Convert lit tests to check disassembled assembly.

Convert lit tests to check disassembled assembly.

Then when we have an integrated assembler, we can check
its disassembly and the result should be the same.
This only touches the tests that invoke llvm-mc currently.
There are other tests which check for .s file output.

There are quite a bit of quirks with llvm-objdump,
which is unfortunate:

(*) The symbolizer doesn't pick up non-section-local
function calls. Some externals were converted to be
local functions. Workaround: where it counts, I just
left a check via .s files and a new --check-prefix.
It's a little better in 3.6.

(*) The symbolizer doesn't pick up global variable names.
I just checked for the relocation addend instead.
Didn't check if it was better in 3.6, but maybe.

(*) We have a bug in bundling lock + instructions.
See
BUG=https://code.google.com/p/nativeclient/issues/detail?id=3929

(*) There's no disassembly for branch lables.
Checks of jump instructions were converted to check
for positive or negative values, depending on whether
it is a forward or backward branch.

BUG=none
R=stichnot@chromium.org

Committed: https://gerrit.chromium.org/gerrit/gitweb?p=native_client/pnacl-subzero.git;a=commit;h=dddc306

[jvoung (off chromium), 2014-08-28 02:21:29]

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

[jvoung (off chromium), 2014-08-28 02:21:36]

https://codereview.chromium.org/509233002/diff/1/tests_lit/llvm2ice_tests/nac...
File tests_lit/llvm2ice_tests/nacl-atomic-fence-all.ll (right):

https://codereview.chromium.org/509233002/diff/1/tests_lit/llvm2ice_tests/nac...
tests_lit/llvm2ice_tests/nacl-atomic-fence-all.ll:190: ; CHECK: lea {{.*}},
g32_d is in bss instead, so this would have been [0]...
I just dropped the check but we can add it back.

[jvoung (off chromium), 2014-08-28 17:31:07]

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
File tests_lit/llvm2ice_tests/64bit.pnacl.ll (right):

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
tests_lit/llvm2ice_tests/64bit.pnacl.ll:7: ; It was only fixed in 3.6.
Okay, I think the reason such a sad bug existed for so long is that usually
Linux, etc. would align function starts.  If you do align the function starts,
then the ret will disassemble correctly, as well as the nop padding that follows
it.

So, I think I'll send out a separate CL to align function starts, since NaCl
also aligns function starts. Then I'll remove these hacks.

[Jim Stichnoth, 2014-08-28 20:19:23]

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
File tests_lit/llvm2ice_tests/64bit.pnacl.ll (left):

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
tests_lit/llvm2ice_tests/64bit.pnacl.ll:43: ; CHECK       sub     esp
Even though this line was wrong ("CHECK" instead of "CHECK:"), did it miss being
converted to the new format?

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
File tests_lit/llvm2ice_tests/64bit.pnacl.ll (right):

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
tests_lit/llvm2ice_tests/64bit.pnacl.ll:53: ; CALLTARGETS-LABEL: pass64BitArg
Would it be clearer to move this -LABEL up with the CHECK-LABEL?

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
tests_lit/llvm2ice_tests/64bit.pnacl.ll:106: ; Bundle padding added by
-triple=i686-none-nacl.
It would be nice if there were a way to match "zero or more" consecutive nop
instructions, but I don't know how...

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
File tests_lit/llvm2ice_tests/convert.ll (right):

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
tests_lit/llvm2ice_tests/convert.ll:41: ; CHECK-LABEL: from_int8:
Change "from_int8:" to "from_int8"?  (here and below)

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
tests_lit/llvm2ice_tests/convert.ll:49: ; CHECK: lea {{.*}} [8]
Can you keep the "lea" out of the compare string?  These should change to "mov"
instructions once we don't need the LEAHACK workaround.

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
File tests_lit/llvm2ice_tests/fp.pnacl.ll (left):

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
tests_lit/llvm2ice_tests/fp.pnacl.ll:28: ; CHECK-NEXT: ret
Why was this line removed?

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
File tests_lit/llvm2ice_tests/nacl-atomic-fence-all.ll (right):

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
tests_lit/llvm2ice_tests/nacl-atomic-fence-all.ll:190: ; CHECK: lea {{.*}},
Should g32_d turn into "dword ptr [12]"?

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
File tests_lit/llvm2ice_tests/shift.ll (right):

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
tests_lit/llvm2ice_tests/shift.ll:29: ; CHECK-LABEL: conv1:
Remove last ':', here and below.

[jvoung (off chromium), 2014-08-29 00:51:20]

Thanks!

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
File tests_lit/llvm2ice_tests/64bit.pnacl.ll (left):

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
tests_lit/llvm2ice_tests/64bit.pnacl.ll:43: ; CHECK       sub     esp
On 2014/08/28 20:19:23, stichnot wrote:
> Even though this line was wrong ("CHECK" instead of "CHECK:"), did it miss
being
> converted to the new format?

Oops, must have accidentally overwritten it with the new CALLTARGETS line.
Fixed.

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
File tests_lit/llvm2ice_tests/64bit.pnacl.ll (right):

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
tests_lit/llvm2ice_tests/64bit.pnacl.ll:7: ; It was only fixed in 3.6.
On 2014/08/28 17:31:07, jvoung wrote:
> Okay, I think the reason such a sad bug existed for so long is that usually
> Linux, etc. would align function starts.  If you do align the function starts,
> then the ret will disassemble correctly, as well as the nop padding that
follows
> it.
> 
> So, I think I'll send out a separate CL to align function starts, since NaCl
> also aligns function starts. Then I'll remove these hacks.

Done.

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
tests_lit/llvm2ice_tests/64bit.pnacl.ll:53: ; CALLTARGETS-LABEL: pass64BitArg
On 2014/08/28 20:19:23, stichnot wrote:
> Would it be clearer to move this -LABEL up with the CHECK-LABEL?

Done.

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
tests_lit/llvm2ice_tests/64bit.pnacl.ll:106: ; Bundle padding added by
-triple=i686-none-nacl.
On 2014/08/28 20:19:23, stichnot wrote:
> It would be nice if there were a way to match "zero or more" consecutive nop
> instructions, but I don't know how...

Acknowledged. Not sure how either =(

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
File tests_lit/llvm2ice_tests/convert.ll (right):

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
tests_lit/llvm2ice_tests/convert.ll:41: ; CHECK-LABEL: from_int8:
On 2014/08/28 20:19:23, stichnot wrote:
> Change "from_int8:" to "from_int8"?  (here and below)

Done.

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
tests_lit/llvm2ice_tests/convert.ll:49: ; CHECK: lea {{.*}} [8]
On 2014/08/28 20:19:23, stichnot wrote:
> Can you keep the "lea" out of the compare string?  These should change to
"mov"
> instructions once we don't need the LEAHACK workaround.

Done.

[I] pray I don't alter these any further, in later CLs.

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
File tests_lit/llvm2ice_tests/fp.pnacl.ll (left):

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
tests_lit/llvm2ice_tests/fp.pnacl.ll:28: ; CHECK-NEXT: ret
On 2014/08/28 20:19:23, stichnot wrote:
> Why was this line removed?

I think to avoid adding -ffunction-section args, but now it works so I've added
them back.

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
File tests_lit/llvm2ice_tests/nacl-atomic-fence-all.ll (right):

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
tests_lit/llvm2ice_tests/nacl-atomic-fence-all.ll:190: ; CHECK: lea {{.*}},
On 2014/08/28 20:19:23, stichnot wrote:
> Should g32_d turn into "dword ptr [12]"?

This is actually going to be "dword ptr [0]", because it's in .data instead of
.bss, I think.

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
File tests_lit/llvm2ice_tests/shift.ll (right):

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
tests_lit/llvm2ice_tests/shift.ll:29: ; CHECK-LABEL: conv1:
On 2014/08/28 20:19:23, stichnot wrote:
> Remove last ':', here and below.

Done.

[Jim Stichnoth, 2014-08-29 16:40:49]

Otherwise LGTM.

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
File tests_lit/llvm2ice_tests/64bit.pnacl.ll (right):

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
tests_lit/llvm2ice_tests/64bit.pnacl.ll:53: ; CALLTARGETS-LABEL: pass64BitArg
On 2014/08/29 00:51:20, jvoung wrote:
> On 2014/08/28 20:19:23, stichnot wrote:
> > Would it be clearer to move this -LABEL up with the CHECK-LABEL?
> 
> Done.

Thanks.  I should have said that this comment applies to pretty much all
CALLTARGETS-LABEL instances in all files. :)

[jvoung (off chromium), 2014-08-29 19:27:48]

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
File tests_lit/llvm2ice_tests/64bit.pnacl.ll (right):

https://codereview.chromium.org/509233002/diff/20001/tests_lit/llvm2ice_tests...
tests_lit/llvm2ice_tests/64bit.pnacl.ll:53: ; CALLTARGETS-LABEL: pass64BitArg
On 2014/08/29 16:40:49, stichnot wrote:
> On 2014/08/29 00:51:20, jvoung wrote:
> > On 2014/08/28 20:19:23, stichnot wrote:
> > > Would it be clearer to move this -LABEL up with the CHECK-LABEL?
> > 
> > Done.
> 
> Thanks.  I should have said that this comment applies to pretty much all
> CALLTARGETS-LABEL instances in all files. :)

Done.

[jvoung (off chromium), 2014-08-29 19:59:08]

Committed patchset #4 (id:60001) manually as dddc306 (tree was closed).