Subzero, MIPS32: Floating point load/store and format conversion instructions

Subzero, MIPS32: Floating point load/store and format conversion instructions

Introducing instructions for floating point format conversion.
These instructions are needed for full implementation of lowerCast
and missing floating point arithmetic instruction frem.

BUG=
R=stichnot@chromium.org

Committed: https://gerrit.chromium.org/gerrit/gitweb?p=native_client/pnacl-subzero.git;a=commit;h=418135a9dd403c33db612edd4db53b48b2a9af0d

[obucinac, 2016-05-31 17:28:43]

Description was changed from

==========
Floating point format conversion instructions

Introducing instructions for floating point format conversion.
These instructions are needed for full implementation of lowerCast
and missing floating point arithmetic instruction frem.

BUG=
==========

to

==========
Subzero, MIPS32: Floating point format conversion instructions

Introducing instructions for floating point format conversion.
These instructions are needed for full implementation of lowerCast
and missing floating point arithmetic instruction frem.

BUG=
==========

[obucinac, 2016-05-31 17:36:29]

obucinac@gmail.com changed reviewers:
+ mohit.bhakkad@imgtec.com, sagar.thakur@imgtec.com

[obucinac, 2016-05-31 17:37:03]

On 2016/05/31 17:36:29, obucinac wrote:
> mailto:obucinac@gmail.com changed reviewers:
> + mailto:mohit.bhakkad@imgtec.com, mailto:sagar.thakur@imgtec.com

This is WIP!

[obucinac, 2016-05-31 17:37:37]

This is WIP!

[obucinac, 2016-06-01 14:08:16]

obucinac@gmail.com changed reviewers:
+ jpp@chromium.org, kschimpf@google.com, stichnot@chromium.org

[obucinac, 2016-06-01 14:09:57]

Instructions for loading/storing values from/to memory, needed as dependency for
casting and lowerLoad/lowerStore.

[obucinac, 2016-06-01 14:11:00]

Description was changed from

==========
Subzero, MIPS32: Floating point format conversion instructions

Introducing instructions for floating point format conversion.
These instructions are needed for full implementation of lowerCast
and missing floating point arithmetic instruction frem.

BUG=
==========

to

==========
Subzero, MIPS32: Floating point load/store and format conversion instructions

Introducing instructions for floating point format conversion.
These instructions are needed for full implementation of lowerCast
and missing floating point arithmetic instruction frem.

BUG=
==========

[John, 2016-06-01 15:01:05]

https://codereview.chromium.org/2024183002/diff/40001/src/IceInstMIPS32.cpp
File src/IceInstMIPS32.cpp (right):

https://codereview.chromium.org/2024183002/diff/40001/src/IceInstMIPS32.cpp#n...
src/IceInstMIPS32.cpp:70: //template <> const char *InstMIPS32Bc1t::Opcode =
"bc1t"; // new base class needed
80-col

https://codereview.chromium.org/2024183002/diff/40001/src/IceInstMIPS32.cpp#n...
src/IceInstMIPS32.cpp:71: //template <> const char *InstMIPS32C_le_d::Opcode =
"c.le.d"; // new base class needed
80-col

https://codereview.chromium.org/2024183002/diff/40001/src/IceInstMIPS32.h
File src/IceInstMIPS32.h (right):

https://codereview.chromium.org/2024183002/diff/40001/src/IceInstMIPS32.h#new...
src/IceInstMIPS32.h:728: //using InstMIPS32Bc1t =
InstMIPS32ThreeAddrFPR<InstMIPS32::Bc1t>; // new base class needed
80-col

https://codereview.chromium.org/2024183002/diff/40001/src/IceInstMIPS32.h#new...
src/IceInstMIPS32.h:729: //using InstMIPS32C_le_d =
InstMIPS32ThreeAddrFPR<InstMIPS32::C_le_d>; // new base class needed
80-col

https://codereview.chromium.org/2024183002/diff/40001/src/IceTargetLoweringMI...
File src/IceTargetLoweringMIPS32.h (right):

https://codereview.chromium.org/2024183002/diff/40001/src/IceTargetLoweringMI...
src/IceTargetLoweringMIPS32.h:194: //void _bc1t(...) {
what do these comments mean?

[obucinac, 2016-06-01 16:38:10]

Unimplemented instructions are removed.

https://codereview.chromium.org/2024183002/diff/40001/src/IceInstMIPS32.cpp
File src/IceInstMIPS32.cpp (right):

https://codereview.chromium.org/2024183002/diff/40001/src/IceInstMIPS32.cpp#n...
src/IceInstMIPS32.cpp:70: //template <> const char *InstMIPS32Bc1t::Opcode =
"bc1t"; // new base class needed
On 2016/06/01 15:01:05, John wrote:
> 80-col

Removed for now.

https://codereview.chromium.org/2024183002/diff/40001/src/IceInstMIPS32.cpp#n...
src/IceInstMIPS32.cpp:71: //template <> const char *InstMIPS32C_le_d::Opcode =
"c.le.d"; // new base class needed
On 2016/06/01 15:01:05, John wrote:
> 80-col

Removed for now.

https://codereview.chromium.org/2024183002/diff/40001/src/IceInstMIPS32.h
File src/IceInstMIPS32.h (right):

https://codereview.chromium.org/2024183002/diff/40001/src/IceInstMIPS32.h#new...
src/IceInstMIPS32.h:728: //using InstMIPS32Bc1t =
InstMIPS32ThreeAddrFPR<InstMIPS32::Bc1t>; // new base class needed
On 2016/06/01 15:01:05, John wrote:
> 80-col

Removed for now.

https://codereview.chromium.org/2024183002/diff/40001/src/IceInstMIPS32.h#new...
src/IceInstMIPS32.h:729: //using InstMIPS32C_le_d =
InstMIPS32ThreeAddrFPR<InstMIPS32::C_le_d>; // new base class needed
On 2016/06/01 15:01:05, John wrote:
> 80-col

Removed for now.

https://codereview.chromium.org/2024183002/diff/40001/src/IceTargetLoweringMI...
File src/IceTargetLoweringMIPS32.h (right):

https://codereview.chromium.org/2024183002/diff/40001/src/IceTargetLoweringMI...
src/IceTargetLoweringMIPS32.h:194: //void _bc1t(...) {
On 2016/06/01 15:01:05, John wrote:
> what do these comments mean?

Removed for now.

[Jim Stichnoth, 2016-06-01 21:39:55]

On 2016/06/01 14:09:57, obucinac wrote:
> Instructions for loading/storing values from/to memory, needed as dependency
for
> casting and lowerLoad/lowerStore.

Is this still a work in progress?  I.e., is there more coming?

Unless this is a prerequisite for some other MIPS lowering CL, I'd be happier if
IceTargetLoweringMIPS32.cpp had actual lowering code that used these new
instructions.

[obucinac, 2016-06-02 11:35:23]

On 2016/06/01 21:39:55, stichnot wrote:
> On 2016/06/01 14:09:57, obucinac wrote:
> > Instructions for loading/storing values from/to memory, needed as dependency
> for
> > casting and lowerLoad/lowerStore.
> 
> Is this still a work in progress?  I.e., is there more coming?
> 
> Unless this is a prerequisite for some other MIPS lowering CL, I'd be happier
if
> IceTargetLoweringMIPS32.cpp had actual lowering code that used these new
> instructions.

This is something we need to share since we all need it to continue our work.

cvt* and trunc* instructions are dependencies for lowerCast, at least for the
cases which are not supported by target helpers, if there are any. In the end we
may shave some of them.

Ldc1, Lwc1, Sdc1, Swc1 are depandency for lowerStore/lowerLoad for floats and
double, and lowerCast for manipulating data.

Lw is dependency for lowerLoad for integers.

These tasks are divided between us, e.g. my tasks are lowerStore/lowerLoad,
while lowerCast belongs to others. We agree to add this as a separate patch, to
have things in place and same code base for these tasks.

[Jim Stichnoth, 2016-06-02 13:46:08]

lgtm

[Jim Stichnoth, 2016-06-02 13:47:10]

Description was changed from

==========
Subzero, MIPS32: Floating point load/store and format conversion instructions

Introducing instructions for floating point format conversion.
These instructions are needed for full implementation of lowerCast
and missing floating point arithmetic instruction frem.

BUG=
==========

to

==========
Subzero, MIPS32: Floating point load/store and format conversion instructions

Introducing instructions for floating point format conversion.
These instructions are needed for full implementation of lowerCast
and missing floating point arithmetic instruction frem.

BUG=
R=stichnot@chromium.org

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

[Jim Stichnoth, 2016-06-02 13:47:12]

Committed patchset #4 (id:60001) manually as
418135a9dd403c33db612edd4db53b48b2a9af0d (presubmit successful).