Adds tests for the new Morton Code class #187

Fletterio · 2025-03-28T23:18:40Z

Checks that the arithmetic and comparisons work as expected + checks it compiles for GPU

XX_Mortons/main.cpp

CMakeLists.txt

…ing 22_CppCOmpat to run tests

22_CppCompat/CIntrinsicsTester.h

12_Mortons/app_resources/common.hlsl

12_Mortons/app_resources/shader.hlsl

devshgraphicsprogramming · 2025-11-28T09:26:24Z

CMakeLists.txt


  	add_subdirectory(70_FLIPFluids)
 	add_subdirectory(71_RayTracingPipeline)
+	add_subdirectory(73_Mortons EXCLUDE_FROM_ALL)


actually can you make it example 14 or 15 because the low numbers is where I keep basic HLSL/C++

devshgraphicsprogramming · 2025-11-28T09:28:27Z

73_Mortons/app_resources/test.comp.hlsl

+[numthreads(256, 1, 1)]
+[shader("compute")]
+void main(uint3 invocationID : SV_DispatchThreadID)
+{
+    if (invocationID.x == 0)
+        fillTestValues(inputTestValues[0], outputTestValues[0]);


just make a 1,1,1 workgroup and always call fillTextValues(inputTestValues[gl_GlobalInvocationID.x],outputTestValues[gl_GlobalInvocationID.x])

devshgraphicsprogramming · 2025-11-28T09:31:27Z

22_CppCompat/CIntrinsicsTester.h

+        // Disabled: current glm implementation is wrong
+        //verifyTestValue("subBorrowResult", expectedTestValues.subBorrow.result, testValues.subBorrow.result, testType);
+        //verifyTestValue("subBorrowBorrow", expectedTestValues.subBorrow.borrow, testValues.subBorrow.borrow, testType);


then use an alternative implementation of subBorrow and don't use GLM's

assigned @Przemog1 in next PR

Actually glm merged my PR, and I pushed a fix to our branch. So it should be usable now

ye so its just updating glm

devshgraphicsprogramming · 2025-11-28T09:32:17Z

73_Mortons/CTester.h

+
+#include <nabla.h>
+#include "app_resources/testCommon.hlsl"
+#include "ITester.h"


why are we reusing things from a different example ?

if you want ITester to be general, please move to appropriate directory like common/include/nbl/examples/testing

see https://github.com/Devsh-Graphics-Programming/Nabla-Examples-and-Tests/tree/08c898d5af460ba6469a78fb625216e27a1bc8a8/common/include/nbl/examples

assigned @Przemog1 in next PR

devshgraphicsprogramming · 2025-11-28T09:35:14Z

73_Mortons/ITester.h

+#ifndef _NBL_EXAMPLES_TESTS_22_CPP_COMPAT_I_TESTER_INCLUDED_
+#define _NBL_EXAMPLES_TESTS_22_CPP_COMPAT_I_TESTER_INCLUDED_
+
+#include <nabla.h>
+#include "app_resources/common.hlsl"
+#include "nbl/application_templates/MonoDeviceApplication.hpp"
+
+using namespace nbl;
+
+class ITester 
+{


please unify with ex22, don't want this much duplicate code

assigned @Przemog1 in next PR

devshgraphicsprogramming · 2025-11-28T09:40:34Z

73_Mortons/ITester.h

+    template<typename InputStruct, typename OutputStruct>
+    OutputStruct dispatch(const InputStruct& input)
+    {
+        // Update input buffer
+        if (!m_inputBufferAllocation.memory->map({ 0ull,m_inputBufferAllocation.memory->getAllocationSize() }, video::IDeviceMemoryAllocation::EMCAF_READ))
+            logFail("Failed to map the Device Memory!\n");
+
+        const video::ILogicalDevice::MappedMemoryRange memoryRange(m_inputBufferAllocation.memory.get(), 0ull, m_inputBufferAllocation.memory->getAllocationSize());
+        if (!m_inputBufferAllocation.memory->getMemoryPropertyFlags().hasFlags(video::IDeviceMemoryAllocation::EMPF_HOST_COHERENT_BIT))
+            m_device->invalidateMappedMemoryRanges(1, &memoryRange);
+
+        std::memcpy(static_cast<InputStruct*>(m_inputBufferAllocation.memory->getMappedPointer()), &input, sizeof(InputStruct));
+
+        m_inputBufferAllocation.memory->unmap();
+
+        // record command buffer
+        m_cmdbuf->reset(video::IGPUCommandBuffer::RESET_FLAGS::NONE);
+        m_cmdbuf->begin(video::IGPUCommandBuffer::USAGE::NONE);
+        m_cmdbuf->beginDebugMarker("test", core::vector4df_SIMD(0, 1, 0, 1));
+        m_cmdbuf->bindComputePipeline(m_pipeline.get());
+        m_cmdbuf->bindDescriptorSets(nbl::asset::EPBP_COMPUTE, m_pplnLayout.get(), 0, 1, &m_ds.get());
+        m_cmdbuf->dispatch(1, 1, 1);
+        m_cmdbuf->endDebugMarker();
+        m_cmdbuf->end();
+
+        video::IQueue::SSubmitInfo submitInfos[1] = {};
+        const video::IQueue::SSubmitInfo::SCommandBufferInfo cmdbufs[] = { {.cmdbuf = m_cmdbuf.get()} };
+        submitInfos[0].commandBuffers = cmdbufs;
+        const video::IQueue::SSubmitInfo::SSemaphoreInfo signals[] = { {.semaphore = m_semaphore.get(), .value = ++m_semaphoreCounter, .stageMask = asset::PIPELINE_STAGE_FLAGS::COMPUTE_SHADER_BIT} };
+        submitInfos[0].signalSemaphores = signals;
+
+        m_api->startCapture();
+        m_queue->submit(submitInfos);
+        m_api->endCapture();
+
+        m_device->waitIdle();
+        OutputStruct output;
+        std::memcpy(&output, static_cast<OutputStruct*>(m_outputBufferAllocation.memory->getMappedPointer()), sizeof(OutputStruct));
+        m_device->waitIdle();
+
+        return output;
+    }


why are we dispatching once per tests, could dispatch all tests in parallel (one invocation one test iteration)!

assigned @Przemog1 in next PR

devshgraphicsprogramming · 2025-11-28T09:44:39Z

73_Mortons/app_resources/common.hlsl

+	/*
+	void fillSecondTestValues(NBL_CONST_REF_ARG(InputTestValues) input)
+	{
+		uint64_t2 Vec2A = { input.coordX, input.coordY };
+		uint64_t2 Vec2B = { input.coordZ, input.coordW };
+
+		uint64_t3 Vec3A = { input.coordX, input.coordY, input.coordZ };
+		uint64_t3 Vec3B = { input.coordY, input.coordZ, input.coordW };
+
+		uint64_t4 Vec4A = { input.coordX, input.coordY, input.coordZ, input.coordW };
+		uint64_t4 Vec4B = { input.coordY, input.coordZ, input.coordW, input.coordX };
+
+		int64_t2 Vec2ASigned = int64_t2(Vec2A);
+		int64_t2 Vec2BSigned = int64_t2(Vec2B);
+
+		int64_t3 Vec3ASigned = int64_t3(Vec3A);
+		int64_t3 Vec3BSigned = int64_t3(Vec3B);
+
+		int64_t4 Vec4ASigned = int64_t4(Vec4A);
+		int64_t4 Vec4BSigned = int64_t4(Vec4B);
+
+		morton::code<false, fullBits_4, 4, emulated_uint64_t> morton_emulated_4A = morton::code<false, fullBits_4, 4, emulated_uint64_t>::create(Vec4A);
+		morton::code<true, fullBits_2, 2, emulated_uint64_t> morton_emulated_2_signed = morton::code<true, fullBits_2, 2, emulated_uint64_t>::create(Vec2ASigned);
+		morton::code<true, fullBits_3, 3, emulated_uint64_t> morton_emulated_3_signed = morton::code<true, fullBits_3, 3, emulated_uint64_t>::create(Vec3ASigned);
+		morton::code<true, fullBits_4, 4, emulated_uint64_t> morton_emulated_4_signed = morton::code<true, fullBits_4, 4, emulated_uint64_t>::create(Vec4ASigned);
+
+		output.mortonEqual_emulated_4 = uint32_t4(morton_emulated_4A.equal<false>(uint16_t4(Vec4B)));
+		
+		output.mortonUnsignedLess_emulated_4 = uint32_t4(morton_emulated_4A.lessThan<false>(uint16_t4(Vec4B)));
+		
+		mortonSignedLess_emulated_2 = uint32_t2(morton_emulated_2_signed.lessThan<false>(int32_t2(Vec2BSigned))); 
+		mortonSignedLess_emulated_3 = uint32_t3(morton_emulated_3_signed.lessThan<false>(int32_t3(Vec3BSigned))); 
+		mortonSignedLess_emulated_4 = uint32_t4(morton_emulated_4_signed.lessThan<false>(int16_t4(Vec4BSigned))); 
+
+		uint16_t castedShift = uint16_t(input.shift);
+
+		arithmetic_right_shift_operator<morton::code<true, fullBits_2, 2, emulated_uint64_t> > rightShiftSignedEmulated2;
+		mortonSignedRightShift_emulated_2 = rightShiftSignedEmulated2(morton_emulated_2_signed, castedShift); 
+		arithmetic_right_shift_operator<morton::code<true, fullBits_3, 3, emulated_uint64_t> > rightShiftSignedEmulated3;
+		mortonSignedRightShift_emulated_3 = rightShiftSignedEmulated3(morton_emulated_3_signed, castedShift); 
+		arithmetic_right_shift_operator<morton::code<true, fullBits_4, 4, emulated_uint64_t> > rightShiftSignedEmulated4;
+		mortonSignedRightShift_emulated_4 = rightShiftSignedEmulated4(morton_emulated_4_signed, castedShift); 
+	}
+	*/


@Fletterio whats this commented out block about ?

I think there was some fucked up reason that was preventing me from running all tests in a single shader (likely some DXC bug) so I think I was in the middle of moving the commented code to a different shader

ok so bug no longer there and we can remove this commented block of code?

Bug is probably still there, I thought I had reported it but it seems not. These are tests that if for some reason you add them to testCommon.hlsl it will fail to compile. I had temporarily commented them out here and the idea was to have many different test shaders so these could be tested apart. Specifically these would be tests for comparison operators for a 4D morton with 16bits per coord stored in an emulated uint64, and tests for the arithmetic right shift operator for a 2D, 3D or 4D morton backed by an emulated uint64.

devshgraphicsprogramming · 2025-11-28T09:45:05Z

73_Mortons/app_resources/common.hlsl

+#ifndef __HLSL_VERSION
+
+constexpr uint64_t smallBitsMask_2 = (uint64_t(1) << smallBits_2) - 1;
+constexpr uint64_t mediumBitsMask_2 = (uint64_t(1) << mediumBits_2) - 1;
+constexpr uint64_t fullBitsMask_2 = (uint64_t(1) << fullBits_2) - 1;
+
+constexpr uint64_t smallBitsMask_3 = (uint64_t(1) << smallBits_3) - 1;
+constexpr uint64_t mediumBitsMask_3 = (uint64_t(1) << mediumBits_3) - 1;
+constexpr uint64_t fullBitsMask_3 = (uint64_t(1) << fullBits_3) - 1;
+
+constexpr uint64_t smallBitsMask_4 = (uint64_t(1) << smallBits_4) - 1;
+constexpr uint64_t mediumBitsMask_4 = (uint64_t(1) << mediumBits_4) - 1;
+constexpr uint64_t fullBitsMask_4 = (uint64_t(1) << fullBits_4) - 1;
+
+#endif


are these variables used anywhere?

No clue what I had in mind here

ok @kevyuu delete them

Fletterio added 3 commits March 23, 2025 19:29

Morton code tests

f2ea51d

Morton codes creating properly

8f4e452

All tests passing, HLSL compiles fine!

0aedfd9

devshgraphicsprogramming reviewed Mar 29, 2025

View reviewed changes

XX_Mortons/main.cpp Outdated Show resolved Hide resolved

devshgraphicsprogramming reviewed Mar 29, 2025

View reviewed changes

CMakeLists.txt Outdated Show resolved Hide resolved

Fletterio added 3 commits April 1, 2025 15:44

Rename example

ea42d5b

Add tests for AddCarry and SUbBorrow intrinsics

2ba08a4

Disable intrinsic tests for uSUbBorrow for the time being, start copy…

f00bbf6

…ing 22_CppCOmpat to run tests

devshgraphicsprogramming reviewed Apr 8, 2025

View reviewed changes

22_CppCompat/CIntrinsicsTester.h Outdated Show resolved Hide resolved

devshgraphicsprogramming reviewed Apr 8, 2025

View reviewed changes

12_Mortons/app_resources/common.hlsl Outdated Show resolved Hide resolved

devshgraphicsprogramming reviewed Apr 8, 2025

View reviewed changes

12_Mortons/app_resources/common.hlsl Outdated Show resolved Hide resolved

devshgraphicsprogramming reviewed Apr 8, 2025

View reviewed changes

12_Mortons/app_resources/shader.hlsl Outdated Show resolved Hide resolved

Fletterio and others added 6 commits April 24, 2025 16:06

Added extensive tests for Morton codes

b2d87c3

Done with tests

c68c336

Merged master

d906bb2

Clarifying comment for blocker issue

f05dec4

Merge branch 'master' into mortons

b13fdc7

Reindex mortons example from 12 to 73

08c898d

devshgraphicsprogramming reviewed Nov 28, 2025

View reviewed changes

Global variable of hlsl to use NBL_CONSTEXPR_INLINE_NSPC_SCOPE_VAR

7f8dd73

Adds tests for the new Morton Code class #187

Are you sure you want to change the base?

Adds tests for the new Morton Code class #187

Conversation

Fletterio commented Mar 28, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

devshgraphicsprogramming Nov 28, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

devshgraphicsprogramming Nov 28, 2025 •

edited

Loading