working through a bunch of avx512 stuff, regspec constructors are const

1 files changed, 151 insertions, 0 deletions

diff --git a/test/protected_mode/evex_generated.rs b/test/protected_mode/evex_generated.rs
index 9c3a06e..bd553d5 100644
--- a/test/protected_mode/evex_generated.rs
+++ b/test/protected_mode/evex_generated.rs
@@ -133,7 +133,9 @@ fn tests_None_0f() {
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x0d, 0x11, 0x0a], "vmovups xmmword [edx]{k5}, xmm1");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x08, 0x11, 0x4a, 0x01], "vmovups xmmword [edx + 0x10], xmm1");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x0d, 0x11, 0x4a, 0x01], "vmovups xmmword [edx + 0x10]{k5}, xmm1");
+    test_invalid(&[0x62, 0xf1, 0x7c, 0x8d, 0x11, 0x0a]);
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x08, 0x12, 0xca], "vmovhlps xmm1, xmm0, xmm2");
+    test_invalid(&[0x62, 0xf1, 0x7c, 0x18, 0x12, 0xca]); // no broadcast
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x08, 0x12, 0x0a], "vmovlps xmm1, xmm0, qword [edx]");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x08, 0x12, 0x4a, 0x01], "vmovlps xmm1, xmm0, qword [edx + 0x8]");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x08, 0x13, 0x0a], "vmovlps qword [edx], xmm1");
@@ -229,12 +231,14 @@ fn tests_None_0f() {
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x08, 0x15, 0x4a, 0x01], "vunpckhps xmm1, xmm0, xmmword [edx + 0x10]");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x0d, 0x15, 0x4a, 0x01], "vunpckhps xmm1{k5}, xmm0, xmmword [edx + 0x10]");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x08, 0x16, 0xca], "vmovlhps xmm1, xmm0, xmm2");
+    test_invalid(&[0x62, 0xf1, 0x7c, 0x18, 0x16, 0xca]); //
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x08, 0x16, 0x0a], "vmovhps xmm1, xmm0, qword [edx]");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x08, 0x16, 0x4a, 0x01], "vmovhps xmm1, xmm0, qword [edx + 0x8]");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x08, 0x17, 0x0a], "vmovhps qword [edx], xmm1");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x08, 0x17, 0x4a, 0x01], "vmovhps qword [edx + 0x8], xmm1");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0xad, 0x28, 0xca], "vmovaps ymm1{k5}{z}, ymm2");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0xad, 0x28, 0x0a], "vmovaps ymm1{k5}{z}, ymmword [edx]");
+    test_invalid(&[0x62, 0xf1, 0x7c, 0xbd, 0x28, 0x0a]); // no broadcast
     test_avx_full(&[0x62, 0xf1, 0x7c, 0xad, 0x28, 0x4a, 0x01], "vmovaps ymm1{k5}{z}, ymmword [edx + 0x20]");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x28, 0x28, 0xca], "vmovaps ymm1, ymm2");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x2d, 0x28, 0xca], "vmovaps ymm1{k5}, ymm2");
@@ -253,6 +257,7 @@ fn tests_None_0f() {
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x4d, 0x28, 0x4a, 0x01], "vmovaps zmm1{k5}, zmmword [edx + 0x40]");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x8d, 0x28, 0xca], "vmovaps xmm1{k5}{z}, xmm2");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x8d, 0x28, 0x0a], "vmovaps xmm1{k5}{z}, xmmword [edx]");
+    test_invalid(&[0x62, 0xf1, 0x7c, 0x9d, 0x29, 0xca]); // no sae/er support on movaps
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x8d, 0x28, 0x4a, 0x01], "vmovaps xmm1{k5}{z}, xmmword [edx + 0x10]");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x08, 0x28, 0xca], "vmovaps xmm1, xmm2");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x0d, 0x28, 0xca], "vmovaps xmm1{k5}, xmm2");
@@ -287,15 +292,22 @@ fn tests_None_0f() {
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x48, 0x2b, 0x4a, 0x01], "vmovntps zmmword [edx + 0x40], zmm1");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x08, 0x2b, 0x0a], "vmovntps xmmword [edx], xmm1");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x08, 0x2b, 0x4a, 0x01], "vmovntps xmmword [edx + 0x10], xmm1");
+    test_invalid(&[0x62, 0xf1, 0x7c, 0x38, 0x2b, 0x0a]); // no zero mask-merge
+    test_invalid(&[0x62, 0xf1, 0x7c, 0xa8, 0x2b, 0x0a]); // no broadcast
+    test_invalid(&[0x62, 0xf1, 0xfc, 0x28, 0x2b, 0x0a]); // no W=1
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x78, 0x2e, 0xca], "vucomiss xmm1{sae}, xmm2");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x28, 0x2e, 0xca], "vucomiss xmm1, xmm2");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x28, 0x2e, 0x0a], "vucomiss xmm1, dword [edx]");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x28, 0x2e, 0x4a, 0x01], "vucomiss xmm1, dword [edx + 0x4]");
+    test_invalid(&[0x62, 0xf1, 0x7c, 0x18, 0x2e, 0x0a]); // no broadcast from memory
+    test_invalid(&[0x62, 0xf1, 0x7d, 0x68, 0x2e, 0x0a]); // no L'L=11
+    test_invalid(&[0x62, 0xf1, 0x7d, 0x88, 0x2e, 0x0a]); // no zero mask-merge
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x78, 0x2f, 0xca], "vcomiss xmm1{sae}, xmm2");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x28, 0x2f, 0xca], "vcomiss xmm1, xmm2");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x28, 0x2f, 0x0a], "vcomiss xmm1, dword [edx]");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x28, 0x2f, 0x4a, 0x01], "vcomiss xmm1, dword [edx + 0x4]");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0xfd, 0x51, 0xca], "vsqrtps zmm1{k5}{z}{rz-sae}, zmm2");
+    test_invalid(&[0x62, 0xf1, 0xfc, 0xfd, 0x51, 0xca]); // requires W=0
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x78, 0x51, 0xca], "vsqrtps zmm1{rz-sae}, zmm2");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x7d, 0x51, 0xca], "vsqrtps zmm1{k5}{rz-sae}, zmm2");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0xbd, 0x51, 0xca], "vsqrtps zmm1{k5}{z}{rd-sae}, zmm2");
@@ -648,6 +660,7 @@ fn tests_None_0f() {
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x0d, 0x59, 0x4a, 0x01], "vmulps xmm1{k5}, xmm0, xmmword [edx + 0x10]");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0xfd, 0x5a, 0xca], "vcvtps2pd zmm1{k5}{z}{sae}, ymm2");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x78, 0x5a, 0xca], "vcvtps2pd zmm1{sae}, ymm2");
+    test_invalid(&[0x62, 0xf1, 0xfc, 0x78, 0x5a, 0xca]); // W=0
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x7d, 0x5a, 0xca], "vcvtps2pd zmm1{k5}{sae}, ymm2");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0xbd, 0x5a, 0x0a], "vcvtps2pd ymm1{k5}{z}, dword [edx]{1to4}");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0xbd, 0x5a, 0x4a, 0x01], "vcvtps2pd ymm1{k5}{z}, dword [edx + 0x4]{1to4}");
@@ -1231,6 +1244,7 @@ fn tests_None_0f() {
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x78, 0xc2, 0xca, 0xcc], "vcmpps k1{sae}, zmm0, zmm2, 0xcc");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x7d, 0xc2, 0xca, 0xcc], "vcmpps k1{k5}{sae}, zmm0, zmm2, 0xcc");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x38, 0xc2, 0x0a, 0xcc], "vcmpps k1, ymm0, dword [edx]{1to8}, 0xcc");
+    test_invalid(&[0x62, 0xf1, 0x7c, 0xb8, 0xc2, 0x0a, 0xcc]); // no zero mask-merge
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x3d, 0xc2, 0x0a, 0xcc], "vcmpps k1{k5}, ymm0, dword [edx]{1to8}, 0xcc");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x38, 0xc2, 0x4a, 0x01, 0xcc], "vcmpps k1, ymm0, dword [edx + 0x4]{1to8}, 0xcc");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x3d, 0xc2, 0x4a, 0x01, 0xcc], "vcmpps k1{k5}, ymm0, dword [edx + 0x4]{1to8}, 0xcc");
@@ -1262,6 +1276,7 @@ fn tests_None_0f() {
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x0d, 0xc2, 0x4a, 0x01, 0xcc], "vcmpps k1{k5}, xmm0, xmmword [edx + 0x10], 0xcc");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0xbd, 0xc6, 0x0a, 0xcc], "vshufps ymm1{k5}{z}, ymm0, dword [edx]{1to8}, 0xcc");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0xbd, 0xc6, 0x4a, 0x01, 0xcc], "vshufps ymm1{k5}{z}, ymm0, dword [edx + 0x4]{1to8}, 0xcc");
+    test_invalid(&[0x62, 0xf1, 0x7c, 0x3d, 0xc6, 0xca, 0xcc]); // no broadcast from register source
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x38, 0xc6, 0x0a, 0xcc], "vshufps ymm1, ymm0, dword [edx]{1to8}, 0xcc");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x3d, 0xc6, 0x0a, 0xcc], "vshufps ymm1{k5}, ymm0, dword [edx]{1to8}, 0xcc");
     test_avx_full(&[0x62, 0xf1, 0x7c, 0x38, 0xc6, 0x4a, 0x01, 0xcc], "vshufps ymm1, ymm0, dword [edx + 0x4]{1to8}, 0xcc");
@@ -16383,4 +16398,140 @@ fn tests_f3_0f38() {
     test_avx_full(&[0x62, 0xf2, 0x7f, 0x2d, 0xab, 0x0a], "v4fnmaddss xmm1{k5}, xmm0, xmmword [edx]");
     test_avx_full(&[0x62, 0xf2, 0x7f, 0x28, 0xab, 0x4a, 0x01], "v4fnmaddss xmm1, xmm0, xmmword [edx + 0x10]");
     test_avx_full(&[0x62, 0xf2, 0x7f, 0x2d, 0xab, 0x4a, 0x01], "v4fnmaddss xmm1{k5}, xmm0, xmmword [edx + 0x10]");
+    test_invalid(&[0x62, 0xf1, 0x7c, 0x8d, 0x11, 0x0a]);
+    test_invalid(&[0x62, 0xf1, 0x7c, 0x18, 0x12, 0xca]); // no broadcast
+    test_invalid(&[0x62, 0xf1, 0x7c, 0x18, 0x16, 0xca]); //
+    test_invalid(&[0x62, 0xf1, 0x7c, 0xbd, 0x28, 0x0a]); // no broadcast
+    test_invalid(&[0x62, 0xf1, 0x7c, 0x9d, 0x29, 0xca]); // no sae/er support on movaps
+    test_invalid(&[0x62, 0xf1, 0x7c, 0x38, 0x2b, 0x0a]); // no zero mask-merge
+    test_invalid(&[0x62, 0xf1, 0x7c, 0xa8, 0x2b, 0x0a]); // no broadcast
+    test_invalid(&[0x62, 0xf1, 0xfc, 0x28, 0x2b, 0x0a]); // no W=1
+    test_invalid(&[0x62, 0xf1, 0x7c, 0x18, 0x2e, 0x0a]); // no broadcast from memory
+    test_invalid(&[0x62, 0xf1, 0x7d, 0x68, 0x2e, 0x0a]); // no L'L=11
+    test_invalid(&[0x62, 0xf1, 0x7d, 0x88, 0x2e, 0x0a]); // no zero mask-merge
+    test_invalid(&[0x62, 0xf1, 0xfc, 0xfd, 0x51, 0xca]); // requires W=0
+    test_invalid(&[0x62, 0xf1, 0xfc, 0x78, 0x5a, 0xca]); // W=0
+    test_invalid(&[0x62, 0xf1, 0x7c, 0xb8, 0xc2, 0x0a, 0xcc]); // no zero mask-merge
+    test_invalid(&[0x62, 0xf1, 0x7c, 0x3d, 0xc6, 0xca, 0xcc]); // no broadcast from register source
+    test_invalid(&[0x62, 0xf1, 0xfd, 0xdd, 0x10, 0x0a]);
+    test_invalid(&[0x62, 0xf1, 0xfd, 0xbd, 0x14, 0xca]); // no broadcast in reg-reg
+    test_invalid(&[0x62, 0xf1, 0xfd, 0xbd, 0x15, 0xca]); // no broadcast in reg-reg
+    test_invalid(&[0x62, 0xf1, 0xfd, 0x38, 0x28, 0x0a]); // no broadcast
+    test_invalid(&[0x62, 0xf1, 0xfd, 0x18, 0x2b, 0x0a]); // no broadcast
+    test_invalid(&[0x62, 0xf1, 0xfd, 0x88, 0x2b, 0x0a]); // no zero mask-merge
+    test_invalid(&[0x62, 0xf1, 0x7d, 0x08, 0x2b, 0x0a]); // no W=-
+    test_invalid(&[0x62, 0xf1, 0xfd, 0x79, 0x2e, 0xca]); // mask reg must be 000
+    test_invalid(&[0x62, 0xf1, 0xfd, 0x18, 0x2e, 0x0a]); // no broadcast from memory
+    test_invalid(&[0x62, 0xf1, 0xfd, 0x68, 0x2e, 0x0a]); // no L'L=11
+    test_invalid(&[0x62, 0xf1, 0xfd, 0x88, 0x2e, 0x0a]); // no zero mask-merge
+    test_invalid(&[0x62, 0xf1, 0xfd, 0x28, 0x5b, 0xca]); // no W=1
+    test_invalid(&[0x62, 0xf1, 0x7d, 0x3d, 0x66, 0xca]); // no broadcast on reg operand (no sae)
+    test_invalid(&[0x62, 0xf1, 0xfd, 0x88, 0x6e, 0xca]); //no zero mask-merge
+    test_invalid(&[0x62, 0xf1, 0x7d, 0x88, 0x6e, 0xca]); // no zero mask-merge
+    test_invalid(&[0x62, 0xf1, 0x7d, 0x88, 0x7e, 0xca]); // no zero mask-merge
+    test_invalid(&[0x62, 0xf1, 0xfd, 0xbd, 0xc2, 0x0a, 0xcc]); // no zero mask-merge
+    test_invalid(&[0x62, 0xf1, 0x7d, 0x78, 0xe6, 0xca]); // requires W=1
+    test_invalid(&[0x62, 0xf1, 0x7d, 0x28, 0xe7, 0xca]); // no reg-reg encoding
+    test_invalid(&[0x62, 0xf1, 0xfd, 0x38, 0xf6, 0x0a]); // no broadcast
+    test_invalid(&[0x62, 0xf1, 0xfd, 0xa8, 0xf6, 0x0a]); // no zero mask-merge
+    test_invalid(&[0x62, 0xf1, 0x7e, 0x6d, 0x10, 0x0a]);
+    test_invalid(&[0x62, 0xf1, 0x7e, 0x6f, 0x10, 0x0a]);
+    test_invalid(&[0x62, 0xf1, 0x7e, 0x38, 0x10, 0x0a]);
+    test_invalid(&[0x62, 0xf1, 0x7e, 0x3d, 0x11, 0xca]);
+    test_invalid(&[0x62, 0xf1, 0x7e, 0xad, 0x11, 0x0a]);
+    test_invalid(&[0x62, 0xf1, 0xfe, 0x29, 0x2a, 0xca]); // mask reg must be 000
+    test_invalid(&[0x62, 0xf1, 0xfe, 0x38, 0x2a, 0x0a]); // no broadcast
+    test_invalid(&[0x62, 0xf1, 0xfe, 0x68, 0x2a, 0x0a]); // no L'L=11
+    test_invalid(&[0x62, 0xf1, 0xfe, 0x29, 0x2c, 0xca]); // mask register must be 000
+    test_invalid(&[0x62, 0xf1, 0xfe, 0x38, 0x2c, 0x0a]); // no broadcast
+    test_invalid(&[0x62, 0xf1, 0x7e, 0x38, 0x2c, 0x0a]); // no broadcast, regardless of W
+    test_invalid(&[0x62, 0xf1, 0xfe, 0x68, 0x2c, 0x0a]); // no L'L=11
+    test_invalid(&[0x62, 0xf1, 0xfe, 0x29, 0x2d, 0xca]); // mask register must be 000
+    test_invalid(&[0x62, 0xf1, 0xfe, 0x38, 0x2d, 0x0a]); // no broadcast
+    test_invalid(&[0x62, 0xf1, 0xfe, 0x68, 0x2d, 0x0a]); // no L'L=11
+    test_invalid(&[0x62, 0xf1, 0x7e, 0x78, 0x5a, 0x0a]); // no broadcast with memory source
+    test_invalid(&[0x62, 0xf1, 0xfe, 0x68, 0x7b, 0x0a]); // no L'L=11
+    test_invalid(&[0x62, 0xf1, 0xfe, 0x88, 0x7e, 0xca]); // no zero mask-merge
+    test_invalid(&[0x62, 0xf1, 0x7e, 0x6d, 0xc2, 0xca, 0xcc]); // do not allow L'L=11
+    test_invalid(&[0x62, 0xf1, 0xff, 0x6d, 0x10, 0x0a]);
+    test_invalid(&[0x62, 0xf1, 0xff, 0x6f, 0x10, 0x0a]);
+    test_invalid(&[0x62, 0xf1, 0xff, 0x3d, 0x11, 0xca]);
+    test_invalid(&[0x62, 0xf1, 0xff, 0xad, 0x11, 0x0a]);
+    test_invalid(&[0x62, 0xf1, 0xff, 0xbd, 0x12, 0x0a]);
+    test_invalid(&[0x62, 0xf1, 0xff, 0x38, 0x51, 0x0a]); // no broadcast
+    test_invalid(&[0x62, 0xf1, 0xff, 0x68, 0x51, 0x0a]); // no L'L=11
+    test_invalid(&[0x62, 0xf1, 0xff, 0xbd, 0x5a, 0x0a]); // no L'L=11 unless for sae
+    test_invalid(&[0x62, 0xf1, 0xff, 0x6f, 0x5a, 0x0a]); // no L'L=11 unless for sae
+    test_invalid(&[0x62, 0xf1, 0x7f, 0x78, 0x5f, 0xca]); // requires W=1
+    test_invalid(&[0x62, 0xf1, 0xff, 0x68, 0x7b, 0x0a]); // no L'L=11
+    test_invalid(&[0x62, 0xf1, 0xff, 0x6d, 0xc2, 0x0a, 0xcc]); // no L'L=11
+    test_invalid(&[0x62, 0xf2, 0xfd, 0x78, 0x13, 0x0a]); // W=0
+    test_invalid(&[0x62, 0xf2, 0x7d, 0xbd, 0x14, 0xca]); // no broadcast
+    test_invalid(&[0x62, 0xf2, 0xfd, 0xad, 0x25, 0xca]); // W must be 1
+    test_invalid(&[0x62, 0xf2, 0xfd, 0xcd, 0x25, 0xca]);
+    test_invalid(&[0x62, 0xf2, 0xfd, 0xad, 0x26, 0x0a]); // no zero-merge
+    test_invalid(&[0x62, 0xf2, 0x7d, 0xad, 0x26, 0x0a]); // no zero-merge
+    test_invalid(&[0x62, 0xf2, 0x7d, 0xdd, 0x27, 0x0a]); // no zero-merge
+    test_invalid(&[0x62, 0xf2, 0x7d, 0x3d, 0x27, 0xca]); // no invalid broadcast mode
+    test_invalid(&[0x62, 0xf2, 0xfd, 0xdd, 0x27, 0x0a]); // no zero-merge
+    test_invalid(&[0x62, 0xf2, 0xfd, 0xbd, 0x28, 0xca]); // no broadcast on register source
+    test_invalid(&[0x62, 0xf2, 0xfd, 0xad, 0x29, 0xca]); // no zero-merge
+    test_invalid(&[0x62, 0xf2, 0xfd, 0x3d, 0x29, 0xca]); // no zero-merge
+    test_invalid(&[0x62, 0xf2, 0x7d, 0x28, 0x2a, 0xca]); // no register source
+    test_invalid(&[0x62, 0xf2, 0x7d, 0xa8, 0x2a, 0x0a]); // no broadcast
+    test_invalid(&[0x62, 0xf2, 0x7d, 0xfd, 0x2d, 0x0a]); // sae is indicated by evex.b, with memory source evex.b implies broadcast as well. vscalefss does not broadcast, so reject.
+    test_invalid(&[0x62, 0xf2, 0x7d, 0xbd, 0x43, 0x0a]); // no broadcast
+    test_invalid(&[0x62, 0xf2, 0x7d, 0x3d, 0x4d, 0xca]); // no sae
+    test_invalid(&[0x62, 0xf2, 0x7d, 0x6d, 0x4d, 0xca]); // no L'L=11
+    test_invalid(&[0x62, 0xf2, 0x7d, 0x3d, 0x4e, 0xca]); // no sae
+    test_invalid(&[0x62, 0xf2, 0x7d, 0x68, 0x4f, 0xca]); // no L'L=11
+    test_invalid(&[0x62, 0xf2, 0x7d, 0xad, 0x63, 0x0a]); // no zero-merge on memory operands
+    test_invalid(&[0x62, 0xf2, 0x7d, 0x38, 0x78, 0xca]); // deny evex.b
+    test_invalid(&[0x62, 0xf2, 0x7d, 0x38, 0x7a, 0xca]); // still no evex.b
+    test_invalid(&[0x62, 0xf2, 0x7d, 0x38, 0x7c, 0xca]); // no broadcast here either
+    test_avx_full(&[0x62, 0xf2, 0x7d, 0x68, 0x99, 0xca], "vfmadd132ss xmm1, xmm0, xmm2"); // no L'L==0 when not sae
+    test_invalid(&[0x62, 0xf2, 0xfd, 0x38, 0xdc, 0x0a]); // no broadcast
+    test_invalid(&[0x62, 0xf2, 0xfd, 0xa8, 0xdc, 0x0a]); // no zero mask-merge
+    test_invalid(&[0x62, 0xf2, 0x7e, 0xad, 0x10, 0x0a]); // cannot set evex.z on stores.
+    test_invalid(&[0x62, 0xf2, 0x7e, 0xcd, 0x12, 0x0a]);
+    test_invalid(&[0x62, 0xf2, 0x7e, 0x88, 0x28, 0xca]); //
+    test_invalid(&[0x62, 0xf2, 0x7e, 0xa8, 0x39, 0xca]); // no zero mask-merge
+    test_invalid(&[0x62, 0xf2, 0x7e, 0x88, 0x3a, 0xca]); // no zero "mask merge", no masking at all
+    test_invalid(&[0x62, 0xf2, 0x7e, 0xbd, 0x72, 0xca]); // no register-register broadcast
+    test_avx_full(&[0x62, 0xf2, 0x7f, 0x08, 0x68, 0xca], "vp2intersectd k1, xmm0, xmm2"); // VP2INTERSECTD_MASKmskw_XMMu32_XMMu32_AVX512, extension: AVX512EVEX
+    test_invalid(&[0x62, 0xf2, 0x7f, 0x09, 0x68, 0xca]); // requires mask reg to be 000
+    test_invalid(&[0x62, 0xf2, 0xff, 0x09, 0x68, 0xca]); // requires mask reg to be 000
+    test_invalid(&[0x62, 0xf3, 0xfd, 0xbd, 0x00, 0xca, 0xcc]); // no broadcast on reg-reg ops
+    test_invalid(&[0x62, 0xf3, 0x7d, 0xbd, 0x04, 0xca, 0xcc]); // no broadcast on reg sources
+    test_invalid(&[0x62, 0xf3, 0x7d, 0x5d, 0x0a, 0x0a, 0xcc]); // no broadcast with memory source
+    test_invalid(&[0x62, 0xf3, 0x7d, 0x6d, 0x0a, 0x0a, 0xcc]); // no broadcast with memory source
+    test_invalid(&[0x62, 0xf3, 0xfd, 0xa8, 0x0b, 0xca, 0xcc]); // no zero-merge without mask reg
+    test_invalid(&[0x62, 0xf3, 0xfd, 0x38, 0x0b, 0x0a, 0xcc]); // no broadcast on memory source
+    test_invalid(&[0x62, 0xf3, 0xfd, 0x68, 0x0b, 0x0a, 0xcc]); // L'L==11 requires sae
+    test_invalid(&[0x62, 0xf3, 0xfd, 0x3d, 0x0f, 0xca, 0xcc]); // no broadcast
+    test_invalid(&[0x62, 0xf3, 0xfd, 0x38, 0x0f, 0x0a, 0xcc]); // still no broadcast
+    test_invalid(&[0x62, 0xf3, 0xfd, 0x88, 0x14, 0xca, 0xcc]); // no zero mask-merge, no masking!
+    test_invalid(&[0x62, 0xf3, 0xfd, 0x18, 0x14, 0xca, 0xcc]); // no broadcast
+    test_invalid(&[0x62, 0xf3, 0xfd, 0x18, 0x15, 0xca, 0xcc]); // no broadcast
+    test_invalid(&[0x62, 0xf3, 0xfd, 0x88, 0x15, 0xca, 0xcc]); // no zero mask-merge
+    test_invalid(&[0x62, 0xf3, 0x7d, 0x18, 0x16, 0xca, 0xcc]); // no broadcast
+    test_invalid(&[0x62, 0xf3, 0x7d, 0x88, 0x16, 0xca, 0xcc]); // no zero mask-merge
+    test_invalid(&[0x62, 0xf3, 0xfd, 0x18, 0x17, 0xca, 0xcc]); // no broadcast
+    test_invalid(&[0x62, 0xf3, 0xfd, 0x88, 0x17, 0xca, 0xcc]); // no zero mask-merge
+    test_invalid(&[0x62, 0xf3, 0x7d, 0x3d, 0x19, 0x0a, 0xcc]); // no zero-merge with memmory dest
+    test_invalid(&[0x62, 0xf3, 0x7d, 0xcd, 0x1b, 0x0a, 0xcc]); // no zero-merge into memory
+    test_invalid(&[0x62, 0xf3, 0x7d, 0x3d, 0x1d, 0x0a, 0xcc]); // no zero-merge into memory
+    test_invalid(&[0x62, 0xf3, 0x7d, 0x6d, 0x1d, 0x0a, 0xcc]); // no L'L==11 for non-sae
+    test_invalid(&[0x62, 0xf3, 0xfd, 0x18, 0x20, 0xca, 0xcc]); //
+    test_invalid(&[0x62, 0xf3, 0xfd, 0x88, 0x20, 0xca, 0xcc]); //
+    test_invalid(&[0x62, 0xf3, 0x7d, 0x18, 0x21, 0xca, 0xcc]); //
+    test_invalid(&[0x62, 0xf3, 0x7d, 0x88, 0x21, 0xca, 0xcc]); //
+    test_invalid(&[0x62, 0xf3, 0xfd, 0x18, 0x22, 0xca, 0xcc]); //
+    test_invalid(&[0x62, 0xf3, 0xfd, 0x88, 0x22, 0xca, 0xcc]); //
+    test_invalid(&[0x62, 0xf3, 0x7d, 0xad, 0x3e, 0xca, 0xcc]); // no zero mask-merge
+    test_invalid(&[0x62, 0xf3, 0x7d, 0x38, 0x42, 0x0a, 0xcc]); // no broadcast
+    test_invalid(&[0x62, 0xf3, 0xfd, 0x29, 0x44, 0xca, 0xcc]); // mask reg must be 000
+    test_invalid(&[0x62, 0xf3, 0xfd, 0x18, 0x44, 0x0a, 0xcc]); //
+    test_invalid(&[0x62, 0xf3, 0xfd, 0x88, 0x44, 0x0a, 0xcc]); //
+    test_invalid(&[0x62, 0xf3, 0xfd, 0xbd, 0x66, 0x0a, 0xcc]); // no zero mask-merge
 }