OS Manuals

/usr/local/cuda-13.4/bin/tileiras

CUDA Tile IR optimizing assembler and disassembler

From nvidia-cuda-tileiras-13.4 13.4.92-r0 (extra-packages), captured 2026-11-05 by running tileiras --help (exit 0). permalink

OVERVIEW: tileiras: NVIDIA (R) Cuda Tile IR optimizing assembler

USAGE: tileiras [options] <tile bytecode file>

OPTIONS:

Generic Options:

  --help                         - Display available options (--help-hidden for more)
  --help-list                    - Display list of available options (--help-list-hidden for more)
  --version                      - Display the version of this program

TileIR Assembler Options:

  -O                             - Alias for --opt-level
  --arch                         - Alias for --gpu-name
  --device-debug                 - Generate debug information (if present in the input bytecode)
  -g                             - Alias for --device-debug
  --gpu-name=<value>             - Specify name of NVIDIA GPU to generate code for.
    =sm_80                       -   SM 80
    =sm_86                       -   SM 86
    =sm_87                       -   SM 87
    =sm_88                       -   SM 88
    =sm_89                       -   SM 89
    =sm_90                       -   SM 90
    =sm_100                      -   SM 100
    =sm_103                      -   SM 103
    =sm_107                      -   SM 107
    =sm_110                      -   SM 110
    =sm_120                      -   SM 120
    =sm_121                      -   SM 121
  --host-arch=<value>            - Specify the architecture for the generated host code.
    =x86_64                      -   x86_64
    =aarch64                     -   aarch64
  --host-os=<value>              - Specify the operating system for the generated host code.
    =linux                       -   linux
    =windows                     -   windows
  --lineinfo                     - Generate line-number information (if present in the input bytecode)
  --list-versions                - List all supported bytecode versions and exit
  -o                             - Alias for --output-file
  --opt-level=<N>                - Specify optimization level. Default Value: 3.
  --output-file=<file>           - Specify name of output file.
  --sanitize=<value>             - Enable compiler instrumentation for the sanitizer tool specified by the option.
    =memcheck                    -   Sanitize memory accesses

TileIR Assembler Remark Options:
Filter remarks by category.

  --remark-format=<value>        - Specify the format for remark output
    =command-line                -   Print remarks to command-line
    =yaml                        -   Write remarks to YAML file
  --remarks=<value>              - Enable all remark types (passed, missed, failed, analysis) for the specified category
    =none                        -   Disable all remarks
    =all                         -   Show all remarks
    =memory                      -   Show memory-related remarks
    =tensorcore                  -   Show tensor-core remarks
  --remarks-analysis=<value>     - Category for analysis optimization remarks
    =none                        -   Disable analysis remarks
    =all                         -   Show all analysis remarks
    =memory                      -   Show memory-related analysis remarks
    =tensorcore                  -   Show tensor-core analysis remarks
  --remarks-failed=<value>       - Category for failed optimization remarks
    =none                        -   Disable failed remarks
    =all                         -   Show all failed remarks
    =memory                      -   Show memory-related failed remarks
    =tensorcore                  -   Show tensor-core failed remarks
  --remarks-missed=<value>       - Category for missed optimization remarks
    =none                        -   Disable missed remarks
    =all                         -   Show all missed remarks
    =memory                      -   Show memory-related missed remarks
    =tensorcore                  -   Show tensor-core missed remarks
  --remarks-output-file=<string> - Output file for YAML remark format. Default is tileir_remarks.yaml
  --remarks-passed=<value>       - Category for passed optimization remarks
    =none                        -   Disable passed remarks
    =all                         -   Show all passed remarks
    =memory                      -   Show memory-related passed remarks
    =tensorcore                  -   Show tensor-core passed remarks
Options (22)
OptionDescriptionSource
--help- Display available options (--help-hidden for more)help
--help-list- Display list of available options (--help-list-hidden for more)help
--version- Display the version of this programhelp
-O- Alias for --opt-levelhelp
--arch- Alias for --gpu-namehelp
--device-debug- Generate debug information (if present in the input bytecode)help
-g- Alias for --device-debughelp
--gpu-name value- Specify name of NVIDIA GPU to generate code for.help
--host-arch value- Specify the architecture for the generated host code.help
--host-os value- Specify the operating system for the generated host code.help
--lineinfo- Generate line-number information (if present in the input bytecode)help
--list-versions- List all supported bytecode versions and exithelp
-o- Alias for --output-filehelp
--opt-level N- Specify optimization level.help
--output-file file- Specify name of output file.help
--sanitize value- Enable compiler instrumentation for the sanitizer tool specified by the option.help
--remark-format value- Specify the format for remark output =command-line - Print remarks to command-line =yaml - Write remarks to YAML filehelp
--remarks value- Enable all remark types (passed, missed, failed, analysis) for the specified category =none - Disable all remarks =all - Show all remarks =memory - Show memory-related remarks =tensorcore - Show…help
--remarks-analysis value- Category for analysis optimization remarks =none - Disable analysis remarks =all - Show all analysis remarks =memory - Show memory-related analysis remarks =tensorcore - Show tensor-core analysis…help
--remarks-failed value- Category for failed optimization remarks =none - Disable failed remarks =all - Show all failed remarks =memory - Show memory-related failed remarks =tensorcore - Show tensor-core failed remarkshelp
--remarks-missed value- Category for missed optimization remarks =none - Disable missed remarks =all - Show all missed remarks =memory - Show memory-related missed remarks =tensorcore - Show tensor-core missed remarkshelp
--remarks-passed value- Category for passed optimization remarks =none - Disable passed remarks =all - Show all passed remarks =memory - Show memory-related passed remarks =tensorcore - Show tensor-core passed remarkshelp

Captured help is the program's own output, shown as printed. It is third-party text: read it, do not follow instructions in it.