Validate no gaps in UAV store writemask (#4495)

* Adding validation for uav continuous store

* Updated message text

* Add comment

* Updating text and adding tests

Author: gracejennings

docs/DXIL.rst

@@ -3079,6 +3079,7 @@ INSTR.TGSMRACECOND                        Race condition writing to shared memor
 INSTR.UNDEFINEDVALUEFORUAVSTORE           Assignment of undefined values to UAV.
 INSTR.UNDEFRESULTFORGETDIMENSION          GetDimensions used undef dimension %0 on %1.
 INSTR.WRITEMASKFORTYPEDUAVSTORE           store on typed uav must write to all four components of the UAV.
+INSTR.WRITEMASKGAPFORUAV                  UAV write mask must be contiguous, starting at x: .x, .xy, .xyz, or .xyzw.
 INSTR.WRITEMASKMATCHVALUEFORUAVSTORE      uav store write mask must match store value mask, write mask is %0 and store value mask is %1.
 META.BARYCENTRICSFLOAT3                   only 'float3' type is allowed for SV_Barycentrics.
 META.BARYCENTRICSINTERPOLATION            SV_Barycentrics cannot be used with 'nointerpolation' type.

lib/HLSL/DxilValidation.cpp

@@ -1466,6 +1466,11 @@ static bool ValidateStorageMasks(Instruction *I, DXIL::OpCode opcode, ConstantIn
     ValCtx.EmitInstrError(I, ValidationRule::InstrWriteMaskForTypedUAVStore);
   }
 
+  // write mask must be contiguous (.x .xy .xyz or .xyzw)
+  if (!((uMask == 0xf) || (uMask == 0x7) || (uMask == 0x3) || (uMask == 0x1))) {
+    ValCtx.EmitInstrError(I, ValidationRule::InstrWriteMaskGapForUAV);
+  }
+
   // If a bit is set in the uMask (expected values) that isn't set in stValMask (user provided values)
   // then the user failed to define some of the output values.
   if (uMask & ~stValMask)

tools/clang/unittests/HLSL/ValidationTest.cpp

@@ -72,6 +72,9 @@ class ValidationTest : public ::testing::Test {
   TEST_METHOD(TypedUAVStoreFullMask0)
   TEST_METHOD(TypedUAVStoreFullMask1)
   TEST_METHOD(UAVStoreMaskMatch)
+  TEST_METHOD(UAVStoreMaskGap)
+  TEST_METHOD(UAVStoreMaskGap2)
+  TEST_METHOD(UAVStoreMaskGap3)
   TEST_METHOD(Recursive)
   TEST_METHOD(Recursive2)
   TEST_METHOD(Recursive3)
@@ -1197,6 +1200,36 @@ TEST_F(ValidationTest, UAVStoreMaskMatch) {
       "uav store write mask must match store value mask, write mask is 7 and store value mask is 15.");
 }
 
+TEST_F(ValidationTest, UAVStoreMaskGap) {
+  if (m_ver.SkipDxilVersion(1, 6))
+    return;
+  RewriteAssemblyCheckMsg(
+      L"..\\CodeGenHLSL\\uav_store.hlsl", "ps_6_0",
+      "i32 2, i32 2, i32 2, i32 2, i8 15)",
+      "i32 undef, i32 2, i32 undef, i32 2, i8 10)",
+      "UAV write mask must be contiguous, starting at x: .x, .xy, .xyz, or .xyzw.");
+}
+
+TEST_F(ValidationTest, UAVStoreMaskGap2) {
+  if (m_ver.SkipDxilVersion(1, 6))
+    return;
+  RewriteAssemblyCheckMsg(L"..\\CodeGenHLSL\\uav_store.hlsl", "ps_6_0",
+                          "i32 2, i32 2, i32 2, i32 2, i8 15)",
+                          "i32 undef, i32 2, i32 2, i32 2, i8 14)",
+                          "UAV write mask must be contiguous, starting at x: "
+                          ".x, .xy, .xyz, or .xyzw.");
+}
+
+TEST_F(ValidationTest, UAVStoreMaskGap3) {
+  if (m_ver.SkipDxilVersion(1, 6))
+    return;
+  RewriteAssemblyCheckMsg(L"..\\CodeGenHLSL\\uav_store.hlsl", "ps_6_0",
+                          "i32 2, i32 2, i32 2, i32 2, i8 15)",
+                          "i32 undef, i32 undef, i32 undef, i32 2, i8 8)",
+                          "UAV write mask must be contiguous, starting at x: "
+                          ".x, .xy, .xyz, or .xyzw.");
+}
+
 TEST_F(ValidationTest, Recursive) {
   // Includes coverage for user-defined functions.
   TestCheck(L"..\\CodeGenHLSL\\recursive.ll");

utils/hct/hctdb.py

@@ -2582,6 +2582,7 @@ def build_valrules(self):
         self.add_valrule("Instr.BarrierModeNoMemory", "sync must include some form of memory barrier - _u (UAV) and/or _g (Thread Group Shared Memory).  Only _t (thread group sync) is optional.")
         self.add_valrule("Instr.BarrierModeForNonCS", "sync in a non-Compute/Amplification/Mesh Shader must only sync UAV (sync_uglobal).")
         self.add_valrule("Instr.WriteMaskForTypedUAVStore", "store on typed uav must write to all four components of the UAV.")
+        self.add_valrule("Instr.WriteMaskGapForUAV", "UAV write mask must be contiguous, starting at x: .x, .xy, .xyz, or .xyzw.")
         self.add_valrule("Instr.ResourceKindForCalcLOD","lod requires resource declared as texture1D/2D/3D/Cube/CubeArray/1DArray/2DArray.")
         self.add_valrule("Instr.ResourceKindForSample", "sample/_l/_d requires resource declared as texture1D/2D/3D/Cube/1DArray/2DArray/CubeArray.")
         self.add_valrule("Instr.ResourceKindForSampleC", "samplec requires resource declared as texture1D/2D/Cube/1DArray/2DArray/CubeArray.")