Auto merge of #109888 - Mark-Simulacrum:balanced-compression, r=pietroalbini

Remove optimal xz settings from CI

This is a companion PR to rust-lang/promote-release#58, which moves the relevant optimal code to rust-lang/promote-release. As mentioned in the comments of that PR, this is expected to cut CI costs (and time, though predominantly felt on fast builders) and reduce wasted resources due to in-practice single-threaded compression not using the full 8+ vCPU builders we have available.

This probably shouldn't land before that PR + a simpleinfra change to enable the recompression of xz artifacts. But if it does land, it's just a matter of a few nightlies with slightly larger artifacts, so not a big deal.

r? `@pietroalbini`

src/ci/run.sh

@@ -58,7 +58,12 @@ RUST_CONFIGURE_ARGS="$RUST_CONFIGURE_ARGS --disable-manage-submodules"
 RUST_CONFIGURE_ARGS="$RUST_CONFIGURE_ARGS --enable-locked-deps"
 RUST_CONFIGURE_ARGS="$RUST_CONFIGURE_ARGS --enable-cargo-native-static"
 RUST_CONFIGURE_ARGS="$RUST_CONFIGURE_ARGS --set rust.codegen-units-std=1"
-RUST_CONFIGURE_ARGS="$RUST_CONFIGURE_ARGS --set dist.compression-profile=best"
+# rust-lang/promote-release will recompress CI artifacts, and while we care
+# about the per-commit artifact sizes, it's not as important that they're
+# highly compressed as it is that the process is fast. Best compression
+# generally implies single-threaded compression which results in wasting most
+# of our CPU resources.
+RUST_CONFIGURE_ARGS="$RUST_CONFIGURE_ARGS --set dist.compression-profile=balanced"
 
 # When building for mingw, limit the number of parallel linker jobs during
 # the LLVM build, as not to run out of memory.

src/tools/rust-installer/src/compression.rs

@@ -89,43 +89,11 @@ impl CompressionFormat {
                         xz2::stream::MtStreamBuilder::new().threads(6).preset(6).encoder().unwrap()
                     }
                     CompressionProfile::Best => {
-                        let mut filters = xz2::stream::Filters::new();
+                        // Note that this isn't actually the best compression settings for the
-                        // the preset is overridden by the other options so it doesn't matter
+                        // produced artifacts, the production artifacts on static.rust-lang.org are
-                        let mut lzma_ops = xz2::stream::LzmaOptions::new_preset(9).unwrap();
+                        // produced by rust-lang/promote-release which hosts recompression logic
-                        // This sets the overall dictionary size, which is also how much memory (baseline)
+                        // and is tuned for optimal compression.
-                        // is needed for decompression.
+                        xz2::stream::MtStreamBuilder::new().threads(6).preset(9).encoder().unwrap()
-                        lzma_ops.dict_size(64 * 1024 * 1024);
-                        // Use the best match finder for compression ratio.
-                        lzma_ops.match_finder(xz2::stream::MatchFinder::BinaryTree4);
-                        lzma_ops.mode(xz2::stream::Mode::Normal);
-                        // Set nice len to the maximum for best compression ratio
-                        lzma_ops.nice_len(273);
-                        // Set depth to a reasonable value, 0 means auto, 1000 is somwhat high but gives
-                        // good results.
-                        lzma_ops.depth(1000);
-                        // 2 is the default and does well for most files
-                        lzma_ops.position_bits(2);
-                        // 0 is the default and does well for most files
-                        lzma_ops.literal_position_bits(0);
-                        // 3 is the default and does well for most files
-                        lzma_ops.literal_context_bits(3);
-
-                        filters.lzma2(&lzma_ops);
-
-                        let mut builder = xz2::stream::MtStreamBuilder::new();
-                        builder.filters(filters);
-
-                        // On 32-bit platforms limit ourselves to 3 threads, otherwise we exceed memory
-                        // usage this process can take. In the future we'll likely only do super-fast
-                        // compression in CI and move this heavyweight processing to promote-release (which
-                        // is always 64-bit and can run on big-memory machines) but for now this lets us
-                        // move forward.
-                        if std::mem::size_of::<usize>() == 4 {
-                            builder.threads(3);
-                        } else {
-                            builder.threads(6);
-                        }
-                        builder.encoder().unwrap()
                     }
                 };
 
@@ -245,20 +213,14 @@ impl Write for CombinedEncoder {
     }
 
     fn flush(&mut self) -> std::io::Result<()> {
-        self.encoders
+        self.encoders.par_iter_mut().map(|w| w.flush()).collect::<std::io::Result<Vec<()>>>()?;
-            .par_iter_mut()
-            .map(|w| w.flush())
-            .collect::<std::io::Result<Vec<()>>>()?;
         Ok(())
     }
 }
 
 impl Encoder for CombinedEncoder {
     fn finish(self: Box<Self>) -> Result<(), Error> {
-        self.encoders
+        self.encoders.into_par_iter().map(|e| e.finish()).collect::<Result<Vec<()>, Error>>()?;
-            .into_par_iter()
-            .map(|e| e.finish())
-            .collect::<Result<Vec<()>, Error>>()?;
         Ok(())
     }
 }