changed trunc and round to simpler forms, instead of the function call that was used before. This improved speed by decent amount

Author: aous72

src/apps/apps.xcodeproj/project.xcworkspace/xcuserdata/aous.xcuserdatad/UserInterfaceState.xcuserstate

@@ -40,7 +40,7 @@
       </Testables>
    </TestAction>
    <LaunchAction
-      buildConfiguration = "Debug"
+      buildConfiguration = "Release"
       selectedDebuggerIdentifier = "Xcode.DebuggerFoundation.Debugger.LLDB"
       selectedLauncherIdentifier = "Xcode.DebuggerFoundation.Launcher.LLDB"
       launchStyle = "0"

src/apps/apps.xcodeproj/xcshareddata/xcschemes/ojph_expand.xcscheme

@@ -52,5 +52,37 @@
             landmarkType = "0">
          </BreakpointContent>
       </BreakpointProxy>
+      <BreakpointProxy
+         BreakpointExtensionID = "Xcode.Breakpoint.FileBreakpoint">
+         <BreakpointContent
+            uuid = "37CC6DAA-5E0A-44BB-9EE3-BBE83978793C"
+            shouldBeEnabled = "No"
+            ignoreCount = "0"
+            continueAfterRunningActions = "No"
+            filePath = "../core/transform/ojph_transform_avx.cpp"
+            startingColumnNumber = "9223372036854775807"
+            endingColumnNumber = "9223372036854775807"
+            startingLineNumber = "288"
+            endingLineNumber = "288"
+            landmarkName = "avx_irrev_horz_wvlt_bwd_tx(dst, lsrc, hsrc, width, even)"
+            landmarkType = "9">
+         </BreakpointContent>
+      </BreakpointProxy>
+      <BreakpointProxy
+         BreakpointExtensionID = "Xcode.Breakpoint.FileBreakpoint">
+         <BreakpointContent
+            uuid = "8046ADD0-A6AF-47B8-A639-114F0AC9B88D"
+            shouldBeEnabled = "No"
+            ignoreCount = "0"
+            continueAfterRunningActions = "No"
+            filePath = "../core/codestream/ojph_codestream.cpp"
+            startingColumnNumber = "9223372036854775807"
+            endingColumnNumber = "9223372036854775807"
+            startingLineNumber = "3031"
+            endingLineNumber = "3031"
+            landmarkName = "subband::push_line()"
+            landmarkType = "7">
+         </BreakpointContent>
+      </BreakpointProxy>
    </Breakpoints>
 </Bucket>

src/apps/apps.xcodeproj/xcuserdata/aous.xcuserdatad/xcdebugger/Breakpoints_v2.xcbkptlist

@@ -3028,7 +3028,7 @@ namespace ojph {
         si32 *dp = lines->i32;
         for (int i = band_rect.siz.w; i > 0; --i)
         {
-          si32 t = (si32)trunc(*sp++ * delta_inv);
+          si32 t = ojph_trunc(*sp++ * delta_inv);
           si32 val = t >= 0 ? t : -t;
           si32 sign = t >= 0 ? 0 : 0x80000000;
           *dp++ = sign | val;

src/core/codestream/ojph_codestream.cpp

@@ -41,6 +41,7 @@
 
 #include <cstdio>
 #include <cstdint>
+#include <cmath>
 
 #include "ojph_defs.h"
 
@@ -122,6 +123,29 @@ namespace ojph {
   #endif
   }
 
+  ////////////////////////////////////////////////////////////////////////////
+  static inline si32 ojph_round(float val)
+  {
+  #ifdef OJPH_COMPILER_MSVC
+    return (si32)(val + (val >= 0.0f ? 0.5f : -0.5f));
+  #elif (defined OJPH_COMPILER_GNUC)
+    return (si32)(val + (val >= 0.0f ? 0.5f : -0.5f));
+  #else
+    return (si32)round(float val);
+  #endif
+  }
+
+  ////////////////////////////////////////////////////////////////////////////
+  static inline si32 ojph_trunc(float val)
+  {
+  #ifdef OJPH_COMPILER_MSVC
+    return (si32)(val);
+  #elif (defined OJPH_COMPILER_GNUC)
+    return (si32)(val);
+  #else
+    return (si32)trunc(float val);
+  #endif
+  }
 
   ////////////////////////////////////////////////////////////////////////////
   // constants

src/core/common/ojph_arch.h

@@ -196,15 +196,15 @@ namespace ojph {
                                        int width)
     {
       for (int i = width; i > 0; --i)
-        *dp++ = (si32)round((*sp++ + 0.5f) * mul);
+        *dp++ = ojph_round((*sp++ + 0.5f) * mul);
     }
 
     //////////////////////////////////////////////////////////////////////////
     void gen_cnvrt_float_to_si32(const float *sp, si32 *dp, float mul,
                                  int width)
     {
       for (int i = width; i > 0; --i)
-        *dp++ = (si32)round(*sp++ * mul);
+        *dp++ = ojph_round(*sp++ * mul);
     }
 
     //////////////////////////////////////////////////////////////////////////