diff options
| -rw-r--r-- | sysdeps/x86_64/fpu/multiarch/svml_s_sinhf4_core_sse4.S | 490 |
1 files changed, 244 insertions, 246 deletions
diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_sinhf4_core_sse4.S b/sysdeps/x86_64/fpu/multiarch/svml_s_sinhf4_core_sse4.S index 0f924915c8..8a3b37f474 100644 --- a/sysdeps/x86_64/fpu/multiarch/svml_s_sinhf4_core_sse4.S +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_sinhf4_core_sse4.S @@ -34,275 +34,273 @@ /* Offsets for data table __svml_ssinh_data_internal */ -#define _sInvLn2 0 -#define _sLn2hi 16 -#define _sLn2lo 32 -#define _sSign 48 -#define _sShifter 64 -#define _iDomainRange 80 -#define _sPC1 96 -#define _sPC2 112 -#define _sPC3 128 -#define _sPC4 144 -#define _sPC5 160 -#define _sPC6 176 -#define _iHalf 192 +#define _sInvLn2 0 +#define _sLn2hi 16 +#define _sLn2lo 32 +#define _sSign 48 +#define _sShifter 64 +#define _iDomainRange 80 +#define _sPC1 96 +#define _sPC2 112 +#define _sPC3 128 +#define _sPC4 144 +#define _sPC5 160 +#define _sPC6 176 +#define _iHalf 192 #include <sysdep.h> - .text - .section .text.sse4,"ax",@progbits + .section .text.sse4, "ax", @progbits ENTRY(_ZGVbN4v_sinhf_sse4) - subq $72, %rsp - cfi_def_cfa_offset(80) - -/* - * Implementation - * Abs argument - */ - movups _sSign+__svml_ssinh_data_internal(%rip), %xmm14 - andps %xmm0, %xmm14 - movaps %xmm14, %xmm10 - -/* - * Load argument - * dM = x/log(2) + RShifter - */ - movups _sInvLn2+__svml_ssinh_data_internal(%rip), %xmm7 - pxor %xmm0, %xmm10 - mulps %xmm10, %xmm7 - -/* - * Check for overflow\underflow - * MORE faster than GE? - */ - movaps %xmm10, %xmm1 - movups _sShifter+__svml_ssinh_data_internal(%rip), %xmm2 - -/* sR = sX - sN*Log2_hi */ - movups _sLn2hi+__svml_ssinh_data_internal(%rip), %xmm3 - addps %xmm2, %xmm7 - -/* - * R - * sN = sM - RShifter - */ - movaps %xmm7, %xmm4 - -/* - * G1,G2 2^N,2^(-N) - * iM now is an EXP(2^N) - */ - pslld $23, %xmm7 - -/* sR = (sX - sN*Log2_hi) - sN*Log2_lo */ - movups _sLn2lo+__svml_ssinh_data_internal(%rip), %xmm5 - subps %xmm2, %xmm4 - mulps %xmm4, %xmm3 - mulps %xmm4, %xmm5 - subps %xmm3, %xmm10 - -/* - * 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) - */ - movups _sPC5+__svml_ssinh_data_internal(%rip), %xmm8 - subps %xmm5, %xmm10 - -/* sR2 = sR^2 */ - movaps %xmm10, %xmm12 - mulps %xmm10, %xmm12 - -/* - * sinh(X) = sG2 + sG1*sinh(dR) + sG2*sR2*(a2+sR2*(a4+a6*sR2) - * sOut = (a4 +a6*sR2) - */ - movups _sPC6+__svml_ssinh_data_internal(%rip), %xmm9 - mulps %xmm12, %xmm8 - mulps %xmm12, %xmm9 - addps _sPC3+__svml_ssinh_data_internal(%rip), %xmm8 - addps _sPC4+__svml_ssinh_data_internal(%rip), %xmm9 - -/* sSinh_r = r^2*(a3+r^2*a5) */ - mulps %xmm12, %xmm8 - -/* sOut = a2+sR2*(a4+a6*sR2) */ - mulps %xmm12, %xmm9 - -/* sSinh_r = r + r*(r^2*(a3+r^2*a5)) */ - mulps %xmm10, %xmm8 - addps _sPC2+__svml_ssinh_data_internal(%rip), %xmm9 - addps %xmm8, %xmm10 - -/* sOut = sR2*(a2+sR2*(a4+a6*sR2) */ - mulps %xmm9, %xmm12 - movdqu _iHalf+__svml_ssinh_data_internal(%rip), %xmm6 - movdqa %xmm6, %xmm13 - psubd %xmm7, %xmm6 - paddd %xmm7, %xmm13 - -/* sG1 = 2^(N-1)+2^(-N-1) */ - movdqa %xmm13, %xmm11 - -/* sG2 = 2^(N-1)-2^(-N-1) */ - subps %xmm6, %xmm13 - addps %xmm6, %xmm11 - -/* sOut = sG2*sR2*(a2+sR2*(a4+a6*sR2) */ - mulps %xmm13, %xmm12 - -/* sOut = sG1*sinh(dR)+sG2*sR2*(a2+sR2*(a4+a6*sR2) */ - mulps %xmm10, %xmm11 - pcmpgtd _iDomainRange+__svml_ssinh_data_internal(%rip), %xmm1 - addps %xmm11, %xmm12 - movmskps %xmm1, %edx - -/* sOut = sG2 + sG1*sinh(dR) + sG2*sR2*(a2+sR2*(a4+a6*sR2) */ - addps %xmm12, %xmm13 - -/* Ret H */ - orps %xmm13, %xmm14 - testl %edx, %edx - -/* Go to special inputs processing branch */ - jne L(SPECIAL_VALUES_BRANCH) - # LOE rbx rbp r12 r13 r14 r15 edx xmm0 xmm14 - -/* Restore registers - * and exit the function - */ + subq $72, %rsp + cfi_def_cfa_offset(80) + + /* + * Implementation + * Abs argument + */ + movups _sSign+__svml_ssinh_data_internal(%rip), %xmm14 + andps %xmm0, %xmm14 + movaps %xmm14, %xmm10 + + /* + * Load argument + * dM = x/log(2) + RShifter + */ + movups _sInvLn2+__svml_ssinh_data_internal(%rip), %xmm7 + pxor %xmm0, %xmm10 + mulps %xmm10, %xmm7 + + /* + * Check for overflow\underflow + * MORE faster than GE? + */ + movaps %xmm10, %xmm1 + movups _sShifter+__svml_ssinh_data_internal(%rip), %xmm2 + + /* sR = sX - sN*Log2_hi */ + movups _sLn2hi+__svml_ssinh_data_internal(%rip), %xmm3 + addps %xmm2, %xmm7 + + /* + * R + * sN = sM - RShifter + */ + movaps %xmm7, %xmm4 + + /* + * G1, G2 2^N, 2^(-N) + * iM now is an EXP(2^N) + */ + pslld $23, %xmm7 + + /* sR = (sX - sN*Log2_hi) - sN*Log2_lo */ + movups _sLn2lo+__svml_ssinh_data_internal(%rip), %xmm5 + subps %xmm2, %xmm4 + mulps %xmm4, %xmm3 + mulps %xmm4, %xmm5 + subps %xmm3, %xmm10 + + /* + * 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) + */ + movups _sPC5+__svml_ssinh_data_internal(%rip), %xmm8 + subps %xmm5, %xmm10 + + /* sR2 = sR^2 */ + movaps %xmm10, %xmm12 + mulps %xmm10, %xmm12 + + /* + * sinh(X) = sG2 + sG1*sinh(dR) + sG2*sR2*(a2+sR2*(a4+a6*sR2) + * sOut = (a4 +a6*sR2) + */ + movups _sPC6+__svml_ssinh_data_internal(%rip), %xmm9 + mulps %xmm12, %xmm8 + mulps %xmm12, %xmm9 + addps _sPC3+__svml_ssinh_data_internal(%rip), %xmm8 + addps _sPC4+__svml_ssinh_data_internal(%rip), %xmm9 + + /* sSinh_r = r^2*(a3+r^2*a5) */ + mulps %xmm12, %xmm8 + + /* sOut = a2+sR2*(a4+a6*sR2) */ + mulps %xmm12, %xmm9 + + /* sSinh_r = r + r*(r^2*(a3+r^2*a5)) */ + mulps %xmm10, %xmm8 + addps _sPC2+__svml_ssinh_data_internal(%rip), %xmm9 + addps %xmm8, %xmm10 + + /* sOut = sR2*(a2+sR2*(a4+a6*sR2) */ + mulps %xmm9, %xmm12 + movdqu _iHalf+__svml_ssinh_data_internal(%rip), %xmm6 + movdqa %xmm6, %xmm13 + psubd %xmm7, %xmm6 + paddd %xmm7, %xmm13 + + /* sG1 = 2^(N-1)+2^(-N-1) */ + movdqa %xmm13, %xmm11 + + /* sG2 = 2^(N-1)-2^(-N-1) */ + subps %xmm6, %xmm13 + addps %xmm6, %xmm11 + + /* sOut = sG2*sR2*(a2+sR2*(a4+a6*sR2) */ + mulps %xmm13, %xmm12 + + /* sOut = sG1*sinh(dR)+sG2*sR2*(a2+sR2*(a4+a6*sR2) */ + mulps %xmm10, %xmm11 + pcmpgtd _iDomainRange+__svml_ssinh_data_internal(%rip), %xmm1 + addps %xmm11, %xmm12 + movmskps %xmm1, %edx + + /* sOut = sG2 + sG1*sinh(dR) + sG2*sR2*(a2+sR2*(a4+a6*sR2) */ + addps %xmm12, %xmm13 + + /* Ret H */ + orps %xmm13, %xmm14 + testl %edx, %edx + + /* Go to special inputs processing branch */ + jne L(SPECIAL_VALUES_BRANCH) + # LOE rbx rbp r12 r13 r14 r15 edx xmm0 xmm14 + + /* Restore registers + * and exit the function + */ L(EXIT): - movaps %xmm14, %xmm0 - addq $72, %rsp - cfi_def_cfa_offset(8) - ret - cfi_def_cfa_offset(80) - -/* Branch to process - * special inputs - */ + movaps %xmm14, %xmm0 + addq $72, %rsp + cfi_def_cfa_offset(8) + ret + cfi_def_cfa_offset(80) + + /* Branch to process + * special inputs + */ L(SPECIAL_VALUES_BRANCH): - movups %xmm0, 32(%rsp) - movups %xmm14, 48(%rsp) - # LOE rbx rbp r12 r13 r14 r15 edx - - xorl %eax, %eax - movq %r12, 16(%rsp) - cfi_offset(12, -64) - movl %eax, %r12d - movq %r13, 8(%rsp) - cfi_offset(13, -72) - movl %edx, %r13d - movq %r14, (%rsp) - cfi_offset(14, -80) - # LOE rbx rbp r15 r12d r13d - -/* Range mask - * bits check - */ + movups %xmm0, 32(%rsp) + movups %xmm14, 48(%rsp) + # LOE rbx rbp r12 r13 r14 r15 edx + + xorl %eax, %eax + movq %r12, 16(%rsp) + cfi_offset(12, -64) + movl %eax, %r12d + movq %r13, 8(%rsp) + cfi_offset(13, -72) + movl %edx, %r13d + movq %r14, (%rsp) + cfi_offset(14, -80) + # LOE rbx rbp 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 rbp r15 r12d r13d + /* Call scalar math function */ + jc L(SCALAR_MATH_CALL) + # LOE rbx rbp r15 r12d r13d -/* Special inputs - * processing loop - */ + /* Special inputs + * processing loop + */ L(SPECIAL_VALUES_LOOP): - incl %r12d - cmpl $4, %r12d - -/* Check bits in range mask */ - jl L(RANGEMASK_CHECK) - # LOE rbx rbp r15 r12d r13d - - movq 16(%rsp), %r12 - cfi_restore(12) - movq 8(%rsp), %r13 - cfi_restore(13) - movq (%rsp), %r14 - cfi_restore(14) - movups 48(%rsp), %xmm14 - -/* Go to exit */ - jmp L(EXIT) - cfi_offset(12, -64) - cfi_offset(13, -72) - cfi_offset(14, -80) - # LOE rbx rbp r12 r13 r14 r15 xmm14 - -/* Scalar math fucntion call - * to process special input - */ + incl %r12d + cmpl $4, %r12d + + /* Check bits in range mask */ + jl L(RANGEMASK_CHECK) + # LOE rbx rbp r15 r12d r13d + + movq 16(%rsp), %r12 + cfi_restore(12) + movq 8(%rsp), %r13 + cfi_restore(13) + movq (%rsp), %r14 + cfi_restore(14) + movups 48(%rsp), %xmm14 + + /* Go to exit */ + jmp L(EXIT) + cfi_offset(12, -64) + cfi_offset(13, -72) + cfi_offset(14, -80) + # LOE rbx rbp r12 r13 r14 r15 xmm14 + + /* 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 rbp r14 r15 r12d r13d xmm0 + movl %r12d, %r14d + movss 32(%rsp, %r14, 4), %xmm0 + call sinhf@PLT + # LOE rbx rbp r14 r15 r12d r13d xmm0 - movss %xmm0, 48(%rsp,%r14,4) + movss %xmm0, 48(%rsp, %r14, 4) -/* Process special inputs in loop */ - jmp L(SPECIAL_VALUES_LOOP) - # LOE rbx rbp r15 r12d r13d + /* Process special inputs in loop */ + jmp L(SPECIAL_VALUES_LOOP) + # LOE rbx rbp r15 r12d r13d END(_ZGVbN4v_sinhf_sse4) - .section .rodata, "a" - .align 16 + .section .rodata, "a" + .align 16 #ifdef __svml_ssinh_data_internal_typedef typedef unsigned int VUINT32; -typedef struct -{ - __declspec(align(16)) VUINT32 _sInvLn2[4][1]; - __declspec(align(16)) VUINT32 _sLn2hi[4][1]; - __declspec(align(16)) VUINT32 _sLn2lo[4][1]; - __declspec(align(16)) VUINT32 _sSign[4][1]; - __declspec(align(16)) VUINT32 _sShifter[4][1]; - __declspec(align(16)) VUINT32 _iDomainRange[4][1]; - __declspec(align(16)) VUINT32 _sPC1[4][1]; - __declspec(align(16)) VUINT32 _sPC2[4][1]; - __declspec(align(16)) VUINT32 _sPC3[4][1]; - __declspec(align(16)) VUINT32 _sPC4[4][1]; - __declspec(align(16)) VUINT32 _sPC5[4][1]; - __declspec(align(16)) VUINT32 _sPC6[4][1]; - __declspec(align(16)) VUINT32 _iHalf[4][1]; +typedef struct { + __declspec(align(16)) VUINT32 _sInvLn2[4][1]; + __declspec(align(16)) VUINT32 _sLn2hi[4][1]; + __declspec(align(16)) VUINT32 _sLn2lo[4][1]; + __declspec(align(16)) VUINT32 _sSign[4][1]; + __declspec(align(16)) VUINT32 _sShifter[4][1]; + __declspec(align(16)) VUINT32 _iDomainRange[4][1]; + __declspec(align(16)) VUINT32 _sPC1[4][1]; + __declspec(align(16)) VUINT32 _sPC2[4][1]; + __declspec(align(16)) VUINT32 _sPC3[4][1]; + __declspec(align(16)) VUINT32 _sPC4[4][1]; + __declspec(align(16)) VUINT32 _sPC5[4][1]; + __declspec(align(16)) VUINT32 _sPC6[4][1]; + __declspec(align(16)) VUINT32 _iHalf[4][1]; } __svml_ssinh_data_internal; #endif __svml_ssinh_data_internal: - .long 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B /* _sInvLn2 */ //k=0 - .align 16 - .long 0x3F317000, 0x3F317000, 0x3F317000, 0x3F317000 /* _sLn2hi */ - .align 16 - .long 0x3805FDF4, 0x3805FDF4, 0x3805FDF4, 0x3805FDF4 /* _sLn2lo */ - .align 16 - .long 0x80000000, 0x80000000, 0x80000000, 0x80000000 /* _sSign */ - .align 16 - .long 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000 /* _sShifter */ - .align 16 - .long 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E /* _iDomainRange */ - .align 16 - .long 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000 /* _sPC1=1 */ - .align 16 - .long 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000 /* _sPC2 */ - .align 16 - .long 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57 /* _sPC3 */ - .align 16 - .long 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72 /* _sPC4 */ - .align 16 - .long 0x3c091461, 0x3c091461, 0x3c091461, 0x3c091461 /* _sPC5 */ - .align 16 - .long 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3 /* _sPC6 */ - // Integer constants - .align 16 - .long 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000 /* _iHalf*/ - .align 16 - .type __svml_ssinh_data_internal,@object - .size __svml_ssinh_data_internal,.-__svml_ssinh_data_internal + .long 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B /* _sInvLn2 */ // k=0 + .align 16 + .long 0x3F317000, 0x3F317000, 0x3F317000, 0x3F317000 /* _sLn2hi */ + .align 16 + .long 0x3805FDF4, 0x3805FDF4, 0x3805FDF4, 0x3805FDF4 /* _sLn2lo */ + .align 16 + .long 0x80000000, 0x80000000, 0x80000000, 0x80000000 /* _sSign */ + .align 16 + .long 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000 /* _sShifter */ + .align 16 + .long 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E /* _iDomainRange */ + .align 16 + .long 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000 /* _sPC1=1 */ + .align 16 + .long 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000 /* _sPC2 */ + .align 16 + .long 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57 /* _sPC3 */ + .align 16 + .long 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72 /* _sPC4 */ + .align 16 + .long 0x3c091461, 0x3c091461, 0x3c091461, 0x3c091461 /* _sPC5 */ + .align 16 + .long 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3 /* _sPC6 */ + // Integer constants + .align 16 + .long 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000 /* _iHalf */ + .align 16 + .type __svml_ssinh_data_internal, @object + .size __svml_ssinh_data_internal, .-__svml_ssinh_data_internal |