From 979a3232f9d23c34b6c8c86fb3913fb34b3da333 Mon Sep 17 00:00:00 2001
From: xctan <xctan@cirno.icu>
Date: Fri, 15 Nov 2024 00:41:28 +0800
Subject: [PATCH] [RV64_DYNAREC] Added more MMX opcodes for vector (#2035)

* [RV64_DYNAREC] Added 0F 68 PUNPCKHBW opcode

* [RV64_DYNAREC] Added 0F 69 PUNPCKHWD opcode

* [RV64_DYNAREC] Added 0F 6A PUNPCKHDQ opcode

* [RV64_DYNAREC] Updated 0F 68-69 PUNPCKHBW/WD opcodes

* [RV64_DYNAREC] Added 0F 60 PUNPCKLBW opcode

* [RV64_DYNAREC] Added 0F 61 PUNPCKLWD opcode

* [RV64_DYNAREC] Added 0F 62 PUNPCKLDQ opcode

* [RV64_DYNAREC] Simplified MMX PUNPCK{L,H}{BW,WD,DQ}
---
 src/dynarec/rv64/dynarec_rv64_0f_vector.c | 101 ++++++++++++++++++++++
 1 file changed, 101 insertions(+)

diff --git a/src/dynarec/rv64/dynarec_rv64_0f_vector.c b/src/dynarec/rv64/dynarec_rv64_0f_vector.c
index f5c3a4d8..195d58f7 100644
--- a/src/dynarec/rv64/dynarec_rv64_0f_vector.c
+++ b/src/dynarec/rv64/dynarec_rv64_0f_vector.c
@@ -483,6 +483,52 @@ uintptr_t dynarec64_0F_vector(dynarec_rv64_t* dyn, uintptr_t addr, uintptr_t ip,
                 VADD_VX(q0, q1, xZR, VECTOR_MASKED);
             }
             break;
+        case 0x60:
+            INST_NAME("PUNPCKLBW Gm, Em");
+            nextop = F8;
+            GETGM_vector(q0);
+            SET_ELEMENT_WIDTH(x1, VECTOR_SEW64, 1);
+            GETEM_vector(q1, 0);
+            SET_ELEMENT_WIDTH(x1, VECTOR_SEW8, 1);
+            v0 = fpu_get_scratch_lmul(dyn, VECTOR_LMUL2);
+            d0 = fpu_get_scratch_lmul(dyn, VECTOR_LMUL2);
+            VWADDU_VX(d0, q0, xZR, VECTOR_UNMASKED);
+            VWADDU_VX(v0, q1, xZR, VECTOR_UNMASKED);
+            SET_ELEMENT_WIDTH(x1, VECTOR_SEW16, 1);
+            VSLL_VI(v0, v0, 8, VECTOR_UNMASKED);
+            VOR_VV(q0, d0, v0, VECTOR_UNMASKED);
+            break;
+        case 0x61:
+            INST_NAME("PUNPCKLWD Gm, Em");
+            nextop = F8;
+            GETGM_vector(q0);
+            SET_ELEMENT_WIDTH(x1, VECTOR_SEW64, 1);
+            GETEM_vector(q1, 0);
+            SET_ELEMENT_WIDTH(x1, VECTOR_SEW16, 1);
+            v0 = fpu_get_scratch_lmul(dyn, VECTOR_LMUL2);
+            d0 = fpu_get_scratch_lmul(dyn, VECTOR_LMUL2);
+            VWADDU_VX(d0, q0, xZR, VECTOR_UNMASKED);
+            VWADDU_VX(v0, q1, xZR, VECTOR_UNMASKED);
+            SET_ELEMENT_WIDTH(x1, VECTOR_SEW32, 1);
+            VSLL_VI(v0, v0, 16, VECTOR_UNMASKED);
+            VOR_VV(q0, d0, v0, VECTOR_UNMASKED);
+            break;
+        case 0x62:
+            INST_NAME("PUNPCKLDQ Gm, Em");
+            nextop = F8;
+            GETGM_vector(q0);
+            SET_ELEMENT_WIDTH(x1, VECTOR_SEW64, 1);
+            GETEM_vector(q1, 0);
+            SET_ELEMENT_WIDTH(x1, VECTOR_SEW32, 1);
+            v0 = fpu_get_scratch_lmul(dyn, VECTOR_LMUL2);
+            d0 = fpu_get_scratch_lmul(dyn, VECTOR_LMUL2);
+            MOV32w(x2, 32);
+            VWADDU_VX(d0, q0, xZR, VECTOR_UNMASKED);
+            VWADDU_VX(v0, q1, xZR, VECTOR_UNMASKED);
+            SET_ELEMENT_WIDTH(x1, VECTOR_SEW64, 1);
+            VSLL_VX(v0, v0, x2, VECTOR_UNMASKED);
+            VOR_VV(q0, d0, v0, VECTOR_UNMASKED);
+            break;
         case 0x63:
             INST_NAME("PACKSSWB Gm, Em");
             nextop = F8;
@@ -515,6 +561,61 @@ uintptr_t dynarec64_0F_vector(dynarec_rv64_t* dyn, uintptr_t addr, uintptr_t ip,
             SET_ELEMENT_WIDTH(x1, VECTOR_SEW8, 1);
             VNCLIPU_WI(q0, d0, 0, VECTOR_UNMASKED);
             break;
+        case 0x68:
+            INST_NAME("PUNPCKHBW Gm, Em");
+            nextop = F8;
+            GETGM_vector(q0);
+            SET_ELEMENT_WIDTH(x1, VECTOR_SEW64, 1);
+            GETEM_vector(q1, 0);
+            SET_ELEMENT_WIDTH(x1, VECTOR_SEW8, 1);
+            v0 = fpu_get_scratch_lmul(dyn, VECTOR_LMUL2);
+            v1 = fpu_get_scratch_lmul(dyn, VECTOR_LMUL1);
+            d0 = fpu_get_scratch_lmul(dyn, VECTOR_LMUL2);
+            VSLIDEDOWN_VI(v0, q0, 4, VECTOR_UNMASKED);
+            VSLIDEDOWN_VI(v1, q1, 4, VECTOR_UNMASKED);
+            VWADDU_VX(d0, v0, xZR, VECTOR_UNMASKED);
+            VWADDU_VX(v0, v1, xZR, VECTOR_UNMASKED);
+            SET_ELEMENT_WIDTH(x1, VECTOR_SEW16, 1);
+            VSLL_VI(v0, v0, 8, VECTOR_UNMASKED);
+            VOR_VV(q0, d0, v0, VECTOR_UNMASKED);
+            break;
+        case 0x69:
+            INST_NAME("PUNPCKHWD Gm, Em");
+            nextop = F8;
+            GETGM_vector(q0);
+            SET_ELEMENT_WIDTH(x1, VECTOR_SEW64, 1);
+            GETEM_vector(q1, 0);
+            SET_ELEMENT_WIDTH(x1, VECTOR_SEW16, 1);
+            v0 = fpu_get_scratch_lmul(dyn, VECTOR_LMUL2);
+            v1 = fpu_get_scratch_lmul(dyn, VECTOR_LMUL1);
+            d0 = fpu_get_scratch_lmul(dyn, VECTOR_LMUL2);
+            VSLIDEDOWN_VI(v0, q0, 2, VECTOR_UNMASKED);
+            VSLIDEDOWN_VI(v1, q1, 2, VECTOR_UNMASKED);
+            VWADDU_VX(d0, v0, xZR, VECTOR_UNMASKED);
+            VWADDU_VX(v0, v1, xZR, VECTOR_UNMASKED);
+            SET_ELEMENT_WIDTH(x1, VECTOR_SEW32, 1);
+            VSLL_VI(v0, v0, 16, VECTOR_UNMASKED);
+            VOR_VV(q0, d0, v0, VECTOR_UNMASKED);
+            break;
+        case 0x6A:
+            INST_NAME("PUNPCKHDQ Gm, Em");
+            nextop = F8;
+            GETGM_vector(q0);
+            SET_ELEMENT_WIDTH(x1, VECTOR_SEW64, 1);
+            GETEM_vector(q1, 0);
+            SET_ELEMENT_WIDTH(x1, VECTOR_SEW32, 1);
+            MOV32w(x2, 32);
+            v0 = fpu_get_scratch_lmul(dyn, VECTOR_LMUL2);
+            v1 = fpu_get_scratch_lmul(dyn, VECTOR_LMUL1);
+            d0 = fpu_get_scratch_lmul(dyn, VECTOR_LMUL2);
+            VSLIDEDOWN_VI(v0, q0, 1, VECTOR_UNMASKED);
+            VSLIDEDOWN_VI(v1, q1, 1, VECTOR_UNMASKED);
+            VWADDU_VX(d0, v0, xZR, VECTOR_UNMASKED);
+            VWADDU_VX(v0, v1, xZR, VECTOR_UNMASKED);
+            SET_ELEMENT_WIDTH(x1, VECTOR_SEW64, 1);
+            VSLL_VX(v0, v0, x2, VECTOR_UNMASKED);
+            VOR_VV(q0, d0, v0, VECTOR_UNMASKED);
+            break;
         case 0x6B:
             INST_NAME("PACKSSDW Gm, Em");
             nextop = F8;