Add option to force filetype=asm for testing

tests_lit/llvm2ice_tests/nacl-other-intrinsics.ll

 ; This tests the NaCl intrinsics not related to atomic operations.
 
 ; RUN: %if --need=target_X8632 --command %p2i --filetype=obj --disassemble \
-; RUN:   --target x8632 -i %s --args -O2 -sandbox \
+; RUN:   --target x8632 --sandbox -i %s --args -O2 \
 ; RUN:   -allow-externally-defined-symbols \
 ; RUN:   | %if --need=target_X8632 --command FileCheck %s
 ; RUN: %if --need=target_X8632 --command %p2i --filetype=obj --disassemble \
-; RUN:   --target x8632 -i %s --args -Om1 -sandbox \
+; RUN:   --target x8632 --sandbox -i %s --args -Om1 \
 ; RUN:   -allow-externally-defined-symbols \
 ; RUN:   | %if --need=target_X8632 --command FileCheck %s
 
 ; Do another run w/ O2 and a different check-prefix (otherwise O2 and Om1
 ; share the same "CHECK" prefix). This separate run helps check that
 ; some code is optimized out.
 ; RUN: %if --need=target_X8632 --command %p2i --filetype=obj --disassemble \
-; RUN:   --target x8632 -i %s --args -O2 -sandbox \
+; RUN:   --target x8632 --sandbox -i %s --args -O2 \
 ; RUN:   -allow-externally-defined-symbols \
 ; RUN:   | %if --need=target_X8632 \
 ; RUN:   --command FileCheck --check-prefix=CHECKO2REM %s
 
 ; Do O2 runs without -sandbox to make sure llvm.nacl.read.tp gets
 ; lowered to __nacl_read_tp instead of gs:0x0.
 ; We also know that because it's O2, it'll have the O2REM optimizations.
 ; RUN: %if --need=target_X8632 --command %p2i --filetype=obj --disassemble \
 ; RUN:   --target x8632 -i %s --args -O2 \
 ; RUN:   -allow-externally-defined-symbols \
@@ skipping 28 matching lines @@
 declare i8* @llvm.stacksave()
 declare void @llvm.stackrestore(i8*)
 
 define internal i32 @test_nacl_read_tp() {
 entry:
   %ptr = call i8* @llvm.nacl.read.tp()
   %__1 = ptrtoint i8* %ptr to i32
   ret i32 %__1
 }
 ; CHECK-LABEL: test_nacl_read_tp
-; CHECK: mov e{{.*}},DWORD PTR gs:0x0
+; CHECK: mov e{{.*}},{{(DWORD PTR )?}}gs:0x0
 ; CHECKO2REM-LABEL: test_nacl_read_tp
-; CHECKO2REM: mov e{{.*}},DWORD PTR gs:0x0
+; CHECKO2REM: mov e{{.*}},{{(DWORD PTR )?}}gs:0x0
 ; CHECKO2UNSANDBOXEDREM-LABEL: test_nacl_read_tp
 ; CHECKO2UNSANDBOXEDREM: call {{.*}} R_{{.*}} __nacl_read_tp
 
 define internal i32 @test_nacl_read_tp_more_addressing() {
 entry:
   %ptr = call i8* @llvm.nacl.read.tp()
   %__1 = ptrtoint i8* %ptr to i32
   %x = add i32 %__1, %__1
   %__3 = inttoptr i32 %x to i32*
   %v = load i32, i32* %__3, align 1
   %v_add = add i32 %v, 1
 
   %ptr2 = call i8* @llvm.nacl.read.tp()
   %__6 = ptrtoint i8* %ptr2 to i32
   %y = add i32 %__6, 4
   %__8 = inttoptr i32 %y to i32*
   %v_add2 = add i32 %v, 4
   store i32 %v_add2, i32* %__8, align 1
   ret i32 %v
 }
 ; CHECK-LABEL: test_nacl_read_tp_more_addressing
-; CHECK: mov e{{.*}},DWORD PTR gs:0x0
-; CHECK: mov e{{.*}},DWORD PTR gs:0x0
+; CHECK: mov e{{.*}},{{(DWORD PTR )?}}gs:0x0
+; CHECK: mov e{{.*}},{{(DWORD PTR )?}}gs:0x0
 ; CHECKO2REM-LABEL: test_nacl_read_tp_more_addressing
-; CHECKO2REM: mov e{{.*}},DWORD PTR gs:0x0
-; CHECKO2REM: mov e{{.*}},DWORD PTR gs:0x0
+; CHECKO2REM: mov e{{.*}},{{(DWORD PTR )?}}gs:0x0
+; CHECKO2REM: mov e{{.*}},{{(DWORD PTR )?}}gs:0x0
 ; CHECKO2UNSANDBOXEDREM-LABEL: test_nacl_read_tp_more_addressing
 ; CHECKO2UNSANDBOXEDREM: call {{.*}} R_{{.*}} __nacl_read_tp
 ; CHECKO2UNSANDBOXEDREM: call {{.*}} R_{{.*}} __nacl_read_tp
 
 define internal i32 @test_nacl_read_tp_dead(i32 %a) {
 entry:
   %ptr = call i8* @llvm.nacl.read.tp()
   ; Not actually using the result of nacl read tp call.
   ; In O2 mode this should be DCE'ed.
   ret i32 %a
@@ skipping 443 matching lines @@
 ; CHECK: lea {{.*}},[esp+0x10]
 ; CHECK: lea {{.*}},[esp+0x10]
 ; CHECK: call
 ; CHECK: mov esp,{{.*}}
 ; CHECK: mov esp,ebp
 ; ARM32-LABEL: test_stacksave_multiple
 ; ARM32: mov {{.*}}, sp
 ; ARM32: mov {{.*}}, sp
 ; ARM32: mov {{.*}}, sp
 ; ARM32: mov sp, {{.*}}