Merge pull request #33 from metaspace/october

Block subsystem update for October (v6.11, v6.12-rc2)

Author: ojeda

flake.nix

@@ -0,0 +1,25 @@
+
+{
+  description = "Shell for building rust-for-linux.com";
+  inputs.nixpkgs.url = "nixpkgs/nixos-24.05";
+  inputs.systems.url = "github:nix-systems/default";
+  inputs.flake-utils = {
+    url = "github:numtide/flake-utils";
+    inputs.systems.follows = "systems";
+  };
+
+  outputs =
+    { nixpkgs, flake-utils, ... }:
+    flake-utils.lib.eachDefaultSystem (
+      system:
+      let
+        pkgs = nixpkgs.legacyPackages.${system};
+      in
+      {
+        devShells.default = pkgs.mkShell { packages = with pkgs; [
+          python3
+          mdbook
+        ]; };
+      }
+    );
+}

src/NVMe-driver.md

@@ -24,6 +24,76 @@ The driver is not currently suitable for general use.
    [slides](https://lpc.events/event/16/contributions/1180/attachments/1017/1961/deck.pdf)
    and [video](https://lpc.events/event/16/contributions/1180/attachments/1017/2249/go)
 
+## 6.12-rc2 Rebase Performance ([`rnvme-v6.12-rc2`](https://git.kernel.org/pub/scm/linux/kernel/git/a.hindborg/linux.git/log/?h=rnvme-v6.12-rc2))
+
+### Setup
+
+ - AMD Ryzen 5 7600
+ - 32 GB 4800 MT/s DDR5 on one channel
+ - 1x Samsung 990 Pro 1TB (PCIe 4.0 x4 16 GT/S)
+ - NixOS 24.05
+
+### Results
+
+- 40 samples
+- Difference of means modeled with t-distribution
+- P95 confidence intervals
+
+![](rnvme/nvme-v6.12-rc2-absolute.svg)
+
+![](rnvme/nvme-v6.12-rc2-relative.svg)
+
+The graph shows
+    <math>
+        <mfrac>
+            <mrow>
+                <mi>R</mi>
+                <mo>-</mo>
+                <mi>C</mi>
+            </mrow>
+            <mrow>
+                <mi>C</mi>
+            </mrow>
+        </mfrac>
+    </math>
+where C is IO/s for the C driver and R is IO/s for the Rust driver. Thus, negative
+means the C driver is faster while positive means the Rust driver is faster.
+
+## 6.11 Rebase Performance ([`rnvme-v6.11`](https://git.kernel.org/pub/scm/linux/kernel/git/a.hindborg/linux.git/log/?h=rnvme-v6.11))
+
+### Setup
+
+ - AMD Ryzen 5 7600
+ - 32 GB 4800 MT/s DDR5 on one channel
+ - 1x Samsung 990 Pro 1TB (PCIe 4.0 x4 16 GT/S)
+ - NixOS 24.05
+
+### Results
+
+- 40 samples
+- Difference of means modeled with t-distribution
+- P95 confidence intervals
+
+![](rnvme/nvme-v6.11-absolute.svg)
+
+![](rnvme/nvme-v6.11-relative.svg)
+
+The graph shows
+    <math>
+        <mfrac>
+            <mrow>
+                <mi>R</mi>
+                <mo>-</mo>
+                <mi>C</mi>
+            </mrow>
+            <mrow>
+                <mi>C</mi>
+            </mrow>
+        </mfrac>
+    </math>
+where C is IO/s for the C driver and R is IO/s for the Rust driver. Thus, negative
+means the C driver is faster while positive means the Rust driver is faster.
+
 ## 6.11-rc2 Rebase Performance ([`rnvme-v6.11-rc2`](https://git.kernel.org/pub/scm/linux/kernel/git/a.hindborg/linux.git/log/?h=rnvme-v6.11-rc2))
 
 ### Setup

src/Null-Block-Driver.md

@@ -67,6 +67,54 @@ in this work:
  - [Mailing List Post](https://lore.kernel.org/all/20230503090708.2524310-1-nmi@metaspace.dk/)
  - [Subset merged in v6.11-rc1](https://lore.kernel.org/all/20240611114551.228679-1-nmi@metaspace.dk/)
 
+## 6.12-rc2 Rebase ([`rnull-v6.12-rc2`](https://git.kernel.org/pub/scm/linux/kernel/git/a.hindborg/linux.git/log/?h=rnull-v6.12-rc2))
+
+Changes from `rnull-v6.10`:
+
+ - Make `QueueData` references pinned.
+
+### Performance
+
+#### Setup
+
+ - AMD Ryzen 5 7600
+ - 32 GB 4800 MT/s DDR5 on one channel
+ - 1x Samsung 990 Pro 1TB (PCIe 4.0 x4 16 GT/S)
+ - NixOS 24.05
+
+#### Results
+
+- Plot shows `(mean_iops_r - mean_iops_c) / mean_iops_c`
+- 40 samples for each configuration
+- Difference of means modeled with t-distribution
+- P95 confidence intervals
+
+![](rnull/rnull-v6.12-rc2.svg)
+
+## 6.11 Rebase ([`rnull-v6.11`](https://git.kernel.org/pub/scm/linux/kernel/git/a.hindborg/linux.git/log/?h=rnull-v6.11))
+
+Changes from `rnull-v6.10`:
+
+ - None.
+
+### Performance
+
+#### Setup
+
+ - AMD Ryzen 5 7600
+ - 32 GB 4800 MT/s DDR5 on one channel
+ - 1x Samsung 990 Pro 1TB (PCIe 4.0 x4 16 GT/S)
+ - NixOS 24.05
+
+#### Results
+
+- Plot shows `(mean_iops_r - mean_iops_c) / mean_iops_c`
+- 40 samples for each configuration
+- Difference of means modeled with t-distribution
+- P95 confidence intervals
+
+![](rnull/rnull-v6.11.svg)
+
 ## 6.11-rc2 Rebase ([`rnull-v6.11-rc2`](https://git.kernel.org/pub/scm/linux/kernel/git/a.hindborg/linux.git/log/?h=rnull-v6.11-rc2))
 
 Changes from `rnull-v6.10`: