gosl
gosl
implements Go as a shader language for GPU compute shaders (using WebGPU), enabling standard Go code to run on the GPU.
The relevant subsets of Go code are specifically marked using //gosl:
comment directives, and this code must only use basic expressions and concrete types that will compile correctly in a shader (see Restrictions below). Method functions and pass-by-reference pointer arguments to struct
types are supported and incur no additional compute cost due to inlining (see notes below for more detail).
A large and complex biologically-based neural network simulation framework called axon has been implemented using gosl
, allowing 1000's of lines of equations and data structures to run through standard Go on the CPU, and accelerated significantly on the GPU. This allows efficient debugging and unit testing of the code in Go, whereas debugging on the GPU is notoriously difficult.
gosl
converts Go code to WGSL which can then be loaded directly into a WebGPU compute shader.
See examples/basic and rand for examples, using the gpu GPU compute shader system. It is also possible in principle to use gosl to generate shader files for any other WebGPU application, but this has not been tested.
You must also install goimports
which is used on the extracted subset of Go code, to get the imports right:
$ go install golang.org/x/tools/cmd/goimports@latest
To install the gosl
command, do:
$ go install cogentcore.org/core/vgpu/gosl@latest
In your Go code, use these comment directives:
//gosl:start <filename>
< Go code to be translated >
//gosl:end <filename>
to bracket code to be processed. The resulting converted code is copied into a shaders
subdirectory created under the current directory where the gosl
command is run, using the filenames specified in the comment directives. Each such filename should correspond to a complete shader program (i.e., a "kernel"), or a file that can be included into other shader programs. Code is appended to the target file names in the order of the source .go files on the command line, so multiple .go files can be combined into one resulting WGSL file.
WGSL specific code, e.g., for the main
compute function or to specify #include
files, can be included either by specifying files with a .wgsl
extension as arguments to the gosl
command, or by using a //gosl:wgsl
comment directive as follows:
//gosl:wgsl <filename>
// <WGSL shader code to be copied>
//gosl:end <filename>
where the WGSL shader code is commented out in the .go file -- it will be copied into the target filename and uncommented. The WGSL code can be surrounded by /*
*/
comment blocks (each on a separate line) for multi-line code (though using a separate .wgsl
file is generally preferable in this case).
For .wgsl
files, their filename is used to determine the shaders
destination file name, and they are automatically appended to the end of the corresponding .wgsl
file generated from the Go
files -- this is where the main
function and associated global variables should be specified.
IMPORTANT: all .go
and .wgsl
files are removed from the shaders
directory prior to processing to ensure everything there is current -- always specify a different source location for any custom .wgsl
files that are included.
Usage
gosl [flags] [path ...]
The flags are:
-debug
enable debugging messages while running
-exclude string
comma-separated list of names of functions to exclude from exporting to HLSL (default "Update,Defaults")
-keep
keep temporary converted versions of the source files, for debugging
-out string
output directory for shader code, relative to where gosl is invoked -- must not be an empty string (default "shaders")
gosl
path args can include filenames, directory names, or Go package paths (e.g., cogentcore.org/core/math32/fastexp.go
loads just that file from the given package) -- files without any //gosl:
comment directives will be skipped up front before any expensive processing, so it is not a problem to specify entire directories where only some files are relevant. Also, you can specify a particular file from a directory, then the entire directory, to ensure that a particular file from that directory appears first -- otherwise alphabetical order is used. gosl
ensures that only one copy of each file is included.
Any struct
types encountered will be checked for 16-byte alignment of sub-types and overall sizes as an even multiple of 16 bytes (4 float32
or int32
values), which is the alignment used in WGSL and glsl shader languages, and the underlying GPU hardware presumably. Look for error messages on the output from the gosl run. This ensures that direct byte-wise copies of data between CPU and GPU will be successful. The fact that gosl
operates directly on the original CPU-side Go code uniquely enables it to perform these alignment checks, which are otherwise a major source of difficult-to-diagnose bugs.
Restrictions
In general shader code should be simple mathematical expressions and data types, with minimal control logic via if
, for
statements, and only using the subset of Go that is consistent with C. Here are specific restrictions:
-
Can only use float32
, [u]int32
for basic types (int
is converted to int32
automatically), and struct
types composed of these same types -- no other Go types (i.e., map
, slices, string
, etc) are compatible. There are strict alignment restrictions on 16 byte (e.g., 4 float32
's) intervals that are enforced via the alignsl
sub-package.
-
WGSL does not support 64 bit float or int.
-
Use slbool.Bool
instead of bool
-- it defines a Go-friendly interface based on a int32
basic type.
-
Alignment and padding of struct
fields is key -- this is automatically checked by gosl
.
-
WGSL does not support enum types, but standard go const
declarations will be converted. Use an int32
or uint32
data type. It will automatically deal with the simple incrementing iota
values, but not more complex cases. Also, for bitflags, define explicitly, not using bitflags
package, and use 0x01
, 0x02
, 0x04
etc instead of 1<<2
-- in theory the latter should be ok but in practice it complains.
-
Cannot use multiple return values, or multiple assignment of variables in a single =
expression.
-
Can use multiple variable names with the same type (e.g., min, max float32
) -- this will be properly converted to the more redundant C form with the type repeated.
-
switch
case
statements are purely self-contained -- no fallthrough
allowed! does support multiple items per case
however.
-
TODO: WGSL does not do multi-pass compiling, so all dependent types must be specified before being used in other ones, and this also precludes referencing the current type within itself. todo: can you just use a forward declaration?
-
WGSL does specify that new variables are initialized to 0, like Go, but also somehow discourages that use-case. It is safer to initialize directly:
val := float32(0) // guaranteed 0 value
var val float32 // ok but generally avoid
Other language features
-
tour-of-wgsl is a good reference to explain things more directly than the spec.
-
ptr<function,MyStruct>
provides a pointer arg
-
private
scope = within the shader code "module", i.e., one thread.
-
function
= within the function, not outside it.
-
workgroup
= shared across workgroup -- coudl be powerful (but slow!) -- need to learn more.
Random numbers: slrand
See slrand for a shader-optimized random number generation package, which is supported by gosl
-- it will convert slrand
calls into appropriate WGSL named function calls. gosl
will also copy the slrand.wgsl
file, which contains the full source code for the RNG, into the destination shaders
directory, so it can be included with a simple local path:
//gosl:wgsl mycode
// #include "slrand.wgsl"
//gosl:end mycode
With sufficiently large N, and ignoring the data copying setup time, around ~80x speedup is typical on a Macbook Pro with M1 processor. The rand
example produces a 175x speedup!
Implementation / Design Notes
Links
Key docs for WGSL as compute shaders: