From 8e3b7f1caaf8305f428ab03d5f637ef8e613a594 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Petr=20Balv=C3=ADn?= <opensource@petrbalvin.org>
Date: Wed, 7 Oct 2026 01:24:31 +0200
Subject: [PATCH] test(verify): prove the amd64 sse families through the
 link-parity kernel

Assisted-by: GLM 5.3 Flash
---
 verify/goobj_link_fold_test.go   | 63 ++++++++++++++++++++++++++++++++
 verify/testdata/linkfold_amd64.s | 58 +++++++++++++++++++++++++++++
 2 files changed, 121 insertions(+)
 create mode 100644 verify/goobj_link_fold_test.go
 create mode 100644 verify/testdata/linkfold_amd64.s

diff --git a/verify/goobj_link_fold_test.go b/verify/goobj_link_fold_test.go
new file mode 100644
index 0000000..91c2c86
--- /dev/null
+++ b/verify/goobj_link_fold_test.go
@@ -0,0 +1,63 @@
+// Copyright (c) 2026 Petr Balvín <opensource@petrbalvin.org> (https://petrbalvin.org)
+// SPDX-License-Identifier: BSD-3-Clause
+
+package verify
+
+import (
+	"os"
+	"os/exec"
+	"testing"
+)
+
+// goobjFoldMainSrc is the Go side of the SSE regression.  The kernel
+// XOR-folds 16-byte blocks and sums the tail bytes through the amd64 SSE
+// families; main derives the same value in Go and panics on any divergence
+// before printing the deterministic line the baseline and the gasm-linked
+// binaries must agree on.
+const goobjFoldMainSrc = `package main
+
+import "encoding/binary"
+
+func fold(buf []byte, seed uint64) uint64
+
+func main() {
+	buf := []byte("gasm link parity vector kernel! 0123456789abcdef tail")
+	low, high := uint64(0), uint64(0)
+	for i := 0; i+16 <= len(buf); i += 16 {
+		low ^= binary.LittleEndian.Uint64(buf[i:])
+		high ^= binary.LittleEndian.Uint64(buf[i+8:])
+	}
+	tail := uint64(0)
+	for _, b := range buf[len(buf)-len(buf)%16:] {
+		tail += uint64(b)
+	}
+	const k = 0x9e3779b97f4a7c15
+	want := (low + high + tail) * k ^ 0xabcdef
+	if got := fold(buf, 0xabcdef); got != want {
+		panic("fold")
+	}
+	println("ok", fold(buf, 0xabcdef))
+}
+`
+
+// TestGOOBJLinkFoldKernel proves the amd64 SSE families are proven end to
+// end: the kernel assembles by gasm into a GOOBJ object, substitutes
+// byte-wise into a real go build archive, relinks with cmd/link and the
+// linked binary computes the same fold as the toolchain-built one, run
+// natively on amd64 hosts.
+func TestGOOBJLinkFoldKernel(t *testing.T) {
+	if testing.Short() {
+		t.Skip("builds the gasm binary and links Go programs")
+	}
+	if os.Getenv("GASM_LINK_PARITY") == "" {
+		t.Skip("deliberate verification: set GASM_LINK_PARITY=1 (just link-parity)")
+	}
+	goBin, err := exec.LookPath("go")
+	if err != nil {
+		t.Skip("no Go toolchain available")
+	}
+	gasmBin := buildLinkParityGasm(t, goBin)
+	t.Run("amd64", func(t *testing.T) {
+		goobjLinkKernel(t, goBin, gasmBin, "amd64", "linkfold_amd64.s", goobjFoldMainSrc)
+	})
+}
diff --git a/verify/testdata/linkfold_amd64.s b/verify/testdata/linkfold_amd64.s
new file mode 100644
index 0000000..498fad0
--- /dev/null
+++ b/verify/testdata/linkfold_amd64.s
@@ -0,0 +1,58 @@
+// Copyright (c) 2026 Petr Balvín <opensource@petrbalvin.org> (https://petrbalvin.org)
+// SPDX-License-Identifier: BSD-3-Clause
+
+#include "textflag.h"
+
+// The link-regression kernel for the amd64 SSE families: fold XOR-folds
+// every 16-byte block of a buffer into one octa with MOVOU and PXOR,
+// folds any remaining bytes with a scalar loop, combines the two octa
+// lanes with PEXTRQ, passes the total to mixq through the ABI0 stack slot
+// and XORs the seed into the returned mix.  mixq multiplies by the odd
+// golden-ratio constant, the same arithmetic the Go side of the regression
+// mirrors bit for bit.  The run claim is native: the host is amd64.
+
+// func mixq(x uint64) uint64
+TEXT ·mixq(SB), NOSPLIT, $0-16
+	MOVQ  x+0(FP), AX
+	MOVQ  $0x9e3779b97f4a7c15, CX
+	IMULQ CX, AX
+	MOVQ  AX, ret+8(FP)
+	RET
+
+// func fold(buf []byte, seed uint64) uint64
+TEXT ·fold(SB), NOSPLIT, $16-40
+	MOVQ    buf+0(FP), SI
+	MOVQ    buf_len+8(FP), CX
+	MOVQ    seed+24(FP), BX
+	PXOR    X1, X1
+	XORQ    AX, AX
+
+blocks:
+	CMPQ    CX, $16
+	JB      tail
+	MOVOU   (SI), X2
+	PXOR    X2, X1
+	ADDQ    $16, SI
+	SUBQ    $16, CX
+	JMP     blocks
+
+tail:
+	TESTQ   CX, CX
+	JE      gathered
+	MOVBQZX (SI), DX
+	ADDQ    DX, AX
+	ADDQ    $1, SI
+	SUBQ    $1, CX
+	JMP     tail
+
+gathered:
+	MOVQ    X1, DX
+	PEXTRQ  $1, X1, CX
+	ADDQ    CX, DX
+	ADDQ    AX, DX
+	MOVQ    DX, 0(SP)
+	CALL    ·mixq(SB)
+	MOVQ    8(SP), DX
+	XORQ    BX, DX
+	MOVQ    DX, ret+32(FP)
+	RET