1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
|
// Select C numeric constant
.radix C
.psr abi64
.psr lsb
// Section has executable code
.section .text, "ax","progbits"
// procedure named 'SharedStub'
.proc SharedStub
// manual bundling
.explicit
.global PrepareAndDispatch
// .exclass PrepareAndDispatch, @fullyvisible
.type PrepareAndDispatch,@function
SharedStub::
// 10 arguments, first 8 are the input arguments of previous
// function call. The 9th one is methodIndex and the 10th is the
// pointer to the remaining input arguments. The last two arguments
// are passed in memory.
.prologue
.save ar.pfs , r41
// allocate 8 input args, 4 local args, and 5 output args
alloc r41 = ar.pfs, 8, 4, 5, 0 // M
.save rp, r40
mov r40 = rp // I
add out4 = 24, sp ;; // I
.save ar.unat, r42
mov r42 = ar.unat // M
.fframe 144
add sp = -144, sp // A
// unwind table already knows gp, don't need to specify anything
add r43 = 0, gp ;; // A
// We have possible 8 integer registers and 8 float registers that could
// be arguments. We also have a stack region from the previous
// stack frame that may hold some stack arguments.
// We need to write the integer registers to a memory region, write
// the float registers to a memory region (making sure we don't step
// on NAT while touching the registers). We also mark the memory
// address of the stack arguments.
// We then call PrepareAndDispatch() specifying the three memory
// region pointers.
.body
add out0 = 0, in0 // A move self ptr
// 144 bytes = 16 byte stack header + 64 byte int space + 64 byte float space
// methodIndex is at 144 + 16 bytes away from current sp
// (current frame + previous frame header)
ld8 out4 = [out4] // M restarg address
add r11 = 160, sp ;; // A address of methodIndex
ld8 out1 = [r11] // M load methodIndex
// sp + 16 is the start of intargs
add out2 = 16, sp // A address of intargs
// the intargs take up 64 bytes, so sp + 16 + 64 is the start of floatargs
add out3 = 80, sp ;; // A address of floatargs
add r11 = 0, out2 ;; // A
st8.spill [r11] = in1, 8 // M
add r10 = 0, out3 ;; // A
st8.spill [r11] = in2, 8 ;; // M
st8.spill [r11] = in3, 8 // M
nop.i 0 ;; // I
st8.spill [r11] = in4, 8 ;; // M
st8.spill [r11] = in5, 8 // M
nop.i 0 ;; // I
st8.spill [r11] = in6, 8 ;; // M
st8.spill [r11] = in7 // M
fclass.nm p14,p15 = f8,@nat ;; // F
(p14) stfd [r10] = f8, 8 // M
(p15) add r10 = 8, r10 // A
fclass.nm p12,p13 = f9,@nat ;; // F
(p12) stfd [r10] = f9, 8 // M
(p13) add r10 = 8, r10 // A
fclass.nm p14,p15 =f10,@nat ;; // F
(p14) stfd [r10] = f10, 8 // M
(p15) add r10 = 8, r10 // A
fclass.nm p12,p13 =f11,@nat ;; // F
(p12) stfd [r10] = f11, 8 // M
(p13) add r10 = 8, r10 // A
fclass.nm p14,p15 =f12,@nat ;; // F
(p14) stfd [r10] = f12, 8 // M
(p15) add r10 = 8, r10 // A
fclass.nm p12,p13 =f13,@nat ;; // F
(p12) stfd [r10] = f13, 8 // M
(p13) add r10 = 8, r10 // A
fclass.nm p14,p15 =f14,@nat ;; // F
(p14) stfd [r10] = f14, 8 // M
(p15) add r10 = 8, r10 // A
fclass.nm p12,p13 =f15,@nat ;; // F
(p12) stfd [r10] = f15, 8 // M
(p13) add r10 = 8, r10 // A
// branch to PrepareAndDispatch
br.call.dptk.few rp = PrepareAndDispatch ;; // B
// epilog
mov ar.unat = r42 // M
mov ar.pfs = r41 // I
mov rp = r40 ;; // I
add gp = 0, r43 // A
add sp = 144, sp // A
br.ret.dptk.few rp ;; // B
.endp
/* Magic indicating no need for an executable stack */
.section .note.GNU-stack, "", @progbits
|