Fix race-condition and visual artifacts issues (#4152)

* Fix race-condition and visual artifacts issues

In PerformanceProfiler::getProfiler() we return a static object
for the profiler implementation, this is not thread-safe, so change
it to thead_local.

There is still some visual artifacts when using slang as the shading
language. We don't know the root cause yet, but found out it's related
to our loop inversion algorithm. So stage this feature for now, and turn
it into an internal option and default off. We will re-enable it after
more investigation on this optimization.

File an new issue 4151 to track it.

* Add '-loop-inversion' to the few tests

8 files changed, 16 insertions, 11 deletions

diff --git a/slang.h b/slang.h
index 3f5ba5c31..3a3b5e620 100644
--- a/slang.h
+++ b/slang.h
@@ -945,6 +945,7 @@ extern "C"
             SaveStdLib,
             SaveStdLibBinSource,
             TrackLiveness,
+            LoopInversion,              // bool, enable loop inversion optimization
 
             // Deprecated
             ParameterBlocksUseRegisterSpaces,
diff --git a/source/core/slang-performance-profiler.cpp b/source/core/slang-performance-profiler.cpp
index f5f74f99d..5b62cbf73 100644
--- a/source/core/slang-performance-profiler.cpp
+++ b/source/core/slang-performance-profiler.cpp
@@ -46,7 +46,7 @@ namespace Slang
 
     PerformanceProfiler* Slang::PerformanceProfiler::getProfiler()
     {
-        static PerformanceProfilerImpl profiler = PerformanceProfilerImpl();
+        thread_local static PerformanceProfilerImpl profiler = PerformanceProfilerImpl();
         return &profiler;
     }
 }
diff --git a/source/slang/slang-lower-to-ir.cpp b/source/slang/slang-lower-to-ir.cpp
index 429660cac..9de1833a0 100644
--- a/source/slang/slang-lower-to-ir.cpp
+++ b/source/slang/slang-lower-to-ir.cpp
@@ -10839,8 +10839,10 @@ RefPtr<IRModule> generateIRForTranslationUnit(
     // - If sccp is unable to eliminate the outer 'if' then we end up with
     //   duplicated code the the conditional value. Users don't tend to put
     //   huge gobs of code in the conditional expression in loops however.
-
-    invertLoops(module);
+    if (compileRequest->getLinkage()->m_optionSet.getBoolOption(CompilerOptionName::LoopInversion))
+    {
+        invertLoops(module);
+    }
 
     // Next, attempt to promote local variables to SSA
     // temporaries and do basic simplifications.
diff --git a/source/slang/slang-options.cpp b/source/slang/slang-options.cpp
index fc2fd42c2..1f27c5efa 100644
--- a/source/slang/slang-options.cpp
+++ b/source/slang/slang-options.cpp
@@ -542,6 +542,7 @@ void initCommandOptions(CommandOptions& options)
         { OptionKind::SaveStdLibBinSource, "-save-stdlib-bin-source","-save-stdlib-bin-source <filename>", "Same as -save-stdlib but output "
         "the data as a C array.\n"},
         { OptionKind::TrackLiveness, "-track-liveness", nullptr, "Enable liveness tracking. Places SLANG_LIVE_START, and SLANG_LIVE_END in output source to indicate value liveness." },
+        { OptionKind::LoopInversion, "-loop-inversion", nullptr, "Enable loop inversion in the code-gen optimization. Default is off" },
     };
     _addOptions(makeConstArrayView(internalOpts), options);
 
@@ -1689,6 +1690,7 @@ SlangResult OptionsParser::_parse(
             case OptionKind::IncompleteLibrary:
             case OptionKind::NoHLSLBinding:
             case OptionKind::NoHLSLPackConstantBufferElements:
+            case OptionKind::LoopInversion:
                 linkage->m_optionSet.set(optionKind, true); break;
                 break;
             case OptionKind::MatrixLayoutRow:
diff --git a/tests/autodiff/path-tracer/pt-loop.slang b/tests/autodiff/path-tracer/pt-loop.slang
index 2e865b80a..ac8bf763d 100644
--- a/tests/autodiff/path-tracer/pt-loop.slang
+++ b/tests/autodiff/path-tracer/pt-loop.slang
@@ -1,7 +1,7 @@
 //Tests automatic synthesis of Differential type requirement.
 
-//TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -shaderobj -output-using-type
-//TEST(compute, vulkan):COMPARE_COMPUTE_EX:-vk -compute -shaderobj -output-using-type
+//TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -shaderobj -output-using-type -Xslang -loop-inversion
+//TEST(compute, vulkan):COMPARE_COMPUTE_EX:-vk -compute -shaderobj -output-using-type -Xslang -loop-inversion
 
 //TEST_INPUT:ubuffer(data=[0 0 0], stride=4):out,name=outputBuffer
 RWStructuredBuffer<float> outputBuffer;
diff --git a/tests/autodiff/reverse-while-loop-2.slang b/tests/autodiff/reverse-while-loop-2.slang
index 9ad9ac466..70b9d5a13 100644
--- a/tests/autodiff/reverse-while-loop-2.slang
+++ b/tests/autodiff/reverse-while-loop-2.slang
@@ -1,6 +1,6 @@
-//TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -shaderobj -output-using-type
-//TEST(compute, vulkan):COMPARE_COMPUTE_EX:-vk -compute -shaderobj -output-using-type
-//TEST(compute):COMPARE_COMPUTE_EX:-cpu -compute -output-using-type -shaderobj
+//TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -shaderobj -output-using-type -Xslang -loop-inversion
+//TEST(compute, vulkan):COMPARE_COMPUTE_EX:-vk -compute -shaderobj -output-using-type -Xslang -loop-inversion
+//TEST(compute):COMPARE_COMPUTE_EX:-cpu -compute -output-using-type -shaderobj -Xslang -loop-inversion
 
 //TEST_INPUT:ubuffer(data=[0 0 0 0 0], stride=4):out,name=outputBuffer
 RWStructuredBuffer<float> outputBuffer;
diff --git a/tests/cross-compile/loop-attribs.slang b/tests/cross-compile/loop-attribs.slang
index 4336698c0..dca3f2331 100644
--- a/tests/cross-compile/loop-attribs.slang
+++ b/tests/cross-compile/loop-attribs.slang
@@ -1,7 +1,7 @@
 // loop-attribs.slang
 // Test that loop attributes are correctly emitted to the resulting HLSL.
 
-//TEST:CROSS_COMPILE:-target dxil-assembly  -entry main -stage fragment -profile sm_6_0 
+//TEST:CROSS_COMPILE:-target dxil-assembly  -entry main -stage fragment -profile sm_6_0 -loop-inversion
 
 float4 main() : SV_Target
 {
@@ -16,4 +16,4 @@ float4 main() : SV_Target
         sum += float(j);
 
     return float4(sum, 0, 0, 0);
-}
\ No newline at end of file
+}
diff --git a/tests/ir/loop-inversion.slang b/tests/ir/loop-inversion.slang
index 7e218a62a..7611c4062 100644
--- a/tests/ir/loop-inversion.slang
+++ b/tests/ir/loop-inversion.slang
@@ -1,4 +1,4 @@
-//TEST():SIMPLE(filecheck=CHECK):-entry computeMain -stage compute -line-directive-mode none -target hlsl
+//TEST():SIMPLE(filecheck=CHECK):-entry computeMain -stage compute -line-directive-mode none -target hlsl -loop-inversion
 //TEST(compute):COMPARE_COMPUTE(filecheck-buffer=OUT):-shaderobj -output-using-type
 //TEST(compute):COMPARE_COMPUTE(filecheck-buffer=OUT):-dx12 -use-dxil -shaderobj -output-using-type
 //TEST(compute):COMPARE_COMPUTE(filecheck-buffer=OUT):-cpu -shaderobj -output-using-type