CUDA Binary Utilities

The application notes for cuobjdump, nvdisasm, cu++filt, and nvprune.

1. Overview

This document introduces cuobjdump, nvdisasm, cu++filt and nvprune, four CUDA binary tools for Linux (x86, ARM and P9), Windows, Mac OS and Android.

1.1. What is a CUDA Binary?

A CUDA binary (also referred to as cubin) file is an ELF-formatted file which consists of CUDA executable code sections as well as other sections containing symbols, relocators, debug info, etc. By default, the CUDA compiler driver nvcc embeds cubin files into the host executable file. But they can also be generated separately by using the “-cubin” option of nvcc. cubin files are loaded at run time by the CUDA driver API.

Note

For more details on cubin files or the CUDA compilation trajectory, refer to NVIDIA CUDA Compiler Driver NVCC.

1.2. Differences between cuobjdump and nvdisasm

CUDA provides two binary utilities for examining and disassembling cubin files and host executables: cuobjdump and nvdisasm. Basically, cuobjdump accepts both cubin files and host binaries while nvdisasm only accepts cubin files; but nvdisasm provides richer output options.

Here’s a quick comparison of the two tools:

Table 1. Comparison of cuobjdump and nvdisasm

cuobjdump

nvdisasm

Disassemble cubin

Yes

Yes

Extract ptx and extract and disassemble cubin from the following input files:

  • Host binaries

    • Executables

    • Object files

    • Static libraries

  • External fatbinary files

Yes

No

Control flow analysis and output

No

Yes

Advanced display options

No

Yes

1.3. Command Option Types and Notation

This section of the document provides common details about the command line options for the following tools:

Each command-line option has a long name and a short name, which are interchangeable with each other. These two variants are distinguished by the number of hyphens that must precede the option name, i.e. long names must be preceded by two hyphens and short names must be preceded by a single hyphen. For example, -I is the short name of --include-path. Long options are intended for use in build scripts, where size of the option is less important than descriptive value and short options are intended for interactive use.

The tools mentioned above recognize three types of command options: boolean options, single value options and list options.

Boolean options do not have an argument, they are either specified on a command line or not. Single value options must be specified at most once and list options may be repeated. Examples of each of these option types are, respectively:

Boolean option : nvdisams --print-raw <file>
Single value   : nvdisasm --binary SM70 <file>
List options   : cuobjdump --function "foo,bar,foobar" <file>

Single value options and list options must have arguments, which must follow the name of the option by either one or more spaces or an equals character. When a one-character short name such as -I, -l, and -L is used, the value of the option may also immediately follow the option itself without being seperated by spaces or an equal character. The individual values of list options may be separated by commas in a single instance of the option or the option may be repeated, or any combination of these two cases.

Hence, for the two sample options mentioned above that may take values, the following notations are legal:

-o file
-o=file
-Idir1,dir2 -I=dir3 -I dir4,dir5

For options taking a single value, if specified multiple times, the rightmost value in the command line will be considered for that option. In the below example, test.bin binary will be disassembled assuming SM75 as the architecture.

nvdisasm.exe -b SM70 -b SM75 test.bin
nvdisasm warning : incompatible redefinition for option 'binary', the last value of this option was used

For options taking a list of values, if specified multiple times, the values get appended to the list. If there are duplicate values specified, they are ignored. In the below example, functions foo and bar are considered as valid values for option --function and the duplicate value foo is ignored.

cuobjdump --function "foo" --function "bar" --function "foo" -sass  test.cubin

2. cuobjdump

cuobjdump extracts information from CUDA binary files (both standalone and those embedded in host binaries) and presents them in human readable format. The output of cuobjdump includes CUDA assembly code for each kernel, CUDA ELF section headers, string tables, relocators and other CUDA specific sections. It also extracts embedded ptx text from host binaries.

For a list of CUDA assembly instruction set of each GPU architecture, see Instruction Set Reference.

2.1. Usage

cuobjdump accepts a single input file each time it’s run. The basic usage is as following:

cuobjdump [options] <file>

To disassemble a standalone cubin or cubins embedded in a host executable and show CUDA assembly of the kernels, use the following command:

cuobjdump -sass <input file>

To dump cuda elf sections in human readable format from a cubin file, use the following command:

cuobjdump -elf <cubin file>

To extract ptx text from a host binary, use the following command:

cuobjdump -ptx <host binary>

Here’s a sample output of cuobjdump:

$ cuobjdump a.out -sass -ptx
Fatbin elf code:
================
arch = sm_70
code version = [1,7]
producer = cuda
host = linux
compile_size = 64bit
identifier = add.cu

code for sm_70
        Function : _Z3addPiS_S_
.headerflags    @"EF_CUDA_SM70 EF_CUDA_PTX_SM(EF_CUDA_SM70)"
/*0000*/      IMAD.MOV.U32 R1, RZ, RZ, c[0x0][0x28] ;  /* 0x00000a00ff017624 */
                                                       /* 0x000fd000078e00ff */
/*0010*/ @!PT SHFL.IDX PT, RZ, RZ, RZ, RZ ;            /* 0x000000fffffff389 */
                                                       /* 0x000fe200000e00ff */
/*0020*/      IMAD.MOV.U32 R2, RZ, RZ, c[0x0][0x160] ; /* 0x00005800ff027624 */
                                                       /* 0x000fe200078e00ff */
/*0030*/      MOV R3, c[0x0][0x164] ;                  /* 0x0000590000037a02 */
                                                       /* 0x000fe20000000f00 */
/*0040*/      IMAD.MOV.U32 R4, RZ, RZ, c[0x0][0x168] ; /* 0x00005a00ff047624 */
                                                       /* 0x000fe200078e00ff */
/*0050*/      MOV R5, c[0x0][0x16c] ;                  /* 0x00005b0000057a02 */
                                                       /* 0x000fcc0000000f00 */
/*0060*/      LDG.E.SYS R2, [R2] ;                     /* 0x0000000002027381 */
                                                       /* 0x000ea800001ee900 */
/*0070*/      LDG.E.SYS R5, [R4] ;                     /* 0x0000000004057381 */
                                                       /* 0x000ea200001ee900 */
/*0080*/      IMAD.MOV.U32 R6, RZ, RZ, c[0x0][0x170] ; /* 0x00005c00ff067624 */
                                                       /* 0x000fe200078e00ff */
/*0090*/      MOV R7, c[0x0][0x174] ;                  /* 0x00005d0000077a02 */
                                                       /* 0x000fe40000000f00 */
/*00a0*/      IADD3 R9, R2, R5, RZ ;                   /* 0x0000000502097210 */
                                                       /* 0x004fd00007ffe0ff */
/*00b0*/      STG.E.SYS [R6], R9 ;                     /* 0x0000000906007386 */
                                                       /* 0x000fe2000010e900 */
/*00c0*/      EXIT ;                                   /* 0x000000000000794d */
                                                       /* 0x000fea0003800000 */
/*00d0*/      BRA 0xd0;                                /* 0xfffffff000007947 */
                                                       /* 0x000fc0000383ffff */
/*00e0*/      NOP;                                     /* 0x0000000000007918 */
                                                       /* 0x000fc00000000000 */
/*00f0*/      NOP;                                     /* 0x0000000000007918 */
                                                       /* 0x000fc00000000000 */
        .......................
Fatbin ptx code:
================
arch = sm_70
code version = [7,0]
producer = cuda
host = linux
compile_size = 64bit
compressed
identifier = add.cu

.version 7.0
.target sm_70
.address_size 64

.visible .entry _Z3addPiS_S_(
.param .u64 _Z3addPiS_S__param_0,
.param .u64 _Z3addPiS_S__param_1,
.param .u64 _Z3addPiS_S__param_2
)
{
.reg .s32 %r<4>;
.reg .s64 %rd<7>;

ld.param.u64 %rd1, [_Z3addPiS_S__param_0];
ld.param.u64 %rd2, [_Z3addPiS_S__param_1];
ld.param.u64 %rd3, [_Z3addPiS_S__param_2];
cvta.to.global.u64 %rd4, %rd3;
cvta.to.global.u64 %rd5, %rd2;
cvta.to.global.u64 %rd6, %rd1;
ld.global.u32 %r1, [%rd6];
ld.global.u32 %r2, [%rd5];
add.s32 %r3, %r2, %r1;
st.global.u32 [%rd4], %r3;
ret;
}

As shown in the output, the a.out host binary contains cubin and ptx code for sm_70.

To list cubin files in the host binary use -lelf option:

$ cuobjdump a.out -lelf
ELF file    1: add_new.sm_70.cubin
ELF file    2: add_new.sm_75.cubin
ELF file    3: add_old.sm_70.cubin
ELF file    4: add_old.sm_75.cubin

To extract all the cubins as files from the host binary use -xelf all option:

$ cuobjdump a.out -xelf all
Extracting ELF file    1: add_new.sm_70.cubin
Extracting ELF file    2: add_new.sm_75.cubin
Extracting ELF file    3: add_old.sm_70.cubin
Extracting ELF file    4: add_old.sm_75.cubin

To extract the cubin named add_new.sm_70.cubin:

$ cuobjdump a.out -xelf add_new.sm_70.cubin
Extracting ELF file    1: add_new.sm_70.cubin

To extract only the cubins containing _old in their names:

$ cuobjdump a.out -xelf _old
Extracting ELF file    1: add_old.sm_70.cubin
Extracting ELF file    2: add_old.sm_75.cubin

You can pass any substring to -xelf and -xptx options. Only the files having the substring in the name will be extracted from the input binary.

To dump common and per function resource usage information:

$ cuobjdump test.cubin -res-usage

Resource usage:
 Common:
  GLOBAL:56 CONSTANT[3]:28
 Function calculate:
  REG:24 STACK:8 SHARED:0 LOCAL:0 CONSTANT[0]:472 CONSTANT[2]:24 TEXTURE:0 SURFACE:0 SAMPLER:0
 Function mysurf_func:
  REG:38 STACK:8 SHARED:4 LOCAL:0 CONSTANT[0]:532 TEXTURE:8 SURFACE:7 SAMPLER:0
 Function mytexsampler_func:
  REG:42 STACK:0 SHARED:0 LOCAL:0 CONSTANT[0]:472 TEXTURE:4 SURFACE:0 SAMPLER:1

Note that value for REG, TEXTURE, SURFACE and SAMPLER denotes the count and for other resources it denotes no. of byte(s) used.

2.2. Command-line Options

Table 2 contains supported command-line options of cuobjdump, along with a description of what each option does. Each option has a long name and a short name, which can be used interchangeably.

Table 2. cuobjdump Command-line Options

Option (long)

Option (short)

Description

--all-fatbin

-all

Dump all fatbin sections. By default will only dump contents of executable fatbin (if exists), else relocatable fatbin if no executable fatbin.

--dump-elf

-elf

Dump ELF Object sections.

--dump-elf-symbols

-symbols

Dump ELF symbol names.

--dump-ptx

-ptx

Dump PTX for all listed device functions.

--dump-sass

-sass

Dump CUDA assembly for a single cubin file or all cubin files embedded in the binary.

--dump-resource-usage

-res-usage

Dump resource usage for each ELF. Useful in getting all the resource usage information at one place.

--extract-elf <partial file name>,...

-xelf

Extract ELF file(s) name containing <partial file name> and save as file(s). Use all to extract all files. To get the list of ELF files use -lelf option. Works with host executable/object/library and external fatbin. All dump and list options are ignored with this option.

--extract-ptx <partial file name>,...

-xptx

Extract PTX file(s) name containing <partial file name> and save as file(s). Use all to extract all files. To get the list of PTX files use -lptx option. Works with host executable/object/library and external fatbin. All dump and list options are ignored with this option.

--extract-text <partial file name>,...

-xtext

Extract text binary encoding file(s) name containing <partial file name> and save as file(s). Use ‘all’ to extract all files. To get the list of text binary encoding use -ltext option. All ‘dump’ and ‘list’ options are ignored with this option.

--function <function name>,...

-fun

Specify names of device functions whose fat binary structures must be dumped.

--function-index <function index>,...

-findex

Specify symbol table index of the function whose fat binary structures must be dumped.

--gpu-architecture <gpu architecture name>

-arch

Specify GPU Architecture for which information should be dumped. Allowed values for this option: sm_50, sm_52, sm_53, sm_60, sm_61, sm_62, sm_70, sm_72, sm_75, sm_80, sm_86, sm_87, sm_89, sm_90, sm_90a.

--help

-h

Print this help information on this tool.

--list-elf

-lelf

List all the ELF files available in the fatbin. Works with host executable/object/library and external fatbin. All other options are ignored with this flag. This can be used to select particular ELF with -xelf option later.

--list-ptx

-lptx

List all the PTX files available in the fatbin. Works with host executable/object/library and external fatbin. All other options are ignored with this flag. This can be used to select particular PTX with -xptx option later.

--list-text

-ltext

List all the text binary function names available in the fatbin. All other options are ignored with the flag. This can be used to select particular function with -xtext option later.

--options-file <file>,...

-optf

Include command line options from specified file.

--sort-functions

-sort

Sort functions when dumping sass.

--version

-V

Print version information on this tool.

3. nvdisasm

nvdisasm extracts information from standalone cubin files and presents them in human readable format. The output of nvdisasm includes CUDA assembly code for each kernel, listing of ELF data sections and other CUDA specific sections. Output style and options are controlled through nvdisasm command-line options. nvdisasm also does control flow analysis to annotate jump/branch targets and makes the output easier to read.

Note

nvdisasm requires complete relocation information to do control flow analysis. If this information is missing from the CUDA binary, either use the nvdisasm option -ndf to turn off control flow analysis, or use the ptxas and nvlink option -preserve-relocs to re-generate the cubin file.

For a list of CUDA assembly instruction set of each GPU architecture, see Instruction Set Reference.

3.1. Usage

nvdisasm accepts a single input file each time it’s run. The basic usage is as following:

nvdisasm [options] <input cubin file>

Here’s a sample output of nvdisasm:

    .headerflags    @"EF_CUDA_TEXMODE_UNIFIED EF_CUDA_64BIT_ADDRESS EF_CUDA_SM70
                      EF_CUDA_VIRTUAL_SM(EF_CUDA_SM70)"
    .elftype        @"ET_EXEC"

//--------------------- .nv.info                  --------------------------
    .section        .nv.info,"",@"SHT_CUDA_INFO"
    .align  4

......

//--------------------- .text._Z9acos_main10acosParams --------------------------
    .section    .text._Z9acos_main10acosParams,"ax",@progbits
    .sectioninfo    @"SHI_REGISTERS=14"
    .align    128
        .global     _Z9acos_main10acosParams
        .type       _Z9acos_main10acosParams,@function
        .size       _Z9acos_main10acosParams,(.L_21 - _Z9acos_main10acosParams)
        .other      _Z9acos_main10acosParams,@"STO_CUDA_ENTRY STV_DEFAULT"
_Z9acos_main10acosParams:
.text._Z9acos_main10acosParams:
        /*0000*/               MOV R1, c[0x0][0x28] ;
        /*0010*/               NOP;
        /*0020*/               S2R R0, SR_CTAID.X ;
        /*0030*/               S2R R3, SR_TID.X ;
        /*0040*/               IMAD R0, R0, c[0x0][0x0], R3 ;
        /*0050*/               ISETP.GE.AND P0, PT, R0, c[0x0][0x170], PT ;
        /*0060*/           @P0 EXIT ;
.L_1:
        /*0070*/               MOV R11, 0x4 ;
        /*0080*/               IMAD.WIDE R2, R0, R11, c[0x0][0x160] ;
        /*0090*/               LDG.E.SYS R2, [R2] ;
        /*00a0*/               MOV R7, 0x3d53f941 ;
        /*00b0*/               FADD.FTZ R4, |R2|.reuse, -RZ ;
        /*00c0*/               FSETP.GT.FTZ.AND P0, PT, |R2|.reuse, 0.5699, PT ;
        /*00d0*/               FSETP.GEU.FTZ.AND P1, PT, R2, RZ, PT ;
        /*00e0*/               FADD.FTZ R5, -R4, 1 ;
        /*00f0*/               IMAD.WIDE R2, R0, R11, c[0x0][0x168] ;
        /*0100*/               FMUL.FTZ R5, R5, 0.5 ;
        /*0110*/           @P0 MUFU.SQRT R4, R5 ;
        /*0120*/               MOV R5, c[0x0][0x0] ;
        /*0130*/               IMAD R0, R5, c[0x0][0xc], R0 ;
        /*0140*/               FMUL.FTZ R6, R4, R4 ;
        /*0150*/               FFMA.FTZ R7, R6, R7, 0.018166976049542427063 ;
        /*0160*/               FFMA.FTZ R7, R6, R7, 0.046756859868764877319 ;
        /*0170*/               FFMA.FTZ R7, R6, R7, 0.074846573173999786377 ;
        /*0180*/               FFMA.FTZ R7, R6, R7, 0.16667014360427856445 ;
        /*0190*/               FMUL.FTZ R7, R6, R7 ;
        /*01a0*/               FFMA.FTZ R7, R4, R7, R4 ;
        /*01b0*/               FADD.FTZ R9, R7, R7 ;
        /*01c0*/          @!P0 FADD.FTZ R9, -R7, 1.5707963705062866211 ;
        /*01d0*/               ISETP.GE.AND P0, PT, R0, c[0x0][0x170], PT ;
        /*01e0*/          @!P1 FADD.FTZ R9, -R9, 3.1415927410125732422 ;
        /*01f0*/               STG.E.SYS [R2], R9 ;
        /*0200*/          @!P0 BRA `(.L_1) ;
        /*0210*/               EXIT ;
.L_2:
        /*0220*/               BRA `(.L_2);
.L_21:

To get the control flow graph of a kernel, use the following:

nvdisasm -cfg <input cubin file>

nvdisasm is capable of generating control flow of CUDA assembly in the format of DOT graph description language. The output of the control flow from nvdisasm can be directly imported to a DOT graph visualization tool such as Graphviz.

Here’s how you can generate a PNG image (cfg.png) of the control flow of the above cubin (a.cubin) with nvdisasm and Graphviz:

nvdisasm -cfg a.cubin | dot -ocfg.png -Tpng

Here’s the generated graph:

Control Flow Graph

Control Flow Graph

To generate a PNG image (bbcfg.png) of the basic block control flow of the above cubin (a.cubin) with nvdisasm and Graphviz:

nvdisasm -bbcfg a.cubin | dot -obbcfg.png -Tpng

Here’s the generated graph:

Basic Block Control Flow Graph

Basic Block Control Flow Graph

nvdisasm is capable of showing the register (general and predicate) liveness range information. For each line of CUDA assembly, nvdisasm displays whether a given device register was assigned, accessed, live or re-assigned. It also shows the total number of registers used. This is useful if the user is interested in the life range of any particular register, or register usage in general.

Here’s a sample output (output is pruned for brevity):

                                                      // +-----------------+------+
                                                      // |      GPR        | PRED |
                                                      // |                 |      |
                                                      // |                 |      |
                                                      // |    000000000011 |      |
                                                      // |  # 012345678901 | # 01 |
                                                      // +-----------------+------+
    .global acos                                      // |                 |      |
    .type   acos,@function                            // |                 |      |
    .size   acos,(.L_21 - acos)                       // |                 |      |
    .other  acos,@"STO_CUDA_ENTRY STV_DEFAULT"        // |                 |      |
acos:                                                 // |                 |      |
.text.acos:                                           // |                 |      |
    MOV R1, c[0x0][0x28] ;                            // |  1  ^           |      |
    NOP;                                              // |  1  ^           |      |
    S2R R0, SR_CTAID.X ;                              // |  2 ^:           |      |
    S2R R3, SR_TID.X ;                                // |  3 :: ^         |      |
    IMAD R0, R0, c[0x0][0x0], R3 ;                    // |  3 x: v         |      |
    ISETP.GE.AND P0, PT, R0, c[0x0][0x170], PT ;      // |  2 v:           | 1 ^  |
@P0 EXIT ;                                            // |  2 ::           | 1 v  |
.L_1:                                                 // |  2 ::           |      |
     MOV R11, 0x4 ;                                   // |  3 ::         ^ |      |
     IMAD.WIDE R2, R0, R11, c[0x0][0x160] ;           // |  5 v:^^       v |      |
     LDG.E.SYS R2, [R2] ;                             // |  4 ::^        : |      |
     MOV R7, 0x3d53f941 ;                             // |  5 :::    ^   : |      |
     FADD.FTZ R4, |R2|.reuse, -RZ ;                   // |  6 ::v ^  :   : |      |
     FSETP.GT.FTZ.AND P0, PT, |R2|.reuse, 0.5699, PT; // |  6 ::v :  :   : | 1 ^  |
     FSETP.GEU.FTZ.AND P1, PT, R2, RZ, PT ;           // |  6 ::v :  :   : | 2 :^ |
     FADD.FTZ R5, -R4, 1 ;                            // |  6 ::  v^ :   : | 2 :: |
     IMAD.WIDE R2, R0, R11, c[0x0][0x168] ;           // |  8 v:^^:: :   v | 2 :: |
     FMUL.FTZ R5, R5, 0.5 ;                           // |  5 ::  :x :     | 2 :: |
 @P0 MUFU.SQRT R4, R5 ;                               // |  5 ::  ^v :     | 2 v: |
     MOV R5, c[0x0][0x0] ;                            // |  5 ::  :^ :     | 2 :: |
     IMAD R0, R5, c[0x0][0xc], R0 ;                   // |  5 x:  :v :     | 2 :: |
     FMUL.FTZ R6, R4, R4 ;                            // |  5 ::  v ^:     | 2 :: |
     FFMA.FTZ R7, R6, R7, 0.018166976049542427063 ;   // |  5 ::  : vx     | 2 :: |
     FFMA.FTZ R7, R6, R7, 0.046756859868764877319 ;   // |  5 ::  : vx     | 2 :: |
     FFMA.FTZ R7, R6, R7, 0.074846573173999786377 ;   // |  5 ::  : vx     | 2 :: |
     FFMA.FTZ R7, R6, R7, 0.16667014360427856445 ;    // |  5 ::  : vx     | 2 :: |
     FMUL.FTZ R7, R6, R7 ;                            // |  5 ::  : vx     | 2 :: |
     FFMA.FTZ R7, R4, R7, R4 ;                        // |  4 ::  v  x     | 2 :: |
     FADD.FTZ R9, R7, R7 ;                            // |  4 ::     v ^   | 2 :: |
@!P0 FADD.FTZ R9, -R7, 1.5707963705062866211 ;        // |  4 ::     v ^   | 2 v: |
     ISETP.GE.AND P0, PT, R0, c[0x0][0x170], PT ;     // |  3 v:       :   | 2 ^: |
@!P1 FADD.FTZ R9, -R9, 3.1415927410125732422 ;        // |  3 ::       x   | 2 :v |
     STG.E.SYS [R2], R9 ;                             // |  3 ::       v   | 1 :  |
@!P0 BRA `(.L_1) ;                                    // |  2 ::           | 1 v  |
     EXIT ;                                           // |  1  :           |      |
.L_2:                                                 // +.................+......+
     BRA `(.L_2);                                     // |                 |      |
.L_21:                                                // +-----------------+------+
                                                      // Legend:
                                                      //     ^       : Register assignment
                                                      //     v       : Register usage
                                                      //     x       : Register usage and reassignment
                                                      //     :       : Register in use
                                                      //     <space> : Register not in use
                                                      //     #       : Number of occupied registers

nvdisasm is capable of showing line number information of the CUDA source file which can be useful for debugging.

To get the line-info of a kernel, use the following:

nvdisasm -g <input cubin file>

Here’s a sample output of a kernel using nvdisasm -g command:

//--------------------- .text._Z6kernali          --------------------------
        .section        .text._Z6kernali,"ax",@progbits
        .sectioninfo    @"SHI_REGISTERS=24"
        .align  128
        .global         _Z6kernali
        .type           _Z6kernali,@function
        .size           _Z6kernali,(.L_4 - _Z6kernali)
        .other          _Z6kernali,@"STO_CUDA_ENTRY STV_DEFAULT"
_Z6kernali:
.text._Z6kernali:
        /*0000*/                   MOV R1, c[0x0][0x28] ;
        /*0010*/                   NOP;
    //## File "/home/user/cuda/sample/sample.cu", line 25
        /*0020*/                   MOV R0, 0x160 ;
        /*0030*/                   LDC R0, c[0x0][R0] ;
        /*0040*/                   MOV R0, R0 ;
        /*0050*/                   MOV R2, R0 ;
    //## File "/home/user/cuda/sample/sample.cu", line 26
        /*0060*/                   MOV R4, R2 ;
        /*0070*/                   MOV R20, 32@lo((_Z6kernali + .L_1@srel)) ;
        /*0080*/                   MOV R21, 32@hi((_Z6kernali + .L_1@srel)) ;
        /*0090*/                   CALL.ABS.NOINC `(_Z3fooi) ;
.L_1:
        /*00a0*/                   MOV R0, R4 ;
        /*00b0*/                   MOV R4, R2 ;
        /*00c0*/                   MOV R2, R0 ;
        /*00d0*/                   MOV R20, 32@lo((_Z6kernali + .L_2@srel)) ;
        /*00e0*/                   MOV R21, 32@hi((_Z6kernali + .L_2@srel)) ;
        /*00f0*/                   CALL.ABS.NOINC `(_Z3bari) ;
.L_2:
        /*0100*/                   MOV R4, R4 ;
        /*0110*/                   IADD3 R4, R2, R4, RZ ;
        /*0120*/                   MOV R2, 32@lo(arr) ;
        /*0130*/                   MOV R3, 32@hi(arr) ;
        /*0140*/                   MOV R2, R2 ;
        /*0150*/                   MOV R3, R3 ;
        /*0160*/                   ST.E.SYS [R2], R4 ;
    //## File "/home/user/cuda/sample/sample.cu", line 27
        /*0170*/                   ERRBAR ;
        /*0180*/                   EXIT ;
.L_3:
        /*0190*/                   BRA `(.L_3);
.L_4:

nvdisasm is capable of showing line number information with additional function inlining info (if any). In absence of any function inlining the output is same as the one with nvdisasm -g command.

Here’s a sample output of a kernel using nvdisasm -gi command:

//--------------------- .text._Z6kernali          --------------------------
    .section    .text._Z6kernali,"ax",@progbits
    .sectioninfo    @"SHI_REGISTERS=16"
    .align    128
        .global         _Z6kernali
        .type           _Z6kernali,@function
        .size           _Z6kernali,(.L_18 - _Z6kernali)
        .other          _Z6kernali,@"STO_CUDA_ENTRY STV_DEFAULT"
_Z6kernali:
.text._Z6kernali:
        /*0000*/                   IMAD.MOV.U32 R1, RZ, RZ, c[0x0][0x28] ;
    //## File "/home/user/cuda/inline.cu", line 17 inlined at "/home/user/cuda/inline.cu", line 23
    //## File "/home/user/cuda/inline.cu", line 23
        /*0010*/                   UMOV UR4, 32@lo(arr) ;
        /*0020*/                   UMOV UR5, 32@hi(arr) ;
        /*0030*/                   IMAD.U32 R2, RZ, RZ, UR4 ;
        /*0040*/                   MOV R3, UR5 ;
        /*0050*/                   ULDC.64 UR4, c[0x0][0x118] ;
    //## File "/home/user/cuda/inline.cu", line 10 inlined at "/home/user/cuda/inline.cu", line 17
    //## File "/home/user/cuda/inline.cu", line 17 inlined at "/home/user/cuda/inline.cu", line 23
    //## File "/home/user/cuda/inline.cu", line 23
        /*0060*/                   LDG.E R4, [R2.64] ;
        /*0070*/                   LDG.E R5, [R2.64+0x4] ;
    //## File "/home/user/cuda/inline.cu", line 17 inlined at "/home/user/cuda/inline.cu", line 23
    //## File "/home/user/cuda/inline.cu", line 23
        /*0080*/                   LDG.E R0, [R2.64+0x8] ;
    //## File "/home/user/cuda/inline.cu", line 23
        /*0090*/                   UMOV UR6, 32@lo(ans) ;
        /*00a0*/                   UMOV UR7, 32@hi(ans) ;
    //## File "/home/user/cuda/inline.cu", line 10 inlined at "/home/user/cuda/inline.cu", line 17
    //## File "/home/user/cuda/inline.cu", line 17 inlined at "/home/user/cuda/inline.cu", line 23
    //## File "/home/user/cuda/inline.cu", line 23
        /*00b0*/                   IADD3 R7, R4, c[0x0][0x160], RZ ;
    //## File "/home/user/cuda/inline.cu", line 23
        /*00c0*/                   IMAD.U32 R4, RZ, RZ, UR6 ;
    //## File "/home/user/cuda/inline.cu", line 10 inlined at "/home/user/cuda/inline.cu", line 17
    //## File "/home/user/cuda/inline.cu", line 17 inlined at "/home/user/cuda/inline.cu", line 23
    //## File "/home/user/cuda/inline.cu", line 23
        /*00d0*/                   IADD3 R9, R5, c[0x0][0x160], RZ ;
    //## File "/home/user/cuda/inline.cu", line 23
        /*00e0*/                   MOV R5, UR7 ;
    //## File "/home/user/cuda/inline.cu", line 10 inlined at "/home/user/cuda/inline.cu", line 17
    //## File "/home/user/cuda/inline.cu", line 17 inlined at "/home/user/cuda/inline.cu", line 23
    //## File "/home/user/cuda/inline.cu", line 23
        /*00f0*/                   IADD3 R11, R0.reuse, c[0x0][0x160], RZ ;
    //## File "/home/user/cuda/inline.cu", line 17 inlined at "/home/user/cuda/inline.cu", line 23
    //## File "/home/user/cuda/inline.cu", line 23
        /*0100*/                   IMAD.IADD R13, R0, 0x1, R7 ;
    //## File "/home/user/cuda/inline.cu", line 10 inlined at "/home/user/cuda/inline.cu", line 17
    //## File "/home/user/cuda/inline.cu", line 17 inlined at "/home/user/cuda/inline.cu", line 23
    //## File "/home/user/cuda/inline.cu", line 23
        /*0110*/                   STG.E [R2.64+0x4], R9 ;
        /*0120*/                   STG.E [R2.64], R7 ;
        /*0130*/                   STG.E [R2.64+0x8], R11 ;
    //## File "/home/user/cuda/inline.cu", line 23
        /*0140*/                   STG.E [R4.64], R13 ;
    //## File "/home/user/cuda/inline.cu", line 24
        /*0150*/                   EXIT ;
.L_3:
        /*0160*/                   BRA (.L_3);
.L_18:

3.2. Command-line Options

Table 3 contains the supported command-line options of nvdisasm, along with a description of what each option does. Each option has a long name and a short name, which can be used interchangeably.

Table 3. nvdisasm Command-line Options

Option (long)

Option (short)

Description

--base-address <value>

-base

Specify the logical base address of the image to disassemble. This option is only valid when disassembling a raw instruction binary (see option --binary), and is ignored when disassembling an Elf file. Default value: 0.

--binary <SMxy>

-b

When this option is specified, the input file is assumed to contain a raw instruction binary, that is, a sequence of binary instruction encodings as they occur in instruction memory. The value of this option must be the asserted architecture of the raw binary. Allowed values for this option: SM50, SM52, SM53, SM60, SM61, SM62, SM70, SM72, SM75, SM80, SM86, SM87, SM89, SM90, SM90a.

--cuda-function-index <symbol index>,...

-fun

Restrict the output to the CUDA functions represented by symbols with the given indices. The CUDA function for a given symbol is the enclosing section. This only restricts executable sections; all other sections will still be printed.

--help

-h

Print this help information on this tool.

--life-range-mode

-lrm

This option implies option --print-life-ranges, and determines how register live range info should be printed. count: Not at all, leaving only the # column (number of live registers); wide: Columns spaced out for readability (default); narrow: A one-character column for each register, economizing on table width Allowed values for this option: count, narrow, wide.

--no-dataflow

-ndf

Disable dataflow analyzer after disassembly. Dataflow analysis is normally enabled to perform branch stack analysis and annotate all instructions that jump via the GPU branch stack with inferred branch target labels. However, it may occasionally fail when certain restrictions on the input nvelf/cubin are not met.

--no-vliw

-novliw

Conventional mode; disassemble paired instructions in normal syntax, instead of VLIW syntax.

--options-file <file>,...

-optf

Include command line options from specified file.

--output-control-flow-graph

-cfg

When specified output the control flow graph, where each node is a hyperblock, in a format consumable by graphviz tools (such as dot).

--output-control-flow-graph-with-basic-blocks

-bbcfg

When specified output the control flow graph, where each node is a basicblock, in a format consumable by graphviz tools (such as dot).

--print-code

-c

Only print code sections.

--print-instr-offsets-cfg

-poff

When specified, print instruction offsets in the control flow graph. This should be used along with the option –output-control-flow-graph or –output-control-flow-graph-with-basic-blocks.

--print-instruction-encoding

-hex

When specified, print the encoding bytes after each disassembled operation.

--print-life-ranges

-plr

Print register life range information in a trailing column in the produced disassembly.

--print-line-info

-g

Annotate disassembly with source line information obtained from .debug_line section, if present.

--print-line-info-inline

-gi

Annotate disassembly with source line information obtained from .debug_line section along with function inlining info, if present.

--print-line-info-ptx

-gp

Annotate disassembly with source line information obtained from .nv_debug_line_sass section, if present.

--print-raw

-raw

Print the disassembly without any attempt to beautify it.

--separate-functions

-sf

Separate the code corresponding with function symbols by some new lines to let them stand out in the printed disassembly.

--version

-V

Print version information on this tool.

4. Instruction Set Reference

This section contains instruction set reference for NVIDIA NVIDIA® GPU architectures.

4.1. Maxwell and Pascal Instruction Set

The Maxwell (Compute Capability 5.x) and the Pascal (Compute Capability 6.x) architectures have the following instruction set format:

(instruction) (destination) (source1), (source2) ...

Valid destination and source locations include:

  • RX for registers

  • SRX for special system-controlled registers

  • PX for condition registers

  • c[X][Y] for constant memory

Table 4 lists valid instructions for the Maxwell and Pascal GPUs.

Table 4. Maxwell and Pascal Instruction Set

Opcode

Description

Floating Point Instructions

FADD

FP32 Add

FCHK

Single Precision FP Divide Range Check

FCMP

FP32 Compare to Zero and Select Source

FFMA

FP32 Fused Multiply and Add

FMNMX

FP32 Minimum/Maximum

FMUL

FP32 Multiply

FSET

FP32 Compare And Set

FSETP

FP32 Compare And Set Predicate

FSWZADD

FP32 Add used for FSWZ emulation

MUFU

Multi Function Operation

RRO

Range Reduction Operator FP

DADD

FP64 Add

DFMA

FP64 Fused Mutiply Add

DMNMX

FP64 Minimum/Maximum

DMUL

FP64 Multiply

DSET

FP64 Compare And Set

DSETP

FP64 Compare And Set Predicate

HADD2

FP16 Add

HFMA2

FP16 Fused Mutiply Add

HMUL2

FP16 Multiply

HSET2

FP16 Compare And Set

HSETP2

FP16 Compare And Set Predicate

Integer Instructions

BFE

Bit Field Extract

BFI

Bit Field Insert

FLO

Find Leading One

IADD

Integer Addition

IADD3

3-input Integer Addition

ICMP

Integer Compare to Zero and Select Source

IMAD

Integer Multiply And Add

IMADSP

Extracted Integer Multiply And Add.

IMNMX

Integer Minimum/Maximum

IMUL

Integer Multiply

ISCADD

Scaled Integer Addition

ISET

Integer Compare And Set

ISETP

Integer Compare And Set Predicate

LEA

Compute Effective Address

LOP

Logic Operation

LOP3

3-input Logic Operation

POPC

Population count

SHF

Funnel Shift

SHL

Shift Left

SHR

Shift Right

XMAD

Integer Short Multiply Add

Conversion Instructions

F2F

Floating Point To Floating Point Conversion

F2I

Floating Point To Integer Conversion

I2F

Integer To Floating Point Conversion

I2I

Integer To Integer Conversion

Movement Instructions

MOV

Move

PRMT

Permute Register Pair

SEL

Select Source with Predicate

SHFL

Warp Wide Register Shuffle

Predicate/CC Instructions

CSET

Test Condition Code And Set

CSETP

Test Condition Code and Set Predicate

PSET

Combine Predicates and Set

PSETP

Combine Predicates and Set Predicate

P2R

Move Predicate Register To Register

R2P

Move Register To Predicate/CC Register

Texture Instructions

TEX

Texture Fetch

TLD

Texture Load

TLD4

Texture Load 4

TXQ

Texture Query

TEXS

Texture Fetch with scalar/non-vec4 source/destinations

TLD4S

Texture Load 4 with scalar/non-vec4 source/destinations

TLDS

Texture Load with scalar/non-vec4 source/destinations

Compute Load/Store Instructions

LD

Load from generic Memory

LDC

Load Constant

LDG

Load from Global Memory

LDL

Load within Local Memory Window

LDS

Local within Shared Memory Window

ST

Store to generic Memory

STG

Store to global Memory

STL

Store to Local Memory

STS

Store to Shared Memory

ATOM

Atomic Operation on generic Memory

ATOMS

Atomic Operation on Shared Memory

RED

Reduction Operation on generic Memory

CCTL

Cache Control

CCTLL

Cache Control

MEMBAR

Memory Barrier

CCTLT

Texture Cache Control

Surface Memory Instructions

SUATOM

Atomic Op on Surface Memory

SULD

Surface Load

SURED

Reduction Op on Surface Memory

SUST

Surface Store

Control Instructions

BRA

Relative Branch

BRX

Relative Branch Indirect

JMP

Absolute Jump

JMX

Absolute Jump Indirect

SSY

Set Synchronization Point

SYNC

Converge threads after conditional branch

CAL

Relative Call

JCAL

Absolute Call

PRET

Pre-Return From Subroutine

RET

Return From Subroutine

BRK

Break

PBK

Pre-Break

CONT

Continue

PCNT

Pre-continue

EXIT

Exit Program

PEXIT

Pre-Exit

BPT

BreakPoint/Trap

Miscellaneous Instructions

NOP

No Operation

CS2R

Move Special Register to Register

S2R

Move Special Register to Register

B2R

Move Barrier To Register

BAR

Barrier Synchronization

R2B

Move Register to Barrier

VOTE

Vote Across SIMD Thread Group

4.2. Volta Instruction Set

The Volta architecture (Compute Capability 7.x) has the following instruction set format:

(instruction) (destination) (source1), (source2) ...

Valid destination and source locations include:

  • RX for registers

  • SRX for special system-controlled registers

  • PX for predicate registers

  • c[X][Y] for constant memory

Table 5 lists valid instructions for the Volta GPUs.

Table 5. Volta Instruction Set

Opcode

Description

Floating Point Instructions

FADD

FP32 Add

FADD32I

FP32 Add

FCHK

Floating-point Range Check

FFMA32I

FP32 Fused Multiply and Add

FFMA

FP32 Fused Multiply and Add

FMNMX

FP32 Minimum/Maximum

FMUL

FP32 Multiply

FMUL32I

FP32 Multiply

FSEL

Floating Point Select

FSET

FP32 Compare And Set

FSETP

FP32 Compare And Set Predicate

FSWZADD

FP32 Swizzle Add

MUFU

FP32 Multi Function Operation

HADD2

FP16 Add

HADD2_32I

FP16 Add

HFMA2

FP16 Fused Mutiply Add

HFMA2_32I

FP16 Fused Mutiply Add

HMMA

Matrix Multiply and Accumulate

HMUL2

FP16 Multiply

HMUL2_32I

FP16 Multiply

HSET2

FP16 Compare And Set

HSETP2

FP16 Compare And Set Predicate

DADD

FP64 Add

DFMA

FP64 Fused Mutiply Add

DMUL

FP64 Multiply

DSETP

FP64 Compare And Set Predicate

Integer Instructions

BMSK

Bitfield Mask

BREV

Bit Reverse

FLO

Find Leading One

IABS

Integer Absolute Value

IADD

Integer Addition

IADD3

3-input Integer Addition

IADD32I

Integer Addition

IDP

Integer Dot Product and Accumulate

IDP4A

Integer Dot Product and Accumulate

IMAD

Integer Multiply And Add

IMMA

Integer Matrix Multiply and Accumulate

IMNMX

Integer Minimum/Maximum

IMUL

Integer Multiply

IMUL32I

Integer Multiply

ISCADD

Scaled Integer Addition

ISCADD32I

Scaled Integer Addition

ISETP

Integer Compare And Set Predicate

LEA

LOAD Effective Address

LOP

Logic Operation

LOP3

Logic Operation

LOP32I

Logic Operation

POPC

Population count

SHF

Funnel Shift

SHL

Shift Left

SHR

Shift Right

VABSDIFF

Absolute Difference

VABSDIFF4

Absolute Difference

Conversion Instructions

F2F

Floating Point To Floating Point Conversion

F2I

Floating Point To Integer Conversion

I2F

Integer To Floating Point Conversion

I2I

Integer To Integer Conversion

I2IP

Integer To Integer Conversion and Packing

FRND

Round To Integer

Movement Instructions

MOV

Move

MOV32I

Move

PRMT

Permute Register Pair

SEL

Select Source with Predicate

SGXT

Sign Extend

SHFL

Warp Wide Register Shuffle

Predicate Instructions

PLOP3

Predicate Logic Operation

PSETP

Combine Predicates and Set Predicate

P2R

Move Predicate Register To Register

R2P

Move Register To Predicate Register

Load/Store Instructions

LD

Load from generic Memory

LDC

Load Constant

LDG

Load from Global Memory

LDL

Load within Local Memory Window

LDS

Load within Shared Memory Window

ST

Store to Generic Memory

STG

Store to Global Memory

STL

Store to Local Memory

STS

Store to Shared Memory

MATCH

Match Register Values Across Thread Group

QSPC

Query Space

ATOM

Atomic Operation on Generic Memory

ATOMS

Atomic Operation on Shared Memory

ATOMG

Atomic Operation on Global Memory

RED

Reduction Operation on Generic Memory

CCTL

Cache Control

CCTLL

Cache Control

ERRBAR

Error Barrier

MEMBAR

Memory Barrier

CCTLT

Texture Cache Control

Texture Instructions

TEX

Texture Fetch

TLD

Texture Load

TLD4

Texture Load 4

TMML

Texture MipMap Level

TXD

Texture Fetch With Derivatives

TXQ

Texture Query

Surface Instructions

SUATOM

Atomic Op on Surface Memory

SULD

Surface Load

SURED

Reduction Op on Surface Memory

SUST

Surface Store

Control Instructions

BMOV

Move Convergence Barrier State

BPT

BreakPoint/Trap

BRA

Relative Branch

BREAK

Break out of the Specified Convergence Barrier

BRX

Relative Branch Indirect

BSSY

Barrier Set Convergence Synchronization Point

BSYNC

Synchronize Threads on a Convergence Barrier

CALL

Call Function

EXIT

Exit Program

JMP

Absolute Jump

JMX

Absolute Jump Indirect

KILL

Kill Thread

NANOSLEEP

Suspend Execution

RET

Return From Subroutine

RPCMOV

PC Register Move

RTT

Return From Trap

WARPSYNC

Synchronize Threads in Warp

YIELD

Yield Control

Miscellaneous Instructions

B2R

Move Barrier To Register

BAR

Barrier Synchronization

CS2R

Move Special Register to Register

DEPBAR

Dependency Barrier

GETLMEMBASE

Get Local Memory Base Address

LEPC

Load Effective PC

NOP

No Operation

PMTRIG

Performance Monitor Trigger

R2B

Move Register to Barrier

S2R

Move Special Register to Register

SETCTAID

Set CTA ID

SETLMEMBASE

Set Local Memory Base Address

VOTE

Vote Across SIMD Thread Group

4.3. Turing Instruction Set

The Turing architecture (Compute Capability 7.3 and 7.5) have the following instruction set format:

(instruction) (destination) (source1), (source2) ...

Valid destination and source locations include:

  • RX for registers

  • URX for uniform registers

  • SRX for special system-controlled registers

  • PX for predicate registers

  • c[X][Y] for constant memory

Table 6 lists valid instructions for the Turing GPUs.

Table 6. Turing Instruction Set

Opcode

Description

Floating Point Instructions

FADD

FP32 Add

FADD32I

FP32 Add

FCHK

Floating-point Range Check

FFMA32I

FP32 Fused Multiply and Add

FFMA

FP32 Fused Multiply and Add

FMNMX

FP32 Minimum/Maximum

FMUL

FP32 Multiply

FMUL32I

FP32 Multiply

FSEL

Floating Point Select

FSET

FP32 Compare And Set

FSETP

FP32 Compare And Set Predicate

FSWZADD

FP32 Swizzle Add

MUFU

FP32 Multi Function Operation

HADD2

FP16 Add

HADD2_32I

FP16 Add

HFMA2

FP16 Fused Mutiply Add

HFMA2_32I

FP16 Fused Mutiply Add

HMMA

Matrix Multiply and Accumulate

HMUL2

FP16 Multiply

HMUL2_32I

FP16 Multiply

HSET2

FP16 Compare And Set

HSETP2

FP16 Compare And Set Predicate

DADD

FP64 Add

DFMA

FP64 Fused Mutiply Add

DMUL

FP64 Multiply

DSETP

FP64 Compare And Set Predicate

Integer Instructions

BMMA

Bit Matrix Multiply and Accumulate

BMSK

Bitfield Mask

BREV

Bit Reverse

FLO

Find Leading One

IABS

Integer Absolute Value

IADD

Integer Addition

IADD3

3-input Integer Addition

IADD32I

Integer Addition

IDP

Integer Dot Product and Accumulate

IDP4A

Integer Dot Product and Accumulate

IMAD

Integer Multiply And Add

IMMA

Integer Matrix Multiply and Accumulate

IMNMX

Integer Minimum/Maximum

IMUL

Integer Multiply

IMUL32I

Integer Multiply

ISCADD

Scaled Integer Addition

ISCADD32I

Scaled Integer Addition

ISETP

Integer Compare And Set Predicate

LEA

LOAD Effective Address

LOP

Logic Operation

LOP3

Logic Operation

LOP32I

Logic Operation

POPC

Population count

SHF

Funnel Shift

SHL

Shift Left

SHR

Shift Right

VABSDIFF

Absolute Difference

VABSDIFF4

Absolute Difference

Conversion Instructions

F2F

Floating Point To Floating Point Conversion

F2I

Floating Point To Integer Conversion

I2F

Integer To Floating Point Conversion

I2I

Integer To Integer Conversion

I2IP

Integer To Integer Conversion and Packing

FRND

Round To Integer

Movement Instructions

MOV

Move

MOV32I

Move

MOVM

Move Matrix with Transposition or Expansion

PRMT

Permute Register Pair

SEL

Select Source with Predicate

SGXT

Sign Extend

SHFL

Warp Wide Register Shuffle

Predicate Instructions

PLOP3

Predicate Logic Operation

PSETP

Combine Predicates and Set Predicate

P2R

Move Predicate Register To Register

R2P

Move Register To Predicate Register

Load/Store Instructions

LD

Load from generic Memory

LDC

Load Constant

LDG

Load from Global Memory

LDL

Load within Local Memory Window

LDS

Load within Shared Memory Window

LDSM

Load Matrix from Shared Memory with Element Size Expansion

ST

Store to Generic Memory

STG

Store to Global Memory

STL

Store to Local Memory

STS

Store to Shared Memory

MATCH

Match Register Values Across Thread Group

QSPC

Query Space

ATOM

Atomic Operation on Generic Memory

ATOMS

Atomic Operation on Shared Memory

ATOMG

Atomic Operation on Global Memory

RED

Reduction Operation on Generic Memory

CCTL

Cache Control

CCTLL

Cache Control

ERRBAR

Error Barrier

MEMBAR

Memory Barrier

CCTLT

Texture Cache Control

Uniform Datapath Instructions

R2UR

Move from Vector Register to a Uniform Register

S2UR

Move Special Register to Uniform Register

UBMSK

Uniform Bitfield Mask

UBREV

Uniform Bit Reverse

UCLEA

Load Effective Address for a Constant

UFLO

Uniform Find Leading One

UIADD3

Uniform Integer Addition

UIADD3.64

Uniform Integer Addition

UIMAD

Uniform Integer Multiplication

UISETP

Integer Compare and Set Uniform Predicate

ULDC

Load from Constant Memory into a Uniform Register

ULEA

Uniform Load Effective Address

ULOP

Logic Operation

ULOP3

Logic Operation

ULOP32I

Logic Operation

UMOV

Uniform Move

UP2UR

Uniform Predicate to Uniform Register

UPLOP3

Uniform Predicate Logic Operation

UPOPC

Uniform Population Count

UPRMT

Uniform Byte Permute

UPSETP

Uniform Predicate Logic Operation

UR2UP

Uniform Register to Uniform Predicate

USEL

Uniform Select

USGXT

Uniform Sign Extend

USHF

Uniform Funnel Shift

USHL

Uniform Left Shift

USHR

Uniform Right Shift

VOTEU

Voting across SIMD Thread Group with Results in Uniform Destination

Texture Instructions

TEX

Texture Fetch

TLD

Texture Load

TLD4

Texture Load 4

TMML

Texture MipMap Level

TXD

Texture Fetch With Derivatives

TXQ

Texture Query

Surface Instructions

SUATOM

Atomic Op on Surface Memory

SULD

Surface Load

SURED

Reduction Op on Surface Memory

SUST

Surface Store

Control Instructions

BMOV

Move Convergence Barrier State

BPT

BreakPoint/Trap

BRA

Relative Branch

BREAK

Break out of the Specified Convergence Barrier

BRX

Relative Branch Indirect

BRXU

Relative Branch with Uniform Register Based Offset

BSSY

Barrier Set Convergence Synchronization Point

BSYNC

Synchronize Threads on a Convergence Barrier

CALL

Call Function

EXIT

Exit Program

JMP

Absolute Jump

JMX

Absolute Jump Indirect

JMXU

Absolute Jump with Uniform Register Based Offset

KILL

Kill Thread

NANOSLEEP

Suspend Execution

RET

Return From Subroutine

RPCMOV

PC Register Move

RTT

Return From Trap

WARPSYNC

Synchronize Threads in Warp

YIELD

Yield Control

Miscellaneous Instructions

B2R

Move Barrier To Register

BAR

Barrier Synchronization

CS2R

Move Special Register to Register

DEPBAR

Dependency Barrier

GETLMEMBASE

Get Local Memory Base Address

LEPC

Load Effective PC

NOP

No Operation

PMTRIG

Performance Monitor Trigger

R2B

Move Register to Barrier

S2R

Move Special Register to Register

SETCTAID

Set CTA ID

SETLMEMBASE

Set Local Memory Base Address

VOTE

Vote Across SIMD Thread Group

4.4. NVIDIA Ampere GPU and Ada Instruction Set

The NVIDIA Ampere GPU and Ada architectures (Compute Capability 8.0 and 8.6) have the following instruction set format:

(instruction) (destination) (source1), (source2) ...

Valid destination and source locations include:

  • RX for registers

  • URX for uniform registers

  • SRX for special system-controlled registers

  • PX for predicate registers

  • UPX for uniform predicate registers

  • c[X][Y] for constant memory

Table 7 lists valid instructions for the NVIDIA Ampere architecrture and Ada GPUs.

Table 7. NVIDIA Ampere GPU and Ada Instruction Set

Opcode

Description

Floating Point Instructions

FADD

FP32 Add

FADD32I

FP32 Add

FCHK

Floating-point Range Check

FFMA32I

FP32 Fused Multiply and Add

FFMA

FP32 Fused Multiply and Add

FMNMX

FP32 Minimum/Maximum

FMUL

FP32 Multiply

FMUL32I

FP32 Multiply

FSEL

Floating Point Select

FSET

FP32 Compare And Set

FSETP

FP32 Compare And Set Predicate

FSWZADD

FP32 Swizzle Add

MUFU

FP32 Multi Function Operation

HADD2

FP16 Add

HADD2_32I

FP16 Add

HFMA2

FP16 Fused Mutiply Add

HFMA2_32I

FP16 Fused Mutiply Add

HMMA

Matrix Multiply and Accumulate

HMNMX2

FP16 Minimum / Maximum

HMUL2

FP16 Multiply

HMUL2_32I

FP16 Multiply

HSET2

FP16 Compare And Set

HSETP2

FP16 Compare And Set Predicate

DADD

FP64 Add

DFMA

FP64 Fused Mutiply Add

DMMA

Matrix Multiply and Accumulate

DMUL

FP64 Multiply

DSETP

FP64 Compare And Set Predicate

Integer Instructions

BMMA

Bit Matrix Multiply and Accumulate

BMSK

Bitfield Mask

BREV

Bit Reverse

FLO

Find Leading One

IABS

Integer Absolute Value

IADD

Integer Addition

IADD3

3-input Integer Addition

IADD32I

Integer Addition

IDP

Integer Dot Product and Accumulate

IDP4A

Integer Dot Product and Accumulate

IMAD

Integer Multiply And Add

IMMA

Integer Matrix Multiply and Accumulate

IMNMX

Integer Minimum/Maximum

IMUL

Integer Multiply

IMUL32I

Integer Multiply

ISCADD

Scaled Integer Addition

ISCADD32I

Scaled Integer Addition

ISETP

Integer Compare And Set Predicate

LEA

LOAD Effective Address

LOP

Logic Operation

LOP3

Logic Operation

LOP32I

Logic Operation

POPC

Population count

SHF

Funnel Shift

SHL

Shift Left

SHR

Shift Right

VABSDIFF

Absolute Difference

VABSDIFF4

Absolute Difference

Conversion Instructions

F2F

Floating Point To Floating Point Conversion

F2I

Floating Point To Integer Conversion

I2F

Integer To Floating Point Conversion

I2I

Integer To Integer Conversion

I2IP

Integer To Integer Conversion and Packing

I2FP

Integer to FP32 Convert and Pack

F2IP

FP32 Down-Convert to Integer and Pack

FRND

Round To Integer

Movement Instructions

MOV

Move

MOV32I

Move

MOVM

Move Matrix with Transposition or Expansion

PRMT

Permute Register Pair

SEL

Select Source with Predicate

SGXT

Sign Extend

SHFL

Warp Wide Register Shuffle

Predicate Instructions

PLOP3

Predicate Logic Operation

PSETP

Combine Predicates and Set Predicate

P2R

Move Predicate Register To Register

R2P

Move Register To Predicate Register

Load/Store Instructions

LD

Load from generic Memory

LDC

Load Constant

LDG

Load from Global Memory

LDGDEPBAR

Global Load Dependency Barrier

LDGSTS

Asynchronous Global to Shared Memcopy

LDL

Load within Local Memory Window

LDS

Load within Shared Memory Window

LDSM

Load Matrix from Shared Memory with Element Size Expansion

ST

Store to Generic Memory

STG

Store to Global Memory

STL

Store to Local Memory

STS

Store to Shared Memory

MATCH

Match Register Values Across Thread Group

QSPC

Query Space

ATOM

Atomic Operation on Generic Memory

ATOMS

Atomic Operation on Shared Memory

ATOMG

Atomic Operation on Global Memory

RED

Reduction Operation on Generic Memory

CCTL

Cache Control

CCTLL

Cache Control

ERRBAR

Error Barrier

MEMBAR

Memory Barrier

CCTLT

Texture Cache Control

Uniform Datapath Instructions

R2UR

Move from Vector Register to a Uniform Register

REDUX

Reduction of a Vector Register into a Uniform Register

S2UR

Move Special Register to Uniform Register

UBMSK

Uniform Bitfield Mask

UBREV

Uniform Bit Reverse

UCLEA

Load Effective Address for a Constant

UF2FP

Uniform FP32 Down-convert and Pack

UFLO

Uniform Find Leading One

UIADD3

Uniform Integer Addition

UIADD3.64

Uniform Integer Addition

UIMAD

Uniform Integer Multiplication

UISETP

Integer Compare and Set Uniform Predicate

ULDC

Load from Constant Memory into a Uniform Register

ULEA

Uniform Load Effective Address

ULOP

Logic Operation

ULOP3

Logic Operation

ULOP32I

Logic Operation

UMOV

Uniform Move

UP2UR

Uniform Predicate to Uniform Register

UPLOP3

Uniform Predicate Logic Operation

UPOPC

Uniform Population Count

UPRMT

Uniform Byte Permute

UPSETP

Uniform Predicate Logic Operation

UR2UP

Uniform Register to Uniform Predicate

USEL

Uniform Select

USGXT

Uniform Sign Extend

USHF

Uniform Funnel Shift

USHL

Uniform Left Shift

USHR

Uniform Right Shift

VOTEU

Voting across SIMD Thread Group with Results in Uniform Destination

Texture Instructions

TEX

Texture Fetch

TLD

Texture Load

TLD4

Texture Load 4

TMML

Texture MipMap Level

TXD

Texture Fetch With Derivatives

TXQ

Texture Query

Surface Instructions

SUATOM

Atomic Op on Surface Memory

SULD

Surface Load

SURED

Reduction Op on Surface Memory

SUST

Surface Store

Control Instructions

BMOV

Move Convergence Barrier State

BPT

BreakPoint/Trap

BRA

Relative Branch

BREAK

Break out of the Specified Convergence Barrier

BRX

Relative Branch Indirect

BRXU

Relative Branch with Uniform Register Based Offset

BSSY

Barrier Set Convergence Synchronization Point

BSYNC

Synchronize Threads on a Convergence Barrier

CALL

Call Function

EXIT

Exit Program

JMP

Absolute Jump

JMX

Absolute Jump Indirect

JMXU

Absolute Jump with Uniform Register Based Offset

KILL

Kill Thread

NANOSLEEP

Suspend Execution

RET

Return From Subroutine

RPCMOV

PC Register Move

WARPSYNC

Synchronize Threads in Warp

YIELD

Yield Control

Miscellaneous Instructions

B2R

Move Barrier To Register

BAR

Barrier Synchronization

CS2R

Move Special Register to Register

DEPBAR

Dependency Barrier

GETLMEMBASE

Get Local Memory Base Address

LEPC

Load Effective PC

NOP

No Operation

PMTRIG

Performance Monitor Trigger

S2R

Move Special Register to Register

SETCTAID

Set CTA ID

SETLMEMBASE

Set Local Memory Base Address

VOTE

Vote Across SIMD Thread Group

4.5. Hopper Instruction Set

The Hopper architecture (Compute Capability 9.0) has the following instruction set format:

(instruction) (destination) (source1), (source2) ...

Valid destination and source locations include:

  • RX for registers

  • URX for uniform registers

  • SRX for special system-controlled registers

  • PX for predicate registers

  • UPX for uniform predicate registers

  • c[X][Y] for constant memory

  • desc[URX][RY] for memory descriptors

Table 8 lists valid instructions for the Hopper GPUs.

Table 8. Hopper Instruction Set

Opcode

Description

Floating Point Instructions

FADD

FP32 Add

FADD32I

FP32 Add

FCHK

Floating-point Range Check

FFMA32I

FP32 Fused Multiply and Add

FFMA

FP32 Fused Multiply and Add

FMNMX

FP32 Minimum/Maximum

FMUL

FP32 Multiply

FMUL32I

FP32 Multiply

FSEL

Floating Point Select

FSET

FP32 Compare And Set

FSETP

FP32 Compare And Set Predicate

FSWZADD

FP32 Swizzle Add

MUFU

FP32 Multi Function Operation

HADD2

FP16 Add

HADD2_32I

FP16 Add

HFMA2

FP16 Fused Mutiply Add

HFMA2_32I

FP16 Fused Mutiply Add

HMMA

Matrix Multiply and Accumulate

HMNMX2

FP16 Minimum / Maximum

HMUL2

FP16 Multiply

HMUL2_32I

FP16 Multiply

HSET2

FP16 Compare And Set

HSETP2

FP16 Compare And Set Predicate

DADD

FP64 Add

DFMA

FP64 Fused Mutiply Add

DMMA

Matrix Multiply and Accumulate

DMUL

FP64 Multiply

DSETP

FP64 Compare And Set Predicate

Integer Instructions

BMMA

Bit Matrix Multiply and Accumulate

BMSK

Bitfield Mask

BREV

Bit Reverse

FLO

Find Leading One

IABS

Integer Absolute Value

IADD

Integer Addition

IADD3

3-input Integer Addition

IADD32I

Integer Addition

IDP

Integer Dot Product and Accumulate

IDP4A

Integer Dot Product and Accumulate

IMAD

Integer Multiply And Add

IMMA

Integer Matrix Multiply and Accumulate

IMNMX

Integer Minimum/Maximum

IMUL

Integer Multiply

IMUL32I

Integer Multiply

ISCADD

Scaled Integer Addition

ISCADD32I

Scaled Integer Addition

ISETP

Integer Compare And Set Predicate

LEA

LOAD Effective Address

LOP

Logic Operation

LOP3

Logic Operation

LOP32I

Logic Operation

POPC

Population count

SHF

Funnel Shift

SHL

Shift Left

SHR

Shift Right

VABSDIFF

Absolute Difference

VABSDIFF4

Absolute Difference

VHMNMX

SIMD FP16 3-Input Minimum / Maximum

VIADD

SIMD Integer Addition

VIADDMNMX

SIMD Integer Addition and Fused Min/Max Comparison

VIMNMX

SIMD Integer Minimum / Maximum

VIMNMX3

SIMD Integer 3-Input Minimum / Maximum

Conversion Instructions

F2F

Floating Point To Floating Point Conversion

F2I

Floating Point To Integer Conversion

I2F

Integer To Floating Point Conversion

I2I

Integer To Integer Conversion

I2IP

Integer To Integer Conversion and Packing

I2FP

Integer to FP32 Convert and Pack

F2IP

FP32 Down-Convert to Integer and Pack

FRND

Round To Integer

Movement Instructions

MOV

Move

MOV32I

Move

MOVM

Move Matrix with Transposition or Expansion

PRMT

Permute Register Pair

SEL

Select Source with Predicate

SGXT

Sign Extend

SHFL

Warp Wide Register Shuffle

Predicate Instructions

PLOP3

Predicate Logic Operation

PSETP

Combine Predicates and Set Predicate

P2R

Move Predicate Register To Register

R2P

Move Register To Predicate Register

Load/Store Instructions

FENCE

Memory Visibility Guarantee for Shared or Global Memory

LD

Load from generic Memory

LDC

Load Constant

LDG

Load from Global Memory

LDGDEPBAR

Global Load Dependency Barrier

LDGMC

Reducing Load

LDGSTS

Asynchronous Global to Shared Memcopy

LDL

Load within Local Memory Window

LDS

Load within Shared Memory Window

LDSM

Load Matrix from Shared Memory with Element Size Expansion

STSM

Store Matrix to Shared Memory

ST

Store to Generic Memory

STG

Store to Global Memory

STL

Store to Local Memory

STS

Store to Shared Memory

STAS

Asynchronous Store to Distributed Shared Memory With Explicit Synchronization

SYNCS

Sync Unit

MATCH

Match Register Values Across Thread Group

QSPC

Query Space

ATOM

Atomic Operation on Generic Memory

ATOMS

Atomic Operation on Shared Memory

ATOMG

Atomic Operation on Global Memory

REDAS

Asynchronous Reduction on Distributed Shared Memory With Explicit Synchronization

REDG

Reduction Operation on Generic Memory

CCTL

Cache Control

CCTLL

Cache Control

ERRBAR

Error Barrier

MEMBAR

Memory Barrier

CCTLT

Texture Cache Control

Uniform Datapath Instructions

R2UR

Move from Vector Register to a Uniform Register

REDUX

Reduction of a Vector Register into a Uniform Register

S2UR

Move Special Register to Uniform Register

UBMSK

Uniform Bitfield Mask

UBREV

Uniform Bit Reverse

UCGABAR_ARV

CGA Barrier Synchronization

UCGABAR_WAIT

CGA Barrier Synchronization

UCLEA

Load Effective Address for a Constant

UF2FP

Uniform FP32 Down-convert and Pack

UFLO

Uniform Find Leading One

UIADD3

Uniform Integer Addition

UIADD3.64

Uniform Integer Addition

UIMAD

Uniform Integer Multiplication

UISETP

Integer Compare and Set Uniform Predicate

ULDC

Load from Constant Memory into a Uniform Register

ULEA

Uniform Load Effective Address

ULEPC

Uniform Load Effective PC

ULOP

Logic Operation

ULOP3

Logic Operation

ULOP32I

Logic Operation

UMOV

Uniform Move

UP2UR

Uniform Predicate to Uniform Register

UPLOP3

Uniform Predicate Logic Operation

UPOPC

Uniform Population Count

UPRMT

Uniform Byte Permute

UPSETP

Uniform Predicate Logic Operation

UR2UP

Uniform Register to Uniform Predicate

USEL

Uniform Select

USETMAXREG

Release, Deallocate and Allocate Registers

USGXT

Uniform Sign Extend

USHF

Uniform Funnel Shift

USHL

Uniform Left Shift

USHR

Uniform Right Shift

VOTEU

Voting across SIMD Thread Group with Results in Uniform Destination

Warpgroup Instructions

BGMMA

Bit Matrix Multiply and Accumulate Across Warps

HGMMA

Matrix Multiply and Accumulate Across a Warpgroup

IGMMA

Integer Matrix Multiply and Accumulate Across a Warpgroup

QGMMA

FP8 Matrix Multiply and Accumulate Across a Warpgroup

WARPGROUP

Warpgroup Synchronization

WARPGROUPSET

Set Warpgroup Counters

Tensor Memory Access Instructions

UBLKCP

Bulk Data Copy

UBLKPF

Bulk Data Prefetch

UBLKRED

Bulk Data Copy from Shared Memory with Reduction

UTMACCTL

TMA Cache Control

UTMACMDFLUSH

TMA Command Flush

UTMALDG

Tensor Load from Global to Shared Memory

UTMAPF

Tensor Prefetch

UTMAREDG

Tensor Store from Shared to Global Memory with Reduction

UTMASTG

Tensor Store from Shared to Global Memory

Texture Instructions

TEX

Texture Fetch

TLD

Texture Load

TLD4

Texture Load 4

TMML

Texture MipMap Level

TXD

Texture Fetch With Derivatives

TXQ

Texture Query

Surface Instructions

SUATOM

Atomic Op on Surface Memory

SULD

Surface Load

SURED

Reduction Op on Surface Memory

SUST

Surface Store

Control Instructions

ACQBULK

Wait for Bulk Release Status Warp State

BMOV

Move Convergence Barrier State

BPT

BreakPoint/Trap

BRA

Relative Branch

BREAK

Break out of the Specified Convergence Barrier

BRX

Relative Branch Indirect

BRXU

Relative Branch with Uniform Register Based Offset

BSSY

Barrier Set Convergence Synchronization Point

BSYNC

Synchronize Threads on a Convergence Barrier

CALL

Call Function

CGAERRBAR

CGA Error Barrier

ELECT

Elect a Leader Thread

ENDCOLLECTIVE

Reset the MCOLLECTIVE mask

EXIT

Exit Program

JMP

Absolute Jump

JMX

Absolute Jump Indirect

JMXU

Absolute Jump with Uniform Register Based Offset

KILL

Kill Thread

NANOSLEEP

Suspend Execution

PREEXIT

Dependent Task Launch Hint

RET

Return From Subroutine

RPCMOV

PC Register Move

WARPSYNC

Synchronize Threads in Warp

YIELD

Yield Control

Miscellaneous Instructions

B2R

Move Barrier To Register

BAR

Barrier Synchronization

CS2R

Move Special Register to Register

DEPBAR

Dependency Barrier

GETLMEMBASE

Get Local Memory Base Address

LEPC

Load Effective PC

NOP

No Operation

PMTRIG

Performance Monitor Trigger

S2R

Move Special Register to Register

SETCTAID

Set CTA ID

SETLMEMBASE

Set Local Memory Base Address

VOTE

Vote Across SIMT Thread Group

5. cu++filt

cu++filt decodes (demangles) low-level identifiers that have been mangled by CUDA C++ into user readable names. For every input alphanumeric word, the output of cu++filt is either the demangled name if the name decodes to a CUDA C++ name, or the original name itself.

5.1. Usage

cu++filt accepts one or more alphanumeric words (consisting of letters, digits, underscores, dollars, or periods) and attepts to decipher them. The basic usage is as following:

cu++filt [options] <symbol(s)>

To demangle an entire file, like a binary, pipe the contents of the file to cu++filt, such as in the following command:

nm <input file> | cu++filt

To demangle function names without printing their parameter types, use the following command :

cu++filt -p <symbol(s)>

To skip a leading underscore from mangled symbols, use the following command:

cu++filt -_ <symbol(s)>

Here’s a sample output of cu++filt:

$ cu++filt _Z1fIiEbl
bool f<int>(long)

As shown in the output, the symbol _Z1fIiEbl was successfully demangled.

To strip all types in the function signature and parameters, use the -p option:

$ cu++filt -p _Z1fIiEbl
f<int>

To skip a leading underscore from a mangled symbol, use the -_ option:

$ cu++filt -_ __Z1fIiEbl
bool f<int>(long)

To demangle an entire file, pipe the contents of the file to cu++filt:

$ nm test.sm_70.cubin | cu++filt
0000000000000000 t hello(char *)
0000000000000070 t hello(char *)::display()
0000000000000000 T hello(int *)

Symbols that cannot be demangled are printed back to stdout as is:

$ cu++filt _ZD2
_ZD2

Multiple symbols can be demangled from the command line:

$ cu++filt _ZN6Scope15Func1Enez _Z3fooIiPFYneEiEvv _ZD2
Scope1::Func1(__int128, long double, ...)
void foo<int, __int128 (*)(long double), int>()
_ZD2

5.2. Command-line Options

Table 9 contains supported command-line options of cu++filt, along with a description of what each option does.

Table 9. cu++filt Command-line Options

Option

Description

-_

Strip underscore. On some systems, the CUDA compiler puts an underscore in front of every name. This option removes the initial underscore. Whether cu++filt removes the underscore by default is target dependent.

-p

When demangling the name of a function, do not display the types of the function’s parameters.

-h

Print a summary of the options to cu++filt and exit.

-v

Print the version information of this tool.

5.3. Library Availability

cu++filt is also available as a static library (libcufilt) that can be linked against an existing project. The following interface describes it’s usage:

char* __cu_demangle(const char *id, char *output_buffer, size_t *length, int *status)

This interface can be found in the file “nv_decode.h” located in the SDK.

Input Parameters

id Input mangled string.

output_buffer Pointer to where the demangled buffer will be stored. This memory must be allocated with malloc. If output-buffer is NULL, memory will be malloc’d to store the demangled name and returned through the function return value. If the output-buffer is too small, it is expanded using realloc.

length It is necessary to provide the size of the output buffer if the user is providing pre-allocated memory. This is needed by the demangler in case the size needs to be reallocated. If the length is non-null, the length of the demangled buffer is placed in length.

status *status is set to one of the following values:

  • 0 - The demangling operation succeeded

  • -1 - A memory allocation failure occurred

  • -2 - Not a valid mangled id

  • -3 - An input validation failure has occurred (one or more arguments are invalid)

Return Value

A pointer to the start of the NUL-terminated demangled name, or NULL if the demangling fails. The caller is responsible for deallocating this memory using free.

Note: This function is thread-safe.

Example Usage

#include <stdio.h>
#include <stdlib.h>
#include "nv_decode.h"

int main()
{
  int     status;
  const char *real_mangled_name="_ZN8clstmp01I5cls01E13clstmp01_mf01Ev";
  const char *fake_mangled_name="B@d_iDentiFier";

  char* realname = __cu_demangle(fake_mangled_name, 0, 0, &status);
  printf("fake_mangled_name:\t result => %s\t status => %d\n", realname, status);
  free(realname);

  size_t size = sizeof(char)*1000;
  realname = (char*)malloc(size);
  __cu_demangle(real_mangled_name, realname, &size, &status);
  printf("real_mangled_name:\t result => %s\t status => %d\n", realname, status);
  free(realname);

  return 0;
}

This prints:

fake_mangled_name:   result => (null)     status => -2
real_mangled_name:   result => clstmp01<cls01>::clstmp01_mf01()   status => 0

6. nvprune

nvprune prunes host object files and libraries to only contain device code for the specified targets.

6.1. Usage

nvprune accepts a single input file each time it’s run, emitting a new output file. The basic usage is as following:

nvprune [options] -o <outfile> <infile>

The input file must be either a relocatable host object or static library (not a host executable), and the output file will be the same format.

Either the –arch or –generate-code option must be used to specify the target(s) to keep. All other device code is discarded from the file. The targets can be either a sm_NN arch (cubin) or compute_NN arch (ptx).

For example, the following will prune libcublas_static.a to only contain sm_70 cubin rather than all the targets which normally exist:

nvprune -arch sm_70 libcublas_static.a -o libcublas_static70.a

Note that this means that libcublas_static70.a will not run on any other architecture, so should only be used when you are building for a single architecture.

6.2. Command-line Options

Table 10 contains supported command-line options of nvprune, along with a description of what each option does. Each option has a long name and a short name, which can be used interchangeably.

Table 10. nvprune Command-line Options

Option (long)

Option (short)

Description

--arch <gpu architecture name>,...

-arch

Specify the name of the NVIDIA GPU architecture which will remain in the object or library.

--generate-code

-gencode

This option is same format as nvcc –generate-code option, and provides a way to specify multiple architectures which should remain in the object or library. Only the ‘code’ values are used as targets to match. Allowed keywords for this option: ‘arch’,’code’.

--no-relocatable-elf

-no-relocatable-elf

Don’t keep any relocatable ELF.

--output-file

-o

Specify name and location of the output file.

--help

-h

Print this help information on this tool.

--options-file <file>,...

-optf

Include command line options from specified file.

--version

-V

Print version information on this tool.

7. Notices

7.1. Notice

This document is provided for information purposes only and shall not be regarded as a warranty of a certain functionality, condition, or quality of a product. NVIDIA Corporation (“NVIDIA”) makes no representations or warranties, expressed or implied, as to the accuracy or completeness of the information contained in this document and assumes no responsibility for any errors contained herein. NVIDIA shall have no liability for the consequences or use of such information or for any infringement of patents or other rights of third parties that may result from its use. This document is not a commitment to develop, release, or deliver any Material (defined below), code, or functionality.

NVIDIA reserves the right to make corrections, modifications, enhancements, improvements, and any other changes to this document, at any time without notice.

Customer should obtain the latest relevant information before placing orders and should verify that such information is current and complete.

NVIDIA products are sold subject to the NVIDIA standard terms and conditions of sale supplied at the time of order acknowledgement, unless otherwise agreed in an individual sales agreement signed by authorized representatives of NVIDIA and customer (“Terms of Sale”). NVIDIA hereby expressly objects to applying any customer general terms and conditions with regards to the purchase of the NVIDIA product referenced in this document. No contractual obligations are formed either directly or indirectly by this document.

NVIDIA products are not designed, authorized, or warranted to be suitable for use in medical, military, aircraft, space, or life support equipment, nor in applications where failure or malfunction of the NVIDIA product can reasonably be expected to result in personal injury, death, or property or environmental damage. NVIDIA accepts no liability for inclusion and/or use of NVIDIA products in such equipment or applications and therefore such inclusion and/or use is at customer’s own risk.

NVIDIA makes no representation or warranty that products based on this document will be suitable for any specified use. Testing of all parameters of each product is not necessarily performed by NVIDIA. It is customer’s sole responsibility to evaluate and determine the applicability of any information contained in this document, ensure the product is suitable and fit for the application planned by customer, and perform the necessary testing for the application in order to avoid a default of the application or the product. Weaknesses in customer’s product designs may affect the quality and reliability of the NVIDIA product and may result in additional or different conditions and/or requirements beyond those contained in this document. NVIDIA accepts no liability related to any default, damage, costs, or problem which may be based on or attributable to: (i) the use of the NVIDIA product in any manner that is contrary to this document or (ii) customer product designs.

No license, either expressed or implied, is granted under any NVIDIA patent right, copyright, or other NVIDIA intellectual property right under this document. Information published by NVIDIA regarding third-party products or services does not constitute a license from NVIDIA to use such products or services or a warranty or endorsement thereof. Use of such information may require a license from a third party under the patents or other intellectual property rights of the third party, or a license from NVIDIA under the patents or other intellectual property rights of NVIDIA.

Reproduction of information in this document is permissible only if approved in advance by NVIDIA in writing, reproduced without alteration and in full compliance with all applicable export laws and regulations, and accompanied by all associated conditions, limitations, and notices.

THIS DOCUMENT AND ALL NVIDIA DESIGN SPECIFICATIONS, REFERENCE BOARDS, FILES, DRAWINGS, DIAGNOSTICS, LISTS, AND OTHER DOCUMENTS (TOGETHER AND SEPARATELY, “MATERIALS”) ARE BEING PROVIDED “AS IS.” NVIDIA MAKES NO WARRANTIES, EXPRESSED, IMPLIED, STATUTORY, OR OTHERWISE WITH RESPECT TO THE MATERIALS, AND EXPRESSLY DISCLAIMS ALL IMPLIED WARRANTIES OF NONINFRINGEMENT, MERCHANTABILITY, AND FITNESS FOR A PARTICULAR PURPOSE. TO THE EXTENT NOT PROHIBITED BY LAW, IN NO EVENT WILL NVIDIA BE LIABLE FOR ANY DAMAGES, INCLUDING WITHOUT LIMITATION ANY DIRECT, INDIRECT, SPECIAL, INCIDENTAL, PUNITIVE, OR CONSEQUENTIAL DAMAGES, HOWEVER CAUSED AND REGARDLESS OF THE THEORY OF LIABILITY, ARISING OUT OF ANY USE OF THIS DOCUMENT, EVEN IF NVIDIA HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. Notwithstanding any damages that customer might incur for any reason whatsoever, NVIDIA’s aggregate and cumulative liability towards customer for the products described herein shall be limited in accordance with the Terms of Sale for the product.

7.2. OpenCL

OpenCL is a trademark of Apple Inc. used under license to the Khronos Group Inc.

7.3. Trademarks

NVIDIA and the NVIDIA logo are trademarks or registered trademarks of NVIDIA Corporation in the U.S. and other countries. Other company and product names may be trademarks of the respective companies with which they are associated.