x86_64: Fix svml_d_tanh4_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>

1 files changed, 1164 insertions, 1166 deletions

diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_tanh4_core_avx2.S b/sysdeps/x86_64/fpu/multiarch/svml_d_tanh4_core_avx2.S
index d2a971ead5..7ddf145b25 100644
--- a/sysdeps/x86_64/fpu/multiarch/svml_d_tanh4_core_avx2.S
+++ b/sysdeps/x86_64/fpu/multiarch/svml_d_tanh4_core_avx2.S
@@ -28,7 +28,7 @@
  *   and to approximate tanh(.) with a polynomial on each of them.
  *
  *   IEEE SPECIAL CONDITIONS:
- *   x = [+,-]0, r = [+,-]0
+ *   x = [+, -]0, r = [+, -]0
  *   x = +Inf,   r = +1
  *   x = -Inf,   r = -1
  *   x = QNaN,   r = QNaN
@@ -72,1208 +72,1206 @@
 
 /* Offsets for data table __svml_dtanh_data_internal
  */
-#define _dbP                          	0
-#define _dbSignMask                   	7680
-#define _dbAbsMask                    	7712
-#define _iExpMantMask                 	7744
-#define _iExpMask                     	7776
-#define _iMinIdxOfsMask               	7808
-#define _iMaxIdxMask                  	7840
+#define _dbP				0
+#define _dbSignMask			7680
+#define _dbAbsMask			7712
+#define _iExpMantMask			7744
+#define _iExpMask			7776
+#define _iMinIdxOfsMask			7808
+#define _iMaxIdxMask			7840
 
 #include <sysdep.h>
 
-        .text
-	.section .text.avx2,"ax",@progbits
+	.section .text.avx2, "ax", @progbits
 ENTRY(_ZGVdN4v_tanh_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
-        lea       _dbP+96+__svml_dtanh_data_internal(%rip), %r8
-        vmovupd   %ymm0, (%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
+	lea	_dbP+96+__svml_dtanh_data_internal(%rip), %r8
+	vmovupd	%ymm0, (%rsp)
 
-/* if VMIN, VMAX is defined for I type */
-        vpxor     %xmm11, %xmm11, %xmm11
+	/* if VMIN, VMAX is defined for I type */
+	vpxor	%xmm11, %xmm11, %xmm11
 
-/*  Constant loading  */
-        vmovups   _iMaxIdxMask+__svml_dtanh_data_internal(%rip), %xmm8
-        vandpd    _dbAbsMask+__svml_dtanh_data_internal(%rip), %ymm0, %ymm1
-        vandpd    _dbSignMask+__svml_dtanh_data_internal(%rip), %ymm0, %ymm2
-        vextractf128 $1, %ymm0, %xmm15
-        vshufps   $221, %xmm15, %xmm0, %xmm14
+	/*  Constant loading  */
+	vmovups	_iMaxIdxMask+__svml_dtanh_data_internal(%rip), %xmm8
+	vandpd	_dbAbsMask+__svml_dtanh_data_internal(%rip), %ymm0, %ymm1
+	vandpd	_dbSignMask+__svml_dtanh_data_internal(%rip), %ymm0, %ymm2
+	vextractf128 $1, %ymm0, %xmm15
+	vshufps	$221, %xmm15, %xmm0, %xmm14
 
-/* Here huge arguments, INF and NaNs are filtered out to callout. */
-        vpand     _iExpMantMask+__svml_dtanh_data_internal(%rip), %xmm14, %xmm12
-        vpsubd    _iMinIdxOfsMask+__svml_dtanh_data_internal(%rip), %xmm12, %xmm9
-        vpcmpgtd  %xmm11, %xmm9, %xmm10
-        vpcmpgtd  %xmm8, %xmm9, %xmm0
-        vpand     %xmm10, %xmm9, %xmm7
-        blendvps  %xmm0, %xmm8, %xmm7
+	/* Here huge arguments, INF and NaNs are filtered out to callout. */
+	vpand	_iExpMantMask+__svml_dtanh_data_internal(%rip), %xmm14, %xmm12
+	vpsubd	_iMinIdxOfsMask+__svml_dtanh_data_internal(%rip), %xmm12, %xmm9
+	vpcmpgtd %xmm11, %xmm9, %xmm10
+	vpcmpgtd %xmm8, %xmm9, %xmm0
+	vpand	%xmm10, %xmm9, %xmm7
+	blendvps %xmm0, %xmm8, %xmm7
 
-/*
- * VSHRIMM( I, iIndex, = iIndex, (17 - 4) );
- * VGATHER_MATRIX( L2D, p, TAB._dbP, iIndex, 0, T_ITEM_SIZE, T_ITEM_GRAN, 13, 0, 0 );
- */
-        vpsrld    $10, %xmm7, %xmm6
-        vmovd     %xmm6, %edx
-        vpcmpgtd  _iExpMask+__svml_dtanh_data_internal(%rip), %xmm12, %xmm13
-        vmovmskps %xmm13, %eax
-        vpextrd   $1, %xmm6, %ecx
-        movslq    %edx, %rdx
-        movslq    %ecx, %rcx
-        vpextrd   $2, %xmm6, %esi
-        vpextrd   $3, %xmm6, %edi
-        movslq    %esi, %rsi
-        movslq    %edi, %rdi
-        vmovupd   -96(%rdx,%r8), %xmm3
-        vmovupd   -96(%rcx,%r8), %xmm4
-        vmovupd   -80(%rcx,%r8), %xmm13
-        vmovupd   -64(%rcx,%r8), %xmm9
-        vmovupd   -80(%rdx,%r8), %xmm14
-        vmovupd   -64(%rdx,%r8), %xmm10
-        vmovupd   -48(%rdx,%r8), %xmm6
-        vinsertf128 $1, -96(%rsi,%r8), %ymm3, %ymm0
-        vinsertf128 $1, -96(%rdi,%r8), %ymm4, %ymm15
-        vmovupd   -48(%rcx,%r8), %xmm3
-        vunpckhpd %ymm15, %ymm0, %ymm0
-        vinsertf128 $1, -80(%rsi,%r8), %ymm14, %ymm12
-        vinsertf128 $1, -64(%rsi,%r8), %ymm10, %ymm8
-        vinsertf128 $1, -80(%rdi,%r8), %ymm13, %ymm11
-        vinsertf128 $1, -64(%rdi,%r8), %ymm9, %ymm7
-        vunpcklpd %ymm11, %ymm12, %ymm15
-        vunpckhpd %ymm11, %ymm12, %ymm14
-        vunpcklpd %ymm7, %ymm8, %ymm13
-        vunpckhpd %ymm7, %ymm8, %ymm12
-        vmovupd   -32(%rdx,%r8), %xmm9
-        vmovupd   -32(%rcx,%r8), %xmm8
-        vinsertf128 $1, -48(%rsi,%r8), %ymm6, %ymm4
-        vinsertf128 $1, -48(%rdi,%r8), %ymm3, %ymm5
-        vunpcklpd %ymm5, %ymm4, %ymm11
-        vunpckhpd %ymm5, %ymm4, %ymm10
-        vmovupd   -16(%rdx,%r8), %xmm3
-        vmovupd   -16(%rcx,%r8), %xmm4
-        vinsertf128 $1, -32(%rsi,%r8), %ymm9, %ymm7
-        vinsertf128 $1, -32(%rdi,%r8), %ymm8, %ymm6
-        vunpcklpd %ymm6, %ymm7, %ymm9
-        vunpckhpd %ymm6, %ymm7, %ymm8
-        vinsertf128 $1, -16(%rsi,%r8), %ymm3, %ymm5
-        vinsertf128 $1, -16(%rdi,%r8), %ymm4, %ymm6
-        vunpcklpd %ymm6, %ymm5, %ymm7
-        vunpckhpd %ymm6, %ymm5, %ymm6
-        vmovupd   (%rdx,%r8), %xmm3
-        vmovupd   (%rcx,%r8), %xmm5
-        vinsertf128 $1, (%rsi,%r8), %ymm3, %ymm4
-        vinsertf128 $1, (%rdi,%r8), %ymm5, %ymm5
-        vunpcklpd %ymm5, %ymm4, %ymm3
-        vaddpd    %ymm3, %ymm1, %ymm1
-        vfmadd213pd %ymm7, %ymm1, %ymm6
-        vfmadd213pd %ymm8, %ymm1, %ymm6
-        vfmadd213pd %ymm9, %ymm1, %ymm6
-        vfmadd213pd %ymm10, %ymm1, %ymm6
-        vfmadd213pd %ymm11, %ymm1, %ymm6
-        vfmadd213pd %ymm12, %ymm1, %ymm6
-        vfmadd213pd %ymm13, %ymm1, %ymm6
-        vfmadd213pd %ymm14, %ymm1, %ymm6
-        vfmadd213pd %ymm15, %ymm1, %ymm6
-        vfmadd213pd %ymm0, %ymm1, %ymm6
-        vorpd     %ymm2, %ymm6, %ymm0
-        testl     %eax, %eax
+	/*
+	 * VSHRIMM( I, iIndex, = iIndex, (17 - 4) );
+	 * VGATHER_MATRIX( L2D, p, TAB._dbP, iIndex, 0, T_ITEM_SIZE, T_ITEM_GRAN, 13, 0, 0 );
+	 */
+	vpsrld	$10, %xmm7, %xmm6
+	vmovd	%xmm6, %edx
+	vpcmpgtd _iExpMask+__svml_dtanh_data_internal(%rip), %xmm12, %xmm13
+	vmovmskps %xmm13, %eax
+	vpextrd	$1, %xmm6, %ecx
+	movslq	%edx, %rdx
+	movslq	%ecx, %rcx
+	vpextrd	$2, %xmm6, %esi
+	vpextrd	$3, %xmm6, %edi
+	movslq	%esi, %rsi
+	movslq	%edi, %rdi
+	vmovupd	-96(%rdx, %r8), %xmm3
+	vmovupd	-96(%rcx, %r8), %xmm4
+	vmovupd	-80(%rcx, %r8), %xmm13
+	vmovupd	-64(%rcx, %r8), %xmm9
+	vmovupd	-80(%rdx, %r8), %xmm14
+	vmovupd	-64(%rdx, %r8), %xmm10
+	vmovupd	-48(%rdx, %r8), %xmm6
+	vinsertf128 $1, -96(%rsi, %r8), %ymm3, %ymm0
+	vinsertf128 $1, -96(%rdi, %r8), %ymm4, %ymm15
+	vmovupd	-48(%rcx, %r8), %xmm3
+	vunpckhpd %ymm15, %ymm0, %ymm0
+	vinsertf128 $1, -80(%rsi, %r8), %ymm14, %ymm12
+	vinsertf128 $1, -64(%rsi, %r8), %ymm10, %ymm8
+	vinsertf128 $1, -80(%rdi, %r8), %ymm13, %ymm11
+	vinsertf128 $1, -64(%rdi, %r8), %ymm9, %ymm7
+	vunpcklpd %ymm11, %ymm12, %ymm15
+	vunpckhpd %ymm11, %ymm12, %ymm14
+	vunpcklpd %ymm7, %ymm8, %ymm13
+	vunpckhpd %ymm7, %ymm8, %ymm12
+	vmovupd	-32(%rdx, %r8), %xmm9
+	vmovupd	-32(%rcx, %r8), %xmm8
+	vinsertf128 $1, -48(%rsi, %r8), %ymm6, %ymm4
+	vinsertf128 $1, -48(%rdi, %r8), %ymm3, %ymm5
+	vunpcklpd %ymm5, %ymm4, %ymm11
+	vunpckhpd %ymm5, %ymm4, %ymm10
+	vmovupd	-16(%rdx, %r8), %xmm3
+	vmovupd	-16(%rcx, %r8), %xmm4
+	vinsertf128 $1, -32(%rsi, %r8), %ymm9, %ymm7
+	vinsertf128 $1, -32(%rdi, %r8), %ymm8, %ymm6
+	vunpcklpd %ymm6, %ymm7, %ymm9
+	vunpckhpd %ymm6, %ymm7, %ymm8
+	vinsertf128 $1, -16(%rsi, %r8), %ymm3, %ymm5
+	vinsertf128 $1, -16(%rdi, %r8), %ymm4, %ymm6
+	vunpcklpd %ymm6, %ymm5, %ymm7
+	vunpckhpd %ymm6, %ymm5, %ymm6
+	vmovupd	(%rdx, %r8), %xmm3
+	vmovupd	(%rcx, %r8), %xmm5
+	vinsertf128 $1, (%rsi, %r8), %ymm3, %ymm4
+	vinsertf128 $1, (%rdi, %r8), %ymm5, %ymm5
+	vunpcklpd %ymm5, %ymm4, %ymm3
+	vaddpd	%ymm3, %ymm1, %ymm1
+	vfmadd213pd %ymm7, %ymm1, %ymm6
+	vfmadd213pd %ymm8, %ymm1, %ymm6
+	vfmadd213pd %ymm9, %ymm1, %ymm6
+	vfmadd213pd %ymm10, %ymm1, %ymm6
+	vfmadd213pd %ymm11, %ymm1, %ymm6
+	vfmadd213pd %ymm12, %ymm1, %ymm6
+	vfmadd213pd %ymm13, %ymm1, %ymm6
+	vfmadd213pd %ymm14, %ymm1, %ymm6
+	vfmadd213pd %ymm15, %ymm1, %ymm6
+	vfmadd213pd %ymm0, %ymm1, %ymm6
+	vorpd	%ymm2, %ymm6, %ymm0
+	testl	%eax, %eax
 
-/* Go to special inputs processing branch */
-        jne       L(SPECIAL_VALUES_BRANCH)
-                                # LOE rbx r12 r13 r14 r15 eax ymm0
+	/* Go to special inputs processing branch */
+	jne	L(SPECIAL_VALUES_BRANCH)
+	# LOE rbx r12 r13 r14 r15 eax ymm0
 
-/* 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   (%rsp), %ymm1
-        vmovupd   %ymm0, 64(%rsp)
-        vmovupd   %ymm1, 32(%rsp)
-                                # LOE rbx r12 r13 r14 r15 eax ymm0
+	vmovupd	(%rsp), %ymm1
+	vmovupd	%ymm0, 64(%rsp)
+	vmovupd	%ymm1, 32(%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      tanh@PLT
-                                # LOE rbx r14 r15 r12d r13d xmm0
+	movl	%r12d, %r14d
+	movsd	32(%rsp, %r14, 8), %xmm0
+	call	tanh@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_tanh_avx2)
 
-        .section .rodata, "a"
-        .align 32
+	.section .rodata, "a"
+	.align	32
 
 #ifdef __svml_dtanh_data_internal_typedef
 typedef unsigned int VUINT32;
-typedef struct
-{
-        __declspec(align(32)) VUINT32 _dbP[60*16][2];
-        __declspec(align(32)) VUINT32 _dbSignMask[4][2];
-        __declspec(align(32)) VUINT32 _dbAbsMask[4][2];
-        __declspec(align(32)) VUINT32 _iExpMantMask[8][1];
-        __declspec(align(32)) VUINT32 _iExpMask[8][1];
-        __declspec(align(32)) VUINT32 _iMinIdxOfsMask[8][1];
-        __declspec(align(32)) VUINT32 _iMaxIdxMask[8][1];
+typedef struct {
+	__declspec(align(32)) VUINT32 _dbP[60*16][2];
+	__declspec(align(32)) VUINT32 _dbSignMask[4][2];
+	__declspec(align(32)) VUINT32 _dbAbsMask[4][2];
+	__declspec(align(32)) VUINT32 _iExpMantMask[8][1];
+	__declspec(align(32)) VUINT32 _iExpMask[8][1];
+	__declspec(align(32)) VUINT32 _iMinIdxOfsMask[8][1];
+	__declspec(align(32)) VUINT32 _iMaxIdxMask[8][1];
 } __svml_dtanh_data_internal;
 #endif
 __svml_dtanh_data_internal:
-        /* Polynomial coefficients */
-        .quad 0x0000000000000000   /* PL0 = +0.000000000000000000000e-01 */
-        .quad 0x0000000000000000   /* PH0 = +0.000000000000000000000e-01 */
-        .quad 0x3FF0000000000000   /* P1  = +1.000000000000000014103e+00 */
-        .quad 0xBD197DEAD79668D3   /* P2  = -2.264132406596103056796e-14 */
-        .quad 0xBFD555555553AF3C   /* P3  = -3.333333333273349741024e-01 */
-        .quad 0xBE052F7CCA134846   /* P4  = -6.165791385711493738399e-10 */
-        .quad 0x3FC11111563849D6   /* P5  = +1.333333655353061107201e-01 */
-        .quad 0xBEB038623673FFB2   /* P6  = -9.668021563879858950855e-07 */
-        .quad 0xBFAB9F685E64022E   /* P7  = -5.395055916051593179252e-02 */
-        .quad 0xBF2A54E2B28F2207   /* P8  = -2.008940439550829012647e-04 */
-        .quad 0x3F97CFB9328A230E   /* P9  = +2.325333949059698582189e-02 */
-        .quad 0xBF75CA6D61723E02   /* P10 = -5.320002811586290441790e-03 */
-        .quad 0x0000000000000000   /* B = +0        */
-        .quad 0x3FF0000000000000   /* A = +1.0      */
-        .quad 0x0000000000000000   /* Align value = +0        */
-        .quad 0x0000000000000000   /* Align value = +0        */
-        .quad 0x3C3708A564FAD29A   /* PL0 = +1.248663375337163807466e-18 */
-        .quad 0x3FC0E6973998DA48   /* PH0 = +1.320370703922029154143e-01 */
-        .quad 0x3FEF712EB25C0888   /* P1  = +9.825662120422444519229e-01 */
-        .quad 0xBFC09B296F7C1EA9   /* P2  = -1.297351641044220078331e-01 */
-        .quad 0xBFD3DD77541EDDA7   /* P3  = -3.103922196855485849143e-01 */
-        .quad 0x3FB58FFCF4309615   /* P4  = +8.422833406128689275566e-02 */
-        .quad 0x3FBD3ABE845DCF49   /* P5  = +1.141776154670967208833e-01 */
-        .quad 0xBFA791DF538C37FA   /* P6  = -4.603479285115947936529e-02 */
-        .quad 0xBFA4F872F69CD6E8   /* P7  = -4.095801601799370195284e-02 */
-        .quad 0x3F9772E49EF6412B   /* P8  = +2.289921970583567527179e-02 */
-        .quad 0x3F8CBC0807393909   /* P9  = +1.403051635784581776625e-02 */
-        .quad 0xBF85F06A30F93319   /* P10 = -1.071246110873285040939e-02 */
-        .quad 0xBFC1000000000000   /* B = -.132813 */
-        .quad 0x3FF0000000000000   /* A = +1        */
-        .quad 0x0000000000000000   /* Align value = +0        */
-        .quad 0x0000000000000000   /* Align value = +0        */
-        .quad 0x3C6004EE5739DEAC   /* PL0 = +6.947247374112211856530e-18 */
-        .quad 0x3FC2DC968E6E0D62   /* PH0 = +1.473568149050193398786e-01 */
-        .quad 0x3FEF4E1E606D96DF   /* P1  = +9.782859691010478680677e-01 */
-        .quad 0xBFC273BD70994AB9   /* P2  = -1.441571044730005866646e-01 */
-        .quad 0xBFD382B548270D2C   /* P3  = -3.048527912726111386771e-01 */
-        .quad 0x3FB7CD2D582A6B29   /* P4  = +9.297450449450351894400e-02 */
-        .quad 0x3FBC1278CCCBF0DB   /* P5  = +1.096568584434324642303e-01 */
-        .quad 0xBFA9C7F5115B86A1   /* P6  = -5.035367810138536095866e-02 */
-        .quad 0xBFA371C21BAF618E   /* P7  = -3.797728145554222910481e-02 */
-        .quad 0x3F9958943F68417E   /* P8  = +2.475196492201935923783e-02 */
-        .quad 0x3F8930D5CFFD4152   /* P9  = +1.230017701132682667572e-02 */
-        .quad 0xBF875CF7ADD31B76   /* P10 = -1.140779017658897660092e-02 */
-        .quad 0xBFC3000000000000   /* B = -.148438 */
-        .quad 0x3FF0000000000000   /* A = +1        */
-        .quad 0x0000000000000000   /* Align value = +0        */
-        .quad 0x0000000000000000   /* Align value = +0        */
-        .quad 0x3C7EABE24E052A1F   /* PL0 = +2.660321779421749543501e-17 */
-        .quad 0x3FC4D04783618C71   /* PH0 = +1.626061812886266111366e-01 */
-        .quad 0x3FEF2765AF97A4B3   /* P1  = +9.735592298067302883212e-01 */
-        .quad 0xBFC443654205FEA5   /* P2  = -1.583067486171689074207e-01 */
-        .quad 0xBFD31F2E208A5B97   /* P3  = -2.987780874040536844467e-01 */
-        .quad 0x3FB9F235BD339878   /* P4  = +1.013520800512156573576e-01 */
-        .quad 0x3FBAD0B0DFCCA141   /* P5  = +1.047468706498238100104e-01 */
-        .quad 0xBFABD1B9600E608E   /* P6  = -5.433444306908184548967e-02 */
-        .quad 0xBFA1CEBEAF07DB58   /* P7  = -3.478046309094534453598e-02 */
-        .quad 0x3F9AFC9FB1D8EFD2   /* P8  = +2.635430834764902126383e-02 */
-        .quad 0x3F8573444F1AB502   /* P9  = +1.047376028449287564018e-02 */
-        .quad 0xBF8874FBC8F24406   /* P10 = -1.194187838544459322219e-02 */
-        .quad 0xBFC5000000000000   /* B = -.164063 */
-        .quad 0x3FF0000000000000   /* A = +1        */
-        .quad 0x0000000000000000   /* Align value = +0        */
-        .quad 0x0000000000000000   /* Align value = +0        */
-        .quad 0x3C7FB199D361A790   /* PL0 = +2.748994907060158996213e-17 */
-        .quad 0x3FC6C170259E21F7   /* PH0 = +1.777782615356639783766e-01 */
-        .quad 0x3FEEFD17479F7C65   /* P1  = +9.683948897253570478266e-01 */
-        .quad 0xBFC609530FE4DF8D   /* P2  = -1.721595599753950294577e-01 */
-        .quad 0xBFD2B3465D71B4DE   /* P3  = -2.921920692959484052676e-01 */
-        .quad 0x3FBBFD2D34AC509B   /* P4  = +1.093319181057403192166e-01 */
-        .quad 0x3FB9778C3C16A0FE   /* P5  = +9.948040453912551395183e-02 */
-        .quad 0xBFADAC4D9E63C665   /* P6  = -5.795519407719210697372e-02 */
-        .quad 0xBFA0139CCAD02D60   /* P7  = -3.139963126894929339124e-02 */
-        .quad 0x3F9C5BF43BA6F19D   /* P8  = +2.769452680671379432854e-02 */
-        .quad 0x3F8190B703350341   /* P9  = +8.576803002712575184772e-03 */
-        .quad 0xBF8936606782858A   /* P10 = -1.231074634444230850234e-02 */
-        .quad 0xBFC7000000000000   /* B = -.179688 */
-        .quad 0x3FF0000000000000   /* A = +1        */
-        .quad 0x0000000000000000   /* Align value = +0        */
-        .quad 0x0000000000000000   /* Align value = +0        */
-        .quad 0x3C6A917CA3624D50   /* PL0 = +1.152216693509785660691e-17 */
-        .quad 0x3FC8AFD7B974FABB   /* PH0 = +1.928662925292508878439e-01 */
-        .quad 0x3FEECF47624A5D03   /* P1  = +9.628025932060214187231e-01 */
-        .quad 0xBFC7C4C2CB4FDE4D   /* P2  = -1.856921665891938814679e-01 */
-        .quad 0xBFD23F69CB2C1F9D   /* P3  = -2.851204380135586155453e-01 */
-        .quad 0x3FBDEC5703A03814   /* P4  = +1.168875106670557712458e-01 */
-        .quad 0x3FB8095003D0CF15   /* P5  = +9.389209836154706616487e-02 */
-        .quad 0xBFAF554B47B10CBB   /* P6  = -6.119761705533607365968e-02 */
-        .quad 0xBF9C89743FE7BC1B   /* P7  = -2.786809577986213853937e-02 */
-        .quad 0x3F9D74725B746E7C   /* P8  = +2.876452143855921824991e-02 */
-        .quad 0x3F7B2D8AFB70B88C   /* P9  = +6.635229968237631511880e-03 */
-        .quad 0xBF89A0A2883EF6CB   /* P10 = -1.251341799058582545252e-02 */
-        .quad 0xBFC9000000000000   /* B = -.195313 */
-        .quad 0x3FF0000000000000   /* A = +1        */
-        .quad 0x0000000000000000   /* Align value = +0        */
-        .quad 0x0000000000000000   /* Align value = +0        */
-        .quad 0x3C7608279E8609CB   /* PL0 = +1.910958764623660748269e-17 */
-        .quad 0x3FCA9B46D2DDC5E3   /* PH0 = +2.078636674519166172015e-01 */
-        .quad 0x3FEE9E0BB72A01A1   /* P1  = +9.567926957534390123919e-01 */
-        .quad 0xBFC974FAD10C5330   /* P2  = -1.988824387305156976885e-01 */
-        .quad 0xBFD1C40ACCBA4044   /* P3  = -2.775904654781735703430e-01 */
-        .quad 0x3FBFBE24E2987853   /* P4  = +1.239951184474830487522e-01 */
-        .quad 0x3FB6885B4345E47F   /* P5  = +8.801813499839460539687e-02 */
-        .quad 0xBFB06563D5670584   /* P6  = -6.404708824176991770896e-02 */
-        .quad 0xBF98CD1D620DF6E2   /* P7  = -2.421995078065365147772e-02 */
-        .quad 0x3F9E44EF3E844D21   /* P8  = +2.955983943054463683119e-02 */
-        .quad 0x3F7325FA0148CAAE   /* P9  = +4.674889165971292322643e-03 */
-        .quad 0xBF89B4C8556C2D92   /* P10 = -1.255184660614964011319e-02 */
-        .quad 0xBFCB000000000000   /* B = -.210938 */
-        .quad 0x3FF0000000000000   /* A = +1        */
-        .quad 0x0000000000000000   /* Align value = +0        */
-        .quad 0x0000000000000000   /* Align value = +0        */
-        .quad 0x3C6F19DAA20F51D5   /* PL0 = +1.348790537832000351176e-17 */
-        .quad 0x3FCC83876CA98E15   /* PH0 = +2.227639465883021474557e-01 */
-        .quad 0x3FEE697B662D07CD   /* P1  = +9.503762241004040620296e-01 */
-        .quad 0xBFCB194C7ED76ACF   /* P2  = -2.117095584242946953999e-01 */
-        .quad 0xBFD141A19E419762   /* P3  = -2.696308179350720680191e-01 */
-        .quad 0x3FC0B89C64BC7B98   /* P4  = +1.306338779331468503007e-01 */
-        .quad 0x3FB4F721150BBFC5   /* P5  = +8.189589275184434216748e-02 */
-        .quad 0xBFB105AAFAB87898   /* P6  = -6.649273511036069461061e-02 */
-        .quad 0xBF94FB3B31248C01   /* P7  = -2.048962104266749732921e-02 */
-        .quad 0x3F9ECD31E588709C   /* P8  = +3.007963145692880855964e-02 */
-        .quad 0x3F664A91A335C105   /* P9  = +2.721104095762541127495e-03 */
-        .quad 0xBF89754E32E1E26E   /* P10 = -1.243077366619723806134e-02 */
-        .quad 0xBFCD000000000000   /* B = -.226563 */
-        .quad 0x3FF0000000000000   /* A = +1        */
-        .quad 0x0000000000000000   /* Align value = +0        */
-        .quad 0x0000000000000000   /* Align value = +0        */
-        .quad 0x3C6AC6C889D8111D   /* PL0 = +1.161245469312620769170e-17 */
-        .quad 0x3FCE6864FE55A3D0   /* PH0 = +2.375608674877001114112e-01 */
-        .quad 0x3FEE31AEE116B82B   /* P1  = +9.435648342384913826391e-01 */
-        .quad 0xBFCCB114B69E808B   /* P2  = -2.241540805525839833707e-01 */
-        .quad 0xBFD0B8AB913BA99D   /* P3  = -2.612713735858507980441e-01 */
-        .quad 0x3FC1823322BED48A   /* P4  = +1.367858810096190233514e-01 */
-        .quad 0x3FB35822B7929893   /* P5  = +7.556359273675842651653e-02 */
-