diff options
Diffstat (limited to 'sysdeps')
| -rw-r--r-- | sysdeps/x86_64/fpu/multiarch/svml_s_hypotf16_core_avx512.S | 274 |
1 files changed, 136 insertions, 138 deletions
diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_hypotf16_core_avx512.S b/sysdeps/x86_64/fpu/multiarch/svml_s_hypotf16_core_avx512.S index db8bff6483..548936fe61 100644 --- a/sysdeps/x86_64/fpu/multiarch/svml_s_hypotf16_core_avx512.S +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_hypotf16_core_avx512.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 @@ -74,166 +74,164 @@ /* Offsets for data table __svml_shypot_data_internal */ -#define _sAbsMask 0 -#define _sHalf 64 -#define _iExpBound 128 +#define _sAbsMask 0 +#define _sHalf 64 +#define _iExpBound 128 #include <sysdep.h> - .text - .section .text.exex512,"ax",@progbits + .section .text.exex512, "ax", @progbits ENTRY(_ZGVeN16vv_hypotf_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 - vgetexpps {sae}, %zmm0, %zmm2 - vgetexpps {sae}, %zmm1, %zmm3 - vmovups _sHalf+__svml_shypot_data_internal(%rip), %zmm6 - vmaxps {sae}, %zmm3, %zmm2, %zmm4 - vmulps {rn-sae}, %zmm0, %zmm0, %zmm2 - vandps _sAbsMask+__svml_shypot_data_internal(%rip), %zmm4, %zmm5 - vfmadd231ps {rn-sae}, %zmm1, %zmm1, %zmm2 - vpcmpd $5, _iExpBound+__svml_shypot_data_internal(%rip), %zmm5, %k0 - vrsqrt14ps %zmm2, %zmm7 - kmovw %k0, %edx - vmulps {rn-sae}, %zmm7, %zmm2, %zmm9 - vmulps {rn-sae}, %zmm7, %zmm6, %zmm8 - vfnmadd231ps {rn-sae}, %zmm9, %zmm9, %zmm2 - vfmadd213ps {rn-sae}, %zmm9, %zmm8, %zmm2 - -/* - * VSCALEF( S, _VRES1, _VRES1, sExp ); - * 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 zmm1 zmm2 - -/* 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 $-64, %rsp + subq $256, %rsp + vgetexpps {sae}, %zmm0, %zmm2 + vgetexpps {sae}, %zmm1, %zmm3 + vmovups _sHalf+__svml_shypot_data_internal(%rip), %zmm6 + vmaxps {sae}, %zmm3, %zmm2, %zmm4 + vmulps {rn-sae}, %zmm0, %zmm0, %zmm2 + vandps _sAbsMask+__svml_shypot_data_internal(%rip), %zmm4, %zmm5 + vfmadd231ps {rn-sae}, %zmm1, %zmm1, %zmm2 + vpcmpd $5, _iExpBound+__svml_shypot_data_internal(%rip), %zmm5, %k0 + vrsqrt14ps %zmm2, %zmm7 + kmovw %k0, %edx + vmulps {rn-sae}, %zmm7, %zmm2, %zmm9 + vmulps {rn-sae}, %zmm7, %zmm6, %zmm8 + vfnmadd231ps {rn-sae}, %zmm9, %zmm9, %zmm2 + vfmadd213ps {rn-sae}, %zmm9, %zmm8, %zmm2 + + /* + * VSCALEF( S, _VRES1, _VRES1, sExp ); + * 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 zmm1 zmm2 + + /* Restore registers + * and exit the function + */ L(EXIT): - vmovaps %zmm2, %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 %zmm2, %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 %zmm1, 128(%rsp) - vmovups %zmm2, 192(%rsp) - # LOE rbx r12 r13 r14 r15 edx zmm2 - - 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 %zmm1, 128(%rsp) + vmovups %zmm2, 192(%rsp) + # LOE rbx r12 r13 r14 r15 edx zmm2 + + 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), %zmm2 - -/* 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 zmm2 - -/* 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), %zmm2 + + /* 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 zmm2 + + /* 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 hypotf@PLT - # LOE rbx r14 r15 r12d r13d xmm0 + movl %r12d, %r14d + movss 64(%rsp, %r14, 4), %xmm0 + movss 128(%rsp, %r14, 4), %xmm1 + call hypotf@PLT + # LOE rbx r14 r15 r12d r13d xmm0 - movss %xmm0, 192(%rsp,%r14,4) + movss %xmm0, 192(%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(_ZGVeN16vv_hypotf_skx) - .section .rodata, "a" - .align 64 + .section .rodata, "a" + .align 64 #ifdef __svml_shypot_data_internal_typedef typedef unsigned int VUINT32; -typedef struct -{ - __declspec(align(64)) VUINT32 _sAbsMask[16][1]; - __declspec(align(64)) VUINT32 _sHalf[16][1]; - __declspec(align(64)) VUINT32 _iExpBound[16][1]; +typedef struct { + __declspec(align(64)) VUINT32 _sAbsMask[16][1]; + __declspec(align(64)) VUINT32 _sHalf[16][1]; + __declspec(align(64)) VUINT32 _iExpBound[16][1]; } __svml_shypot_data_internal; #endif __svml_shypot_data_internal: - .long 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff /* _sAbsMask */ - .align 64 - .long 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000 /* _sHalf */ - /* fma based algorithm*/ - .align 64 - .long 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000 /* _iExpBound */ - .align 64 - .type __svml_shypot_data_internal,@object - .size __svml_shypot_data_internal,.-__svml_shypot_data_internal + .long 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff /* _sAbsMask */ + .align 64 + .long 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000 /* _sHalf */ + /* fma based algorithm*/ + .align 64 + .long 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000, 0x427C0000 /* _iExpBound */ + .align 64 + .type __svml_shypot_data_internal, @object + .size __svml_shypot_data_internal, .-__svml_shypot_data_internal |