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
---
 tests/bugs/gh-7441.slang | 74 ++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 74 insertions(+)
 create mode 100644 tests/bugs/gh-7441.slang

(limited to 'tests/bugs')

diff --git a/tests/bugs/gh-7441.slang b/tests/bugs/gh-7441.slang
new file mode 100644
index 000000000..94ca7965b
--- /dev/null
+++ b/tests/bugs/gh-7441.slang
@@ -0,0 +1,74 @@
+//TEST(compute):COMPARE_COMPUTE_EX(filecheck-buffer=BUFFER):-cuda -compute -output-using-type
+//TEST(compute):COMPARE_COMPUTE_EX(filecheck-buffer=BUFFER-MTL):-metal -compute -output-using-type
+//TEST:REFLECTION(filecheck=REFLECT):-stage compute -entry computeMain -target cuda -no-codegen
+//TEST:REFLECTION(filecheck=REFLECT-MTL):-stage compute -entry computeMain -target metal -no-codegen
+
+
+// Test struct for bool layout analysis
+struct TestType
+{
+    uint value;
+    bool f_bool;
+    bool1 f_bool1;
+    bool pad1;
+    bool2 f_bool2;
+    bool pad2;
+    bool3 f_bool3;
+    bool4 f_bool4;
+    uint END;
+};
+
+//TEST_INPUT:ubuffer(data=[0 0 0 0 0 0], stride=4):out,name=buffer
+RWStructuredBuffer<TestType> buffer;
+
+[shader("compute")]
+[numthreads(1, 1, 1)]
+void computeMain(uint3 tid: SV_DispatchThreadID)
+{
+    uint i = tid.x;
+
+    // Initialize test data
+    buffer[i].value = 7;
+    buffer[i].f_bool = true;
+    buffer[i].f_bool1 = bool1(true);
+    buffer[i].pad1 = false;
+    buffer[i].f_bool2 = bool2(true, true);
+    buffer[i].pad2 = false;
+    buffer[i].f_bool3 = bool3(true, false, true);
+    buffer[i].f_bool4 = bool4(true, false, true, false);
+    buffer[i].END = 0x12345678;
+}
+
+// BUFFER: 7
+// BUFFER-NEXT: 101
+// BUFFER-NEXT: 1000101
+// BUFFER-NEXT: 100
+// BUFFER-NEXT: 10001
+// BUFFER-NEXT: 12345678
+
+// Expected output for Metal (different struct layout)
+// BUFFER-MTL: 7
+// BUFFER-MTL-NEXT: 101
+// BUFFER-MTL-NEXT: 101
+// BUFFER-MTL-NEXT: 10001
+// BUFFER-MTL-NEXT: 10001
+// BUFFER-MTL-NEXT: 12345678
+
+// REFLECT:         "name": "f_bool1",
+// REFLECT:         "binding": {"kind": "uniform", "offset": 5, "size": 1, "elementStride": 1}
+// REFLECT:         "name": "f_bool2",
+// REFLECT:         "binding": {"kind": "uniform", "offset": 8, "size": 2, "elementStride": 1}
+// REFLECT:         "name": "f_bool3",
+// REFLECT:         "binding": {"kind": "uniform", "offset": 11, "size": 3, "elementStride": 1}
+// REFLECT:         "name": "f_bool4",
+// REFLECT:         "binding": {"kind": "uniform", "offset": 16, "size": 4, "elementStride": 1}
+
+// Metal-specific reflection (different bool3 layout)
+// REFLECT-MTL:         "name": "f_bool1",
+// REFLECT-MTL:         "binding": {"kind": "uniform", "offset": 5, "size": 1, "elementStride": 1}
+// REFLECT-MTL:         "name": "f_bool2",
+// REFLECT-MTL:         "binding": {"kind": "uniform", "offset": 8, "size": 2, "elementStride": 1}
+// REFLECT-MTL:         "name": "f_bool3",
+// REFLECT-MTL:         "binding": {"kind": "uniform", "offset": 12, "size": 4, "elementStride": 1}
+// REFLECT-MTL:         "name": "f_bool4",
+// REFLECT-MTL:         "binding": {"kind": "uniform", "offset": 16, "size": 4, "elementStride": 1}
-- 
cgit v1.2.3