Working patches with latest branch and z/OS specific formatting for patches

VyoJ · VyoJ · commit d4b22d37ee9c · 2025-07-31T03:29:17.000-05:00
diff --git a/buildenv b/buildenv
@@ -7,14 +7,14 @@ export ZOPEN_DEV_TAG="master"
 export ZOPEN_NAME="llamacpp-master"
 export ZOPEN_RUNTIME_DEPS="ncurses"
 
-# rm -f "llama"
-# ln -s "llama.cpp" "llama"
-# ln -s "llama.cpp" $ZOPEN_NAME
+rm -f "llama"
+ln -s "llama.cpp" "llama"
+ln -s "llama.cpp" $ZOPEN_NAME
 
 export ZOPEN_COMP="CLANG"
 # set env variables
 # export CURL_HOME="/data/zopen/usr/local/zopen/curl/curl"
-# export BLAS_HOME="/usr/lpp/cbclib"
+export BLAS_HOME="/usr/lpp/cbclib"
 
 export ZOPEN_CONFIGURE="cmake"
 export ZOPEN_CONFIGURE_OPTS="-B ../build --install-prefix \"\$ZOPEN_INSTALL_DIR/\" -DCURL_LIBRARY=\$CURL_HOME/lib/libcurl.a -DCURL_INCLUDE_DIR=\$CURL_HOME/include -DBUILD_SHARED_LIBS_DEFAULT=OFF -DBUILD_SHARED_LIBS=OFF -DGGML_STATIC=ON -DGGML_BACKEND_DL=OFF -DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS -DBLAS_INCLUDE_DIRS=\$BLAS_HOME/include/openblas -DBLAS_LIBRARIES=\$BLAS_HOME/lib/libopenblas.so -DLLAMA_BUILD_TESTS=ON ."
@@ -76,4 +76,4 @@ zopen_get_version()
   # Modify to echo the version of your tool/library
   # Rather than hardcoding the version, obtain the version by running the tool/library
   echo "1.0.0"
-}
+}
diff --git a/examples/frontend/src/components/ChatMode.js b/examples/frontend/src/components/ChatMode.js
@@ -139,7 +139,7 @@ const ChatMode = ({ conversationStarted, setConversationStarted }) => {
             onClick={handleChatSubmit}
             disabled={isTyping}
           >
-            â¤
+            Ã¢ÂÂ¤
           </button>
         </div>
       </div>
diff --git a/patches/arg.cpp.patch b/patches/arg.cpp.patch
@@ -10,7 +10,7 @@ index 40af7e57..46fec792 100644
 +#   include <cstdlib>
 +#   endif
  #endif
-
+ 
  using json = nlohmann::ordered_json;
 @@ -195,6 +198,8 @@ bool common_has_curl() {
  #   endif
diff --git a/patches/ggml-backend-reg.cpp.patch b/patches/ggml-backend-reg.cpp.patch
@@ -4,12 +4,12 @@ index 405d8e3..b3682a9 100644
 +++ b/ggml/src/ggml-backend-reg.cpp
 @@ -556,7 +556,9 @@ static ggml_backend_reg_t ggml_backend_load_best(const char * name, bool silent,
  }
-
+ 
  void ggml_backend_load_all() {
 -    ggml_backend_load_all_from_path(nullptr);
 +#ifdef GGML_BACKEND_DL
 +     ggml_backend_load_all_from_path(nullptr);
 +#endif
  }
-
+ 
  void ggml_backend_load_all_from_path(const char * dir_path) {
diff --git a/patches/ggml-cpu.c.patch b/patches/ggml-cpu.c.patch
@@ -3,10 +3,11 @@ index c7426df..e074799 100644
 --- a/ggml/src/ggml-cpu/ggml-cpu.c
 +++ b/ggml/src/ggml-cpu/ggml-cpu.c
 @@ -17,7 +17,7 @@
-
+ 
  #if defined(_MSC_VER) || defined(__MINGW32__)
  #include <malloc.h> // using malloc.h with MSC/MINGW
 -#elif !defined(__FreeBSD__) && !defined(__NetBSD__) && !defined(__OpenBSD__)
 +#elif !defined(__FreeBSD__) && !defined(__NetBSD__) && !defined(__OpenBSD__) && !defined(__MVS__)
  #include <alloca.h>
  #endif
+ 
diff --git a/patches/ggml-cpu.cpp.patch b/patches/ggml-cpu.cpp.patch
@@ -5,10 +5,10 @@ index e013e8b..6290b22 100644
 @@ -33,6 +33,38 @@
  #    include <sys/types.h>
  #endif
-
-+#define CVT_PTR        0x10
-+#define CVTRCEP_OFFSET         0x490
-+#define RCEAFC_OFFSET  0x088
+ 
++#define CVT_PTR 	0x10
++#define CVTRCEP_OFFSET 	0x490
++#define RCEAFC_OFFSET 	0x088
 +
 +typedef unsigned data_area_ptr_assign_type;
 +
@@ -17,7 +17,7 @@ index e013e8b..6290b22 100644
 +#if defined(_LP64)
 +        data_area_ptr_assign_type lower;
 +#endif
-+       data_area_ptr_assign_type assign;
++	data_area_ptr_assign_type assign;
 +    };
 +    char* deref;
 +} data_area_ptr;
@@ -39,15 +39,15 @@ index e013e8b..6290b22 100644
 +}
 +
  // ggml-backend interface
-
+ 
  std::vector<ggml_backend_buffer_type_t>& ggml_backend_cpu_get_extra_buffers_type() {
 @@ -43,7 +75,7 @@ std::vector<ggml_backend_buffer_type_t>& ggml_backend_cpu_get_extra_buffers_type
          if (ggml_backend_amx_buffer_type()) {
              bufts.push_back(ggml_backend_amx_buffer_type());
          }
 -#endif
-+#endif
-
++#endif	
+ 
  #ifdef GGML_USE_CPU_KLEIDIAI
          if (ggml_backend_cpu_kleidiai_buffer_type()) {
 @@ -340,6 +372,9 @@ static void ggml_backend_cpu_device_get_memory(ggml_backend_dev_t dev, size_t *
@@ -59,4 +59,4 @@ index e013e8b..6290b22 100644
 +    *free = get_free_memory();
  #else
      long pages = sysconf(_SC_PHYS_PAGES);
-     long page_size = sysconf(_SC_PAGE_SIZE);
+     long page_size = sysconf(_SC_PAGE_SIZE);
diff --git a/patches/ggml.c.patch b/patches/ggml.c.patch
@@ -5,7 +5,7 @@ index 97da26b3..212faf9c 100644
 @@ -14,9 +14,13 @@
  #include <hbwmalloc.h>
  #endif
-
+ 
 +#if defined(__MVS__)
 +#include <zos-base.h>
 +#endif
@@ -16,19 +16,19 @@ index 97da26b3..212faf9c 100644
 +#elif !defined(__FreeBSD__) && !defined(__NetBSD__) && !defined(__OpenBSD__) && !defined(__MVS__)
  #include <alloca.h>
  #endif
-
+ 
 @@ -299,6 +303,15 @@ void * ggml_aligned_malloc(size_t size) {
-
+ 
  #if defined(_MSC_VER) || defined(__MINGW32__)
      return _aligned_malloc(size, alignment);
 +#elif defined(__MVS__)
 +//    void * aligned_memory = NULL;
 +//    if (size ==0)
-+//         size = 1;
++//	    size = 1;
 +//    aligned_memory = malloc(size);
 +//    int result = 0;
 +//    if (aligned_memory == NULL)
-+//         result = errno;
++//	    result = errno;
 +    return __aligned_malloc(size, alignment);
  #else
      if (size == 0) {
@@ -41,4 +41,4 @@ index 97da26b3..212faf9c 100644
 +    __aligned_free(ptr);
  #elif GGML_USE_CPU_HBM
      if (ptr != NULL) {
-         hbw_free(ptr);
+         hbw_free(ptr);
diff --git a/patches/llama-context.cpp.patch b/patches/llama-context.cpp.patch
@@ -7,31 +7,31 @@ index 06e93b19..0db8530f 100644
                  }
              }
 -
-+
++	
 +            #ifndef __MVS__
              std::fill(output_ids.begin(), output_ids.end(), -1);
 -
-+           #else
-+           for (auto& id: output_ids) {
-+               id = -1;
-+           }
++	    #else
++	    for (auto& id: output_ids) {
++		id = -1;
++	    }
              for (uint32_t i = 0; i < n_outputs; ++i) {
                  output_ids[out_ids[i]] = i;
              }
-+           #endif
++	    #endif
          }
      }
-
+ 
 @@ -1264,8 +1269,10 @@ uint32_t llama_context::output_reserve(int32_t n_outputs) {
      embd   = has_embd   ? output_base + logits_size : nullptr;
-
+ 
      // set all ids as invalid (negative)
 -    std::fill(output_ids.begin(), output_ids.end(), -1);
 -
 +    //std::fill(output_ids.begin(), output_ids.end(), -1);
 +    for (auto& id: output_ids) {
-+       id = -1;
++	id = -1;
 +    }
      this->n_outputs = 0;
-
-     return n_outputs_max;
+ 
+     return n_outputs_max;
diff --git a/patches/llama-hparams.h.patch b/patches/llama-hparams.h.patch
@@ -0,0 +1,12 @@
+diff --git a/src/llama-hparams.h b/src/llama-hparams.h
+index 8b7e2a11..530dfea2 100644
+--- a/src/llama-hparams.h
++++ b/src/llama-hparams.h
+@@ -3,6 +3,7 @@
+ #include "llama.h"
+ 
+ #include <array>
++#include <algorithm>
+ 
+ // bump if necessary
+ #define LLAMA_MAX_LAYERS  512
diff --git a/patches/miniaudio.h.patch b/patches/miniaudio.h.patch
@@ -4,7 +4,7 @@ index c74bebe..72cbafd 100644
 +++ b/vendor/miniaudio/miniaudio.h
 @@ -3873,8 +3873,8 @@ typedef ma_uint16 wchar_t;
      #define MA_POSIX
-
+ 
      /*
 -    Use the MA_NO_PTHREAD_IN_HEADER option at your own risk. This is intentionally undocumented.
 -    You can use this to avoid including pthread.h in the header section. The downside is that it
@@ -16,9 +16,9 @@ index c74bebe..72cbafd 100644
 @@ -16176,7 +16176,7 @@ static ma_result ma_thread_create__posix(ma_thread* pThread, ma_thread_priority
      int result;
      pthread_attr_t* pAttr = NULL;
-
+ 
 -#if !defined(__EMSCRIPTEN__) && !defined(__3DS__)
 +#if !defined(__EMSCRIPTEN__) && !defined(__3DS__) && !defined(__MVS__)
      /* Try setting the thread priority. It's not critical if anything fails here. */
      pthread_attr_t attr;
-     if (pthread_attr_init(&attr) == 0) {
+     if (pthread_attr_init(&attr) == 0) {
diff --git a/patches/run.cpp.patch b/patches/run.cpp.patch
@@ -3,14 +3,14 @@ index 6fe728c6..9ee09d80 100644
 --- a/tools/run/run.cpp
 +++ b/tools/run/run.cpp
 @@ -22,6 +22,9 @@
-
+ 
  #if defined(LLAMA_USE_CURL)
  #    include <curl/curl.h>
 +#    if defined(__MVS__)
 +#    include <cstdlib>
 +#    endif
  #endif
-
+ 
  #include <signal.h>
 @@ -507,6 +510,7 @@ class HttpClient {
          curl_easy_setopt(curl, CURLOPT_FOLLOWLOCATION, 1L);
@@ -19,4 +19,4 @@ index 6fe728c6..9ee09d80 100644
 +	curl_easy_setopt(curl, CURLOPT_CAINFO, std::getenv("SSL_CERT_PATH"));
          return curl_easy_perform(curl);
      }
-
+ 
diff --git a/patches/simd.inprogresspatch b/patches/simd.inprogresspatch
diff --git a/patches/stb_image.h.patch b/patches/stb_image.h.patch

Original file line number	Diff line number	Diff line change
`@@ -139,7 +139,7 @@ const ChatMode = ({ conversationStarted, setConversationStarted }) => {`
`139`	`139`	`onClick={handleChatSubmit}`
`140`	`140`	`disabled={isTyping}`
`141`	`141`	`>`
`142`		`- â¤`
	`142`	`+ Ã¢ÂÂ¤`
`143`	`143`	`</button>`
`144`	`144`	`</div>`
`145`	`145`	`</div>`
Original file line number	Diff line number	Diff line change
`@@ -7,31 +7,31 @@ index 06e93b19..0db8530f 100644`
`7`	`7`	`}`
`8`	`8`	`}`
`9`	`9`	`-`
`10`		`-+`
	`10`	`++`
`11`	`11`	`+ #ifndef __MVS__`
`12`	`12`	`std::fill(output_ids.begin(), output_ids.end(), -1);`
`13`	`13`	`-`
`14`		`-+ #else`
`15`		`-+ for (auto& id: output_ids) {`
`16`		`-+ id = -1;`
`17`		`-+ }`
	`14`	`++ #else`
	`15`	`++ for (auto& id: output_ids) {`
	`16`	`++ id = -1;`
	`17`	`++ }`
`18`	`18`	`for (uint32_t i = 0; i < n_outputs; ++i) {`
`19`	`19`	`output_ids[out_ids[i]] = i;`
`20`	`20`	`}`
`21`		`-+ #endif`
	`21`	`++ #endif`
`22`	`22`	`}`
`23`	`23`	`}`
`24`		`-`
	`24`	`+`
`25`	`25`	`@@ -1264,8 +1269,10 @@ uint32_t llama_context::output_reserve(int32_t n_outputs) {`
`26`	`26`	`embd = has_embd ? output_base + logits_size : nullptr;`
`27`		`-`
	`27`	`+`
`28`	`28`	`// set all ids as invalid (negative)`
`29`	`29`	`- std::fill(output_ids.begin(), output_ids.end(), -1);`
`30`	`30`	`-`
`31`	`31`	`+ //std::fill(output_ids.begin(), output_ids.end(), -1);`
`32`	`32`	`+ for (auto& id: output_ids) {`
`33`		`-+ id = -1;`
	`33`	`++ id = -1;`
`34`	`34`	`+ }`
`35`	`35`	`this->n_outputs = 0;`
`36`		`-`
`37`		`- return n_outputs_max;`
	`36`	`+`
	`37`	`+ return n_outputs_max;`