Chrome Version: 56.0.2924.87
OS: Chrome OS

---

I happened to be looking at some disassembly in Chrome as part of http://crosbug.com/p/62731

I was looking at this snippet of code:

      char* ubuf = new char[ulength];
      if (!port::Snappy_Uncompress(data, n, ubuf)) {

Here's what the compiler did:

   0xad622360 <+536>:   ldr     r0, [r7, #32]      // Put number of bytes to pass to new in R0 (first param)
   0xad622362 <+538>:   bl      0xad23868c         // the call to new
   0xad622366 <+542>:   ldr     r1, [r7, #16]      // put "data" in r1, but eventually goes to r0
   0xad622368 <+544>:   mov     r6, r0             // store ubuf in r6
   0xad62236a <+546>:   mov     r2, r6             // ubuf is the 3rd param to Snappy_Uncompress
   0xad62236c <+548>:   mov     r0, r1             // finally "data" goes into r0.
   0xad62236e <+550>:   ldr     r1, [r7, #8]       // so r7 + 8 is "n"
   0xad622370 <+552>:   bl      0xad60cc04 <_ZN7leveldb4port17Snappy_UncompressEPKcjPc>

It seems like there's an extra needless register shuffle going on here.  Specifically there was no real reason that the compiler needed to store "data" in r1 and then later shuffle it over to r0.