add vcompressb, vcompressw
diff --git a/gen/gen_avx512.cpp b/gen/gen_avx512.cpp
index 1c8cf9b..8a11b7d 100644
--- a/gen/gen_avx512.cpp
+++ b/gen/gen_avx512.cpp
@@ -228,6 +228,9 @@
{ 0x8B, "vpcompressd", T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N4 },
{ 0x8B, "vpcompressq", T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW1 | T_N8 },
+
+ { 0x63, "vcompressb", T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N1 },
+ { 0x63, "vcompressw", T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW1 | T_N2 },
};
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
const Tbl *p = &tbl[i];
diff --git a/test/misc.cpp b/test/misc.cpp
index 5fa5ea7..e7db693 100644
--- a/test/misc.cpp
+++ b/test/misc.cpp
@@ -199,5 +199,44 @@
CYBOZU_TEST_EQUAL(c.getSize(), n);
CYBOZU_TEST_EQUAL_ARRAY(c.getCode(), tbl, n);
}
+CYBOZU_TEST_AUTO(vcompressb_w)
+{
+ struct Code : Xbyak::CodeGenerator {
+ Code()
+ {
+ vcompressb(ptr[rax + 64], xmm1);
+ vcompressb(xmm30 | k5, xmm1);
+ vcompressb(ptr[rax + 64], ymm1);
+ vcompressb(ymm30 | k3 |T_z, ymm1);
+ vcompressb(ptr[rax + 64], zmm1);
+ vcompressb(zmm30 | k2 |T_z, zmm1);
+ vcompressw(ptr[rax + 64], xmm1);
+ vcompressw(xmm30 | k5, xmm1);
+ vcompressw(ptr[rax + 64], ymm1);
+ vcompressw(ymm30 | k3 |T_z, ymm1);
+ vcompressw(ptr[rax + 64], zmm1);
+ vcompressw(zmm30 | k2 |T_z, zmm1);
+ }
+ } c;
+ const uint8_t tbl[] = {
+ 0x62, 0xf2, 0x7d, 0x08, 0x63, 0x48, 0x40,
+ 0x62, 0x92, 0x7d, 0x0d, 0x63, 0xce,
+ 0x62, 0xf2, 0x7d, 0x28, 0x63, 0x48, 0x40,
+ 0x62, 0x92, 0x7d, 0xab, 0x63, 0xce,
+ 0x62, 0xf2, 0x7d, 0x48, 0x63, 0x48, 0x40,
+ 0x62, 0x92, 0x7d, 0xca, 0x63, 0xce,
+
+
+ 0x62, 0xf2, 0xfd, 0x08, 0x63, 0x48, 0x20,
+ 0x62, 0x92, 0xfd, 0x0d, 0x63, 0xce,
+ 0x62, 0xf2, 0xfd, 0x28, 0x63, 0x48, 0x20,
+ 0x62, 0x92, 0xfd, 0xab, 0x63, 0xce,
+ 0x62, 0xf2, 0xfd, 0x48, 0x63, 0x48, 0x20,
+ 0x62, 0x92, 0xfd, 0xca, 0x63, 0xce,
+ };
+ const size_t n = sizeof(tbl) / sizeof(tbl[0]);
+ CYBOZU_TEST_EQUAL(c.getSize(), n);
+ CYBOZU_TEST_EQUAL_ARRAY(c.getCode(), tbl, n);
+}
#endif
diff --git a/xbyak/xbyak_mnemonic.h b/xbyak/xbyak_mnemonic.h
index 5591bb3..79aa8a2 100644
--- a/xbyak/xbyak_mnemonic.h
+++ b/xbyak/xbyak_mnemonic.h
@@ -1662,8 +1662,10 @@
void vcmpps(const Opmask& k, const Xmm& x, const Operand& op, uint8 imm) { opAVX_K_X_XM(k, x, op, T_0F | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX, 0xC2, imm); }
void vcmpsd(const Opmask& k, const Xmm& x, const Operand& op, uint8 imm) { opAVX_K_X_XM(k, x, op, T_N8 | T_F2 | T_0F | T_EW1 | T_SAE_Z | T_MUST_EVEX, 0xC2, imm); }
void vcmpss(const Opmask& k, const Xmm& x, const Operand& op, uint8 imm) { opAVX_K_X_XM(k, x, op, T_N4 | T_F3 | T_0F | T_EW0 | T_SAE_Z | T_MUST_EVEX, 0xC2, imm); }
+void vcompressb(const Operand& op, const Xmm& x) { opAVX_X_XM_IMM(x, op, T_N1 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x63); }
void vcompresspd(const Operand& op, const Xmm& x) { opAVX_X_XM_IMM(x, op, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX, 0x8A); }
void vcompressps(const Operand& op, const Xmm& x) { opAVX_X_XM_IMM(x, op, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x8A); }
+void vcompressw(const Operand& op, const Xmm& x) { opAVX_X_XM_IMM(x, op, T_N2 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX, 0x63); }
void vcvtpd2qq(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F | T_EW1 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B64, 0x7B); }
void vcvtpd2udq(const Xmm& x, const Operand& op) { opCvt2(x, op, T_0F | T_YMM | T_MUST_EVEX | T_EW1 | T_B64 | T_ER_Z, 0x79); }
void vcvtpd2uqq(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F | T_EW1 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B64, 0x79); }
