Merge branch GitHub 'master' into GitLab master

StandAlone/StandAlone.cpp

@@ -827,7 +827,7 @@ int C_DECL main(int argc, char* argv[])
         // Print out all the resulting infologs
         for (int w = 0; w < NumWorkItems; ++w) {
             if (Work[w]) {
-                if (printShaderNames)
+                if (printShaderNames || Work[w]->results.size() > 0)
                     PutsIfNonEmpty(Work[w]->name.c_str());
                 PutsIfNonEmpty(Work[w]->results.c_str());
                 delete Work[w];

Test/baseResults/100LimitsConf.vert.out

+100Limits.vert
 ERROR: 0:24: 'limitation' : while loops not available 
 ERROR: 0:26: 'limitation' : do-while loops not available 
 ERROR: 0:28: 'limitations' : inductive-loop init-declaration requires the form "type-specifier loop-index = constant-expression" 

Test/baseResults/constFold.frag.out

@@ -257,6 +257,19 @@ ERROR: node is still EOpNull!
 0:120                1.000000
 0:120          Constant:
 0:120            3 (const int)
+0:126  Function Definition: foo3( (global void)
+0:126    Function Parameters: 
+0:128    Sequence
+0:128      Sequence
+0:128        move second child to first child (temp 3X2 matrix of float)
+0:128          'r32' (temp 3X2 matrix of float)
+0:128          Constant:
+0:128            43.000000
+0:128            64.000000
+0:128            51.000000
+0:128            76.000000
+0:128            59.000000
+0:128            88.000000
 0:?   Linker Objects
 0:?     'a' (const int)
 0:?       1 (const int)
@@ -331,6 +344,18 @@ ERROR: node is still EOpNull!
 0:?       4.000000
 0:?       5.000000
 0:?     'a4' (global 2-element array of float)
+0:?     'mm2' (const 2X2 matrix of float)
+0:?       1.000000
+0:?       2.000000
+0:?       3.000000
+0:?       4.000000
+0:?     'mm32' (const 3X2 matrix of float)
+0:?       10.000000
+0:?       11.000000
+0:?       12.000000
+0:?       13.000000
+0:?       14.000000
+0:?       15.000000
 
 
 Linked fragment stage:
@@ -584,6 +609,19 @@ ERROR: node is still EOpNull!
 0:120                1.000000
 0:120          Constant:
 0:120            3 (const int)
+0:126  Function Definition: foo3( (global void)
+0:126    Function Parameters: 
+0:128    Sequence
+0:128      Sequence
+0:128        move second child to first child (temp 3X2 matrix of float)
+0:128          'r32' (temp 3X2 matrix of float)
+0:128          Constant:
+0:128            43.000000
+0:128            64.000000
+0:128            51.000000
+0:128            76.000000
+0:128            59.000000
+0:128            88.000000
 0:?   Linker Objects
 0:?     'a' (const int)
 0:?       1 (const int)
@@ -658,4 +696,16 @@ ERROR: node is still EOpNull!
 0:?       4.000000
 0:?       5.000000
 0:?     'a4' (global 2-element array of float)
+0:?     'mm2' (const 2X2 matrix of float)
+0:?       1.000000
+0:?       2.000000
+0:?       3.000000
+0:?       4.000000
+0:?     'mm32' (const 3X2 matrix of float)
+0:?       10.000000
+0:?       11.000000
+0:?       12.000000
+0:?       13.000000
+0:?       14.000000
+0:?       15.000000
 

Test/baseResults/spv.image.frag.out

@@ -333,7 +333,7 @@ Linked fragment stage:
              214:          96 Load 98(i2DMS)
              215:   29(ivec2) Load 142(ic2D)
              217:  125(fvec4) Load 127(v)
-                              ImageWrite 214 215 216
+                              ImageWrite 214 215 217 Sample 216
              218:         106 Load 108(i2DMSArray)
              219:    7(ivec3) Load 152(ic3D)
              221:  125(fvec4) ImageRead 218 219 Sample 220
@@ -343,7 +343,7 @@ Linked fragment stage:
              224:         106 Load 108(i2DMSArray)
              225:    7(ivec3) Load 152(ic3D)
              227:  125(fvec4) Load 127(v)
-                              ImageWrite 224 225 226
+                              ImageWrite 224 225 227 Sample 226
                               Store 229(ui) 19
              233:      6(int) Load 132(ic1D)
              236:    235(ptr) ImageTexelPointer 232(ii1D) 233 0

Test/constFold.frag

@@ -119,3 +119,11 @@ void foo2()
     float f = vec4(7.8 < 2.4 ? -1.333 : 1.444).a;
     f = vec4(inv.x < 2.4 ? -1.0 : 1.0).a;  // not folded, ensuring no propagation
 }
+
+const mat2 mm2 = mat2(1.0, 2.0, 3.0, 4.0);
+const mat3x2 mm32 = mat3x2(10.0, 11.0, 12.0, 13.0, 14.0, 15.0);
+
+void foo3()
+{
+    mat3x2 r32 = mm2 * mm32;
+}

Test/spv.bitCast.frag

+#version 450
+
+uniform int   i1;
+uniform ivec2 i2;
+uniform ivec3 i3;
+uniform ivec4 i4;
+
+uniform uint  u1;
+uniform uvec2 u2;
+uniform uvec3 u3;
+uniform uvec4 u4;
+
+uniform float f1;
+uniform vec2  f2;
+uniform vec3  f3;
+uniform vec4  f4;
+
+out vec4 fragColor;
+
+void main()
+{
+	ivec4 idata = ivec4(0);
+	idata.x     += floatBitsToInt(f1);
+	idata.xy    += floatBitsToInt(f2);
+	idata.xyz   += floatBitsToInt(f3);
+	idata       += floatBitsToInt(f4);
+
+	uvec4 udata = uvec4(0);
+	udata.x     += floatBitsToUint(f1);
+	udata.xy    += floatBitsToUint(f2);
+	udata.xyz   += floatBitsToUint(f3);
+	udata       += floatBitsToUint(f4);
+
+	vec4 fdata = vec4(0.0);
+	fdata.x     += intBitsToFloat(i1);
+	fdata.xy    += intBitsToFloat(i2);
+	fdata.xyz   += intBitsToFloat(i3);
+	fdata       += intBitsToFloat(i4);
+    fdata.x     += uintBitsToFloat(u1);
+	fdata.xy    += uintBitsToFloat(u2);
+	fdata.xyz   += uintBitsToFloat(u3);
+	fdata       += uintBitsToFloat(u4);
+
+    fragColor = (idata == udata) ? fdata : fdata + vec4(0.2);
+}
\ No newline at end of file

Test/spv.layoutNested.vert

+#version 450
+
+// should get 3 SPV types for S: no layout, 140, and 430
+struct S
+{
+	highp uvec3 a;
+	mediump mat2 b[4];
+	lowp uint c;
+};
+
+layout(set = 0, binding = 0, std140) uniform Block140
+{
+	mediump int u;
+	S s[2][3];
+	mediump vec2 v;
+} inst140;
+
+layout(set = 0, binding = 1, std430) buffer Block430
+{
+	mediump int u;
+	S s[2][3];
+	mediump vec2 v;
+} inst430;
+
+S s;
+
+// should get 5 SPV types for T: no layout, 140/row, 140/col, 430/row, and 430/col
+struct T {
+    mat2 m;
+    int a;
+};
+
+T t;
+
+struct Nestor {
+    T nestorT;
+};
+
+layout(set = 1, binding = 0, std140) uniform Bt1
+{
+    layout(row_major) Nestor nt;
+} Btn1;
+
+layout(set = 1, binding = 0, std140) uniform Bt2
+{
+    layout(column_major) Nestor nt;
+} Btn2;
+
+layout(row_major, set = 1, binding = 0, std140) uniform Bt3
+{
+    layout(column_major) Nestor ntcol;
+    Nestor ntrow;                         // should be row major decoration version of Nestor
+} Btn3;
+
+layout(set = 1, binding = 0, std430) buffer bBt1
+{
+    layout(row_major) Nestor nt;
+} bBtn1;
+
+layout(set = 1, binding = 0, std430) buffer bBt2
+{
+    layout(column_major) Nestor nt;
+} bBtn2;
+
+layout(set = 1, binding = 0, std430) buffer bBt3
+{
+    layout(row_major) Nestor ntcol;
+    Nestor ntrow;                         // should be col major decoration version of Nestor
+} bBtn3;
+
+void main()
+{
+}

Test/test-spirv-list

@@ -30,6 +30,7 @@ spv.accessChain.frag
 spv.aggOps.frag
 spv.always-discard.frag
 spv.always-discard2.frag
+spv.bitCast.frag
 spv.bool.vert
 spv.conditionalDiscard.frag
 spv.conversion.frag
@@ -47,6 +48,7 @@ spv.forwardFun.frag
 spv.functionCall.frag
 spv.functionSemantics.frag
 spv.interpOps.frag
+spv.layoutNested.vert
 spv.length.frag
 spv.localAggregates.frag
 spv.loops.frag

glslang/Include/Types.h

@@ -186,15 +186,15 @@ enum TLayoutPacking {
     ElpShared,      // default, but different than saying nothing
     ElpStd140,
     ElpStd430,
-    ElpPacked
-    // If expanding, see bitfield width below
+    ElpPacked,
+    ElpCount        // If expanding, see bitfield width below
 };
 
 enum TLayoutMatrix {
     ElmNone,
     ElmRowMajor,
-    ElmColumnMajor  // default, but different than saying nothing
-    // If expanding, see bitfield width below
+    ElmColumnMajor, // default, but different than saying nothing
+    ElmCount        // If expanding, see bitfield width below
 };
 
 // Union of geometry shader and tessellation shader geometry types.

glslang/Include/revision.h

@@ -2,5 +2,5 @@
 // For the version, it uses the latest git tag followed by the number of commits.
 // For the date, it uses the current date (when then script is run).
 
-#define GLSLANG_REVISION "SPIRV99.845"
-#define GLSLANG_DATE "13-Dec-2015"
+#define GLSLANG_REVISION "SPIRV99.861"
+#define GLSLANG_DATE "20-Dec-2015"

glslang/MachineIndependent/Initialize.cpp

@@ -3048,7 +3048,7 @@ void IdentifyBuiltIns(int version, EProfile profile, int spv, EShLanguage langua
         SpecialQualifier("gl_FragColor",        EvqFragColor,  EbvFragColor,        symbolTable);
         SpecialQualifier("gl_FragDepth",        EvqFragDepth,  EbvFragDepth,        symbolTable);
         SpecialQualifier("gl_FragDepthEXT",     EvqFragDepth,  EbvFragDepth,        symbolTable);
-        SpecialQualifier("gl_HelperInvocation", EvqIn,         EbvHelperInvocation, symbolTable);
+        SpecialQualifier("gl_HelperInvocation", EvqVaryingIn,  EbvHelperInvocation, symbolTable);
 
         BuiltInVariable("gl_ClipDistance",    EbvClipDistance,   symbolTable);
         BuiltInVariable("gl_CullDistance",    EbvCullDistance,   symbolTable);

glslang/MachineIndependent/ParseHelper.cpp

@@ -5345,8 +5345,11 @@ void TParseContext::fixBlockUniformOffsets(TQualifier& qualifier, TTypeList& typ
         const TSourceLoc& memberLoc = typeList[member].loc;
 
         // "When align is applied to an array, it effects only the start of the array, not the array's internal stride."
-        
-        int memberAlignment = intermediate.getBaseAlignment(*typeList[member].type, memberSize, qualifier.layoutPacking == ElpStd140);
+
+        // modify just the children's view of matrix layout, if there is one for this member
+        TLayoutMatrix subMatrixLayout = typeList[member].type->getQualifier().layoutMatrix;
+        int memberAlignment = intermediate.getBaseAlignment(*typeList[member].type, memberSize, qualifier.layoutPacking == ElpStd140,
+                                                            subMatrixLayout != ElmNone ? subMatrixLayout == ElmRowMajor : qualifier.layoutMatrix == ElmRowMajor);
         if (memberQualifier.hasOffset()) {
             // "The specified offset must be a multiple 
             // of the base alignment of the type of the block member it qualifies, or a compile-time error results."

glslang/MachineIndependent/Scan.cpp

@@ -309,7 +309,7 @@ struct str_hash
         unsigned long hash = 5381;
         int c;
 
-        while (c = *str++)
+        while ((c = *str++))
             hash = ((hash << 5) + hash) + c;
 
         return hash;

glslang/MachineIndependent/linkValidate.cpp

@@ -859,7 +859,7 @@ int TIntermediate::getBaseAlignmentScalar(const TType& type, int& size)
 //
 // The size is returned in the 'size' parameter
 // Return value is the alignment of the type.
-int TIntermediate::getBaseAlignment(const TType& type, int& size, bool std140)
+int TIntermediate::getBaseAlignment(const TType& type, int& size, bool std140, bool rowMajor)
 {
     int alignment;
 
@@ -921,7 +921,7 @@ int TIntermediate::getBaseAlignment(const TType& type, int& size, bool std140)
     if (type.isArray()) {
         // TODO: perf: this might be flattened by using getCumulativeArraySize(), and a deref that discards all arrayness
         TType derefType(type, 0);
-        alignment = getBaseAlignment(derefType, size, std140);
+        alignment = getBaseAlignment(derefType, size, std140, rowMajor);
         if (std140)
             alignment = std::max(baseAlignmentVec4Std140, alignment);
         RoundToPow2(size, alignment);
@@ -937,7 +937,10 @@ int TIntermediate::getBaseAlignment(const TType& type, int& size, bool std140)
         int maxAlignment = std140 ? baseAlignmentVec4Std140 : 0;
         for (size_t m = 0; m < memberList.size(); ++m) {
             int memberSize;
-            int memberAlignment = getBaseAlignment(*memberList[m].type, memberSize, std140);
+            // modify just the children's view of matrix layout, if there is one for this member
+            TLayoutMatrix subMatrixLayout = memberList[m].type->getQualifier().layoutMatrix;
+            int memberAlignment = getBaseAlignment(*memberList[m].type, memberSize, std140,
+                                                   (subMatrixLayout != ElmNone) ? (subMatrixLayout == ElmRowMajor) : rowMajor);
             maxAlignment = std::max(maxAlignment, memberAlignment);
             RoundToPow2(size, memberAlignment);         
             size += memberSize;
@@ -968,11 +971,11 @@ int TIntermediate::getBaseAlignment(const TType& type, int& size, bool std140)
         // rule 5: deref to row, not to column, meaning the size of vector is num columns instead of num rows
         TType derefType(type, 0, type.getQualifier().layoutMatrix == ElmRowMajor);
             
-        alignment = getBaseAlignment(derefType, size, std140);
+        alignment = getBaseAlignment(derefType, size, std140, rowMajor);
         if (std140)
             alignment = std::max(baseAlignmentVec4Std140, alignment);
         RoundToPow2(size, alignment);
-        if (type.getQualifier().layoutMatrix == ElmRowMajor)
+        if (rowMajor)
             size *= type.getMatrixRows();
         else
             size *= type.getMatrixCols();

glslang/MachineIndependent/localintermediate.h

@@ -305,7 +305,7 @@ public:
     }
     int addXfbBufferOffset(const TType&);
     unsigned int computeTypeXfbSize(const TType&, bool& containsDouble) const;
-    static int getBaseAlignment(const TType&, int& size, bool std140);
+    static int getBaseAlignment(const TType&, int& size, bool std140, bool rowMajor);
 
 protected:
     void error(TInfoSink& infoSink, const char*);

glslang/MachineIndependent/reflection.cpp

@@ -123,7 +123,10 @@ public:
         int memberSize;
         int offset = 0;
         for (int m = 0; m <= index; ++m) {
-            int memberAlignment = intermediate.getBaseAlignment(*memberList[m].type, memberSize, type.getQualifier().layoutPacking == ElpStd140);
+            // modify just the children's view of matrix layout, if there is one for this member
+            TLayoutMatrix subMatrixLayout = memberList[m].type->getQualifier().layoutMatrix;
+            int memberAlignment = intermediate.getBaseAlignment(*memberList[m].type, memberSize, type.getQualifier().layoutPacking == ElpStd140,
+                                                                subMatrixLayout != ElmNone ? subMatrixLayout == ElmRowMajor : type.getQualifier().layoutMatrix == ElmRowMajor);
             RoundToPow2(offset, memberAlignment);
             if (m < index)
                 offset += memberSize;
@@ -141,7 +144,8 @@ public:
         int lastOffset = getOffset(blockType, lastIndex);
 
         int lastMemberSize;
-        intermediate.getBaseAlignment(*memberList[lastIndex].type, lastMemberSize, blockType.getQualifier().layoutPacking == ElpStd140);
+        intermediate.getBaseAlignment(*memberList[lastIndex].type, lastMemberSize, blockType.getQualifier().layoutPacking == ElpStd140,
+                                      blockType.getQualifier().layoutMatrix == ElmRowMajor);
 
         return lastOffset + lastMemberSize;
     }