refactor to use workgroup2 stuff in cpp and average metering

Author: keptsecret

26_Autoexposure/app_resources/avg_luma_meter.comp.hlsl

@@ -20,14 +20,15 @@ groupshared float32_t sdata[WORKGROUP_SIZE];
 struct SharedAccessor
 {
     using type = float32_t;
-    void get(const uint32_t index, NBL_REF_ARG(uint32_t) value)
+    template<typename AccessType, typename IndexType>
+    void get(const uint32_t ix, NBL_REF_ARG(AccessType) value)
     {
-        value = sdata[index];
+        value = sdata[ix];
     }
-
-    void set(const uint32_t index, const uint32_t value)
+    template<typename AccessType, typename IndexType>
+    void set(const uint32_t ix, const AccessType value)
     {
-        sdata[index] = value;
+        sdata[ix] = value;
     }
 
     void workgroupExecutionAndMemoryBarrier()
@@ -62,7 +63,7 @@ void main(uint32_t3 ID : SV_GroupThreadID, uint32_t3 GroupID : SV_GroupID)
     SharedAccessor sdata;
     TexAccessor tex;
 
-    using LumaMeter = luma_meter::geom_meter< WORKGROUP_SIZE, SUBGROUP_SIZE, PtrAccessor, SharedAccessor, TexAccessor>;
+    using LumaMeter = luma_meter::geom_meter<wg_config_t, PtrAccessor, SharedAccessor, TexAccessor, device_capabilities>;
     LumaMeter meter = LumaMeter::create(pushData.lumaMin, pushData.lumaMax, pushData.sampleCount, pushData.rcpFirstPassWGCount);
 
     uint32_t texWidth, texHeight;

26_Autoexposure/app_resources/avg_luma_tonemap.comp.hlsl

@@ -25,14 +25,15 @@ groupshared float32_t sdata[WORKGROUP_SIZE];
 struct SharedAccessor
 {
     using type = float32_t;
-    void get(const uint32_t index, NBL_REF_ARG(uint32_t) value)
+    template<typename AccessType, typename IndexType>
+    void get(const uint32_t ix, NBL_REF_ARG(AccessType) value)
     {
-        value = sdata[index];
+        value = sdata[ix];
     }
-
-    void set(const uint32_t index, const uint32_t value)
+    template<typename AccessType, typename IndexType>
+    void set(const uint32_t ix, const AccessType value)
     {
-        sdata[index] = value;
+        sdata[ix] = value;
     }
 
     void workgroupExecutionAndMemoryBarrier()
@@ -67,7 +68,7 @@ void main(uint32_t3 ID : SV_GroupThreadID, uint32_t3 GroupID : SV_GroupID)
     SharedAccessor sdata;
     TexAccessor tex;
 
-    using LumaMeter = luma_meter::geom_meter< WORKGROUP_SIZE, SUBGROUP_SIZE, PtrAccessor, SharedAccessor, TexAccessor>;
+    using LumaMeter = luma_meter::geom_meter<wg_config_t, PtrAccessor, SharedAccessor, TexAccessor, device_capabilities>;
     LumaMeter meter = LumaMeter::create(pushData.lumaMin, pushData.lumaMax, pushData.sampleCount, pushData.rcpFirstPassWGCount);
 
     float32_t EV = meter.gatherLuma(val_accessor);

26_Autoexposure/app_resources/common.hlsl

@@ -7,6 +7,7 @@
 
 #include "nbl/builtin/hlsl/cpp_compat.hlsl"
 #include "nbl/builtin/hlsl/luma_meter/common.hlsl"
+#include "nbl/builtin/hlsl/workgroup2/arithmetic_config.hlsl"
 
 namespace nbl
 {
@@ -42,6 +43,21 @@ struct AutoexposurePushData
 #error "Define SUBGROUP_SIZE!"
 #endif
 
+#ifndef WG_CONFIG_T
+#error "Define WG_CONFIG_T!"
+#endif
+
+using wg_config_t = WG_CONFIG_T;
+
+struct device_capabilities
+{
+#ifdef NATIVE_SUBGROUP_ARITHMETIC
+    NBL_CONSTEXPR_STATIC_INLINE bool shaderSubgroupArithmetic = true;
+#else
+    NBL_CONSTEXPR_STATIC_INLINE bool shaderSubgroupArithmetic = false;
+#endif
+};
+
 #endif
 
 }

26_Autoexposure/app_resources/present.frag.hlsl

@@ -4,8 +4,6 @@
 
 #pragma wave shader_stage(fragment)
 
-#include "app_resources/common.hlsl"
-
 // vertex shader is provided by the fullScreenTriangle extension
 #include <nbl/builtin/hlsl/ext/FullScreenTriangle/SVertexAttributes.hlsl>
 using namespace nbl::hlsl;

26_Autoexposure/main.cpp

@@ -11,6 +11,7 @@
 #include "nbl/ext/FullScreenTriangle/FullScreenTriangle.h"
 
 #include "nbl/builtin/hlsl/luma_meter/common.hlsl"
+#include "nbl/builtin/hlsl/workgroup2/arithmetic_config.hlsl"
 #include "app_resources/common.hlsl"
 
 using namespace nbl;
@@ -317,18 +318,26 @@ class AutoexposureApp final : public SimpleWindowedApplication, public BuiltinRe
 				const uint32_t workgroupSize = m_physicalDevice->getLimits().maxComputeWorkGroupInvocations;
 				const uint32_t subgroupSize = m_physicalDevice->getLimits().maxSubgroupSize;
 
+				const uint32_t configItemsPerInvoc = MeterMode == MeteringMode::AVERAGE ? 1 : workgroupSize / BinCount;
+				workgroup2::SArithmeticConfiguration wgConfig;
+				wgConfig.init(hlsl::findMSB(workgroupSize), hlsl::log2(float(subgroupSize)), configItemsPerInvoc);
+
 				struct MacroDefines
 				{
 					std::string identifier;
 					std::string definition;
 				};
-				const MacroDefines definesBuf[2] = {
+				constexpr uint32_t NumDefines = 4;
+				const MacroDefines definesBuf[NumDefines] = {
 					{ "WORKGROUP_SIZE", std::to_string(workgroupSize) },
-					{ "SUBGROUP_SIZE", std::to_string(subgroupSize) }
+					{ "SUBGROUP_SIZE", std::to_string(subgroupSize) },
+					{"WG_CONFIG_T", wgConfig.getConfigTemplateStructString()},
+                    {"NATIVE_SUBGROUP_ARITHMETIC", "1"}
 				};
 
+				const uint32_t defineCount = m_physicalDevice->getLimits().shaderSubgroupArithmetic ? NumDefines : NumDefines - 1;
 				std::vector<IShaderCompiler::SMacroDefinition> defines;
-				for (uint32_t i = 0; i < 2; i++)
+				for (uint32_t i = 0; i < defineCount; i++)
 					defines.emplace_back(definesBuf[i].identifier, definesBuf[i].definition);
 			    options.preprocessorOptions.extraDefines = defines;