diff options
Diffstat (limited to 'print.html')
| -rw-r--r-- | print.html | 346 |
1 files changed, 335 insertions, 11 deletions
@@ -2829,6 +2829,7 @@ renice -n 5 -p PID <li><a href="trace_profile/./callgrind.html">callgrind</a></li> <li><a href="trace_profile/./valgrind.html">valgrind</a></li> <li><a href="trace_profile/./vtune.html">vtune</a></li> +<li><a href="trace_profile/./tracy.html">tracy</a></li> </ul> <div style="break-before: page; page-break-before: always;"></div><h1 id="usrbintime1"><a class="header" href="#usrbintime1">/usr/bin/time(1)</a></h1> <pre><code class="language-markdown"># statistics of process run @@ -3276,6 +3277,231 @@ main: main.c vtune: main $(VTUNE)/bin64/vtune -collect hotspots -start-paused -- ./main </code></pre> +<div style="break-before: page; page-break-before: always;"></div><h1 id="tracy1"><a class="header" href="#tracy1">tracy(1)</a></h1> +<p><a href="https://github.com/wolfpld/tracy">Tracy</a> is a frame profiler, supporting manual code instrumentation +and providing a sampling profiler.</p> +<p>One can either record and visualize the profiling data live using +<code>tracy-profiler</code> or record the profiling data to a file using <code>tracy-capture</code>.</p> +<pre><code>tracy-profiler [file] [-p port] + +tracy-capture -o file [-f] [-p port] + -f overwrite <file> if it exists +</code></pre> +<h2 id="example-7"><a class="header" href="#example-7">Example</a></h2> +<p>The example showcases different cases:</p> +<ol> +<li>Use tracy from a single binary. In that case the <code>TracyClient.cpp</code> can be +directly linked / included in the instrumented binary.</li> +<li>Use tracy from different binaries (eg main executable + shared library). In +this case the <code>TracyClient.cpp</code> should be compiled into its own shared +library, such that there is a single tracy client.</li> +<li>Use tracy from different binaries on windows. In this case the +<code>TracyClient.cpp</code> must be compiled again into a separate shared library, +while defining <code>TRACY_EXPORTS</code>. The code being instrumented must be compiled +with <code>TRACY_IMPORTS</code> defined.</li> +</ol> +<p>An instrumented <code>c++</code> example:</p> +<pre><code class="language-cpp">#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(); +} +</code></pre> +<p>An instrumented <code>c</code> example:</p> +<pre><code class="language-c">#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); +} +</code></pre> +<p>Raw build commands to demonstrate compiling tracy w/o <code>cmake</code>, in case we need +to integrate it into a different build system.</p> +<pre><code class="language-makefile">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) +</code></pre> +<blockquote> +<p>Find <code>get-tracy.sh</code> <a href="https://github.com/johannst/notes/blob/master/src/trace_profile/tracy/get-tracy.sh">here</a>.</p> +</blockquote> <div style="break-before: page; page-break-before: always;"></div><h1 id="debug-1"><a class="header" href="#debug-1">Debug</a></h1> <ul> <li><a href="debug/./gdb.html">gdb</a></li> @@ -3774,7 +4000,7 @@ $MOOSE=moose host:port tty </code></pre> -<h2 id="example-7"><a class="header" href="#example-7">Example</a></h2> +<h2 id="example-8"><a class="header" href="#example-8">Example</a></h2> <pre><code class="language-markdown"># Start gdbserver. gdbserver localhost:1234 /bin/ls @@ -3941,6 +4167,7 @@ objdump -C --disassemble=foo::bar <bin> <li><a href="development/./c++.html">c++</a></li> <li><a href="development/./glibc.html">glibc</a></li> <li><a href="development/./gcc.html">gcc</a></li> +<li><a href="development/./gas.html">gas</a></li> <li><a href="development/./git.html">git</a></li> <li><a href="development/./cmake.html">cmake</a></li> <li><a href="development/./make.html">make</a></li> @@ -5015,6 +5242,55 @@ run1: <li>C ABI (x86_64) - <a href="https://gitlab.com/x86-psABIs/x86-64-ABI">SystemV ABI</a></li> <li>C++ ABI - <a href="https://itanium-cxx-abi.github.io/cxx-abi">C++ Itanium ABI</a></li> </ul> +<div style="break-before: page; page-break-before: always;"></div><h1 id="gas"><a class="header" href="#gas">gas</a></h1> +<h2 id="frequently-used-directives"><a class="header" href="#frequently-used-directives">Frequently used directives</a></h2> +<ul> +<li> +<p><code>.byte</code>, <code>.2byte</code>, <code>.4byte</code>, <code>.8byte</code> to define a N byte value</p> +<pre><code class="language-x86asm">.byte 0xaa +.2byte 0xaabb +.2byte 0xaa, 0xbb +.4byte 0xaabbccdd +.8byte 0xaabbccdd11223344 +</code></pre> +</li> +<li> +<p><code>.ascii</code> to define an ascii string</p> +<pre><code class="language-x86asm">.ascii "foo" ; allocates 3 bytes +</code></pre> +</li> +<li> +<p><code>.asciz</code> to define an ascii string with <code>'\0'</code> terminator</p> +<pre><code class="language-x86asm">.asciz "foo" ; allocates 4 bytes (str + \0) +</code></pre> +</li> +<li> +<p><code>.macro</code> to define assembler macros. Arguments are accessed with the +<code>\arg</code> syntax.</p> +<pre><code class="language-x86asm">.macro defstr name str +\name: + .ascii "\str" +\name\()_len: + .8byte . - \name +.endm + +; use as +defstr foo, "foobar" +</code></pre> +<blockquote> +<p>Use <code>\()</code> to concatenate macro argument and literal.</p> +</blockquote> +</li> +<li> +<p><a href="https://sourceware.org/binutils/docs/as">GNU Assembler</a></p> +</li> +<li> +<p><a href="https://sourceware.org/binutils/docs/as/Pseudo-Ops.html#Pseudo-Ops">GNU Assembler Directives</a></p> +</li> +<li> +<p><a href="https://sourceware.org/binutils/docs/as/i386_002dDependent.html">GNU Assembler <code>x86_64</code> dependent features</a></p> +</li> +</ul> <div style="break-before: page; page-break-before: always;"></div><h1 id="git1"><a class="header" href="#git1">git(1)</a></h1> <h2 id="working-areas"><a class="header" href="#working-areas">Working areas</a></h2> <pre><code class="language-text">+-------------------+ --- stash -----> +-------+ @@ -6091,7 +6367,7 @@ link time <code>-lgcov</code>.</p> generated for a single file for example such as</p> <pre><code class="language-shell">gcov <SRC FILE | OBJ FILE> </code></pre> -<h2 id="example-8"><a class="header" href="#example-8">Example</a></h2> +<h2 id="example-9"><a class="header" href="#example-9">Example</a></h2> <pre><code class="language-cpp">#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 </code></pre> -<h2 id="example-9"><a class="header" href="#example-9">Example</a></h2> +<h2 id="example-10"><a class="header" href="#example-10">Example</a></h2> <p>Watch cpu frequency.</p> <pre><code class="language-sh">watch -n1 "cpupower -c all frequency-info -f -m | xargs -n2 -d'\n'" </code></pre> @@ -7112,7 +7388,7 @@ locally and <code>-R</code> means that requests are issued remotely.</p> drop into shell)</li> <li><code>-f</code> run <code>ssh</code> command in the background</li> </ul> -<h3 id="example-10"><a class="header" href="#example-10">Example</a></h3> +<h3 id="example-11"><a class="header" href="#example-11">Example</a></h3> <pre><code class="language-sh"># 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 <blockquote> <p>See <a href="https://github.com/johannst/mini-kvm-rs/blob/main/guest/guest64-msr.S">guest64-msr.S</a> as an example.</p> </blockquote> +<h4 id="some-interesting-msrs"><a class="header" href="#some-interesting-msrs">Some interesting MSRs</a></h4> +<ul> +<li><code>C000_0082: IA32_LSTAR</code> target address for <a href="https://www.felixcloutier.com/x86/syscall"><code>syscall</code></a> instruction +in <strong>IA-32e</strong> (64 bit) mode.</li> +<li><code>C000_0100: IA32_FS_BASE</code> storage for <strong>%fs</strong> segment base address.</li> +<li><code>C000_0101: IA32_GS_BASE</code> storage for <strong>%gs</strong> segment base address.</li> +<li><code>C000_0102: IA32_KERNEL_GS_BASE</code> additional register, <a href="https://www.felixcloutier.com/x86/swapgs"><code>swapgs</code></a> +swaps <strong>GS_BASE</strong> and <strong>KERNEL_GS_BASE</strong>, 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 <a href="https://web.git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/x86/entry/entry_64.S?h=v6.13#n87"><code>entry_SYSCALL_64</code></a>.</li> +</ul> +<h2 id="current-privilege-level"><a class="header" href="#current-privilege-level">Current privilege level</a></h2> +<p>The current privilege level can be found at any time in the last two bits of the +code segment selector <code>cs</code>. The following shows an example debugging an entry +and exit of a syscall in x86_64-linux.</p> +<pre><code>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) +</code></pre> <h2 id="size-directives"><a class="header" href="#size-directives">Size directives</a></h2> <p>Explicitly specify size of the operation.</p> <pre><code class="language-x86asm">mov byte ptr [rax], 0xff // save 1 byte(s) at [rax] @@ -8373,6 +8684,18 @@ mov al, 0xaa mov cx, 0x10 rep stosb </code></pre> +<h2 id="att-syntax-for-intel-syntax-users"><a class="header" href="#att-syntax-for-intel-syntax-users">AT&T syntax for intel syntax users</a></h2> +<pre><code class="language-x86asm">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 +</code></pre> <h2 id="time-stamp-counter---rdtsc"><a class="header" href="#time-stamp-counter---rdtsc">Time stamp counter - <code>rdtsc</code></a></h2> <pre><code class="language-c">static inline uint64_t rdtsc() { uint32_t eax, edx; @@ -8566,34 +8889,35 @@ must must save these registers in case they are used.</p> <li>gnu assembler <a href="https://sourceware.org/binutils/docs/as"><code>gas</code></a></li> <li>intel syntax</li> </ul> -<pre><code class="language-x86asm"># file: greet.s +<pre><code class="language-x86asm">// 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 </code></pre> <blockquote> -<p>Syscall numbers are defined in <code>/usr/include/asm/unistd.h</code>.</p> +<p>Files with <code>.S</code> suffix are pre-processed, while files with <code>.s</code> suffix are not.</p> </blockquote> <p>To compile and run:</p> -<pre><code class="language-bash">> gcc -o greet greet.s -nostartfiles -nostdlib && ./greet +<pre><code class="language-bash">> gcc -o greet greet.S -nostartfiles -nostdlib && ./greet Hi ASM-World! </code></pre> <h2 id="mbr-boot-sectors-example"><a class="header" href="#mbr-boot-sectors-example">MBR boot sectors example</a></h2> |