llama : define magic numbers as integer constants (#1518) (#1520)

The underlying representation of multibyte character literals is
implementation-defined. This could, at least in principle, cause
cross-build data export/import issues independent of endianness.

Define magic numbers as integer literals to be on the safe side.

Signed-off-by: Juuso Alasuutari <redacted>

diff --git a/llama.cpp b/llama.cpp
index b38d55dda0e27915f412cfe1a55475f5c3b1000e..4cbc8d6b63752359f7de4b2c20600bfe9f4c703f 100644 (file)
--- a/llama.cpp
+++ b/llama.cpp
@@ -427,26 +427,30 @@ struct llama_file_loader {
     }
     void read_magic() {
         uint32_t magic = file.read_u32();
-        uint32_t version = 0;
 
-        if (magic != 'ggml') {
-            version = file.read_u32();
+        if (magic == LLAMA_FILE_MAGIC_GGML) {
+            file_version = LLAMA_FILE_VERSION_GGML;
+            return;
         }
 
-        if (magic == 'ggml' && version == 0) {
-            file_version = LLAMA_FILE_VERSION_GGML;
-        } else if (magic == 'ggmf' && version == 1) {
-            file_version = LLAMA_FILE_VERSION_GGMF_V1;
-        } else if (magic == 'ggjt' && version == 1) {
-            file_version = LLAMA_FILE_VERSION_GGJT_V1;
-        } else if (magic == 'ggjt' && version == 2) {
-            file_version = LLAMA_FILE_VERSION_GGJT_V2;
-        } else if (magic == 'ggjt' && version == 3) {
-            file_version = LLAMA_FILE_VERSION_GGJT_V3;
-        } else {
-            throw format("unknown (magic, version) combination: %08x, %08x; is this really a GGML file?",
-                         magic, version);
+        uint32_t version = file.read_u32();
+
+        switch (magic) {
+            case LLAMA_FILE_MAGIC_GGMF:
+                switch (version) {
+                    case 1: file_version = LLAMA_FILE_VERSION_GGMF_V1; return;
+                }
+                break;
+            case LLAMA_FILE_MAGIC_GGJT:
+                switch (version) {
+                    case 1: file_version = LLAMA_FILE_VERSION_GGJT_V1; return;
+                    case 2: file_version = LLAMA_FILE_VERSION_GGJT_V2; return;
+                    case 3: file_version = LLAMA_FILE_VERSION_GGJT_V3; return;
+                }
         }
+
+        throw format("unknown (magic, version) combination: %08x, %08x; is this really a GGML file?",
+                     magic, version);
     }
     void read_hparams() {
         hparams.n_vocab = file.read_u32();
@@ -2290,7 +2294,7 @@ int llama_apply_lora_from_file_internal(struct llama_context * ctx, const char *
     {
         uint32_t magic;
         fin.read((char *) &magic, sizeof(magic));
-        if (magic != 'ggla') {
+        if (magic != LLAMA_FILE_MAGIC_GGLA) {
             fprintf(stderr, "%s: bad file magic\n", __func__);
             return 1;
         }

diff --git a/llama.h b/llama.h
index 0a63d034b38edf6353a19d7167994da7d2e94f32..37bae5357400f466d9d7e011cfca1bbe55d39445 100644 (file)
--- a/llama.h
+++ b/llama.h
@@ -19,10 +19,16 @@
 #    define LLAMA_API
 #endif
 
+#define LLAMA_FILE_MAGIC_GGJT        0x67676a74u // 'ggjt'
+#define LLAMA_FILE_MAGIC_GGLA        0x67676c61u // 'ggla'
+#define LLAMA_FILE_MAGIC_GGMF        0x67676d66u // 'ggmf'
+#define LLAMA_FILE_MAGIC_GGML        0x67676d6cu // 'ggml'
+#define LLAMA_FILE_MAGIC_GGSN        0x6767736eu // 'ggsn'
+
 #define LLAMA_FILE_VERSION           3
-#define LLAMA_FILE_MAGIC             'ggjt'
-#define LLAMA_FILE_MAGIC_UNVERSIONED 'ggml'
-#define LLAMA_SESSION_MAGIC          'ggsn'
+#define LLAMA_FILE_MAGIC             LLAMA_FILE_MAGIC_GGJT
+#define LLAMA_FILE_MAGIC_UNVERSIONED LLAMA_FILE_MAGIC_GGML
+#define LLAMA_SESSION_MAGIC          LLAMA_FILE_MAGIC_GGSN
 #define LLAMA_SESSION_VERSION        1
 
 #ifdef __cplusplus