diff --git a/tools/test/xregs_via_sig/c2x2c_amd64.S b/tools/test/xregs_via_sig/c2x2c_amd64.S
--- a/tools/test/xregs_via_sig/c2x2c_amd64.S
+++ b/tools/test/xregs_via_sig/c2x2c_amd64.S
@@ -5,9 +5,37 @@
  * $FreeBSD$
  */
 
+#include <machine/specialreg.h>
+
+
+	.global xregs_bank
+	.type xregs_bank, @function
+xregs_bank:
+	movl	$1, %eax
+	cpuid
+	andl	$(CPUID2_AVX|CPUID2_OSXSAVE), %ecx
+	cmpl	$(CPUID2_AVX|CPUID2_OSXSAVE), %ecx
+	jne	sse
+	xorl	%ecx, %ecx
+	xgetbv
+	andl	$XFEATURE_AVX, %eax
+	cmpl	$XFEATURE_AVX, %eax
+	jne	sse
+	movq	$1, %rax
+	retq
+sse:
+	xor	%rax, %rax
+	retq
+
+	.size xregs_bank, . - xregs_bank
+
+
 	.global cpu_to_xregs
 	.type cpu_to_xregs, @function
 cpu_to_xregs:
+	call	xregs_bank
+	cmpl	$1, %eax
+	je	cpu_to_avx
 	movdqu	%xmm0, (%rdi)
 	movdqu	%xmm1, 1 * 16(%rdi)
 	movdqu	%xmm2, 2 * 16(%rdi)
@@ -26,12 +54,34 @@
 	movdqu	%xmm15, 15 * 16(%rdi)
 	retq
 
+cpu_to_avx:
+	vmovdqu	%ymm0, (%rdi)
+	vmovdqu	%ymm1, 1 * 32(%rdi)
+	vmovdqu	%ymm2, 2 * 32(%rdi)
+	vmovdqu	%ymm3, 3 * 32(%rdi)
+	vmovdqu	%ymm4, 4 * 32(%rdi)
+	vmovdqu	%ymm5, 5 * 32(%rdi)
+	vmovdqu	%ymm6, 6 * 32(%rdi)
+	vmovdqu	%ymm7, 7 * 32(%rdi)
+	vmovdqu	%ymm8, 8 * 32(%rdi)
+	vmovdqu	%ymm9, 9 * 32(%rdi)
+	vmovdqu	%ymm10, 10 * 32(%rdi)
+	vmovdqu	%ymm11, 11 * 32(%rdi)
+	vmovdqu	%ymm12, 12 * 32(%rdi)
+	vmovdqu	%ymm13, 13 * 32(%rdi)
+	vmovdqu	%ymm14, 14 * 32(%rdi)
+	vmovdqu	%ymm15, 15 * 32(%rdi)
+	retq
+
 	.size cpu_to_xregs, . - cpu_to_xregs
 
 
 	.global xregs_to_cpu
 	.type xregs_to_cpu, @function
 xregs_to_cpu:
+	call	xregs_bank
+	cmpl	$1, %eax
+	je	avx_to_cpu
 	movdqu	(%rdi), %xmm0
 	movdqu	1 * 16(%rdi), %xmm1
 	movdqu	2 * 16(%rdi), %xmm2
@@ -50,6 +100,25 @@
 	movdqu	15 * 16(%rdi), %xmm15
 	retq
 
+avx_to_cpu:
+	vmovdqu	(%rdi), %ymm0
+	vmovdqu	1 * 32(%rdi), %ymm1
+	vmovdqu	2 * 32(%rdi), %ymm2
+	vmovdqu	3 * 32(%rdi), %ymm3
+	vmovdqu	4 * 32(%rdi), %ymm4
+	vmovdqu	5 * 32(%rdi), %ymm5
+	vmovdqu	6 * 32(%rdi), %ymm6
+	vmovdqu	7 * 32(%rdi), %ymm7
+	vmovdqu	8 * 32(%rdi), %ymm8
+	vmovdqu	9 * 32(%rdi), %ymm9
+	vmovdqu	10 * 32(%rdi), %ymm10
+	vmovdqu	11 * 32(%rdi), %ymm11
+	vmovdqu	12 * 32(%rdi), %ymm12
+	vmovdqu	13 * 32(%rdi), %ymm13
+	vmovdqu	14 * 32(%rdi), %ymm14
+	vmovdqu	15 * 32(%rdi), %ymm15
+	retq
+
 	.size xregs_to_cpu, . - xregs_to_cpu
 
 	.section        .note.GNU-stack,"",@progbits
diff --git a/tools/test/xregs_via_sig/xregs_via_sig.c b/tools/test/xregs_via_sig/xregs_via_sig.c
--- a/tools/test/xregs_via_sig/xregs_via_sig.c
+++ b/tools/test/xregs_via_sig/xregs_via_sig.c
@@ -49,15 +49,25 @@
 #endif
 
 #if defined(__amd64__)
-#define	XREGSRNAM	"xmm"
 #define	NREGS		16
+#define	NBYTES		32
+
+static const char *xregs_names[] = {
+	"xmm",
+	"ymm",
+};
+
 #elif defined(__aarch64__)
 #define	XREGSRNAM	"q"
 #define	NREGS		32
+#define	NBYTES		16
+#endif
+#if !defined(XREGSRNAM)
+int xregs_bank(void);
 #endif
 
 struct xregsreg {
-	uint8_t xregs_bytes[16];
+	uint8_t xregs_bytes[NBYTES];
 };
 
 struct xregs {
@@ -69,6 +79,17 @@
 
 static atomic_uint sigs;
 
+static const char *
+xregs_name(void)
+{
+#if !defined(XREGSRNAM)
+	return (xregs_names[xregs_bank()]);
+#else
+	return (XREGSRNAM);
+#endif
+}
+
+
 static void
 sigusr1_handler(int sig __unused, siginfo_t *si __unused, void *m __unused)
 {
@@ -126,7 +147,7 @@
 		r2 = &xregs2->xregsreg[i];
 		for (j = 0; j < nitems(r1->xregs_bytes); j++) {
 			if (r1->xregs_bytes[j] != r2->xregs_bytes[j]) {
-				printf("%%%s%u\n", XREGSRNAM, i);
+				printf("%%%s%u\n", xregs_name(), i);
 				dump_xregs(r1);
 				dump_xregs(r2);
 				break;