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
|
; RUN: llvm-upgrade %s | llvm-as | llc
; Test that a negative constant smaller than 64 bits (e.g., int)
; is correctly implemented with sign-extension.
; In particular, the current code generated is:
;
; main:
; .L_main_LL_0:
; save %o6, -224, %o6
; setx .G_fmtArg_1, %o1, %o0
; setuw 1, %o1 ! i = 1
; setuw 4294967295, %o3 ! THE BUG: 0x00000000ffffffff
; setsw 0, %i0
; add %i6, 1999, %o2 ! fval
; add %o1, %g0, %o1
; add %o0, 0, %o0
; mulx %o1, %o3, %o1 ! ERROR: 0xffffffff; should be -1
; add %o1, 3, %o1 ! ERROR: 0x100000002; should be 0x2
; mulx %o1, 12, %o3 !
; add %o2, %o3, %o3 ! produces bad address!
; call printf
; nop
; jmpl %i7+8, %g0
; restore %g0, 0, %g0
;
; llc produces:
; ioff = 2 fval = 0xffffffff7fffec90 &fval[2] = 0xb7fffeca8
; instead of:
; ioff = 2 fval = 0xffffffff7fffec90 &fval[2] = 0xffffffff7fffeca8
;
%Results = type { float, float, float }
%fmtArg = internal global [39 x sbyte] c"ioff = %u\09fval = 0x%p\09&fval[2] = 0x%p\0A\00"; <[39 x sbyte]*> [#uses=1]
implementation
declare int "printf"(sbyte*, ...)
int "main"()
begin
%fval = alloca %Results, uint 4
%i = add uint 1, 0 ; i = 1
%iscale = mul uint %i, 4294967295 ; i*-1 = -1
%ioff = add uint %iscale, 3 ; 3+(-i) = 2
%ioff = cast uint %ioff to long
%fptr = getelementptr %Results* %fval, long %ioff ; &fval[2]
%castFmt = getelementptr [39 x sbyte]* %fmtArg, long 0, long 0
call int (sbyte*, ...)* %printf(sbyte* %castFmt, uint %ioff, %Results* %fval, %Results* %fptr)
ret int 0
end
|
