Last change for mips16 prolog/epilog cleanup and optimization.
Some tiny cosmetic code changes to follow. Because of the wide
ranging nature of the patch a full 24 test cycle was needed to
check against regression. This was the smallest patch I could
make to progress from the earlier ones in the series. 

llvm-svn: 197350
diff --git a/llvm/test/CodeGen/Mips/helloworld.ll b/llvm/test/CodeGen/Mips/helloworld.ll
index d363006..709c12e 100644
--- a/llvm/test/CodeGen/Mips/helloworld.ll
+++ b/llvm/test/CodeGen/Mips/helloworld.ll
@@ -25,10 +25,9 @@
 ; SR32:  .set noreorder
 ; SR32:  .set nomacro
 ; SR32:  .set noat
-; SR:	save 	$ra, $16, $17, [[FS:[0-9]+]]
+; SR:	save 	$ra, 24 # 16 bit inst
 ; PE:    .ent main
-; PE:    .align  2
-; PE-NEXT:	li	$[[T1:[0-9]+]], %hi(_gp_disp)
+; PE:	li	$[[T1:[0-9]+]], %hi(_gp_disp)
 ; PE-NEXT: 	addiu	$[[T2:[0-9]+]], $pc, %lo(_gp_disp)
 ; PE:	        sll	$[[T3:[0-9]+]], $[[T1]], 16
 ; C1:	lw	${{[0-9]+}}, %got($.str)(${{[0-9]+}})
@@ -37,7 +36,7 @@
 ; C2:	move	$25, ${{[0-9]+}}
 ; C1:	move 	$gp, ${{[0-9]+}}
 ; C1:	jalrc 	${{[0-9]+}}
-; SR:	restore 	$ra, $16, $17, [[FS]]
+; SR:	restore $ra,	24 # 16 bit inst
 ; PE:	li	$2, 0
 ; PE:	jrc 	$ra