Commit Graph

1336 Commits

Author SHA1 Message Date
David Neto
5f69f75126 Support SPV_GOOGLE_decorate_string and SPV_GOOGLE_hlsl_functionality1
This commit add assembling, disassembling, and basic validation for two
Google extensions to better support HLSL translation.
2018-03-05 13:34:13 -05:00
Steven Perron
9ba50e34f2 Avoid generating duplicate names when merging types
The merging types we do not remove other information related to the
types.  We simply leave it duplicated, and hope it is removed later.
This is what happens with decorations.  They are removed in the next
phase of remove duplicates.  However, for OpNames that is not the case.
We end up with two different names for the same id, which does not make
sense.

The solution is to remove the names and decorations for the type being
removed instead of rewriting them to refer to the other type.

Note that it is possible that if the first type does not have a name,
then the types will end up with no name.  That is fine because the names
should not have any semantic significance anyway.

The was identified in issue #1372, but this does not fix that issue.
2018-03-05 12:02:50 -05:00
Pierre Moreau
6cd6e5ebef Define Disassemble only when Effcee is used in fold_test 2018-03-02 16:40:52 -05:00
David Neto
a942ff40f5 Android.mk: Generate enum mappings from unified1 grammar
Some tokens are only showing up in the unified1 grammar.
So enum string mappings have to be generated from that grammar, not
the grammar from the (deprecated) include/spirv/1.2 grammar.

Example: capabilities FragmentFullyCovered, Float16ImageAMD
2018-03-02 15:27:22 -05:00
David Neto
fe21921629 Start v2018.2 development 2018-03-02 14:30:49 -05:00
David Neto
6432a129f8 Finalize v2018.1 2018-03-02 14:30:49 -05:00
Alan Baker
52bceb3569 Handles more cases of redundant selects
* Handles OpConstantNull and vector types
 * vector selects (except against a null) are converted to vector
 shuffles
* Added tests
2018-03-02 14:28:08 -05:00
David Neto
a7cec7843c Update CHANGES 2018-03-02 14:01:28 -05:00
Alan Baker
824625760b Fixes #1361. Mark all non-constant global values as varying in CCP
* Also mark function parameters as varying
* Conservatively mark assignment instructions as varying if any input is
varying after attempting to fold
* Added a test to catch this case
2018-03-01 15:24:41 -05:00
Arseny Kapoulkine
8b27ba834d Vulkan BuiltIn variables can't have Location/Component decorations
As per Vulkan spec, BuiltIn variables can't have Location or Component
decorations. On some drivers, these can lead to driver crashing when
compiling the shader pipeline; for example, NVidia/AMD desktop drivers:
https://github.com/KhronosGroup/glslang/issues/1182.

This change adds validation and tests to catch this.
2018-03-01 15:00:08 -05:00
Alan Baker
ce5941a642 Fixes #1357. Support null constants better in folding
* getFloatConstantKind() now handles OpConstantNull
* PerformOperation() now handles OpConstantNull for vectors
* Fixed some instances where we would attempt to merge a division by 0
* added tests
2018-02-28 23:12:27 -05:00
GregF
bdaf8d56fb Opt: Add constant folding for FToI and IToF 2018-02-28 23:08:52 -05:00
Alan Baker
9457cabbce Fixes #1354. Do not merge integer division.
* Removes merging of div with a div or mul for integers
* Updated tests
2018-02-28 13:33:21 -05:00
Steven Perron
588f4fcc95 Add more folding rules for vector shuffle.
Adds rule to fold OpVectorShuffle with constant inputs.

Adds rules to fold OpCompositeExtrac being fed by an OpVectorShuffle.
2018-02-27 21:20:22 -05:00
Victor Lomuller
90e1637ce4 Remove Function::GetBlocks pushed by accident 2018-02-27 21:07:10 -05:00
Steven Perron
2cb589cc14 Remove uses DCEInst and call ADCE
The algorithm used in DCEInst to remove dead code is very slow.  It is
fine if you only want to remove a small number of instructions, but, if
you need to remove a large number of instructions, then the algorithm in
ADCE is much faster.

This PR removes the calls to DCEInst in the load-store removal passes
and adds a pass of ADCE afterwards.

A number of different iterations of the order of optimization, and I
believe this is the best I could find.

The results I have on 3 sets of shaders are:

Legalization:

Set 1: 5.39 -> 5.01
Set 2: 13.98 -> 8.38
Set 3: 98.00 -> 96.26

Performance passes:

Set 1: 6.90 -> 5.23
Set 2: 10.11 -> 6.62
Set 3: 253.69 -> 253.74

Size reduction passes:

Set 1: 7.16 -> 7.25
Set 2: 17.17 -> 16.81
Set 3: 112.06 -> 107.71

Note that the third set's compile time is large because of the large
number of basic blocks, not so much because of the number of
instructions.  That is why we don't see much gain there.
2018-02-27 21:06:08 -05:00
David Neto
0c13467161 Consistently include latest spirv.h header file.
Use indirection through latest_version_spirv.h

Also, when generating enum tables, use the unified1 JSON grammar since
it now has FragmentFullyCoveredEXT but the other JSON grammars don't.
They are starting to fall behind.
2018-02-27 18:47:29 -05:00
Alan Baker
802cf053c7 Merge arithmetic with non-trivial constant operands
Adding basis of arithmetic merging

* Refactored constant collection in ConstantManager
* New rules:
 * consecutive negates
 * negate of arithmetic op with a constant
 * consecutive muls
 * reciprocal of div

* Removed IRContext::CanFoldFloatingPoint
 * replaced by Instruction::IsFloatingPointFoldingAllowed
* Fixed some bad tests
* added some header comments

Added PerformIntegerOperation

* minor fixes to constants and tests
* fixed IntMultiplyBy1 to work with 64 bit ints
* added tests for integer mul merging

Adding test for vector integer multiply merging

Adding support for merging integer add and sub through negate

* Added tests

Adding rules to merge mult with preceding divide

* Has a couple tests, but needs more
* Added more comments

Fixed bug in integer division folding

* Will no longer merge through integer division if there would be a
remainder in the division
* Added a bunch more tests

Adding rules to merge divide and multiply through divide

* Improved comments
* Added tests

Adding rules to handle mul or div of a negation

* Added tests

Changes for review

* Early exit if no constants are involved in more functions
* fixed some comments
* removed unused declaration
* clarified some logic

Adding new rules for add and subtract

* Fold adds of adds, subtracts or negates
* Fold subtracts of adds, subtracts or negates
* Added tests
2018-02-27 13:02:13 -05:00
Stephen McGroarty
20b8cdb7c6 Make IR builder use the type manager for constants
This change makes the IR builder use the type manager to generate
OpTypeInts when creating OpConstants. This avoids dangling references
being stored by the created OpConstants.
2018-02-27 12:59:26 -05:00
Pierre Moreau
9394272c98 linker: merge debug annotations from category c)
Fixes: https://github.com/KhronosGroup/SPIRV-Tools/issues/1218
2018-02-27 12:31:50 -05:00
Pierre Moreau
bdd6617faa linker: Allow modules to be partially linked
Fixes: https://github.com/KhronosGroup/SPIRV-Tools/issues/1144
2018-02-27 12:21:13 -05:00
Pierre Moreau
64298bfd15 tools/linker: Allow setting --verify-ids on the command line 2018-02-27 12:06:24 -05:00
David Neto
b08b94ec0a Update CHANGES 2018-02-27 11:48:40 -05:00
Victor Lomuller
3497a94460 Add loop unswitch pass.
It moves all conditional branching and switch whose conditions are loop
invariant and uniform. Before performing the loop unswitch we check that
the loop does not contain any instruction that would prevent it
(barriers, group instructions etc.).
2018-02-27 08:52:46 -05:00
Stephen McGroarty
e354984b09 Unroller support for multiple induction variables
Support for multiple induction variables within a loop and support for
loop condition operands <= and >=.
2018-02-27 11:50:08 +00:00
Steven Perron
94af58a350 Clean up variables before sroa
In some shaders there are a lot of very large and deeply nested
structures.  This creates a lot of work for scalar replacement.  Also,
since commit ca4457b we have been very aggressive as rewriting
variables.  This has causes a large increase in compile time in creating
and then deleting the instructions.

To help low the costs, I want to run a cleanup of some of the easy loads
and stores to remove.  This reduces the number of symbols sroa has to
work on.  It also reduces the amount of code the simplifier has to
simplify because it was not generated by sroa.

To confirm the improvement, I ran numbers on three different sets of
shaders:

Time to run --legalize-hlsl:

Set #1: 55.89s -> 12.0s
Set #2: 1m44s -> 1m40.5s
Set #3: 6.8s -> 5.7s

Time to run -O

Set #1: 18.8s -> 10.9s
Set #2: 5m44s -> 4m17s
Set #3: 7.8s -> 7.8s

Contributes to #1328.
2018-02-22 21:40:58 -05:00
Steven Perron
3f19c2031a Preserve analysies in the simplification pass
Fixes a bug at the same time.  In `UpdateDefUse`, if the definition
already exists, we are not suppose to analyse it again.  When you do
the entries for the definition are deleted, and we don't want that.
The check for this was wrong.
2018-02-22 16:06:30 -05:00
Lei Zhang
432dc40412 Appveyor: remove VS2015 configuration to reduce build time
We already have VS2013 and VS2017, which should be good guards.
2018-02-22 15:17:04 -05:00
GregF
46a9ec9d23 Opt: Check for side-effects in DCEInst()
This function now checks for side-effects before adding operand
instructions to the dead instruction work list.

Because this fix puts more pressure on IsCombinatorInstruction() to
be correct, this commit adds all OpConstant* and OpType* instructions
to combinator_ops_ set.

Fixes #1341.
2018-02-22 12:24:13 -05:00
Alan Baker
01760d2f0f Fixes #1338. Handle OpConstantNull in branch/switch conditions
* No longer assume the branch/switch condition must be bool or int
constants (respectively)
* Added a couple unit tests for each case
2018-02-21 10:22:39 -05:00
Steven Perron
51ecc7318f Reduce instruction create and deletion during inlining.
When inlining a function call the instructions in the same basic block
as the call get cloned.  The clone is added to the set of new blocks
containing the inlined code, and the original instructions are deleted.

This PR will change this so that we simply move the instructions to the
new blocks.  This saves on the creation and deletion of the
instructions.

Contributes to #1328.
2018-02-21 09:50:47 -05:00
Steven Perron
c1b936637e Add Insert-extract elimination back into legalization passes.
Fixes #1326.
2018-02-21 09:46:51 -05:00
Arseny Kapoulkine
309be423cc Add folding for redundant add/sub/mul/div/mix operations
This change implements instruction folding for arithmetic operations
that are redundant, specifically:

  x + 0 = 0 + x = x
  x - 0 = x
  0 - x = -x
  x * 0 = 0 * x = 0
  x * 1 = 1 * x = x
  0 / x = 0
  x / 1 = x
  mix(a, b, 0) = a
  mix(a, b, 1) = b

Cache ExtInst import id in feature manager

This allows us to avoid string lookups during optimization; for now we
just cache GLSL std450 import id but I can imagine caching more sets as
they become utilized by the optimizer.

Add tests for add/sub/mul/div/mix folding

The tests cover scalar float/double cases, and some vector cases.

Since most of the code for floating point folding is shared, the tests
for vector folding are not as exhaustive as scalar.

To test sub->negate folding I had to implement a custom fixture.
2018-02-20 18:29:27 -05:00
Steven Perron
fa3ac3cc33 Revert "Preserve analysies in the simplification pass"
This reverts commit ec3bbf093e.
2018-02-20 18:21:25 -05:00
Steven Perron
ec3bbf093e Preserve analysies in the simplification pass
Building the def-use chains is very expensive, so we do not want to
invalidate them it if is not necessary.  At the moment, it seems like
most optimizatoins are good at not invalidating the def-use chains, but
simplification does.

This PR get the simlification pass to keep the analysies valid.

Contributes to #1328.
2018-02-20 14:45:08 -05:00
Diego Novillo
6c75050136 Speed up Phi insertion.
On some shader code we have in our testsuite, Phi insertion is showing
massive compile time slowdowns, particularly during destruction.  The
specific shader I was looking at has about 600 variables to keep track
of and around 3200 basic blocks.  The algorithm is currently O(var x
blocks), which means maps with around 2M entries.  This was taking about
8 minutes of compile time.

This patch changes the tracking of stored variables to be more sparse.
Instead of having every basic block contain all the tracked variables in
the map, they now have only the variables actually stored in that block.

This speeds up deallocation, which brings down compile time to about
1m20s.

Note that this is not the definite fix for this.  I will re-write Phi
insertion to use a standard SSA rewriting algorithm
(https://github.com/KhronosGroup/SPIRV-Tools/issues/893).

This contributes to
https://github.com/KhronosGroup/SPIRV-Tools/issues/1328.
2018-02-20 12:04:06 -05:00
Lei Zhang
ac9ab0dacf Travis: require MacOS to build and test again 2018-02-20 11:25:38 -05:00
Steven Perron
9d95a91a9f Fix folding insert feeding extract
I mixed up two cases when folding an OpCompositeExtract that is feed by
and OpCompositeInsert.  The specific cases are demonstracted in the new
test.  I mixed up the conditions for the cases, and treated one like the
other.

Fixes #1323.
2018-02-20 11:22:51 -05:00
Alan Baker
c3f34d8bf3 Fixes #1300. Adding checks for bad CCP transitions and unsettled values
* Now track propagation status and assert on bad statuses
 * Added helper methods to access instruction propagation status
* Modified the phi meet operator to properly reflect the paper it is
based on
* Modified SSA edge addition so that all edge are added, but only on
state changes
* Fixed a bug in instruction simulation where interesting conditional
branches would not mark the interesting edge as executed
 * Added a test to catch this bug
* Added an ostream operator for SSAPropagator::PropStatus
2018-02-18 19:41:34 -05:00
Andrew Woloszyn
e543b195df Removed warnings from hex_float.h
Bitcasting FloatProxy<->uint_type was hitting a warning
with g++8.0.1. Replace bitcasts with new casting traits for FloatProxy.
2018-02-16 21:15:51 -05:00
Steven Perron
04cd63e5b9 Make better use of simplification pass
The simplification pass works better after all of the dead branches are
removed.  So swapping them around in the legalization passes.  Also
adding the simplification pass to performance passes right after dead
branch elimination.

Added CCP to the legalization passes so we can propagate the constants
into the branchs, and remove as many branches a possible.  CCP is
designed to still get opportunities even if the branches are dead, so it
is a good place for it.

Fixes #1118
2018-02-16 20:46:49 -05:00
Arseny Kapoulkine
1054413600 Add constant folding rules for floating-point comparison
This change handles all 6 regular comparison types in two variations,
ordered (true if values are ordered *and* comparison is true) and
unordered (true if values are unordered *or* comparison is true).

Ordered comparison matches the default floating-point behavior on host
but we use std::isnan to check ordering explicitly anyway.

This change also slightly reworks the floating-point folding support
code to make it possible to define a folding operation that returns
boolean instead of floating point.

These tests exhaustively test ordered/unordered comparisons for
float/double.

Since for NaN inputs the comparison result doesn't depend on the
comparison function, we just test == and !=; NaN inputs result in true
unordered comparisons and false ordered comparisons.
2018-02-16 20:41:22 -05:00
Arseny Kapoulkine
27d23a92a0 Remove constants from constant manager in KillInst
Registering a constant in constant manager establishes a relation
between instruction that defined it and constant object. On complex
shaders this could result in the constant definition getting removed as
part of one of the DCE pass, and a subsequent simplification pass trying
to use the defining instruction for the constant.

To fix this, we now remove associated constant entries from constant
manager when killing constant instructions; the constant object is still
registered and can be remapped to a new instruction later.

GetDefiningInstruction shouldn't ever return nullptr after this change
so add an assertion to check for that.
2018-02-16 20:37:12 -05:00
Steven Perron
50f307f889 Simplify OpPhi instructions referencing unreachable continues
In dead branch elimination, we already recognize unreachable continue
blocks, and update OpPhi instruction accordingly.  This change adds an
extra check: if the head block has exactly 1 other incoming edge, then
replace the OpPhi with the value from that edge.

Fixes #1314.
2018-02-16 18:58:03 -05:00
Steven Perron
3756b387f3 Get CCP to use the constant floating point rules.
Fixes #1311
2018-02-16 13:49:47 -05:00
Lei Zhang
efe286cd32 SubgroupBallotKHR can enable SubgroupSize & SubgroupLocalInvocationId 2018-02-16 10:02:18 -05:00
Nerijus Baliūnas
5d442fad2f fix typo 2018-02-15 12:00:02 -05:00
David Neto
d2c0fce361 Invoke cmake via CMAKE_COMMAND variable
Need to do this in case cmake is not on the path.
This should fix the Android NDK build, as in when building the NDK
itself.
2018-02-15 11:34:50 -05:00
Lei Zhang
f3a10470d3
Avoid using static unordered_map (#1304)
unordered_map is not POD. Using it as static may cause problems
when operator new() and operator delete() is customized.

Also changed some function signatures to use const char* instead
of std::string, which will give caller the flexibility to avoid
creating a std::string.
2018-02-15 10:19:15 -05:00
Arseny Kapoulkine
32a8e04c7d Add folding of redundant OpSelect insns
We can fold OpSelect into one of the operands in two cases:

- condition is constant
- both results are the same

Even if the original shader doesn't have either of these, if-conversion
pass sometimes ends up generating instructions like

   %7127 = OpSelect %int %3220 %7058 %7058

And this optimization cleans them up.
2018-02-15 10:03:22 -05:00