/* lM now is an EXP(2^N) */
|
||||
vpandq _lExpMask+__svml_dcosh_data_internal(%rip), %zmm2, %zmm3
|
||||
vpaddq %zmm3, %zmm11, %zmm4
|
||||
vpsubq %zmm3, %zmm0, %zmm0
|
||||
vsubpd {rn-sae}, %zmm0, %zmm4, %zmm14
|
||||
vaddpd {rn-sae}, %zmm0, %zmm4, %zmm13
|
||||
|
||||
/* dM=r^2*(a3 +a5*r^2) */
|
||||
vmulpd {rn-sae}, %zmm1, %zmm12, %zmm0
|
||||
vfmadd213pd {rn-sae}, %zmm13, %zmm13, %zmm6
|
||||
/* dM=r^2*(a3 +a5*r^2) */
|
||||
vmulpd {rn-sae}, %zmm1, %zmm12, %zmm0
|
||||
vfmadd213pd {rn-sae}, %zmm13, %zmm13, %zmm6
|
||||
|
||||
/* dM= r + r^3*(a3 +a5*r^2) */
|
||||
vfmadd213pd {rn-sae}, %zmm5, %zmm5, %zmm0
|
||||
vfmadd213pd {rn-sae}, %zmm6, %zmm14, %zmm0
|
||||
testl %edx, %edx
|
||||
/* dM= r + r^3*(a3 +a5*r^2) */
|
||||
vfmadd213pd {rn-sae}, %zmm5, %zmm5, %zmm0
|
||||
vfmadd213pd {rn-sae}, %zmm6, %zmm14, %zmm0
|
||||
testl %edx, %edx
|
||||
|
||||
/* Go to special inputs processing branch */
|
||||
jne L(SPECIAL_VALUES_BRANCH)
|
||||
# LOE rbx r12 r13 r14 r15 edx zmm0 zmm10
|
||||
/* Go to special inputs processing branch */
|
||||
jne L(SPECIAL_VALUES_BRANCH)
|
||||
# LOE rbx r12 r13 r14 r15 edx zmm0 zmm10
|
||||
|
||||
/* 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):
|
||||
vmovups %zmm10, 64(%rsp)
|
||||
vmovups %zmm0, 128(%rsp)
|
||||
# LOE rbx r12 r13 r14 r15 edx zmm0
|
||||
vmovups %zmm10, 64(%rsp)
|
||||
vmovups %zmm0, 128(%rsp)
|
||||
# LOE rbx r12 r13 r14 r15 edx zmm0
|
||||
|
||||
xorl %eax, %eax
|
||||
# LOE rbx r12 r13 r14 r15 eax edx
|
||||
xorl %eax, %eax
|
||||
# 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: -64; DW_OP_and; DW_OP_const4s: -176; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x50, 0xff, 0xff, 0xff, 0x22
|
||||
movl %eax, %r12d
|
||||
movq %r13, 8(%rsp)
|
||||
/* DW_CFA_expression: r13 (r13) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -184; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x48, 0xff, 0xff, 0xff, 0x22
|
||||
movl %edx, %r13d
|
||||
movq %r14, (%rsp)
|
||||
/* DW_CFA_expression: r14 (r14) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -192; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x40, 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: -64; DW_OP_and; DW_OP_const4s: -176; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x50, 0xff, 0xff, 0xff, 0x22
|
||||
movl %eax, %r12d
|
||||
movq %r13, 8(%rsp)
|
||||
/* DW_CFA_expression: r13 (r13) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -184; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x48, 0xff, 0xff, 0xff, 0x22
|
||||
movl %edx, %r13d
|
||||
movq %r14, (%rsp)
|
||||
/* DW_CFA_expression: r14 (r14) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -192; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x40, 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 $8, %r12d
|
||||
incl %r12d
|
||||
cmpl $8, %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)
|
||||
vmovups 128(%rsp), %zmm0
|
||||
movq 16(%rsp), %r12
|
||||
cfi_restore(12)
|
||||
movq 8(%rsp), %r13
|
||||
cfi_restore(13)
|
||||
movq (%rsp), %r14
|
||||
cfi_restore(14)
|
||||
vmovups 128(%rsp), %zmm0
|
||||
|
||||
/* Go to exit */
|
||||
jmp L(EXIT)
|
||||
/* DW_CFA_expression: r12 (r12) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -176; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x50, 0xff, 0xff, 0xff, 0x22
|
||||
/* DW_CFA_expression: r13 (r13) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -184; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x48, 0xff, 0xff, 0xff, 0x22
|
||||
/* DW_CFA_expression: r14 (r14) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -192; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x40, 0xff, 0xff, 0xff, 0x22
|
||||
# LOE rbx r12 r13 r14 r15 zmm0
|
||||
/* Go to exit */
|
||||
jmp L(EXIT)
|
||||
/* DW_CFA_expression: r12 (r12) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -176; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x50, 0xff, 0xff, 0xff, 0x22
|
||||
/* DW_CFA_expression: r13 (r13) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -184; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x48, 0xff, 0xff, 0xff, 0x22
|
||||
/* DW_CFA_expression: r14 (r14) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -192; DW_OP_plus) */
|
||||
.cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x40, 0xff, 0xff, 0xff, 0x22
|
||||
# LOE rbx r12 r13 r14 r15 zmm0
|
||||
|
||||
/* Scalar math fucntion call
|
||||
* to process special input
|
||||
*/
|
||||
/* Scalar math fucntion call
|
||||
* to process special input
|
||||
*/
|
||||
|
||||
L(SCALAR_MATH_CALL):
|
||||
movl %r12d, %r14d
|
||||
movsd 64(%rsp,%r14,8), %xmm0
|
||||
call cosh@PLT
|
||||
# LOE rbx r14 r15 r12d r13d xmm0
|
||||
movl %r12d, %r14d
|
||||
movsd 64(%rsp, %r14, 8), %xmm0
|
||||
call cosh@PLT
|
||||
# LOE rbx r14 r15 r12d r13d xmm0
|
||||
|
||||
movsd %xmm0, 128(%rsp,%r14,8)
|
||||
movsd %xmm0, 128(%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(_ZGVeN8v_cosh_skx)
|
||||
|
||||
.section .rodata, "a"
|
||||
.align 64
|
||||
.section .rodata, "a"
|
||||
.align 64
|
||||
|
||||
#ifdef __svml_dcosh_data_internal_typedef
|
||||
typedef unsigned int VUINT32;
|
||||
typedef struct
|
||||
{
|
||||
__declspec(align(64)) VUINT32 _dTp_h[(1<<4)][2];
|
||||
__declspec(align(64)) VUINT32 _dTn_h[(1<<4)][2];
|
||||
__declspec(align(64)) VUINT32 _dbShifter_UISA[8][2];
|
||||
__declspec(align(64)) VUINT32 _dPC2_UISA[8][2];
|
||||
__declspec(align(64)) VUINT32 _dPC3_UISA[8][2];
|
||||
__declspec(align(64)) VUINT32 _dPC4_UISA[8][2];
|
||||
__declspec(align(64)) VUINT32 _dPC5_UISA[8][2];
|
||||
__declspec(align(64)) VUINT32 _dPC6_UISA[8][2];
|
||||
__declspec(align(64)) VUINT32 _dPC7_UISA[8][2];
|
||||
__declspec(align(64)) VUINT32 _dbInvLn2[8][2];
|
||||
__declspec(align(64)) VUINT32 _dbLn2hi[8][2];
|
||||
__declspec(align(64)) VUINT32 _dbLn2lo[8][2];
|
||||
__declspec(align(64)) VUINT32 _dbShifter[8][2];
|
||||
__declspec(align(64)) VUINT32 _dPC2[8][2];
|
||||
__declspec(align(64)) VUINT32 _dPC3[8][2];
|
||||
__declspec(align(64)) VUINT32 _dPC4[8][2];
|
||||
__declspec(align(64)) VUINT32 _lExpMask[8][2];
|
||||
__declspec(align(64)) VUINT32 _dSign[8][2]; //0x8000000000000000
|
||||
__declspec(align(64)) VUINT32 _iDomainRange[16][1];
|
||||
typedef struct {
|
||||
__declspec(align(64)) VUINT32 _dTp_h[(1<<4)][2];
|
||||
__declspec(align(64)) VUINT32 _dTn_h[(1<<4)][2];
|
||||
__declspec(align(64)) VUINT32 _dbShifter_UISA[8][2];
|
||||
__declspec(align(64)) VUINT32 _dPC2_UISA[8][2];
|
||||
__declspec(align(64)) VUINT32 _dPC3_UISA[8][2];
|
||||
__declspec(align(64)) VUINT32 _dPC4_UISA[8][2];
|
||||
__declspec(align(64)) VUINT32 _dPC5_UISA[8][2];
|
||||
__declspec(align(64)) VUINT32 _dPC6_UISA[8][2];
|
||||
__declspec(align(64)) VUINT32 _dPC7_UISA[8][2];
|
||||
__declspec(align(64)) VUINT32 _dbInvLn2[8][2];
|
||||
__declspec(align(64)) VUINT32 _dbLn2hi[8][2];
|
||||
__declspec(align(64)) VUINT32 _dbLn2lo[8][2];
|
||||
__declspec(align(64)) VUINT32 _dbShifter[8][2];
|
||||
__declspec(align(64)) VUINT32 _dPC2[8][2];
|
||||
__declspec(align(64)) VUINT32 _dPC3[8][2];
|
||||
__declspec(align(64)) VUINT32 _dPC4[8][2];
|
||||
__declspec(align(64)) VUINT32 _lExpMask[8][2];
|
||||
__declspec(align(64)) VUINT32 _dSign[8][2]; // 0x8000000000000000
|
||||
__declspec(align(64)) VUINT32 _iDomainRange[16][1];
|
||||
} __svml_dcosh_data_internal;
|
||||
#endif
|
||||
__svml_dcosh_data_internal:
|
||||
/*== _dTp_h ==*/
|
||||
.quad 0x3fe0000000000000, 0x3fe0b5586cf9890f, 0x3fe172b83c7d517b, 0x3fe2387a6e756238
|
||||
.quad 0x3fe306fe0a31b715, 0x3fe3dea64c123422, 0x3fe4bfdad5362a27, 0x3fe5ab07dd485429
|
||||
.quad 0x3fe6a09e667f3bcd, 0x3fe7a11473eb0187, 0x3fe8ace5422aa0db, 0x3fe9c49182a3f090
|
||||
.quad 0x3feae89f995ad3ad, 0x3fec199bdd85529c, 0x3fed5818dcfba487, 0x3feea4afa2a490da
|
||||
/*== dTn_h ==*/
|
||||
.align 64
|
||||
.quad 0x3fe0000000000000, 0x3fdea4afa2a490da, 0x3fdd5818dcfba487, 0x3fdc199bdd85529c
|
||||
.quad 0x3fdae89f995ad3ad, 0x3fd9c49182a3f090, 0x3fd8ace5422aa0db, 0x3fd7a11473eb0187
|
||||
.quad 0x3fd6a09e667f3bcd, 0x3fd5ab07dd485429, 0x3fd4bfdad5362a27, 0x3fd3dea64c123422
|
||||
.quad 0x3fd306fe0a31b715, 0x3fd2387a6e756238, 0x3fd172b83c7d517b, 0x3fd0b5586cf9890f
|
||||
.align 64
|
||||
.quad 0x42F8000000000000, 0x42F8000000000000, 0x42F8000000000000, 0x42F8000000000000, 0x42F8000000000000, 0x42F8000000000000, 0x42F8000000000000, 0x42F8000000000000 /* _dbShifter_UISA */
|
||||
.align 64
|
||||
.quad 0x3fe0000000000004, 0x3fe0000000000004, 0x3fe0000000000004, 0x3fe0000000000004, 0x3fe0000000000004, 0x3fe0000000000004, 0x3fe0000000000004, 0x3fe0000000000004 /* _dPC2_UISA */
|
||||
.align 64
|
||||
.quad 0x3fc5555555555543, 0x3fc5555555555543, 0x3fc5555555555543, 0x3fc5555555555543, 0x3fc5555555555543, 0x3fc5555555555543, 0x3fc5555555555543, 0x3fc5555555555543 /* _dPC3_UISA */
|
||||
.align 64
|
||||
.quad 0x3fa5555555484f37, 0x3fa5555555484f37, 0x3fa5555555484f37, 0x3fa5555555484f37, 0x3fa5555555484f37, 0x3fa5555555484f37, 0x3fa5555555484f37, 0x3fa5555555484f37 /* _dPC4_UISA */
|
||||
.align 64
|
||||
.quad 0x3f81111111286a0c, 0x3f81111111286a0c, 0x3f81111111286a0c, 0x3f81111111286a0c, 0x3f81111111286a0c, 0x3f81111111286a0c, 0x3f81111111286a0c, 0x3f81111111286a0c /* _dPC5_UISA */
|
||||
.align 64
|
||||
.quad 0x3f56c183da08f116, 0x3f56c183da08f116, 0x3f56c183da08f116, 0x3f56c183da08f116, 0x3f56c183da08f116, 0x3f56c183da08f116, 0x3f56c183da08f116, 0x3f56c183da08f116 /* _dPC6_UISA */
|
||||
.align 64
|
||||
.quad 0x3f2a018d76da03da, 0x3f2a018d76da03da, 0x3f2a018d76da03da, 0x3f2a018d76da03da, 0x3f2a018d76da03da, 0x3f2a018d76da03da, 0x3f2a018d76da03da, 0x3f2a018d76da03da /* _dPC7_UISA */
|
||||
/*== _dbT ==*/
|
||||
.align 64
|
||||
.quad 0x3ff71547652b82fe, 0x3ff71547652b82fe, 0x3ff71547652b82fe, 0x3ff71547652b82fe, 0x3ff71547652b82fe, 0x3ff71547652b82fe, 0x3ff71547652b82fe, 0x3ff71547652b82fe /* _dbInvLn2 = 1/log(2) */
|
||||
.align 64
|
||||
.quad 0x3FE62E42FEFC0000, 0x3FE62E42FEFC0000, 0x3FE62E42FEFC0000, 0x3FE62E42FEFC0000, 0x3FE62E42FEFC0000, 0x3FE62E42FEFC0000, 0x3FE62E42FEFC0000, 0x3FE62E42FEFC0000 /* _dbLn2hi = log(2) hi*/
|
||||
.align 64
|
||||
.quad 0xBDAC610CA86C3899, 0xBDAC610CA86C3899, 0xBDAC610CA86C3899, 0xBDAC610CA86C3899, 0xBDAC610CA86C3899, 0xBDAC610CA86C3899, 0xBDAC610CA86C3899, 0xBDAC610CA86C3899 /* _dbLn2lo = log(2) lo*/
|
||||
.align 64
|
||||
.quad 0x42B8000000000000, 0x42B8000000000000, 0x42B8000000000000, 0x42B8000000000000, 0x42B8000000000000, 0x42B8000000000000, 0x42B8000000000000, 0x42B8000000000000 /* _dbShifter */
|
||||
.align 64
|
||||
.quad 0x3FDFFFFFFFFFFDBD, 0x3FDFFFFFFFFFFDBD, 0x3FDFFFFFFFFFFDBD, 0x3FDFFFFFFFFFFDBD, 0x3FDFFFFFFFFFFDBD, 0x3FDFFFFFFFFFFDBD, 0x3FDFFFFFFFFFFDBD, 0x3FDFFFFFFFFFFDBD /* _dPC2 */
|
||||
.align 64
|
||||
.quad 0x3FC5555570813E14, 0x3FC5555570813E14, 0x3FC5555570813E14, 0x3FC5555570813E14, 0x3FC5555570813E14, 0x3FC5555570813E14, 0x3FC5555570813E14, 0x3FC5555570813E14 /* _dPC3 */
|
||||
.align 64
|
||||
.quad 0x3FA55555CF16D299, 0x3FA55555CF16D299, 0x3FA55555CF16D299, 0x3FA55555CF16D299, 0x3FA55555CF16D299, 0x3FA55555CF16D299, 0x3FA55555CF16D299, 0x3FA55555CF16D299 /* _dPC4 */
|
||||
.align 64
|
||||
.quad 0x7ff0000000000000, 0x7ff0000000000000, 0x7ff0000000000000, 0x7ff0000000000000, 0x7ff0000000000000, 0x7ff0000000000000, 0x7ff0000000000000, 0x7ff0000000000000 /* _lExpMask */
|
||||
.align 64
|
||||
.quad 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000 /* _dSign*/
|
||||
.align 64
|
||||
.long 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99 /* _iDomainRange 0x40861d9ac12a3e85 =(1021*2^K-0.5)*log(2)/2^K -needed for quick exp*/
|
||||
.align 64
|
||||
.type __svml_dcosh_data_internal,@object
|
||||
.size __svml_dcosh_data_internal,.-__svml_dcosh_data_internal
|
||||
/* _dTp_h */
|
||||
.quad 0x3fe0000000000000, 0x3fe0b5586cf9890f, 0x3fe172b83c7d517b, 0x3fe2387a6e756238
|
||||
.quad 0x3fe306fe0a31b715, 0x3fe3dea64c123422, 0x3fe4bfdad5362a27, 0x3fe5ab07dd485429
|
||||
.quad 0x3fe6a09e667f3bcd, 0x3fe7a11473eb0187, 0x3fe8ace5422aa0db, 0x3fe9c49182a3f090
|
||||
.quad 0x3feae89f995ad3ad, 0x3fec199bdd85529c, 0x3fed5818dcfba487, 0x3feea4afa2a490da
|
||||
/* dTn_h */
|
||||
.align 64
|
||||
.quad 0x3fe0000000000000, 0x3fdea4afa2a490da, 0x3fdd5818dcfba487, 0x3fdc199bdd85529c
|
||||
.quad 0x3fdae89f995ad3ad, 0x3fd9c49182a3f090, 0x3fd8ace5422aa0db, 0x3fd7a11473eb0187
|
||||
.quad 0x3fd6a09e667f3bcd, 0x3fd5ab07dd485429, 0x3fd4bfdad5362a27, 0x3fd3dea64c123422
|
||||
.quad 0x3fd306fe0a31b715, 0x3fd2387a6e756238, 0x3fd172b83c7d517b, 0x3fd0b5586cf9890f
|
||||
.align 64
|
||||
.quad 0x42F8000000000000, 0x42F8000000000000, 0x42F8000000000000, 0x42F8000000000000, 0x42F8000000000000, 0x42F8000000000000, 0x42F8000000000000, 0x42F8000000000000 /* _dbShifter_UISA */
|
||||
.align 64
|
||||
.quad 0x3fe0000000000004, 0x3fe0000000000004, 0x3fe0000000000004, 0x3fe0000000000004, 0x3fe0000000000004, 0x3fe0000000000004, 0x3fe0000000000004, 0x3fe0000000000004 /* _dPC2_UISA */
|
||||
.align 64
|
||||
.quad 0x3fc5555555555543, 0x3fc5555555555543, 0x3fc5555555555543, 0x3fc5555555555543, 0x3fc5555555555543, 0x3fc5555555555543, 0x3fc5555555555543, 0x3fc5555555555543 /* _dPC3_UISA */
|
||||
.align 64
|
||||
.quad 0x3fa5555555484f37, 0x3fa5555555484f37, 0x3fa5555555484f37, 0x3fa5555555484f37, 0x3fa5555555484f37, 0x3fa5555555484f37, 0x3fa5555555484f37, 0x3fa5555555484f37 /* _dPC4_UISA */
|
||||
.align 64
|
||||
.quad 0x3f81111111286a0c, 0x3f81111111286a0c, 0x3f81111111286a0c, 0x3f81111111286a0c, 0x3f81111111286a0c, 0x3f81111111286a0c, 0x3f81111111286a0c, 0x3f81111111286a0c /* _dPC5_UISA */
|
||||
.align 64
|
||||
.quad 0x3f56c183da08f116, 0x3f56c183da08f116, 0x3f56c183da08f116, 0x3f56c183da08f116, 0x3f56c183da08f116, 0x3f56c183da08f116, 0x3f56c183da08f116, 0x3f56c183da08f116 /* _dPC6_UISA */
|
||||
.align 64
|
||||
.quad 0x3f2a018d76da03da, 0x3f2a018d76da03da, 0x3f2a018d76da03da, 0x3f2a018d76da03da, 0x3f2a018d76da03da, 0x3f2a018d76da03da, 0x3f2a018d76da03da, 0x3f2a018d76da03da /* _dPC7_UISA */
|
||||
/* _dbT */
|
||||
.align 64
|
||||
.quad 0x3ff71547652b82fe, 0x3ff71547652b82fe, 0x3ff71547652b82fe, 0x3ff71547652b82fe, 0x3ff71547652b82fe, 0x3ff71547652b82fe, 0x3ff71547652b82fe, 0x3ff71547652b82fe /* _dbInvLn2 = 1/log(2) */
|
||||
.align 64
|
||||
.quad 0x3FE62E42FEFC0000, 0x3FE62E42FEFC0000, 0x3FE62E42FEFC0000, 0x3FE62E42FEFC0000, 0x3FE62E42FEFC0000, 0x3FE62E42FEFC0000, 0x3FE62E42FEFC0000, 0x3FE62E42FEFC0000 /* _dbLn2hi = log(2) hi */
|
||||
.align 64
|
||||
.quad 0xBDAC610CA86C3899, 0xBDAC610CA86C3899, 0xBDAC610CA86C3899, 0xBDAC610CA86C3899, 0xBDAC610CA86C3899, 0xBDAC610CA86C3899, 0xBDAC610CA86C3899, 0xBDAC610CA86C3899 /* _dbLn2lo = log(2) lo */
|
||||
.align 64
|
||||
.quad 0x42B8000000000000, 0x42B8000000000000, 0x42B8000000000000, 0x42B8000000000000, 0x42B8000000000000, 0x42B8000000000000, 0x42B8000000000000, 0x42B8000000000000 /* _dbShifter */
|
||||
.align 64
|
||||
.quad 0x3FDFFFFFFFFFFDBD, 0x3FDFFFFFFFFFFDBD, 0x3FDFFFFFFFFFFDBD, 0x3FDFFFFFFFFFFDBD, 0x3FDFFFFFFFFFFDBD, 0x3FDFFFFFFFFFFDBD, 0x3FDFFFFFFFFFFDBD, 0x3FDFFFFFFFFFFDBD /* _dPC2 */
|
||||
.align 64
|
||||
.quad 0x3FC5555570813E14, 0x3FC5555570813E14, 0x3FC5555570813E14, 0x3FC5555570813E14, 0x3FC5555570813E14, 0x3FC5555570813E14, 0x3FC5555570813E14, 0x3FC5555570813E14 /* _dPC3 */
|
||||
.align 64
|
||||
.quad 0x3FA55555CF16D299, 0x3FA55555CF16D299, 0x3FA55555CF16D299, 0x3FA55555CF16D299, 0x3FA55555CF16D299, 0x3FA55555CF16D299, 0x3FA55555CF16D299, 0x3FA55555CF16D299 /* _dPC4 */
|
||||
.align 64
|
||||
.quad 0x7ff0000000000000, 0x7ff0000000000000, 0x7ff0000000000000, 0x7ff0000000000000, 0x7ff0000000000000, 0x7ff0000000000000, 0x7ff0000000000000, 0x7ff0000000000000 /* _lExpMask */
|
||||
.align 64
|
||||
.quad 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000 /* _dSign */
|
||||
.align 64
|
||||
.long 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99, 0x40861d99 /* _iDomainRange 0x40861d9ac12a3e85 =(1021*2^K-0.5)*log(2)/2^K -needed for quick exp */
|
||||
.align 64
|
||||
.type __svml_dcosh_data_internal, @object
|
||||
.size __svml_dcosh_data_internal, .-__svml_dcosh_data_internal
|
||||
|
|
|
|||
Loading…
Reference in New Issue