windows-nt/Source/XPSP1/NT/base/ntos/ke/ia64/fpmisc.s

590 lines
8.8 KiB
ArmAsm
Raw Permalink Normal View History

2020-09-26 03:20:57 -05:00
#include "ksia64.h"
//++
//
// VOID
// run_fms (
// IN ULONGLONG *fpsr,
// OUT FLOAT128 *fr1,
// IN FLOAT128 *fr2,
// IN FLOAT128 *fr3,
// IN FLOAT128 *fr4
// )
//
// Routine Description:
//
// This function runs FMS operation with the specified inputs and FPSR.
//
//--
LEAF_ENTRY(run_fms)
alloc r31=ar.pfs,5,2,0,0 // r32, r33, r34, r35, r36, r37, r38
ARGPTR (r32)
ARGPTR (r33)
ARGPTR (r34)
ARGPTR (r35)
ARGPTR (r36)
// &fpsr is in r32
// &fr1 (output) is in r33
// &fr2 (input) is in r34
// &fr3 (input) is in r35
// &fr4 (input) is in r36
// save old FPSR in r37
mov r37 = ar40
nop.i 0;;
// load new fpsr in r38
ld8 r38 = [r32];;
// set new value of FPSR
mov ar40 = r38
nop.i 0;;
// load first input argument into f8
ldf.fill f8 = [r34]
// load second input argument into f9
ldf.fill f9 = [r35]
nop.i 0;;
// load third input argument into f10
ldf.fill f10 = [r36]
nop.m 0
nop.i 0;;
nop.m 0
(p0) fms.s0 f11 = f8, f9, f10 // f11 = f8 * f9 - f10
nop.i 0;;
// store result
stf.spill [r33] = f11
// save new FPSR in r38
mov r38 = ar40
nop.i 0;;
// store new fpsr from r38
st8 [r32] = r38
// restore FPSR
mov ar40 = r37
nop.i 0;;
nop.m 0
nop.i 0
// return
LEAF_RETURN
LEAF_EXIT(run_fms)
//++
//
// VOID
// thmF (
// IN ULONGLONG *fpsr,
// OUT FLOAT128 *fr1,
// IN FLOAT128 *fr2,
// IN FLOAT128 *fr3
// )
//
// Routine Description:
//
//--
LEAF_ENTRY(thmF)
alloc r31=ar.pfs,4,4,0,0 // r32, r33, r34, r35, r36, r37, r38, r39
ARGPTR (r32)
ARGPTR (r33)
ARGPTR (r34)
ARGPTR (r35)
// &fpsr is in r32
// &a is in r33
// &b is in r34
// &div is in r35 (the address of the divide result)
// save old FPSR in r36
mov r36 = ar40
// save predicates in r37
mov r37 = pr;;
// load new fpsr in r39
ld8 r39 = [r32];;
// set new value of FPSR
mov ar40 = r39
nop.i 0;;
nop.m 0
// clear predicates
movl r38 = 0x0000000000000001;;
nop.m 0
// load clear predicates from r38
mov pr = r38,0x1ffff
nop.i 0;;
// load a, the first argument, in f6
ldf.fill f6 = [r33]
// load b, the second argument, in f7
ldf.fill f7 = [r34]
nop.i 0;;
nop.m 0
// Step (1)
// y0 = 1 / b in f8
frcpa.s0 f8,p2=f6,f7
nop.i 0;;
nop.m 0
// Step (2)
// e0 = 1 - b * y0 in f9
(p2) fnma.s1 f9=f7,f8,f1
nop.i 0
nop.m 0
// Step (10)
// q0 = a * y0 in f10
(p2) fma.s1 f10=f6,f8,f0
nop.i 0;;
nop.m 0
// Step (3)
// y1 = y0 + e0 * y0 in f8
(p2) fma.s1 f8=f9,f8,f8
nop.i 0
nop.m 0
// Step (4)
// e1 = e0 * e0 in f9
(p2) fma.s1 f9=f9,f9,f0
nop.i 0
nop.m 0
// Step (11)
// r0 = a - b * q0 in f11
(p2) fnma.s1 f11=f7,f10,f6
nop.i 0;;
nop.m 0
// Step (5)
// y2 = y1 + e1 * y1 in f8
(p2) fma.s1 f8=f8,f9,f8
nop.i 0;;
nop.m 0
// Step (6)
// e2 = 1 - b * y2 in f9
(p2) fnma.s1 f9=f7,f8,f1
nop.i 0;;
nop.m 0
// Step (7)
// y3 = y2 + e2 * y2 in f8
(p2) fma.s1 f8=f8,f9,f8
nop.i 0;;
nop.m 0
// Step (8)
// e3 = 1 - b * y3 in f9
(p2) fnma.s1 f9=f7,f8,f1
nop.i 0
nop.m 0
// Step (12)
// q1 = q0 + r0 * y3 in f10
(p2) fma.s1 f10=f11,f8,f10
nop.i 0;;
nop.m 0
// Step (9)
// y4 = y3 + e3 * y3 in f8
(p2) fma.s1 f8=f8,f9,f8
nop.i 0
nop.m 0
// Step (13)
// r1 = a - b * q1 in f11
(p2) fnma.s1 f11=f7,f10,f6
nop.i 0;;
nop.m 0
// Step (14)
// q2 = q1 + r1 * y4 in f8
(p2) fma.s0 f8=f11,f8,f10
nop.i 0;;
// save new FPSR in r39
mov r39 = ar40;;
// store new fpsr from r39
st8 [r32] = r39
// restore predicates from r37
mov pr = r37,0x1ffff;;
// store result
stf.spill [r35]=f8
// restore FPSR
mov ar40 = r36
// return
LEAF_RETURN
LEAF_EXIT(thmF)
//++
//
// VOID
// thmL (
// IN ULONGLONG *fpsr,
// OUT FLOAT128 *fr1,
// IN FLOAT128 *fr2
// )
//
// Routine Description:
//
//--
LEAF_ENTRY(thmL)
alloc r31=ar.pfs,3,5,0,0 // r32, r33, r34, r35, r36, r37, r38, r39
ARGPTR (r32)
ARGPTR (r33)
ARGPTR (r34)
// &fpsr is in r32
// &a is in r33
// &sqrt is in r34 (the address of the sqrt result)
// save old FPSR in r35
mov r35 = ar40
// save predicates in r36
mov r36 = pr;;
// load new fpsr in r38
ld8 r38 = [r32];;
// set new value of FPSR
mov ar40 = r38
nop.i 0;;
nop.m 0
// clear predicates
movl r37 = 0x0000000000000001;;
nop.m 0
// load clear predicates from r37
mov pr = r37,0x1ffff
nop.i 0;;
// load the argument a in f6
ldf.fill f6 = [r33]
nop.m 0
nop.i 0;;
nop.m 0
// Step (1)
// y0 = 1/sqrt(a) in f8
frsqrta.s0 f8,p2=f6
nop.i 0;;
nop.m 0
// Step (2)
// load 1/2 in f7; h = 1/2 * a in f9
(p2) movl r39 = 0x0fffe;;
(p2) setf.exp f7 = r39
nop.i 0;;
nop.m 0
(p2) fma.s1 f9=f7,f6,f0
nop.i 0;;
nop.m 0
// Step (3)
// t1 = y0 * y0 in f10
(p2) fma.s1 f10=f8,f8,f0
nop.i 0;;
nop.m 0
// Step (4)
// t2 = 1/2 - t1 * h in f10
(p2) fnma.s1 f10=f10,f9,f7
nop.i 0;;
nop.m 0
// Step (5)
// y1 = y0 + t2 * y0 in f8
(p2) fma.s1 f8=f10,f8,f8
nop.i 0;;
nop.m 0
// Step (6)
// t3 = y1 * h in f10
(p2) fma.s1 f10=f8,f9,f0
nop.i 0;;
nop.m 0
// Step (7)
// t4 = 1/2 - t3 * y1 in f10
(p2) fnma.s1 f10=f10,f8,f7
nop.i 0;;
nop.m 0
// Step (8)
// y2 = y1 + t4 * y1 in f8
(p2) fma.s1 f8=f10,f8,f8
nop.i 0;;
nop.m 0
// Step (9)
// S = a * y2 in f10
(p2) fma.s1 f10=f6,f8,f0
nop.i 0;;
nop.m 0
// Step (10)
// t5 = y2 * h in f9
(p2) fma.s1 f9=f8,f9,f0
nop.i 0;;
nop.m 0
// Step (11)
// H = 1/2 * y2 in f11
(p2) fma.s1 f11=f7,f8,f0
nop.i 0;;
nop.m 0
// Step (13)
// t6 = 1/2 - t5 * y2 in f7
(p2) fnma.s1 f7=f9,f8,f7
nop.i 0;;
nop.m 0
// Step (12)
// d = a - S * S in f8
(p2) fnma.s1 f8=f10,f10,f6
nop.i 0;;
nop.m 0
// Step (14)
// S1 = S + d * H in f8
(p2) fma.s1 f8=f8,f11,f10
nop.i 0;;
nop.m 0
// Step (15)
// H1 = H + t6 * h in f7
(p2) fma.s1 f7=f11,f7,f11
nop.i 0;;
nop.m 0
// Step (16)
// d1 = a - S1 * S1 in f6
(p2) fnma.s1 f6=f8,f8,f6
nop.i 0;;
nop.m 0
// Step (17)
// R = S1 + d1 * H1 in f8
(p2) fma.s0 f8=f6,f7,f8
nop.i 0;;
// save new FPSR in r38
mov r38 = ar40;;
// store new fpsr from r38
st8 [r32] = r38
// restore predicates from r36
mov pr = r36,0x1ffff;;
// store result
stf.spill [r34]=f8
// restore FPSR
mov ar40 = r35
// return
LEAF_RETURN
LEAF_EXIT(thmL)
//++
//
// VOID
// KiEmulateLoadFloat80(
// IN PVOID UnalignedAddress,
// OUT PVOID FloatData
// );
//
//--
LEAF_ENTRY(KiEmulateLoadFloat80)
ARGPTR(a0)
ARGPTR(a1)
ldfe ft0 = [a0]
;;
stf.spill [a1] = ft0
LEAF_RETURN
LEAF_EXIT(KiEmulateLoadFloat80)
//++
//
// VOID
// KiEmulateLoadFloatInt(
// IN PVOID UnalignedAddress,
// OUT PVOID FloatData
// );
//
//--
LEAF_ENTRY(KiEmulateLoadFloatInt)
ARGPTR(a0)
ARGPTR(a1)
ldf8 ft0 = [a0]
;;
stf.spill [a1] = ft0
LEAF_RETURN
LEAF_EXIT(KiEmulateLoadFloatInt)
//++
//
// VOID
// KiEmulateLoadFloat32(
// IN PVOID UnalignedAddress,
// OUT PVOID FloatData
// );
//
//--
LEAF_ENTRY(KiEmulateLoadFloat32)
ARGPTR(a0)
ARGPTR(a1)
ldfs ft0 = [a0]
;;
stf.spill [a1] = ft0
LEAF_RETURN
LEAF_EXIT(KiEmulateLoadFloat32)
//++
//
// VOID
// KiEmulateLoadFloat64(
// IN PVOID UnalignedAddress,
// OUT PVOID FloatData
// );
//
//--
LEAF_ENTRY(KiEmulateLoadFloat64)
ARGPTR(a0)
ARGPTR(a1)
ldfd ft0 = [a0]
;;
stf.spill [a1] = ft0
LEAF_RETURN
LEAF_EXIT(KiEmulateLoadFloat64)
//++
//
// VOID
// KiEmulateStoreFloat80(
// IN PVOID UnalignedAddress,
// OUT PVOID FloatData
// );
//
//--
LEAF_ENTRY(KiEmulateStoreFloat80)
ARGPTR(a0)
ARGPTR(a1)
ldf.fill ft0 = [a1]
;;
stfe [a0] = ft0
LEAF_RETURN
LEAF_EXIT(KiEmulateStoreFloat80)
//++
//
// VOID
// KiEmulateStoreFloatInt(
// IN PVOID UnalignedAddress,
// OUT PVOID FloatData
// );
//
//--
LEAF_ENTRY(KiEmulateStoreFloatInt)
ARGPTR(a0)
ARGPTR(a1)
ldf.fill ft0 = [a1]
;;
stfd [a0] = ft0
LEAF_RETURN
LEAF_EXIT(KiEmulateStoreFloatInt)
//++
//
// VOID
// KiEmulateStoreFloat32(
// IN PVOID UnalignedAddress,
// OUT PVOID FloatData
// );
//
//--
LEAF_ENTRY(KiEmulateStoreFloat32)
ARGPTR(a0)
ARGPTR(a1)
ldf.fill ft0 = [a1]
;;
stfs [a0] = ft0
LEAF_RETURN
LEAF_EXIT(KiEmulateStoreFloat32)
//++
//
// VOID
// KiEmulateStoreFloat64(
// IN PVOID UnalignedAddress,
// OUT PVOID FloatData
// );
//
//--
LEAF_ENTRY(KiEmulateStoreFloat64)
ARGPTR(a0)
ARGPTR(a1)
ldf.fill ft0 = [a1]
;;
stfd [a0] = ft0
LEAF_RETURN
LEAF_EXIT(KiEmulateStoreFloat64)