Add PALIGNR x86 instruction (asm & semantic)

3 files changed, 35 insertions, 0 deletions

diff --git a/miasm2/arch/x86/arch.py b/miasm2/arch/x86/arch.py
index e86665f7..ad5df770 100644
--- a/miasm2/arch/x86/arch.py
+++ b/miasm2/arch/x86/arch.py
@@ -2404,6 +2404,11 @@ class x86_rm_xmm_m64(x86_rm_mm):
     is_mm = False
     is_xmm = True
 
+class x86_rm_xmm_m128(x86_rm_mm):
+    msize = 128
+    is_mm = False
+    is_xmm = True
+
 
 class x86_rm_xmm_reg(x86_rm_mm):
     msize = None
@@ -3281,6 +3286,7 @@ rm_arg_mm_reg = bs(l=0, cls=(x86_rm_mm_reg,), fname='rmarg')
 rm_arg_xmm = bs(l=0, cls=(x86_rm_xmm,), fname='rmarg')
 rm_arg_xmm_m32 = bs(l=0, cls=(x86_rm_xmm_m32,), fname='rmarg')
 rm_arg_xmm_m64 = bs(l=0, cls=(x86_rm_xmm_m64,), fname='rmarg')
+rm_arg_xmm_m128 = bs(l=0, cls=(x86_rm_xmm_m128,), fname='rmarg')
 rm_arg_xmm_reg = bs(l=0, cls=(x86_rm_xmm_reg,), fname='rmarg')
 
 rm_arg_bnd = bs(l=0, cls=(x86_rm_bnd,), fname='rmarg')
@@ -4212,6 +4218,11 @@ addop("cvttsd2si",[bs8(0x0f), bs8(0x2c), pref_f2]
 addop("cvttss2si",[bs8(0x0f), bs8(0x2c), pref_f3]
       + rmmod(reg, rm_arg_xmm_m32))
 
+addop("palignr", [bs8(0x0f), bs8(0x73), bs8(0x0f), no_xmm_pref] +
+      rmmod(mm_reg, rm_arg_mm_m64) + [u08], [mm_reg, rm_arg_mm_m64, u08])
+addop("palignr", [bs8(0x0f), bs8(0x3a), bs8(0x0f), pref_66] +
+      rmmod(xmm_reg, rm_arg_xmm_m128) + [u08], [xmm_reg, rm_arg_xmm_m128, u08])
+
 addop("psrlq", [bs8(0x0f), bs8(0x73), no_xmm_pref] +
       rmmod(d2, rm_arg_mm) + [u08], [rm_arg_mm, u08])
 addop("psrlq", [bs8(0x0f), bs8(0x73), pref_66] +
diff --git a/miasm2/arch/x86/sem.py b/miasm2/arch/x86/sem.py
index 6f2c7947..b2403604 100644
--- a/miasm2/arch/x86/sem.py
+++ b/miasm2/arch/x86/sem.py
@@ -4133,6 +4133,26 @@ def bndmov(ir, instr, dst, src):
     # Implemented as a NOP, because BND side effects are not yet supported
     return [], []
 
+def palignr(ir, instr, dst, src, imm):
+    # dst.src >> imm * 8 [:dst.size]
+
+    shift = int(imm) * 8
+    if shift == 0:
+        result = src
+    elif shift == src.size:
+        result = dst
+    elif shift > src.size:
+        result = dst >> m2_expr.ExprInt(shift - src.size, dst.size)
+    else:
+        # shift < src.size
+        result = m2_expr.ExprCompose(
+            src[shift:],
+            dst[:shift],
+        )
+
+    return [m2_expr.ExprAff(dst, result)], []
+
+
 mnemo_func = {'mov': mov,
               'xchg': xchg,
               'movzx': movzx,
@@ -4569,6 +4589,7 @@ mnemo_func = {'mov': mov,
               "pslld": pslld,
               "psllq": psllq,
               "pslldq": pslldq,
+              "palignr": palignr,
 
               "pmaxub": pmaxub,
               "pmaxuw": pmaxuw,
diff --git a/test/arch/x86/arch.py b/test/arch/x86/arch.py
index 4d5d5a1c..c1818bc9 100644
--- a/test/arch/x86/arch.py
+++ b/test/arch/x86/arch.py
@@ -2802,6 +2802,9 @@ reg_tests = [
     (m32, "00000000    PCMPGTQ    XMM0, XMM5",
     "660f3837C5"),
 
+    (m64, "00000000    PALIGNR    XMM1, XMM2, 0xC",
+     "660f3a0fca0c"),
+
 
     (m32, "00000000    PUNPCKHBW  MM2, QWORD PTR [EDX]",
     "0F6812"),