; ; MIT License ; ----------- ; ; Copyright (c) 2002-2019 Advanced Micro Devices, Inc. ; ; Permission is hereby granted, free of charge, to any person obtaining a copy ; of this Software and associated documentaon files (the "Software"), to deal ; in the Software without restriction, including without limitation the rights ; to use, copy, modify, merge, publish, distribute, sublicense, and/or sell ; copies of the Software, and to permit persons to whom the Software is ; furnished to do so, subject to the following conditions: ; ; The above copyright notice and this permission notice shall be included in ; all copies or substantial portions of the Software. ; ; THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR ; IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, ; FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE ; AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER ; LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, ; OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN ; THE SOFTWARE. ; ; log10.asm ; ; An implementation of the log10 libm function. ; ; Prototype: ; ; double log10(double x); ; ; ; Algorithm: ; Similar to one presnted in log.asm ; .const ALIGN 16 __real_ninf DQ 0fff0000000000000h ; -inf DQ 0000000000000000h __real_inf DQ 7ff0000000000000h ; +inf DQ 0000000000000000h __real_neg_qnan DQ 0fff8000000000000h ; neg qNaN DQ 0000000000000000h __real_qnanbit DQ 0008000000000000h DQ 0000000000000000h __int_1023 DQ 00000000000003ffh DQ 0000000000000000h __mask_001 DQ 0000000000000001h DQ 0000000000000000h __mask_mant DQ 000FFFFFFFFFFFFFh ; mask for mantissa bits DQ 0000000000000000h __mask_mant_top8 DQ 000ff00000000000h ; mask for top 8 mantissa bits DQ 0000000000000000h __mask_mant9 DQ 0000080000000000h ; mask for 9th mantissa bit DQ 0000000000000000h __real_log10_e DQ 3fdbcb7b1526e50eh DQ 0000000000000000h __real_log10_e_lead DQ 3fdbcb7800000000h ; log10e_lead 4.34293746948242187500e-01 DQ 0000000000000000h __real_log10_e_tail DQ 3ea8a93728719535h ; log10e_tail 7.3495500964015109100644e-7 DQ 0000000000000000h __real_log10_2_lead DQ 3fd3441350000000h DQ 0000000000000000h __real_log10_2_tail DQ 3e03ef3fde623e25h DQ 0000000000000000h __real_two DQ 4000000000000000h ; 2 DQ 0000000000000000h __real_one DQ 3ff0000000000000h ; 1 DQ 0000000000000000h __real_half DQ 3fe0000000000000h ; 1/2 DQ 0000000000000000h __mask_100 DQ 0000000000000100h DQ 0000000000000000h __real_1_over_512 DQ 3f60000000000000h DQ 0000000000000000h __real_1_over_2 DQ 3fe0000000000000h DQ 0000000000000000h __real_1_over_3 DQ 3fd5555555555555h DQ 0000000000000000h __real_1_over_4 DQ 3fd0000000000000h DQ 0000000000000000h __real_1_over_5 DQ 3fc999999999999ah DQ 0000000000000000h __real_1_over_6 DQ 3fc5555555555555h DQ 0000000000000000h __real_neg_1023 DQ 0c08ff80000000000h DQ 0000000000000000h __mask_2045 DQ 00000000000007fdh DQ 0000000000000000h __real_threshold DQ 3fb0000000000000h ; .0625 DQ 0000000000000000h __real_near_one_lt DQ 3fee000000000000h ; .9375 DQ 0000000000000000h __real_near_one_gt DQ 3ff1000000000000h ; 1.0625 DQ 0000000000000000h __real_min_norm DQ 0010000000000000h DQ 0000000000000000h __real_notsign DQ 7ffFFFFFFFFFFFFFh ; ^sign bit DQ 0000000000000000h __real_ca1 DQ 3fb55555555554e6h ; 8.33333333333317923934e-02 DQ 0000000000000000h __real_ca2 DQ 3f89999999bac6d4h ; 1.25000000037717509602e-02 DQ 0000000000000000h __real_ca3 DQ 3f62492307f1519fh ; 2.23213998791944806202e-03 DQ 0000000000000000h __real_ca4 DQ 3f3c8034c85dfff0h ; 4.34887777707614552256e-04 DQ 0000000000000000h __mask_lower DQ 0ffffffff00000000h DQ 0000000000000000h ; these codes and the ones in the corresponding .c file have to match __flag_x_zero DD 00000001 __flag_x_neg DD 00000002 __flag_x_nan DD 00000003 EXTRN __log10_256_lead:QWORD EXTRN __log10_256_tail:QWORD EXTRN __log_F_inv_qword:QWORD EXTRN __use_fma3_lib:DWORD ; local variable storage offsets save_xmm6 EQU 20h dummy_space EQU 30h stack_size EQU 058h include fm.inc fname TEXTEQU fname_special TEXTEQU <_log10_special> EXTERN fname_special:PROC .code ALIGN 16 PUBLIC fname fname PROC FRAME StackAllocate stack_size SaveXmm xmm6, save_xmm6 .ENDPROLOG cmp DWORD PTR __use_fma3_lib, 0 jne Llog10_fma3 Llog10_sse2: ; compute exponent part movapd xmm3, xmm0 movapd xmm4, xmm0 psrlq xmm3, 52 movd rax, xmm0 psubq xmm3, XMMWORD PTR __int_1023 ; xmm3 <-- unbiased exponent ; NaN or inf movapd xmm5, xmm0 andpd xmm5, XMMWORD PTR __real_inf comisd xmm5, QWORD PTR __real_inf je Llog10_sse2_x_is_inf_or_nan movapd xmm2, xmm0 cvtdq2pd xmm6, xmm3 ; xmm6 <-- unbiased exp as double pand xmm2, XMMWORD PTR __mask_mant subsd xmm4, QWORD PTR __real_one comisd xmm6, QWORD PTR __real_neg_1023 je Llog10_sse2_denormal_adjust Llog10_sse2_continue_common: andpd xmm4, XMMWORD PTR __real_notsign ; compute index into the log tables mov r9, rax and rax, QWORD PTR __mask_mant_top8 and r9, QWORD PTR __mask_mant9 shl r9, 1 add rax, r9 movd xmm1, rax ; near one codepath comisd xmm4, QWORD PTR __real_threshold jb Llog10_sse2_near_one ; F, Y shr rax, 44 por xmm2, XMMWORD PTR __real_half por xmm1, XMMWORD PTR __real_half lea r9, QWORD PTR __log_F_inv_qword ; check for negative numbers or zero xorpd xmm5, xmm5 comisd xmm0, xmm5 jbe Llog10_sse2_x_is_zero_or_neg ; f = F - Y, r = f * inv subsd xmm1, xmm2 mulsd xmm1, QWORD PTR [r9+rax*8] movapd xmm2, xmm1 movapd xmm0, xmm1 lea r9, QWORD PTR __log10_256_lead ; poly movsd xmm3, QWORD PTR __real_1_over_6 movsd xmm1, QWORD PTR __real_1_over_3 mulsd xmm3, xmm2 mulsd xmm1, xmm2 mulsd xmm0, xmm2 movapd xmm4, xmm0 addsd xmm3, QWORD PTR __real_1_over_5 addsd xmm1, QWORD PTR __real_1_over_2 mulsd xmm4, xmm0 mulsd xmm3, xmm2 mulsd xmm1, xmm0 addsd xmm3, QWORD PTR __real_1_over_4 addsd xmm1, xmm2 mulsd xmm3, xmm4 addsd xmm1, xmm3 movsd xmm5, QWORD PTR __real_log10_2_tail mulsd xmm1, QWORD PTR __real_log10_e ; m*log(10) + log10(G) - poly mulsd xmm5, xmm6 subsd xmm5, xmm1 movsd xmm0, QWORD PTR [r9+rax*8] lea rdx, QWORD PTR __log10_256_tail movsd xmm2, QWORD PTR [rdx+rax*8] movsd xmm4, QWORD PTR __real_log10_2_lead mulsd xmm4, xmm6 addsd xmm0, xmm4 addsd xmm2, xmm5 addsd xmm0, xmm2 RestoreXmm xmm6, save_xmm6 StackDeallocate stack_size ret ALIGN 16 Llog10_sse2_near_one: ; r = x - 1.0 movsd xmm2, QWORD PTR __real_two subsd xmm0, QWORD PTR __real_one ; r addsd xmm2, xmm0 movapd xmm1, xmm0 divsd xmm1, xmm2 ; r/(2+r) = u/2 movsd xmm4, QWORD PTR __real_ca2 movsd xmm5, QWORD PTR __real_ca4 movapd xmm6, xmm0 mulsd xmm6, xmm1 ; correction addsd xmm1, xmm1 ; u movapd xmm2, xmm1 mulsd xmm2, xmm1 ; u^2 mulsd xmm4, xmm2 mulsd xmm5, xmm2 addsd xmm4, QWORD PTR __real_ca1 addsd xmm5, QWORD PTR __real_ca3 mulsd xmm2, xmm1 ; u^3 mulsd xmm4, xmm2 mulsd xmm2, xmm2 mulsd xmm2, xmm1 ; u^7 mulsd xmm5, xmm2 movsd xmm2, QWORD PTR __real_log10_e_tail addsd xmm4, xmm5 subsd xmm4, xmm6 movsd xmm6, QWORD PTR __real_log10_e_lead movapd xmm3, xmm0 pand xmm3, XMMWORD PTR __mask_lower subsd xmm0, xmm3 addsd xmm4, xmm0 movapd xmm0, xmm3 movapd xmm1, xmm4 mulsd xmm4, xmm2 mulsd xmm0, xmm2 mulsd xmm1, xmm6 mulsd xmm3, xmm6 addsd xmm0, xmm4 addsd xmm0, xmm1 addsd xmm0, xmm3 RestoreXmm xmm6, save_xmm6 StackDeallocate stack_size ret Llog10_sse2_denormal_adjust: por xmm2, XMMWORD PTR __real_one subsd xmm2, QWORD PTR __real_one movsd xmm5, xmm2 pand xmm2, XMMWORD PTR __mask_mant movd rax, xmm2 psrlq xmm5, 52 psubd xmm5, XMMWORD PTR __mask_2045 cvtdq2pd xmm6, xmm5 jmp Llog10_sse2_continue_common ALIGN 16 Llog10_sse2_x_is_zero_or_neg: jne Llog10_sse2_x_is_neg movsd xmm1, QWORD PTR __real_ninf mov r8d, DWORD PTR __flag_x_zero call fname_special jmp Llog10_sse2_finish ALIGN 16 Llog10_sse2_x_is_neg: movsd xmm1, QWORD PTR __real_neg_qnan mov r8d, DWORD PTR __flag_x_neg call fname_special jmp Llog10_sse2_finish ALIGN 16 Llog10_sse2_x_is_inf_or_nan: cmp rax, QWORD PTR __real_inf je Llog10_sse2_finish cmp rax, QWORD PTR __real_ninf je Llog10_sse2_x_is_neg or rax, QWORD PTR __real_qnanbit movd xmm1, rax mov r8d, DWORD PTR __flag_x_nan call fname_special jmp Llog10_sse2_finish ALIGN 16 Llog10_sse2_finish: RestoreXmm xmm6, save_xmm6 StackDeallocate stack_size ret ALIGN 16 Llog10_fma3: ; compute exponent part xor rax,rax vpsrlq xmm3,xmm0,52 vmovq rax,xmm0 vpsubq xmm3,xmm3,QWORD PTR __int_1023 vcvtdq2pd xmm6,xmm3 ; xmm6 <-- (double)xexp ; NaN or Inf? vpand xmm5,xmm0,__real_inf vcomisd xmm5,QWORD PTR __real_inf je Llog10_fma3_x_is_inf_or_nan ; negative number or zero? vpxor xmm5,xmm5,xmm5 vcomisd xmm0,xmm5 jbe Llog10_fma3_x_is_zero_or_neg vpand xmm2,xmm0,__mask_mant vsubsd xmm4,xmm0,QWORD PTR __real_one ; Subnormal? vcomisd xmm6,QWORD PTR __real_neg_1023 je Llog10_fma3_denormal_adjust Llog10_fma3_continue_common: ; compute index into the log tables vpand xmm1,xmm0,DWORD PTR __mask_mant_top8 vpand xmm3,xmm0,DWORD PTR __mask_mant9 vpsllq xmm3,xmm3,1 vpaddq xmm1,xmm3,xmm1 vmovq rax,xmm1 ; near one codepath vpand xmm4,xmm4,DWORD PTR __real_notsign vcomisd xmm4,QWORD PTR __real_threshold jb Llog10_fma3_near_one ; F,Y shr rax,44 vpor xmm2,xmm2,DWORD PTR __real_half vpor xmm1,xmm1,DWORD PTR __real_half lea r9,DWORD PTR __log_F_inv_qword ; f = F - Y,r = f * inv vsubsd xmm1,xmm1,xmm2 vmulsd xmm1,xmm1,QWORD PTR [r9 + rax * 8] lea r9,DWORD PTR __log10_256_lead ; poly vmulsd xmm0,xmm1,xmm1 ; r*r vmovsd xmm3,QWORD PTR __real_1_over_6 vmovsd xmm5,QWORD PTR __real_1_over_3 vfmadd213sd xmm3,xmm1,QWORD PTR __real_1_over_5 ; r*1/6 + 1/5 vfmadd213sd xmm5,xmm1,QWORD PTR __real_half ; 1/2+r*1/3 movsd xmm4,xmm0 ; r*r vfmadd213sd xmm3 ,xmm1,QWORD PTR __real_1_over_4 ; 1/4+(1/5*r+r*r*1/6) vmulsd xmm4,xmm0,xmm0 ; r*r*r*r vfmadd231sd xmm1,xmm5,xmm0 ; r*r*(1/2+r*1/3) + r vfmadd231sd xmm1,xmm3,xmm4 vmulsd xmm1,xmm1,QWORD PTR __real_log10_e ; m*log(2) + log(G) - poly*log10_e vmovsd xmm5,QWORD PTR __real_log10_2_tail vfmsub213sd xmm5,xmm6,xmm1 movsd xmm0,QWORD PTR [r9 + rax * 8] lea rdx,DWORD PTR __log10_256_tail movsd xmm2,QWORD PTR [rdx + rax * 8] vaddsd xmm2,xmm2,xmm5 vfmadd231sd xmm0,xmm6,QWORD PTR __real_log10_2_lead vaddsd xmm0,xmm0,xmm2 AVXRestoreXmm xmm6, save_xmm6 StackDeallocate stack_size ret ALIGN 16 Llog10_fma3_near_one: ; r = x - 1.0 vmovsd xmm2,QWORD PTR __real_two vsubsd xmm0,xmm0,QWORD PTR __real_one ; r vaddsd xmm2,xmm2,xmm0 vdivsd xmm1,xmm0,xmm2 ; r/(2+r) = u/2 vmovsd xmm4,QWORD PTR __real_ca2 vmovsd xmm5,QWORD PTR __real_ca4 vmulsd xmm6,xmm0,xmm1 ; correction vaddsd xmm1,xmm1,xmm1 ; u vmulsd xmm2,xmm1,xmm1 ; u^2 vfmadd213sd xmm4,xmm2,QWORD PTR __real_ca1 vfmadd213sd xmm5,xmm2,QWORD PTR __real_ca3 vmulsd xmm2,xmm2,xmm1 ; u^3 vmulsd xmm4,xmm4,xmm2 vmulsd xmm2,xmm2,xmm2 vmulsd xmm2,xmm2,xmm1 ; u^7 vmulsd xmm5,xmm5,xmm2 vaddsd xmm4,xmm4,xmm5 vsubsd xmm4,xmm4,xmm6 vpand xmm3,xmm0,QWORD PTR __mask_lower vsubsd xmm0,xmm0,xmm3 vaddsd xmm4,xmm4,xmm0 vmulsd xmm1,xmm4,QWORD PTR __real_log10_e_lead vmulsd xmm4,xmm4,QWORD PTR __real_log10_e_tail vmulsd xmm0,xmm3,QWORD PTR __real_log10_e_tail vmulsd xmm3,xmm3,QWORD PTR __real_log10_e_lead vaddsd xmm0,xmm0,xmm4 vaddsd xmm0,xmm0,xmm1 vaddsd xmm0,xmm0,xmm3 AVXRestoreXmm xmm6, save_xmm6 StackDeallocate stack_size ret Llog10_fma3_denormal_adjust: vpor xmm2,xmm2,QWORD PTR __real_one vsubsd xmm2,xmm2,QWORD PTR __real_one vpsrlq xmm5,xmm2,52 vpand xmm2,xmm2,QWORD PTR __mask_mant vmovapd xmm0,xmm2 vpsubd xmm5,xmm5,DWORD PTR __mask_2045 vcvtdq2pd xmm6,xmm5 jmp Llog10_fma3_continue_common ALIGN 16 Llog10_fma3_x_is_zero_or_neg: jne Llog10_fma3_x_is_neg vmovsd xmm1,QWORD PTR __real_ninf mov r8d,DWORD PTR __flag_x_zero call fname_special AVXRestoreXmm xmm6, save_xmm6 StackDeallocate stack_size ret ALIGN 16 Llog10_fma3_x_is_neg: vmovsd xmm1,QWORD PTR __real_neg_qnan mov r8d,DWORD PTR __flag_x_neg call fname_special AVXRestoreXmm xmm6, save_xmm6 StackDeallocate stack_size ret ALIGN 16 Llog10_fma3_x_is_inf_or_nan: cmp rax,QWORD PTR __real_inf je Llog10_fma3_finish cmp rax,QWORD PTR __real_ninf je Llog10_fma3_x_is_neg or rax,QWORD PTR __real_qnanbit movd xmm1,rax mov r8d,DWORD PTR __flag_x_nan call fname_special jmp Llog10_fma3_finish ALIGN 16 Llog10_fma3_finish: AVXRestoreXmm xmm6, save_xmm6 StackDeallocate stack_size ret fname endp END