diff --git a/src/renderer_d3d11.cpp b/src/renderer_d3d11.cpp
index bb7516686..fc968bdf8 100644
--- a/src/renderer_d3d11.cpp
+++ b/src/renderer_d3d11.cpp
@@ -6420,8 +6420,8 @@ BX_PRAGMA_DIAGNOSTIC_POP();
 						deviceCtx->VSSetConstantBuffers(0, 1, &vsh->m_buffer);
 
 						const ShaderD3D11* fsh = program.m_fsh;
-						if (NULL != m_currentColor
-						||  fsh->m_hasDepthOp)
+						if (NULL != fsh
+						&& (NULL != m_currentColor || fsh->m_hasDepthOp) )
 						{
 							deviceCtx->PSSetShader(fsh->m_pixelShader, NULL, 0);
 							deviceCtx->PSSetConstantBuffers(0, 1, &fsh->m_buffer);
@@ -6448,10 +6448,13 @@ BX_PRAGMA_DIAGNOSTIC_POP();
 							commit(*vcb);
 						}
 
-						UniformBuffer* fcb = program.m_fsh->m_constantBuffer;
-						if (NULL != fcb)
+						if (NULL != program.m_fsh)
 						{
-							commit(*fcb);
+							UniformBuffer* fcb = program.m_fsh->m_constantBuffer;
+							if (NULL != fcb)
+							{
+								commit(*fcb);
+							}
 						}
 					}
 
diff --git a/src/renderer_d3d12.cpp b/src/renderer_d3d12.cpp
index 70890ac95..f7fb64a40 100644
--- a/src/renderer_d3d12.cpp
+++ b/src/renderer_d3d12.cpp
@@ -2826,7 +2826,11 @@ data.NumQualityLevels = 0;
 			murmur.add(_stencil);
 			murmur.add(program.m_vsh->m_hash);
 			murmur.add(program.m_vsh->m_attrMask, sizeof(program.m_vsh->m_attrMask) );
-			murmur.add(program.m_fsh->m_hash);
+			if (NULL != program.m_fsh)
+			{
+				murmur.add(program.m_fsh->m_hash);
+			}
+
 			for (uint32_t ii = 0; ii < _numStreams; ++ii)
 			{
 				murmur.add(_vertexDecls[ii]->m_hash);
@@ -2851,66 +2855,76 @@ data.NumQualityLevels = 0;
 			desc.VS.pShaderBytecode = program.m_vsh->m_code->data;
 			desc.VS.BytecodeLength  = program.m_vsh->m_code->size;
 
- 			const Memory* temp = alloc(program.m_fsh->m_code->size);
- 			bx::memSet(temp->data, 0, temp->size);
- 			bx::MemoryReader rd(program.m_fsh->m_code->data, program.m_fsh->m_code->size);
- 			bx::StaticMemoryBlockWriter wr(temp->data, temp->size);
+			const Memory* temp = NULL;
 
-			DxbcContext dxbc;
-			bx::Error err;
-			read(&rd, dxbc, &err);
-
-			bool patchShader = !dxbc.shader.aon9;
-			if (BX_ENABLED(BGFX_CONFIG_DEBUG)
-			&&  patchShader)
+			if (NULL != program.m_fsh)
 			{
-				union { uint32_t offset; void* ptr; } cast = { 0 };
-				filter(dxbc.shader, dxbc.shader, patchCb0, cast.ptr);
+ 				temp = alloc(program.m_fsh->m_code->size);
+ 				bx::memSet(temp->data, 0, temp->size);
+ 				bx::MemoryReader rd(program.m_fsh->m_code->data, program.m_fsh->m_code->size);
+ 				bx::StaticMemoryBlockWriter wr(temp->data, temp->size);
 
-				write(&wr, dxbc, &err);
+				DxbcContext dxbc;
+				bx::Error err;
+				read(&rd, dxbc, &err);
 
-				dxbcHash(temp->data + 20, temp->size - 20, temp->data + 4);
-
-				patchShader = 0 == bx::memCmp(program.m_fsh->m_code->data, temp->data, 16);
-				BX_CHECK(patchShader, "DXBC fragment shader patching error (ShaderHandle: %d).", program.m_fsh - m_shaders);
-
-				if (!patchShader)
+				bool patchShader = !dxbc.shader.aon9;
+				if (BX_ENABLED(BGFX_CONFIG_DEBUG)
+				&&  patchShader)
 				{
-					for (uint32_t ii = 20; ii < temp->size; ii += 16)
-					{
-						if (0 != bx::memCmp(&program.m_fsh->m_code->data[ii], &temp->data[ii], 16) )
-						{
-// 							bx::debugPrintfData(&program.m_fsh->m_code->data[ii], temp->size-ii, "");
-// 							bx::debugPrintfData(&temp->data[ii], temp->size-ii, "");
-							break;
-						}
-					}
+					union { uint32_t offset; void* ptr; } cast = { 0 };
+					filter(dxbc.shader, dxbc.shader, patchCb0, cast.ptr);
 
+					write(&wr, dxbc, &err);
+
+					dxbcHash(temp->data + 20, temp->size - 20, temp->data + 4);
+
+					patchShader = 0 == bx::memCmp(program.m_fsh->m_code->data, temp->data, 16);
+					BX_CHECK(patchShader, "DXBC fragment shader patching error (ShaderHandle: %d).", program.m_fsh - m_shaders);
+
+					if (!patchShader)
+					{
+						for (uint32_t ii = 20; ii < temp->size; ii += 16)
+						{
+							if (0 != bx::memCmp(&program.m_fsh->m_code->data[ii], &temp->data[ii], 16) )
+							{
+	// 							bx::debugPrintfData(&program.m_fsh->m_code->data[ii], temp->size-ii, "");
+	// 							bx::debugPrintfData(&temp->data[ii], temp->size-ii, "");
+								break;
+							}
+						}
+
+						desc.PS.pShaderBytecode = program.m_fsh->m_code->data;
+						desc.PS.BytecodeLength  = program.m_fsh->m_code->size;
+					}
+				}
+
+				if (patchShader)
+				{
+					bx::memCopy(temp->data, program.m_fsh->m_code->data, program.m_fsh->m_code->size);
+
+					bx::seek(&wr, 0, bx::Whence::Begin);
+					union { uint32_t offset; void* ptr; } cast =
+					{
+						uint32_t(program.m_vsh->m_size)/16
+					};
+					filter(dxbc.shader, dxbc.shader, patchCb0, cast.ptr);
+					write(&wr, dxbc, &err);
+					dxbcHash(temp->data + 20, temp->size - 20, temp->data + 4);
+
+					desc.PS.pShaderBytecode = temp->data;
+					desc.PS.BytecodeLength  = temp->size;
+				}
+				else
+				{
 					desc.PS.pShaderBytecode = program.m_fsh->m_code->data;
 					desc.PS.BytecodeLength  = program.m_fsh->m_code->size;
 				}
 			}
-
-			if (patchShader)
-			{
-				bx::memCopy(temp->data, program.m_fsh->m_code->data, program.m_fsh->m_code->size);
-
-				bx::seek(&wr, 0, bx::Whence::Begin);
-				union { uint32_t offset; void* ptr; } cast =
-				{
-					uint32_t(program.m_vsh->m_size)/16
-				};
-				filter(dxbc.shader, dxbc.shader, patchCb0, cast.ptr);
-				write(&wr, dxbc, &err);
-				dxbcHash(temp->data + 20, temp->size - 20, temp->data + 4);
-
-				desc.PS.pShaderBytecode = temp->data;
-				desc.PS.BytecodeLength  = temp->size;
-			}
 			else
 			{
-				desc.PS.pShaderBytecode = program.m_fsh->m_code->data;
-				desc.PS.BytecodeLength  = program.m_fsh->m_code->size;
+				desc.PS.pShaderBytecode = NULL;
+				desc.PS.BytecodeLength  = 0;
 			}
 
 			desc.DS.pShaderBytecode = NULL;
@@ -3018,7 +3032,10 @@ data.NumQualityLevels = 0;
 
 			m_pipelineStateCache.add(hash, pso);
 
-			release(temp);
+			if (NULL != temp)
+			{
+				release(temp);
+			}
 
 			ID3DBlob* blob;
 			HRESULT hr = pso->GetCachedBlob(&blob);
@@ -6285,10 +6302,13 @@ data.NumQualityLevels = 0;
 							commit(*vcb);
 						}
 
-						UniformBuffer* fcb = program.m_fsh->m_constantBuffer;
-						if (NULL != fcb)
+						if (NULL != program.m_fsh)
 						{
-							commit(*fcb);
+							UniformBuffer* fcb = program.m_fsh->m_constantBuffer;
+							if (NULL != fcb)
+							{
+								commit(*fcb);
+							}
 						}
 
 						hasPredefined = 0 < program.m_numPredefined;
diff --git a/src/renderer_d3d9.cpp b/src/renderer_d3d9.cpp
index 7bd7491cd..264d96345 100644
--- a/src/renderer_d3d9.cpp
+++ b/src/renderer_d3d9.cpp
@@ -1027,7 +1027,7 @@ namespace bgfx { namespace d3d9
 
 		void createProgram(ProgramHandle _handle, ShaderHandle _vsh, ShaderHandle _fsh) override
 		{
-			m_program[_handle.idx].create(m_shaders[_vsh.idx], m_shaders[_fsh.idx]);
+			m_program[_handle.idx].create(&m_shaders[_vsh.idx], isValid(_fsh) ? &m_shaders[_fsh.idx] : NULL);
 		}
 
 		void destroyProgram(ProgramHandle _handle) override
@@ -4105,7 +4105,10 @@ namespace bgfx { namespace d3d9
 					{
 						ProgramD3D9& program = m_program[programIdx];
 						device->SetVertexShader(program.m_vsh->m_vertexShader);
-						device->SetPixelShader(program.m_fsh->m_pixelShader);
+						device->SetPixelShader(NULL == program.m_fsh
+							? NULL
+							: program.m_fsh->m_pixelShader
+							);
 					}
 
 					programChanged =
@@ -4124,10 +4127,13 @@ namespace bgfx { namespace d3d9
 							commit(*vcb);
 						}
 
-						UniformBuffer* fcb = program.m_fsh->m_constantBuffer;
-						if (NULL != fcb)
+						if (NULL != program.m_fsh)
 						{
-							commit(*fcb);
+							UniformBuffer* fcb = program.m_fsh->m_constantBuffer;
+							if (NULL != fcb)
+							{
+								commit(*fcb);
+							}
 						}
 					}
 
@@ -4406,11 +4412,11 @@ namespace bgfx { namespace d3d9
 					);
 
 				const D3DADAPTER_IDENTIFIER9& identifier = m_identifier;
-				tvm.printf(0, pos++, 0x8b, " Device: %s (%s)", identifier.Description, identifier.Driver);
+				tvm.printf(0, pos++, 0x8f, " Device: %s (%s)", identifier.Description, identifier.Driver);
 
 				char processMemoryUsed[16];
 				bx::prettify(processMemoryUsed, BX_COUNTOF(processMemoryUsed), bx::getProcessMemoryUsed() );
-				tvm.printf(0, pos++, 0x8b, " Memory: %s (process) ", processMemoryUsed);
+				tvm.printf(0, pos++, 0x8f, " Memory: %s (process) ", processMemoryUsed);
 
 				pos = 10;
 				tvm.printf(10, pos++, 0x8b, "        Frame: %7.3f, % 7.3f \x1f, % 7.3f \x1e [ms] / % 6.2f FPS "
diff --git a/src/renderer_d3d9.h b/src/renderer_d3d9.h
index b9321403e..248b61467 100644
--- a/src/renderer_d3d9.h
+++ b/src/renderer_d3d9.h
@@ -259,17 +259,19 @@ namespace bgfx { namespace d3d9
 
 	struct ProgramD3D9
 	{
-		void create(const ShaderD3D9& _vsh, const ShaderD3D9& _fsh)
+		void create(const ShaderD3D9* _vsh, const ShaderD3D9* _fsh)
 		{
-			BX_CHECK(NULL != _vsh.m_vertexShader, "Vertex shader doesn't exist.");
-			m_vsh = &_vsh;
+			m_vsh = _vsh;
+			m_fsh = _fsh;
 
-			BX_CHECK(NULL != _fsh.m_pixelShader, "Fragment shader doesn't exist.");
-			m_fsh = &_fsh;
+			bx::memCopy(&m_predefined[0], _vsh->m_predefined, _vsh->m_numPredefined*sizeof(PredefinedUniform) );
+			m_numPredefined = _vsh->m_numPredefined;
 
-			bx::memCopy(&m_predefined[0], _vsh.m_predefined, _vsh.m_numPredefined*sizeof(PredefinedUniform) );
-			bx::memCopy(&m_predefined[_vsh.m_numPredefined], _fsh.m_predefined, _fsh.m_numPredefined*sizeof(PredefinedUniform) );
-			m_numPredefined = _vsh.m_numPredefined + _fsh.m_numPredefined;
+			if (NULL != _fsh)
+			{
+				bx::memCopy(&m_predefined[_vsh->m_numPredefined], _fsh->m_predefined, _fsh->m_numPredefined*sizeof(PredefinedUniform) );
+				m_numPredefined += _fsh->m_numPredefined;
+			}
 		}
 
 		void destroy()