feat(hw-001): Hardware Abstraction Layer — unified ternary backend, 21 tests

hardware_abstraction.zig (744 lines): compile-time backend selection
(CPU_SCALAR/CPU_SIMD/FPGA/GPU), SIMD capability detection
(AVX-512/AVX2/SSE4/NEON via builtin.cpu.arch + featureSetHas),
ScalarBackend (bind/unbind/bundle2/bundle3/cosine/hamming/permute/matvec),
SimdBackend @vector(8,i8) (bind/bundle2/cosine/matvec),
HardwareAbstraction unified dispatch, PerfCounters tracking,
MemoryAnalysis (16x vs f32, 8x vs f16), BackendConfig force/auto,
21 tests, build.zig wired (test-hardware-abstraction step).

Unlocks: CORE-004 (JIT Compilation), HW-003 (FPGA Acceleration)
Tech tree: 50/56 (89%), Hardware branch: 1/3 (33%)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

Author: gHashTag

.ralph/TECH_TREE.md

@@ -83,6 +83,9 @@
 |CORE-003|Bytecode VM|core|+500% execution speed vs interpreter|
 |**INF-001**|**GGUF Parser**|**inference**|**gguf_parser.zig (850 lines): GGUF v3 binary parser, ByteReader, 13 value types, tensor info, Q4_0/Q8_0 dequant, f16-to-f32, model config extraction, GGUFBuilder for round-trip tests, 20 tests, build.zig wired**|
 |**INF-002**|**Transformer Forward Pass**|**inference**|**transformer_forward.zig (960 lines): LLaMA-style transformer, RMSNorm, RoPE cache, SIMD matVec, GQA attention, SwiGLU FFN, KV cache, generation loop, top-p sampling, inference stats, 18 tests, build.zig wired**|
+|----|------|--------|------|
+|**HW-001**|**Hardware Abstraction Layer**|**hardware**|**hardware_abstraction.zig (~750 lines): compile-time backend selection (CPU_SCALAR/CPU_SIMD/FPGA/GPU), SIMD capability detection (AVX-512/AVX2/SSE4/NEON), ScalarBackend + SimdBackend @Vector(8,i8), unified dispatch, PerfCounters, MemoryAnalysis (16x compression), 21 tests, build.zig wired**|
+|----|------|--------|------|
 |DEP-001|Docker Container|deployment|Portable deployment|
 |DEP-002|Fly.io Integration|deployment|Global edge deployment|
 |OPT-T01|Ternary Weight Quantization|optimization|20x weight compression|
@@ -94,11 +97,11 @@
 ## 🔒 Locked (waiting for dependencies)
 | ID | Name | Branch | Needs (missing) |
 |----|------|--------|----------------|
-|CORE-004|JIT Compilation|core|HW-001 ❌|
+|CORE-004|JIT Compilation|core|HW-001 ✅ — **UNLOCKED**|
 |INF-005|Speculative Decoding v2|inference|INF-003 ❌, INF-004 ❌|
 |OPT-003|Weight Streaming|optimization|OPT-002 ❌|
 |DEP-004|Multi-Region Replication|deployment|DEP-003 ❌|
-|HW-003|FPGA Acceleration|hardware|HW-001 ❌|
+|HW-003|FPGA Acceleration|hardware|HW-001 ✅ — **UNLOCKED**|
 
 ## 📊 Branch Progress
 | Branch | Done | Total | % |
@@ -107,19 +110,19 @@
 |**Inference**|**4**|**5**|**80%**|
 |Deployment|2|4|50%|
 |**Optimization**|**16**|**16**|**100%**|
-|Hardware|0|3|0%|
+|**Hardware**|**1**|**3**|**33%**|
 |**Math**|**5**|**5**|**100%**|
 |**Development**|**3**|**3**|**100%**|
 |**Symbolic**|**5**|**5**|**100%**|
 |Visualization|1|1|100%|
 |**Nexus**|**10**|**10**|**100%**|
 |Multilingual|3|3|100%|
-|**Total**|**49**|**56**|**88%**|
+|**Total**|**50**|**56**|**89%**|
 
 ## 🎯 Recommended Next (highest ROI)
-1. **DEP-001** Docker Container — portable deployment, enables CI testing
-2. **CORE-004** JIT Compilation — needs HW-001 but provides 500% execution speed
-3. **DEP-003** Auto-Scaling — elastic infrastructure, prerequisite for DEP-004
+1. **CORE-004** JIT Compilation — HW-001 ✅ UNLOCKED, provides 500% execution speed
+2. **HW-003** FPGA Acceleration — HW-001 ✅ UNLOCKED, native ternary hardware
+3. **DEP-001** Docker Container — portable deployment, enables CI testing
 
 ---
 φ² + 1/φ² = 3 | TRINITY

build.zig

@@ -1895,4 +1895,17 @@ pub fn build(b: *std.Build) void {
     const gen_tfm_fwd_step = b.step("test-transformer-forward", "Test INF-002 Transformer Forward Pass — native LLM inference");
     gen_tfm_fwd_step.dependOn(&run_gen_tfm_fwd_tests.step);
     test_step.dependOn(&run_gen_tfm_fwd_tests.step);
+
+    // Generated Hardware Abstraction tests (HW-001: Unified ternary backend interface)
+    const gen_hw_abs_tests = b.addTest(.{
+        .root_module = b.createModule(.{
+            .root_source_file = b.path("generated/hardware_abstraction.zig"),
+            .target = target,
+            .optimize = optimize,
+        }),
+    });
+    const run_gen_hw_abs_tests = b.addRunArtifact(gen_hw_abs_tests);
+    const gen_hw_abs_step = b.step("test-hardware-abstraction", "Test HW-001 Hardware Abstraction Layer — unified ternary backend");
+    gen_hw_abs_step.dependOn(&run_gen_hw_abs_tests.step);
+    test_step.dependOn(&run_gen_hw_abs_tests.step);
 }