Merge pull request #455 from KhronosGroup/fix-452

Implement OpFRem.

reference/opt/shaders-hlsl/asm/frag/frem.asm.frag

@@ -0,0 +1,29 @@
+static float4 FragColor;
+static float4 vA;
+static float4 vB;
+
+struct SPIRV_Cross_Input
+{
+    float4 vA : TEXCOORD0;
+    float4 vB : TEXCOORD1;
+};
+
+struct SPIRV_Cross_Output
+{
+    float4 FragColor : SV_Target0;
+};
+
+void frag_main()
+{
+    FragColor = fmod(vA, vB);
+}
+
+SPIRV_Cross_Output main(SPIRV_Cross_Input stage_input)
+{
+    vA = stage_input.vA;
+    vB = stage_input.vB;
+    frag_main();
+    SPIRV_Cross_Output stage_output;
+    stage_output.FragColor = FragColor;
+    return stage_output;
+}

reference/opt/shaders-msl/asm/frag/frem.asm.frag

@@ -0,0 +1,23 @@
+#include <metal_stdlib>
+#include <simd/simd.h>
+
+using namespace metal;
+
+struct main0_in
+{
+    float4 vB [[user(locn1)]];
+    float4 vA [[user(locn0)]];
+};
+
+struct main0_out
+{
+    float4 FragColor [[color(0)]];
+};
+
+fragment main0_out main0(main0_in in [[stage_in]])
+{
+    main0_out out = {};
+    out.FragColor = fmod(in.vA, in.vB);
+    return out;
+}
+

reference/opt/shaders/asm/frag/frem.asm.frag

@@ -0,0 +1,13 @@
+#version 310 es
+precision mediump float;
+precision highp int;
+
+layout(location = 0) out vec4 FragColor;
+layout(location = 0) in vec4 vA;
+layout(location = 1) in vec4 vB;
+
+void main()
+{
+    FragColor = vA - vB * trunc(vA / vB);
+}
+

reference/shaders-hlsl/asm/frag/frem.asm.frag

@@ -0,0 +1,29 @@
+static float4 FragColor;
+static float4 vA;
+static float4 vB;
+
+struct SPIRV_Cross_Input
+{
+    float4 vA : TEXCOORD0;
+    float4 vB : TEXCOORD1;
+};
+
+struct SPIRV_Cross_Output
+{
+    float4 FragColor : SV_Target0;
+};
+
+void frag_main()
+{
+    FragColor = fmod(vA, vB);
+}
+
+SPIRV_Cross_Output main(SPIRV_Cross_Input stage_input)
+{
+    vA = stage_input.vA;
+    vB = stage_input.vB;
+    frag_main();
+    SPIRV_Cross_Output stage_output;
+    stage_output.FragColor = FragColor;
+    return stage_output;
+}

reference/shaders-msl/asm/frag/frem.asm.frag

@@ -0,0 +1,23 @@
+#include <metal_stdlib>
+#include <simd/simd.h>
+
+using namespace metal;
+
+struct main0_in
+{
+    float4 vB [[user(locn1)]];
+    float4 vA [[user(locn0)]];
+};
+
+struct main0_out
+{
+    float4 FragColor [[color(0)]];
+};
+
+fragment main0_out main0(main0_in in [[stage_in]])
+{
+    main0_out out = {};
+    out.FragColor = fmod(in.vA, in.vB);
+    return out;
+}
+

reference/shaders/asm/frag/frem.asm.frag

@@ -0,0 +1,13 @@
+#version 310 es
+precision mediump float;
+precision highp int;
+
+layout(location = 0) out vec4 FragColor;
+layout(location = 0) in vec4 vA;
+layout(location = 1) in vec4 vB;
+
+void main()
+{
+    FragColor = vA - vB * trunc(vA / vB);
+}
+

shaders-hlsl/asm/frag/frem.asm.frag

@@ -0,0 +1,41 @@
+; SPIR-V
+; Version: 1.0
+; Generator: Khronos Glslang Reference Front End; 3
+; Bound: 16
+; Schema: 0
+               OpCapability Shader
+          %1 = OpExtInstImport "GLSL.std.450"
+               OpMemoryModel Logical GLSL450
+               OpEntryPoint Fragment %main "main" %FragColor %vA %vB
+               OpExecutionMode %main OriginUpperLeft
+               OpSource ESSL 310
+               OpName %main "main"
+               OpName %FragColor "FragColor"
+               OpName %vA "vA"
+               OpName %vB "vB"
+               OpDecorate %FragColor RelaxedPrecision
+               OpDecorate %FragColor Location 0
+               OpDecorate %vA RelaxedPrecision
+               OpDecorate %vA Location 0
+               OpDecorate %12 RelaxedPrecision
+               OpDecorate %vB RelaxedPrecision
+               OpDecorate %vB Location 1
+               OpDecorate %14 RelaxedPrecision
+               OpDecorate %15 RelaxedPrecision
+       %void = OpTypeVoid
+          %3 = OpTypeFunction %void
+      %float = OpTypeFloat 32
+    %v4float = OpTypeVector %float 4
+%_ptr_Output_v4float = OpTypePointer Output %v4float
+  %FragColor = OpVariable %_ptr_Output_v4float Output
+%_ptr_Input_v4float = OpTypePointer Input %v4float
+         %vA = OpVariable %_ptr_Input_v4float Input
+         %vB = OpVariable %_ptr_Input_v4float Input
+       %main = OpFunction %void None %3
+          %5 = OpLabel
+         %12 = OpLoad %v4float %vA
+         %14 = OpLoad %v4float %vB
+         %15 = OpFRem %v4float %12 %14
+               OpStore %FragColor %15
+               OpReturn
+               OpFunctionEnd

shaders-msl/asm/frag/frem.asm.frag

@@ -0,0 +1,41 @@
+; SPIR-V
+; Version: 1.0
+; Generator: Khronos Glslang Reference Front End; 3
+; Bound: 16
+; Schema: 0
+               OpCapability Shader
+          %1 = OpExtInstImport "GLSL.std.450"
+               OpMemoryModel Logical GLSL450
+               OpEntryPoint Fragment %main "main" %FragColor %vA %vB
+               OpExecutionMode %main OriginUpperLeft
+               OpSource ESSL 310
+               OpName %main "main"
+               OpName %FragColor "FragColor"
+               OpName %vA "vA"
+               OpName %vB "vB"
+               OpDecorate %FragColor RelaxedPrecision
+               OpDecorate %FragColor Location 0
+               OpDecorate %vA RelaxedPrecision
+               OpDecorate %vA Location 0
+               OpDecorate %12 RelaxedPrecision
+               OpDecorate %vB RelaxedPrecision
+               OpDecorate %vB Location 1
+               OpDecorate %14 RelaxedPrecision
+               OpDecorate %15 RelaxedPrecision
+       %void = OpTypeVoid
+          %3 = OpTypeFunction %void
+      %float = OpTypeFloat 32
+    %v4float = OpTypeVector %float 4
+%_ptr_Output_v4float = OpTypePointer Output %v4float
+  %FragColor = OpVariable %_ptr_Output_v4float Output
+%_ptr_Input_v4float = OpTypePointer Input %v4float
+         %vA = OpVariable %_ptr_Input_v4float Input
+         %vB = OpVariable %_ptr_Input_v4float Input
+       %main = OpFunction %void None %3
+          %5 = OpLabel
+         %12 = OpLoad %v4float %vA
+         %14 = OpLoad %v4float %vB
+         %15 = OpFRem %v4float %12 %14
+               OpStore %FragColor %15
+               OpReturn
+               OpFunctionEnd

shaders/asm/frag/frem.asm.frag

@@ -0,0 +1,41 @@
+; SPIR-V
+; Version: 1.0
+; Generator: Khronos Glslang Reference Front End; 3
+; Bound: 16
+; Schema: 0
+               OpCapability Shader
+          %1 = OpExtInstImport "GLSL.std.450"
+               OpMemoryModel Logical GLSL450
+               OpEntryPoint Fragment %main "main" %FragColor %vA %vB
+               OpExecutionMode %main OriginUpperLeft
+               OpSource ESSL 310
+               OpName %main "main"
+               OpName %FragColor "FragColor"
+               OpName %vA "vA"
+               OpName %vB "vB"
+               OpDecorate %FragColor RelaxedPrecision
+               OpDecorate %FragColor Location 0
+               OpDecorate %vA RelaxedPrecision
+               OpDecorate %vA Location 0
+               OpDecorate %12 RelaxedPrecision
+               OpDecorate %vB RelaxedPrecision
+               OpDecorate %vB Location 1
+               OpDecorate %14 RelaxedPrecision
+               OpDecorate %15 RelaxedPrecision
+       %void = OpTypeVoid
+          %3 = OpTypeFunction %void
+      %float = OpTypeFloat 32
+    %v4float = OpTypeVector %float 4
+%_ptr_Output_v4float = OpTypePointer Output %v4float
+  %FragColor = OpVariable %_ptr_Output_v4float Output
+%_ptr_Input_v4float = OpTypePointer Input %v4float
+         %vA = OpVariable %_ptr_Input_v4float Input
+         %vB = OpVariable %_ptr_Input_v4float Input
+       %main = OpFunction %void None %3
+          %5 = OpLabel
+         %12 = OpLoad %v4float %vA
+         %14 = OpLoad %v4float %vB
+         %15 = OpFRem %v4float %12 %14
+               OpStore %FragColor %15
+               OpReturn
+               OpFunctionEnd

spirv_glsl.cpp

@@ -5909,6 +5909,28 @@ void CompilerGLSL::emit_instruction(const Instruction &instruction)
 		BFOP(mod);
 		break;
 
+	case OpFRem:
+	{
+		if (is_legacy())
+			SPIRV_CROSS_THROW("OpFRem requires trunc() and is only supported on non-legacy targets. A workaround is "
+			                  "needed for legacy.");
+
+		uint32_t result_type = ops[0];
+		uint32_t result_id = ops[1];
+		uint32_t op0 = ops[2];
+		uint32_t op1 = ops[3];
+
+		// Needs special handling.
+		bool forward = should_forward(op0) && should_forward(op1);
+		auto expr = join(to_enclosed_expression(op0), " - ", to_enclosed_expression(op1), " * ", "trunc(",
+		                 to_enclosed_expression(op0), " / ", to_enclosed_expression(op1), ")");
+
+		emit_op(result_type, result_id, expr, forward);
+		inherit_expression_dependencies(result_id, op0);
+		inherit_expression_dependencies(result_id, op1);
+		break;
+	}
+
 	// Relational
 	case OpAny:
 		UFOP(any);

spirv_hlsl.cpp

@@ -3224,6 +3224,10 @@ void CompilerHLSL::emit_instruction(const Instruction &instruction)
 		break;
 	}
 
+	case OpFRem:
+		emit_binary_func_op(ops[0], ops[1], ops[2], ops[3], "fmod");
+		break;
+
 	case OpImage:
 	{
 		uint32_t result_type = ops[0];

spirv_msl.cpp

@@ -18,8 +18,8 @@
 #include "GLSL.std.450.h"
 
 #include <algorithm>
-#include <numeric>
 #include <assert.h>
+#include <numeric>
 
 using namespace spv;
 using namespace spirv_cross;
@@ -66,8 +66,7 @@ void CompilerMSL::build_implicit_builtins()
 
 			auto &var = id.get<SPIRVariable>();
 
-			if (var.storage == StorageClassInput &&
-			    meta[var.self].decoration.builtin &&
+			if (var.storage == StorageClassInput && meta[var.self].decoration.builtin &&
 			    meta[var.self].decoration.builtin_type == BuiltInFragCoord)
 			{
 				builtin_frag_coord_id = var.self;
@@ -1486,6 +1485,10 @@ void CompilerMSL::emit_instruction(const Instruction &instruction)
 		UFOP(popcount);
 		break;
 
+	case OpFRem:
+		BFOP(fmod);
+		break;
+
 	// Atomics
 	case OpAtomicExchange:
 	{