From 28758e0e427ceca196937dc90efe3ab1cb35bd70 Mon Sep 17 00:00:00 2001
From: Ellie Hermaszewska <ellieh@nvidia.com>
Date: Thu, 17 Jul 2025 14:59:33 +0800
Subject: Perf improvements to IR serialization (#7751)

* option to use riff as serialization backend

* option to use riff as serialization backend

* perf

* shuffle code

* perf improvements to deserialization

* formatting

* remove bit_cast

* correct IR verification

* neaten serialized format

* fix peek module info

* formatting

* remove temporary profiling code

* cleanup

* fix wasm build

* more explicit sizes

* deserialize via fossil on 32 bit wasm

* Make serialized modules Int size agnostic

* reorder stable names to allow range based check for 64 bit constants

* format

* review comments

* fix build

* fix

* c++17 compat slang-common.h
---
 source/core/slang-common.h                 | 19 +++++++++++++++++--
 source/core/slang-performance-profiler.cpp | 21 ++++++++++++++-------
 2 files changed, 31 insertions(+), 9 deletions(-)

(limited to 'source/core')

diff --git a/source/core/slang-common.h b/source/core/slang-common.h
index 9248e250f..015c3233c 100644
--- a/source/core/slang-common.h
+++ b/source/core/slang-common.h
@@ -5,7 +5,9 @@
 
 #include <assert.h>
 #include <cstddef>
+#include <cstring>
 #include <stdint.h>
+#include <type_traits>
 
 #define VARIADIC_TEMPLATE
 
@@ -199,6 +201,19 @@ inline bool isBitSet(T value, T bitToTest)
     static_assert(sizeof(T) <= sizeof(uint32_t), "Only support up to 32 bit enums");
     return (T)((uint32_t)value & (uint32_t)bitToTest) == bitToTest;
 }
+
+template<typename To, typename From>
+typename std::enable_if_t<
+    sizeof(To) == sizeof(From) && std::is_trivially_copyable_v<From> &&
+        std::is_trivially_copyable_v<To> && std::is_trivially_constructible_v<To>,
+    To>
+bitCast(const From& src)
+{
+    To dst;
+    std::memcpy(&dst, &src, sizeof(To));
+    return dst;
+}
+
 } // namespace Slang
 
 // SLANG_DEFER
@@ -349,7 +364,7 @@ public:
 #define SLANG_ASSERT(VALUE)               \
     do                                    \
     {                                     \
-        if (!(VALUE))                     \
+        if (!(VALUE)) [[unlikely]]        \
             SLANG_ASSERT_FAILURE(#VALUE); \
     } while (0)
 #else
@@ -357,7 +372,7 @@ public:
 #endif
 
 #define SLANG_RELEASE_ASSERT(VALUE) \
-    if (VALUE)                      \
+    if (VALUE) [[likely]]           \
     {                               \
     }                               \
     else                            \
diff --git a/source/core/slang-performance-profiler.cpp b/source/core/slang-performance-profiler.cpp
index 9bc532680..bd046f4cc 100644
--- a/source/core/slang-performance-profiler.cpp
+++ b/source/core/slang-performance-profiler.cpp
@@ -35,15 +35,22 @@ public:
         char buffer[512];
         for (const auto& func : data)
         {
-            memset(buffer, 0, sizeof(buffer));
-            snprintf(buffer, sizeof(buffer), "[*] %30s", func.key);
-            out << buffer << " \t";
-            auto milliseconds =
-                std::chrono::duration_cast<std::chrono::milliseconds>(func.value.duration);
-            out << func.value.invocationCount << " \t"
-                << static_cast<uint64_t>(milliseconds.count()) << "ms\n";
+            auto microseconds =
+                std::chrono::duration_cast<std::chrono::microseconds>(func.value.duration);
+            double milliseconds = microseconds.count() / 1000.0;
+            snprintf(
+                buffer,
+                sizeof(buffer),
+                "[*] %30s \t%d \t%8.2fms\n",
+                func.key,
+                func.value.invocationCount,
+                milliseconds);
+
+            out << buffer;
         }
     }
+
+
     virtual void clear() override { data.clear(); }
     virtual void dispose() override { data = decltype(data)(); }
 };
-- 
cgit v1.2.3