Subzero. X8664. Fixes filetype=asm.

Subzero. X8664. Fixes filetype=asm.

Fixes filetype=asm for x8664.

BUG= https://bugs.chromium.org/p/nativeclient/issues/detail?id=4077
R=stichnot@chromium.org

Committed: https://gerrit.chromium.org/gerrit/gitweb?p=native_client/pnacl-subzero.git;a=commit;h=008f4ce5417c97891ea02b54b691b39aa2e2a0af

[John, 2015-12-22 18:34:35]

jpp@chromium.org changed reviewers:
+ sehr@chromium.org, stichnot@chromium.org

[Jim Stichnoth, 2015-12-22 20:38:07]

If you force filetype=asm:

diff --git a/pydir/crosstest_generator.py b/pydir/crosstest_generator.py
index ef32a08..4ee1089 100755
--- a/pydir/crosstest_generator.py
+++ b/pydir/crosstest_generator.py
@@ -65,7 +65,7 @@ def main():
   for v in arch_attrs.values():
     flat_attrs += v
   arch_flags = { 'x8632': [],
-                 'x8664': [],
+                 'x8664': ['--filetype=asm'],
                  # ARM doesn't have an ELF writer yet, and iasm does not
                  # support sandboxing yet.
                  'arm32': ['--filetype=asm'] }


make check-xtest fails:

Failing Tests (5):
    subzero_crosstests :: Output/test_arith_x8664_native_O2_sse2.xtest
    subzero_crosstests :: Output/test_arith_x8664_native_Om1_sse2.xtest
    subzero_crosstests :: Output/test_cast_x8664_native_Om1_sse2.xtest
    subzero_crosstests :: Output/test_stacksave_x8664_native_O2_sse2.xtest
    subzero_crosstests :: Output/test_stacksave_x8664_native_Om1_sse2.xtest

  Expected Passes    : 117
  Unexpected Failures: 5

https://codereview.chromium.org/1543573002/diff/160001/src/IceTargetLoweringX...
File src/IceTargetLoweringX8664Traits.h (right):

https://codereview.chromium.org/1543573002/diff/160001/src/IceTargetLoweringX...
src/IceTargetLoweringX8664Traits.h:403: static int32_t getGprForType(Type Ty,
int32_t RegNum) {
Yuck. :(

No matter, though, the x86-32 version is also yucky and I meant to fix that all
anyway.

My thought was to just build a giant constexpr table, though that might be
tricky with our current x-macros setup.

https://codereview.chromium.org/1543573002/diff/160001/src/IceTargetLoweringX...
File src/IceTargetLoweringX86BaseImpl.h (right):

https://codereview.chromium.org/1543573002/diff/160001/src/IceTargetLoweringX...
src/IceTargetLoweringX86BaseImpl.h:4272: void
TargetX86Base<Machine>::lowerIndirectJump(Variable *Target) {
While you're at it, could you rename this argument?  "Target" is usually of type
TargetLowering.

[John, 2015-12-23 18:23:25]

Patchset #10 (id:180001) has been deleted

[John, 2015-12-23 18:30:44]

https://codereview.chromium.org/1543573002/diff/160001/src/IceTargetLoweringX...
File src/IceTargetLoweringX8664Traits.h (right):

https://codereview.chromium.org/1543573002/diff/160001/src/IceTargetLoweringX...
src/IceTargetLoweringX8664Traits.h:403: static int32_t getGprForType(Type Ty,
int32_t RegNum) {
On 2015/12/22 20:38:06, stichnot wrote:
> Yuck. :(
> 
> No matter, though, the x86-32 version is also yucky and I meant to fix that
all
> anyway.
> 
> My thought was to just build a giant constexpr table, though that might be
> tricky with our current x-macros setup.

One could build a [BaseReg] -> {Reg_i8, Reg_i16, Reg_i32, Reg_i64} table, and
then do a

return Table[getBaseReg(RegNum)][Ty];

this clean "implementation" would come at the expense of even moar indirection
than we already have (good luck debugging that table.) Therefore, I decided for
this implementation. We might even have a bigger executable with this big a**
switch, but in the end this method is just a bunch of 

return RegNum - FirstInClass(RegNum) + FirstOfType(Ty)

statements, where FirstRegInClass is a constexpr.

https://codereview.chromium.org/1543573002/diff/160001/src/IceTargetLoweringX...
File src/IceTargetLoweringX86BaseImpl.h (right):

https://codereview.chromium.org/1543573002/diff/160001/src/IceTargetLoweringX...
src/IceTargetLoweringX86BaseImpl.h:4272: void
TargetX86Base<Machine>::lowerIndirectJump(Variable *Target) {
On 2015/12/22 20:38:07, stichnot wrote:
> While you're at it, could you rename this argument?  "Target" is usually of
type
> TargetLowering.

Done.

[Jim Stichnoth, 2015-12-23 21:39:26]

lgtm

[John, 2015-12-24 21:22:21]

Description was changed from

==========
Subzero. X8664. Fixes filetype=asm.

Fixes filetype=asm for x8664.

BUG= https://bugs.chromium.org/p/nativeclient/issues/detail?id=4077
==========

to

==========
Subzero. X8664. Fixes filetype=asm.

Fixes filetype=asm for x8664.

BUG= https://bugs.chromium.org/p/nativeclient/issues/detail?id=4077
R=stichnot@chromium.org

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

[John, 2015-12-24 21:22:22]

Committed patchset #12 (id:240001) manually as
008f4ce5417c97891ea02b54b691b39aa2e2a0af (presubmit successful).