diff options
Diffstat (limited to 'sysdeps/x86_64/fpu/multiarch/svml_s_hypotf8_core_avx2.S')
-rw-r--r-- | sysdeps/x86_64/fpu/multiarch/svml_s_hypotf8_core_avx2.S | 334 |
1 files changed, 166 insertions, 168 deletions
diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_hypotf8_core_avx2.S b/sysdeps/x86_64/fpu/multiarch/svml_s_hypotf8_core_avx2.S index 620fb0290f..fc97828008 100644 --- a/sysdeps/x86_64/fpu/multiarch/svml_s_hypotf8_core_avx2.S +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_hypotf8_core_avx2.S @@ -63,7 +63,7 @@ * result = s - h * d * * EP version of the function can be implemented as y[i]=sqrt(a[i]^2+b[i]^2) - * with all intermediate operations done in target precision for i=1,..,n. + * with all intermediate operations done in target precision for i=1, .., n. * It can return result y[i]=0 in case a[i]^2 and b[i]^2 underflow in target * precision (for some i). It can return result y[i]=NAN in case * a[i]^2+b[i]^2 overflow in target precision, for some i. It can return @@ -74,196 +74,194 @@ /* Offsets for data table __svml_shypot_data_internal */ -#define _sHiLoMask 0 -#define _sAbsMask 32 -#define _sHalf 64 -#define _LowBoundary 96 -#define _HighBoundary 128 +#define _sHiLoMask 0 +#define _sAbsMask 32 +#define _sHalf 64 +#define _LowBoundary 96 +#define _HighBoundary 128 #include <sysdep.h> - .text - .section .text.avx2,"ax",@progbits + .section .text.avx2, "ax", @progbits ENTRY(_ZGVdN8vv_hypotf_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 - -/* - * Implementation - * Multiprecision branch for _HA_ only - * No multiprecision branch for _LA_ - * _z = _VARG1 * _VARG1 + _VARG2 * _VARG2 - */ - vmulps %ymm0, %ymm0, %ymm8 - -/* - * Variables - * Defines - * Constants loading - */ - vmovups _sHalf+__svml_shypot_data_internal(%rip), %ymm7 - -/* Check _z exponent to be withing borders [1E3 ; 60A] else goto Callout */ - vmovups _LowBoundary+__svml_shypot_data_internal(%rip), %ymm2 - vfmadd231ps %ymm1, %ymm1, %ymm8 - -/* _s0 ~ 1.0/sqrt(_z) */ - vrsqrtps %ymm8, %ymm6 - vpcmpgtd %ymm8, %ymm2, %ymm3 - -/* First iteration */ - vmulps %ymm8, %ymm6, %ymm9 - vmulps %ymm7, %ymm6, %ymm2 - vfnmadd231ps %ymm9, %ymm2, %ymm7 - vfmadd213ps %ymm9, %ymm7, %ymm9 - -/* Second iteration */ - vfmadd132ps %ymm7, %ymm2, %ymm2 - vpcmpgtd _HighBoundary+__svml_shypot_data_internal(%rip), %ymm8, %ymm4 - vpor %ymm4, %ymm3, %ymm5 - -/* Finish second iteration in native precision for _LA_ */ - vfmsub231ps %ymm9, %ymm9, %ymm8 - vmovmskps %ymm5, %edx - vfnmadd213ps %ymm9, %ymm8, %ymm2 - -/* 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 ymm1 ymm2 - -/* Restore registers - * and exit the function - */ + 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 + + /* + * Implementation + * Multiprecision branch for _HA_ only + * No multiprecision branch for _LA_ + * _z = _VARG1 * _VARG1 + _VARG2 * _VARG2 + */ + vmulps %ymm0, %ymm0, %ymm8 + + /* + * Variables + * Defines + * Constants loading + */ + vmovups _sHalf+__svml_shypot_data_internal(%rip), %ymm7 + + /* Check _z exponent to be withing borders [1E3 ; 60A] else goto Callout */ + vmovups _LowBoundary+__svml_shypot_data_internal(%rip), %ymm2 + vfmadd231ps %ymm1, %ymm1, %ymm8 + + /* _s0 ~ 1.0/sqrt(_z) */ + vrsqrtps %ymm8, %ymm6 + vpcmpgtd %ymm8, %ymm2, %ymm3 + + /* First iteration */ + vmulps %ymm8, %ymm6, %ymm9 + vmulps %ymm7, %ymm6, %ymm2 + vfnmadd231ps %ymm9, %ymm2, %ymm7 + vfmadd213ps %ymm9, %ymm7, %ymm9 + + /* Second iteration */ + vfmadd132ps %ymm7, %ymm2, %ymm2 + vpcmpgtd _HighBoundary+__svml_shypot_data_internal(%rip), %ymm8, %ymm4 + vpor %ymm4, %ymm3, %ymm5 + + /* Finish second iteration in native precision for _LA_ */ + vfmsub231ps %ymm9, %ymm9, %ymm8 + vmovmskps %ymm5, %edx + vfnmadd213ps %ymm9, %ymm8, %ymm2 + + /* 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 ymm1 ymm2 + + /* Restore registers + * and exit the function + */ L(EXIT): - vmovaps %ymm2, %ymm0 - 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 %ymm2, %ymm0 + 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 %ymm0, 32(%rsp) - vmovups %ymm1, 64(%rsp) - vmovups %ymm2, 96(%rsp) - # LOE rbx r12 r13 r14 r15 edx ymm2 - - 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 - -/* Range mask - * bits check - */ + vmovups %ymm0, 32(%rsp) + vmovups %ymm1, 64(%rsp) + vmovups %ymm2, 96(%rsp) + # LOE rbx r12 r13 r14 r15 edx ymm2 + + 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 + + /* 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 $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 96(%rsp), %ymm2 - -/* 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 ymm2 - -/* Scalar math fucntion call - * to process special input - */ + 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 96(%rsp), %ymm2 + + /* 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 ymm2 + + /* Scalar math fucntion call + * to process special input + */ L(SCALAR_MATH_CALL): - movl %r12d, %r14d - movss 32(%rsp,%r14,4), %xmm0 - movss 64(%rsp,%r14,4), %xmm1 - call hypotf@PLT - # LOE rbx r14 r15 r12d r13d xmm0 + movl %r12d, %r14d + movss 32(%rsp, %r14, 4), %xmm0 + movss 64(%rsp, %r14, 4), %xmm1 + call hypotf@PLT + # LOE rbx r14 r15 r12d r13d xmm0 - movss %xmm0, 96(%rsp,%r14,4) + movss %xmm0, 96(%rsp, %r14, 4) -/* 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(_ZGVdN8vv_hypotf_avx2) - .section .rodata, "a" - .align 32 + .section .rodata, "a" + .align 32 #ifdef __svml_shypot_data_internal_typedef typedef unsigned int VUINT32; -typedef struct -{ - __declspec(align(32)) VUINT32 _sHiLoMask[8][1]; - __declspec(align(32)) VUINT32 _sAbsMask[8][1]; - __declspec(align(32)) VUINT32 _sHalf[8][1]; - __declspec(align(32)) VUINT32 _LowBoundary[8][1]; - __declspec(align(32)) VUINT32 _HighBoundary[8][1]; +typedef struct { + __declspec(align(32)) VUINT32 _sHiLoMask[8][1]; + __declspec(align(32)) VUINT32 _sAbsMask[8][1]; + __declspec(align(32)) VUINT32 _sHalf[8][1]; + __declspec(align(32)) VUINT32 _LowBoundary[8][1]; + __declspec(align(32)) VUINT32 _HighBoundary[8][1]; } __svml_shypot_data_internal; #endif __svml_shypot_data_internal: - /* legacy algorithm */ - .long 0xFFF80000, 0xFFF80000, 0xFFF80000, 0xFFF80000, 0xFFF80000, 0xFFF80000, 0xFFF80000, 0xFFF80000 /* _sHiLoMask */ - .align 32 - .long 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff /* _sAbsMask */ - .align 32 - .long 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000 /* _sHalf */ - .align 32 - .long 0x1E300000, 0x1E300000, 0x1E300000, 0x1E300000, 0x1E300000, 0x1E300000, 0x1E300000, 0x1E300000 /* _LowBoundary */ - .align 32 - .long 0x60A00000, 0x60A00000, 0x60A00000, 0x60A00000, 0x60A00000, 0x60A00000, 0x60A00000, 0x60A00000 /* _HighBoundary */ - .align 32 - .type __svml_shypot_data_internal,@object - .size __svml_shypot_data_internal,.-__svml_shypot_data_internal + /* legacy algorithm */ + .long 0xFFF80000, 0xFFF80000, 0xFFF80000, 0xFFF80000, 0xFFF80000, 0xFFF80000, 0xFFF80000, 0xFFF80000 /* _sHiLoMask */ + .align 32 + .long 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff /* _sAbsMask */ + .align 32 + .long 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000 /* _sHalf */ + .align 32 + .long 0x1E300000, 0x1E300000, 0x1E300000, 0x1E300000, 0x1E300000, 0x1E300000, 0x1E300000, 0x1E300000 /* _LowBoundary */ + .align 32 + .long 0x60A00000, 0x60A00000, 0x60A00000, 0x60A00000, 0x60A00000, 0x60A00000, 0x60A00000, 0x60A00000 /* _HighBoundary */ + .align 32 + .type __svml_shypot_data_internal, @object + .size __svml_shypot_data_internal, .-__svml_shypot_data_internal |