diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_atan24_core_avx2.S b/sysdeps/x86_64/fpu/multiarch/svml_d_atan24_core_avx2.S index 9438c1b751..1b601576cc 100644 --- a/sysdeps/x86_64/fpu/multiarch/svml_d_atan24_core_avx2.S +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_atan24_core_avx2.S @@ -30,422 +30,421 @@ /* Offsets for data table __svml_datan2_data_internal */ -#define dPI 0 -#define dPIO2 32 -#define dA19 64 -#define dA18 96 -#define dA17 128 -#define dA16 160 -#define dA15 192 -#define dA14 224 -#define dA13 256 -#define dA12 288 -#define dA11 320 -#define dA10 352 -#define dA09 384 -#define dA08 416 -#define dA07 448 -#define dA06 480 -#define dA05 512 -#define dA04 544 -#define dA03 576 -#define dA02 608 -#define dA01 640 -#define dA00 672 -#define dSIGN_MASK 704 -#define iCHK_WORK_SUB 736 -#define iCHK_WORK_CMP 768 -#define dABS_MASK 800 -#define dZERO 832 +#define dPI 0 +#define dPIO2 32 +#define dA19 64 +#define dA18 96 +#define dA17 128 +#define dA16 160 +#define dA15 192 +#define dA14 224 +#define dA13 256 +#define dA12 288 +#define dA11 320 +#define dA10 352 +#define dA09 384 +#define dA08 416 +#define dA07 448 +#define dA06 480 +#define dA05 512 +#define dA04 544 +#define dA03 576 +#define dA02 608 +#define dA01 640 +#define dA00 672 +#define dSIGN_MASK 704 +#define iCHK_WORK_SUB 736 +#define iCHK_WORK_CMP 768 +#define dABS_MASK 800 +#define dZERO 832 #include - .text - .section .text.avx2,"ax",@progbits + .section .text.avx2, "ax", @progbits ENTRY(_ZGVdN4vv_atan2_avx2) - pushq %rbp - cfi_def_cfa_offset(16) - movq %rsp, %rbp - cfi_def_cfa(6, 16) - cfi_offset(6, -16) - andq $-32, %rsp - subq $128, %rsp - xorl %edx, %edx + pushq %rbp + cfi_def_cfa_offset(16) + movq %rsp, %rbp + cfi_def_cfa(6, 16) + cfi_offset(6, -16) + andq $-32, %rsp + subq $128, %rsp + xorl %edx, %edx -/* - * #define NO_VECTOR_ZERO_ATAN2_ARGS - * Declarations - * Variables - * Constants - * The end of declarations - * Implementation - * Get r0~=1/B - * Cannot be replaced by VQRCP(D, dR0, dB); - * Argument Absolute values - */ - vmovupd dABS_MASK+__svml_datan2_data_internal(%rip), %ymm5 + /* + * #define NO_VECTOR_ZERO_ATAN2_ARGS + * Declarations + * Variables + * Constants + * The end of declarations + * Implementation + * Get r0~=1/B + * Cannot be replaced by VQRCP(D, dR0, dB); + * Argument Absolute values + */ + vmovupd dABS_MASK+__svml_datan2_data_internal(%rip), %ymm5 -/* Argument signs */ - vmovupd dSIGN_MASK+__svml_datan2_data_internal(%rip), %ymm4 - vmovups iCHK_WORK_SUB+__svml_datan2_data_internal(%rip), %xmm13 - vmovupd %ymm0, (%rsp) - vmovapd %ymm1, %ymm8 - vandpd %ymm5, %ymm8, %ymm2 - vandpd %ymm5, %ymm0, %ymm1 - vcmpnlt_uqpd %ymm2, %ymm1, %ymm15 + /* Argument signs */ + vmovupd dSIGN_MASK+__svml_datan2_data_internal(%rip), %ymm4 + vmovups iCHK_WORK_SUB+__svml_datan2_data_internal(%rip), %xmm13 + vmovupd %ymm0, (%rsp) + vmovapd %ymm1, %ymm8 + vandpd %ymm5, %ymm8, %ymm2 + vandpd %ymm5, %ymm0, %ymm1 + vcmpnlt_uqpd %ymm2, %ymm1, %ymm15 -/* - * 1) If yx then a=-x, b=y, PIO2=Pi/2 - */ - vorpd %ymm4, %ymm2, %ymm6 - vblendvpd %ymm15, %ymm6, %ymm1, %ymm3 - vblendvpd %ymm15, %ymm1, %ymm2, %ymm6 - vdivpd %ymm6, %ymm3, %ymm14 - vmovups iCHK_WORK_CMP+__svml_datan2_data_internal(%rip), %xmm3 - vmovupd %ymm6, 32(%rsp) - vandpd %ymm4, %ymm0, %ymm7 - vandpd %ymm4, %ymm8, %ymm5 - vandpd dPIO2+__svml_datan2_data_internal(%rip), %ymm15, %ymm4 + /* + * 1) If yx then a=-x, b=y, PIO2=Pi/2 + */ + vorpd %ymm4, %ymm2, %ymm6 + vblendvpd %ymm15, %ymm6, %ymm1, %ymm3 + vblendvpd %ymm15, %ymm1, %ymm2, %ymm6 + vdivpd %ymm6, %ymm3, %ymm14 + vmovups iCHK_WORK_CMP+__svml_datan2_data_internal(%rip), %xmm3 + vmovupd %ymm6, 32(%rsp) + vandpd %ymm4, %ymm0, %ymm7 + vandpd %ymm4, %ymm8, %ymm5 + vandpd dPIO2+__svml_datan2_data_internal(%rip), %ymm15, %ymm4 -/* Check if y and x are on main path. */ - vextractf128 $1, %ymm2, %xmm9 - vextractf128 $1, %ymm1, %xmm10 - vshufps $221, %xmm9, %xmm2, %xmm11 - vshufps $221, %xmm10, %xmm1, %xmm12 - vpsubd %xmm13, %xmm11, %xmm0 - vpsubd %xmm13, %xmm12, %xmm9 - vpcmpgtd %xmm3, %xmm0, %xmm15 - vpcmpeqd %xmm3, %xmm0, %xmm6 - vpcmpgtd %xmm3, %xmm9, %xmm10 - vpcmpeqd %xmm3, %xmm9, %xmm3 - vpor %xmm6, %xmm15, %xmm11 - vpor %xmm3, %xmm10, %xmm12 + /* Check if y and x are on main path. */ + vextractf128 $1, %ymm2, %xmm9 + vextractf128 $1, %ymm1, %xmm10 + vshufps $221, %xmm9, %xmm2, %xmm11 + vshufps $221, %xmm10, %xmm1, %xmm12 + vpsubd %xmm13, %xmm11, %xmm0 + vpsubd %xmm13, %xmm12, %xmm9 + vpcmpgtd %xmm3, %xmm0, %xmm15 + vpcmpeqd %xmm3, %xmm0, %xmm6 + vpcmpgtd %xmm3, %xmm9, %xmm10 + vpcmpeqd %xmm3, %xmm9, %xmm3 + vpor %xmm6, %xmm15, %xmm11 + vpor %xmm3, %xmm10, %xmm12 -/* Polynomial. */ - vmulpd %ymm14, %ymm14, %ymm10 - vpor %xmm12, %xmm11, %xmm3 - vmovupd dA18+__svml_datan2_data_internal(%rip), %ymm9 - vmovupd dA17+__svml_datan2_data_internal(%rip), %ymm12 - vmovupd dA16+__svml_datan2_data_internal(%rip), %ymm15 - vmulpd %ymm10, %ymm10, %ymm11 + /* Polynomial. */ + vmulpd %ymm14, %ymm14, %ymm10 + vpor %xmm12, %xmm11, %xmm3 + vmovupd dA18+__svml_datan2_data_internal(%rip), %ymm9 + vmovupd dA17+__svml_datan2_data_internal(%rip), %ymm12 + vmovupd dA16+__svml_datan2_data_internal(%rip), %ymm15 + vmulpd %ymm10, %ymm10, %ymm11 -/* if x<0, dPI = Pi, else dPI =0 */ - vcmple_oqpd dZERO+__svml_datan2_data_internal(%rip), %ymm8, %ymm13 - vmovmskps %xmm3, %eax - vmulpd %ymm11, %ymm11, %ymm0 - vandpd __svml_datan2_data_internal(%rip), %ymm13, %ymm6 - vmovupd dA19+__svml_datan2_data_internal(%rip), %ymm13 - vfmadd213pd dA14+__svml_datan2_data_internal(%rip), %ymm0, %ymm9 - vfmadd213pd dA13+__svml_datan2_data_internal(%rip), %ymm0, %ymm12 - vfmadd213pd dA12+__svml_datan2_data_internal(%rip), %ymm0, %ymm15 - vfmadd213pd dA15+__svml_datan2_data_internal(%rip), %ymm0, %ymm13 - vfmadd213pd dA10+__svml_datan2_data_internal(%rip), %ymm0, %ymm9 - vfmadd213pd dA09+__svml_datan2_data_internal(%rip), %ymm0, %ymm12 - vfmadd213pd dA08+__svml_datan2_data_internal(%rip), %ymm0, %ymm15 - vfmadd213pd dA11+__svml_datan2_data_internal(%rip), %ymm0, %ymm13 - vfmadd213pd dA06+__svml_datan2_data_internal(%rip), %ymm0, %ymm9 - vfmadd213pd dA05+__svml_datan2_data_internal(%rip), %ymm0, %ymm12 - vfmadd213pd dA04+__svml_datan2_data_internal(%rip), %ymm0, %ymm15 - vfmadd213pd dA07+__svml_datan2_data_internal(%rip), %ymm0, %ymm13 - vfmadd213pd dA02+__svml_datan2_data_internal(%rip), %ymm0, %ymm9 - vfmadd213pd dA01+__svml_datan2_data_internal(%rip), %ymm0, %ymm12 - vfmadd213pd dA03+__svml_datan2_data_internal(%rip), %ymm0, %ymm13 + /* if x<0, dPI = Pi, else dPI =0 */ + vcmple_oqpd dZERO+__svml_datan2_data_internal(%rip), %ymm8, %ymm13 + vmovmskps %xmm3, %eax + vmulpd %ymm11, %ymm11, %ymm0 + vandpd __svml_datan2_data_internal(%rip), %ymm13, %ymm6 + vmovupd dA19+__svml_datan2_data_internal(%rip), %ymm13 + vfmadd213pd dA14+__svml_datan2_data_internal(%rip), %ymm0, %ymm9 + vfmadd213pd dA13+__svml_datan2_data_internal(%rip), %ymm0, %ymm12 + vfmadd213pd dA12+__svml_datan2_data_internal(%rip), %ymm0, %ymm15 + vfmadd213pd dA15+__svml_datan2_data_internal(%rip), %ymm0, %ymm13 + vfmadd213pd dA10+__svml_datan2_data_internal(%rip), %ymm0, %ymm9 + vfmadd213pd dA09+__svml_datan2_data_internal(%rip), %ymm0, %ymm12 + vfmadd213pd dA08+__svml_datan2_data_internal(%rip), %ymm0, %ymm15 + vfmadd213pd dA11+__svml_datan2_data_internal(%rip), %ymm0, %ymm13 + vfmadd213pd dA06+__svml_datan2_data_internal(%rip), %ymm0, %ymm9 + vfmadd213pd dA05+__svml_datan2_data_internal(%rip), %ymm0, %ymm12 + vfmadd213pd dA04+__svml_datan2_data_internal(%rip), %ymm0, %ymm15 + vfmadd213pd dA07+__svml_datan2_data_internal(%rip), %ymm0, %ymm13 + vfmadd213pd dA02+__svml_datan2_data_internal(%rip), %ymm0, %ymm9 + vfmadd213pd dA01+__svml_datan2_data_internal(%rip), %ymm0, %ymm12 + vfmadd213pd dA03+__svml_datan2_data_internal(%rip), %ymm0, %ymm13 -/* A00=1.0, account for it later VQFMA(D, dP4, dP4, dR8, dA00); */ - vmulpd %ymm15, %ymm0, %ymm0 - vfmadd213pd %ymm9, %ymm10, %ymm13 - vfmadd213pd %ymm0, %ymm10, %ymm12 - vfmadd213pd %ymm12, %ymm11, %ymm13 + /* A00=1.0, account for it later VQFMA(D, dP4, dP4, dR8, dA00); */ + vmulpd %ymm15, %ymm0, %ymm0 + vfmadd213pd %ymm9, %ymm10, %ymm13 + vfmadd213pd %ymm0, %ymm10, %ymm12 + vfmadd213pd %ymm12, %ymm11, %ymm13 -/* - * Reconstruction. - * dP=(R+R*dP) + dPIO2 - */ - vfmadd213pd %ymm14, %ymm14, %ymm13 - vaddpd %ymm13, %ymm4, %ymm14 - vorpd %ymm5, %ymm14, %ymm0 - vaddpd %ymm0, %ymm6, %ymm9 - vorpd %ymm7, %ymm9, %ymm0 + /* + * Reconstruction. + * dP=(R+R*dP) + dPIO2 + */ + vfmadd213pd %ymm14, %ymm14, %ymm13 + vaddpd %ymm13, %ymm4, %ymm14 + vorpd %ymm5, %ymm14, %ymm0 + vaddpd %ymm0, %ymm6, %ymm9 + vorpd %ymm7, %ymm9, %ymm0 -/* Special branch for fast (vector) processing of zero arguments */ - testl %eax, %eax + /* Special branch for fast (vector) processing of zero arguments */ + testl %eax, %eax -/* Go to auxilary branch */ - jne L(AUX_BRANCH) - # LOE rbx r12 r13 r14 r15 edx xmm3 ymm0 ymm1 ymm2 ymm4 ymm5 ymm6 ymm7 ymm8 + /* Go to auxilary branch */ + jne L(AUX_BRANCH) + # LOE rbx r12 r13 r14 r15 edx xmm3 ymm0 ymm1 ymm2 ymm4 ymm5 ymm6 ymm7 ymm8 -/* Return from auxilary branch - * for out of main path inputs - */ + /* Return from auxilary branch + * for out of main path inputs + */ L(AUX_BRANCH_RETURN): -/* - * Special branch for fast (vector) processing of zero arguments - * The end of implementation - */ - testl %edx, %edx + /* + * Special branch for fast (vector) processing of zero arguments + * The end of implementation + */ + testl %edx, %edx -/* Go to special inputs processing branch */ - jne L(SPECIAL_VALUES_BRANCH) - # LOE rbx r12 r13 r14 r15 edx ymm0 ymm8 + /* Go to special inputs processing branch */ + jne L(SPECIAL_VALUES_BRANCH) + # LOE rbx r12 r13 r14 r15 edx ymm0 ymm8 -/* 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 %ymm8, 64(%rsp) - vmovupd %ymm0, 96(%rsp) - vmovupd %ymm1, 32(%rsp) - # LOE rbx r12 r13 r14 r15 edx ymm0 + vmovupd (%rsp), %ymm1 + vmovupd %ymm8, 64(%rsp) + vmovupd %ymm0, 96(%rsp) + vmovupd %ymm1, 32(%rsp) + # LOE rbx r12 r13 r14 r15 edx ymm0 - 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: -32; DW_OP_and; DW_OP_const4s: -112; DW_OP_plus) */ - .cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x90, 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: -32; DW_OP_and; DW_OP_const4s: -120; DW_OP_plus) */ - .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x88, 0xff, 0xff, 0xff, 0x22 - movl %edx, %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: -128; DW_OP_plus) */ - .cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x80, 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: -112; DW_OP_plus) */ + .cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x90, 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: -32; DW_OP_and; DW_OP_const4s: -120; DW_OP_plus) */ + .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x88, 0xff, 0xff, 0xff, 0x22 + movl %edx, %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: -128; DW_OP_plus) */ + .cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x80, 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 96(%rsp), %ymm0 + movq 16(%rsp), %r12 + cfi_restore(12) + movq 8(%rsp), %r13 + cfi_restore(13) + movq (%rsp), %r14 + cfi_restore(14) + vmovupd 96(%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: -112; DW_OP_plus) */ - .cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x90, 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: -120; DW_OP_plus) */ - .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x88, 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: -128; DW_OP_plus) */ - .cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x80, 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: -112; DW_OP_plus) */ + .cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x90, 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: -120; DW_OP_plus) */ + .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x88, 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: -128; DW_OP_plus) */ + .cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x80, 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 - movsd 64(%rsp,%r14,8), %xmm1 - call atan2@PLT - # LOE rbx r14 r15 r12d r13d xmm0 + movl %r12d, %r14d + movsd 32(%rsp, %r14, 8), %xmm0 + movsd 64(%rsp, %r14, 8), %xmm1 + call atan2@PLT + # LOE rbx r14 r15 r12d r13d xmm0 - movsd %xmm0, 96(%rsp,%r14,8) + movsd %xmm0, 96(%rsp, %r14, 8) -/* Process special inputs in loop */ - jmp L(SPECIAL_VALUES_LOOP) - cfi_restore(12) - cfi_restore(13) - cfi_restore(14) - # LOE rbx r15 r12d r13d + /* Process special inputs in loop */ + jmp L(SPECIAL_VALUES_LOOP) + cfi_restore(12) + cfi_restore(13) + cfi_restore(14) + # LOE rbx r15 r12d r13d -/* Auxilary branch - * for out of main path inputs - */ + /* Auxilary branch + * for out of main path inputs + */ L(AUX_BRANCH): - vmovupd (%rsp), %ymm11 + vmovupd (%rsp), %ymm11 -/* Check if at least on of Y or Y is zero: iAXAYZERO */ - vmovupd dZERO+__svml_datan2_data_internal(%rip), %ymm10 + /* Check if at least on of Y or Y is zero: iAXAYZERO */ + vmovupd dZERO+__svml_datan2_data_internal(%rip), %ymm10 -/* Check if both X & Y are not NaNs: iXYnotNAN */ - vcmpordpd %ymm8, %ymm8, %ymm12 - vcmpordpd %ymm11, %ymm11, %ymm13 - vcmpeqpd %ymm10, %ymm2, %ymm2 - vcmpeqpd %ymm10, %ymm1, %ymm1 - vandpd %ymm13, %ymm12, %ymm14 - vorpd %ymm1, %ymm2, %ymm2 - vextractf128 $1, %ymm14, %xmm15 - vextractf128 $1, %ymm2, %xmm11 - vshufps $221, %xmm15, %xmm14, %xmm9 - vshufps $221, %xmm11, %xmm2, %xmm12 + /* Check if both X & Y are not NaNs: iXYnotNAN */ + vcmpordpd %ymm8, %ymm8, %ymm12 + vcmpordpd %ymm11, %ymm11, %ymm13 + vcmpeqpd %ymm10, %ymm2, %ymm2 + vcmpeqpd %ymm10, %ymm1, %ymm1 + vandpd %ymm13, %ymm12, %ymm14 + vorpd %ymm1, %ymm2, %ymm2 + vextractf128 $1, %ymm14, %xmm15 + vextractf128 $1, %ymm2, %xmm11 + vshufps $221, %xmm15, %xmm14, %xmm9 + vshufps $221, %xmm11, %xmm2, %xmm12 -/* - * Path for zero arguments (at least one of both) - * Check if both args are zeros (den. is zero) - */ - vcmpeqpd 32(%rsp), %ymm10, %ymm2 + /* + * Path for zero arguments (at least one of both) + * Check if both args are zeros (den. is zero) + */ + vcmpeqpd 32(%rsp), %ymm10, %ymm2 -/* Check if at least on of Y or Y is zero and not NaN: iAXAYZEROnotNAN */ - vpand %xmm9, %xmm12, %xmm1 + /* Check if at least on of Y or Y is zero and not NaN: iAXAYZEROnotNAN */ + vpand %xmm9, %xmm12, %xmm1 -/* Exclude from previous callout mask zero (and not NaN) arguments */ - vpandn %xmm3, %xmm1, %xmm3 + /* Exclude from previous callout mask zero (and not NaN) arguments */ + vpandn %xmm3, %xmm1, %xmm3 -/* Go to callout */ - vmovmskps %xmm3, %edx + /* Go to callout */ + vmovmskps %xmm3, %edx -/* Set sPIO2 to zero if den. is zero */ - vblendvpd %ymm2, %ymm10, %ymm4, %ymm4 - vorpd %ymm5, %ymm4, %ymm5 + /* Set sPIO2 to zero if den. is zero */ + vblendvpd %ymm2, %ymm10, %ymm4, %ymm4 + vorpd %ymm5, %ymm4, %ymm5 -/* Res = sign(Y)*(X<0)?(PIO2+PI):PIO2 */ - vextractf128 $1, %ymm10, %xmm2 - vextractf128 $1, %ymm8, %xmm3 - vshufps $221, %xmm2, %xmm10, %xmm4 - vshufps $221, %xmm3, %xmm8, %xmm9 - vpcmpgtd %xmm9, %xmm4, %xmm12 - vpshufd $80, %xmm12, %xmm11 - vpshufd $250, %xmm12, %xmm13 - vinsertf128 $1, %xmm13, %ymm11, %ymm14 - vandpd %ymm6, %ymm14, %ymm6 - vaddpd %ymm6, %ymm5, %ymm2 - vorpd %ymm7, %ymm2, %ymm2 + /* Res = sign(Y)*(X<0)?(PIO2+PI):PIO2 */ + vextractf128 $1, %ymm10, %xmm2 + vextractf128 $1, %ymm8, %xmm3 + vshufps $221, %xmm2, %xmm10, %xmm4 + vshufps $221, %xmm3, %xmm8, %xmm9 + vpcmpgtd %xmm9, %xmm4, %xmm12 + vpshufd $80, %xmm12, %xmm11 + vpshufd $250, %xmm12, %xmm13 + vinsertf128 $1, %xmm13, %ymm11, %ymm14 + vandpd %ymm6, %ymm14, %ymm6 + vaddpd %ymm6, %ymm5, %ymm2 + vorpd %ymm7, %ymm2, %ymm2 -/* Merge results from main and spec path */ - vpshufd $80, %xmm1, %xmm7 - vpshufd $250, %xmm1, %xmm1 - vinsertf128 $1, %xmm1, %ymm7, %ymm3 - vblendvpd %ymm3, %ymm2, %ymm0, %ymm0 + /* Merge results from main and spec path */ + vpshufd $80, %xmm1, %xmm7 + vpshufd $250, %xmm1, %xmm1 + vinsertf128 $1, %xmm1, %ymm7, %ymm3 + vblendvpd %ymm3, %ymm2, %ymm0, %ymm0 -/* Return to main vector processing path */ - jmp L(AUX_BRANCH_RETURN) - # LOE rbx r12 r13 r14 r15 edx ymm0 ymm8 + /* Return to main vector processing path */ + jmp L(AUX_BRANCH_RETURN) + # LOE rbx r12 r13 r14 r15 edx ymm0 ymm8 END(_ZGVdN4vv_atan2_avx2) - .section .rodata, "a" - .align 32 + .section .rodata, "a" + .align 32 #ifdef __svml_datan2_data_internal_typedef typedef unsigned int VUINT32; typedef struct { - __declspec(align(32)) VUINT32 dPI[4][2]; - __declspec(align(32)) VUINT32 dPIO2[4][2]; - __declspec(align(32)) VUINT32 dA19[4][2]; - __declspec(align(32)) VUINT32 dA18[4][2]; - __declspec(align(32)) VUINT32 dA17[4][2]; - __declspec(align(32)) VUINT32 dA16[4][2]; - __declspec(align(32)) VUINT32 dA15[4][2]; - __declspec(align(32)) VUINT32 dA14[4][2]; - __declspec(align(32)) VUINT32 dA13[4][2]; - __declspec(align(32)) VUINT32 dA12[4][2]; - __declspec(align(32)) VUINT32 dA11[4][2]; - __declspec(align(32)) VUINT32 dA10[4][2]; - __declspec(align(32)) VUINT32 dA09[4][2]; - __declspec(align(32)) VUINT32 dA08[4][2]; - __declspec(align(32)) VUINT32 dA07[4][2]; - __declspec(align(32)) VUINT32 dA06[4][2]; - __declspec(align(32)) VUINT32 dA05[4][2]; - __declspec(align(32)) VUINT32 dA04[4][2]; - __declspec(align(32)) VUINT32 dA03[4][2]; - __declspec(align(32)) VUINT32 dA02[4][2]; - __declspec(align(32)) VUINT32 dA01[4][2]; - __declspec(align(32)) VUINT32 dA00[4][2]; - __declspec(align(32)) VUINT32 dSIGN_MASK[4][2]; - __declspec(align(32)) VUINT32 iCHK_WORK_SUB[8][1]; - __declspec(align(32)) VUINT32 iCHK_WORK_CMP[8][1]; - __declspec(align(32)) VUINT32 dABS_MASK[4][2]; - __declspec(align(32)) VUINT32 dZERO[4][2]; + __declspec(align(32)) VUINT32 dPI[4][2]; + __declspec(align(32)) VUINT32 dPIO2[4][2]; + __declspec(align(32)) VUINT32 dA19[4][2]; + __declspec(align(32)) VUINT32 dA18[4][2]; + __declspec(align(32)) VUINT32 dA17[4][2]; + __declspec(align(32)) VUINT32 dA16[4][2]; + __declspec(align(32)) VUINT32 dA15[4][2]; + __declspec(align(32)) VUINT32 dA14[4][2]; + __declspec(align(32)) VUINT32 dA13[4][2]; + __declspec(align(32)) VUINT32 dA12[4][2]; + __declspec(align(32)) VUINT32 dA11[4][2]; + __declspec(align(32)) VUINT32 dA10[4][2]; + __declspec(align(32)) VUINT32 dA09[4][2]; + __declspec(align(32)) VUINT32 dA08[4][2]; + __declspec(align(32)) VUINT32 dA07[4][2]; + __declspec(align(32)) VUINT32 dA06[4][2]; + __declspec(align(32)) VUINT32 dA05[4][2]; + __declspec(align(32)) VUINT32 dA04[4][2]; + __declspec(align(32)) VUINT32 dA03[4][2]; + __declspec(align(32)) VUINT32 dA02[4][2]; + __declspec(align(32)) VUINT32 dA01[4][2]; + __declspec(align(32)) VUINT32 dA00[4][2]; + __declspec(align(32)) VUINT32 dSIGN_MASK[4][2]; + __declspec(align(32)) VUINT32 iCHK_WORK_SUB[8][1]; + __declspec(align(32)) VUINT32 iCHK_WORK_CMP[8][1]; + __declspec(align(32)) VUINT32 dABS_MASK[4][2]; + __declspec(align(32)) VUINT32 dZERO[4][2]; } __svml_datan2_data_internal; #endif __svml_datan2_data_internal: - .quad 0x400921FB54442D18, 0x400921FB54442D18, 0x400921FB54442D18, 0x400921FB54442D18 //dPI - .align 32 - .quad 0x3FF921FB54442D18, 0x3FF921FB54442D18, 0x3FF921FB54442D18, 0x3FF921FB54442D18 //dPIO2 - .align 32 - .quad 0xBEF4FDB537ABC7A3, 0xBEF4FDB537ABC7A3, 0xBEF4FDB537ABC7A3, 0xBEF4FDB537ABC7A3 // dA19 - .align 32 - .quad 0x3F2CED0A36665209, 0x3F2CED0A36665209, 0x3F2CED0A36665209, 0x3F2CED0A36665209 // dA18 - .align 32 - .quad 0xBF52E67C93954C23, 0xBF52E67C93954C23, 0xBF52E67C93954C23, 0xBF52E67C93954C23 // dA17 - .align 32 - .quad 0x3F6F5A1DAE82AFB3, 0x3F6F5A1DAE82AFB3, 0x3F6F5A1DAE82AFB3, 0x3F6F5A1DAE82AFB3 // dA16 - .align 32 - .quad 0xBF82B2EC618E4BAD, 0xBF82B2EC618E4BAD, 0xBF82B2EC618E4BAD, 0xBF82B2EC618E4BAD // dA15 - .align 32 - .quad 0x3F914F4C661116A5, 0x3F914F4C661116A5, 0x3F914F4C661116A5, 0x3F914F4C661116A5 // dA14 - .align 32 - .quad 0xBF9A5E83B081F69C, 0xBF9A5E83B081F69C, 0xBF9A5E83B081F69C, 0xBF9A5E83B081F69C // dA13 - .align 32 - .quad 0x3FA169980CB6AD4F, 0x3FA169980CB6AD4F, 0x3FA169980CB6AD4F, 0x3FA169980CB6AD4F // dA12 - .align 32 - .quad 0xBFA4EFA2E563C1BC, 0xBFA4EFA2E563C1BC, 0xBFA4EFA2E563C1BC, 0xBFA4EFA2E563C1BC // dA11 - .align 32 - .quad 0x3FA7EC0FBC50683B, 0x3FA7EC0FBC50683B, 0x3FA7EC0FBC50683B, 0x3FA7EC0FBC50683B // dA10 - .align 32 - .quad 0xBFAAD261EAA09954, 0xBFAAD261EAA09954, 0xBFAAD261EAA09954, 0xBFAAD261EAA09954 // dA09 - .align 32 - .quad 0x3FAE1749BD612DCF, 0x3FAE1749BD612DCF, 0x3FAE1749BD612DCF, 0x3FAE1749BD612DCF // dA08 - .align 32 - .quad 0xBFB11084009435E0, 0xBFB11084009435E0, 0xBFB11084009435E0, 0xBFB11084009435E0 // dA07 - .align 32 - .quad 0x3FB3B12A49295651, 0x3FB3B12A49295651, 0x3FB3B12A49295651, 0x3FB3B12A49295651 // dA06 - .align 32 - .quad 0xBFB745D009BADA94, 0xBFB745D009BADA94, 0xBFB745D009BADA94, 0xBFB745D009BADA94 // dA05 - .align 32 - .quad 0x3FBC71C707F7D5B5, 0x3FBC71C707F7D5B5, 0x3FBC71C707F7D5B5, 0x3FBC71C707F7D5B5 // dA04 - .align 32 - .quad 0xBFC2492491EE55C7, 0xBFC2492491EE55C7, 0xBFC2492491EE55C7, 0xBFC2492491EE55C7 // dA03 - .align 32 - .quad 0x3FC999999997EE34, 0x3FC999999997EE34, 0x3FC999999997EE34, 0x3FC999999997EE34 // dA02 - .align 32 - .quad 0xBFD55555555553C5, 0xBFD55555555553C5, 0xBFD55555555553C5, 0xBFD55555555553C5 // dA01 - .align 32 - .quad 0x3FF0000000000000, 0x3FF0000000000000, 0x3FF0000000000000, 0x3FF0000000000000 // dA00 - .align 32 - .quad 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000 //dSIGN_MASK - .align 32 - .long 0x80300000, 0x80300000, 0x80300000, 0x80300000, 0x80300000, 0x80300000, 0x80300000, 0x80300000 //iCHK_WORK_SUB - .align 32 - .long 0xfdd00000, 0xfdd00000, 0xfdd00000, 0xfdd00000, 0xfdd00000, 0xfdd00000, 0xfdd00000, 0xfdd00000 //iCHK_WORK_CMP - .align 32 - .quad 0x7fffffffffffffff, 0x7fffffffffffffff, 0x7fffffffffffffff, 0x7fffffffffffffff //dABS_MASK - .align 32 - .quad 0x0000000000000000, 0x0000000000000000, 0x0000000000000000, 0x0000000000000000 //dZERO - .align 32 - .type __svml_datan2_data_internal,@object - .size __svml_datan2_data_internal,.-__svml_datan2_data_internal + .quad 0x400921FB54442D18, 0x400921FB54442D18, 0x400921FB54442D18, 0x400921FB54442D18 // dPI + .align 32 + .quad 0x3FF921FB54442D18, 0x3FF921FB54442D18, 0x3FF921FB54442D18, 0x3FF921FB54442D18 // dPIO2 + .align 32 + .quad 0xBEF4FDB537ABC7A3, 0xBEF4FDB537ABC7A3, 0xBEF4FDB537ABC7A3, 0xBEF4FDB537ABC7A3 // dA19 + .align 32 + .quad 0x3F2CED0A36665209, 0x3F2CED0A36665209, 0x3F2CED0A36665209, 0x3F2CED0A36665209 // dA18 + .align 32 + .quad 0xBF52E67C93954C23, 0xBF52E67C93954C23, 0xBF52E67C93954C23, 0xBF52E67C93954C23 // dA17 + .align 32 + .quad 0x3F6F5A1DAE82AFB3, 0x3F6F5A1DAE82AFB3, 0x3F6F5A1DAE82AFB3, 0x3F6F5A1DAE82AFB3 // dA16 + .align 32 + .quad 0xBF82B2EC618E4BAD, 0xBF82B2EC618E4BAD, 0xBF82B2EC618E4BAD, 0xBF82B2EC618E4BAD // dA15 + .align 32 + .quad 0x3F914F4C661116A5, 0x3F914F4C661116A5, 0x3F914F4C661116A5, 0x3F914F4C661116A5 // dA14 + .align 32 + .quad 0xBF9A5E83B081F69C, 0xBF9A5E83B081F69C, 0xBF9A5E83B081F69C, 0xBF9A5E83B081F69C // dA13 + .align 32 + .quad 0x3FA169980CB6AD4F, 0x3FA169980CB6AD4F, 0x3FA169980CB6AD4F, 0x3FA169980CB6AD4F // dA12 + .align 32 + .quad 0xBFA4EFA2E563C1BC, 0xBFA4EFA2E563C1BC, 0xBFA4EFA2E563C1BC, 0xBFA4EFA2E563C1BC // dA11 + .align 32 + .quad 0x3FA7EC0FBC50683B, 0x3FA7EC0FBC50683B, 0x3FA7EC0FBC50683B, 0x3FA7EC0FBC50683B // dA10 + .align 32 + .quad 0xBFAAD261EAA09954, 0xBFAAD261EAA09954, 0xBFAAD261EAA09954, 0xBFAAD261EAA09954 // dA09 + .align 32 + .quad 0x3FAE1749BD612DCF, 0x3FAE1749BD612DCF, 0x3FAE1749BD612DCF, 0x3FAE1749BD612DCF // dA08 + .align 32 + .quad 0xBFB11084009435E0, 0xBFB11084009435E0, 0xBFB11084009435E0, 0xBFB11084009435E0 // dA07 + .align 32 + .quad 0x3FB3B12A49295651, 0x3FB3B12A49295651, 0x3FB3B12A49295651, 0x3FB3B12A49295651 // dA06 + .align 32 + .quad 0xBFB745D009BADA94, 0xBFB745D009BADA94, 0xBFB745D009BADA94, 0xBFB745D009BADA94 // dA05 + .align 32 + .quad 0x3FBC71C707F7D5B5, 0x3FBC71C707F7D5B5, 0x3FBC71C707F7D5B5, 0x3FBC71C707F7D5B5 // dA04 + .align 32 + .quad 0xBFC2492491EE55C7, 0xBFC2492491EE55C7, 0xBFC2492491EE55C7, 0xBFC2492491EE55C7 // dA03 + .align 32 + .quad 0x3FC999999997EE34, 0x3FC999999997EE34, 0x3FC999999997EE34, 0x3FC999999997EE34 // dA02 + .align 32 + .quad 0xBFD55555555553C5, 0xBFD55555555553C5, 0xBFD55555555553C5, 0xBFD55555555553C5 // dA01 + .align 32 + .quad 0x3FF0000000000000, 0x3FF0000000000000, 0x3FF0000000000000, 0x3FF0000000000000 // dA00 + .align 32 + .quad 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000 // dSIGN_MASK + .align 32 + .long 0x80300000, 0x80300000, 0x80300000, 0x80300000, 0x80300000, 0x80300000, 0x80300000, 0x80300000 // iCHK_WORK_SUB + .align 32 + .long 0xfdd00000, 0xfdd00000, 0xfdd00000, 0xfdd00000, 0xfdd00000, 0xfdd00000, 0xfdd00000, 0xfdd00000 // iCHK_WORK_CMP + .align 32 + .quad 0x7fffffffffffffff, 0x7fffffffffffffff, 0x7fffffffffffffff, 0x7fffffffffffffff // dABS_MASK + .align 32 + .quad 0x0000000000000000, 0x0000000000000000, 0x0000000000000000, 0x0000000000000000 // dZERO + .align 32 + .type __svml_datan2_data_internal, @object + .size __svml_datan2_data_internal, .-__svml_datan2_data_internal