blob: a42c37e224bfc9d48c34d8126c7134fd25773327 (
plain)
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
|
# Chapter 14: Narrow types
## Narrow or sub-word types
As a prelude to arrays, especially arrays of bytes common in all networking
codes, Simple needs some way to manipulate sub-word integer types.
Computations are always done in 64 bits for both integer and float ops, but
will be truncated (rounded) on a variable or field assignment, and (sign)
extended on a load.
```java
i8 x = 123456789;
if( x != 21 )
return false;
return true;
```
The variable `x` is limited to 8 bits signed. Assigning the vary large value
`123456789` (hex `0x75BCD15`) simply truncates to 8 bits yielding `21 (0x15)`.
The value is sign-extended on a load.
```java
i16 x = 123456789;
if( x != -13035 )
return false;
return true;
```
The variable `x` is limited to 16 bits signed. Assigning the vary large value
`0x75BCD15` simply truncates to 16 bits yielding `0xCD15`. The value is
sign-extended on a load yielding `-13035 (0xFFFF FFFF FFFF CD15)`.
```java
u16 x = 123456789;
if( x != 52501 )
return false;
return true;
```
The variable `x` is limited to 16 bits unsigned. Assigning the vary large value
`0x75BCD15` simply truncates to 16 bits yielding `0xCD15`. The value is
zero-extended on a load yielding `52501 (0x0000 0000 0000 CD15)`.
Similarly, floats can be limited to IEEE754 32-bit standard format and will be
rounded when assigned, and inflated to 64-bits when loaded.
## Type Implementation
The TypeInteger class is reworked to support a full range of min/max values.
At this time, only some power-of-2 sized ranges are exposed but the optimizer
internally supports all ranges. The MEET operation takes the min-of-mins and
max-of-maxes. The DUAL operation inverts the min and max.
The TypeFloat class is also reworked to support 32-bit and 64-bit sizes.
## Nodes
Added this Chapter are some bit-manipulation Nodes representing common hardware
ops to mask, shift, sign-extend and round bits. The truncation and extension
logic uses these. They are otherwise very similar to the previous
`Add/Sub/Mul/Div` Nodes but with slightly different constant math and
slightly different `idealize()` calls.
|
