Subzero: Fix emission of 16-bit immediates.

Subzero: Fix emission of 16-bit immediates.

The operand type needs to be propagated into EmitImmediate() and EmitComplex() so that we know whether to emit the 2-byte or 4-byte form.

BUG= none
R=jvoung@chromium.org

Committed: https://gerrit.chromium.org/gerrit/gitweb?p=native_client/pnacl-subzero.git;a=commit;h=94c4c8e7390924485238bd99cd74ed2bb4b04a3b

[Jim Stichnoth, 2014-09-28 17:21:18]

stichnot@chromium.org changed reviewers:
+ jvoung@chromium.org

[Jim Stichnoth, 2014-09-28 18:14:29]

https://codereview.chromium.org/607353002/diff/20001/tests_lit/assembler/x86/...
File tests_lit/assembler/x86/immediate_encodings.ll (right):

https://codereview.chromium.org/607353002/diff/20001/tests_lit/assembler/x86/...
tests_lit/assembler/x86/immediate_encodings.ll:88: ; CHECK: 66 35 00 ff  xor ax,
65280
Oddly, this part works correctly in master.

[jvoung (off chromium), 2014-09-29 01:16:56]

LGTM

Oops thanks!

Looks like I can also repro a runtime error "Illegal instruction" if I add a
test case to tests_arith.cpp where one of the operands is a constant (watching
out for shift ones though...).

https://codereview.chromium.org/607353002/diff/20001/tests_lit/assembler/x86/...
File tests_lit/assembler/x86/immediate_encodings.ll (right):

https://codereview.chromium.org/607353002/diff/20001/tests_lit/assembler/x86/...
tests_lit/assembler/x86/immediate_encodings.ll:88: ; CHECK: 66 35 00 ff  xor ax,
65280
On 2014/09/28 18:14:28, stichnot wrote:
> Oddly, this part works correctly in master.

I think that llvm-objdump semi-silently ignored the error.

     124:       66 35 00 ff                                     xor     ax,
65280
     12a:       66 83 c0 01                                     add     ax, 1
     12e:       0f b7 c0                                        movzx   eax, ax

124 + 6 bytes is 12a, but it doesn't show the last two bytes! The llvm-objdump
does print the warning:

  
/...native_client/toolchain_build/out/llvm_x86_64_linux_work/Release+Asserts/bin/llvm-objdump:
warning: invalid instruction encoding

[Jim Stichnoth, 2014-09-29 03:30:25]

Committed patchset #2 (id:20001) manually as 94c4c8e (presubmit successful).