[machinst x64]: add shuffle implementation

cranelift/codegen/src/isa/x64/inst/args.rs

@@ -425,6 +425,7 @@ pub enum SseOpcode {
     Pmulld,
     Pmullw,
     Pmuludq,
+    Pshufb,
     Pshufd,
     Psllw,
     Pslld,
@@ -557,7 +558,7 @@ impl SseOpcode {
             | SseOpcode::Ucomisd
             | SseOpcode::Xorpd => SSE2,
 
-            SseOpcode::Pabsb | SseOpcode::Pabsw | SseOpcode::Pabsd => SSSE3,
+            SseOpcode::Pabsb | SseOpcode::Pabsw | SseOpcode::Pabsd | SseOpcode::Pshufb => SSSE3,
 
             SseOpcode::Insertps
             | SseOpcode::Pextrb
@@ -672,6 +673,7 @@ impl fmt::Debug for SseOpcode {
             SseOpcode::Pmulld => "pmulld",
             SseOpcode::Pmullw => "pmullw",
             SseOpcode::Pmuludq => "pmuludq",
+            SseOpcode::Pshufb => "pshufb",
             SseOpcode::Pshufd => "pshufd",
             SseOpcode::Psllw => "psllw",
             SseOpcode::Pslld => "pslld",

cranelift/codegen/src/isa/x64/inst/emit.rs

@@ -1797,6 +1797,7 @@ pub(crate) fn emit(
                 SseOpcode::Pmulld => (LegacyPrefixes::_66, 0x0F3840, 3),
                 SseOpcode::Pmullw => (LegacyPrefixes::_66, 0x0FD5, 2),
                 SseOpcode::Pmuludq => (LegacyPrefixes::_66, 0x0FF4, 2),
+                SseOpcode::Pshufb => (LegacyPrefixes::_66, 0x0F3800, 3),
                 SseOpcode::Psubb => (LegacyPrefixes::_66, 0x0FF8, 2),
                 SseOpcode::Psubd => (LegacyPrefixes::_66, 0x0FFA, 2),
                 SseOpcode::Psubq => (LegacyPrefixes::_66, 0x0FFB, 2),

cranelift/codegen/src/isa/x64/inst/emit_tests.rs

@@ -3243,6 +3243,12 @@ fn test_x64_emit() {
         "pxor    %xmm11, %xmm2",
     ));
 
+    insns.push((
+        Inst::xmm_rm_r(SseOpcode::Pshufb, RegMem::reg(xmm11), w_xmm2),
+        "66410F3800D3",
+        "pshufb  %xmm11, %xmm2",
+    ));
+
     // XMM_Mov_R_M: float stores
     insns.push((
         Inst::xmm_mov_r_m(SseOpcode::Movss, xmm15, Amode::imm_reg(128, r12), None),