Adding testing suite

.github/workflows/ci.yml

@@ -0,0 +1,42 @@
+name: Build & Test
+
+on:
+  push:
+    branches: [master]
+  pull_request:
+    branches: [master]
+
+jobs:
+  linux:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - name: Build
+        run: make
+      - name: Test
+        run: make test
+
+  macos:
+    runs-on: macos-latest
+    steps:
+      - uses: actions/checkout@v4
+      - name: Build
+        run: make
+      - name: Test
+        run: make test
+
+  windows:
+    runs-on: windows-latest
+    defaults:
+      run:
+        shell: msys2 {0}
+    steps:
+      - uses: actions/checkout@v4
+      - uses: msys2/setup-msys2@v2
+        with:
+          msystem: MINGW64
+          install: mingw-w64-x86_64-gcc make
+      - name: Build
+        run: make
+      - name: Test
+        run: make test

Makefile

@@ -25,7 +25,25 @@ $(TARGET): $(OBJECTS)
 %.o: %.c
 	$(CC) $(CFLAGS) -c $< -o $@
 
+# ---- Test Suite ----
+TCFLAGS = -std=c99 -Wall -Wextra -O0 -g -Isrc -Isrc/fe -Isrc/ir -Isrc/amdgpu
+TSRC    = tests/tmain.c tests/tsmoke.c tests/tcomp.c tests/tenc.c \
+          tests/ttabs.c tests/ttypes.c tests/terrs.c tests/tphase.c
+TOBJS   = $(TSRC:.c=.o)
+COBJS   = src/ir/bir.o src/ir/bir_print.o src/ir/bir_lower.o src/ir/bir_mem2reg.o \
+          src/amdgpu/encode.o src/amdgpu/enc_tab.o src/amdgpu/isel.o src/amdgpu/emit.o \
+          src/fe/lexer.o src/fe/parser.o src/fe/preproc.o src/fe/sema.o
+
+test: $(TARGET) trunner
+	./trunner --all
+
+trunner: $(TOBJS) $(COBJS)
+	$(CC) $(TCFLAGS) -o $@ $^
+
+tests/%.o: tests/%.c
+	$(CC) $(TCFLAGS) -c $< -o $@
+
 clean:
-	rm -f $(OBJECTS) $(TARGET) $(TARGET).exe
+	rm -f $(OBJECTS) $(TARGET) $(TARGET).exe trunner trunner.exe $(TOBJS)
 
-.PHONY: all clean
+.PHONY: all clean test

README.md

@@ -6,6 +6,8 @@ This is what happens when you look at NVIDIA's walled garden and think "how hard
 
 **UPDATE: RDNA 2, 3 and 4 now supported.** `--gfx1030` / `--gfx1100` / `--gfx1200`
 
+**UPDATE: HSA runtime launcher added.** Compile kernels and dispatch them on real AMD hardware — no HIP, no hipcc. See [Runtime Launcher](#runtime-launcher).
+
 ## What It Does
 
 Takes CUDA C source code, the same `.cu` files you'd feed to `nvcc`, and compiles them to AMD RDNA 2 (gfx1030), RDNA 3 (gfx1100) and RDNA 4 (gfx1200) binaries. No LLVM. No HIP translation layer. No "convert your CUDA to something else first." Just a lexer, a parser, an IR, and a hand-written instruction selection backend that would make a compiler textbook weep.

tests/tcomp.c

@@ -0,0 +1,58 @@
+/* tcomp.c -- compile matrix
+ * Every .cu file, every target. The brute-force approach to confidence. */
+
+#include "tharns.h"
+
+static char obuf[TH_BUFSZ];
+
+/* ---- Helpers ---- */
+
+static const char *targets[] = { "--gfx1030", "", "--gfx1200" };
+static const char *tnames[]  = { "gfx1030", "gfx1100", "gfx1200" };
+
+static int compile_cu(const char *cu, const char *extra)
+{
+    char cmd[TH_BUFSZ];
+    const char *out = "test_out.hsaco";
+
+    for (int t = 0; t < 3; t++) {
+        snprintf(cmd, TH_BUFSZ,
+                 BC_BIN " --amdgpu-bin %s %s %s -o %s",
+                 targets[t], extra, cu, out);
+        int rc = th_run(cmd, obuf, TH_BUFSZ);
+        if (rc != 0) {
+            printf("  target %s failed: %s\n", tnames[t], obuf);
+            return -1;
+        }
+        if (!th_exist(out)) {
+            printf("  target %s: output missing\n", tnames[t]);
+            return -1;
+        }
+        remove(out);
+    }
+    return 0;
+}
+
+/* ---- compile: individual .cu files ---- */
+
+#define COMP_TEST(name, cu, extra) \
+    static void name(void) { \
+        CHECK(compile_cu(cu, extra) == 0); \
+        PASS(); \
+    } \
+    TH_REG("compile", name)
+
+COMP_TEST(cmp_vecadd,  "tests/vector_add.cu",         "")
+COMP_TEST(cmp_canon,   "tests/canonical.cu",           "")
+COMP_TEST(cmp_feat,    "tests/cuda_features.cu",       "")
+COMP_TEST(cmp_tier12,  "tests/test_tier12.cu",         "")
+COMP_TEST(cmp_notgpt,  "tests/notgpt.cu",              "")
+COMP_TEST(cmp_stress,  "tests/stress.cu",              "")
+COMP_TEST(cmp_math,    "tests/mymathhomework.cu",      "")
+COMP_TEST(cmp_launch,  "tests/test_launch_bounds.cu",  "")
+COMP_TEST(cmp_coop,    "tests/test_coop_groups.cu",    "")
+COMP_TEST(cmp_preproc, "tests/test_preproc.cu",        "")
+COMP_TEST(cmp_includ,  "tests/test_include.cu",        "-I tests")
+COMP_TEST(cmp_tmpl,    "tests/templates.cu",           "")
+COMP_TEST(cmp_unsign,  "tests/test_unsigned.cu",       "")
+COMP_TEST(cmp_shr2d,   "tests/test_shared2d.cu",      "")