LightNews LightNews
banner
serpent7776.bsky.social
@serpent7776.bsky.social
130 followers 10 following 480 posts
The Syntax Sorcerer aka I like programming languages.
Posts Replies Media Videos
serpent7776.bsky.social
#TIL PAUSE x86 instruction
#assembly
Improves the performance of spin-wait loops. When executing a “spin-wait loop,” processors will suffer a severe performance penalty when exiting the loop because it detects a possible memory order violation. The PAUSE instruction provides a hint to the processor that the code sequence is a spin-wait loop. The processor uses this hint to avoid the memory order violation in most situations, which greatly improves processor performance. For this reason, it is recommended that a PAUSE instruction be placed in all spin-wait loops. https://www.felixcloutier.com/x86/pause
November 19, 2025 at 5:04 PM
serpent7776.bsky.social
#cplusplus value categories.
2x2 matrix of c++ value categories: on vertical axis: can move / can't move on horizontal axis: has identity / no identity lvalue: has identity, can't move prvalue: no identity, can move xvalue: has identity, can move rvalue: can move (either xvalue or prvalue) glvalue: has identity (either lvalue or xvalue) https://blog.knatten.org/2018/03/09/lvalues-rvalues-glvalues-prvalues-xvalues-help/
November 18, 2025 at 4:24 PM
serpent7776.bsky.social
#programming
Corporate needs you to find the difference between this picture and this picture meme: - "loop" on the left picture - "recursion" on the right picture - They're the same picture
November 12, 2025 at 6:01 PM
serpent7776.bsky.social
#cplusplus
World if meme subtitle: if C++ was inspired by modula rather than simula.
October 31, 2025 at 7:28 PM
serpent7776.bsky.social
Vector iota function in #zig
fn iota(comptime len: comptime_int, init: u32) @Vector(len, u32) { var result: @Vector(len, u32) = undefined; inline for (0..len) |i| { result[i] = init + @as(u32, @intCast(i)); } return result; }
October 25, 2025 at 8:40 PM
serpent7776.bsky.social
plus scan function (aka prefix sum) in #zig.
fn plus_scan(comptime N: usize, comptime T: type, v: @Vector(N, T)) @Vector(N, T) { var result: @Vector(N, T) = undefined; result[0] = v[0]; inline for (1..N) |n| { result[n] = result[n-1] + v[n]; } return result; }
October 25, 2025 at 5:18 PM
serpent7776.bsky.social
#zig pub v export keywords.
In Zig, pub and export serve different purposes related to visibility and linkage of functions, variables, and other declarations. pub: This keyword makes a function, variable, or type publicly accessible outside the current file or module. It controls Zig's module-level visibility, allowing other Zig code that imports the module to access the declaration. However, pub does not make the symbol available at the binary or linker level outside the Zig compilation unit. It keeps the symbol within Zig’s modular namespace. export: This keyword makes the function or variable available externally at the linker level with C calling conventions. It ensures the symbol is globally visible outside the compiled object or library, typically for use by C or other languages, or to be the entry point in a binary. export also implies that the function uses the C calling convention and restricts use of certain Zig-only features (e.g., comptime parameters). It is used to create interfaces visible from outside the Zig code.
October 25, 2025 at 4:34 PM
serpent7776.bsky.social
#zig wtf is x deduced as usize in the loop here? This is nonsense.
const std = @import("std"); fn f(x: u32) void { std.debug.print("{d}\n", .{x}); } pub fn main() void { const b: u32 = 1; const e: u32 = 10; for (b..e) |x| { f(x); } }
October 25, 2025 at 1:33 PM
serpent7776.bsky.social
Perplexity, you OK?
Your answer was upgraded to Pro Search Pro considers 10x more sources. Subscribe to increase your daily limit. -1 enhanced queries remaining today.
October 22, 2025 at 8:26 PM
serpent7776.bsky.social
Who tf designed snap remove?
I accidentally messed up Firefox snap and had to remove and install and it removed my profile data.
remove Remove snaps from the system The remove command removes the named snap instance from the system. By default all the snap revisions are removed, including their data and the common data directory. When a --revision option is passed only the specified revision is removed. https://manpages.ubuntu.com/manpages/bionic/en/man8/snap.8.html
October 21, 2025 at 8:32 PM
serpent7776.bsky.social
Yoohoo! You've earned the Hacktoberfest 2025: Level 1 badge! Hacktoberfest 2025: Level 1 New avatar level unlocked, looking good! Share your progress and keep the PRs coming
October 20, 2025 at 7:11 PM
serpent7776.bsky.social
Why not #assembly
Bilbo - Why shouldn’t I keep it meme: After all why not why shouldn't I solve travelling salesman problem in SIMD
October 18, 2025 at 9:25 PM
serpent7776.bsky.social
Seriously #Go?
When a byte-slice-to-string conversion is used in a map element retrieval expression (the map lookup, e.g., value := m[string(bytes)]), Go does not allocate new memory for the string conversion. This is an optimization that avoids unnecessary allocations because the string is used only transiently to look up the map entry. However, when the byte-slice-to-string conversion is used in an L-value expression where the map element is assigned or updated (an L-value means the left side of an assignment), like m[string(bytes)] = newValue, Go does allocate memory for the string conversion. This is because the map needs a stable, persistent string key to store internally in the map's data structure, so it must own a copy.
October 17, 2025 at 8:49 PM
serpent7776.bsky.social
Uh oh, #clickhouse having some OOM issues.
Elapsed: 1.113 sec. Processed 120.62 million rows, 1.69 GB (108.35 million rows/s., 1.52 GB/s.) Peak memory usage: 2.15 MiB. Received exception from server (version 25.9.3): Code: 241. DB::Exception: Received from localhost:9000. DB::Exception: (total) memory limit exceeded: would use 3.36 GiB (attempt to allocate chunk of 0.00 B), current RSS: 1.01 GiB, maximum: 3.36 GiB. OvercommitTracker decision: Query was selected to stop by OvercommitTracker: While executing AggregatingTransform. (MEMORY_LIMIT_EXCEEDED)
October 11, 2025 at 9:05 PM
serpent7776.bsky.social
wtf arm?
Table showing alignment requirements of primitive types on various platforms. All agree except for int128, which is 16 on all listed platforms, except for arm, which is 8.
October 8, 2025 at 9:42 PM
serpent7776.bsky.social
Do you tune #postgresql to lower the value of random_page_cost?
seq_page_cost and random_page_cost The default values of seq_page_cost and random_page_cost are 1.0 and 4.0, respectively. This means that PostgreSQL assumes that the random scan is four times slower than the sequential scan. In other words, the default value of PostgreSQL is based on using HDDs. On the other hand, in recent days, the default value of random_page_cost is too large because SSDs are mostly used. If the default value of random_page_cost is used despite using an SSD, the planner may select ineffective plans. Therefore, when using an SSD, it is better to change the value of random_page_cost to 1.0. This blog reported the problem when using the default value of random_page_cost. https://www.interdb.jp/pg/pgsql03/02.html
September 30, 2025 at 5:25 PM
serpent7776.bsky.social
#zig error messages are pretty crap. There's nowhere a source location where error occurred.
Why do I need to pass an extra flag to get basic diagnostics?
/usr/lib/zig/std/Io/Writer.zig:1771:5: error: invalid format string 'd' for type 'bool' @compileError("invalid format string '" ++ fmt ++ "' for type '" ++ @typeName(@TypeOf(value)) ++ "'"); ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ referenced by: printValue__anon_17174: /usr/lib/zig/std/Io/Writer.zig:1049:40 print__anon_15630: /usr/lib/zig/std/Io/Writer.zig:700:25 11 reference(s) hidden; use '-freference-trace=13' to see all references
September 29, 2025 at 8:50 PM
serpent7776.bsky.social
Measuring per-commit vxdiff performance.
Plot showing time taken by vxdiff to compare two big images for each commit.
September 27, 2025 at 7:33 PM
serpent7776.bsky.social
I like those non-goals.
Non-goals 1. Convincing people to use VictoriaMetrics components when there are better suited solutions exist for their tasks, since users will become angry at VictoriaMetrics after they discover better solutions. 2. Breaking links to VictoriaMetrics docs , since users will be unhappy seeing 404 page or unexpected results after they click some old link somewhere on the Internet or in the internal knowledge base. 3. Breaking backwards compatibility in new releases, since users will be unhappy when their working setup breaks after the upgrade.
September 19, 2025 at 8:53 PM
serpent7776.bsky.social
#SQL is weird.
SQL is a weird concept. You write your application in one language, say JavaScript, and then send commands in a completely different language, called SQL, to the database. The database then compiles and optimizes that SQL command, runs it, and returns your data. It seems terribly inefficient and, yet, your application might do this hundreds of times per second. It’s madness! https://fly.io/blog/sqlite-virtual-machine/
September 13, 2025 at 9:14 PM
serpent7776.bsky.social
Everything got faster, but the relative ratios flipped.
Everything got faster, but the relative ratios also completely flipped. For a 2020 server, a single channel of DDR4–3200 had nearly twice the throughput of a x16 PCIe slot, and PCIe Gen3x16 had more throughput than a 100 Gbps Ethernet link. By 2023, DDR advanced 50% with the transition to DDR5, but PCIe throughput quadrupled with a two-generation jump from Gen3 to Gen5, and Ethernet throughput increased eightfold, to 800 Gbps on a single link. A single 800G Ethernet link has nearly twice the throughput of the widest possible Gen5 PCIe interface, which is difficult for traditional architectures where NICs connect computing resources to the network. https://blog.enfabrica.net/the-next-step-in-high-performance-distributed-computing-systems-4f98f13064ac
September 12, 2025 at 8:04 PM
serpent7776.bsky.social
Don't tie yourself to a platform.
Don't tie yourself to a platform.
September 11, 2025 at 8:58 PM
serpent7776.bsky.social
All software should be possible to write by one person.
All software should be possible to write by one person.
September 10, 2025 at 8:55 PM
serpent7776.bsky.social
Cross function in #K
I'm sure this can be shorter.
{{[x;y;w] (a;b):w; (x@a),(y@b)}[x;y]@'+!((#x),#y)}["abc";"xyzw"] ("ax" "ay" "az" "aw" "bx" "by" "bz" "bw" "cx" "cy" "cz" "cw")
September 8, 2025 at 7:40 PM
serpent7776.bsky.social
Naming things is hard #postgresql
/* * SendProcSignal * Send a signal to a Postgres process * Not to be confused with ProcSendSignal */
September 3, 2025 at 8:56 PM