[xen master] x86emul: handle AVX512-FP16 conversion to/from (packed) int16 insns

[patchbot@xxxxxxx]

commit 59c2c9c2a003d61a2c5f2e7f88342db765c283c3
Author:     Jan Beulich <jbeulich@xxxxxxxx>
AuthorDate: Mon Jun 5 15:00:15 2023 +0200
Commit:     Jan Beulich <jbeulich@xxxxxxxx>
CommitDate: Mon Jun 5 15:00:15 2023 +0200

    x86emul: handle AVX512-FP16 conversion to/from (packed) int16 insns
    
    These are easiest in that they have same-size source and destination
    vectors, yet they're different from other conversion insns in that they
    use opcodes which have different meaning in the 0F encoding space
    ({,V}H{ADD,SUB}P{S,D}), hence requiring a little bit of overriding.
    
    Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
    Acked-by: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
---
 tools/tests/x86_emulator/evex-disp8.c  | 6 ++++++
 tools/tests/x86_emulator/predicates.c  | 6 ++++++
 xen/arch/x86/x86_emulate/decode.c      | 8 +++++++-
 xen/arch/x86/x86_emulate/x86_emulate.c | 8 ++++++++
 4 files changed, 27 insertions(+), 1 deletion(-)

diff --git a/tools/tests/x86_emulator/evex-disp8.c 
b/tools/tests/x86_emulator/evex-disp8.c
index b5329fcfc3..381cb488fe 100644
--- a/tools/tests/x86_emulator/evex-disp8.c
+++ b/tools/tests/x86_emulator/evex-disp8.c
@@ -612,6 +612,12 @@ static const struct test avx512_fp16_all[] = {
     INSN(cmpph,           , 0f3a, c2,    vl, fp16, vl),
     INSN(cmpsh,         f3, 0f3a, c2,    el, fp16, el),
     INSN(comish,          , map5, 2f,    el, fp16, el),
+    INSN(cvtph2uw,        , map5, 7d,    vl, fp16, vl),
+    INSN(cvtph2w,       66, map5, 7d,    vl, fp16, vl),
+    INSN(cvttph2uw,       , map5, 7c,    vl, fp16, vl),
+    INSN(cvttph2w,      66, map5, 7c,    vl, fp16, vl),
+    INSN(cvtuw2ph,      f2, map5, 7d,    vl, fp16, vl),
+    INSN(cvtw2ph,       f3, map5, 7d,    vl, fp16, vl),
     INSN(divph,           , map5, 5e,    vl, fp16, vl),
     INSN(divsh,         f3, map5, 5e,    el, fp16, el),
     INSNX(fcmaddcph,    f2, map6, 56, 1, vl,    d, vl),
diff --git a/tools/tests/x86_emulator/predicates.c 
b/tools/tests/x86_emulator/predicates.c
index acbd32a5be..614432a42d 100644
--- a/tools/tests/x86_emulator/predicates.c
+++ b/tools/tests/x86_emulator/predicates.c
@@ -2064,6 +2064,12 @@ static const struct evex {
     { { 0x5f }, 2, T, R, pfx_no, W0, Ln }, /* vmaxph */
     { { 0x5f }, 2, T, R, pfx_f3, W0, LIG }, /* vmaxsh */
     { { 0x6e }, 2, T, R, pfx_66, WIG, L0 }, /* vmovw */
+    { { 0x7c }, 2, T, R, pfx_no, W0, Ln }, /* vcvttph2uw */
+    { { 0x7c }, 2, T, R, pfx_66, W0, Ln }, /* vcvttph2w */
+    { { 0x7d }, 2, T, R, pfx_no, W0, Ln }, /* vcvtph2uw */
+    { { 0x7d }, 2, T, R, pfx_66, W0, Ln }, /* vcvtph2w */
+    { { 0x7d }, 2, T, R, pfx_f3, W0, Ln }, /* vcvtw2ph */
+    { { 0x7d }, 2, T, R, pfx_f2, W0, Ln }, /* vcvtuwph */
     { { 0x7e }, 2, T, W, pfx_66, WIG, L0 }, /* vmovw */
 }, evex_map6[] = {
     { { 0x2c }, 2, T, R, pfx_66, W0, Ln }, /* vscalefph */
diff --git a/xen/arch/x86/x86_emulate/decode.c 
b/xen/arch/x86/x86_emulate/decode.c
index aef64d9854..a5baa3ed62 100644
--- a/xen/arch/x86/x86_emulate/decode.c
+++ b/xen/arch/x86/x86_emulate/decode.c
@@ -247,7 +247,7 @@ static const struct twobyte_table {
     [0x78 ... 0x79] = { DstImplicit|SrcMem|ModRM|Mov, simd_other, d8s_vl },
     [0x7a] = { DstImplicit|SrcMem|ModRM|Mov, simd_packed_fp, d8s_vl },
     [0x7b] = { DstImplicit|SrcMem|ModRM|Mov, simd_other, d8s_dq64 },
-    [0x7c ... 0x7d] = { DstImplicit|SrcMem|ModRM, simd_other },
+    [0x7c ... 0x7d] = { DstImplicit|SrcMem|ModRM, simd_other, d8s_vl },
     [0x7e] = { DstMem|SrcImplicit|ModRM|Mov, simd_none, d8s_dq64 },
     [0x7f] = { DstMem|SrcImplicit|ModRM|Mov, simd_packed_int, d8s_vl },
     [0x80 ... 0x8f] = { DstImplicit|SrcImm },
@@ -1485,6 +1485,12 @@ int x86emul_decode(struct x86_emulate_state *s,
                     s->fp16 = true;
                 s->simd_size = simd_none;
                 break;
+
+            case 0x7c: /* vcvttph2{,u}w */
+            case 0x7d: /* vcvtph2{,u}w / vcvt{,u}w2ph */
+                d = DstReg | SrcMem | TwoOp;
+                s->fp16 = true;
+                break;
             }
 
             /* Like above re-use twobyte_table[] here. */
diff --git a/xen/arch/x86/x86_emulate/x86_emulate.c 
b/xen/arch/x86/x86_emulate/x86_emulate.c
index 5c1ec668de..eb40938b0f 100644
--- a/xen/arch/x86/x86_emulate/x86_emulate.c
+++ b/xen/arch/x86/x86_emulate/x86_emulate.c
@@ -7781,6 +7781,14 @@ x86_emulate(
         generate_exception_if(evex.w, X86_EXC_UD);
         goto avx512f_all_fp;
 
+    case X86EMUL_OPC_EVEX   (5, 0x7c): /* vcvttph2uw [xyz]mm/mem,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(5, 0x7c): /* vcvttph2w [xyz]mm/mem,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX   (5, 0x7d): /* vcvtph2uw [xyz]mm/mem,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(5, 0x7d): /* vcvtph2w [xyz]mm/mem,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_F3(5, 0x7d): /* vcvtw2ph [xyz]mm/mem,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_F2(5, 0x7d): /* vcvtuw2ph [xyz]mm/mem,[xyz]mm{k} */
+        op_bytes = 16 << evex.lr;
+        /* fall through */
     case X86EMUL_OPC_EVEX_66(6, 0x2c): /* vscalefph 
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
     case X86EMUL_OPC_EVEX_66(6, 0x42): /* vgetexpph [xyz]mm/mem,[xyz]mm{k} */
     case X86EMUL_OPC_EVEX_66(6, 0x96): /* vfmaddsub132ph 
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
--
generated by git-patchbot for /home/xen/git/xen.git#master