File: SHL.md

package info (click to toggle)
intel-graphics-compiler 1.0.17791.18-1
  • links: PTS, VCS
  • area: main
  • in suites: sid
  • size: 102,312 kB
  • sloc: cpp: 935,343; lisp: 286,143; ansic: 16,196; python: 3,279; yacc: 2,487; lex: 1,642; pascal: 300; sh: 174; makefile: 27
file content (109 lines) | stat: -rw-r--r-- 2,399 bytes parent folder | download | duplicates (2)
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
<!---======================= begin_copyright_notice ============================

Copyright (C) 2020-2022 Intel Corporation

SPDX-License-Identifier: MIT

============================= end_copyright_notice ==========================-->

## Opcode

  SHL = 0x24

## Format

| | | | | | |
| --- | --- | --- | --- | --- | --- |
| 0x24(SHL) | Exec_size | Pred | Dst | Src0 | Src1 |


## Semantics


```

                    for (i = 0; i < exec_size; ++i) {
                      if (ChEn[i]) {
                        dst[i] = src0[i] << src1[i];
                      }
                    }
```

## Description





```
    Performs component-wise logic left shift of <src0> and stores the result into <dst>.
```


- **Exec_size(ub):** Execution size

  - Bit[2..0]: size of the region for source and destination operands

    - 0b000:  1 element (scalar)
    - 0b001:  2 elements
    - 0b010:  4 elements
    - 0b011:  8 elements
    - 0b100:  16 elements
    - 0b101:  32 elements
  - Bit[7..4]: execution mask (explicit control over the enabled channels)

    - 0b0000:  M1
    - 0b0001:  M2
    - 0b0010:  M3
    - 0b0011:  M4
    - 0b0100:  M5
    - 0b0101:  M6
    - 0b0110:  M7
    - 0b0111:  M8
    - 0b1000:  M1_NM
    - 0b1001:  M2_NM
    - 0b1010:  M3_NM
    - 0b1011:  M4_NM
    - 0b1100:  M5_NM
    - 0b1101:  M6_NM
    - 0b1110:  M7_NM
    - 0b1111:  M8_NM

- **Pred(uw):** Predication control


- **Dst(vec_operand):** The destination operand. Operand class: general,indirect


- **Src0(vec_operand):** The first source operand. Operand class: general,indirect,immediate


- **Src1(vec_operand):** The second source operand. Operand class: general,indirect,immediate


#### Properties
- **Supported Types:** B,D,Q,UB,UD,UQ,UW,W
- **Saturation:** Yes
- **Source Modifier:** arithmetic


#### Operand type maps
- **Type map**
  -  **Dst types:** UD, D, UW, W, UB, B
  -  **Src types:** UD, D, UW, W, UB, B


## Text
```
[(<P>)] SHL[.sat] (<exec_size>) <dst> <src0> <src1>
```

## Notes





  Dst, src0, and src1 must have integer type, though src0 and src1 may be different type. The last 5 (or 6 for Q/UQ type destinations) LSB bits of src1 are used as an unsigned integer value for the shift operation. If saturation is enabled, the shifted result must be within 33 bits or the result is undefined.
    Q/UQ is not supported on all platforms without int64 (ICL, TGL, DG2, MTL, etc.)