all: AVX-512 (#217)

Extends avo to support most AVX-512 instruction sets.

The instruction type is extended to support suffixes. The K family of opmask
registers is added to the register package, and the operand package is updated
to support the new operand types. Move instruction deduction in `Load` and
`Store` is extended to support KMOV* and VMOV* forms.

Internal code generation packages were overhauled. Instruction database loading
required various messy changes to account for the additional complexities of the
AVX-512 instruction sets. The internal/api package was added to introduce a
separation between instruction forms in the database, and the functions avo
provides to create them. This was required since with instruction suffixes there
is no longer a one-to-one mapping between instruction constructors and opcodes.

AVX-512 bloated generated source code size substantially, initially increasing
compilation and CI test times to an unacceptable level. Two changes were made to
address this:

1.  Instruction constructors in the `x86` package moved to an optab-based
    approach. This compiles substantially faster than the verbose code
    generation we had before.

2.  The most verbose code-generated tests are moved under build tags and
    limited to a stress test mode. Stress test builds are run on
    schedule but not in regular CI.

An example of AVX-512 accelerated 16-lane MD5 is provided to demonstrate and
test the new functionality.

Updates #20 #163 #229

Co-authored-by: Vaughn Iverson <vsivsi@yahoo.com>
This commit is contained in:
Michael McLoughlin
2021-11-12 18:35:36 -08:00
parent 2867bd7e01
commit b76e849b5c
71 changed files with 257395 additions and 61474 deletions

View File

@@ -3,6 +3,7 @@ package gen
import (
"fmt"
"github.com/mmcloughlin/avo/internal/api"
"github.com/mmcloughlin/avo/internal/inst"
"github.com/mmcloughlin/avo/internal/prnt"
"github.com/mmcloughlin/avo/printer"
@@ -26,34 +27,41 @@ func (b *build) Generate(is []inst.Instruction) ([]byte, error) {
b.Printf("package build\n\n")
b.Printf("import (\n")
b.Printf("\t\"%s/operand\"\n", pkg)
b.Printf("\t\"%s/x86\"\n", pkg)
b.Printf("\t%q\n", api.ImportPath(api.IRPackage))
b.Printf("\t%q\n", api.ImportPath(api.OperandPackage))
b.Printf("\t%q\n", api.ImportPath("x86"))
b.Printf(")\n\n")
for _, i := range is {
b.instruction(i)
// Helper to reduce source code size a little.
b.Printf("func (c *Context) addinstruction(i *ir.Instruction, err error) {\n")
b.Printf("if err == nil { c.Instruction(i) }")
b.Printf(" else { c.adderror(err) }\n")
b.Printf("}\n\n")
// Generate build functions.
fns := api.InstructionsFunctions(is)
for _, fn := range fns {
b.function(fn)
}
return b.Result()
}
func (b *build) instruction(i inst.Instruction) {
s := params(i)
d := doc(i)
func (b *build) function(fn *api.Function) {
s := fn.Signature()
d := fn.Doc()
// Context method.
methoddoc := append([]string{}, d...)
methoddoc = append(methoddoc, fmt.Sprintf("Construct and append a %s instruction to the active function.", i.Opcode))
methoddoc = append(methoddoc, fmt.Sprintf("Construct and append a %s instruction to the active function.", fn.Opcode()))
b.Comment(methoddoc...)
b.Printf("func (c *Context) %s(%s) {\n", i.Opcode, s.ParameterList())
b.Printf("if inst, err := x86.%s(%s); err == nil", i.Opcode, s.Arguments())
b.Printf(" { c.Instruction(inst) }")
b.Printf(" else { c.adderror(err) }\n")
b.Printf("func (c *Context) %s(%s) {\n", fn.Name(), s.ParameterList())
b.Printf("c.addinstruction(x86.%s(%s))", fn.Name(), s.Arguments())
b.Printf("}\n\n")
// Global version.
globaldoc := append([]string{}, methoddoc...)
globaldoc = append(globaldoc, "Operates on the global context.")
b.Comment(globaldoc...)
b.Printf("func %s(%s) { ctx.%s(%s) }\n\n", i.Opcode, s.ParameterList(), i.Opcode, s.Arguments())
b.Printf("func %s(%s) { ctx.%s(%s) }\n\n", fn.Name(), s.ParameterList(), fn.Name(), s.Arguments())
}