diff options
author | Sunil K Pandey <skpgkp2@gmail.com> | 2021-12-29 08:53:16 -0800 |
---|---|---|
committer | Sunil K Pandey <skpgkp2@gmail.com> | 2021-12-29 11:37:42 -0800 |
commit | ef7ea9c1327be11c54462cb539ea8854979a21f3 (patch) | |
tree | ddf578d007fe3fe3d15e42f63f82cc432123c47e /sysdeps/x86_64/fpu/multiarch/svml_s_coshf8_core_avx2.S | |
parent | 8b726453d560eef536c9ff730f5f5be05b4ad6a7 (diff) | |
download | glibc-ef7ea9c1327be11c54462cb539ea8854979a21f3.tar.gz glibc-ef7ea9c1327be11c54462cb539ea8854979a21f3.tar.xz glibc-ef7ea9c1327be11c54462cb539ea8854979a21f3.zip |
x86-64: Add vector cosh/coshf implementation to libmvec
Implement vectorized cosh/coshf containing SSE, AVX, AVX2 and AVX512 versions for libmvec as per vector ABI. It also contains accuracy and ABI tests for vector cosh/coshf with regenerated ulps. Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
Diffstat (limited to 'sysdeps/x86_64/fpu/multiarch/svml_s_coshf8_core_avx2.S')
-rw-r--r-- | sysdeps/x86_64/fpu/multiarch/svml_s_coshf8_core_avx2.S | 308 |
1 files changed, 308 insertions, 0 deletions
diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_coshf8_core_avx2.S b/sysdeps/x86_64/fpu/multiarch/svml_s_coshf8_core_avx2.S new file mode 100644 index 0000000000..9149061e7e --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_coshf8_core_avx2.S @@ -0,0 +1,308 @@ +/* Function coshf vectorized with AVX2. + Copyright (C) 2021 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + https://www.gnu.org/licenses/. */ + +/* + * ALGORITHM DESCRIPTION: + * + * Compute cosh(x) as (exp(x)+exp(-x))/2, + * where exp is calculated as + * exp(M*ln2 + ln2*(j/2^k) + r) = 2^M * 2^(j/2^k) * exp(r) + * + * Special cases: + * + * cosh(NaN) = quiet NaN, and raise invalid exception + * cosh(INF) = that INF + * cosh(0) = 1 + * cosh(x) overflows for big x and returns MAXLOG+log(2) + * + */ + +/* Offsets for data table __svml_scosh_data_internal + */ +#define _sInvLn2 0 +#define _sLn2hi 32 +#define _sLn2lo 64 +#define _sSign 96 +#define _sShifter 128 +#define _iDomainRange 160 +#define _sPC1 192 +#define _sPC2 224 +#define _sPC3 256 +#define _sPC4 288 +#define _sPC5 320 +#define _sPC6 352 +#define _iHalf 384 + +#include <sysdep.h> + + .text + .section .text.avx2,"ax",@progbits +ENTRY(_ZGVdN8v_coshf_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 + vmovups _sSign+__svml_scosh_data_internal(%rip), %ymm2 + vmovups _sShifter+__svml_scosh_data_internal(%rip), %ymm7 + +/* + * Load argument + * dM = x/log(2) + RShifter + */ + vmovups _sInvLn2+__svml_scosh_data_internal(%rip), %ymm10 + vmovups _sLn2hi+__svml_scosh_data_internal(%rip), %ymm8 + vmovups _iDomainRange+__svml_scosh_data_internal(%rip), %ymm3 + +/* + * sinh(r) = r*((a1=1)+r^2*(a3+r^2*(a5+{v1 r^2*a7})))) = r + r*(r^2*(a3+r^2*(a5+r^2*a7))) .... + * sSinh_r = (a3+r^2*a5) + */ + vmovups _sPC5+__svml_scosh_data_internal(%rip), %ymm15 + vmovups _iHalf+__svml_scosh_data_internal(%rip), %ymm11 + vmovaps %ymm0, %ymm1 + +/* + * Implementation + * Abs argument + */ + vandnps %ymm1, %ymm2, %ymm0 + vfmadd213ps %ymm7, %ymm0, %ymm10 + +/* + * R + * sN = sM - RShifter + */ + vsubps %ymm7, %ymm10, %ymm9 + +/* + * G1,G2 2^N,2^(-N) + * iM now is an EXP(2^N) + */ + vpslld $23, %ymm10, %ymm12 + +/* Check for overflow\underflow */ + vpcmpgtd %ymm3, %ymm0, %ymm4 + vpcmpeqd %ymm3, %ymm0, %ymm5 + +/* sR = sX - sN*Log2_hi */ + vfnmadd231ps %ymm8, %ymm9, %ymm0 + vpaddd %ymm12, %ymm11, %ymm13 + vpsubd %ymm12, %ymm11, %ymm14 + vpor %ymm5, %ymm4, %ymm6 + +/* sR = (sX - sN*Log2_hi) - sN*Log2_lo */ + vfnmadd231ps _sLn2lo+__svml_scosh_data_internal(%rip), %ymm9, %ymm0 + +/* sG1 = 2^(N-1)-2^(-N-1) */ + vsubps %ymm14, %ymm13, %ymm4 + +/* sG2 = 2^(N-1)+2^(-N-1) */ + vaddps %ymm14, %ymm13, %ymm3 + +/* sR2 = sR^2,shaffled */ + vmulps %ymm0, %ymm0, %ymm2 + vfmadd213ps _sPC3+__svml_scosh_data_internal(%rip), %ymm2, %ymm15 + +/* sSinh_r = r^2*(a3+r^2*a5) */ + vmulps %ymm15, %ymm2, %ymm13 + +/* sSinh_r = r + r*(r^2*(a3+r^2*a5)) */ + vfmadd213ps %ymm0, %ymm0, %ymm13 + +/* + * sinh(X) = sG2 + sG1*sinh(dR) + sG2*sR2*(a2+sR2*(a4+a6*sR2) + * sOut = (a4 +a6*sR2) + */ + vmovups _sPC6+__svml_scosh_data_internal(%rip), %ymm0 + vfmadd213ps _sPC4+__svml_scosh_data_internal(%rip), %ymm2, %ymm0 + +/* sOut = a2+sR2*(a4+a6*sR2) */ + vfmadd213ps _sPC2+__svml_scosh_data_internal(%rip), %ymm2, %ymm0 + +/* sOut = sR2*(a2+sR2*(a4+a6*sR2) */ + vmulps %ymm0, %ymm2, %ymm15 + +/* sOut = sG2*sR2*(a2+sR2*(a4+a6*sR2) */ + vmulps %ymm15, %ymm3, %ymm14 + +/* sOut = sG1*sinh(dR)+sG2*sR2*(a2+sR2*(a4+a6*sR2) */ + vfmadd213ps %ymm14, %ymm13, %ymm4 + vmovmskps %ymm6, %edx + +/* sOut = sG2 + sG1*sinh(dR) + sG2*sR2*(a2+sR2*(a4+a6*sR2) */ + vaddps %ymm4, %ymm3, %ymm0 + +/* Ret H */ + testl %edx, %edx + +/* Go to special inputs processing branch */ + jne L(SPECIAL_VALUES_BRANCH) + # LOE rbx r12 r13 r14 r15 edx ymm0 ymm1 + +/* 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) + +/* Branch to process + * special inputs + */ + +L(SPECIAL_VALUES_BRANCH): + vmovups %ymm1, 32(%rsp) + vmovups %ymm0, 64(%rsp) + # LOE rbx r12 r13 r14 r15 edx ymm0 + + 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: -80; DW_OP_plus) */ + .cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xb0, 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: -88; DW_OP_plus) */ + .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xa8, 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: -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 + */ + +L(RANGEMASK_CHECK): + btl %r12d, %r13d + +/* Call scalar math function */ + jc L(SCALAR_MATH_CALL) + # LOE rbx r15 r12d r13d + +/* Special inputs + * processing loop + */ + +L(SPECIAL_VALUES_LOOP): + incl %r12d + cmpl $8, %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 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 + +/* Scalar math fucntion call + * to process special input + */ + +L(SCALAR_MATH_CALL): + movl %r12d, %r14d + movss 32(%rsp,%r14,4), %xmm0 + call coshf@PLT + # LOE rbx r14 r15 r12d r13d xmm0 + + movss %xmm0, 64(%rsp,%r14,4) + +/* Process special inputs in loop */ + jmp L(SPECIAL_VALUES_LOOP) + # LOE rbx r15 r12d r13d +END(_ZGVdN8v_coshf_avx2) + + .section .rodata, "a" + .align 32 + +#ifdef __svml_scosh_data_internal_typedef +typedef unsigned int VUINT32; +typedef struct +{ + __declspec(align(32)) VUINT32 _sInvLn2[8][1]; + __declspec(align(32)) VUINT32 _sLn2hi[8][1]; + __declspec(align(32)) VUINT32 _sLn2lo[8][1]; + __declspec(align(32)) VUINT32 _sSign[8][1]; + __declspec(align(32)) VUINT32 _sShifter[8][1]; + __declspec(align(32)) VUINT32 _iDomainRange[8][1]; + __declspec(align(32)) VUINT32 _sPC1[8][1]; + __declspec(align(32)) VUINT32 _sPC2[8][1]; + __declspec(align(32)) VUINT32 _sPC3[8][1]; + __declspec(align(32)) VUINT32 _sPC4[8][1]; + __declspec(align(32)) VUINT32 _sPC5[8][1]; + __declspec(align(32)) VUINT32 _sPC6[8][1]; + __declspec(align(32)) VUINT32 _iHalf[8][1]; +} __svml_scosh_data_internal; +#endif +__svml_scosh_data_internal: + .long 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B /* _sInvLn2 */ //k=0 + .align 32 + .long 0x3F317000, 0x3F317000, 0x3F317000, 0x3F317000, 0x3F317000, 0x3F317000, 0x3F317000, 0x3F317000 /* _sLn2hi */ + .align 32 + .long 0x3805fdf4, 0x3805fdf4, 0x3805fdf4, 0x3805fdf4, 0x3805fdf4, 0x3805fdf4, 0x3805fdf4, 0x3805fdf4 /* _sLn2lo */ + .align 32 + .long 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000 /* _sSign */ + .align 32 + .long 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000 /* _sShifter */ + .align 32 + .long 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E /* _iDomainRange */ + .align 32 + .long 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000 /* _sPC1=1 */ + .align 32 + .long 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000 /* _sPC2 */ + .align 32 + .long 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57 /* _sPC3 */ + .align 32 + .long 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72 /* _sPC4 */ + .align 32 + .long 0x3c091461, 0x3c091461, 0x3c091461, 0x3c091461, 0x3c091461, 0x3c091461, 0x3c091461, 0x3c091461 /* _sPC5 */ + .align 32 + .long 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3 /* _sPC6 */ + // Integer constants + .align 32 + .long 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000 /* _iHalf*/ + .align 32 + .type __svml_scosh_data_internal,@object + .size __svml_scosh_data_internal,.-__svml_scosh_data_internal |