Subzero: Fold the load instruction into the next cast instruction.

tests_lit/llvm2ice_tests/8bit.pnacl.ll

 ; This tries to be a comprehensive test of i8 operations.
 
 ; RUN: %p2i --filetype=obj --disassemble -i %s --args -O2 | FileCheck %s
 ; RUN: %p2i --filetype=obj --disassemble -i %s --args -Om1 | FileCheck %s
 
 declare void @useInt(i32 %x)
 
 define internal i32 @add8Bit(i32 %a, i32 %b) {
 entry:
   %a_8 = trunc i32 %a to i8
@@ skipping 317 matching lines @@
 ; CHECK-DAG: mov {{.*}},{{[a-d]}}l
 ; CHECK-DAG: mov {{.*}},BYTE PTR
 ; CHECK-DAG: mov BYTE PTR {{.*}}
 
 @global8 = internal global [1 x i8] c"\01", align 4
 
 define i32 @load_i8(i32 %addr_arg) {
 entry:
   %addr = inttoptr i32 %addr_arg to i8*
   %ret = load i8* %addr, align 1
-  %ret_ext = zext i8 %ret to i32
+  %ret2 = sub i8 %ret, 0
+  %ret_ext = zext i8 %ret2 to i32
   ret i32 %ret_ext
 }
 ; CHECK-LABEL: load_i8
 ; CHECK: mov {{[a-d]l}},BYTE PTR
 
 define i32 @load_i8_global(i32 %addr_arg) {
 entry:
   %addr = bitcast [1 x i8]* @global8 to i8*
   %ret = load i8* %addr, align 1
-  %ret_ext = zext i8 %ret to i32
+  %ret2 = sub i8 %ret, 0
+  %ret_ext = zext i8 %ret2 to i32
   ret i32 %ret_ext
 }
 ; CHECK-LABEL: load_i8_global
 ; CHECK: mov {{[a-d]l}},BYTE PTR
 
 define void @store_i8(i32 %addr_arg, i32 %val) {
 entry:
   %val_trunc = trunc i32 %val to i8
   %addr = inttoptr i32 %addr_arg to i8*
   store i8 %val_trunc, i8* %addr, align 1
   ret void
 }
 ; CHECK-LABEL: store_i8
 ; CHECK: mov BYTE PTR {{.*}},{{[a-d]l}}
 
 define void @store_i8_const(i32 %addr_arg) {
 entry:
   %addr = inttoptr i32 %addr_arg to i8*
   store i8 123, i8* %addr, align 1
   ret void
 }
 ; CHECK-LABEL: store_i8_const
 ; CHECK: mov BYTE PTR {{.*}},0x7b