ti@rapidis.com wrote:

> I've done both. Though mostly instrumetation-based, since I was
> interested in measures for which specific lines in a few big functions
> were most costly.

> Quantify does this easily, and I don't think VS
> 2005's profiler does it.

ti@rapidis.com wrote:

> I'm trying to figure out what the JIT will inline at what it won't. 2
> questions:
>
> 1. Will the JIT inline this function? (from actual software):
>
> public int GetEdgeCostIndex(bool isConnector, bool isForward)

> 2. How do I get CorDbg to show dissaembly for the JIT's optimized code?

> I tried the sample program below. The TestMethod() function should a be
> prime candidate for inlining, but cordbg doesn't show that.
>
> class Program
> {
> static void Main(string[] args)
> {
> for (int i = 0; i < 1000000; i++)
> Program.TestMethod();
>
> System.Console.WriteLine("x");
> }
> public static void TestMethod()
> {
> return;
> }
> }

>>> 00ca0070 833d8410280200 cmp dword ptr ds:[2281084h],0

ti@rapidis.com wrote:

> In that context, do you know of a good way to get a brakedown of which
> statements in a big function are the time consuming ones? The Profiler
> in VSTE will only tell me the time for the function itself.

> 2. How do I get CorDbg to show dissaembly for the JIT's optimized code?
> I tried the sample program below. The TestMethod() function should a be
> prime candidate for inlining, but cordbg doesn't show that.

<ti@rapidis.com> wrote in message
news:1153422653.173475.112870@s13g2000cwa.googlegroups.com...
>A related issue:
>
> I'm trying to figure out what the JIT will inline at what it won't. 2
> questions:
>
> 1. Will the JIT inline this function? (from actual software):
>
> public int GetEdgeCostIndex(bool isConnector, bool isForward)
> {
> if (isConnector)
> {
> if (isForward)
> return ForwardConnectorIndex;
> else
> return BackwardConnectorIndex;
> }
> else
> {
> if (isForward)
> return ForwardLinkIndex;
> else
> return BackwardLinkIndex;
> }
> }
>
> 2. How do I get CorDbg to show dissaembly for the JIT's optimized code?
> I tried the sample program below. The TestMethod() function should a be
> prime candidate for inlining, but cordbg doesn't show that.
>
> class Program
> {
> static void Main(string[] args)
> {
> for (int i = 0; i < 1000000; i++)
> Program.TestMethod();
>
> System.Console.WriteLine("x");
> }
> public static void TestMethod()
> {
> return;
> }
> }
>
> Below is seen a session in cordbg with a release build of the above
> program. Note that the function is not inlined. What gives??
>
> Microsoft (R) Common Language Runtime Test Debugger Shell Version
> 2.0.50727.42 (RTM.050727-4200)
> Copyright (C) Microsoft Corporation. All rights reserved.
>
> (cordbg) mode JitOptimizations 1
> JIT's will produce optimized code
>
>
> (cordbg) run csinlined.exe
> Process 3856/0xf10 created.
> Warning: couldn't load symbols for
> C:\WINDOWS\assembly\GAC_32\mscorlib\2.0.0.0__b77a5c561934e089\mscorlib.dll
> [thread 0x7b8] Thread created.
>
> 007: for (int i = 0; i < 1000000; i++)
> (cordbg) dis 100
> Function CsInlined.Program.Main (code starts at 0x11a9ba8).
> Offsets are relative to function start.
> [0000] push esi
> [0001] push eax
> [0002] mov dword ptr [esp],ecx
> [0005] cmp dword ptr ds:[00962DDCh],0
> [000c] je 00000007
> [000e] call 78EE8748
> [0013] xor esi,esi
> *[0015] xor esi,esi
> [0017] nop
> [0018] jmp 00000009
> [001a] call dword ptr ds:[00963044h]
> [0020] inc esi
> [0021] cmp esi,0F4240h
> [0027] jl FFFFFFF3
> [0029] mov ecx,dword ptr ds:[02302098h]
> [002f] call dword ptr ds:[011D1364h]
> [0035] nop
> [0036] pop ecx
> [0037] pop esi
> [0038] ret
> (cordbg) ss
> [0021] cmp esi,0F4240h
> (cordbg) ss
> [0027] jl FFFFFFF3
> (cordbg) ss
> [001a] call dword ptr ds:[00963044h]
> (cordbg) ss
> [000e] nop
> (cordbg) dis 100
> Function CsInlined.Program.TestMethod (code starts at 0x11a9bf8).
> Offsets are relative to function start.
> [0000] cmp dword ptr ds:[00962DDCh],0
> [0007] je 00000007
> [0009] call 78EE86FD
> *[000e] nop
> [000f] ret
> (cordbg)
>