[dynarmic] use (v)pshufb (SSSE3+) for VREV32/64 emits (#2851)

Signed-off-by: lizzie <lizzie@eden-emu.dev>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/2851
Reviewed-by: crueter <crueter@eden-emu.dev>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Reviewed-by: Caio Oliveira <caiooliveirafarias0@gmail.com>
Co-authored-by: lizzie <lizzie@eden-emu.dev>
Co-committed-by: lizzie <lizzie@eden-emu.dev>

src/dynarmic/src/dynarmic/backend/x64/emit_x64_vector.cpp

@@ -3521,48 +3521,47 @@ void EmitX64::EmitVectorReverseElementsInHalfGroups8(EmitContext& ctx, IR::Inst*
 
 void EmitX64::EmitVectorReverseElementsInWordGroups8(EmitContext& ctx, IR::Inst* inst) {
     auto args = ctx.reg_alloc.GetArgumentInfo(inst);
-
     const Xbyak::Xmm data = ctx.reg_alloc.UseScratchXmm(args[0]);
+    if (code.HasHostFeature(HostFeature::AVX)) {
+        code.vpshufb(data, data, code.Const(xword, 0x0405060700010203, 0x0c0d0e0f08090a0b));
+    } else if (code.HasHostFeature(HostFeature::SSSE3)) {
+        code.pshufb(data, code.Const(xword, 0x0405060700010203, 0x0c0d0e0f08090a0b));
+    } else {
         const Xbyak::Xmm tmp = ctx.reg_alloc.ScratchXmm();
-
-    // TODO: PSHUFB
-
         code.movdqa(tmp, data);
         code.psllw(tmp, 8);
         code.psrlw(data, 8);
         code.por(data, tmp);
         code.pshuflw(data, data, 0b10110001);
         code.pshufhw(data, data, 0b10110001);
-
+    }
     ctx.reg_alloc.DefineValue(inst, data);
 }
 
 void EmitX64::EmitVectorReverseElementsInWordGroups16(EmitContext& ctx, IR::Inst* inst) {
     auto args = ctx.reg_alloc.GetArgumentInfo(inst);
-
     const Xbyak::Xmm data = ctx.reg_alloc.UseScratchXmm(args[0]);
-
     code.pshuflw(data, data, 0b10110001);
     code.pshufhw(data, data, 0b10110001);
-
     ctx.reg_alloc.DefineValue(inst, data);
 }
 
 void EmitX64::EmitVectorReverseElementsInLongGroups8(EmitContext& ctx, IR::Inst* inst) {
     auto args = ctx.reg_alloc.GetArgumentInfo(inst);
-
     const Xbyak::Xmm data = ctx.reg_alloc.UseScratchXmm(args[0]);
+    if (code.HasHostFeature(HostFeature::AVX)) {
+        code.vpshufb(data, data, code.Const(xword, 0x0001020304050607, 0x08090a0b0c0d0e0f));
+    } else if (code.HasHostFeature(HostFeature::SSSE3)) {
+        code.pshufb(data, code.Const(xword, 0x0001020304050607, 0x08090a0b0c0d0e0f));
+    } else {
         const Xbyak::Xmm tmp = ctx.reg_alloc.ScratchXmm();
-
-    // TODO: PSHUFB
-
         code.movdqa(tmp, data);
         code.psllw(tmp, 8);
         code.psrlw(data, 8);
         code.por(data, tmp);
         code.pshuflw(data, data, 0b00011011);
         code.pshufhw(data, data, 0b00011011);
-
+    }
     ctx.reg_alloc.DefineValue(inst, data);
 }
 

src/dynarmic/tests/A64/a64.cpp

@@ -91,6 +91,28 @@ TEST_CASE("A64: CLZ", "[a64]") {
     REQUIRE(jit.GetVector(5) == Vector{0x0, 0x0000001e0000001f});
 }
 
+TEST_CASE("A64: VREV", "[a64]") {
+    A64TestEnv env;
+    A64::UserConfig jit_user_config{};
+    jit_user_config.callbacks = &env;
+    A64::Jit jit{jit_user_config};
+    oaknut::VectorCodeGenerator code{env.code_mem, nullptr};
+    code.REV32(V0.B16(), V5.B16());
+    code.REV32(V1.H8(), V5.H8());
+    code.REV64(V2.B16(), V5.B16());
+    code.REV64(V3.H8(), V5.H8());
+    code.REV64(V4.S4(), V5.S4());
+    jit.SetPC(0);
+    jit.SetVector(5, {0x1020304050607080, 0x90A0B0C0D0E0F000});
+    env.ticks_left = env.code_mem.size();
+    CheckedRun([&]() { jit.Run(); });
+    REQUIRE(jit.GetVector(0) == Vector{0x4030201080706050, 0xc0b0a09000f0e0d0});
+    REQUIRE(jit.GetVector(1) == Vector{0x3040102070805060, 0xb0c090a0f000d0e0});
+    REQUIRE(jit.GetVector(2) == Vector{0x8070605040302010, 0x00f0e0d0c0b0a090});
+    REQUIRE(jit.GetVector(3) == Vector{0x7080506030401020, 0xf000d0e0b0c090a0});
+    REQUIRE(jit.GetVector(4) == Vector{0x5060708010203040, 0xd0e0f00090a0b0c0});
+}
+
 TEST_CASE("A64: UADDL{V,P}", "[a64]") {
     A64TestEnv env;
     A64::UserConfig jit_user_config{};