Add initial integrated assembler w/ some Xmm ops.

tests_lit/llvm2ice_tests/nop-insertion.ll

 ; This is a smoke test of nop insertion.
 
+; Don't use integrated-as because this currently depends on the # variant
+; assembler comment.
 ; RUN: %llvm2ice -rng-seed=1 -nop-insertion -nop-insertion-percentage=50 \
-; RUN:    -max-nops-per-instruction=1 %s | FileCheck %s --check-prefix=PROB50
+; RUN:    -max-nops-per-instruction=1 -integrated-as=false %s \
+; RUN:    | FileCheck %s --check-prefix=PROB50
 ; RUN: %llvm2ice -rng-seed=1 -nop-insertion -nop-insertion-percentage=90 \
-; RUN:    -max-nops-per-instruction=1 %s | FileCheck %s --check-prefix=PROB90
+; RUN:    -max-nops-per-instruction=1 -integrated-as=false %s \
+; RUN:    | FileCheck %s --check-prefix=PROB90
 ; RUN: %llvm2ice -rng-seed=1 -nop-insertion -nop-insertion-percentage=50 \
-; RUN:    -max-nops-per-instruction=2 %s | FileCheck %s --check-prefix=MAXNOPS2
+; RUN:    -max-nops-per-instruction=2 -integrated-as=false %s \
+; RUN:    | FileCheck %s --check-prefix=MAXNOPS2
 
 define <4 x i32> @mul_v4i32(<4 x i32> %a, <4 x i32> %b) {
 entry:
   %res = mul <4 x i32> %a, %b
   ret <4 x i32> %res
 ; PROB50-LABEL: mul_v4i32:
 ; PROB50: nop # variant = 3
 ; PROB50: sub esp, 60
 ; PROB50: nop # variant = 4
 ; PROB50: movups xmmword ptr [esp+32], xmm0
@@ skipping 65 matching lines @@
 ; MAXNOPS2: pshufd xmm1, xmm0, 216
 ; MAXNOPS2: nop # variant = 3
 ; MAXNOPS2: movups xmmword ptr [esp], xmm1
 ; MAXNOPS2: nop # variant = 0
 ; MAXNOPS2: movups xmm0, xmmword ptr [esp]
 ; MAXNOPS2: nop # variant = 2
 ; MAXNOPS2: add esp, 60
 ; MAXNOPS2: nop # variant = 4
 ; MAXNOPS2: ret
 }