From bdda8a90cdd44ca822b21233ac988f92d1f20826 Mon Sep 17 00:00:00 2001
From: "Harsh Aggarwal (NVIDIA)" <haaggarwal@nvidia.com>
Date: Fri, 1 Aug 2025 14:48:53 +0530
Subject: Fix 7441: CUDA boolean vector layout to use 1-byte elements (#7862)

* Fix 7441: CUDA boolean vector layout to use 1-byte elements

Boolean vectors (bool1, bool2, bool3, bool4) were incorrectly implemented
as integer-based types using 4 bytes per element instead of actual 1-byte
boolean elements on CUDA targets.

Changes:
- Update CUDA prelude to define boolean vectors as structs with bool fields
  instead of typedef aliases to integer vectors
- Implement CUDALayoutRulesImpl::GetVectorLayout to use 1-byte alignment
  for boolean vectors, matching actual CUDA memory layout behavior
- Update make_bool functions to populate struct fields correctly

This ensures boolean vectors have the same memory layout as bool[4] arrays:
- bool1: 1 byte (was 4 bytes)
- bool2: 2 bytes (was 8 bytes)
- bool3: 3 bytes (was 12 bytes)
- bool4: 4 bytes (was 16 bytes)

Fixes memory layout mismatch between Slang reflection API and actual
CUDA compilation, achieving 75% memory savings for boolean vector usage.

* Fix CI issues -

Add and update associated functions and operators

* Make boolX same as uchar

* Use align construct on struct for boolX

* Improve Test case for robust alignment checks

* Formatting

* Disable selected slangpy tests

* add metal check which is slightly different than cuda

* Test-1

* Test-2

* Test-3

* Test-4

* ReflectionChange

* cleanup and update

* _slang_select with plain bool is needed for reverse-loop-checkpoint-test
---
 source/slang/slang-reflection-json.cpp | 6 +++++-
 source/slang/slang-type-layout.cpp     | 8 --------
 2 files changed, 5 insertions(+), 9 deletions(-)

(limited to 'source')

diff --git a/source/slang/slang-reflection-json.cpp b/source/slang/slang-reflection-json.cpp
index c253c4d66..26cdafa46 100644
--- a/source/slang/slang-reflection-json.cpp
+++ b/source/slang/slang-reflection-json.cpp
@@ -64,6 +64,7 @@ static void emitReflectionVarBindingInfoJSON(
     SlangParameterCategory category,
     SlangUInt index,
     SlangUInt count,
+    SlangUInt stride,
     SlangUInt space = 0)
 {
     if (category == SLANG_PARAMETER_CATEGORY_UNIFORM)
@@ -73,6 +74,8 @@ static void emitReflectionVarBindingInfoJSON(
         writer << "\"offset\": " << index;
         writer << ", ";
         writer << "\"size\": " << count;
+        writer << ", ";
+        writer << "\"elementStride\": " << stride;
     }
     else
     {
@@ -197,6 +200,7 @@ static void emitReflectionVarBindingInfoJSON(
             auto index = var->getOffset(category);
             auto space = var->getBindingSpace(category);
             auto count = typeLayout->getSize(category);
+            auto elementStride = typeLayout->getElementStride(category);
 
             // Query the paramater usage for the specified entry point.
             // Note: both `request` and `entryPointIndex` may be invalid here, but that should just
@@ -216,7 +220,7 @@ static void emitReflectionVarBindingInfoJSON(
 
             writer << "{";
 
-            emitReflectionVarBindingInfoJSON(writer, category, index, count, space);
+            emitReflectionVarBindingInfoJSON(writer, category, index, count, elementStride, space);
 
             if (usedAvailable)
             {
diff --git a/source/slang/slang-type-layout.cpp b/source/slang/slang-type-layout.cpp
index 3982b579d..275a178d0 100644
--- a/source/slang/slang-type-layout.cpp
+++ b/source/slang/slang-type-layout.cpp
@@ -582,14 +582,6 @@ struct CUDALayoutRulesImpl : DefaultLayoutRulesImpl
         SimpleLayoutInfo elementInfo,
         size_t elementCount) override
     {
-        // Special case bool
-        if (elementType == BaseType::Bool)
-        {
-            SimpleLayoutInfo fixInfo(elementInfo);
-            fixInfo.size = sizeof(int32_t);
-            fixInfo.alignment = sizeof(int32_t);
-            return GetVectorLayout(BaseType::Int, fixInfo, elementCount);
-        }
 
         const auto elementSize = elementInfo.size.getFiniteValue();
 
-- 
cgit v1.2.3