Intel encodings for nearest/floor/ceil/trunc.

These floating point rounding operations all use the roundss/roundsd
instructions that are available in SSE 4.1.

lib/cretonne/meta/gen_binemit.py

@@ -38,6 +38,7 @@ def gen_recipe(recipe, fmt):
     with fmt.indented(
             'if let InstructionData::{} {{'.format(iform.name),
             '}'):
+        fmt.line('opcode,')
         for f in iform.imm_fields:
             fmt.line('{},'.format(f.member))
         if want_args:

lib/cretonne/meta/isa/intel/encodings.py

@@ -11,9 +11,10 @@ from . import settings as cfg
 from . import instructions as x86
 from .legalize import intel_expand
 from base.legalize import narrow, expand
+from .settings import use_sse41
 
 try:
-    from typing import TYPE_CHECKING
+    from typing import TYPE_CHECKING, Any  # noqa
     if TYPE_CHECKING:
         from cdsl.instructions import MaybeBoundInst  # noqa
 except ImportError:
@@ -82,7 +83,7 @@ def enc_i32_i64_ld_st(inst, w_bit, recipe, *args, **kwargs):
 
 
 def enc_flt(inst, recipe, *args, **kwargs):
-    # type: (MaybeBoundInst, r.TailRecipe, *int, **int) -> None
+    # type: (MaybeBoundInst, r.TailRecipe, *int, **Any) -> None
     """
     Add encodings for floating point instruction `inst` to both I32 and I64.
     """
@@ -363,6 +364,16 @@ enc_flt(base.fdemote.f32.f64, r.furm, 0xf2, 0x0f, 0x5a)
 enc_flt(base.sqrt.f32, r.furm, 0xf3, 0x0f, 0x51)
 enc_flt(base.sqrt.f64, r.furm, 0xf2, 0x0f, 0x51)
 
+# Rounding. The recipe looks at the opcode to pick an immediate.
+for inst in [
+        base.nearest,
+        base.floor,
+        base.ceil,
+        base.trunc]:
+    enc_flt(inst.f32, r.furmi_rnd, 0x66, 0x0f, 0x3a, 0x0a, isap=use_sse41)
+    enc_flt(inst.f64, r.furmi_rnd, 0x66, 0x0f, 0x3a, 0x0b, isap=use_sse41)
+
+
 # Binary arithmetic ops.
 for inst,           opc in [
         (base.fadd, 0x58),

lib/cretonne/meta/isa/intel/recipes.py

@@ -289,6 +289,21 @@ frurm = TailRecipe(
         modrm_rr(in_reg0, out_reg0, sink);
         ''')
 
+# XX /r, RMI form for one of the roundXX SSE 4.1 instructions.
+furmi_rnd = TailRecipe(
+        'furmi_rnd', Unary, size=2, ins=FPR, outs=FPR,
+        emit='''
+        PUT_OP(bits, rex2(in_reg0, out_reg0), sink);
+        modrm_rr(in_reg0, out_reg0, sink);
+        sink.put1(match opcode {
+            Opcode::Nearest => 0b00,
+            Opcode::Floor => 0b01,
+            Opcode::Ceil => 0b10,
+            Opcode::Trunc => 0b11,
+            x => panic!("{} unexpected for furmi_rnd", opcode),
+        });
+        ''')
+
 # XX /r, for regmove instructions.
 rmov = TailRecipe(
         'ur', RegMove, size=1, ins=GPR, outs=(),