Merge pull request #1161 from LoopDawg/matmul-truncate
WIP: HLSL: matrix and vector truncations for m*v, v*m, m*m
diff --git a/Test/baseResults/hlsl.mul-truncate.frag.out b/Test/baseResults/hlsl.mul-truncate.frag.out
new file mode 100644
index 0000000..a1c3c03
--- /dev/null
+++ b/Test/baseResults/hlsl.mul-truncate.frag.out
@@ -0,0 +1,413 @@
+hlsl.mul-truncate.frag
+Shader version: 500
+gl_FragCoord origin is upper left
+0:? Sequence
+0:17 Function Definition: @main( ( temp 4-component vector of float)
+0:17 Function Parameters:
+0:? Sequence
+0:19 Sequence
+0:19 move second child to first child ( temp float)
+0:19 'r00' ( temp float)
+0:19 dot-product ( temp float)
+0:19 v2: direct index for structure (layout( row_major std140) uniform 2-component vector of float)
+0:19 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:19 Constant:
+0:19 7 (const uint)
+0:19 Construct vec2 ( in 2-component vector of float)
+0:19 v3: direct index for structure (layout( row_major std140) uniform 3-component vector of float)
+0:19 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:19 Constant:
+0:19 6 (const uint)
+0:20 Sequence
+0:20 move second child to first child ( temp float)
+0:20 'r01' ( temp float)
+0:20 dot-product ( temp float)
+0:20 Construct vec2 ( in 2-component vector of float)
+0:20 v4: direct index for structure (layout( row_major std140) uniform 4-component vector of float)
+0:20 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:20 Constant:
+0:20 5 (const uint)
+0:20 v2: direct index for structure (layout( row_major std140) uniform 2-component vector of float)
+0:20 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:20 Constant:
+0:20 7 (const uint)
+0:23 Sequence
+0:23 move second child to first child ( temp 4-component vector of float)
+0:23 'r10' ( temp 4-component vector of float)
+0:23 matrix-times-vector ( temp 4-component vector of float)
+0:23 Construct mat3x4 ( uniform 3X4 matrix of float)
+0:23 m44: direct index for structure (layout( row_major std140) uniform 4X4 matrix of float)
+0:23 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:23 Constant:
+0:23 0 (const uint)
+0:23 v3: direct index for structure (layout( row_major std140) uniform 3-component vector of float)
+0:23 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:23 Constant:
+0:23 6 (const uint)
+0:24 Sequence
+0:24 move second child to first child ( temp 4-component vector of float)
+0:24 'r11' ( temp 4-component vector of float)
+0:24 matrix-times-vector ( temp 4-component vector of float)
+0:24 m34: direct index for structure (layout( row_major std140) uniform 3X4 matrix of float)
+0:24 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:24 Constant:
+0:24 2 (const uint)
+0:24 Construct vec3 ( uniform 3-component vector of float)
+0:24 v4: direct index for structure (layout( row_major std140) uniform 4-component vector of float)
+0:24 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:24 Constant:
+0:24 5 (const uint)
+0:27 Sequence
+0:27 move second child to first child ( temp 4-component vector of float)
+0:27 'r20' ( temp 4-component vector of float)
+0:27 vector-times-matrix ( temp 4-component vector of float)
+0:27 v3: direct index for structure (layout( row_major std140) uniform 3-component vector of float)
+0:27 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:27 Constant:
+0:27 6 (const uint)
+0:27 Construct mat4x3 ( uniform 4X3 matrix of float)
+0:27 m44: direct index for structure (layout( row_major std140) uniform 4X4 matrix of float)
+0:27 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:27 Constant:
+0:27 0 (const uint)
+0:28 Sequence
+0:28 move second child to first child ( temp 4-component vector of float)
+0:28 'r21' ( temp 4-component vector of float)
+0:28 vector-times-matrix ( temp 4-component vector of float)
+0:28 Construct vec3 ( uniform 3-component vector of float)
+0:28 v4: direct index for structure (layout( row_major std140) uniform 4-component vector of float)
+0:28 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:28 Constant:
+0:28 5 (const uint)
+0:28 m43: direct index for structure (layout( row_major std140) uniform 4X3 matrix of float)
+0:28 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:28 Constant:
+0:28 1 (const uint)
+0:36 Branch: Return with expression
+0:36 add ( temp 4-component vector of float)
+0:36 add ( temp 4-component vector of float)
+0:36 add ( temp 4-component vector of float)
+0:36 add ( temp 4-component vector of float)
+0:36 add ( temp 4-component vector of float)
+0:36 'r10' ( temp 4-component vector of float)
+0:36 'r11' ( temp 4-component vector of float)
+0:36 'r20' ( temp 4-component vector of float)
+0:36 'r21' ( temp 4-component vector of float)
+0:36 'r00' ( temp float)
+0:36 'r01' ( temp float)
+0:17 Function Definition: main( ( temp void)
+0:17 Function Parameters:
+0:? Sequence
+0:17 move second child to first child ( temp 4-component vector of float)
+0:? '@entryPointOutput' (layout( location=0) out 4-component vector of float)
+0:17 Function Call: @main( ( temp 4-component vector of float)
+0:? Linker Objects
+0:? 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:? '@entryPointOutput' (layout( location=0) out 4-component vector of float)
+
+
+Linked fragment stage:
+
+
+Shader version: 500
+gl_FragCoord origin is upper left
+0:? Sequence
+0:17 Function Definition: @main( ( temp 4-component vector of float)
+0:17 Function Parameters:
+0:? Sequence
+0:19 Sequence
+0:19 move second child to first child ( temp float)
+0:19 'r00' ( temp float)
+0:19 dot-product ( temp float)
+0:19 v2: direct index for structure (layout( row_major std140) uniform 2-component vector of float)
+0:19 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:19 Constant:
+0:19 7 (const uint)
+0:19 Construct vec2 ( in 2-component vector of float)
+0:19 v3: direct index for structure (layout( row_major std140) uniform 3-component vector of float)
+0:19 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:19 Constant:
+0:19 6 (const uint)
+0:20 Sequence
+0:20 move second child to first child ( temp float)
+0:20 'r01' ( temp float)
+0:20 dot-product ( temp float)
+0:20 Construct vec2 ( in 2-component vector of float)
+0:20 v4: direct index for structure (layout( row_major std140) uniform 4-component vector of float)
+0:20 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:20 Constant:
+0:20 5 (const uint)
+0:20 v2: direct index for structure (layout( row_major std140) uniform 2-component vector of float)
+0:20 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:20 Constant:
+0:20 7 (const uint)
+0:23 Sequence
+0:23 move second child to first child ( temp 4-component vector of float)
+0:23 'r10' ( temp 4-component vector of float)
+0:23 matrix-times-vector ( temp 4-component vector of float)
+0:23 Construct mat3x4 ( uniform 3X4 matrix of float)
+0:23 m44: direct index for structure (layout( row_major std140) uniform 4X4 matrix of float)
+0:23 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:23 Constant:
+0:23 0 (const uint)
+0:23 v3: direct index for structure (layout( row_major std140) uniform 3-component vector of float)
+0:23 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:23 Constant:
+0:23 6 (const uint)
+0:24 Sequence
+0:24 move second child to first child ( temp 4-component vector of float)
+0:24 'r11' ( temp 4-component vector of float)
+0:24 matrix-times-vector ( temp 4-component vector of float)
+0:24 m34: direct index for structure (layout( row_major std140) uniform 3X4 matrix of float)
+0:24 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:24 Constant:
+0:24 2 (const uint)
+0:24 Construct vec3 ( uniform 3-component vector of float)
+0:24 v4: direct index for structure (layout( row_major std140) uniform 4-component vector of float)
+0:24 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:24 Constant:
+0:24 5 (const uint)
+0:27 Sequence
+0:27 move second child to first child ( temp 4-component vector of float)
+0:27 'r20' ( temp 4-component vector of float)
+0:27 vector-times-matrix ( temp 4-component vector of float)
+0:27 v3: direct index for structure (layout( row_major std140) uniform 3-component vector of float)
+0:27 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:27 Constant:
+0:27 6 (const uint)
+0:27 Construct mat4x3 ( uniform 4X3 matrix of float)
+0:27 m44: direct index for structure (layout( row_major std140) uniform 4X4 matrix of float)
+0:27 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:27 Constant:
+0:27 0 (const uint)
+0:28 Sequence
+0:28 move second child to first child ( temp 4-component vector of float)
+0:28 'r21' ( temp 4-component vector of float)
+0:28 vector-times-matrix ( temp 4-component vector of float)
+0:28 Construct vec3 ( uniform 3-component vector of float)
+0:28 v4: direct index for structure (layout( row_major std140) uniform 4-component vector of float)
+0:28 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:28 Constant:
+0:28 5 (const uint)
+0:28 m43: direct index for structure (layout( row_major std140) uniform 4X3 matrix of float)
+0:28 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:28 Constant:
+0:28 1 (const uint)
+0:36 Branch: Return with expression
+0:36 add ( temp 4-component vector of float)
+0:36 add ( temp 4-component vector of float)
+0:36 add ( temp 4-component vector of float)
+0:36 add ( temp 4-component vector of float)
+0:36 add ( temp 4-component vector of float)
+0:36 'r10' ( temp 4-component vector of float)
+0:36 'r11' ( temp 4-component vector of float)
+0:36 'r20' ( temp 4-component vector of float)
+0:36 'r21' ( temp 4-component vector of float)
+0:36 'r00' ( temp float)
+0:36 'r01' ( temp float)
+0:17 Function Definition: main( ( temp void)
+0:17 Function Parameters:
+0:? Sequence
+0:17 move second child to first child ( temp 4-component vector of float)
+0:? '@entryPointOutput' (layout( location=0) out 4-component vector of float)
+0:17 Function Call: @main( ( temp 4-component vector of float)
+0:? Linker Objects
+0:? 'anon@0' (layout( row_major std140) uniform block{layout( row_major std140) uniform 4X4 matrix of float m44, layout( row_major std140) uniform 4X3 matrix of float m43, layout( row_major std140) uniform 3X4 matrix of float m34, layout( row_major std140) uniform 2X4 matrix of float m24, layout( row_major std140) uniform 4X2 matrix of float m42, layout( row_major std140) uniform 4-component vector of float v4, layout( row_major std140) uniform 3-component vector of float v3, layout( row_major std140) uniform 2-component vector of float v2})
+0:? '@entryPointOutput' (layout( location=0) out 4-component vector of float)
+
+// Module Version 10000
+// Generated by (magic number): 80002
+// Id's are bound by 139
+
+ Capability Shader
+ 1: ExtInstImport "GLSL.std.450"
+ MemoryModel Logical GLSL450
+ EntryPoint Fragment 4 "main" 137
+ ExecutionMode 4 OriginUpperLeft
+ Source HLSL 500
+ Name 4 "main"
+ Name 9 "@main("
+ Name 12 "r00"
+ Name 20 "Matrix"
+ MemberName 20(Matrix) 0 "m44"
+ MemberName 20(Matrix) 1 "m43"
+ MemberName 20(Matrix) 2 "m34"
+ MemberName 20(Matrix) 3 "m24"
+ MemberName 20(Matrix) 4 "m42"
+ MemberName 20(Matrix) 5 "v4"
+ MemberName 20(Matrix) 6 "v3"
+ MemberName 20(Matrix) 7 "v2"
+ Name 22 ""
+ Name 36 "r01"
+ Name 48 "r10"
+ Name 74 "r11"
+ Name 86 "r20"
+ Name 109 "r21"
+ Name 137 "@entryPointOutput"
+ MemberDecorate 20(Matrix) 0 RowMajor
+ MemberDecorate 20(Matrix) 0 Offset 0
+ MemberDecorate 20(Matrix) 0 MatrixStride 16
+ MemberDecorate 20(Matrix) 1 RowMajor
+ MemberDecorate 20(Matrix) 1 Offset 64
+ MemberDecorate 20(Matrix) 1 MatrixStride 16
+ MemberDecorate 20(Matrix) 2 RowMajor
+ MemberDecorate 20(Matrix) 2 Offset 112
+ MemberDecorate 20(Matrix) 2 MatrixStride 16
+ MemberDecorate 20(Matrix) 3 RowMajor
+ MemberDecorate 20(Matrix) 3 Offset 176
+ MemberDecorate 20(Matrix) 3 MatrixStride 16
+ MemberDecorate 20(Matrix) 4 RowMajor
+ MemberDecorate 20(Matrix) 4 Offset 240
+ MemberDecorate 20(Matrix) 4 MatrixStride 16
+ MemberDecorate 20(Matrix) 5 Offset 272
+ MemberDecorate 20(Matrix) 6 Offset 288
+ MemberDecorate 20(Matrix) 7 Offset 304
+ Decorate 20(Matrix) Block
+ Decorate 22 DescriptorSet 0
+ Decorate 137(@entryPointOutput) Location 0
+ 2: TypeVoid
+ 3: TypeFunction 2
+ 6: TypeFloat 32
+ 7: TypeVector 6(float) 4
+ 8: TypeFunction 7(fvec4)
+ 11: TypePointer Function 6(float)
+ 13: TypeMatrix 7(fvec4) 4
+ 14: TypeVector 6(float) 3
+ 15: TypeMatrix 14(fvec3) 4
+ 16: TypeMatrix 7(fvec4) 3
+ 17: TypeMatrix 7(fvec4) 2
+ 18: TypeVector 6(float) 2
+ 19: TypeMatrix 18(fvec2) 4
+ 20(Matrix): TypeStruct 13 15 16 17 19 7(fvec4) 14(fvec3) 18(fvec2)
+ 21: TypePointer Uniform 20(Matrix)
+ 22: 21(ptr) Variable Uniform
+ 23: TypeInt 32 1
+ 24: 23(int) Constant 7
+ 25: TypePointer Uniform 18(fvec2)
+ 28: 23(int) Constant 6
+ 29: TypePointer Uniform 14(fvec3)
+ 37: 23(int) Constant 5
+ 38: TypePointer Uniform 7(fvec4)
+ 47: TypePointer Function 7(fvec4)
+ 49: 23(int) Constant 0
+ 50: TypePointer Uniform 13
+ 53: 6(float) Constant 1065353216
+ 54: 6(float) Constant 0
+ 75: 23(int) Constant 2
+ 76: TypePointer Uniform 16
+ 116: 23(int) Constant 1
+ 117: TypePointer Uniform 15
+ 136: TypePointer Output 7(fvec4)
+137(@entryPointOutput): 136(ptr) Variable Output
+ 4(main): 2 Function None 3
+ 5: Label
+ 138: 7(fvec4) FunctionCall 9(@main()
+ Store 137(@entryPointOutput) 138
+ Return
+ FunctionEnd
+ 9(@main(): 7(fvec4) Function None 8
+ 10: Label
+ 12(r00): 11(ptr) Variable Function
+ 36(r01): 11(ptr) Variable Function
+ 48(r10): 47(ptr) Variable Function
+ 74(r11): 47(ptr) Variable Function
+ 86(r20): 47(ptr) Variable Function
+ 109(r21): 47(ptr) Variable Function
+ 26: 25(ptr) AccessChain 22 24
+ 27: 18(fvec2) Load 26
+ 30: 29(ptr) AccessChain 22 28
+ 31: 14(fvec3) Load 30
+ 32: 6(float) CompositeExtract 31 0
+ 33: 6(float) CompositeExtract 31 1
+ 34: 18(fvec2) CompositeConstruct 32 33
+ 35: 6(float) Dot 27 34
+ Store 12(r00) 35
+ 39: 38(ptr) AccessChain 22 37
+ 40: 7(fvec4) Load 39
+ 41: 6(float) CompositeExtract 40 0
+ 42: 6(float) CompositeExtract 40 1
+ 43: 18(fvec2) CompositeConstruct 41 42
+ 44: 25(ptr) AccessChain 22 24
+ 45: 18(fvec2) Load 44
+ 46: 6(float) Dot 43 45
+ Store 36(r01) 46
+ 51: 50(ptr) AccessChain 22 49
+ 52: 13 Load 51
+ 55: 6(float) CompositeExtract 52 0 0
+ 56: 6(float) CompositeExtract 52 0 1
+ 57: 6(float) CompositeExtract 52 0 2
+ 58: 6(float) CompositeExtract 52 0 3
+ 59: 6(float) CompositeExtract 52 1 0
+ 60: 6(float) CompositeExtract 52 1 1
+ 61: 6(float) CompositeExtract 52 1 2
+ 62: 6(float) CompositeExtract 52 1 3
+ 63: 6(float) CompositeExtract 52 2 0
+ 64: 6(float) CompositeExtract 52 2 1
+ 65: 6(float) CompositeExtract 52 2 2
+ 66: 6(float) CompositeExtract 52 2 3
+ 67: 7(fvec4) CompositeConstruct 55 56 57 58
+ 68: 7(fvec4) CompositeConstruct 59 60 61 62
+ 69: 7(fvec4) CompositeConstruct 63 64 65 66
+ 70: 16 CompositeConstruct 67 68 69
+ 71: 29(ptr) AccessChain 22 28
+ 72: 14(fvec3) Load 71
+ 73: 7(fvec4) MatrixTimesVector 70 72
+ Store 48(r10) 73
+ 77: 76(ptr) AccessChain 22 75
+ 78: 16 Load 77
+ 79: 38(ptr) AccessChain 22 37
+ 80: 7(fvec4) Load 79
+ 81: 6(float) CompositeExtract 80 0
+ 82: 6(float) CompositeExtract 80 1
+ 83: 6(float) CompositeExtract 80 2
+ 84: 14(fvec3) CompositeConstruct 81 82 83
+ 85: 7(fvec4) MatrixTimesVector 78 84
+ Store 74(r11) 85
+ 87: 29(ptr) AccessChain 22 28
+ 88: 14(fvec3) Load 87
+ 89: 50(ptr) AccessChain 22 49
+ 90: 13 Load 89
+ 91: 6(float) CompositeExtract 90 0 0
+ 92: 6(float) CompositeExtract 90 0 1
+ 93: 6(float) CompositeExtract 90 0 2
+ 94: 6(float) CompositeExtract 90 1 0
+ 95: 6(float) CompositeExtract 90 1 1
+ 96: 6(float) CompositeExtract 90 1 2
+ 97: 6(float) CompositeExtract 90 2 0
+ 98: 6(float) CompositeExtract 90 2 1
+ 99: 6(float) CompositeExtract 90 2 2
+ 100: 6(float) CompositeExtract 90 3 0
+ 101: 6(float) CompositeExtract 90 3 1
+ 102: 6(float) CompositeExtract 90 3 2
+ 103: 14(fvec3) CompositeConstruct 91 92 93
+ 104: 14(fvec3) CompositeConstruct 94 95 96
+ 105: 14(fvec3) CompositeConstruct 97 98 99
+ 106: 14(fvec3) CompositeConstruct 100 101 102
+ 107: 15 CompositeConstruct 103 104 105 106
+ 108: 7(fvec4) VectorTimesMatrix 88 107
+ Store 86(r20) 108
+ 110: 38(ptr) AccessChain 22 37
+ 111: 7(fvec4) Load 110
+ 112: 6(float) CompositeExtract 111 0
+ 113: 6(float) CompositeExtract 111 1
+ 114: 6(float) CompositeExtract 111 2
+ 115: 14(fvec3) CompositeConstruct 112 113 114
+ 118: 117(ptr) AccessChain 22 116
+ 119: 15 Load 118
+ 120: 7(fvec4) VectorTimesMatrix 115 119
+ Store 109(r21) 120
+ 121: 7(fvec4) Load 48(r10)
+ 122: 7(fvec4) Load 74(r11)
+ 123: 7(fvec4) FAdd 121 122
+ 124: 7(fvec4) Load 86(r20)
+ 125: 7(fvec4) FAdd 123 124
+ 126: 7(fvec4) Load 109(r21)
+ 127: 7(fvec4) FAdd 125 126
+ 128: 6(float) Load 12(r00)
+ 129: 7(fvec4) CompositeConstruct 128 128 128 128
+ 130: 7(fvec4) FAdd 127 129
+ 131: 6(float) Load 36(r01)
+ 132: 7(fvec4) CompositeConstruct 131 131 131 131
+ 133: 7(fvec4) FAdd 130 132
+ ReturnValue 133
+ FunctionEnd
diff --git a/Test/hlsl.mul-truncate.frag b/Test/hlsl.mul-truncate.frag
new file mode 100644
index 0000000..6baf2ac
--- /dev/null
+++ b/Test/hlsl.mul-truncate.frag
@@ -0,0 +1,37 @@
+
+// Test v*v, v*m, m*v, and m*m argument clamping.
+
+cbuffer Matrix
+{
+ float4x4 m44;
+ float4x3 m43;
+ float3x4 m34;
+ float2x4 m24;
+ float4x2 m42;
+ float4 v4;
+ float3 v3;
+ float2 v2;
+}
+
+float4 main() : SV_Target0
+{
+ // v*v:
+ float r00 = mul(v2, v3); // float = float2*float3; // clamp to float2 dot product
+ float r01 = mul(v4, v2); // float = float4*float2; // clamp to float2 dot product
+
+ // v*m
+ float4 r10 = mul(v3, m44); // float4 = float3 * float4x4; // clamp mat to float3x4;
+ float4 r11 = mul(v4, m34); // truncate vector to vec3
+
+ // m*v
+ float4 r20 = mul(m44, v3); // float4 = float4x4 * float3; // clamp mat to float4x3;
+ float4 r21 = mul(m43, v4); // truncate vector to vec3
+
+ // // m*m
+ // float2x3 r30 = mul(m24, m33); // float2x3 = float2x4 * float3x3;
+ // float3x4 r31 = mul(m33, m24); // float3x4 = float3x3 * float2x4;
+ // float3x2 r32 = mul(m33, m42); // float3x2 = float3x3 * float4x2;
+ // float4x3 r33 = mul(m42, m33); // float4x3 = float4x2 * float3x3;
+
+ return r10 + r11 + r20 + r21 + r00 + r01; // + r30[0].x + r31[0] + r32[0].x + transpose(r33)[0];
+}
diff --git a/glslang/Include/Types.h b/glslang/Include/Types.h
index b5b91f5..d23b615 100644
--- a/glslang/Include/Types.h
+++ b/glslang/Include/Types.h
@@ -1152,6 +1152,7 @@
sampler.clear();
qualifier.clear();
qualifier.storage = q;
+ assert(!(isMatrix() && vectorSize != 0)); // prevent vectorSize != 0 on matrices
}
// for explicit precision qualifier
TType(TBasicType t, TStorageQualifier q, TPrecisionQualifier p, int vs = 1, int mc = 0, int mr = 0,
@@ -1164,6 +1165,7 @@
qualifier.storage = q;
qualifier.precision = p;
assert(p >= EpqNone && p <= EpqHigh);
+ assert(!(isMatrix() && vectorSize != 0)); // prevent vectorSize != 0 on matrices
}
// for turning a TPublicType into a TType, using a shallow copy
explicit TType(const TPublicType& p) :
diff --git a/gtests/Hlsl.FromFile.cpp b/gtests/Hlsl.FromFile.cpp
index 2e4b146..b409d81 100644
--- a/gtests/Hlsl.FromFile.cpp
+++ b/gtests/Hlsl.FromFile.cpp
@@ -238,6 +238,7 @@
{"hlsl.matrixSwizzle.vert", "ShaderFunction"},
{"hlsl.memberFunCall.frag", "main"},
{"hlsl.mintypes.frag", "main"},
+ {"hlsl.mul-truncate.frag", "main"},
{"hlsl.multiEntry.vert", "RealEntrypoint"},
{"hlsl.multiReturn.frag", "main"},
{"hlsl.matrixindex.frag", "main"},
diff --git a/hlsl/hlslParseHelper.cpp b/hlsl/hlslParseHelper.cpp
index adb955f..ce9afc7 100755
--- a/hlsl/hlslParseHelper.cpp
+++ b/hlsl/hlslParseHelper.cpp
@@ -5008,6 +5008,12 @@
bool builtIn = false;
int thisDepth = 0;
+ // For mat mul, the situation is unusual: we have to compare vector sizes to mat row or col sizes,
+ // and clamp the opposite arg. Since that's complex, we farm it off to a separate method.
+ // It doesn't naturally fall out of processing an argument at a time in isolation.
+ if (function->getName() == "mul")
+ addGenMulArgumentConversion(loc, *function, arguments);
+
TIntermAggregate* aggregate = arguments ? arguments->getAsAggregate() : nullptr;
// TODO: this needs improvement: there's no way at present to look up a signature in
@@ -5170,6 +5176,68 @@
}
//
+// HLSL allows mismatched dimensions on vec*mat, mat*vec, vec*vec, and mat*mat. This is a
+// situation not well suited to resolution in intrinsic selection, but we can do so here, since we
+// can look at both arguments insert explicit shape changes here, if required.
+//
+void HlslParseContext::addGenMulArgumentConversion(const TSourceLoc& loc, TFunction& call, TIntermTyped*& args)
+{
+ TIntermAggregate* argAggregate = args ? args->getAsAggregate() : nullptr;
+
+ if (argAggregate == nullptr || argAggregate->getSequence().size() != 2) {
+ // It really ought to have two arguments.
+ error(loc, "expected: mul arguments", "", "");
+ return;
+ }
+
+ TIntermTyped* arg0 = argAggregate->getSequence()[0]->getAsTyped();
+ TIntermTyped* arg1 = argAggregate->getSequence()[1]->getAsTyped();
+
+ if (arg0->isVector() && arg1->isVector()) {
+ // For:
+ // vec * vec: it's handled during intrinsic selection, so while we could do it here,
+ // we can also ignore it, which is easier.
+ } else if (arg0->isVector() && arg1->isMatrix()) {
+ // vec * mat: we clamp the vec if the mat col is smaller, else clamp the mat col.
+ if (arg0->getVectorSize() < arg1->getMatrixCols()) {
+ // vec is smaller, so truncate larger mat dimension
+ const TType truncType(arg1->getBasicType(), arg1->getQualifier().storage, arg1->getQualifier().precision,
+ 0, arg0->getVectorSize(), arg1->getMatrixRows());
+ arg1 = addConstructor(loc, arg1, truncType);
+ } else if (arg0->getVectorSize() > arg1->getMatrixCols()) {
+ // vec is larger, so truncate vec to mat size
+ const TType truncType(arg0->getBasicType(), arg0->getQualifier().storage, arg0->getQualifier().precision,
+ arg1->getMatrixCols());
+ arg0 = addConstructor(loc, arg0, truncType);
+ }
+ } else if (arg0->isMatrix() && arg1->isVector()) {
+ // mat * vec: we clamp the vec if the mat col is smaller, else clamp the mat col.
+ if (arg1->getVectorSize() < arg0->getMatrixRows()) {
+ // vec is smaller, so truncate larger mat dimension
+ const TType truncType(arg0->getBasicType(), arg0->getQualifier().storage, arg0->getQualifier().precision,
+ 0, arg0->getMatrixCols(), arg1->getVectorSize());
+ arg0 = addConstructor(loc, arg0, truncType);
+ } else if (arg1->getVectorSize() > arg0->getMatrixRows()) {
+ // vec is larger, so truncate vec to mat size
+ const TType truncType(arg1->getBasicType(), arg1->getQualifier().storage, arg1->getQualifier().precision,
+ arg0->getMatrixRows());
+ arg1 = addConstructor(loc, arg1, truncType);
+ }
+ } else if (arg0->isMatrix() && arg1->isMatrix()) {
+ // mat * mat
+ } else {
+ // It's something with scalars: we'll just leave it alone.
+ }
+
+ // Put arguments back.
+ argAggregate->getSequence()[0] = arg0;
+ argAggregate->getSequence()[1] = arg1;
+
+ call[0].type = &arg0->getWritableType();
+ call[1].type = &arg1->getWritableType();
+}
+
+//
// Add any needed implicit conversions for function-call arguments to input parameters.
//
void HlslParseContext::addInputArgumentConversions(const TFunction& function, TIntermTyped*& arguments)
@@ -7015,6 +7083,7 @@
}
}
+
//
// Look up a function name in the symbol table, and make sure it is a function.
//
diff --git a/hlsl/hlslParseHelper.h b/hlsl/hlslParseHelper.h
index c1752f1..67d192d 100755
--- a/hlsl/hlslParseHelper.h
+++ b/hlsl/hlslParseHelper.h
@@ -141,6 +141,7 @@
void checkNoShaderLayouts(const TSourceLoc&, const TShaderQualifiers&);
const TFunction* findFunction(const TSourceLoc& loc, TFunction& call, bool& builtIn, int& thisDepth, TIntermTyped*& args);
+ void addGenMulArgumentConversion(const TSourceLoc& loc, TFunction& call, TIntermTyped*& args);
void declareTypedef(const TSourceLoc&, const TString& identifier, const TType&);
void declareStruct(const TSourceLoc&, TString& structName, TType&);
TSymbol* lookupUserType(const TString&, TType&);
