From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: by sourceware.org (Postfix, from userid 7852) id 84CB53858438; Tue, 8 Mar 2022 05:45:51 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 84CB53858438 Content-Type: text/plain; charset="us-ascii" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit From: Sunil Pandey To: glibc-cvs@sourceware.org Subject: [glibc] x86_64: Fix svml_s_atan2f16_core_avx512.S code formatting X-Act-Checkin: glibc X-Git-Author: Sunil K Pandey X-Git-Refname: refs/heads/master X-Git-Oldrev: 4a3a3e0a038ac128b179f6464876343f0ad9404f X-Git-Newrev: 00dec7a2577a3286522189e2737a83bb2b1e2376 Message-Id: <20220308054551.84CB53858438@sourceware.org> Date: Tue, 8 Mar 2022 05:45:51 +0000 (GMT) X-BeenThere: glibc-cvs@sourceware.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Glibc-cvs mailing list List-Unsubscribe: , List-Archive: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 08 Mar 2022 05:45:51 -0000 https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=00dec7a2577a3286522189e2737a83bb2b1e2376 commit 00dec7a2577a3286522189e2737a83bb2b1e2376 Author: Sunil K Pandey Date: Mon Mar 7 10:47:10 2022 -0800 x86_64: Fix svml_s_atan2f16_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 Diff: --- .../fpu/multiarch/svml_s_atan2f16_core_avx512.S | 669 ++++++++++----------- 1 file changed, 334 insertions(+), 335 deletions(-) diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_atan2f16_core_avx512.S b/sysdeps/x86_64/fpu/multiarch/svml_s_atan2f16_core_avx512.S index a4e9a02b38..56aa5bb917 100644 --- a/sysdeps/x86_64/fpu/multiarch/svml_s_atan2f16_core_avx512.S +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_atan2f16_core_avx512.S @@ -30,370 +30,369 @@ /* Offsets for data table __svml_satan2_data_internal */ -#define sZERO 0 -#define sONE 64 -#define sSIGN_MASK 128 -#define sABS_MASK 192 -#define sPIO2 256 -#define sPI 320 -#define sPC8 384 -#define sPC7 448 -#define sPC6 512 -#define sPC5 576 -#define sPC4 640 -#define sPC3 704 -#define sPC2 768 -#define sPC1 832 -#define sPC0 896 -#define iCHK_WORK_SUB 960 -#define iCHK_WORK_CMP 1024 +#define sZERO 0 +#define sONE 64 +#define sSIGN_MASK 128 +#define sABS_MASK 192 +#define sPIO2 256 +#define sPI 320 +#define sPC8 384 +#define sPC7 448 +#define sPC6 512 +#define sPC5 576 +#define sPC4 640 +#define sPC3 704 +#define sPC2 768 +#define sPC1 832 +#define sPC0 896 +#define iCHK_WORK_SUB 960 +#define iCHK_WORK_CMP 1024 #include - .text - .section .text.exex512,"ax",@progbits + .section .text.exex512, "ax", @progbits ENTRY(_ZGVeN16vv_atan2f_skx) - pushq %rbp - cfi_def_cfa_offset(16) - movq %rsp, %rbp - cfi_def_cfa(6, 16) - cfi_offset(6, -16) - andq $-64, %rsp - subq $256, %rsp - xorl %edx, %edx - -/* - * #define NO_VECTOR_ZERO_ATAN2_ARGS - * Declarations - * Variables - * Constants - * The end of declarations - * Implementation - * Arguments signs - */ - vmovups sABS_MASK+__svml_satan2_data_internal(%rip), %zmm6 - vmovups sONE+__svml_satan2_data_internal(%rip), %zmm3 - -/* Testing on working interval. */ - vmovups iCHK_WORK_SUB+__svml_satan2_data_internal(%rip), %zmm9 - vmovups iCHK_WORK_CMP+__svml_satan2_data_internal(%rip), %zmm14 - -/* - * 1) If yx then a=-x, b=y, PIO2=Pi/2 - */ - vmovups sPIO2+__svml_satan2_data_internal(%rip), %zmm4 - vpternlogd $255, %zmm13, %zmm13, %zmm13 - vmovaps %zmm1, %zmm8 - vandps %zmm6, %zmm8, %zmm2 - vandps %zmm6, %zmm0, %zmm1 - vorps sSIGN_MASK+__svml_satan2_data_internal(%rip), %zmm2, %zmm5 - vpsubd %zmm9, %zmm2, %zmm10 - vpsubd %zmm9, %zmm1, %zmm12 - vxorps %zmm2, %zmm8, %zmm7 - vxorps %zmm1, %zmm0, %zmm6 - vcmpps $17, {sae}, %zmm2, %zmm1, %k1 - vpcmpgtd %zmm10, %zmm14, %k2 - vpcmpgtd %zmm12, %zmm14, %k3 - vmovups sPC6+__svml_satan2_data_internal(%rip), %zmm14 - vblendmps %zmm1, %zmm5, %zmm11{%k1} - vblendmps %zmm2, %zmm1, %zmm5{%k1} - vxorps %zmm4, %zmm4, %zmm4{%k1} - -/* - * Division a/b. - * Enabled when FMA is available and - * performance is better with NR iteration - */ - vrcp14ps %zmm5, %zmm15 - vfnmadd231ps {rn-sae}, %zmm5, %zmm15, %zmm3 - vfmadd213ps {rn-sae}, %zmm15, %zmm3, %zmm15 - vmulps {rn-sae}, %zmm15, %zmm11, %zmm3 - vfnmadd231ps {rn-sae}, %zmm5, %zmm3, %zmm11 - vfmadd213ps {rn-sae}, %zmm3, %zmm11, %zmm15 - vmovups sPC8+__svml_satan2_data_internal(%rip), %zmm11 - vpternlogd $255, %zmm3, %zmm3, %zmm3 - -/* Polynomial. */ - vmulps {rn-sae}, %zmm15, %zmm15, %zmm9 - vpandnd %zmm10, %zmm10, %zmm13{%k2} - vmulps {rn-sae}, %zmm9, %zmm9, %zmm10 - vfmadd231ps {rn-sae}, %zmm10, %zmm11, %zmm14 - vmovups sPC5+__svml_satan2_data_internal(%rip), %zmm11 - vpandnd %zmm12, %zmm12, %zmm3{%k3} - vpord %zmm3, %zmm13, %zmm3 - vmovups sPC4+__svml_satan2_data_internal(%rip), %zmm13 - vmovups sPC7+__svml_satan2_data_internal(%rip), %zmm12 - vptestmd %zmm3, %zmm3, %k0 - vfmadd213ps {rn-sae}, %zmm13, %zmm10, %zmm14 - vfmadd231ps {rn-sae}, %zmm10, %zmm12, %zmm11 - vmovups sPC3+__svml_satan2_data_internal(%rip), %zmm12 - vmovups sPC2+__svml_satan2_data_internal(%rip), %zmm13 - -/* Special branch for fast (vector) processing of zero arguments */ - kortestw %k0, %k0 - vfmadd213ps {rn-sae}, %zmm12, %zmm10, %zmm11 - vmovups sPC1+__svml_satan2_data_internal(%rip), %zmm12 - vfmadd213ps {rn-sae}, %zmm13, %zmm10, %zmm14 - vmovups sPC0+__svml_satan2_data_internal(%rip), %zmm13 - vfmadd213ps {rn-sae}, %zmm12, %zmm10, %zmm11 - vfmadd213ps {rn-sae}, %zmm13, %zmm10, %zmm14 - vfmadd213ps {rn-sae}, %zmm14, %zmm9, %zmm11 - -/* Reconstruction. */ - vfmadd213ps {rn-sae}, %zmm4, %zmm15, %zmm11 - -/* if x<0, sPI = Pi, else sPI =0 */ - vmovups __svml_satan2_data_internal(%rip), %zmm15 - vorps %zmm7, %zmm11, %zmm9 - vcmpps $18, {sae}, %zmm15, %zmm8, %k4 - vmovups sPI+__svml_satan2_data_internal(%rip), %zmm11 - vaddps {rn-sae}, %zmm11, %zmm9, %zmm9{%k4} - vorps %zmm6, %zmm9, %zmm10 - -/* Go to auxilary branch */ - jne L(AUX_BRANCH) - # LOE rbx r12 r13 r14 r15 edx zmm0 zmm1 zmm2 zmm3 zmm4 zmm5 zmm6 zmm7 zmm8 zmm10 zmm11 - -/* Return from auxilary branch - * for out of main path inputs - */ + pushq %rbp + cfi_def_cfa_offset(16) + movq %rsp, %rbp + cfi_def_cfa(6, 16) + cfi_offset(6, -16) + andq $-64, %rsp + subq $256, %rsp + xorl %edx, %edx + + /* + * #define NO_VECTOR_ZERO_ATAN2_ARGS + * Declarations + * Variables + * Constants + * The end of declarations + * Implementation + * Arguments signs + */ + vmovups sABS_MASK+__svml_satan2_data_internal(%rip), %zmm6 + vmovups sONE+__svml_satan2_data_internal(%rip), %zmm3 + + /* Testing on working interval. */ + vmovups iCHK_WORK_SUB+__svml_satan2_data_internal(%rip), %zmm9 + vmovups iCHK_WORK_CMP+__svml_satan2_data_internal(%rip), %zmm14 + + /* + * 1) If yx then a=-x, b=y, PIO2=Pi/2 + */ + vmovups sPIO2+__svml_satan2_data_internal(%rip), %zmm4 + vpternlogd $255, %zmm13, %zmm13, %zmm13 + vmovaps %zmm1, %zmm8 + vandps %zmm6, %zmm8, %zmm2 + vandps %zmm6, %zmm0, %zmm1 + vorps sSIGN_MASK+__svml_satan2_data_internal(%rip), %zmm2, %zmm5 + vpsubd %zmm9, %zmm2, %zmm10 + vpsubd %zmm9, %zmm1, %zmm12 + vxorps %zmm2, %zmm8, %zmm7 + vxorps %zmm1, %zmm0, %zmm6 + vcmpps $17, {sae}, %zmm2, %zmm1, %k1 + vpcmpgtd %zmm10, %zmm14, %k2 + vpcmpgtd %zmm12, %zmm14, %k3 + vmovups sPC6+__svml_satan2_data_internal(%rip), %zmm14 + vblendmps %zmm1, %zmm5, %zmm11{%k1} + vblendmps %zmm2, %zmm1, %zmm5{%k1} + vxorps %zmm4, %zmm4, %zmm4{%k1} + + /* + * Division a/b. + * Enabled when FMA is available and + * performance is better with NR iteration + */ + vrcp14ps %zmm5, %zmm15 + vfnmadd231ps {rn-sae}, %zmm5, %zmm15, %zmm3 + vfmadd213ps {rn-sae}, %zmm15, %zmm3, %zmm15 + vmulps {rn-sae}, %zmm15, %zmm11, %zmm3 + vfnmadd231ps {rn-sae}, %zmm5, %zmm3, %zmm11 + vfmadd213ps {rn-sae}, %zmm3, %zmm11, %zmm15 + vmovups sPC8+__svml_satan2_data_internal(%rip), %zmm11 + vpternlogd $255, %zmm3, %zmm3, %zmm3 + + /* Polynomial. */ + vmulps {rn-sae}, %zmm15, %zmm15, %zmm9 + vpandnd %zmm10, %zmm10, %zmm13{%k2} + vmulps {rn-sae}, %zmm9, %zmm9, %zmm10 + vfmadd231ps {rn-sae}, %zmm10, %zmm11, %zmm14 + vmovups sPC5+__svml_satan2_data_internal(%rip), %zmm11 + vpandnd %zmm12, %zmm12, %zmm3{%k3} + vpord %zmm3, %zmm13, %zmm3 + vmovups sPC4+__svml_satan2_data_internal(%rip), %zmm13 + vmovups sPC7+__svml_satan2_data_internal(%rip), %zmm12 + vptestmd %zmm3, %zmm3, %k0 + vfmadd213ps {rn-sae}, %zmm13, %zmm10, %zmm14 + vfmadd231ps {rn-sae}, %zmm10, %zmm12, %zmm11 + vmovups sPC3+__svml_satan2_data_internal(%rip), %zmm12 + vmovups sPC2+__svml_satan2_data_internal(%rip), %zmm13 + + /* Special branch for fast (vector) processing of zero arguments */ + kortestw %k0, %k0 + vfmadd213ps {rn-sae}, %zmm12, %zmm10, %zmm11 + vmovups sPC1+__svml_satan2_data_internal(%rip), %zmm12 + vfmadd213ps {rn-sae}, %zmm13, %zmm10, %zmm14 + vmovups sPC0+__svml_satan2_data_internal(%rip), %zmm13 + vfmadd213ps {rn-sae}, %zmm12, %zmm10, %zmm11 + vfmadd213ps {rn-sae}, %zmm13, %zmm10, %zmm14 + vfmadd213ps {rn-sae}, %zmm14, %zmm9, %zmm11 + + /* Reconstruction. */ + vfmadd213ps {rn-sae}, %zmm4, %zmm15, %zmm11 + + /* if x<0, sPI = Pi, else sPI =0 */ + vmovups __svml_satan2_data_internal(%rip), %zmm15 + vorps %zmm7, %zmm11, %zmm9 + vcmpps $18, {sae}, %zmm15, %zmm8, %k4 + vmovups sPI+__svml_satan2_data_internal(%rip), %zmm11 + vaddps {rn-sae}, %zmm11, %zmm9, %zmm9{%k4} + vorps %zmm6, %zmm9, %zmm10 + + /* Go to auxilary branch */ + jne L(AUX_BRANCH) + # LOE rbx r12 r13 r14 r15 edx zmm0 zmm1 zmm2 zmm3 zmm4 zmm5 zmm6 zmm7 zmm8 zmm10 zmm11 + + /* 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 zmm0 zmm8 zmm10 + /* Go to special inputs processing branch */ + jne L(SPECIAL_VALUES_BRANCH) + # LOE rbx r12 r13 r14 r15 edx zmm0 zmm8 zmm10 -/* Restore registers - * and exit the function - */ + /* Restore registers + * and exit the function + */ L(EXIT): - vmovaps %zmm10, %zmm0 - 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 - */ + vmovaps %zmm10, %zmm0 + 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 + */ L(SPECIAL_VALUES_BRANCH): - vmovups %zmm0, 64(%rsp) - vmovups %zmm8, 128(%rsp) - vmovups %zmm10, 192(%rsp) - # LOE rbx r12 r13 r14 r15 edx zmm10 - - 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: -240; DW_OP_plus) */ - .cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x10, 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: -248; DW_OP_plus) */ - .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x08, 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: -256; DW_OP_plus) */ - .cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x00, 0xff, 0xff, 0xff, 0x22 - # LOE rbx r15 r12d r13d - -/* Range mask - * bits check - */ + vmovups %zmm0, 64(%rsp) + vmovups %zmm8, 128(%rsp) + vmovups %zmm10, 192(%rsp) + # LOE rbx r12 r13 r14 r15 edx zmm10 + + 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: -240; DW_OP_plus) */ + .cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x10, 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: -248; DW_OP_plus) */ + .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x08, 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: -256; DW_OP_plus) */ + .cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x00, 0xff, 0xff, 0xff, 0x22 + # LOE rbx r15 r12d r13d + + /* 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 $16, %r12d - -/* 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 192(%rsp), %zmm10 - -/* 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: -240; DW_OP_plus) */ - .cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x10, 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: -248; DW_OP_plus) */ - .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x08, 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: -256; DW_OP_plus) */ - .cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x00, 0xff, 0xff, 0xff, 0x22 - # LOE rbx r12 r13 r14 r15 zmm10 - -/* Scalar math fucntion call - * to process special input - */ + incl %r12d + cmpl $16, %r12d + + /* 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 192(%rsp), %zmm10 + + /* 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: -240; DW_OP_plus) */ + .cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x10, 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: -248; DW_OP_plus) */ + .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x08, 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: -256; DW_OP_plus) */ + .cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x00, 0xff, 0xff, 0xff, 0x22 + # LOE rbx r12 r13 r14 r15 zmm10 + + /* Scalar math fucntion call + * to process special input + */ L(SCALAR_MATH_CALL): - movl %r12d, %r14d - movss 64(%rsp,%r14,4), %xmm0 - movss 128(%rsp,%r14,4), %xmm1 - call atan2f@PLT - # LOE rbx r14 r15 r12d r13d xmm0 - - movss %xmm0, 192(%rsp,%r14,4) - -/* 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 - */ + movl %r12d, %r14d + movss 64(%rsp, %r14, 4), %xmm0 + movss 128(%rsp, %r14, 4), %xmm1 + call atan2f@PLT + # LOE rbx r14 r15 r12d r13d xmm0 -L(AUX_BRANCH): -/* Check if at least on of Y or Y is zero: iAXAYZERO */ - vmovups __svml_satan2_data_internal(%rip), %zmm9 + movss %xmm0, 192(%rsp, %r14, 4) -/* Check if both X & Y are not NaNs: iXYnotNAN */ - vcmpps $3, {sae}, %zmm8, %zmm8, %k1 - vcmpps $3, {sae}, %zmm0, %zmm0, %k2 - vpcmpd $4, %zmm9, %zmm2, %k3 - vpcmpd $4, %zmm9, %zmm1, %k4 + /* Process special inputs in loop */ + jmp L(SPECIAL_VALUES_LOOP) + cfi_restore(12) + cfi_restore(13) + cfi_restore(14) + # LOE rbx r15 r12d r13d -/* - * Path for zero arguments (at least one of both) - * Check if both args are zeros (den. is zero) - */ - vcmpps $4, {sae}, %zmm9, %zmm5, %k5 - -/* Res = sign(Y)*(X<0)?(PIO2+PI):PIO2 */ - vpcmpgtd %zmm8, %zmm9, %k6 - vpternlogd $255, %zmm14, %zmm14, %zmm14 - vpternlogd $255, %zmm12, %zmm12, %zmm12 - vpternlogd $255, %zmm13, %zmm13, %zmm13 - vpandnd %zmm2, %zmm2, %zmm14{%k3} - vpternlogd $255, %zmm2, %zmm2, %zmm2 - vpandnd %zmm1, %zmm1, %zmm2{%k4} - vpord %zmm2, %zmm14, %zmm15 - vpternlogd $255, %zmm2, %zmm2, %zmm2 - vpandnd %zmm5, %zmm5, %zmm2{%k5} - -/* Set sPIO2 to zero if den. is zero */ - vpandnd %zmm4, %zmm2, %zmm4 - vpandd %zmm2, %zmm9, %zmm5 - vpord %zmm5, %zmm4, %zmm2 - vorps %zmm7, %zmm2, %zmm7 - vaddps {rn-sae}, %zmm11, %zmm7, %zmm7{%k6} - vorps %zmm6, %zmm7, %zmm6 - vpandnd %zmm8, %zmm8, %zmm12{%k1} - vpandnd %zmm0, %zmm0, %zmm13{%k2} - vandps %zmm13, %zmm12, %zmm12 - -/* Check if at least on of Y or Y is zero and not NaN: iAXAYZEROnotNAN */ - vpandd %zmm12, %zmm15, %zmm1 - -/* Exclude from previous callout mask zero (and not NaN) arguments */ - vpandnd %zmm3, %zmm1, %zmm3 - -/* Go to callout */ - vptestmd %zmm3, %zmm3, %k0 - kmovw %k0, %edx - -/* Merge results from main and spec path */ - vpandnd %zmm10, %zmm1, %zmm10 - vpandd %zmm1, %zmm6, %zmm11 - vpord %zmm11, %zmm10, %zmm10 - -/* Return to main vector processing path */ - jmp L(AUX_BRANCH_RETURN) - # LOE rbx r12 r13 r14 r15 edx zmm0 zmm8 zmm10 + /* Auxilary branch + * for out of main path inputs + */ + +L(AUX_BRANCH): + /* Check if at least on of Y or Y is zero: iAXAYZERO */ + vmovups __svml_satan2_data_internal(%rip), %zmm9 + + /* Check if both X & Y are not NaNs: iXYnotNAN */ + vcmpps $3, {sae}, %zmm8, %zmm8, %k1 + vcmpps $3, {sae}, %zmm0, %zmm0, %k2 + vpcmpd $4, %zmm9, %zmm2, %k3 + vpcmpd $4, %zmm9, %zmm1, %k4 + + /* + * Path for zero arguments (at least one of both) + * Check if both args are zeros (den. is zero) + */ + vcmpps $4, {sae}, %zmm9, %zmm5, %k5 + + /* Res = sign(Y)*(X<0)?(PIO2+PI):PIO2 */ + vpcmpgtd %zmm8, %zmm9, %k6 + vpternlogd $255, %zmm14, %zmm14, %zmm14 + vpternlogd $255, %zmm12, %zmm12, %zmm12 + vpternlogd $255, %zmm13, %zmm13, %zmm13 + vpandnd %zmm2, %zmm2, %zmm14{%k3} + vpternlogd $255, %zmm2, %zmm2, %zmm2 + vpandnd %zmm1, %zmm1, %zmm2{%k4} + vpord %zmm2, %zmm14, %zmm15 + vpternlogd $255, %zmm2, %zmm2, %zmm2 + vpandnd %zmm5, %zmm5, %zmm2{%k5} + + /* Set sPIO2 to zero if den. is zero */ + vpandnd %zmm4, %zmm2, %zmm4 + vpandd %zmm2, %zmm9, %zmm5 + vpord %zmm5, %zmm4, %zmm2 + vorps %zmm7, %zmm2, %zmm7 + vaddps {rn-sae}, %zmm11, %zmm7, %zmm7{%k6} + vorps %zmm6, %zmm7, %zmm6 + vpandnd %zmm8, %zmm8, %zmm12{%k1} + vpandnd %zmm0, %zmm0, %zmm13{%k2} + vandps %zmm13, %zmm12, %zmm12 + + /* Check if at least on of Y or Y is zero and not NaN: iAXAYZEROnotNAN */ + vpandd %zmm12, %zmm15, %zmm1 + + /* Exclude from previous callout mask zero (and not NaN) arguments */ + vpandnd %zmm3, %zmm1, %zmm3 + + /* Go to callout */ + vptestmd %zmm3, %zmm3, %k0 + kmovw %k0, %edx + + /* Merge results from main and spec path */ + vpandnd %zmm10, %zmm1, %zmm10 + vpandd %zmm1, %zmm6, %zmm11 + vpord %zmm11, %zmm10, %zmm10 + + /* Return to main vector processing path */ + jmp L(AUX_BRANCH_RETURN) + # LOE rbx r12 r13 r14 r15 edx zmm0 zmm8 zmm10 END(_ZGVeN16vv_atan2f_skx) - .section .rodata, "a" - .align 64 + .section .rodata, "a" + .align 64 #ifdef __svml_satan2_data_internal_typedef typedef unsigned int VUINT32; typedef struct { - __declspec(align(64)) VUINT32 sZERO[16][1]; - __declspec(align(64)) VUINT32 sONE[16][1]; - __declspec(align(64)) VUINT32 sSIGN_MASK[16][1]; - __declspec(align(64)) VUINT32 sABS_MASK[16][1]; - __declspec(align(64)) VUINT32 sPIO2[16][1]; - __declspec(align(64)) VUINT32 sPI[16][1]; - __declspec(align(64)) VUINT32 sPC8[16][1]; - __declspec(align(64)) VUINT32 sPC7[16][1]; - __declspec(align(64)) VUINT32 sPC6[16][1]; - __declspec(align(64)) VUINT32 sPC5[16][1]; - __declspec(align(64)) VUINT32 sPC4[16][1]; - __declspec(align(64)) VUINT32 sPC3[16][1]; - __declspec(align(64)) VUINT32 sPC2[16][1]; - __declspec(align(64)) VUINT32 sPC1[16][1]; - __declspec(align(64)) VUINT32 sPC0[16][1]; - __declspec(align(64)) VUINT32 iCHK_WORK_SUB[16][1]; - __declspec(align(64)) VUINT32 iCHK_WORK_CMP[16][1]; + __declspec(align(64)) VUINT32 sZERO[16][1]; + __declspec(align(64)) VUINT32 sONE[16][1]; + __declspec(align(64)) VUINT32 sSIGN_MASK[16][1]; + __declspec(align(64)) VUINT32 sABS_MASK[16][1]; + __declspec(align(64)) VUINT32 sPIO2[16][1]; + __declspec(align(64)) VUINT32 sPI[16][1]; + __declspec(align(64)) VUINT32 sPC8[16][1]; + __declspec(align(64)) VUINT32 sPC7[16][1]; + __declspec(align(64)) VUINT32 sPC6[16][1]; + __declspec(align(64)) VUINT32 sPC5[16][1]; + __declspec(align(64)) VUINT32 sPC4[16][1]; + __declspec(align(64)) VUINT32 sPC3[16][1]; + __declspec(align(64)) VUINT32 sPC2[16][1]; + __declspec(align(64)) VUINT32 sPC1[16][1]; + __declspec(align(64)) VUINT32 sPC0[16][1]; + __declspec(align(64)) VUINT32 iCHK_WORK_SUB[16][1]; + __declspec(align(64)) VUINT32 iCHK_WORK_CMP[16][1]; } __svml_satan2_data_internal; #endif __svml_satan2_data_internal: - .long 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000 // sZERO - .align 64 - .long 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000 // sONE - .align 64 - .long 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000 // sSIGN_MASK - .align 64 - .long 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF // sABS_MASK - .align 64 - .long 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB // sPIO2 - .align 64 - .long 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB // sPI - .align 64 - .long 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0 // sA08 - .align 64 - .long 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631 // sA07 - .align 64 - .long 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384 // sA06 - .align 64 - .long 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629 // sA05 - .align 64 - .long 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474 // sA04 - .align 64 - .long 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8 // sA03 - .align 64 - .long 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F // sA02 - .align 64 - .long 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49 // sA01 - .align 64 - .long 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000 // sA00 - .align 64 - .long 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000 //iCHK_WORK_SUB - .align 64 - .long 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000 //iCHK_WORK_CMP - .align 64 - .type __svml_satan2_data_internal,@object - .size __svml_satan2_data_internal,.-__svml_satan2_data_internal + .long 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000 // sZERO + .align 64 + .long 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000 // sONE + .align 64 + .long 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000 // sSIGN_MASK + .align 64 + .long 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF // sABS_MASK + .align 64 + .long 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB // sPIO2 + .align 64 + .long 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB, 0x40490FDB // sPI + .align 64 + .long 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0, 0x3B322CC0 // sA08 + .align 64 + .long 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631, 0xBC7F2631 // sA07 + .align 64 + .long 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384, 0x3D2BC384 // sA06 + .align 64 + .long 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629, 0xBD987629 // sA05 + .align 64 + .long 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474, 0x3DD96474 // sA04 + .align 64 + .long 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8, 0xBE1161F8 // sA03 + .align 64 + .long 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F, 0x3E4CB79F // sA02 + .align 64 + .long 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49, 0xBEAAAA49 // sA01 + .align 64 + .long 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000 // sA00 + .align 64 + .long 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000, 0x81000000 // iCHK_WORK_SUB + .align 64 + .long 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000, 0xFC000000 // iCHK_WORK_CMP + .align 64 + .type __svml_satan2_data_internal, @object + .size __svml_satan2_data_internal, .-__svml_satan2_data_internal