File: exec.cpp

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 (123 lines) | stat: -rw-r--r-- 3,910 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
110
111
112
113
114
115
116
117
118
119
120
121
122
123
/*========================== begin_copyright_notice ============================

Copyright (C) 2021-2024 Intel Corporation

SPDX-License-Identifier: MIT

============================= end_copyright_notice ===========================*/

#include <cm-cl/atomic.h>
#include <cm-cl/exec.h>

using namespace cm;

CM_NODEBUG CM_INLINE uint __spirv_BuiltInWorkDim() { return 3; }

CM_NODEBUG CM_INLINE ulong __spirv_BuiltInGlobalSize(int dim) {
  return static_cast<ulong>(cm::exec::get_local_size(dim)) *
         cm::exec::get_group_count(dim);
}

CM_NODEBUG CM_INLINE ulong __spirv_BuiltInGlobalInvocationId(int dim) {
  return cm::exec::get_local_id(dim) +
         static_cast<ulong>(cm::exec::get_group_id(dim)) *
             cm::exec::get_local_size(dim);
}

CM_NODEBUG CM_INLINE ulong __spirv_BuiltInWorkgroupSize(int dim) {
  return cm::exec::get_local_size(dim);
}

CM_NODEBUG CM_INLINE ulong __spirv_BuiltInEnqueuedWorkgroupSize(int dim) {
  return cm::exec::get_local_size(dim);
}

CM_NODEBUG CM_INLINE ulong __spirv_BuiltInLocalInvocationId(int dim) {
  return cm::exec::get_local_id(dim);
}

CM_NODEBUG CM_INLINE ulong __spirv_BuiltInNumWorkgroups(int dim) {
  return cm::exec::get_group_count(dim);
}

CM_NODEBUG CM_INLINE ulong __spirv_BuiltInWorkgroupId(int dim) {
  return cm::exec::get_group_id(dim);
}

CM_NODEBUG CM_INLINE ulong __spirv_BuiltInGlobalOffset(int dim) {
  // VC does not support global offset yet
  return 0;
}

CM_NODEBUG CM_INLINE ulong __spirv_BuiltInLocalInvocationIndex() {
  using namespace cm::exec;
  return get_local_size(0) * get_local_size(1) * get_local_id(2) +
         get_local_size(0) * get_local_id(1) + get_local_id(0);
}

CM_NODEBUG CM_INLINE ulong __spirv_BuiltInGlobalLinearId() {
  using namespace cm::exec;
  uint group_id = get_group_count(0) * get_group_count(1) * get_group_id(2) +
                  get_group_count(0) * get_group_id(1) + get_group_id(0);
  uint local_size = get_local_size(0) * get_local_size(1) * get_local_size(2);
  uint local_id = __spirv_BuiltInLocalInvocationIndex();

  return group_id * local_size + local_id;
}

CM_NODEBUG CM_INLINE uint __spirv_BuiltInSubgroupSize() {
  // VC code always uses SubgroupSize == 1
  return 1;
}

CM_NODEBUG CM_INLINE uint __spirv_BuiltInSubgroupMaxSize() {
  // VC code always uses SubgroupSize == 1
  return 1;
}

CM_NODEBUG CM_INLINE uint __spirv_BuiltInNumSubgroups() {
  return __spirv_BuiltInWorkgroupSize(0) * __spirv_BuiltInWorkgroupSize(1) *
         __spirv_BuiltInWorkgroupSize(2);
}

CM_NODEBUG CM_INLINE uint __spirv_BuiltInNumEnqueuedSubgroups() {
  return __spirv_BuiltInNumSubgroups();
}

CM_NODEBUG CM_INLINE uint __spirv_BuiltInSubgroupId() {
  return __spirv_BuiltInLocalInvocationIndex();
}

CM_NODEBUG CM_INLINE uint __spirv_BuiltInSubgroupLocalInvocationId() {
  // VC code always uses SubgroupSize == 1
  return 0;
}

// When __SubDeviceID is declared as an extern int, it is lowered to LLVM-IR
// like:
//
// @__SubDeviceID = external addrspace(1) global i32, align 4
//
// This global address is being then patched by the runtime and can be set to
// null when implicit scaling is disabled. One may wonder why `__SubDeviceID` is
// not declared as an `extern int*` In this case this would end up as a pointer
// to pointer in LLVM-IR. It would generate two loads and wouldn't be consistent
// with runtime behavior.
__global extern int __SubDeviceID;

// This variable has to be patched by GenXInitBiFConstant pass which will
// convert it into a constant with a target specific initialization value
int __cm_cl_MaxHWThreadIDPerSubDevice;

int __spirv_BuiltInSubDeviceIDINTEL() {
  __global volatile int *P = &__SubDeviceID;
  if (!P)
    return 0;
  return *P;
}

int __spirv_BuiltInGlobalHWThreadIDINTEL() {
  int SubDeviceId = __spirv_BuiltInSubDeviceIDINTEL();
  return cm::detail::__cm_cl_hw_thread_id() +
         SubDeviceId * __cm_cl_MaxHWThreadIDPerSubDevice;
}