ggml: fix compilation error s390x (llama/12848)

* ggml: fixes #12846 compilation error

Signed-off-by: Aaron Teo <[email protected]>

Co-authored-by: Aleksei Nikiforov <[email protected]>

* ggml: add documentation for code change

Signed-off-by: Aaron Teo <[email protected]>

Co-authored-by: Aleksei Nikiforov <[email protected]>

* ggml: refactor to type-cast and update documentation

Signed-off-by: Aaron Teo <[email protected]>

Co-authored-by: Aleksei Nikiforov <[email protected]>

* ggml: update documentation to provide full issue link

Signed-off-by: Aaron Teo <[email protected]>

Co-authored-by: Aleksei Nikiforov <[email protected]>

---------

Co-authored-by: Aleksei Nikiforov <[email protected]>

ggml/src/ggml-cpu/simd-mappings.h

@@ -855,13 +855,17 @@ static inline __vector float __lzs_f16cx4_load(const ggml_fp16_t * x) {
         tmp[i] = GGML_FP16_TO_FP32(x[i]);
     }
 
-    return vec_xl(0, tmp);
+    // note: keep type-cast here to prevent compiler bugs
+    // see: https://github.com/ggml-org/llama.cpp/issues/12846
+    return vec_xl(0, (const float *)(tmp));
 }
 
 static inline void __lzs_f16cx4_store(ggml_fp16_t * x, __vector float y) {
     float arr[4];
 
-    vec_xst(y, 0, arr);
+    // note: keep type-cast here to prevent compiler bugs
+    // see: https://github.com/ggml-org/llama.cpp/issues/12846
+    vec_xst(y, 0, (float *)(arr));
 
     for (int i = 0; i < 4; i++) {
         x[i] = GGML_FP32_TO_FP16(arr[i]);