mirror of
https://github.com/FEX-Emu/FEX.git
synced 2026-10-26 15:01:06 +01:00
Paulo Matos
b5241e0f60
3DNow Reciprocal estimations did not have enough accuracy. Tests were enabled to check for accurate values of reciprocals. * where needed, reciprocal accuracy was increased. * 3DNow sqrt reciprocal fixed for negative values. * New helper VFCopySign IR op added. Fixes #4319.
30 lines
1.1 KiB
Plaintext
30 lines
1.1 KiB
Plaintext
%ifndef CHECK_PRECISION_INC
|
|
%define CHECK_PRECISION_INC
|
|
|
|
; CheckPrecision.inc - NASM include file containing macro to check precision
|
|
; of single floating-point number.
|
|
|
|
;; Clobbers xmm12, xmm13, xmm14, xmm15
|
|
;; Returns result in al.
|
|
;; Arguments are in memory locations.
|
|
%macro check_relerr 3; %1=REF %2=X %3=TOLERANCE
|
|
movss xmm12, dword [ %1 ] ; xmm12 has REF
|
|
movss xmm13, dword [ %2 ] ; xmm13 has X
|
|
movss xmm15, dword [rel abs_mask_float] ; xmm15 has the abs float mask
|
|
movaps xmm14, xmm12 ; xmm14 has REF
|
|
subss xmm14, xmm13 ; xmm14 = REF - X
|
|
andps xmm12, xmm15 ; xmm12 = abs(REF)
|
|
mulss xmm12, dword [ %3 ] ; xmm12 = abs(REF) * tolerance
|
|
movaps xmm13, xmm14 ; xmm13 = REF - X
|
|
andps xmm13, xmm15 ; xmm13 = abs(REF - X)
|
|
|
|
xor eax, eax ; clears eax
|
|
comiss xmm12, xmm13 ; compares xmm12 and xmm13
|
|
setnb al ; stores xmm12 >= xmm13, i.e. abs(REF) * tolerance >= abs(REF - X)
|
|
%endmacro
|
|
|
|
%macro define_check_data_constants 0
|
|
abs_mask_float:
|
|
dd 0x7fffffff ; Bitmask to get absolute value of a float (single precision)
|
|
%endmacro
|
|
%endif |