chore: go mod tidy / vendor / make deps

vendor/github.com/pjbgf/sha1cd/sha1cdblock_amd64.go

@@ -4,39 +4,47 @@
 package sha1cd
 
 import (
-	"math"
-	"unsafe"
+	"runtime"
 
+	"github.com/klauspost/cpuid/v2"
 	shared "github.com/pjbgf/sha1cd/internal"
 )
 
-// blockAMD64 hashes the message p into the current state in dig.
+var hasSHANI = (runtime.GOARCH == "amd64" &&
+	cpuid.CPU.Supports(cpuid.AVX) &&
+	cpuid.CPU.Supports(cpuid.SHA) &&
+	cpuid.CPU.Supports(cpuid.SSE3) &&
+	cpuid.CPU.Supports(cpuid.SSE4))
+
+// blockAMD64 hashes the message p into the current state in h.
 // Both m1 and cs are used to store intermediate results which are used by the collision detection logic.
 //
 //go:noescape
-func blockAMD64(dig *digest, p sliceHeader, m1 []uint32, cs [][5]uint32)
+func blockAMD64(h []uint32, p []byte, m1 []uint32, cs [][5]uint32)
 
 func block(dig *digest, p []byte) {
+	if forceGeneric || !hasSHANI {
+		blockGeneric(dig, p)
+		return
+	}
+
 	m1 := [shared.Rounds]uint32{}
 	cs := [shared.PreStepState][shared.WordBuffers]uint32{}
 
 	for len(p) >= shared.Chunk {
-		// Only send a block to be processed, as the collission detection
-		// works on a block by block basis.
-		ips := sliceHeader{
-			base: uintptr(unsafe.Pointer(&p[0])),
-			len:  int(math.Min(float64(len(p)), float64(shared.Chunk))),
-			cap:  shared.Chunk,
-		}
+		// The assembly code only supports processing a block at a time,
+		// so adjust the chunk accordingly.
+		chunk := p[:shared.Chunk]
 
-		blockAMD64(dig, ips, m1[:], cs[:])
+		blockAMD64(dig.h[:], chunk, m1[:], cs[:])
+		rectifyCompressionState(m1, &cs)
 
 		col := checkCollision(m1, cs, dig.h)
 		if col {
 			dig.col = true
 
-			blockAMD64(dig, ips, m1[:], cs[:])
-			blockAMD64(dig, ips, m1[:], cs[:])
+			blockAMD64(dig.h[:], chunk, m1[:], cs[:])
+			blockAMD64(dig.h[:], chunk, m1[:], cs[:])
 		}
 
 		p = p[shared.Chunk:]