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
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
//! NVPTX intrinsics (experimental)
//!
//! These intrinsics form the foundation of the CUDA
//! programming model.
//!
//! The reference is the [CUDA C Programming Guide][cuda_c]. Relevant is also
//! the [LLVM NVPTX Backend documentation][llvm_docs].
//!
//! [cuda_c]:
//! http://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html
//! [llvm_docs]:
//! https://llvm.org/docs/NVPTXUsage.html

use crate::ffi::c_void;

#[allow(improper_ctypes)]
extern "C" {
    #[link_name = "llvm.nvvm.barrier0"]
    fn syncthreads() -> ();
    #[link_name = "llvm.nvvm.read.ptx.sreg.ntid.x"]
    fn block_dim_x() -> i32;
    #[link_name = "llvm.nvvm.read.ptx.sreg.ntid.y"]
    fn block_dim_y() -> i32;
    #[link_name = "llvm.nvvm.read.ptx.sreg.ntid.z"]
    fn block_dim_z() -> i32;
    #[link_name = "llvm.nvvm.read.ptx.sreg.ctaid.x"]
    fn block_idx_x() -> i32;
    #[link_name = "llvm.nvvm.read.ptx.sreg.ctaid.y"]
    fn block_idx_y() -> i32;
    #[link_name = "llvm.nvvm.read.ptx.sreg.ctaid.z"]
    fn block_idx_z() -> i32;
    #[link_name = "llvm.nvvm.read.ptx.sreg.nctaid.x"]
    fn grid_dim_x() -> i32;
    #[link_name = "llvm.nvvm.read.ptx.sreg.nctaid.y"]
    fn grid_dim_y() -> i32;
    #[link_name = "llvm.nvvm.read.ptx.sreg.nctaid.z"]
    fn grid_dim_z() -> i32;
    #[link_name = "llvm.nvvm.read.ptx.sreg.tid.x"]
    fn thread_idx_x() -> i32;
    #[link_name = "llvm.nvvm.read.ptx.sreg.tid.y"]
    fn thread_idx_y() -> i32;
    #[link_name = "llvm.nvvm.read.ptx.sreg.tid.z"]
    fn thread_idx_z() -> i32;
}

/// Synchronizes all threads in the block.
#[inline]
pub unsafe fn _syncthreads() -> () {
    syncthreads()
}

/// x-th thread-block dimension.
#[inline]
pub unsafe fn _block_dim_x() -> i32 {
    block_dim_x()
}

/// y-th thread-block dimension.
#[inline]
pub unsafe fn _block_dim_y() -> i32 {
    block_dim_y()
}

/// z-th thread-block dimension.
#[inline]
pub unsafe fn _block_dim_z() -> i32 {
    block_dim_z()
}

/// x-th thread-block index.
#[inline]
pub unsafe fn _block_idx_x() -> i32 {
    block_idx_x()
}

/// y-th thread-block index.
#[inline]
pub unsafe fn _block_idx_y() -> i32 {
    block_idx_y()
}

/// z-th thread-block index.
#[inline]
pub unsafe fn _block_idx_z() -> i32 {
    block_idx_z()
}

/// x-th block-grid dimension.
#[inline]
pub unsafe fn _grid_dim_x() -> i32 {
    grid_dim_x()
}

/// y-th block-grid dimension.
#[inline]
pub unsafe fn _grid_dim_y() -> i32 {
    grid_dim_y()
}

/// z-th block-grid dimension.
#[inline]
pub unsafe fn _grid_dim_z() -> i32 {
    grid_dim_z()
}

/// x-th thread index.
#[inline]
pub unsafe fn _thread_idx_x() -> i32 {
    thread_idx_x()
}

/// y-th thread index.
#[inline]
pub unsafe fn _thread_idx_y() -> i32 {
    thread_idx_y()
}

/// z-th thread index.
#[inline]
pub unsafe fn _thread_idx_z() -> i32 {
    thread_idx_z()
}

/// Generates the trap instruction `TRAP`
#[inline]
pub unsafe fn trap() -> ! {
    crate::intrinsics::abort()
}

// Basic CUDA syscall declarations.
extern "C" {
    /// Print formatted output from a kernel to a host-side output stream.
    ///
    /// Syscall arguments:
    /// * `status`: The status value that is returned by `vprintf`.
    /// * `format`: A pointer to the format specifier input (uses common `printf` format).
    /// * `valist`: A pointer to the valist input.
    ///
    /// ```
    /// #[repr(C)]
    /// struct PrintArgs(f32, f32, f32, i32);
    ///
    /// vprintf(
    ///     "int(%f + %f) = int(%f) = %d\n".as_ptr(),
    ///     transmute(&PrintArgs(a, b, a + b, (a + b) as i32)),
    /// );
    /// ```
    ///
    /// Sources:
    /// [Programming Guide](https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#formatted-output),
    /// [PTX Interoperability](https://docs.nvidia.com/cuda/ptx-writers-guide-to-interoperability/index.html#system-calls).
    pub fn vprintf(format: *const u8, valist: *const c_void) -> i32;

    /// Allocate memory dynamically from a fixed-size heap in global memory.
    ///
    /// The CUDA in-kernel `malloc()` function allocates at least `size` bytes
    /// from the device heap and returns a pointer to the allocated memory
    /// or `NULL` if insufficient memory exists to fulfill the request.
    ///
    /// The returned pointer is guaranteed to be aligned to a 16-byte boundary.
    ///
    /// The memory allocated by a given CUDA thread via `malloc()` remains allocated
    /// for the lifetime of the CUDA context, or until it is explicitly released
    /// by a call to `free()`. It can be used by any other CUDA threads
    /// even from subsequent kernel launches.
    ///
    /// Sources:
    /// [Programming Guide](https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#dynamic-global-memory-allocation-and-operations),
    /// [PTX Interoperability](https://docs.nvidia.com/cuda/ptx-writers-guide-to-interoperability/index.html#system-calls).
    // FIXME(denzp): assign `malloc` and `nothrow` attributes.
    pub fn malloc(size: usize) -> *mut c_void;

    /// Free previously dynamically allocated memory.
    ///
    /// The CUDA in-kernel `free()` function deallocates the memory pointed to by `ptr`,
    /// which must have been returned by a previous call to `malloc()`. If `ptr` is NULL,
    /// the call to `free()` is ignored.
    ///
    /// Any CUDA thread may free memory allocated by another thread, but care should be taken
    /// to ensure that the same pointer is not freed more than once. Repeated calls to `free()`
    /// with the same `ptr` has undefined behavior.
    ///
    /// Sources:
    /// [Programming Guide](https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#dynamic-global-memory-allocation-and-operations),
    /// [PTX Interoperability](https://docs.nvidia.com/cuda/ptx-writers-guide-to-interoperability/index.html#system-calls).
    // FIXME(denzp): assign `nothrow` attribute.
    pub fn free(ptr: *mut c_void);

    // Internal declaration of the syscall. Exported variant has
    // the `char_size` parameter set to `1` (single char size in bytes).
    fn __assertfail(
        message: *const u8,
        file: *const u8,
        line: u32,
        function: *const u8,
        char_size: usize,
    );
}

/// Syscall to be used whenever the *assert expression produces a `false` value*.
///
/// Syscall arguments:
/// * `message`: The pointer to the string that should be output.
/// * `file`: The pointer to the file name string associated with the assert.
/// * `line`: The line number associated with the assert.
/// * `function`: The pointer to the function name string associated with the assert.
///
/// Source:
/// [PTX Interoperability](https://docs.nvidia.com/cuda/ptx-writers-guide-to-interoperability/index.html#system-calls).
#[inline]
pub unsafe fn __assert_fail(message: *const u8, file: *const u8, line: u32, function: *const u8) {
    __assertfail(message, file, line, function, 1)
}