AuroraMiddleware/glslang
1
0
Fork 0
mirror of https://github.com/KhronosGroup/glslang synced 2024-11-08 11:30:06 +00:00
Code Issues Packages Projects Releases Wiki Activity

HLSL: add intrinsic function implicit promotions

Browse Source 
This PR handles implicit promotions for intrinsics when there is no exact match,
such as for example clamp(int, bool, float).  In this case the int and bool will
be promoted to a float, and the clamp(float, float, float) form used.

These promotions can be mixed with shape conversions, e.g, clamp(int, bool2, float2).

Output conversions are handled either via the existing addOutputArgumentConversion
function, which this PR generalizes to handle either aggregates or unaries, or by
intrinsic decomposition.  If there are methods or intrinsics to be decomposed,
then decomposition is responsible for any output conversions, which turns out to
happen automatically in all current cases.  This can be revisited once inout
conversions are in place.

Some cases of actual ambiguity were fixed in several tests, e.g, spv.register.autoassign.*

Some intrinsics with only uint versions were expanded to signed ints natively, where the
underlying AST and SPIR-V supports that.  E.g, countbits.  This avoids extraneous
conversion nodes.

A new function promoteAggregate is added, and used by findFunction.  This is essentially
a generalization of the "promote 1st or 2nd arg" algorithm in promoteBinary.

The actual selection proceeds in three steps, as described in the comments in
hlslParseContext::findFunction:

1. Attempt an exact match.  If found, use it.
2. If not, obtain the operator from step 1, and promote arguments.
3. Re-select the intrinsic overload from the results of step 2.
This commit is contained in:
steve-lunarg 2016-11-02 12:42:34 -06:00
parent 1c573fbcfb
commit ef33ec0925
19 changed files with 5453 additions and 3261 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4189
Test/baseResults/hlsl.intrinsics.frag.out
View File

File diff suppressed because it is too large Load Diff

184
Test/baseResults/hlsl.intrinsics.promote.down.frag.out Normal file
View File

@ -0,0 +1,184 @@
hlsl.intrinsics.promote.down.frag
Shader version: 450
gl_FragCoord origin is upper left
0:? Sequence
0:15 Function Definition: main( (temp structure{temp 4-component vector of float color})
0:15 Function Parameters:
0:? Sequence
0:16 Sequence
0:16 move second child to first child (temp uint)
0:16 'r00' (temp uint)
0:16 bitCount (temp uint)
0:16 Convert float to uint (temp uint)
0:16 f: direct index for structure (layout(offset=8 ) uniform float)
0:16 'anon@0' (layout(row_major std140 ) uniform block{layout(offset=0 ) uniform int i, layout(offset=4 ) uniform uint u, layout(offset=8 ) uniform float f, layout(offset=12 ) uniform bool b, layout(offset=16 ) uniform 2-component vector of int i2, layout(offset=24 ) uniform 2-component vector of uint u2, layout(offset=32 ) uniform 2-component vector of float f2, layout(offset=40 ) uniform 2-component vector of bool b2})
0:16 Constant:
0:16 2 (const uint)
0:17 Sequence
0:17 move second child to first child (temp 2-component vector of uint)
0:17 'r01' (temp 2-component vector of uint)
0:17 bitFieldReverse (temp 2-component vector of uint)
0:17 Convert float to uint (temp 2-component vector of uint)
0:17 f2: direct index for structure (layout(offset=32 ) uniform 2-component vector of float)
0:17 'anon@0' (layout(row_major std140 ) uniform block{layout(offset=0 ) uniform int i, layout(offset=4 ) uniform uint u, layout(offset=8 ) uniform float f, layout(offset=12 ) uniform bool b, layout(offset=16 ) uniform 2-component vector of int i2, layout(offset=24 ) uniform 2-component vector of uint u2, layout(offset=32 ) uniform 2-component vector of float f2, layout(offset=40 ) uniform 2-component vector of bool b2})
0:17 Constant:
0:17 6 (const uint)
0:20 move second child to first child (temp 4-component vector of float)
0:20 color: direct index for structure (temp 4-component vector of float)
0:20 'ps_output' (temp structure{temp 4-component vector of float color})
0:20 Constant:
0:20 0 (const int)
0:? Constant:
0:? 0.000000
0:? 0.000000
0:? 0.000000
0:? 0.000000
0:21 Sequence
0:21 Sequence
0:21 move second child to first child (temp 4-component vector of float)
0:? 'color' (layout(location=0 ) out 4-component vector of float)
0:21 color: direct index for structure (temp 4-component vector of float)
0:21 'ps_output' (temp structure{temp 4-component vector of float color})
0:21 Constant:
0:21 0 (const int)
0:21 Branch: Return
0:? Linker Objects
0:? 'color' (layout(location=0 ) out 4-component vector of float)
0:? 'anon@0' (layout(row_major std140 ) uniform block{layout(offset=0 ) uniform int i, layout(offset=4 ) uniform uint u, layout(offset=8 ) uniform float f, layout(offset=12 ) uniform bool b, layout(offset=16 ) uniform 2-component vector of int i2, layout(offset=24 ) uniform 2-component vector of uint u2, layout(offset=32 ) uniform 2-component vector of float f2, layout(offset=40 ) uniform 2-component vector of bool b2})
Linked fragment stage:
Shader version: 450
gl_FragCoord origin is upper left
0:? Sequence
0:15 Function Definition: main( (temp structure{temp 4-component vector of float color})
0:15 Function Parameters:
0:? Sequence
0:16 Sequence
0:16 move second child to first child (temp uint)
0:16 'r00' (temp uint)
0:16 bitCount (temp uint)
0:16 Convert float to uint (temp uint)
0:16 f: direct index for structure (layout(offset=8 ) uniform float)
0:16 'anon@0' (layout(row_major std140 ) uniform block{layout(offset=0 ) uniform int i, layout(offset=4 ) uniform uint u, layout(offset=8 ) uniform float f, layout(offset=12 ) uniform bool b, layout(offset=16 ) uniform 2-component vector of int i2, layout(offset=24 ) uniform 2-component vector of uint u2, layout(offset=32 ) uniform 2-component vector of float f2, layout(offset=40 ) uniform 2-component vector of bool b2})
0:16 Constant:
0:16 2 (const uint)
0:17 Sequence
0:17 move second child to first child (temp 2-component vector of uint)
0:17 'r01' (temp 2-component vector of uint)
0:17 bitFieldReverse (temp 2-component vector of uint)
0:17 Convert float to uint (temp 2-component vector of uint)
0:17 f2: direct index for structure (layout(offset=32 ) uniform 2-component vector of float)
0:17 'anon@0' (layout(row_major std140 ) uniform block{layout(offset=0 ) uniform int i, layout(offset=4 ) uniform uint u, layout(offset=8 ) uniform float f, layout(offset=12 ) uniform bool b, layout(offset=16 ) uniform 2-component vector of int i2, layout(offset=24 ) uniform 2-component vector of uint u2, layout(offset=32 ) uniform 2-component vector of float f2, layout(offset=40 ) uniform 2-component vector of bool b2})
0:17 Constant:
0:17 6 (const uint)
0:20 move second child to first child (temp 4-component vector of float)
0:20 color: direct index for structure (temp 4-component vector of float)
0:20 'ps_output' (temp structure{temp 4-component vector of float color})
0:20 Constant:
0:20 0 (const int)
0:? Constant:
0:? 0.000000
0:? 0.000000
0:? 0.000000
0:? 0.000000
0:21 Sequence
0:21 Sequence
0:21 move second child to first child (temp 4-component vector of float)
0:? 'color' (layout(location=0 ) out 4-component vector of float)
0:21 color: direct index for structure (temp 4-component vector of float)
0:21 'ps_output' (temp structure{temp 4-component vector of float color})
0:21 Constant:
0:21 0 (const int)
0:21 Branch: Return
0:? Linker Objects
0:? 'color' (layout(location=0 ) out 4-component vector of float)
0:? 'anon@0' (layout(row_major std140 ) uniform block{layout(offset=0 ) uniform int i, layout(offset=4 ) uniform uint u, layout(offset=8 ) uniform float f, layout(offset=12 ) uniform bool b, layout(offset=16 ) uniform 2-component vector of int i2, layout(offset=24 ) uniform 2-component vector of uint u2, layout(offset=32 ) uniform 2-component vector of float f2, layout(offset=40 ) uniform 2-component vector of bool b2})
// Module Version 10000
// Generated by (magic number): 80001
// Id's are bound by 45
Capability Shader
1: ExtInstImport "GLSL.std.450"
MemoryModel Logical GLSL450
EntryPoint Fragment 4 "main" 41
ExecutionMode 4 OriginUpperLeft
Name 4 "main"
Name 8 "r00"
Name 14 "$Global"
MemberName 14($Global) 0 "i"
MemberName 14($Global) 1 "u"
MemberName 14($Global) 2 "f"
MemberName 14($Global) 3 "b"
MemberName 14($Global) 4 "i2"
MemberName 14($Global) 5 "u2"
MemberName 14($Global) 6 "f2"
MemberName 14($Global) 7 "b2"
Name 16 ""
Name 24 "r01"
Name 32 "PS_OUTPUT"
MemberName 32(PS_OUTPUT) 0 "color"
Name 34 "ps_output"
Name 41 "color"
MemberDecorate 14($Global) 0 Offset 0
MemberDecorate 14($Global) 1 Offset 4
MemberDecorate 14($Global) 2 Offset 8
MemberDecorate 14($Global) 3 Offset 12
MemberDecorate 14($Global) 4 Offset 16
MemberDecorate 14($Global) 5 Offset 24
MemberDecorate 14($Global) 6 Offset 32
MemberDecorate 14($Global) 7 Offset 40
Decorate 14($Global) Block
Decorate 16 DescriptorSet 0
Decorate 41(color) Location 0
2: TypeVoid
3: TypeFunction 2
6: TypeInt 32 0
7: TypePointer Function 6(int)
9: TypeInt 32 1
10: TypeFloat 32
11: TypeVector 9(int) 2
12: TypeVector 6(int) 2
13: TypeVector 10(float) 2
14($Global): TypeStruct 9(int) 6(int) 10(float) 6(int) 11(ivec2) 12(ivec2) 13(fvec2) 12(ivec2)
15: TypePointer Uniform 14($Global)
16: 15(ptr) Variable Uniform
17: 9(int) Constant 2
18: TypePointer Uniform 10(float)
23: TypePointer Function 12(ivec2)
25: 9(int) Constant 6
26: TypePointer Uniform 13(fvec2)
31: TypeVector 10(float) 4
32(PS_OUTPUT): TypeStruct 31(fvec4)
33: TypePointer Function 32(PS_OUTPUT)
35: 9(int) Constant 0
36: 10(float) Constant 0
37: 31(fvec4) ConstantComposite 36 36 36 36
38: TypePointer Function 31(fvec4)
40: TypePointer Output 31(fvec4)
41(color): 40(ptr) Variable Output
4(main): 2 Function None 3
5: Label
8(r00): 7(ptr) Variable Function
24(r01): 23(ptr) Variable Function
34(ps_output): 33(ptr) Variable Function
19: 18(ptr) AccessChain 16 17
20: 10(float) Load 19
21: 6(int) ConvertFToU 20
22: 6(int) BitCount 21
Store 8(r00) 22
27: 26(ptr) AccessChain 16 25
28: 13(fvec2) Load 27
29: 12(ivec2) ConvertFToU 28
30: 12(ivec2) BitReverse 29
Store 24(r01) 30
39: 38(ptr) AccessChain 34(ps_output) 35
Store 39 37
42: 38(ptr) AccessChain 34(ps_output) 35
43: 31(fvec4) Load 42
Store 41(color) 43
Return
FunctionEnd

1313
Test/baseResults/hlsl.intrinsics.promote.frag.out Normal file
View File

File diff suppressed because it is too large Load Diff

337
Test/baseResults/hlsl.intrinsics.promote.outputs.frag.out Normal file
View File

@ -0,0 +1,337 @@
hlsl.intrinsics.promote.outputs.frag
Shader version: 450
gl_FragCoord origin is upper left
0:? Sequence
0:20 Function Definition: main( (temp structure{temp 4-component vector of float color})
0:20 Function Parameters:
0:? Sequence
0:37 clamp (temp float)
0:37 fpos: direct index for structure (layout(offset=52 ) uniform float)
0:37 'anon@0' (layout(row_major std140 ) uniform block{layout(offset=0 ) uniform int i, layout(offset=4 ) uniform uint u, layout(offset=8 ) uniform float f, layout(offset=12 ) uniform bool b, layout(offset=16 ) uniform 2-component vector of int i2, layout(offset=24 ) uniform 2-component vector of uint u2, layout(offset=32 ) uniform 2-component vector of float f2, layout(offset=40 ) uniform 2-component vector of bool b2, layout(offset=48 ) uniform uint upos, layout(offset=52 ) uniform float fpos})
0:37 Constant:
0:37 9 (const uint)
0:37 Constant:
0:37 0.000000
0:37 Constant:
0:37 1.000000
0:40 Sequence
0:40 move second child to first child (temp uint)
0:40 'sizeQueryTemp' (temp uint)
0:40 textureSize (temp uint)
0:40 'g_tTex1df4' (uniform texture1D)
0:40 move second child to first child (temp int)
0:40 'WidthI' (temp int)
0:40 Convert uint to int (temp int)
0:40 'sizeQueryTemp' (temp uint)
0:41 Sequence
0:41 move second child to first child (temp uint)
0:41 'sizeQueryTemp' (temp uint)
0:41 textureSize (temp uint)
0:41 'g_tTex1df4' (uniform texture1D)
0:41 Constant:
0:41 6 (const uint)
0:41 move second child to first child (temp int)
0:41 'WidthI' (temp int)
0:41 Convert uint to int (temp int)
0:41 'sizeQueryTemp' (temp uint)
0:41 move second child to first child (temp uint)
0:41 'NumberOfLevelsU' (temp uint)
0:41 textureQueryLevels (temp uint)
0:41 'g_tTex1df4' (uniform texture1D)
0:42 Sequence
0:42 move second child to first child (temp uint)
0:42 'sizeQueryTemp' (temp uint)
0:42 textureSize (temp uint)
0:42 'g_tTex1df4' (uniform texture1D)
0:42 Constant:
0:42 6 (const uint)
0:42 move second child to first child (temp uint)
0:42 'WidthU' (temp uint)
0:42 'sizeQueryTemp' (temp uint)
0:42 move second child to first child (temp int)
0:42 'NumberOfLevelsI' (temp int)
0:42 Convert uint to int (temp int)
0:42 textureQueryLevels (temp uint)
0:42 'g_tTex1df4' (uniform texture1D)
0:43 Sequence
0:43 move second child to first child (temp uint)
0:43 'sizeQueryTemp' (temp uint)
0:43 textureSize (temp uint)
0:43 'g_tTex1df4' (uniform texture1D)
0:43 Constant:
0:43 6 (const uint)
0:43 move second child to first child (temp int)
0:43 'WidthI' (temp int)
0:43 Convert uint to int (temp int)
0:43 'sizeQueryTemp' (temp uint)
0:43 move second child to first child (temp int)
0:43 'NumberOfLevelsI' (temp int)
0:43 Convert uint to int (temp int)
0:43 textureQueryLevels (temp uint)
0:43 'g_tTex1df4' (uniform texture1D)
0:47 move second child to first child (temp 4-component vector of float)
0:47 color: direct index for structure (temp 4-component vector of float)
0:47 'ps_output' (temp structure{temp 4-component vector of float color})
0:47 Constant:
0:47 0 (const int)
0:47 Constant:
0:47 0.000000
0:47 0.000000
0:47 0.000000
0:47 0.000000
0:48 Sequence
0:48 Sequence
0:48 move second child to first child (temp 4-component vector of float)
0:? 'color' (layout(location=0 ) out 4-component vector of float)
0:48 color: direct index for structure (temp 4-component vector of float)
0:48 'ps_output' (temp structure{temp 4-component vector of float color})
0:48 Constant:
0:48 0 (const int)
0:48 Branch: Return
0:? Linker Objects
0:? 'color' (layout(location=0 ) out 4-component vector of float)
0:? 'g_tTexbfs' (layout(r32f ) uniform samplerBuffer)
0:? 'g_tTex1df4' (uniform texture1D)
0:? 'anon@0' (layout(row_major std140 ) uniform block{layout(offset=0 ) uniform int i, layout(offset=4 ) uniform uint u, layout(offset=8 ) uniform float f, layout(offset=12 ) uniform bool b, layout(offset=16 ) uniform 2-component vector of int i2, layout(offset=24 ) uniform 2-component vector of uint u2, layout(offset=32 ) uniform 2-component vector of float f2, layout(offset=40 ) uniform 2-component vector of bool b2, layout(offset=48 ) uniform uint upos, layout(offset=52 ) uniform float fpos})
Linked fragment stage:
Shader version: 450
gl_FragCoord origin is upper left
0:? Sequence
0:20 Function Definition: main( (temp structure{temp 4-component vector of float color})
0:20 Function Parameters:
0:? Sequence
0:37 clamp (temp float)
0:37 fpos: direct index for structure (layout(offset=52 ) uniform float)
0:37 'anon@0' (layout(row_major std140 ) uniform block{layout(offset=0 ) uniform int i, layout(offset=4 ) uniform uint u, layout(offset=8 ) uniform float f, layout(offset=12 ) uniform bool b, layout(offset=16 ) uniform 2-component vector of int i2, layout(offset=24 ) uniform 2-component vector of uint u2, layout(offset=32 ) uniform 2-component vector of float f2, layout(offset=40 ) uniform 2-component vector of bool b2, layout(offset=48 ) uniform uint upos, layout(offset=52 ) uniform float fpos})
0:37 Constant:
0:37 9 (const uint)
0:37 Constant:
0:37 0.000000
0:37 Constant:
0:37 1.000000
0:40 Sequence
0:40 move second child to first child (temp uint)
0:40 'sizeQueryTemp' (temp uint)
0:40 textureSize (temp uint)
0:40 'g_tTex1df4' (uniform texture1D)
0:40 move second child to first child (temp int)
0:40 'WidthI' (temp int)
0:40 Convert uint to int (temp int)
0:40 'sizeQueryTemp' (temp uint)
0:41 Sequence
0:41 move second child to first child (temp uint)
0:41 'sizeQueryTemp' (temp uint)
0:41 textureSize (temp uint)
0:41 'g_tTex1df4' (uniform texture1D)
0:41 Constant:
0:41 6 (const uint)
0:41 move second child to first child (temp int)
0:41 'WidthI' (temp int)
0:41 Convert uint to int (temp int)
0:41 'sizeQueryTemp' (temp uint)
0:41 move second child to first child (temp uint)
0:41 'NumberOfLevelsU' (temp uint)
0:41 textureQueryLevels (temp uint)
0:41 'g_tTex1df4' (uniform texture1D)
0:42 Sequence
0:42 move second child to first child (temp uint)
0:42 'sizeQueryTemp' (temp uint)
0:42 textureSize (temp uint)
0:42 'g_tTex1df4' (uniform texture1D)
0:42 Constant:
0:42 6 (const uint)
0:42 move second child to first child (temp uint)
0:42 'WidthU' (temp uint)
0:42 'sizeQueryTemp' (temp uint)
0:42 move second child to first child (temp int)
0:42 'NumberOfLevelsI' (temp int)
0:42 Convert uint to int (temp int)
0:42 textureQueryLevels (temp uint)
0:42 'g_tTex1df4' (uniform texture1D)
0:43 Sequence
0:43 move second child to first child (temp uint)
0:43 'sizeQueryTemp' (temp uint)
0:43 textureSize (temp uint)
0:43 'g_tTex1df4' (uniform texture1D)
0:43 Constant:
0:43 6 (const uint)
0:43 move second child to first child (temp int)
0:43 'WidthI' (temp int)
0:43 Convert uint to int (temp int)
0:43 'sizeQueryTemp' (temp uint)
0:43 move second child to first child (temp int)
0:43 'NumberOfLevelsI' (temp int)
0:43 Convert uint to int (temp int)
0:43 textureQueryLevels (temp uint)
0:43 'g_tTex1df4' (uniform texture1D)
0:47 move second child to first child (temp 4-component vector of float)
0:47 color: direct index for structure (temp 4-component vector of float)
0:47 'ps_output' (temp structure{temp 4-component vector of float color})
0:47 Constant:
0:47 0 (const int)
0:47 Constant:
0:47 0.000000
0:47 0.000000
0:47 0.000000
0:47 0.000000
0:48 Sequence
0:48 Sequence
0:48 move second child to first child (temp 4-component vector of float)
0:? 'color' (layout(location=0 ) out 4-component vector of float)
0:48 color: direct index for structure (temp 4-component vector of float)
0:48 'ps_output' (temp structure{temp 4-component vector of float color})
0:48 Constant:
0:48 0 (const int)
0:48 Branch: Return
0:? Linker Objects
0:? 'color' (layout(location=0 ) out 4-component vector of float)
0:? 'g_tTexbfs' (layout(r32f ) uniform samplerBuffer)
0:? 'g_tTex1df4' (uniform texture1D)
0:? 'anon@0' (layout(row_major std140 ) uniform block{layout(offset=0 ) uniform int i, layout(offset=4 ) uniform uint u, layout(offset=8 ) uniform float f, layout(offset=12 ) uniform bool b, layout(offset=16 ) uniform 2-component vector of int i2, layout(offset=24 ) uniform 2-component vector of uint u2, layout(offset=32 ) uniform 2-component vector of float f2, layout(offset=40 ) uniform 2-component vector of bool b2, layout(offset=48 ) uniform uint upos, layout(offset=52 ) uniform float fpos})
// Module Version 10000
// Generated by (magic number): 80001
// Id's are bound by 76
Capability Shader
Capability Sampled1D
Capability SampledBuffer
Capability ImageQuery
1: ExtInstImport "GLSL.std.450"
MemoryModel Logical GLSL450
EntryPoint Fragment 4 "main" 68
ExecutionMode 4 OriginUpperLeft
Name 4 "main"
Name 12 "$Global"
MemberName 12($Global) 0 "i"
MemberName 12($Global) 1 "u"
MemberName 12($Global) 2 "f"
MemberName 12($Global) 3 "b"
MemberName 12($Global) 4 "i2"
MemberName 12($Global) 5 "u2"
MemberName 12($Global) 6 "f2"
MemberName 12($Global) 7 "b2"
MemberName 12($Global) 8 "upos"
MemberName 12($Global) 9 "fpos"
Name 14 ""
Name 23 "sizeQueryTemp"
Name 26 "g_tTex1df4"
Name 30 "WidthI"
Name 33 "sizeQueryTemp"
Name 39 "NumberOfLevelsU"
Name 42 "sizeQueryTemp"
Name 45 "WidthU"
Name 47 "NumberOfLevelsI"
Name 51 "sizeQueryTemp"
Name 60 "PS_OUTPUT"
MemberName 60(PS_OUTPUT) 0 "color"
Name 62 "ps_output"
Name 68 "color"
Name 75 "g_tTexbfs"
MemberDecorate 12($Global) 0 Offset 0
MemberDecorate 12($Global) 1 Offset 4
MemberDecorate 12($Global) 2 Offset 8
MemberDecorate 12($Global) 3 Offset 12
MemberDecorate 12($Global) 4 Offset 16
MemberDecorate 12($Global) 5 Offset 24
MemberDecorate 12($Global) 6 Offset 32
MemberDecorate 12($Global) 7 Offset 40
MemberDecorate 12($Global) 8 Offset 48
MemberDecorate 12($Global) 9 Offset 52
Decorate 12($Global) Block
Decorate 14 DescriptorSet 0
Decorate 26(g_tTex1df4) DescriptorSet 0
Decorate 68(color) Location 0
Decorate 75(g_tTexbfs) DescriptorSet 0
2: TypeVoid
3: TypeFunction 2
6: TypeInt 32 1
7: TypeInt 32 0
8: TypeFloat 32
9: TypeVector 6(int) 2
10: TypeVector 7(int) 2
11: TypeVector 8(float) 2
12($Global): TypeStruct 6(int) 7(int) 8(float) 7(int) 9(ivec2) 10(ivec2) 11(fvec2) 10(ivec2) 7(int) 8(float)
13: TypePointer Uniform 12($Global)
14: 13(ptr) Variable Uniform
15: 6(int) Constant 9
16: TypePointer Uniform 8(float)
19: 8(float) Constant 0
20: 8(float) Constant 1065353216
22: TypePointer Function 7(int)
24: TypeImage 8(float) 1D sampled format:Unknown
25: TypePointer UniformConstant 24
26(g_tTex1df4): 25(ptr) Variable UniformConstant
29: TypePointer Function 6(int)
35: 7(int) Constant 6
59: TypeVector 8(float) 4
60(PS_OUTPUT): TypeStruct 59(fvec4)
61: TypePointer Function 60(PS_OUTPUT)
63: 6(int) Constant 0
64: 59(fvec4) ConstantComposite 19 19 19 19
65: TypePointer Function 59(fvec4)
67: TypePointer Output 59(fvec4)
68(color): 67(ptr) Variable Output
72: TypeImage 8(float) Buffer sampled format:R32f
73: TypeSampledImage 72
74: TypePointer UniformConstant 73
75(g_tTexbfs): 74(ptr) Variable UniformConstant
4(main): 2 Function None 3
5: Label
23(sizeQueryTemp): 22(ptr) Variable Function
30(WidthI): 29(ptr) Variable Function
33(sizeQueryTemp): 22(ptr) Variable Function
39(NumberOfLevelsU): 22(ptr) Variable Function
42(sizeQueryTemp): 22(ptr) Variable Function
45(WidthU): 22(ptr) Variable Function
47(NumberOfLevelsI): 29(ptr) Variable Function
51(sizeQueryTemp): 22(ptr) Variable Function
62(ps_output): 61(ptr) Variable Function
17: 16(ptr) AccessChain 14 15
18: 8(float) Load 17
21: 8(float) ExtInst 1(GLSL.std.450) 43(FClamp) 18 19 20
27: 24 Load 26(g_tTex1df4)
28: 6(int) ImageQuerySize 27
Store 23(sizeQueryTemp) 28
31: 7(int) Load 23(sizeQueryTemp)
32: 6(int) Bitcast 31
Store 30(WidthI) 32
34: 24 Load 26(g_tTex1df4)
36: 6(int) ImageQuerySizeLod 34 35
Store 33(sizeQueryTemp) 36
37: 7(int) Load 33(sizeQueryTemp)
38: 6(int) Bitcast 37
Store 30(WidthI) 38
40: 24 Load 26(g_tTex1df4)
41: 6(int) ImageQueryLevels 40
Store 39(NumberOfLevelsU) 41
43: 24 Load 26(g_tTex1df4)
44: 6(int) ImageQuerySizeLod 43 35
Store 42(sizeQueryTemp) 44
46: 7(int) Load 42(sizeQueryTemp)
Store 45(WidthU) 46
48: 24 Load 26(g_tTex1df4)
49: 6(int) ImageQueryLevels 48
50: 6(int) Bitcast 49
Store 47(NumberOfLevelsI) 50
52: 24 Load 26(g_tTex1df4)
53: 6(int) ImageQuerySizeLod 52 35
Store 51(sizeQueryTemp) 53
54: 7(int) Load 51(sizeQueryTemp)
55: 6(int) Bitcast 54
Store 30(WidthI) 55
56: 24 Load 26(g_tTex1df4)
57: 6(int) ImageQueryLevels 56
58: 6(int) Bitcast 57
Store 47(NumberOfLevelsI) 58
66: 65(ptr) AccessChain 62(ps_output) 63
Store 66 64
69: 65(ptr) AccessChain 62(ps_output) 63
70: 59(fvec4) Load 69
Store 68(color) 70
Return
FunctionEnd

2242
Test/baseResults/hlsl.intrinsics.vert.out
View File

File diff suppressed because it is too large Load Diff

27
Test/baseResults/spv.register.autoassign-2.frag.out
View File

@ -5,7 +5,7 @@ Linked fragment stage:
// Module Version 10000
// Generated by (magic number): 80001
// Id's are bound by 30
// Id's are bound by 31
Capability Shader
1: ExtInstImport "GLSL.std.450"
@ -16,14 +16,14 @@ Linked fragment stage:
Name 9 "Color"
Name 12 "g_tScene[0]"
Name 16 "g_tSamp"
Name 24 "g_tScene[1]"
Name 25 "g_tScene[1]"
Decorate 9(Color) Location 0
Decorate 12(g_tScene[0]) DescriptorSet 0
Decorate 12(g_tScene[0]) Binding 10
Decorate 16(g_tSamp) DescriptorSet 0
Decorate 16(g_tSamp) Binding 5
Decorate 24(g_tScene[1]) DescriptorSet 0
Decorate 24(g_tScene[1]) Binding 11
Decorate 25(g_tScene[1]) DescriptorSet 0
Decorate 25(g_tScene[1]) Binding 11
2: TypeVoid
3: TypeFunction 2
6: TypeFloat 32
@ -39,19 +39,20 @@ Linked fragment stage:
18: TypeSampledImage 10
20: TypeVector 6(float) 2
21: 6(float) Constant 1050253722
22: 20(fvec2) ConstantComposite 21 21
24(g_tScene[1]): 11(ptr) Variable UniformConstant
22: 6(float) Constant 1053609165
23: 20(fvec2) ConstantComposite 21 22
25(g_tScene[1]): 11(ptr) Variable UniformConstant
4(main): 2 Function None 3
5: Label
13: 10 Load 12(g_tScene[0])
17: 14 Load 16(g_tSamp)
19: 18 SampledImage 13 17
23: 7(fvec4) ImageSampleImplicitLod 19 22
25: 10 Load 24(g_tScene[1])
26: 14 Load 16(g_tSamp)
27: 18 SampledImage 25 26
28: 7(fvec4) ImageSampleImplicitLod 27 22
29: 7(fvec4) FAdd 23 28
Store 9(Color) 29
24: 7(fvec4) ImageSampleImplicitLod 19 23
26: 10 Load 25(g_tScene[1])
27: 14 Load 16(g_tSamp)
28: 18 SampledImage 26 27
29: 7(fvec4) ImageSampleImplicitLod 28 23
30: 7(fvec4) FAdd 24 29
Store 9(Color) 30
Return
FunctionEnd

4
Test/hlsl.intrinsics.frag
View File

@ -33,7 +33,7 @@ float PixelShaderFunctionS(float inF0, float inF1, float inF2, uint inU0, uint i
clip(inF0);
float r014 = cos(inF0);
float r015 = cosh(inF0);
uint r016 = countbits(7);
int r016 = countbits(7);
float r017 = ddx(inF0);
float r018 = ddx_coarse(inF0);
float r019 = ddx_fine(inF0);
@ -111,7 +111,7 @@ float2 PixelShaderFunction2(float2 inF0, float2 inF1, float2 inF2, uint2 inU0, u
clip(inF0);
float2 r013 = cos(inF0);
float2 r015 = cosh(inF0);
uint2 r016 = countbits(int2(7,3));
int2 r016 = countbits(int2(7,3));
float2 r017 = ddx(inF0);
float2 r018 = ddx_coarse(inF0);
float2 r019 = ddx_fine(inF0);

22
Test/hlsl.intrinsics.promote.down.frag Normal file
View File

@ -0,0 +1,22 @@
struct PS_OUTPUT { float4 color : SV_Target0; };
int i;
uint u;
float f;
bool b;
int2 i2;
uint2 u2;
float2 f2;
bool2 b2;
PS_OUTPUT main()
{
uint r00 = countbits(f);
uint2 r01 = reversebits(f2);
PS_OUTPUT ps_output;
ps_output.color = float4(0,0,0,0);
return ps_output;
};

79
Test/hlsl.intrinsics.promote.frag Normal file
View File

@ -0,0 +1,79 @@
struct PS_OUTPUT { float4 color : SV_Target0; };
int i;
uint u;
float f;
bool b;
int2 i2;
uint2 u2;
float2 f2;
bool2 b2;
Buffer <float> g_tTexbfs;
Texture1D <float4> g_tTex1df4;
uint upos;
float fpos;
PS_OUTPUT main()
{
// Same shapes:
float r00 = max(b, f);
uint r01 = max(b, u);
int r02 = max(b, i);
float r03 = max(i, f);
float r04 = max(u, f);
float2 r10 = max(b2, f2);
uint2 r11 = max(b2, u2);
int2 r12 = max(b2, i2);
float2 r13 = max(i2, f2);
float2 r14 = max(u2, f2);
float2 r20 = clamp(i2, u2, f2); // 3 args, converts all to best type.
uint2 r21 = clamp(b2, u2, b2);
float2 r22 = clamp(b2, f2, b2);
// Mixed shapes:
float2 r30 = max(b, f2);
uint2 r31 = max(b, u2);
int2 r32 = max(b, i2);
float2 r33 = max(i, f2);
float2 r34 = max(u, f2);
float2 r40 = clamp(i, u2, f2); // 3 args, converts all to best type.
uint2 r41 = clamp(b2, u, b2);
float2 r42 = clamp(b2, f, b);
int2 r43 = clamp(i, i2, u2);
float r50 = g_tTexbfs.Load(upos);
float r51 = g_tTexbfs.Load(fpos);
int MipLevel;
uint WidthU;
uint HeightU;
uint ElementsU;
uint DepthU;
uint NumberOfLevelsU;
uint NumberOfSamplesU;
int WidthI;
int HeightI;
int ElementsI;
int DepthI;
int NumberOfLevelsI;
int NumberOfSamplesI;
g_tTex1df4 . GetDimensions(WidthI);
g_tTex1df4 . GetDimensions(6, WidthI, NumberOfLevelsU);
g_tTex1df4 . GetDimensions(6, WidthU, NumberOfLevelsI);
g_tTex1df4 . GetDimensions(6, WidthI, NumberOfLevelsI);
// max(i2, f2);
PS_OUTPUT ps_output;
ps_output.color = r00;
return ps_output;
};

49
Test/hlsl.intrinsics.promote.outputs.frag Normal file
View File

@ -0,0 +1,49 @@
struct PS_OUTPUT { float4 color : SV_Target0; };
int i;
uint u;
float f;
bool b;
int2 i2;
uint2 u2;
float2 f2;
bool2 b2;
Buffer <float> g_tTexbfs;
Texture1D <float4> g_tTex1df4;
uint upos;
float fpos;
PS_OUTPUT main()
{
int MipLevel;
uint WidthU;
uint HeightU;
uint ElementsU;
uint DepthU;
uint NumberOfLevelsU;
uint NumberOfSamplesU;
int WidthI;
int HeightI;
int ElementsI;
int DepthI;
int NumberOfLevelsI;
int NumberOfSamplesI;
saturate(fpos);
// Test output promotions
g_tTex1df4 . GetDimensions(WidthI);
g_tTex1df4 . GetDimensions(6, WidthI, NumberOfLevelsU);
g_tTex1df4 . GetDimensions(6, WidthU, NumberOfLevelsI);
g_tTex1df4 . GetDimensions(6, WidthI, NumberOfLevelsI);
// max(i2, f2);
PS_OUTPUT ps_output;
ps_output.color = 0;
return ps_output;
};

4
Test/spv.register.autoassign-2.frag
View File

@ -10,6 +10,6 @@ struct PS_OUTPUT
void main(out PS_OUTPUT psout)
{
psout.Color = g_tScene[0].Sample(g_tSamp, 0.3) +
g_tScene[1].Sample(g_tSamp, 0.3);
psout.Color = g_tScene[0].Sample(g_tSamp, float2(0.3,0.4)) +
g_tScene[1].Sample(g_tSamp, float2(0.3,0.4));
}

4
Test/spv.register.autoassign.rangetest.frag
View File

@ -10,6 +10,6 @@ struct PS_OUTPUT
void main(out PS_OUTPUT psout)
{
psout.Color = g_tScene[0].Sample(g_tSamp, 0.3) +
g_tScene[1].Sample(g_tSamp, 0.3);
psout.Color = g_tScene[0].Sample(g_tSamp, float2(0.3, 0.3)) +
g_tScene[1].Sample(g_tSamp, float2(0.3, 0.3));
}

103
glslang/MachineIndependent/Intermediate.cpp
View File

@ -45,6 +45,7 @@
#include "propagateNoContraction.h"
#include <cfloat>
#include <utility>
namespace glslang {
@ -575,6 +576,27 @@ TIntermTyped* TIntermediate::addConversion(TOperator op, const TType& type, TInt
case EOpDivAssign:
case EOpModAssign:
case EOpAtan:
case EOpClamp:
case EOpCross:
case EOpDistance:
case EOpDot:
case EOpDst:
case EOpFaceForward:
case EOpFma:
case EOpFrexp:
case EOpLdexp:
case EOpMix:
case EOpLit:
case EOpMax:
case EOpMin:
case EOpModf:
case EOpPow:
case EOpReflect:
case EOpRefract:
case EOpSmoothStep:
case EOpStep:
case EOpSequence:
case EOpConstructStruct:
@ -833,6 +855,9 @@ bool TIntermediate::canImplicitlyPromote(TBasicType from, TBasicType to, TOperat
if (profile == EEsProfile || version == 110)
return false;
if (from == to)
return true;
// TODO: Move more policies into language-specific handlers.
// Some languages allow more general (or potentially, more specific) conversions under some conditions.
if (source == EShSourceHlsl) {
@ -901,6 +926,8 @@ bool TIntermediate::canImplicitlyPromote(TBasicType from, TBasicType to, TOperat
return version >= 400;
case EbtUint:
return true;
case EbtBool:
return (source == EShSourceHlsl);
default:
return false;
}
@ -908,6 +935,8 @@ bool TIntermediate::canImplicitlyPromote(TBasicType from, TBasicType to, TOperat
switch (from) {
case EbtInt:
return true;
case EbtBool:
return (source == EShSourceHlsl);
default:
return false;
}
@ -1747,6 +1776,9 @@ bool TIntermediate::promote(TIntermOperator* node)
if (node->getAsBinaryNode())
return promoteBinary(*node->getAsBinaryNode());
if (node->getAsAggregate())
return promoteAggregate(*node->getAsAggregate());
return false;
}
@ -2190,6 +2222,77 @@ bool TIntermediate::promoteBinary(TIntermBinary& node)
return true;
}
//
// See TIntermediate::promote
//
bool TIntermediate::promoteAggregate(TIntermAggregate& node)
{
TOperator op = node.getOp();
TIntermSequence& args = node.getSequence();
const int numArgs = args.size();
// Presently, only hlsl does intrinsic promotions.
if (getSource() != EShSourceHlsl)
return true;
// set of opcodes that can be promoted in this manner.
switch (op) {
case EOpAtan:
case EOpClamp:
case EOpCross:
case EOpDistance:
case EOpDot:
case EOpDst:
case EOpFaceForward:
// case EOpFindMSB: TODO: ??
// case EOpFindLSB: TODO: ??
case EOpFma:
case EOpMod:
case EOpFrexp:
case EOpLdexp:
case EOpMix:
case EOpLit:
case EOpMax:
case EOpMin:
case EOpModf:
// case EOpGenMul: TODO: ??
case EOpPow:
case EOpReflect:
case EOpRefract:
// case EOpSinCos: TODO: ??
case EOpSmoothStep:
case EOpStep:
break;
default:
return true;
}
// TODO: array and struct behavior
// Try converting all nodes to the given node's type
TIntermSequence convertedArgs(numArgs, nullptr);
// Try to convert all types to the nonConvArg type.
for (int nonConvArg = 0; nonConvArg < numArgs; ++nonConvArg) {
// Try converting all args to this arg's type
for (int convArg = 0; convArg < numArgs; ++convArg) {
convertedArgs[convArg] = addConversion(op, args[nonConvArg]->getAsTyped()->getType(),
args[convArg]->getAsTyped());
}
// If we successfully converted all the args, use the result.
if (std::all_of(convertedArgs.begin(), convertedArgs.end(),
[](const TIntermNode* node) { return node != nullptr; })) {
std::swap(args, convertedArgs);
return true;
}
}
return false;
}
void TIntermBinary::updatePrecision()
{
#ifdef AMD_EXTENSIONS

3
glslang/MachineIndependent/linkValidate.cpp
View File

@ -370,6 +370,9 @@ void TIntermediate::mergeErrorCheck(TInfoSink& infoSink, const TIntermSymbol& sy
//
void TIntermediate::finalCheck(TInfoSink& infoSink)
{
if (getTreeRoot() == nullptr)
return;
if (source == EShSourceGlsl && numEntryPoints < 1)
error(infoSink, "Missing entry point: Each stage requires one entry point");

3
glslang/MachineIndependent/localintermediate.h
View File

@ -381,6 +381,7 @@ public:
int addXfbBufferOffset(const TType&);
unsigned int computeTypeXfbSize(const TType&, bool& containsDouble) const;
static int getBaseAlignment(const TType&, int& size, int& stride, bool std140, bool rowMajor);
bool promote(TIntermOperator*);
protected:
TIntermSymbol* addSymbol(int Id, const TString&, const TType&, const TConstUnionArray&, TIntermTyped* subtree, const TSourceLoc&);
@ -395,10 +396,10 @@ protected:
bool userOutputUsed() const;
static int getBaseAlignmentScalar(const TType&, int& size);
bool isSpecializationOperation(const TIntermOperator&) const;
bool promote(TIntermOperator*);
bool promoteUnary(TIntermUnary&);
bool promoteBinary(TIntermBinary&);
void addSymbolLinkageNode(TIntermAggregate*& linkage, TSymbolTable&, const TString&);
bool promoteAggregate(TIntermAggregate&);
const EShLanguage language; // stage, known at construction time
EShSource source; // source language, known a bit later

3
gtests/Hlsl.FromFile.cpp
View File

@ -134,6 +134,9 @@ INSTANTIATE_TEST_CASE_P(
{"hlsl.intrinsics.negative.comp", "ComputeShaderFunction"},
{"hlsl.intrinsics.negative.frag", "PixelShaderFunction"},
{"hlsl.intrinsics.negative.vert", "VertexShaderFunction"},
{"hlsl.intrinsics.promote.frag", "main"},
{"hlsl.intrinsics.promote.down.frag", "main"},
{"hlsl.intrinsics.promote.outputs.frag", "main"},
{"hlsl.layout.frag", "main"},
{"hlsl.load.2dms.dx10.frag", "main"},
{"hlsl.load.array.dx10.frag", "main"},

128
hlsl/hlslParseHelper.cpp
View File

@ -2555,7 +2555,7 @@ TIntermTyped* HlslParseContext::handleFunctionCall(const TSourceLoc& loc, TFunct
//
const TFunction* fnCandidate;
bool builtIn;
fnCandidate = findFunction(loc, *function, builtIn);
fnCandidate = findFunction(loc, *function, builtIn, arguments);
if (fnCandidate) {
// This is a declared function that might map to
// - a built-in operator,
@ -2597,21 +2597,27 @@ TIntermTyped* HlslParseContext::handleFunctionCall(const TSourceLoc& loc, TFunct
}
}
// for decompositions, since we want to operate on the function node, not the aggregate holding
// output conversions.
const TIntermTyped* fnNode = result;
decomposeIntrinsic(loc, result, arguments); // HLSL->AST intrinsic decompositions
decomposeSampleMethods(loc, result, arguments); // HLSL->AST sample method decompositions
decomposeGeometryMethods(loc, result, arguments); // HLSL->AST geometry method decompositions
// Convert 'out' arguments. If it was a constant folded built-in, it won't be an aggregate anymore.
// Built-ins with a single argument aren't called with an aggregate, but they also don't have an output.
// Also, build the qualifier list for user function calls, which are always called with an aggregate.
if (result->getAsAggregate()) {
// We don't do this is if there has been a decomposition, which will have added its own conversions
// for output parameters.
if (result == fnNode && result->getAsAggregate()) {
TQualifierList& qualifierList = result->getAsAggregate()->getQualifierList();
for (int i = 0; i < fnCandidate->getParamCount(); ++i) {
TStorageQualifier qual = (*fnCandidate)[i].type->getQualifier().storage;
qualifierList.push_back(qual);
}
result = addOutputArgumentConversions(*fnCandidate, *result->getAsAggregate());
result = addOutputArgumentConversions(*fnCandidate, *result->getAsOperator());
}
decomposeIntrinsic(loc, result, arguments); // HLSL->AST intrinsic decompositions
decomposeSampleMethods(loc, result, arguments); // HLSL->AST sample method decompositions
decomposeGeometryMethods(loc, result, arguments); // HLSL->AST geometry method decompositions
}
}
@ -2724,9 +2730,19 @@ void HlslParseContext::addInputArgumentConversions(const TFunction& function, TI
//
// Returns a node of a subtree that evaluates to the return value of the function.
//
TIntermTyped* HlslParseContext::addOutputArgumentConversions(const TFunction& function, TIntermAggregate& intermNode)
TIntermTyped* HlslParseContext::addOutputArgumentConversions(const TFunction& function, TIntermOperator& intermNode)
{
TIntermSequence& arguments = intermNode.getSequence();
assert (intermNode.getAsAggregate() != nullptr || intermNode.getAsUnaryNode() != nullptr);
const TSourceLoc& loc = intermNode.getLoc();
TIntermSequence argSequence; // temp sequence for unary node args
if (intermNode.getAsUnaryNode())
argSequence.push_back(intermNode.getAsUnaryNode()->getOperand());
TIntermSequence& arguments = argSequence.empty() ? intermNode.getAsAggregate()->getSequence() : argSequence;
const auto needsConversion = [&](int argNum) {
return function[argNum].type->getQualifier().isParamOutput() &&
(*function[argNum].type != arguments[argNum]->getAsTyped()->getType() ||
@ -2759,8 +2775,8 @@ TIntermTyped* HlslParseContext::addOutputArgumentConversions(const TFunction& fu
if (intermNode.getBasicType() != EbtVoid) {
// do the "tempRet = function(...), " bit from above
tempRet = makeInternalVariable("tempReturn", intermNode.getType());
TIntermSymbol* tempRetNode = intermediate.addSymbol(*tempRet, intermNode.getLoc());
conversionTree = intermediate.addAssign(EOpAssign, tempRetNode, &intermNode, intermNode.getLoc());
TIntermSymbol* tempRetNode = intermediate.addSymbol(*tempRet, loc);
conversionTree = intermediate.addAssign(EOpAssign, tempRetNode, &intermNode, loc);
} else
conversionTree = &intermNode;
@ -2775,7 +2791,7 @@ TIntermTyped* HlslParseContext::addOutputArgumentConversions(const TFunction& fu
// Make a temporary for what the function expects the argument to look like.
TVariable* tempArg = makeInternalVariable("tempArg", *function[i].type);
tempArg->getWritableType().getQualifier().makeTemporary();
TIntermSymbol* tempArgNode = intermediate.addSymbol(*tempArg, intermNode.getLoc());
TIntermSymbol* tempArgNode = intermediate.addSymbol(*tempArg, loc);
// This makes the deepest level, the member-wise copy
TIntermTyped* tempAssign = handleAssign(arguments[i]->getLoc(), EOpAssign, arguments[i]->getAsTyped(), tempArgNode);
@ -2783,17 +2799,18 @@ TIntermTyped* HlslParseContext::addOutputArgumentConversions(const TFunction& fu
conversionTree = intermediate.growAggregate(conversionTree, tempAssign, arguments[i]->getLoc());
// replace the argument with another node for the same tempArg variable
arguments[i] = intermediate.addSymbol(*tempArg, intermNode.getLoc());
arguments[i] = intermediate.addSymbol(*tempArg, loc);
}
}
// Finalize the tree topology (see bigger comment above).
if (tempRet) {
// do the "..., tempRet" bit from above
TIntermSymbol* tempRetNode = intermediate.addSymbol(*tempRet, intermNode.getLoc());
conversionTree = intermediate.growAggregate(conversionTree, tempRetNode, intermNode.getLoc());
TIntermSymbol* tempRetNode = intermediate.addSymbol(*tempRet, loc);
conversionTree = intermediate.growAggregate(conversionTree, tempRetNode, loc);
}
conversionTree = intermediate.setAggregateOperator(conversionTree, EOpComma, intermNode.getType(), intermNode.getLoc());
conversionTree = intermediate.setAggregateOperator(conversionTree, EOpComma, intermNode.getType(), loc);
return conversionTree;
}
@ -4339,7 +4356,8 @@ void HlslParseContext::mergeObjectLayoutQualifiers(TQualifier& dst, const TQuali
//
// Return the function symbol if found, otherwise nullptr.
//
const TFunction* HlslParseContext::findFunction(const TSourceLoc& loc, const TFunction& call, bool& builtIn)
const TFunction* HlslParseContext::findFunction(const TSourceLoc& loc, const TFunction& call, bool& builtIn,
TIntermNode* args)
{
// const TFunction* function = nullptr;
@ -4445,9 +4463,81 @@ const TFunction* HlslParseContext::findFunction(const TSourceLoc& loc, const TFu
// send to the generic selector
const TFunction* bestMatch = selectFunction(candidateList, call, convertible, better, tie);
if (bestMatch == nullptr)
if (bestMatch == nullptr) {
error(loc, "no matching overloaded function found", call.getName().c_str(), "");
else if (tie)
return nullptr;
}
// For builtins, we can convert across the arguments. This will happen in several steps:
// Step 1: If there's an exact match, use it.
// Step 2a: Otherwise, get the operator from the best match and promote arguments:
// Step 2b: reconstruct the TFunction based on the new arg types
// Step 3: Re-select after type promotion is applied, to find proper candidate.
if (builtIn) {
// Step 1: If there's an exact match, use it.
if (call.getMangledName() == bestMatch->getMangledName())
return bestMatch;
// Step 2a: Otherwise, get the operator from the best match and promote arguments as if we
// are that kind of operator.
if (args != nullptr) {
// The arg list can be a unary node, or an aggregate. We have to handle both.
// We will use the normal promote() facilities, which require an interm node.
TIntermOperator* promote = nullptr;
if (call.getParamCount() == 1) {
promote = new TIntermUnary(bestMatch->getBuiltInOp());
promote->getAsUnaryNode()->setOperand(args->getAsTyped());
} else {
promote = new TIntermAggregate(bestMatch->getBuiltInOp());
promote->getAsAggregate()->getSequence().swap(args->getAsAggregate()->getSequence());
}
if (! intermediate.promote(promote))
return nullptr;
// Obtain the promoted arg list.
if (call.getParamCount() == 1) {
args = promote->getAsUnaryNode()->getOperand();
} else {
promote->getAsAggregate()->getSequence().swap(args->getAsAggregate()->getSequence());
}
}
// Step 2b: reconstruct the TFunction based on the new arg types
TFunction convertedCall(&call.getName(), call.getType(), call.getBuiltInOp());
if (args->getAsAggregate()) {
// Handle aggregates: put all args into the new function call
for (int arg=0; arg<int(args->getAsAggregate()->getSequence().size()); ++arg) {
// TODO: But for constness, we could avoid the new & shallowCopy, and use the pointer directly.
TParameter param = { 0, new TType };
param.type->shallowCopy(args->getAsAggregate()->getSequence()[arg]->getAsTyped()->getType());
convertedCall.addParameter(param);
}
} else if (args->getAsUnaryNode()) {
// Handle unaries: put all args into the new function call
TParameter param = { 0, new TType };
param.type->shallowCopy(args->getAsUnaryNode()->getOperand()->getAsTyped()->getType());
convertedCall.addParameter(param);
} else if (args->getAsTyped()) {
// Handle bare e.g, floats, not in an aggregate.
TParameter param = { 0, new TType };
param.type->shallowCopy(args->getAsTyped()->getType());
convertedCall.addParameter(param);
} else {
assert(0); // unknown argument list.
return nullptr;
}
// Step 3: Re-select after type promotion, to find proper candidate
// send to the generic selector
bestMatch = selectFunction(candidateList, convertedCall, convertible, better, tie);
// At this point, there should be no tie.
}
if (tie)
error(loc, "ambiguous best function under implicit type conversion", call.getName().c_str(), "");
return bestMatch;

4
hlsl/hlslParseHelper.h
View File

@ -84,7 +84,7 @@ public:
void decomposeGeometryMethods(const TSourceLoc&, TIntermTyped*& node, TIntermNode* arguments);
TIntermTyped* handleLengthMethod(const TSourceLoc&, TFunction*, TIntermNode*);
void addInputArgumentConversions(const TFunction&, TIntermNode*&) const;
TIntermTyped* addOutputArgumentConversions(const TFunction&, TIntermAggregate&);
TIntermTyped* addOutputArgumentConversions(const TFunction&, TIntermOperator&);
void builtInOpCheck(const TSourceLoc&, const TFunction&, TIntermOperator&);
TFunction* handleConstructorCall(const TSourceLoc&, const TType&);
void handleSemantic(TSourceLoc, TQualifier&, const TString& semantic);
@ -125,7 +125,7 @@ public:
void mergeObjectLayoutQualifiers(TQualifier& dest, const TQualifier& src, bool inheritOnly);
void checkNoShaderLayouts(const TSourceLoc&, const TShaderQualifiers&);
const TFunction* findFunction(const TSourceLoc& loc, const TFunction& call, bool& builtIn);
const TFunction* findFunction(const TSourceLoc& loc, const TFunction& call, bool& builtIn, TIntermNode* args);
void declareTypedef(const TSourceLoc&, TString& identifier, const TType&, TArraySizes* typeArray = 0);
TIntermNode* declareVariable(const TSourceLoc&, TString& identifier, TType&, TIntermTyped* initializer = 0);
TIntermTyped* addConstructor(const TSourceLoc&, TIntermNode*, const TType&);

16
hlsl/hlslParseables.cpp
View File

@ -558,8 +558,8 @@ void TBuiltInParseablesHlsl::initialize(int /*version*/, EProfile /*profile*/, c
{ "AllMemoryBarrier", nullptr, nullptr, "-", "-", EShLangCS },
{ "AllMemoryBarrierWithGroupSync", nullptr, nullptr, "-", "-", EShLangCS },
{ "any", "S", "B", "SVM", "BFIU", EShLangAll },
{ "asdouble", "S", "D", "S,", "U,", EShLangAll },
{ "asdouble", "V2", "D", "V2,", "U,", EShLangAll },
{ "asdouble", "S", "D", "S,", "UI,", EShLangAll },
{ "asdouble", "V2", "D", "V2,", "UI,", EShLangAll },
{ "asfloat", nullptr, "F", "SVM", "BFIU", EShLangAll },
{ "asin", nullptr, nullptr, "SVM", "F", EShLangAll },
{ "asint", nullptr, "I", "SVM", "FU", EShLangAll },
@ -572,7 +572,7 @@ void TBuiltInParseablesHlsl::initialize(int /*version*/, EProfile /*profile*/, c
{ "clip", "-", "-", "SVM", "F", EShLangPS },
{ "cos", nullptr, nullptr, "SVM", "F", EShLangAll },
{ "cosh", nullptr, nullptr, "SVM", "F", EShLangAll },
{ "countbits", nullptr, nullptr, "SV", "U", EShLangAll },
{ "countbits", nullptr, nullptr, "SV", "UI", EShLangAll },
{ "cross", nullptr, nullptr, "V3,", "F,", EShLangAll },
{ "D3DCOLORtoUBYTE4", "V4", "I", "V4", "F", EShLangAll },
{ "ddx", nullptr, nullptr, "SVM", "F", EShLangPS },
@ -636,9 +636,9 @@ void TBuiltInParseablesHlsl::initialize(int /*version*/, EProfile /*profile*/, c
{ "log10", nullptr, nullptr, "SVM", "F", EShLangAll },
{ "log2", nullptr, nullptr, "SVM", "F", EShLangAll },
{ "mad", nullptr, nullptr, "SVM,,", "DFUI,,", EShLangAll },
{ "max", nullptr, nullptr, "SVM,", "FI,", EShLangAll },
{ "min", nullptr, nullptr, "SVM,", "FI,", EShLangAll },
{ "modf", nullptr, nullptr, "SVM,>", "FI,", EShLangAll },
{ "max", nullptr, nullptr, "SVM,", "FIU,", EShLangAll },
{ "min", nullptr, nullptr, "SVM,", "FIU,", EShLangAll },
{ "modf", nullptr, nullptr, "SVM,>", "FIU,", EShLangAll },
{ "msad4", "V4", "U", "S,V2,V4", "U,,", EShLangAll },
{ "mul", "S", nullptr, "S,S", "FI,", EShLangAll },
{ "mul", "V", nullptr, "S,V", "FI,", EShLangAll },
@ -665,7 +665,7 @@ void TBuiltInParseablesHlsl::initialize(int /*version*/, EProfile /*profile*/, c
{ "rcp", nullptr, nullptr, "SVM", "FD", EShLangAll },
{ "reflect", nullptr, nullptr, "V,", "F,", EShLangAll },
{ "refract", nullptr, nullptr, "V,V,S", "F,,", EShLangAll },
{ "reversebits", nullptr, nullptr, "SV", "U", EShLangAll },
{ "reversebits", nullptr, nullptr, "SV", "UI", EShLangAll },
{ "round", nullptr, nullptr, "SVM", "F", EShLangAll },
{ "rsqrt", nullptr, nullptr, "SVM", "F", EShLangAll },
{ "saturate", nullptr, nullptr , "SVM", "F", EShLangAll },
@ -735,7 +735,7 @@ void TBuiltInParseablesHlsl::initialize(int /*version*/, EProfile /*profile*/, c
// RWTexture loads
{ "Load", "V4", nullptr, "!#,V", "FIU,I", EShLangAll },
// (RW)Buffer loads
{ "Load", "V4", nullptr, "~*1,V", "FIU,I", EShLangAll },
{ "Load", "V4", nullptr, "~*1,V", "FIU,I", EShLangAll },
{ "Gather", /*!O*/ "V4", nullptr, "%@,S,V", "FIU,S,F", EShLangAll },
{ "Gather", /* O*/ "V4", nullptr, "%@,S,V,V", "FIU,S,F,I", EShLangAll },