x86-64: Add vector sinh/sinhf implementation to libmvec

Implement vectorized sinh/sinhf containing SSE, AVX, AVX2 and
AVX512 versions for libmvec as per vector ABI.  It also contains
accuracy and ABI tests for vector sinh/sinhf with regenerated ulps.

Reviewed-by: H.J. Lu <hjl.tools@gmail.com>

1 files changed, 309 insertions, 0 deletions

diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_sinhf8_core_avx2.S b/sysdeps/x86_64/fpu/multiarch/svml_s_sinhf8_core_avx2.S
new file mode 100644
index 0000000000..ea13fb60d4
--- /dev/null
+++ b/sysdeps/x86_64/fpu/multiarch/svml_s_sinhf8_core_avx2.S
@@ -0,0 +1,309 @@
+/* Function sinhf 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 sinh(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:
+ *
+ *   sinh(NaN) = quiet NaN, and raise invalid exception
+ *   sinh(INF) = that INF
+ *   sinh(x)   = x for subnormals
+ *   sinh(x) overflows for big x and returns MAXLOG+log(2)
+ *
+ */
+
+/* Offsets for data table __svml_ssinh_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_sinhf_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   _sInvLn2+__svml_ssinh_data_internal(%rip), %ymm7
+        vmovups   _sShifter+__svml_ssinh_data_internal(%rip), %ymm4
+        vmovups   _sLn2hi+__svml_ssinh_data_internal(%rip), %ymm5
+
+/*
+ * sinh(X) = sG2 + sG1*sinh(dR) + sG2*sR2*(a2+sR2*(a4+a6*sR2)
+ * sOut = (a4 +a6*sR2)
+ */
+        vmovups   _sPC6+__svml_ssinh_data_internal(%rip), %ymm14
+
+/*
+ * 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_ssinh_data_internal(%rip), %ymm12
+        vmovups   _iHalf+__svml_ssinh_data_internal(%rip), %ymm8
+        vmovaps   %ymm0, %ymm2
+
+/*
+ *  Implementation
+ *  Abs argument
+ */
+        vandps    _sSign+__svml_ssinh_data_internal(%rip), %ymm2, %ymm1
+        vxorps    %ymm2, %ymm1, %ymm0
+
+/*
+ *  Load argument
+ * dM = x/log(2) + RShifter
+ */
+        vfmadd213ps %ymm4, %ymm0, %ymm7
+
+/*
+ *  R
+ * sN = sM - RShifter
+ */
+        vsubps    %ymm4, %ymm7, %ymm6
+
+/*
+ *  G1,G2 2^N,2^(-N)
+ * iM now is an EXP(2^N)
+ */
+        vpslld    $23, %ymm7, %ymm9
+
+/*
+ * Check for overflow\underflow
+ * MORE faster than GE?
+ */
+        vpcmpgtd  _iDomainRange+__svml_ssinh_data_internal(%rip), %ymm0, %ymm3
+
+/* sR = sX - sN*Log2_hi */
+        vfnmadd231ps %ymm5, %ymm6, %ymm0
+        vpaddd    %ymm9, %ymm8, %ymm10
+        vpsubd    %ymm9, %ymm8, %ymm11
+
+/* sR = (sX - sN*Log2_hi) - sN*Log2_lo */
+        vfnmadd231ps _sLn2lo+__svml_ssinh_data_internal(%rip), %ymm6, %ymm0
+
+/* sR2 = sR^2 */
+        vmulps    %ymm0, %ymm0, %ymm13
+        vfmadd213ps _sPC4+__svml_ssinh_data_internal(%rip), %ymm13, %ymm14
+        vfmadd213ps _sPC3+__svml_ssinh_data_internal(%rip), %ymm13, %ymm12
+
+/* sOut = a2+sR2*(a4+a6*sR2) */
+        vfmadd213ps _sPC2+__svml_ssinh_data_internal(%rip), %ymm13, %ymm14
+
+/* sSinh_r = r^2*(a3+r^2*a5) */
+        vmulps    %ymm12, %ymm13, %ymm12
+
+/* sOut = sR2*(a2+sR2*(a4+a6*sR2) */
+        vmulps    %ymm14, %ymm13, %ymm15
+
+/* sSinh_r = r + r*(r^2*(a3+r^2*a5)) */
+        vfmadd213ps %ymm0, %ymm0, %ymm12
+        vmovmskps %ymm3, %edx
+
+/* sG1 = 2^(N-1)+2^(-N-1) */
+        vaddps    %ymm11, %ymm10, %ymm3
+
+/* sG2 = 2^(N-1)-2^(-N-1) */
+        vsubps    %ymm11, %ymm10, %ymm10
+
+/* sOut = sG2*sR2*(a2+sR2*(a4+a6*sR2) */
+        vmulps    %ymm15, %ymm10, %ymm0
+
+/* sOut = sG1*sinh(dR)+sG2*sR2*(a2+sR2*(a4+a6*sR2) */
+        vfmadd213ps %ymm0, %ymm12, %ymm3
+
+/* sOut = sG2 + sG1*sinh(dR) + sG2*sR2*(a2+sR2*(a4+a6*sR2) */
+        vaddps    %ymm3, %ymm10, %ymm4
+
+/*  Ret H  */
+        vorps     %ymm4, %ymm1, %ymm0
+        testl     %edx, %edx
+
+/* Go to special inputs processing branch */
+        jne       L(SPECIAL_VALUES_BRANCH)
+                                # LOE rbx r12 r13 r14 r15 edx ymm0 ymm2
+
+/* 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   %ymm2, 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      sinhf@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_sinhf_avx2)
+
+        .section .rodata, "a"
+        .align 32
+
+#ifdef __svml_ssinh_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_ssinh_data_internal;
+#endif
+__svml_ssinh_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_ssinh_data_internal,@object
+        .size	__svml_ssinh_data_internal,.-__svml_ssinh_data_internal