Commit Graph

238 Commits

Author SHA1 Message Date
David Neto
ee1b3bb3bb Assembler support for image operands from Rev31
Rev32 and later add many more image operands, and
rearrange their values.
2015-10-26 12:55:33 -04:00
Lei Zhang
184c76dbaf Let EncodeAndDecodeSuccessfully remove preamble comments. 2015-10-26 12:55:33 -04:00
Lei Zhang
6d41581c93 Clean up code for encoding literal operands. 2015-10-26 12:55:33 -04:00
David Neto
3fca4cddee Remove SPV_OPERAND_TYPE_VARIABLE_MEMORY_ACCESS
If a memory mask operand is present, it is a mask.  The mask appears
only once, so just use SPV_OPERAND_TYPE_OPTIONAL_MEMORY_MASK.

The "variable literals" aspect comes into play as follows: if the
Aligned bit is set in the mask, then the parser will be made to
expect the alignment value as a literal number operand that follows
the mask.  That is done through mask operand expansion.
2015-10-26 12:55:33 -04:00
David Neto
5bf88fcc95 Assembler: mask expressions where 1 bits imply operands
Properly support a memory access mask with a combination
of bits, including the Aligned bit. When the Aligned bit is
set, the parser should expect an alignment value literal operand.
2015-10-26 12:55:33 -04:00
David Neto
fbf5cf4591 Document the syntax of mask expressions 2015-10-26 12:55:33 -04:00
David Neto
36b0c0f6b3 Assembler support for simple mask expressions
For example, support combining mask enums with "|",
such as "NotNaN|AllowRecip" for the fast math mode.

This is supported for mask values that don't modify the
expected operand pattern:
 - fast math mode
 - function control
 - loop control
 - selection control

TODO: disassembler support to print them as mask expressions.
2015-10-26 12:55:33 -04:00
David Neto
388c40d9c6 Generalize spvOperandTableNameLookup to take string length.
This is preparation for parsing mask expressions.
2015-10-26 12:55:33 -04:00
Dejan Mircevski
e3a19c0d63 Forbid !<integer> preceding or succeeding '='. 2015-10-26 12:55:33 -04:00
David Neto
b30a0c529b Assembler test for Dim, ImageFormat enums.
Adds SPV_OPERAND_TYPE_SAMPLER_IMAGE_FORMAT, and the
translation tables for its values from the spv::ImageFormat values.
2015-10-26 12:55:33 -04:00
David Neto
8382de2670 Document inability to use names from 3.12, 1.13
You can't use names from 3.12 Image Channel Order and
3.13 Image Channel Data Type since in the intstruction grammar,
they are only used as return values, but never named arguments
to instructions.
2015-10-26 12:55:33 -04:00
David Neto
f7ee0ca5da Use generalized EnumCase in AccessQualifier assembly tests
This is in preparation for coming tests that will also
use the templated EnumCase instead of making their
own structs.

Also reformat AccessQualifier test.
2015-10-26 12:55:33 -04:00
Dejan Mircevski
ba569fb1aa Handle IDs among !<integer> operands. 2015-10-26 12:55:33 -04:00
David Neto
5f33d31d89 Assembler test for SamplerAddressingMode, SamplerFilterMode
Uses OpConstantSampler instruction.
2015-10-26 12:55:33 -04:00
David Neto
5494dd4334 Assembler test for Storage Class enum values
This covers the storage classes in SPIR-V Rev31.  Rev32 has more.
2015-10-26 12:55:33 -04:00
Lei Zhang
b41d150b7f Support multiple word literal numbers as constants.
Add a new operand type SPV_OPERAND_TYPE_MULTIWORD_LITERAL_NUMBER
to represent those operands that can expand into multiple words.
Now only OpConstant and OpSpecConstant have such kind of operand.
2015-10-26 12:55:33 -04:00
David Neto
cebad036f7 Update CHANGES and Future Work 2015-10-26 12:55:33 -04:00
David Neto
fadbf6220e Assembler test for ExecutionScope, with OpGroupIAdd
Use spvOperandTypeStr to generalize the error message
for failure to parse an ID.
Extend spvOperandTypeStr to cover SPV_OPERAND_TYPE_EXECUTION_SCOPE.
2015-10-26 12:55:33 -04:00
David Neto
abe51c4142 Assembler test for OpTypePipe
Covers all AccessQualifier enums.

In Rev32, OpTypePipe no longer has a type argument.
Added TODOs for this.
2015-10-26 12:55:33 -04:00
David Neto
4a29131dcd Assembler test for single-valued Memory Access flag. 2015-10-26 12:55:33 -04:00
David Neto
f4fde6c60d Assembler test for single-valued FunctionControl mask
Fixed name of "Inline" enumerant.

Support for combinations of function control masks should come later.
2015-10-26 12:55:33 -04:00
Lei Zhang
97afd5c03c Remove useless variable. 2015-10-26 12:55:33 -04:00
David Neto
c0b32ecb54 Assembler test for single-valued LoopControl mask
Support for combining masks should come later.
2015-10-26 12:55:33 -04:00
David Neto
ad18c77ac9 Assembler test for single-valued SelectionControl mask
Support for combining masks should come later.
2015-10-26 12:55:33 -04:00
David Neto
d7aa15ff16 Clarify the syntax. Some named enumerants are unusable
You can't use a named enumerant if it's only meaningful
in an operand supplied as an ID to a target instruction.
The place where you'd use the name is something like an
OpConstant, but there's not enough context to bring those
names into scope, unless you're willing to tolerate
potential collisions.

Occurs for the names in:
- 3.25 Memory Semantics
- 3.27 Scope ID
- 3.29 Kernel Enqueue Flags
- 3.30 Kernel Profiling Info
2015-10-26 12:55:33 -04:00
Lei Zhang
41bf0733c6 Simplify code for handling literals in disassembler. 2015-10-26 12:55:33 -04:00
David Neto
2136ff70cc Move the syntax description into its own file 2015-10-26 12:55:33 -04:00
David Neto
c09932eb1e Add patch for generating opcode.inc syntax table. 2015-10-26 12:55:33 -04:00
Lei Zhang
a94701db39 Run clang-format. 2015-10-26 12:55:33 -04:00
David Neto
e8e3712c3b Assembler test for OpExecutionMode 2015-10-26 12:55:33 -04:00
David Neto
f6498bfc73 Assembler test for OpEntryPoint 2015-10-26 12:55:33 -04:00
Dejan Mircevski
58fc88e2ee Simplify format indication in CompileSuccesfully(). 2015-10-26 12:55:33 -04:00
Lei Zhang
4005670363 Limit the use of spvCheck and spvCheckReturn to validator.
spvCheck is indeed just an if-statement. Defining such a macro
doesn't help much.
2015-10-26 12:55:33 -04:00
David Neto
7114ffea06 Update readme with latest news. 2015-10-26 12:55:33 -04:00
Lei Zhang
4e092d3ac3 Move bitwisecast.h to source/ and add include guard. 2015-10-26 12:55:33 -04:00
David Neto
a48678ab92 Fix an infinite loop during message generation.
Rename getWord to spvGetWord and unit test it.
2015-10-26 12:55:33 -04:00
Lei Zhang
dbdf6e122c Add option to change assembly syntax format in the dis tool. 2015-10-26 12:55:33 -04:00
Dejan Mircevski
f79519c926 Handle operands when OpCode is !<integer>. 2015-10-26 12:55:33 -04:00
David Neto
e994e2e283 Add TODO for capability dependencies for Rev32+ 2015-10-26 12:55:33 -04:00
David Neto
4caa47c3d4 Assembler test for OpMemoryModel 2015-10-26 12:55:33 -04:00
Lei Zhang
29e667eb6c Add library function parameter to set assembly format for disassembler. 2015-10-26 12:55:33 -04:00
Lei Zhang
65ae325774 Add option to change assembly syntax format in the as tool. 2015-10-26 12:55:33 -04:00
David Neto
c6402d64f4 Assembler tests for OpCapability.
Adds test file for instructions in the Mode-Setting section
of the SPIR-V spec.
2015-10-26 12:55:33 -04:00
David Neto
55bdfcb0bc Assembler test for LinkageAttributes decorations 2015-10-26 12:54:39 -04:00
David Neto
37547b2304 Assembler tests for all decorations except Linkage attributes
Fixes typos in various parser tables.

Updates readme.md with the fixes.
2015-10-26 12:54:39 -04:00
Lei Zhang
e78a7c19ff Run clang-format. 2015-10-26 12:54:39 -04:00
Lei Zhang
06efdc59e1 Add an API parameter to choose assembly syntax format for assembler.
Added a new enum for supported assembly syntax formats:
Canonical Assembly Format (CAF) and Assignment Assembly Format (AAF).

Updated assembler interface functions to support choice of assembly
syntax format.
2015-10-26 12:54:39 -04:00
Dejan Mircevski
92a225b371 Simplify checking for !<integer> first word. 2015-10-26 12:54:39 -04:00
David Neto
a08350a2c3 Assembler test for OpSourceExtension.
Add test utility function MakeVector(std::string) to encode
a string as a vector of 32-bit words, as per the SPIR-V literal
string encoding.
2015-10-26 12:54:39 -04:00
Andrew Woloszyn
4b4acded77 Added missing diagnostic when disassembling empty file.
Added diagnostic messages for what should be an internal failure
that never happens. I figure if we return "failed" for something the
user cannot control we should print a message for it.
2015-10-26 12:54:39 -04:00