diff --git a/README.md b/README.md index 2ece294b7..beae920dc 100644 --- a/README.md +++ b/README.md @@ -238,12 +238,17 @@ In order to build llama.cpp you have three different options. cmake --build . --config Release ``` -- Using `Zig`: +- Using `Zig` (version 0.11 or later): + + Building for optimization levels and CPU features can be accomplished using standard build arguments, for example AVX2, FMA, F16C, + it's also possible to cross compile for other operating systems and architectures: ```bash - zig build -Doptimize=ReleaseFast + zig build -Doptimize=ReleaseFast -Dtarget=x86_64-windows-gnu -Dcpu=x86_64+avx2+fma+f16c ``` + The `zig targets` command will give you valid options to use. + - Using `gmake` (FreeBSD): 1. Install and activate [DRM in FreeBSD](https://wiki.freebsd.org/Graphics) diff --git a/build.zig b/build.zig index 04c88d8a2..d6dcb37aa 100644 --- a/build.zig +++ b/build.zig @@ -1,5 +1,6 @@ // Compatible with Zig Version 0.11.0 const std = @import("std"); +const ArrayList = std.ArrayList; const Compile = std.Build.Step.Compile; const ConfigHeader = std.Build.Step.ConfigHeader; const Mode = std.builtin.Mode; @@ -11,8 +12,20 @@ const Maker = struct { optimize: Mode, config_header: *ConfigHeader, - const cflags = .{"-std=c11"}; - const cxxflags = .{"-std=c++11"}; + cflags: ArrayList([]const u8), + cxxflags: ArrayList([]const u8), + objs: ArrayList(*Compile), + + fn addCFlag(m: *Maker, flag: []const u8) void { + m.cflags.append(flag) catch @panic("OOM"); + } + fn addCxxFlag(m: *Maker, flag: []const u8) void { + m.cxxflags.append(flag) catch @panic("OOM"); + } + fn addFlag(m: *Maker, flag: []const u8) void { + m.addCFlag(flag); + m.addCxxFlag(flag); + } fn init(builder: *std.build.Builder) Maker { const commit_hash = @embedFile(".git/refs/heads/master"); @@ -23,25 +36,34 @@ const Maker = struct { .BUILD_COMMIT = commit_hash[0 .. commit_hash.len - 1], // omit newline }, ); - return Maker{ + var m = Maker{ .builder = builder, .target = builder.standardTargetOptions(.{}), .optimize = builder.standardOptimizeOption(.{}), .config_header = config_header, + .cflags = ArrayList([]const u8).init(builder.allocator), + .cxxflags = ArrayList([]const u8).init(builder.allocator), + .objs = ArrayList(*Compile).init(builder.allocator), }; + m.addCFlag("-std=c11"); + m.addCxxFlag("-std=c++11"); + return m; } fn obj(m: *const Maker, name: []const u8, src: []const u8) *Compile { const o = m.builder.addObject(.{ .name = name, .target = m.target, .optimize = m.optimize }); if (std.mem.endsWith(u8, src, ".c")) { - o.addCSourceFiles(&.{src}, &cflags); + o.addCSourceFiles(&.{src}, m.cflags.items); o.linkLibC(); } else { - o.addCSourceFiles(&.{src}, &cxxflags); + o.addCSourceFiles(&.{src}, m.cxxflags.items); o.linkLibCpp(); } o.addIncludePath(.{ .path = "." }); o.addIncludePath(.{ .path = "./examples" }); + if (o.target.isWindows()) { + o.want_lto = false; // https://github.com/ziglang/zig/issues/15958 + } return o; } @@ -49,32 +71,37 @@ const Maker = struct { const e = m.builder.addExecutable(.{ .name = name, .target = m.target, .optimize = m.optimize }); e.addIncludePath(.{ .path = "." }); e.addIncludePath(.{ .path = "./examples" }); - e.addCSourceFiles(&.{src}, &cxxflags); + e.addCSourceFiles(&.{src}, m.cxxflags.items); for (deps) |d| e.addObject(d); + for (m.objs.items) |o| e.addObject(o); e.linkLibC(); e.linkLibCpp(); e.addConfigHeader(m.config_header); m.builder.installArtifact(e); - - // Currently a bug is preventing correct linking for optimized builds for Windows: - // https://github.com/ziglang/zig/issues/15958 if (e.target.isWindows()) { - e.want_lto = false; + e.want_lto = false; // https://github.com/ziglang/zig/issues/15958 } return e; } }; pub fn build(b: *std.build.Builder) void { - const make = Maker.init(b); + var make = Maker.init(b); + + if (b.option(bool, "k-quants", "Enable K-quants, (default: true)") orelse true) { + make.addFlag("-DGGML_USE_K_QUANTS"); + const k_quants = make.obj("k_quants", "k_quants.c"); + make.objs.append(k_quants) catch @panic("OOM"); + } const ggml = make.obj("ggml", "ggml.c"); const ggml_alloc = make.obj("ggml-alloc", "ggml-alloc.c"); const llama = make.obj("llama", "llama.cpp"); const common = make.obj("common", "examples/common.cpp"); + const console = make.obj("common", "examples/console.cpp"); const grammar_parser = make.obj("grammar-parser", "examples/grammar-parser.cpp"); - _ = make.exe("main", "examples/main/main.cpp", &.{ ggml, ggml_alloc, llama, common, grammar_parser }); + _ = make.exe("main", "examples/main/main.cpp", &.{ ggml, ggml_alloc, llama, common, console, grammar_parser }); _ = make.exe("quantize", "examples/quantize/quantize.cpp", &.{ ggml, ggml_alloc, llama }); _ = make.exe("perplexity", "examples/perplexity/perplexity.cpp", &.{ ggml, ggml_alloc, llama, common }); _ = make.exe("embedding", "examples/embedding/embedding.cpp", &.{ ggml, ggml_alloc, llama, common });