Commit Graph

85 Commits

Author SHA1 Message Date
qining
27e04a004d Fix spec const construtor for matrix and vector
Fix issue: #237

1. The code generated for matrix constructor should 1) build column
vectors first, 2) build matrix with the vectors.

2. When there is only one scalar type constituent in vector's
constructor, we should populate the constituent to fill all the slots in
the vector. As for matrix, the single constituent should be populated to
the diagonal positions (top-left to bottom-right diagonal).

remove createSpvConstantFromConstSubTree()
2016-04-14 17:52:57 -04:00
qining
189b2033a4 Refine the code and address comments 2016-04-12 23:19:28 -04:00
qining
e24aa5edbb SpecOp bool->uint/int and uint<->int conversion
Bool -> uint/int with OpSpecConstantOp OpSelect instruction.

uint <-> int conversion with OpSpecConstantOp OpIAdd instruction.

Note, implicit conversion: `const uint = an_int_spec_constant` is not
supported. Explicit type casting is required: `const uint =
uint(an_int_spec_constant)`
2016-04-12 22:01:55 -04:00
John Kessenich
1c7e70763b Merge branch 'master' into hlsl-frontend 2016-04-03 20:36:48 -06:00
John Kessenich
79845ad8e7 Merge pull request #217 from baldurk/vs2010-compile-fixes
VS2010 compile fixes
2016-04-03 14:06:22 -06:00
Eric Engestrom
6a6d6dda95 fix spelling mistakes 2016-04-03 01:17:13 +01:00
baldurk
bd9f8351f4 Specify explicit return type on lambda function 2016-04-02 13:38:42 +02:00
baldurk
0dfbe3f90d Change {parameter} lists into explicit std::vector temporaries 2016-04-02 13:38:28 +02:00
qining
5c61d8e0f9 fix format; remove unnecessary parameters; rename some spec op mode guard class; remove support of float point comparison and composite type comparison; update the tests. 2016-03-31 13:57:28 -04:00
qining
135452061a Spec Constant Operations
Approach:
Add a flag in `Builder` to indicate 'spec constant mode' and 'normal
mode'. When the builder is in 'normal mode', nothing changed. When the
builder is in 'spec constant mode', binary, unary and other instruction
creation rountines will be redirected to `createSpecConstantOp()` to
create instrution at module level with `OpSpecConstantOp <original
opcode> <operands>`.

'spec constant mode' should be enabled if and only if we are creating
spec constants. So a flager setter/recover guard is added when handling
binary/unary nodes in `createSpvConstantsFromConstSubTree()`.

Note when handling spec constants which are represented as ConstantUnion
Node, we should not use `OpSpecConstantOp` to initialize the composite
constant, so builder is set to 'normal mode'.

Tests:
Tests are added in Test/spv.specConstantOperations.vert, including:

1) Arithmetic, shift opeations for both scalar and composite type spec constants.
2) Size conversion from/to float and double for both scalar and vector.
3) Bitwise and/or/xor for both scalar and vector.
4) Unary negate/not for both scalar and vector.
5) Vector swizzles.
6) Comparisons for scalars.
7) == and != for composite type spec constants

Issues:
1) To implement == and != for composite type spec constants, the Spec needs
to allow OpAll, OpAny, OpFOrdEqual, OpFUnordEqual, OpOrdNotEqual,
OpFUnordNotEqual. Currently none of them are allowed in the Spec.
2016-03-30 16:18:26 -04:00
John Kessenich
4d65ee31a6 Generalize "main" to a settable entry point name. 2016-03-12 18:17:47 -07:00
qining
95aa52737d Change to traditional 'for' loop 2016-03-09 21:40:41 -05:00
qining
da39733f28 Remove decoration of undefined IDs
Fix issue #185 by removing OpDecorate instructions whose target IDs are
defined in unreachable blocks and thus not dumped in the generated
SPIR-V code.
2016-03-09 19:54:03 -05:00
John Kessenich
52d08596ec Merge pull request #175 from rdb/master
Fix compilation issues with MSVC 2010
2016-02-27 21:29:14 -07:00
baldurk
227e026dbf MSVC warning fix - conversion from size_t to int, possible loss of data 2016-02-25 21:19:49 +01:00
rdb
32084e889d Fix compilation issues with MSVC 2010
(mostly by eliminating use of range-based for loops and std::tie)
2016-02-23 22:17:38 +01:00
Mark Adams
18b637f9dc Fix warnings/errors for strict aliasing & function prototypes
This fixes various issues related to gcc's strict-aliasing warning
by using unions. It also handles various cases hit with
gcc's missing-declarations warning.
2016-02-23 12:17:11 -05:00
John Kessenich
6c292d3ba7 SPV: Implement Vulkan version of GLSL (KHR_vulkan_glsl). 2016-02-15 21:46:55 -07:00
John Kessenich
019f08fcd8 SPV: Fix issue #159: use ExplicitLod forms for non-fragment stages. 2016-02-15 15:40:42 -07:00
John Kessenich
5e80113939 SPV Capabilities: SampleRateShading, SparseResidency, MinLod, and ImageQuery. 2016-02-15 11:09:46 -07:00
John Kessenich
b0364dcc3e SPV Capabilities: Image types.
SampledBuffer
ImageBuffer
Sampled1D
Image1D
SampledCubeArray
ImageCubeArray
SampledRect
ImageRect
InputAttachment
ImageMSArray
StorageImageMultisample
2016-02-14 17:37:30 -07:00
John Kessenich
3c52207e8a SPV capabilities: Add int16, int64, float16, and float64. 2016-02-14 17:11:15 -07:00
John Kessenich
103bef9d74 SPV: Handle GLSL bool loads from a uniform buffer as a conversion from int -> bool.
SPIR-V bool is abstract; it has no bit pattern for storage with transparent memory.
OpenGL's convention is a bool in a uniform buffer is 32-bit int with non-0 being 'true'.
2016-02-08 21:41:30 -07:00
John Kessenich
32cfd49b68 SPV: RelaxedPrecision: Plumb this through the full AST->SPV translator. 2016-02-02 13:55:29 -07:00
Mark Adams
d5ac538e23 Fix two new Windows build warnings/errors 2016-02-01 19:13:06 -08:00
Dejan Mircevski
97605c86fd Fix #137 by avoiding a C++11 feature.
Apparently, older MSVC versions don't support brace-initializers for
function arguments.

Thanks @baldurk for a suggestion on his branch.
2016-01-20 10:25:28 -05:00
Dejan Mircevski
ed55bcd9f8 Don't remove SPIR-V blocks before codegen.
A removed block releases its instructions, so Module::idToInstruction
suddenly contains dangling references.  The original motivation for
block removal was to skip some unreachable blocks, but that's already
achieved by InReadableOrder.cpp.

Also updated stale comments.
2016-01-19 21:18:14 -05:00
Dejan Mircevski
e7f6cac1bd Merge branch 'topo' into loopgen-after-readable-order 2016-01-19 16:47:41 -05:00
Dejan Mircevski
454796e008 Call addPredecessor() on OpSwitch blocks. 2016-01-19 10:10:15 -05:00
Dejan Mircevski
5fe789b4af Add Block::successors. 2016-01-19 10:10:15 -05:00
John Kessenich
3e9add360d Merge pull request #133 from AWoloszyn/spirv-memory
Free memory associated with SPIR-V generation and the pragmaTable.
2016-01-18 10:48:07 -07:00
Andrew Woloszyn
b7946d16bb Free memory associated with SPIR-V generation. 2016-01-18 10:29:41 -05:00
John Kessenich
68f1431a55 Merge pull request #121 from amdrexu/feature
Parser & SPV: Implement two extensions regarding GLSL sparse texture.
2016-01-15 00:52:18 -07:00
Dejan Mircevski
dba2826328 Merge branch 'master' of github.com:google/glslang into loopgen
Change-Id: Ie8236430bb9e30a9be2e0c1573c42183c2f4d0d4
2016-01-11 16:26:18 -05:00
Dejan Mircevski
832c65c33b Fix back-branch target for do-while loops.
To ensure back branches always go to a header block, create a header
block even for !testFirst loops.  Then unify common code between the
testFirst/!testFirst cases.

Generate the header-block code first, so update golden files.

Realize that certain infinite loops generate invalid SPIR-V, so put a
TODO to instead abort code generation in such cases.

Change-Id: I1e173c8f73daad186cfc666b7d72bd563ed7665d
2016-01-11 15:57:11 -05:00
Dejan Mircevski
7819bee82c Switch loops stack to use LoopBlocks.
Also remove dead code.

Change-Id: I2c0177d8cab48b7d6f9442715aecb7951597f3c8
2016-01-11 09:35:22 -05:00
Dejan Mircevski
e537b8b488 Fix unreachable-block removal.
Add a test for loop without a condition.

Change-Id: Idd7fc462218a84b1e745207e2975a3f2897d30a0
2016-01-10 19:37:00 -05:00
Dejan Mircevski
9c6734c8df First cut at new loop codegen.
Change-Id: Id3bdf8b7a5606e7ce5d856ef225d5ddbe59a584b
2016-01-10 12:15:13 -05:00
Mark Adams
364c21c8c0 Fix several build warnings/error encountered with VS2013
This also fixes the newlines for spirv.hpp to be consistent with the rest of the files.
2016-01-06 13:41:02 -05:00
Rex Xu
48edadfd24 Parser & SPV: Implement two extensions regarding GLSL sparse texture.
Implement extension "GL_ARB_sparse_texture2".
Implement extension "GL_ARB_sparse_texture_clamp".
2016-01-05 16:07:02 +08:00
John Kessenich
e23c9849c2 SPV: Handle type punning in createCompositeCompare(). 2016-01-04 23:59:09 -07:00
John Kessenich
c9e0a42b92 SPV: Handle stride decorations for arrays of arrays, and using multiple type instances when strides are used. 2015-12-29 21:27:24 -07:00
John Kessenich
50e57560a1 SPV: Don't make MissingFunctionality() exit anymore. 2015-12-21 21:21:11 -07:00
John Kessenich
2211835b4d SPV: Implement composite comparisons (reductions across hierchical compare). 2015-12-21 20:54:09 -07:00
John Kessenich
76f7139789 SPV: Modify last merge: Have promoteScalar() generate the correct vectorType. 2015-12-09 19:08:42 -07:00
Rex Xu
e723b45bca SPV: Scalar smear operation should use type of the scalar to make the
expected vector type when the provided vector type is incompatible with
the scalar.
2015-12-09 16:05:06 +08:00
John Kessenich
55e7d11ce8 SPIR-V: Move from Version .99 Rev 31 to Version 1.0, Rev 2. 2015-11-15 22:31:41 -07:00
John Kessenich
f685df8c16 Add comment that was missed in previous commit. 2015-10-13 10:55:08 -06:00
John Kessenich
716312771c SPV constants and constOffsets completion: isConstant() and disassembler to see results.
Expand to full isConstant() implementation.
Fix disassembler to generate texture look-up masks.
2015-10-13 10:45:22 -06:00
Rex Xu
86e608133c Generate correct image operand mask for Offset and ConstOffset(#77) 2015-10-11 19:37:48 +08:00