j2k-cuda-runtime 0.7.1

CUDA codec engine and Driver API runtime for j2k device adapters
Documentation 
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

// SPDX-License-Identifier: MIT OR Apache-2.0

use crate::{
    context::CudaContext,
    driver::CuFunction,
    error::CudaError,
    execution::cuda_kernel_param,
    kernels::{
        htj2k_codeblock_sample_launch_geometry, htj2k_encode_codeblock_launch_geometry, CudaKernel,
    },
    memory::CudaDeviceBuffer,
};

use super::types::{CudaHtj2kEncodeCodeblocksLaunch, CudaHtj2kEncodeMultiInputLaunch};

impl CudaContext {
    pub(super) fn launch_htj2k_encode_codeblocks(
        &self,
        request: &CudaHtj2kEncodeCodeblocksLaunch<'_>,
    ) -> Result<(), CudaError> {
        let function = self.htj2k_encode_kernel_function(CudaKernel::Htj2kEncodeCodeblocks)?;
        let mut coefficients_ptr = request.coefficients.device_ptr();
        let mut output_ptr = request.output.device_ptr();
        let mut jobs_ptr = request.jobs.device_ptr();
        let mut vlc_table0_ptr = request.tables.vlc_table0.device_ptr();
        let mut vlc_table1_ptr = request.tables.vlc_table1.device_ptr();
        let mut uvlc_table_ptr = request.tables.uvlc_table.device_ptr();
        let mut statuses_ptr = request.statuses.device_ptr();
        let mut job_count_u64 =
            u64::try_from(request.job_count).map_err(|_| CudaError::LengthTooLarge {
                len: request.job_count,
            })?;
        let mut params = cuda_kernel_params!(
            coefficients_ptr,
            output_ptr,
            jobs_ptr,
            vlc_table0_ptr,
            vlc_table1_ptr,
            uvlc_table_ptr,
            statuses_ptr,
            job_count_u64
        );
        let geometry = htj2k_encode_codeblock_launch_geometry(request.job_count).ok_or(
            CudaError::LengthTooLarge {
                len: request.job_count,
            },
        )?;
        self.launch_kernel_async(function, geometry, &mut params)
    }

    pub(super) fn launch_htj2k_encode_codeblocks_multi_input(
        &self,
        request: &CudaHtj2kEncodeMultiInputLaunch<'_>,
    ) -> Result<(), CudaError> {
        self.launch_htj2k_encode_multi_input_kernel(
            CudaKernel::Htj2kEncodeCodeblocksMultiInput,
            request,
        )
    }

    pub(super) fn launch_htj2k_encode_codeblocks_multi_input_cleanup(
        &self,
        request: &CudaHtj2kEncodeMultiInputLaunch<'_>,
    ) -> Result<(), CudaError> {
        self.launch_htj2k_encode_multi_input_kernel(
            CudaKernel::Htj2kEncodeCodeblocksMultiInputCleanup,
            request,
        )
    }

    pub(super) fn launch_htj2k_encode_codeblocks_multi_input_cleanup_64(
        &self,
        request: &CudaHtj2kEncodeMultiInputLaunch<'_>,
    ) -> Result<(), CudaError> {
        self.launch_htj2k_encode_multi_input_kernel(
            CudaKernel::Htj2kEncodeCodeblocksMultiInputCleanup64,
            request,
        )
    }

    fn launch_htj2k_encode_multi_input_kernel(
        &self,
        kernel: CudaKernel,
        request: &CudaHtj2kEncodeMultiInputLaunch<'_>,
    ) -> Result<(), CudaError> {
        let function = self.htj2k_encode_kernel_function(kernel)?;
        let mut output_ptr = request.output.device_ptr();
        let mut jobs_ptr = request.jobs.device_ptr();
        let mut vlc_table0_ptr = request.tables.vlc_table0.device_ptr();
        let mut vlc_table1_ptr = request.tables.vlc_table1.device_ptr();
        let mut uvlc_table_ptr = request.tables.uvlc_table.device_ptr();
        let mut statuses_ptr = request.statuses.device_ptr();
        let mut job_count_u64 =
            u64::try_from(request.job_count).map_err(|_| CudaError::LengthTooLarge {
                len: request.job_count,
            })?;
        let mut params = cuda_kernel_params!(
            output_ptr,
            jobs_ptr,
            vlc_table0_ptr,
            vlc_table1_ptr,
            uvlc_table_ptr,
            statuses_ptr,
            job_count_u64
        );
        let geometry = htj2k_encode_codeblock_launch_geometry(request.job_count).ok_or(
            CudaError::LengthTooLarge {
                len: request.job_count,
            },
        )?;
        self.launch_kernel_async(function, geometry, &mut params)
    }

    pub(crate) fn launch_htj2k_compact_codeblocks(
        &self,
        scratch: &CudaDeviceBuffer,
        compact: &CudaDeviceBuffer,
        jobs: &CudaDeviceBuffer,
        job_count: usize,
    ) -> Result<(), CudaError> {
        let function = self.htj2k_encode_kernel_function(CudaKernel::Htj2kCompactCodeblocks)?;
        let mut scratch_ptr = scratch.device_ptr();
        let mut compact_ptr = compact.device_ptr();
        let mut jobs_ptr = jobs.device_ptr();
        let mut job_count_u64 =
            u64::try_from(job_count).map_err(|_| CudaError::LengthTooLarge { len: job_count })?;
        let mut params = cuda_kernel_params!(scratch_ptr, compact_ptr, jobs_ptr, job_count_u64);
        let geometry = htj2k_codeblock_sample_launch_geometry(job_count)
            .ok_or(CudaError::LengthTooLarge { len: job_count })?;
        self.launch_kernel_async(function, geometry, &mut params)
    }

    fn htj2k_encode_kernel_function(&self, kernel: CudaKernel) -> Result<CuFunction, CudaError> {
        if kernel.is_htj2k_encode_codeblock_stage() {
            self.inner.cuda_oxide_htj2k_encode_kernel_function(kernel)
        } else {
            self.inner.cuda_oxide_j2k_encode_kernel_function(kernel)
        }
    }
}