From 2cad8341019659a65fc6e94992165b3d7b7a37db Mon Sep 17 00:00:00 2001 From: johannst Date: Sun, 23 Mar 2025 23:51:01 +0000 Subject: deploy: 916b73bee95494c205ba67e4a50e6a525afc3a3c --- print.html | 346 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 335 insertions(+), 11 deletions(-) (limited to 'print.html') diff --git a/print.html b/print.html index 28df4d6..ef23c83 100644 --- a/print.html +++ b/print.html @@ -2829,6 +2829,7 @@ renice -n 5 -p PID
  • callgrind
  • valgrind
  • vtune
    • +
  • tracy

    • /usr/bin/time(1)

    # statistics of process run
@@ -3276,6 +3277,231 @@ main: main.c
 vtune: main
 	$(VTUNE)/bin64/vtune -collect hotspots -start-paused -- ./main
    +

    tracy(1)

    +

    Tracy is a frame profiler, supporting manual code instrumentation +and providing a sampling profiler.

    +

    One can either record and visualize the profiling data live using +tracy-profiler or record the profiling data to a file using tracy-capture.

    +
    tracy-profiler [file] [-p port]
+
+tracy-capture -o file [-f] [-p port]
+    -f   overwrite <file> if it exists
+
    +

    Example

    +

    The example showcases different cases:

    +
      +
    1. Use tracy from a single binary. In that case the TracyClient.cpp can be +directly linked / included in the instrumented binary.
      2. +
    2. Use tracy from different binaries (eg main executable + shared library). In +this case the TracyClient.cpp should be compiled into its own shared +library, such that there is a single tracy client.
      3. +
    3. Use tracy from different binaries on windows. In this case the +TracyClient.cpp must be compiled again into a separate shared library, +while defining TRACY_EXPORTS. The code being instrumented must be compiled +with TRACY_IMPORTS defined.
      4. +
    +

    An instrumented c++ example:

    +
    #include <chrono>
+#include <thread>
+
+#include <tracy/Tracy.hpp>
+
+#ifdef USE_FOO
+extern "C" void foo_comp_hook(int64_t);
+#endif
+
+void init() {
+  // Create a named zone (active for the current scope).
+  // Name will be used when rendering the zone in the thread timeline.
+  ZoneScopedN("init()");
+  // Set explicit color for the rendered zone.
+  ZoneColor(0xff0000);
+
+  std::this_thread::sleep_for(std::chrono::seconds(1));
+}
+
+void comp(const char* name) {
+  // Track call count.
+  static int64_t ccnt = 0;
+  ccnt += 1;
+
+  // Create an unnamed zone for the current scope.
+  ZoneScoped;
+  // Name the zone by formatting the name dynamically.
+  // This name is shown for the zone in the thread timeline, however
+  // in the zone statistics they are all accounted under one common
+  // zone "comp".
+  ZoneNameF("comp(%s)", name);
+  // Additional text to attach to the zone.
+  ZoneTextF("text(%s)", name);
+  // Additional value to attach to the zone measurement.
+  ZoneValue(ccnt);
+
+  // Statistics for dynamic names, text and values can be looked at in the zone
+  // statistics.There measurements can be grouped by different categories.
+
+  // Add a simple plot.
+  TracyPlot("comp-plot", ccnt % 4);
+
+  std::this_thread::sleep_for(std::chrono::milliseconds(100));
+
+#ifdef USE_FOO
+  foo_comp_hook(ccnt);
+#endif
+}
+
+void post_comp() {
+  // Create an unnamed zone for the current scope and capture callstack (max
+  // depth 10). Capturing callstack requires platform with TRACY_HAS_CALLSTACK
+  // support.
+  ZoneScopedS(10);
+  // Name the zone, w/o formatting.
+  const char name[] = "post_comp()";
+  ZoneName(name, sizeof(name));
+
+  // Add trace messages to the timeline.
+  TracyMessageL("start sleep in post_comp()");
+  std::this_thread::sleep_for(std::chrono::milliseconds(50));
+  TracyMessageL("end sleep in post_comp()");
+}
+
+void fini() {
+  // Create a named zone with an explicit color.
+  ZoneScopedNC("fini()", 0x00ff00);
+  std::this_thread::sleep_for(std::chrono::seconds(1));
+}
+
+int main() {
+  // Create a named zone.
+  ZoneScopedN("main()");
+
+  init();
+
+  int step = 0;
+  while (step++ < 10) {
+    // Create a frame message, this start a new frame with the name
+    // "step" and end the previous frame with the name "step".
+    FrameMarkNamed("step");
+    // Create a named scope.
+    ZoneScopedN("step()");
+    comp("a");
+    comp("b");
+    comp("c");
+    post_comp();
+  }
+
+  fini();
+}
+
    +

    An instrumented c example:

    +
    #include <stdint.h>
+#include <inttypes.h>
+#include <stdio.h>
+
+#include <tracy/TracyC.h>
+
+static void comp_helper(int64_t i) {
+  char buf[64];
+  int cnt = snprintf(buf, sizeof(buf), "helper(%" PRId64 ")", i);
+
+  // Create an active unnamed zone.
+  TracyCZone(ctx, 1);
+
+  // Name the zone.
+  TracyCZoneName(ctx, buf, cnt);
+  // Add custom text to the zone measurement.
+  TracyCZoneText(ctx, buf, cnt);
+  // Add custom value to the zone measurement.
+  TracyCZoneValue(ctx, i);
+
+  for (int ii = 0; ii < i * 100000; ++ii) {
+    /* fake work */
+  }
+
+  // End the zone measurement.
+  TracyCZoneEnd(ctx);
+}
+
+void foo_comp_hook(int64_t cnt) {
+  // Create an active named zone.
+  TracyCZoneN(ctx, "foo", 1);
+
+  for (int i = 0; i < cnt; ++i) {
+    // Plot value.
+    TracyCPlot("foo_comp_hook", cnt + i);
+
+    comp_helper(i);
+  }
+
+  // Configure plot "foo", probably best done once during initialization..
+  TracyCPlotConfig("foo", TracyPlotFormatNumber, 1 /* step */, 1 /* fill */,
+                   0xff0000);
+  // Plot value.
+  TracyCPlot("foo", cnt);
+
+  // End the zone measurement.
+  TracyCZoneEnd(ctx);
+}
+
    +

    Raw build commands to demonstrate compiling tracy w/o cmake, in case we need +to integrate it into a different build system.

    +
    B := BUILD
+
+main: $(B)/main-static $(B)/main-dynamic $(B)/main-dynamic-win
+tracy: $(B)/tracy
+.PHONY: main tracy
+
+# -- TRACY STATIC ---------------------------------------------------------------
+
+$(B)/main-static: main.cpp | $(B)
+	clang++ -DTRACY_ENABLE -I$(B)/tracy/public -o $@ $^ $(B)/tracy/public/TracyClient.cpp
+
+# -- TRACY DYNAMIC --------------------------------------------------------------
+
+$(B)/main-dynamic: main.cpp $(B)/foo.so $(B)/TracyClient.so | $(B)
+	clang++ -DTRACY_ENABLE -I$(B)/tracy/public -DUSE_FOO -o $@ $^
+
+$(B)/foo.so: foo.c $(B)/TracyClient.so
+	clang -DTRACY_ENABLE -I$(B)/tracy/public -fPIC -shared -o $@ $^
+
+$(B)/TracyClient.so: $(B)/tracy/public/TracyClient.cpp
+	clang++ -DTRACY_ENABLE -I$(B)/tracy/public -fPIC -shared -o $@ $^
+
+# -- TRACY DYNAMIC WINDOWS ------------------------------------------------------
+
+$(B)/main-dynamic-win: main.cpp $(B)/foo.dll $(B)/TracyClient.dll
+	@# eg run with wine
+	zig c++ -target x86_64-windows -DTRACY_ENABLE -DTRACY_IMPORTS -DUSE_FOO -o $@ $^ -I $(B)/tracy/public
+
+$(B)/foo.dll: foo.c $(B)/TracyClient.dll
+	zig c++ -target x86_64-windows -DTRACY_ENABLE -DTRACY_IMPORTS -fPIC -shared -o $@ $^ -I $(B)/tracy/public
+
+$(B)/TracyClient.dll: $(B)/tracy/public/TracyClient.cpp
+	@# win libs from 'pragma comment(lib, ..)'
+	zig c++ -target x86_64-windows -DTRACY_ENABLE -DTRACY_EXPORTS -fPIC -shared -o $@ $^ -lws2_32 -ldbghelp -ladvapi32 -luser32
+
+# -- TRACY ----------------------------------------------------------------------
+
+# Get latest tracy and build profiler.
+$(B)/tracy: $(B)
+	cd $(B); bash $(CURDIR)/get-tracy.sh
+.PHONY: $(B)/tracy
+
+$B:
+	mkdir -p $(B)
+.PHONY: $(B)
+
+# -- CLEAN ----------------------------------------------------------------------
+
+clean:
+	$(RM) $(B)/*.so $(B)/*.dll $(B)/*.pdb $(B)/*.lib $(B)/main*
+
+distclean:
+	rm -rf $(B)
+
    +
    +

    Find get-tracy.sh here.

    +

    Debug

    +

    gas

    +

    Frequently used directives

    +

    git(1)

    Working areas

    +-------------------+ --- stash -----> +-------+
@@ -6091,7 +6367,7 @@ link time -lgcov.
    
 generated for a single file for example such as
    
 
    gcov <SRC FILE | OBJ FILE>
 
    
-
    Example
    
+
    Example
    
 
    #include <cstdio>
 
 void tell_me(int desc) {
@@ -6997,7 +7273,7 @@ cpupower -c all frequency-info -g
 # Change frequency governor to POWERSAVE (eg).
 cpupower -c all frequency-set -g powersave
 
    
-
    Example
    
+
    Example
    
 
    Watch cpu frequency.
    
 
    watch -n1 "cpupower -c all frequency-info -f -m | xargs -n2 -d'\n'"
 
    
@@ -7112,7 +7388,7 @@ locally and -R means that requests are issued remotely.
    
 drop into shell)
 
  • -f run ssh command in the background
    • 
 
-
    Example
    
+
    Example
    
 
    # Forward requests on localhost:8080 to moose:1234 and keep ssh in forground
 # but dont drop into a shell.
 ssh -N -L 8080:1234 moose
@@ -8307,6 +8583,41 @@ wrmsr     // Write MSR register, effectively does MSR[ECX] <- EDX:EAX
 
    
 
    See guest64-msr.S as an example.
    
 
    
+
    Some interesting MSRs
    
+
      
+
    • C000_0082: IA32_LSTAR target address for syscall instruction
+in IA-32e (64 bit) mode.
      • 
+
    • C000_0100: IA32_FS_BASE storage for %fs segment base address.
      • 
+
    • C000_0101: IA32_GS_BASE storage for %gs segment base address.
      • 
+
    • C000_0102: IA32_KERNEL_GS_BASE additional register, swapgs
+swaps GS_BASE and KERNEL_GS_BASE, without altering any register state.
+Can be used to swap in a pointer to a kernel data structure on syscall entry,
+as for example in entry_SYSCALL_64.
      • 
+
    
+
    Current privilege level
    
+
    The current privilege level can be found at any time in the last two bits of the
+code segment selector cs. The following shows an example debugging an entry
+and exit of a syscall in x86_64-linux.
    
+
    Breakpoint 1, entry_SYSCALL_64 () at arch/x86/entry/entry_64.S:90
+90		swapgs
+(gdb) info r rax rcx cs
+rax            0x0                 0                ; syscall nr
+rcx            0x7feb16399e56      140647666916950  ; ret addr
+cs             0x10                16               ; cs & 0x3 -> 0 (ring0,kernel)
+
+(gdb) c
+Breakpoint 2, entry_SYSCALL_64 () at arch/x86/entry/entry_64.S:217
+217		sysretq
+(gdb) info r rcx cs
+rcx            0x7feb16399e56      140647666916950  ; ret addr
+cs             0x10                16               ; cs & 0x3 -> 0 (ring0,kernel)
+
+(gdb) b *$rcx
+(gdb) s
+Breakpoint 3, 0x00007feb16399e56 in ?? ()
+(gdb) info r cs
+cs             0x33                51  ; cs & 0x3 -> 3 (ring3,user)
+
    
 
    Size directives
    
 
    Explicitly specify size of the operation.
    
 
    mov  byte ptr [rax], 0xff    // save 1 byte(s) at [rax]
@@ -8373,6 +8684,18 @@ mov al, 0xaa
 mov cx, 0x10
 rep stosb
 
    
+
    AT&T syntax for intel syntax users
    
+
    mov %rax, %rbx           // mov rbx, rax
+mov $12, %rax            // mov rax, 12
+
+mov (%rsp), %rax         // mov rax, [rsp]
+mov 8(%rsp), %rax        // mov rax, [rsp + 8]
+mov (%rsp,%rcx,4), %rax  // mov rax, [rsp + 8 * rcx]
+mov 0x100, %rax          // mov rax, [0x100]
+mov (0x100), %rax        // mov rax, [0x100]
+
+mov %gs:8, %rax          // mov rax, gs:8
+
    
 
    Time stamp counter - rdtsc
    
 
    static inline uint64_t rdtsc() {
   uint32_t eax, edx;
@@ -8566,34 +8889,35 @@ must must save these registers in case they are used.
    
 
  • gnu assembler gas
    • 
 
  • intel syntax
    • 
 
-
    # file: greet.s
+
    // file: greet.S
+#include <asm/unistd.h>
 
     .intel_syntax noprefix
 
     .section .text, "ax", @progbits
     .global _start
 _start:
-    mov rdi, 1                      # fd
+    mov rdi, 1                      # fd (stdout)
     lea rsi, [rip + greeting]       # buf
     mov rdx, [rip + greeting_len]   # count
-    mov rax, 1                      # write(2) syscall nr
+    mov rax, __NR_write             # write(2) syscall nr
     syscall
 
-    mov rdi, 0                      # exit code
+    mov rdi, __NR_exit              # exit code
     mov rax, 60                     # exit(2) syscall nr
     syscall
 
     .section .rdonly, "a", @progbits
 greeting:
-    .asciz "Hi ASM-World!\n"
+    .ascii "Hi ASM-World!\n"
 greeting_len:
     .int .-greeting
 
    
 
    
-
    Syscall numbers are defined in /usr/include/asm/unistd.h.
    
+
    Files with .S suffix are pre-processed, while files with .s suffix are not.
    
 
    
 
    To compile and run:
    
-
    > gcc -o greet greet.s -nostartfiles -nostdlib && ./greet
+
    > gcc -o greet greet.S -nostartfiles -nostdlib && ./greet
 Hi ASM-World!
 
    
 
    MBR boot sectors example
    
-- 
cgit v1.2.3