RPCS3/glslang
1
0
Fork 0
mirror of https://github.com/RPCS3/glslang.git synced 2025-01-25 20:54:24 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity
glslang/Test/baseResults/hlsl.structarray.flatten.geom.out

280 lines
18 KiB
Plaintext
Raw Normal View History

HLSL: Recursive composite flattening This PR implements recursive type flattening. For example, an array of structs of other structs can be flattened to individual member variables at the shader interface. This is sufficient for many purposes, e.g, uniforms containing opaque types, but is not sufficient for geometry shader arrayed inputs. That will be handled separately with structure splitting, which is not implemented by this PR. In the meantime, that case is detected and triggers an error. The recursive flattening extends the following three aspects of single-level flattening: - Flattening of structures to individual members with names such as "foo[0].samp[1]"; - Turning constant references to the nested composite type into a reference to a particular flattened member. - Shadow copies between arrays of flattened members and the nested composite type. Previous single-level flattening only flattened at the shader interface, and that is unchanged by this PR. Internally, shadow copies are, such as if the type is passed to a function. Also, the reasons for flattening are unchanged. Uniforms containing opaque types, and interface struct types are flattened. (The latter will change with structure splitting). One existing test changes: hlsl.structin.vert, which did in fact contain a nested composite type to be flattened. Two new tests are added: hlsl.structarray.flatten.frag, and hlsl.structarray.flatten.geom (currently issues an error until type splitting is online). The process of arriving at the individual member from chained postfix expressions is more complex than it was with one level. See large-ish comment above HlslParseContext::flatten() for details.
2016-11-28 17:09:54 -07:00
hlsl.structarray.flatten.geom
HLSL: Emit the OpSource HLSL instruction for HLSL, using new headers.
2017-04-07 15:33:08 -06:00
Shader version: 500
HLSL: Recursive composite flattening This PR implements recursive type flattening. For example, an array of structs of other structs can be flattened to individual member variables at the shader interface. This is sufficient for many purposes, e.g, uniforms containing opaque types, but is not sufficient for geometry shader arrayed inputs. That will be handled separately with structure splitting, which is not implemented by this PR. In the meantime, that case is detected and triggers an error. The recursive flattening extends the following three aspects of single-level flattening: - Flattening of structures to individual members with names such as "foo[0].samp[1]"; - Turning constant references to the nested composite type into a reference to a particular flattened member. - Shadow copies between arrays of flattened members and the nested composite type. Previous single-level flattening only flattened at the shader interface, and that is unchanged by this PR. Internally, shadow copies are, such as if the type is passed to a function. Also, the reasons for flattening are unchanged. Uniforms containing opaque types, and interface struct types are flattened. (The latter will change with structure splitting). One existing test changes: hlsl.structin.vert, which did in fact contain a nested composite type to be flattened. Two new tests are added: hlsl.structarray.flatten.frag, and hlsl.structarray.flatten.geom (currently issues an error until type splitting is online). The process of arriving at the individual member from chained postfix expressions is more complex than it was with one level. See large-ish comment above HlslParseContext::flatten() for details.
2016-11-28 17:09:54 -07:00
invocations = -1
max_vertices = 4
input primitive = lines
output primitive = triangle_strip
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
0:? Sequence
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:16 Function Definition: @main(struct-VertexData-vf4-vf4-vf21[2];struct-PS_IN-vf4-vf4-vf21; ( temp void)
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
0:16 Function Parameters:
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:16 'vin' ( in 2-element array of structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:16 'outStream' ( out structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
HLSL: Recursive composite flattening This PR implements recursive type flattening. For example, an array of structs of other structs can be flattened to individual member variables at the shader interface. This is sufficient for many purposes, e.g, uniforms containing opaque types, but is not sufficient for geometry shader arrayed inputs. That will be handled separately with structure splitting, which is not implemented by this PR. In the meantime, that case is detected and triggers an error. The recursive flattening extends the following three aspects of single-level flattening: - Flattening of structures to individual members with names such as "foo[0].samp[1]"; - Turning constant references to the nested composite type into a reference to a particular flattened member. - Shadow copies between arrays of flattened members and the nested composite type. Previous single-level flattening only flattened at the shader interface, and that is unchanged by this PR. Internally, shadow copies are, such as if the type is passed to a function. Also, the reasons for flattening are unchanged. Uniforms containing opaque types, and interface struct types are flattened. (The latter will change with structure splitting). One existing test changes: hlsl.structin.vert, which did in fact contain a nested composite type to be flattened. Two new tests are added: hlsl.structarray.flatten.frag, and hlsl.structarray.flatten.geom (currently issues an error until type splitting is online). The process of arriving at the individual member from chained postfix expressions is more complex than it was with one level. See large-ish comment above HlslParseContext::flatten() for details.
2016-11-28 17:09:54 -07:00
0:? Sequence
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:19 move second child to first child ( temp 4-component vector of float)
0:19 color: direct index for structure ( temp 4-component vector of float)
0:19 'vout' ( temp structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
0:19 Constant:
0:19 1 (const int)
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:19 color: direct index for structure ( temp 4-component vector of float)
0:19 direct index ( temp structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:19 'vin' ( in 2-element array of structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
0:19 Constant:
0:19 1 (const int)
0:19 Constant:
0:19 1 (const int)
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:20 move second child to first child ( temp 2-component vector of float)
0:20 uv: direct index for structure ( temp 2-component vector of float)
0:20 'vout' ( temp structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
0:20 Constant:
0:20 2 (const int)
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:20 uv: direct index for structure ( temp 2-component vector of float)
0:20 direct index ( temp structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:20 'vin' ( in 2-element array of structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
0:20 Constant:
0:20 1 (const int)
0:20 Constant:
0:20 2 (const int)
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:21 move second child to first child ( temp 4-component vector of float)
0:21 position: direct index for structure ( temp 4-component vector of float)
0:21 'vout' ( temp structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
0:21 Constant:
0:21 0 (const int)
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:21 position: direct index for structure ( temp 4-component vector of float)
0:21 direct index ( temp structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:21 'vin' ( in 2-element array of structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
0:21 Constant:
0:21 1 (const int)
0:21 Constant:
0:21 0 (const int)
0:22 Sequence
HLSL: fix GS implementation for EP wrapping The prior implementation of GS did not work with the new EP wrapping architecture. This fixes it: the Append() method now looks up the actual output rather than the internal sanitized temporary type, and writes to that.
2017-03-29 20:01:13 -06:00
0:22 Sequence
0:22 move second child to first child ( temp 4-component vector of float)
HLSL: Move debug naming to a simpler, more consistent, scheme. This will help in expanding flattening and reducing splitting.
2017-08-09 14:29:29 -06:00
0:? 'outStream.position' ( out 4-component vector of float Position)
HLSL: fix GS implementation for EP wrapping The prior implementation of GS did not work with the new EP wrapping architecture. This fixes it: the Append() method now looks up the actual output rather than the internal sanitized temporary type, and writes to that.
2017-03-29 20:01:13 -06:00
0:22 position: direct index for structure ( temp 4-component vector of float)
0:22 'vout' ( temp structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:22 Constant:
0:22 0 (const int)
0:22 move second child to first child ( temp 4-component vector of float)
HLSL: Flatten structs for all non-arrayed I/O interfaces.
2017-08-11 00:17:26 -06:00
0:? 'outStream.color' (layout( location=0) out 4-component vector of float)
HLSL: fix GS implementation for EP wrapping The prior implementation of GS did not work with the new EP wrapping architecture. This fixes it: the Append() method now looks up the actual output rather than the internal sanitized temporary type, and writes to that.
2017-03-29 20:01:13 -06:00
0:22 color: direct index for structure ( temp 4-component vector of float)
0:22 'vout' ( temp structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:22 Constant:
0:22 1 (const int)
0:22 move second child to first child ( temp 2-component vector of float)
HLSL: Flatten structs for all non-arrayed I/O interfaces.
2017-08-11 00:17:26 -06:00
0:? 'outStream.uv' (layout( location=1) out 2-component vector of float)
HLSL: fix GS implementation for EP wrapping The prior implementation of GS did not work with the new EP wrapping architecture. This fixes it: the Append() method now looks up the actual output rather than the internal sanitized temporary type, and writes to that.
2017-03-29 20:01:13 -06:00
0:22 uv: direct index for structure ( temp 2-component vector of float)
0:22 'vout' ( temp structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:22 Constant:
0:22 2 (const int)
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:22 EmitVertex ( temp void)
0:16 Function Definition: main( ( temp void)
HLSL: Wrap the entry-point; need to write 'in' args, and support 'inout' args. This needs some render testing, but is destined to be part of master. This also leads to a variety of other simplifications. - IO are global symbols, so only need one list of linkage nodes (deferred) - no longer need parse-context-wide 'inEntryPoint' state, entry-point is localized - several parts of splitting/flattening are now localized
2017-01-19 15:41:47 -07:00
0:16 Function Parameters:
0:? Sequence
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:16 move second child to first child ( temp 2-element array of structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:? 'vin' ( temp 2-element array of structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:? 'vin' (layout( location=0) in 2-element array of structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:16 Function Call: @main(struct-VertexData-vf4-vf4-vf21[2];struct-PS_IN-vf4-vf4-vf21; ( temp void)
0:? 'vin' ( temp 2-element array of structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:? 'outStream' ( temp structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
HLSL: Recursive composite flattening This PR implements recursive type flattening. For example, an array of structs of other structs can be flattened to individual member variables at the shader interface. This is sufficient for many purposes, e.g, uniforms containing opaque types, but is not sufficient for geometry shader arrayed inputs. That will be handled separately with structure splitting, which is not implemented by this PR. In the meantime, that case is detected and triggers an error. The recursive flattening extends the following three aspects of single-level flattening: - Flattening of structures to individual members with names such as "foo[0].samp[1]"; - Turning constant references to the nested composite type into a reference to a particular flattened member. - Shadow copies between arrays of flattened members and the nested composite type. Previous single-level flattening only flattened at the shader interface, and that is unchanged by this PR. Internally, shadow copies are, such as if the type is passed to a function. Also, the reasons for flattening are unchanged. Uniforms containing opaque types, and interface struct types are flattened. (The latter will change with structure splitting). One existing test changes: hlsl.structin.vert, which did in fact contain a nested composite type to be flattened. Two new tests are added: hlsl.structarray.flatten.frag, and hlsl.structarray.flatten.geom (currently issues an error until type splitting is online). The process of arriving at the individual member from chained postfix expressions is more complex than it was with one level. See large-ish comment above HlslParseContext::flatten() for details.
2016-11-28 17:09:54 -07:00
0:? Linker Objects
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:? 'vin' (layout( location=0) in 2-element array of structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
HLSL: Move debug naming to a simpler, more consistent, scheme. This will help in expanding flattening and reducing splitting.
2017-08-09 14:29:29 -06:00
0:? 'outStream.position' ( out 4-component vector of float Position)
HLSL: Flatten structs for all non-arrayed I/O interfaces.
2017-08-11 00:17:26 -06:00
0:? 'outStream.color' (layout( location=0) out 4-component vector of float)
0:? 'outStream.uv' (layout( location=1) out 2-component vector of float)
HLSL: Recursive composite flattening This PR implements recursive type flattening. For example, an array of structs of other structs can be flattened to individual member variables at the shader interface. This is sufficient for many purposes, e.g, uniforms containing opaque types, but is not sufficient for geometry shader arrayed inputs. That will be handled separately with structure splitting, which is not implemented by this PR. In the meantime, that case is detected and triggers an error. The recursive flattening extends the following three aspects of single-level flattening: - Flattening of structures to individual members with names such as "foo[0].samp[1]"; - Turning constant references to the nested composite type into a reference to a particular flattened member. - Shadow copies between arrays of flattened members and the nested composite type. Previous single-level flattening only flattened at the shader interface, and that is unchanged by this PR. Internally, shadow copies are, such as if the type is passed to a function. Also, the reasons for flattening are unchanged. Uniforms containing opaque types, and interface struct types are flattened. (The latter will change with structure splitting). One existing test changes: hlsl.structin.vert, which did in fact contain a nested composite type to be flattened. Two new tests are added: hlsl.structarray.flatten.frag, and hlsl.structarray.flatten.geom (currently issues an error until type splitting is online). The process of arriving at the individual member from chained postfix expressions is more complex than it was with one level. See large-ish comment above HlslParseContext::flatten() for details.
2016-11-28 17:09:54 -07:00
Linked geometry stage:
HLSL: Emit the OpSource HLSL instruction for HLSL, using new headers.
2017-04-07 15:33:08 -06:00
Shader version: 500
HLSL: Recursive composite flattening This PR implements recursive type flattening. For example, an array of structs of other structs can be flattened to individual member variables at the shader interface. This is sufficient for many purposes, e.g, uniforms containing opaque types, but is not sufficient for geometry shader arrayed inputs. That will be handled separately with structure splitting, which is not implemented by this PR. In the meantime, that case is detected and triggers an error. The recursive flattening extends the following three aspects of single-level flattening: - Flattening of structures to individual members with names such as "foo[0].samp[1]"; - Turning constant references to the nested composite type into a reference to a particular flattened member. - Shadow copies between arrays of flattened members and the nested composite type. Previous single-level flattening only flattened at the shader interface, and that is unchanged by this PR. Internally, shadow copies are, such as if the type is passed to a function. Also, the reasons for flattening are unchanged. Uniforms containing opaque types, and interface struct types are flattened. (The latter will change with structure splitting). One existing test changes: hlsl.structin.vert, which did in fact contain a nested composite type to be flattened. Two new tests are added: hlsl.structarray.flatten.frag, and hlsl.structarray.flatten.geom (currently issues an error until type splitting is online). The process of arriving at the individual member from chained postfix expressions is more complex than it was with one level. See large-ish comment above HlslParseContext::flatten() for details.
2016-11-28 17:09:54 -07:00
invocations = 1
max_vertices = 4
input primitive = lines
output primitive = triangle_strip
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
0:? Sequence
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:16 Function Definition: @main(struct-VertexData-vf4-vf4-vf21[2];struct-PS_IN-vf4-vf4-vf21; ( temp void)
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
0:16 Function Parameters:
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:16 'vin' ( in 2-element array of structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:16 'outStream' ( out structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
HLSL: Recursive composite flattening This PR implements recursive type flattening. For example, an array of structs of other structs can be flattened to individual member variables at the shader interface. This is sufficient for many purposes, e.g, uniforms containing opaque types, but is not sufficient for geometry shader arrayed inputs. That will be handled separately with structure splitting, which is not implemented by this PR. In the meantime, that case is detected and triggers an error. The recursive flattening extends the following three aspects of single-level flattening: - Flattening of structures to individual members with names such as "foo[0].samp[1]"; - Turning constant references to the nested composite type into a reference to a particular flattened member. - Shadow copies between arrays of flattened members and the nested composite type. Previous single-level flattening only flattened at the shader interface, and that is unchanged by this PR. Internally, shadow copies are, such as if the type is passed to a function. Also, the reasons for flattening are unchanged. Uniforms containing opaque types, and interface struct types are flattened. (The latter will change with structure splitting). One existing test changes: hlsl.structin.vert, which did in fact contain a nested composite type to be flattened. Two new tests are added: hlsl.structarray.flatten.frag, and hlsl.structarray.flatten.geom (currently issues an error until type splitting is online). The process of arriving at the individual member from chained postfix expressions is more complex than it was with one level. See large-ish comment above HlslParseContext::flatten() for details.
2016-11-28 17:09:54 -07:00
0:? Sequence
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:19 move second child to first child ( temp 4-component vector of float)
0:19 color: direct index for structure ( temp 4-component vector of float)
0:19 'vout' ( temp structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
0:19 Constant:
0:19 1 (const int)
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:19 color: direct index for structure ( temp 4-component vector of float)
0:19 direct index ( temp structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:19 'vin' ( in 2-element array of structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
0:19 Constant:
0:19 1 (const int)
0:19 Constant:
0:19 1 (const int)
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:20 move second child to first child ( temp 2-component vector of float)
0:20 uv: direct index for structure ( temp 2-component vector of float)
0:20 'vout' ( temp structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
0:20 Constant:
0:20 2 (const int)
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:20 uv: direct index for structure ( temp 2-component vector of float)
0:20 direct index ( temp structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:20 'vin' ( in 2-element array of structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
0:20 Constant:
0:20 1 (const int)
0:20 Constant:
0:20 2 (const int)
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:21 move second child to first child ( temp 4-component vector of float)
0:21 position: direct index for structure ( temp 4-component vector of float)
0:21 'vout' ( temp structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
0:21 Constant:
0:21 0 (const int)
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:21 position: direct index for structure ( temp 4-component vector of float)
0:21 direct index ( temp structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:21 'vin' ( in 2-element array of structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
0:21 Constant:
0:21 1 (const int)
0:21 Constant:
0:21 0 (const int)
0:22 Sequence
HLSL: fix GS implementation for EP wrapping The prior implementation of GS did not work with the new EP wrapping architecture. This fixes it: the Append() method now looks up the actual output rather than the internal sanitized temporary type, and writes to that.
2017-03-29 20:01:13 -06:00
0:22 Sequence
0:22 move second child to first child ( temp 4-component vector of float)
HLSL: Move debug naming to a simpler, more consistent, scheme. This will help in expanding flattening and reducing splitting.
2017-08-09 14:29:29 -06:00
0:? 'outStream.position' ( out 4-component vector of float Position)
HLSL: fix GS implementation for EP wrapping The prior implementation of GS did not work with the new EP wrapping architecture. This fixes it: the Append() method now looks up the actual output rather than the internal sanitized temporary type, and writes to that.
2017-03-29 20:01:13 -06:00
0:22 position: direct index for structure ( temp 4-component vector of float)
0:22 'vout' ( temp structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:22 Constant:
0:22 0 (const int)
0:22 move second child to first child ( temp 4-component vector of float)
HLSL: Flatten structs for all non-arrayed I/O interfaces.
2017-08-11 00:17:26 -06:00
0:? 'outStream.color' (layout( location=0) out 4-component vector of float)
HLSL: fix GS implementation for EP wrapping The prior implementation of GS did not work with the new EP wrapping architecture. This fixes it: the Append() method now looks up the actual output rather than the internal sanitized temporary type, and writes to that.
2017-03-29 20:01:13 -06:00
0:22 color: direct index for structure ( temp 4-component vector of float)
0:22 'vout' ( temp structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:22 Constant:
0:22 1 (const int)
0:22 move second child to first child ( temp 2-component vector of float)
HLSL: Flatten structs for all non-arrayed I/O interfaces.
2017-08-11 00:17:26 -06:00
0:? 'outStream.uv' (layout( location=1) out 2-component vector of float)
HLSL: fix GS implementation for EP wrapping The prior implementation of GS did not work with the new EP wrapping architecture. This fixes it: the Append() method now looks up the actual output rather than the internal sanitized temporary type, and writes to that.
2017-03-29 20:01:13 -06:00
0:22 uv: direct index for structure ( temp 2-component vector of float)
0:22 'vout' ( temp structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:22 Constant:
0:22 2 (const int)
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:22 EmitVertex ( temp void)
0:16 Function Definition: main( ( temp void)
HLSL: Wrap the entry-point; need to write 'in' args, and support 'inout' args. This needs some render testing, but is destined to be part of master. This also leads to a variety of other simplifications. - IO are global symbols, so only need one list of linkage nodes (deferred) - no longer need parse-context-wide 'inEntryPoint' state, entry-point is localized - several parts of splitting/flattening are now localized
2017-01-19 15:41:47 -07:00
0:16 Function Parameters:
0:? Sequence
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:16 move second child to first child ( temp 2-element array of structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:? 'vin' ( temp 2-element array of structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:? 'vin' (layout( location=0) in 2-element array of structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:16 Function Call: @main(struct-VertexData-vf4-vf4-vf21[2];struct-PS_IN-vf4-vf4-vf21; ( temp void)
0:? 'vin' ( temp 2-element array of structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
0:? 'outStream' ( temp structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
HLSL: Recursive composite flattening This PR implements recursive type flattening. For example, an array of structs of other structs can be flattened to individual member variables at the shader interface. This is sufficient for many purposes, e.g, uniforms containing opaque types, but is not sufficient for geometry shader arrayed inputs. That will be handled separately with structure splitting, which is not implemented by this PR. In the meantime, that case is detected and triggers an error. The recursive flattening extends the following three aspects of single-level flattening: - Flattening of structures to individual members with names such as "foo[0].samp[1]"; - Turning constant references to the nested composite type into a reference to a particular flattened member. - Shadow copies between arrays of flattened members and the nested composite type. Previous single-level flattening only flattened at the shader interface, and that is unchanged by this PR. Internally, shadow copies are, such as if the type is passed to a function. Also, the reasons for flattening are unchanged. Uniforms containing opaque types, and interface struct types are flattened. (The latter will change with structure splitting). One existing test changes: hlsl.structin.vert, which did in fact contain a nested composite type to be flattened. Two new tests are added: hlsl.structarray.flatten.frag, and hlsl.structarray.flatten.geom (currently issues an error until type splitting is online). The process of arriving at the individual member from chained postfix expressions is more complex than it was with one level. See large-ish comment above HlslParseContext::flatten() for details.
2016-11-28 17:09:54 -07:00
0:? Linker Objects
GLSL output: Removed fixed-size buffer; fixes #769. Makes some white-space differences in most output, plus a few cases where more could have been put out but was cut short by the previous fix-sized buffer.
2017-03-14 19:43:13 -06:00
0:? 'vin' (layout( location=0) in 2-element array of structure{ temp 4-component vector of float position, temp 4-component vector of float color, temp 2-component vector of float uv})
HLSL: Move debug naming to a simpler, more consistent, scheme. This will help in expanding flattening and reducing splitting.
2017-08-09 14:29:29 -06:00
0:? 'outStream.position' ( out 4-component vector of float Position)
HLSL: Flatten structs for all non-arrayed I/O interfaces.
2017-08-11 00:17:26 -06:00
0:? 'outStream.color' (layout( location=0) out 4-component vector of float)
0:? 'outStream.uv' (layout( location=1) out 2-component vector of float)
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
// Module Version 10000
SPV: GeneratorVersion: bump version number because of atomic decrement change.
2017-10-07 11:33:11 -06:00
// Generated by (magic number): 80002
HLSL: Flatten structs for all non-arrayed I/O interfaces.
2017-08-11 00:17:26 -06:00
// Id's are bound by 58
HLSL: Recursive composite flattening This PR implements recursive type flattening. For example, an array of structs of other structs can be flattened to individual member variables at the shader interface. This is sufficient for many purposes, e.g, uniforms containing opaque types, but is not sufficient for geometry shader arrayed inputs. That will be handled separately with structure splitting, which is not implemented by this PR. In the meantime, that case is detected and triggers an error. The recursive flattening extends the following three aspects of single-level flattening: - Flattening of structures to individual members with names such as "foo[0].samp[1]"; - Turning constant references to the nested composite type into a reference to a particular flattened member. - Shadow copies between arrays of flattened members and the nested composite type. Previous single-level flattening only flattened at the shader interface, and that is unchanged by this PR. Internally, shadow copies are, such as if the type is passed to a function. Also, the reasons for flattening are unchanged. Uniforms containing opaque types, and interface struct types are flattened. (The latter will change with structure splitting). One existing test changes: hlsl.structin.vert, which did in fact contain a nested composite type to be flattened. Two new tests are added: hlsl.structarray.flatten.frag, and hlsl.structarray.flatten.geom (currently issues an error until type splitting is online). The process of arriving at the individual member from chained postfix expressions is more complex than it was with one level. See large-ish comment above HlslParseContext::flatten() for details.
2016-11-28 17:09:54 -07:00
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
Capability Geometry
1: ExtInstImport "GLSL.std.450"
MemoryModel Logical GLSL450
HLSL: Flatten structs for all non-arrayed I/O interfaces.
2017-08-11 00:17:26 -06:00
EntryPoint Geometry 4 "main" 38 41 45 50
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
ExecutionMode 4 InputLines
ExecutionMode 4 Invocations 1
ExecutionMode 4 OutputTriangleStrip
ExecutionMode 4 OutputVertices 4
HLSL: Emit the OpSource HLSL instruction for HLSL, using new headers.
2017-04-07 15:33:08 -06:00
Source HLSL 500
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
Name 4 "main"
HLSL: Wrap the entry-point; need to write 'in' args, and support 'inout' args. This needs some render testing, but is destined to be part of master. This also leads to a variety of other simplifications. - IO are global symbols, so only need one list of linkage nodes (deferred) - no longer need parse-context-wide 'inEntryPoint' state, entry-point is localized - several parts of splitting/flattening are now localized
2017-01-19 15:41:47 -07:00
Name 9 "VertexData"
MemberName 9(VertexData) 0 "position"
MemberName 9(VertexData) 1 "color"
MemberName 9(VertexData) 2 "uv"
Name 14 "PS_IN"
MemberName 14(PS_IN) 0 "position"
MemberName 14(PS_IN) 1 "color"
MemberName 14(PS_IN) 2 "uv"
Name 19 "@main(struct-VertexData-vf4-vf4-vf21[2];struct-PS_IN-vf4-vf4-vf21;"
Name 17 "vin"
Name 18 "outStream"
HLSL: Make the entry-point shadow function have non-IO params and return. This also removes an no longer needed makeTemporary() and rationalizes makeTypeNonIo()'s interface.
2017-02-01 18:09:17 -07:00
Name 21 "vout"
HLSL: Move debug naming to a simpler, more consistent, scheme. This will help in expanding flattening and reducing splitting.
2017-08-09 14:29:29 -06:00
Name 38 "outStream.position"
HLSL: Flatten structs for all non-arrayed I/O interfaces.
2017-08-11 00:17:26 -06:00
Name 41 "outStream.color"
Name 45 "outStream.uv"
Name 48 "vin"
Name 50 "vin"
Name 52 "outStream"
Name 53 "param"
Name 55 "param"
HLSL: Move debug naming to a simpler, more consistent, scheme. This will help in expanding flattening and reducing splitting.
2017-08-09 14:29:29 -06:00
Decorate 38(outStream.position) BuiltIn Position
HLSL: Flatten structs for all non-arrayed I/O interfaces.
2017-08-11 00:17:26 -06:00
Decorate 41(outStream.color) Location 0
Decorate 45(outStream.uv) Location 1
Decorate 50(vin) Location 0
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
2: TypeVoid
3: TypeFunction 2
6: TypeFloat 32
7: TypeVector 6(float) 4
8: TypeVector 6(float) 2
HLSL: Wrap the entry-point; need to write 'in' args, and support 'inout' args. This needs some render testing, but is destined to be part of master. This also leads to a variety of other simplifications. - IO are global symbols, so only need one list of linkage nodes (deferred) - no longer need parse-context-wide 'inEntryPoint' state, entry-point is localized - several parts of splitting/flattening are now localized
2017-01-19 15:41:47 -07:00
9(VertexData): TypeStruct 7(fvec4) 7(fvec4) 8(fvec2)
10: TypeInt 32 0
11: 10(int) Constant 2
12: TypeArray 9(VertexData) 11
13: TypePointer Function 12
14(PS_IN): TypeStruct 7(fvec4) 7(fvec4) 8(fvec2)
15: TypePointer Function 14(PS_IN)
16: TypeFunction 2 13(ptr) 15(ptr)
HLSL: Make the entry-point shadow function have non-IO params and return. This also removes an no longer needed makeTemporary() and rationalizes makeTypeNonIo()'s interface.
2017-02-01 18:09:17 -07:00
22: TypeInt 32 1
23: 22(int) Constant 1
24: TypePointer Function 7(fvec4)
28: 22(int) Constant 2
29: TypePointer Function 8(fvec2)
33: 22(int) Constant 0
HLSL: fix GS implementation for EP wrapping The prior implementation of GS did not work with the new EP wrapping architecture. This fixes it: the Append() method now looks up the actual output rather than the internal sanitized temporary type, and writes to that.
2017-03-29 20:01:13 -06:00
37: TypePointer Output 7(fvec4)
HLSL: Move debug naming to a simpler, more consistent, scheme. This will help in expanding flattening and reducing splitting.
2017-08-09 14:29:29 -06:00
38(outStream.position): 37(ptr) Variable Output
HLSL: Flatten structs for all non-arrayed I/O interfaces.
2017-08-11 00:17:26 -06:00
41(outStream.color): 37(ptr) Variable Output
44: TypePointer Output 8(fvec2)
45(outStream.uv): 44(ptr) Variable Output
49: TypePointer Input 12
50(vin): 49(ptr) Variable Input
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
4(main): 2 Function None 3
5: Label
HLSL: Flatten structs for all non-arrayed I/O interfaces.
2017-08-11 00:17:26 -06:00
48(vin): 13(ptr) Variable Function
52(outStream): 15(ptr) Variable Function
53(param): 13(ptr) Variable Function
55(param): 15(ptr) Variable Function
51: 12 Load 50(vin)
Store 48(vin) 51
54: 12 Load 48(vin)
Store 53(param) 54
56: 2 FunctionCall 19(@main(struct-VertexData-vf4-vf4-vf21[2];struct-PS_IN-vf4-vf4-vf21;) 53(param) 55(param)
57: 14(PS_IN) Load 55(param)
Store 52(outStream) 57
HLSL: Wrap the entry-point; need to write 'in' args, and support 'inout' args. This needs some render testing, but is destined to be part of master. This also leads to a variety of other simplifications. - IO are global symbols, so only need one list of linkage nodes (deferred) - no longer need parse-context-wide 'inEntryPoint' state, entry-point is localized - several parts of splitting/flattening are now localized
2017-01-19 15:41:47 -07:00
Return
FunctionEnd
19(@main(struct-VertexData-vf4-vf4-vf21[2];struct-PS_IN-vf4-vf4-vf21;): 2 Function None 16
17(vin): 13(ptr) FunctionParameter
18(outStream): 15(ptr) FunctionParameter
20: Label
HLSL: Make the entry-point shadow function have non-IO params and return. This also removes an no longer needed makeTemporary() and rationalizes makeTypeNonIo()'s interface.
2017-02-01 18:09:17 -07:00
21(vout): 15(ptr) Variable Function
25: 24(ptr) AccessChain 17(vin) 23 23
26: 7(fvec4) Load 25
27: 24(ptr) AccessChain 21(vout) 23
Store 27 26
30: 29(ptr) AccessChain 17(vin) 23 28
31: 8(fvec2) Load 30
32: 29(ptr) AccessChain 21(vout) 28
Store 32 31
34: 24(ptr) AccessChain 17(vin) 23 33
35: 7(fvec4) Load 34
36: 24(ptr) AccessChain 21(vout) 33
Store 36 35
HLSL: fix GS implementation for EP wrapping The prior implementation of GS did not work with the new EP wrapping architecture. This fixes it: the Append() method now looks up the actual output rather than the internal sanitized temporary type, and writes to that.
2017-03-29 20:01:13 -06:00
39: 24(ptr) AccessChain 21(vout) 33
40: 7(fvec4) Load 39
HLSL: Move debug naming to a simpler, more consistent, scheme. This will help in expanding flattening and reducing splitting.
2017-08-09 14:29:29 -06:00
Store 38(outStream.position) 40
HLSL: Flatten structs for all non-arrayed I/O interfaces.
2017-08-11 00:17:26 -06:00
42: 24(ptr) AccessChain 21(vout) 23
43: 7(fvec4) Load 42
Store 41(outStream.color) 43
46: 29(ptr) AccessChain 21(vout) 28
47: 8(fvec2) Load 46
Store 45(outStream.uv) 47
HLSL: inter-stage structure splitting. This adds structure splitting, which among other things will enable GS support where input structs are passed, and thus become input arrays of structs in the GS inputs. That is a common GS case. The salient points of this PR are: * Structure splitting has been changed from "always between stages" to "only into the VS and out of the PS". It had previously happened between stages because it's not legal to pass a struct containing a builtin IO variable. * Structs passed between stages are now split into a struct containing ONLY user types, and a collection of loose builtin IO variables, if any. The user-part is passed as a normal struct between stages, which is valid SPIR-V now that the builtin IO is removed. * Internal to the shader, a sanitized struct (with IO qualifiers removed) is used, so that e.g, functions can work unmodified. * If a builtin IO such as Position occurs in an arrayed struct, for example as an input to a GS, the array reference is moved to the split-off loose variable, which is given the array dimension itself. When passing things around inside the shader, such as over a function call, the the original type is used in a sanitized form that removes the builtIn qualifications and makes them temporaries. This means internal function calls do not have to change. However, the type when returned from the shader will be member-wise copied from the internal sanitized one to the external type. The sanitized type is used in variable declarations. When copying split types and unsplit, if a sub-struct contains only user variables, it is copied as a single entity to avoid more AST verbosity. Above strategy arrived at with talks with @johnkslang. This is a big complex change. I'm inclined to leave it as a WIP until it can get some exposure to real world cases.
2016-12-14 15:22:25 -07:00
EmitVertex
Return
FunctionEnd
Reference in New Issue Copy Permalink