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>

.github/workflows/packages.yml

@@ -561,6 +561,7 @@ jobs:
         run: go test ./...
   segmentio-asm:
     runs-on: ubuntu-latest
+    if: false # skip: https://github.com/mmcloughlin/avo/issues/229
     steps:
       - name: Install Go
         uses: actions/setup-go@37335c7bb261b353407cff977110895fa0b4f7d8 # v2.1.3

.github/workflows/stress.yml

@@ -0,0 +1,50 @@
+name: stress
+
+permissions:
+  contents: read
+
+on:
+  schedule:
+    - cron: '33 3 * * 6'
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    env:
+        GOFLAGS: -tags=stress
+    steps:
+    - name: Install Go
+      uses: actions/setup-go@37335c7bb261b353407cff977110895fa0b4f7d8 # v2.1.3
+      with:
+        go-version: 1.17.x
+    - name: Configure Go Environment
+      run: |
+        echo GOPATH=${{ runner.workspace }} >> $GITHUB_ENV
+        echo ${{ runner.workspace }}/bin >> $GITHUB_PATH
+    - name: Go Environment
+      run: go env
+    - name: Checkout code
+      uses: actions/checkout@5a4ac9002d0be2fb38bd78e4b4dbde5606d7042f # v2.3.4
+      with:
+        persist-credentials: false
+    - name: Bootstrap
+      run: ./script/bootstrap
+
+    - name: Generate
+      run: ./script/generate
+    - name: Status
+      run: git status
+
+    - name: Build
+      run: go build ./...
+    - name: Test
+      run: go test -bench . ./...
+    - name: Coverage
+      run: ./script/coverage
+
+    - name: Upload Stress Unit Test Coverage
+      uses: codecov/codecov-action@51d810878be5422784e86451c0e7c14e5860ec47 # v2.0.2
+      with:
+        token: ${{ secrets.CODECOV_TOKEN }}
+        files: unittests.coverprofile
+        flags: stress

.golangci.yml

@@ -1,3 +1,6 @@
+run:
+  timeout: 5m
+
 linters:
   enable-all: true
   disable:
@@ -27,3 +30,5 @@ issues:
     - Error return value of .((os\.)?std(out|err)\..*|.*Close|.*Flush|os\.Remove(All)?|.*printf?|os\.(Un)?Setenv). is not checked
     # gocritic: https://github.com/go-critic/go-critic/issues/762
     - ' with `(len|cap|real|imag)`'
+    # We want to allow all caps in certain cases.
+    - "ALL_CAPS in Go names"

README.md

@@ -189,6 +189,7 @@ Implementations of full algorithms:
 * **[sha1](examples/sha1):** [SHA-1](https://en.wikipedia.org/wiki/SHA-1) cryptographic hash.
 * **[fnv1a](examples/fnv1a):** [FNV-1a](https://en.wikipedia.org/wiki/Fowler%E2%80%93Noll%E2%80%93Vo_hash_function#FNV-1a_hash) hash function.
 * **[dot](examples/dot):** Vector dot product.
+* **[md5x16](examples/md5x16):** AVX-512 accelerated [MD5](https://en.wikipedia.org/wiki/MD5).
 * **[geohash](examples/geohash):** Integer [geohash](https://en.wikipedia.org/wiki/Geohash) encoding.
 * **[stadtx](examples/stadtx):** [`StadtX` hash](https://github.com/demerphq/BeagleHash) port from [dgryski/go-stadtx](https://github.com/dgryski/go-stadtx).
 

build/context.go

@@ -15,6 +15,9 @@ import (
 	"github.com/mmcloughlin/avo/reg"
 )
 
+//go:generate avogen -output zinstructions.go build
+//go:generate avogen -output zinstructions_test.go buildtest
+
 // Context maintains state for incrementally building an avo File.
 type Context struct {
 	pkg      *packages.Package
@@ -175,8 +178,6 @@ func (c *Context) activefunc() *ir.Function {
 	return c.function
 }
 
-//go:generate avogen -output zinstructions.go build
-
 // StaticGlobal adds a new static data section to the file and returns a pointer to it.
 func (c *Context) StaticGlobal(name string) operand.Mem {
 	c.global = ir.NewStaticGlobal(name)

build/global.go

@@ -98,6 +98,9 @@ func YMM() reg.VecVirtual { return ctx.YMM() }
 // ZMM allocates and returns a 512-bit vector register.
 func ZMM() reg.VecVirtual { return ctx.ZMM() }
 
+// K allocates and returns an opmask register.
+func K() reg.OpmaskVirtual { return ctx.K() }
+
 // Param returns a the named argument of the active function.
 func Param(name string) gotypes.Component { return ctx.Param(name) }