sycl : add flash-attn support for head size 512 (#21654)

* sycl : add flash-attn support for head size 512

This patch extends the SYCL Flash Attention implementation to support head sizes (DKQ/DV) of 512.

Changes:
- Added DKQ/DV 512 cases to both tile and vector Flash Attention kernels.
- Updated kernel selection logic to allow vector kernels for head sizes up to 512 (previously 256).
- Removed unused/redundant AMD and RDNA-specific configuration functions in `fattn-tile.hpp`.
- Refactored `ggml_backend_sycl_buffer_init_tensor` to use a switch statement for clearer tensor extra buffer initialization.
- Added necessary template instances for the new 512 head size across various quantization types.

* remove defunct mxfp4 reorder from setting buffer type

diff --git a/ggml/src/ggml-sycl/fattn-tile.cpp b/ggml/src/ggml-sycl/fattn-tile.cpp
index 9d4f019cf517366d556fcccecc15be0ee64da37b..9449d75784d0e7731492989db147fdbe9e83bb7b 100644 (file)
--- a/ggml/src/ggml-sycl/fattn-tile.cpp
+++ b/ggml/src/ggml-sycl/fattn-tile.cpp
@@ -44,6 +44,10 @@ void ggml_sycl_flash_attn_ext_tile(ggml_backend_sycl_context & ctx, ggml_tensor
             GGML_ASSERT(V->ne[0] == K->ne[0]);
             ggml_sycl_flash_attn_ext_tile_case<256, 256>(ctx, dst);
         } break;
+        case 512: {
+            GGML_ASSERT(V->ne[0] == K->ne[0]);
+            ggml_sycl_flash_attn_ext_tile_case<512, 512>(ctx, dst);
+        } break;
         case 576: {
             GGML_ASSERT(V->ne[0] == 512);
             ggml_sycl_flash_attn_ext_tile_case<576, 512>(ctx, dst);

diff --git a/ggml/src/ggml-sycl/fattn-tile.hpp b/ggml/src/ggml-sycl/fattn-tile.hpp
index b4d4e0ae90e3ff97d98ed3b8158f0021c158bbb7..9ba5296968d5af828c8a7e3119604832d6fc685c 100644 (file)
--- a/ggml/src/ggml-sycl/fattn-tile.hpp
+++ b/ggml/src/ggml-sycl/fattn-tile.hpp
@@ -67,6 +67,12 @@ static constexpr uint32_t ggml_sycl_fattn_tile_get_config_fp16(const int DKQ, co
     GGML_SYCL_FATTN_TILE_CONFIG_CASE(256, 256, 16, 256, 2,  64,  64)
     GGML_SYCL_FATTN_TILE_CONFIG_CASE(256, 256, 32, 256, 2,  64,  64)
 
+    GGML_SYCL_FATTN_TILE_CONFIG_CASE(512, 512,  2,  64, 2,  64,  64)
+    GGML_SYCL_FATTN_TILE_CONFIG_CASE(512, 512,  4, 128, 2,  64,  64)
+    GGML_SYCL_FATTN_TILE_CONFIG_CASE(512, 512,  8, 256, 2,  64,  64)
+    GGML_SYCL_FATTN_TILE_CONFIG_CASE(512, 512, 16, 256, 2,  64,  64)
+    GGML_SYCL_FATTN_TILE_CONFIG_CASE(512, 512, 32, 256, 2,  64,  64)
+
     GGML_SYCL_FATTN_TILE_CONFIG_CASE(576, 512,  4, 128, 2,  64,  64)
     GGML_SYCL_FATTN_TILE_CONFIG_CASE(576, 512,  8, 256, 2,  64,  64)
     GGML_SYCL_FATTN_TILE_CONFIG_CASE(576, 512, 16, 256, 2,  64,  64)
@@ -124,6 +130,12 @@ static constexpr uint32_t ggml_sycl_fattn_tile_get_config_fp32(const int DKQ, co
     GGML_SYCL_FATTN_TILE_CONFIG_CASE(256, 256, 16, 256, 2,  32, 128)
     GGML_SYCL_FATTN_TILE_CONFIG_CASE(256, 256, 32, 256, 2,  32,  64)
 
+    GGML_SYCL_FATTN_TILE_CONFIG_CASE(512, 512,  2, 128, 2,  64,  64)
+    GGML_SYCL_FATTN_TILE_CONFIG_CASE(512, 512,  4, 128, 2,  64,  64)
+    GGML_SYCL_FATTN_TILE_CONFIG_CASE(512, 512,  8, 256, 2,  64,  64)
+    GGML_SYCL_FATTN_TILE_CONFIG_CASE(512, 512, 16, 256, 2,  64,  64)
+    GGML_SYCL_FATTN_TILE_CONFIG_CASE(512, 512, 32, 256, 2,  64,  64)
+
     GGML_SYCL_FATTN_TILE_CONFIG_CASE(576, 512,  4, 128, 2,  32,  64)
     GGML_SYCL_FATTN_TILE_CONFIG_CASE(576, 512,  8, 256, 2,  32,  64)
     GGML_SYCL_FATTN_TILE_CONFIG_CASE(576, 512, 16, 256, 2,  32,  64)
@@ -131,134 +143,6 @@ static constexpr uint32_t ggml_sycl_fattn_tile_get_config_fp32(const int DKQ, co
     return 0;
 }
 
-static constexpr uint32_t ggml_sycl_fattn_tile_get_config_amd(const int DKQ, const int DV, const int ncols) {
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 40,  40,  2,  64, 2,  32,  40)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 40,  40,  4, 128, 2,  32,  40)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 40,  40,  8, 256, 2,  32,  40)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 40,  40, 16, 256, 2,  32,  40)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 40,  40, 32, 256, 2,  32,  40)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 40,  40, 64, 256, 2,  32,  40)
-
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 64,  64,  2,  64, 3,  32,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 64,  64,  4, 128, 3,  64,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 64,  64,  8, 128, 2,  32,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 64,  64, 16, 256, 2, 128,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 64,  64, 32, 256, 2,  64,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 64,  64, 64, 256, 2,  64,  64)
-
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 72,  72,  2,  64, 2,  32,  72)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 72,  72,  4, 128, 2,  32,  72)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 72,  72,  8, 256, 2,  32,  72)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 72,  72, 16, 256, 2,  32,  72)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 72,  72, 32, 256, 2,  32,  72)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 72,  72, 64, 256, 2,  32,  72)
-
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 80,  80,  2,  64, 2,  32,  40)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 80,  80,  4, 128, 2,  32,  40)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 80,  80,  8, 256, 2,  32,  40)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 80,  80, 16, 256, 2,  32,  40)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 80,  80, 32, 256, 2,  32,  40)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 80,  80, 64, 256, 2,  32,  40)
-
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 96,  96,  2,  64, 2,  32,  48)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 96,  96,  4, 128, 2,  32,  48)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 96,  96,  8, 256, 2,  32,  48)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 96,  96, 16, 256, 2,  32,  48)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 96,  96, 32, 256, 2,  32,  48)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 96,  96, 64, 256, 2,  32,  48)
-
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(112, 112,  2,  64, 2,  32,  56)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(112, 112,  4, 128, 2,  32,  56)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(112, 112,  8, 256, 2,  32,  56)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(112, 112, 16, 256, 2,  32,  56)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(112, 112, 32, 256, 2,  32,  56)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(112, 112, 64, 256, 2,  32,  56)
-
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(128, 128,  2, 256, 2, 128,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(128, 128,  4, 128, 2,  64, 128)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(128, 128,  8, 256, 2,  64, 128)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(128, 128, 16, 256, 2,  64, 128)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(128, 128, 32, 256, 2,  64,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(128, 128, 64, 256, 2,  64,  32)
-
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(256, 256,  2, 256, 2, 128,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(256, 256,  4, 256, 2,  64, 128)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(256, 256,  8, 256, 2,  64, 128)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(256, 256, 16, 256, 2,  32, 128)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(256, 256, 32, 256, 2,  32, 128)
-
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(576, 512,  4, 128, 2,  64,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(576, 512,  8, 256, 2,  64,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(576, 512, 16, 256, 2,  64,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(576, 512, 32, 512, 1, 128,  64)
-
-    return 0;
-}
-
-static constexpr uint32_t ggml_sycl_fattn_tile_get_config_amd_rdna(const int DKQ, const int DV, const int ncols) {
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 40,  40,  2,  64, 2,  32,  40)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 40,  40,  4, 128, 2,  32,  40)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 40,  40,  8, 256, 2,  32,  40)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 40,  40, 16, 256, 2,  32,  40)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 40,  40, 32, 256, 2,  32,  40)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 40,  40, 64, 256, 2,  32,  40)
-
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 64,  64,  2,  64, 8,  32,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 64,  64,  4,  64, 8,  32,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 64,  64,  8, 128, 5, 128,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 64,  64, 16, 128, 5, 128,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 64,  64, 32, 128, 4,  64,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 64,  64, 64, 128, 5,  64,  64)
-
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 72,  72,  2,  64, 2,  32,  72)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 72,  72,  4, 128, 2,  32,  72)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 72,  72,  8, 256, 2,  32,  72)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 72,  72, 16, 256, 2,  32,  72)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 72,  72, 32, 256, 2,  32,  72)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 72,  72, 64, 256, 2,  32,  72)
-
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 80,  80,  2,  64, 2,  32,  40)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 80,  80,  4, 128, 2,  32,  40)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 80,  80,  8, 256, 2,  32,  40)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 80,  80, 16, 256, 2,  32,  40)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 80,  80, 32, 256, 2,  32,  40)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 80,  80, 64, 256, 2,  32,  40)
-
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 96,  96,  2,  64, 2,  32,  48)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 96,  96,  4, 128, 2,  32,  48)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 96,  96,  8, 256, 2,  32,  48)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 96,  96, 16, 256, 2,  32,  48)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 96,  96, 32, 256, 2,  32,  48)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE( 96,  96, 64, 256, 2,  32,  48)
-
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(112, 112,  2,  64, 2,  32,  56)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(112, 112,  4, 128, 2,  32,  56)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(112, 112,  8, 256, 2,  32,  56)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(112, 112, 16, 256, 2,  32,  56)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(112, 112, 32, 256, 2,  32,  56)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(112, 112, 64, 256, 2,  32,  56)
-
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(128, 128,  2,  64, 8,  32,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(128, 128,  4, 128, 8,  64,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(128, 128,  8, 128, 8,  64,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(128, 128, 16, 256, 3, 128, 128)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(128, 128, 32, 256, 3, 128,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(128, 128, 64, 256, 3,  64,  64)
-
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(256, 256,  2,  64, 8,  32,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(256, 256,  4, 128, 6,  32, 256)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(256, 256,  8, 128, 6,  32, 256)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(256, 256, 16, 256, 5,  32, 256)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(256, 256, 32, 256, 3,  64, 128)
-
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(576, 512,  4, 128, 2,  64,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(576, 512,  8, 256, 2,  64,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(576, 512, 16, 256, 4,  64,  64)
-    GGML_SYCL_FATTN_TILE_CONFIG_CASE(576, 512, 32, 256, 2, 128,  64)
-
-    return 0;
-}
-
 static constexpr uint32_t ggml_sycl_fattn_tile_get_config(const int DKQ, const int DV, const int ncols, const int cc) {
     if(fast_fp16_available(cc))
         return ggml_sycl_fattn_tile_get_config_fp16(DKQ, DV, ncols);
@@ -1293,6 +1177,16 @@ static void launch_fattn_tile_switch_ncols2(ggml_backend_sycl_context & ctx, ggm
             launch_fattn_tile_switch_ncols1<DKQ, DV, 4, use_logit_softcap>(ctx, dst);
             return;
         }
+        // ncols2=2 and ncols2=1 fallbacks only for cases where ncols=2 config exists (DKQ == DV).
+        // For DKQ == 576, DV == 512 only GQA-optimized variants are implemented.
+        if constexpr (DKQ == DV) {
+            if (use_gqa_opt && gqa_ratio % 2 == 0) {
+                launch_fattn_tile_switch_ncols1<DKQ, DV, 2, use_logit_softcap>(ctx, dst);
+                return;
+            }
+            launch_fattn_tile_switch_ncols1<DKQ, DV, 1, use_logit_softcap>(ctx, dst);
+            return;
+        }
     }
 
     if constexpr (DV <= 256) {
@@ -1347,5 +1241,6 @@ extern DECL_FATTN_TILE_CASE( 96,  96);
 extern DECL_FATTN_TILE_CASE(112, 112);
 extern DECL_FATTN_TILE_CASE(128, 128);
 extern DECL_FATTN_TILE_CASE(256, 256);
+extern DECL_FATTN_TILE_CASE(512, 512);
 extern DECL_FATTN_TILE_CASE(576, 512);
 

diff --git a/ggml/src/ggml-sycl/fattn-vec.hpp b/ggml/src/ggml-sycl/fattn-vec.hpp
index 48c389052f435b5187e9e2fe5c4d8babfc6f363c..8031acfdff8831475fdbf80e0398dc553e7b9c4f 100644 (file)
--- a/ggml/src/ggml-sycl/fattn-vec.hpp
+++ b/ggml/src/ggml-sycl/fattn-vec.hpp
@@ -664,4 +664,11 @@ EXTERN_DECL_FATTN_VEC_CASES(256, GGML_TYPE_Q5_0)
 EXTERN_DECL_FATTN_VEC_CASES(256, GGML_TYPE_Q5_1)
 EXTERN_DECL_FATTN_VEC_CASES(256, GGML_TYPE_Q8_0)
 
+EXTERN_DECL_FATTN_VEC_CASES(512, GGML_TYPE_F16)
+EXTERN_DECL_FATTN_VEC_CASES(512, GGML_TYPE_Q4_0)
+EXTERN_DECL_FATTN_VEC_CASES(512, GGML_TYPE_Q4_1)
+EXTERN_DECL_FATTN_VEC_CASES(512, GGML_TYPE_Q5_0)
+EXTERN_DECL_FATTN_VEC_CASES(512, GGML_TYPE_Q5_1)
+EXTERN_DECL_FATTN_VEC_CASES(512, GGML_TYPE_Q8_0)
+
 #endif // GGML_SYCL_FATTN_VEC_HPP

diff --git a/ggml/src/ggml-sycl/fattn.cpp b/ggml/src/ggml-sycl/fattn.cpp
index c276ed898270ac2e9e977732227fb9428fee4f01..7c6e6112fdcdd488aa2ff41279444f77f7dcaec2 100644 (file)
--- a/ggml/src/ggml-sycl/fattn.cpp
+++ b/ggml/src/ggml-sycl/fattn.cpp
@@ -34,6 +34,7 @@
     FATTN_VEC_CASE( 64, type_K, type_V)       \
     FATTN_VEC_CASE(128, type_K, type_V)       \
     FATTN_VEC_CASE(256, type_K, type_V)       \
+    FATTN_VEC_CASE(512, type_K, type_V)       \
 
 static void ggml_sycl_flash_attn_ext_vec(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     ggml_tensor * Q = dst->src[0];
@@ -141,6 +142,7 @@ static best_fattn_kernel ggml_sycl_get_best_fattn_kernel(const int device, const
         case 128:
         case 112:
         case 256:
+        case 512:
             if (V->ne[0] != K->ne[0]) {
                 return BEST_FATTN_KERNEL_NONE;
             }
@@ -185,7 +187,7 @@ static best_fattn_kernel ggml_sycl_get_best_fattn_kernel(const int device, const
     }
 
     // For small batch sizes the vector kernel may be preferable over the kernels optimized for large batch sizes:
-    const bool can_use_vector_kernel = Q->ne[0] <= 256 && Q->ne[0] % 64 == 0 && K->ne[1] % FATTN_KQ_STRIDE == 0;
+    const bool can_use_vector_kernel = Q->ne[0] <= 512 && Q->ne[0] % 64 == 0 && K->ne[1] % FATTN_KQ_STRIDE == 0;
 
     // Todo: Use the XMX kernel if possible:
 

diff --git a/ggml/src/ggml-sycl/ggml-sycl.cpp b/ggml/src/ggml-sycl/ggml-sycl.cpp
index e80ead9aea4715082e67c01dca6b6991ff98b373..7f9b2df524ed903d85d740cc80c065c382e76591 100644 (file)
--- a/ggml/src/ggml-sycl/ggml-sycl.cpp
+++ b/ggml/src/ggml-sycl/ggml-sycl.cpp
@@ -411,11 +411,22 @@ ggml_backend_sycl_buffer_init_tensor(ggml_backend_buffer_t buffer,
         assert(tensor->view_src->buffer->buft == buffer->buft);
         return GGML_STATUS_SUCCESS;
     }
-    if ((tensor->type == GGML_TYPE_Q4_0 || tensor->type == GGML_TYPE_Q8_0 || tensor->type == GGML_TYPE_Q4_K || tensor->type == GGML_TYPE_Q6_K) &&
-        !g_ggml_sycl_disable_optimize) {
-        ggml_tensor_extra_gpu * extra = new ggml_tensor_extra_gpu{};
-        tensor->extra                 = extra;
-        ctx->tensor_extras.push_back(extra);  //used to release it when destroy ctx.
+
+    if (!g_ggml_sycl_disable_optimize) {
+        // set reorder extra buffer based on supported type
+        switch (tensor->type) {
+            case GGML_TYPE_Q4_0:
+            case GGML_TYPE_Q8_0:
+            case GGML_TYPE_Q4_K:
+            case GGML_TYPE_Q6_K:{
+                ggml_tensor_extra_gpu * extra = new ggml_tensor_extra_gpu{};
+                tensor->extra                 = extra;
+                ctx->tensor_extras.push_back(extra);
+                break;
+            }
+            default:
+                break;
+        }
     }
 
     if (ggml_is_quantized(tensor->type)) {

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq512-dv512.cpp b/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq512-dv512.cpp
new file mode 100644 (file)
index 0000000..9a6a187
--- /dev/null
+++ b/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq512-dv512.cpp
@@ -0,0 +1,6 @@
+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../fattn-tile.hpp"
+
+DECL_FATTN_TILE_CASE(512, 512);
+

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-f16.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-f16.cpp
index 32cf4f2859ba7a5aa530fb762e460d9fb129286b..43ef94c118cda1b0ab693649de4d0f9a10d61922 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-f16.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-f16.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_F16, GGML_TYPE_F16);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_F16, GGML_TYPE_F16);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_F16, GGML_TYPE_F16);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_F16, GGML_TYPE_F16);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q4_0.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q4_0.cpp
index a61a19021bbd7e89eec94e722cc9fae29cbd9c79..9404061d456d7040fafea2f2fe335766d2928890 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q4_0.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q4_0.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_F16, GGML_TYPE_Q4_0);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_F16, GGML_TYPE_Q4_0);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_F16, GGML_TYPE_Q4_0);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_F16, GGML_TYPE_Q4_0);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q4_1.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q4_1.cpp
index 63b74fb347a4cd6f7c5c210bcd277c13f6275f58..a8bb9f52d0c880b5dd8d9c531d7807e83df86faf 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q4_1.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q4_1.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_F16, GGML_TYPE_Q4_1);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_F16, GGML_TYPE_Q4_1);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_F16, GGML_TYPE_Q4_1);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_F16, GGML_TYPE_Q4_1);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q5_0.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q5_0.cpp
index 46e2d9853c53c89689a44149a044b66b7b57e9d9..7d61f6ab0afedbeb5d3c8b2250a631a21f747476 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q5_0.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q5_0.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_F16, GGML_TYPE_Q5_0);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_F16, GGML_TYPE_Q5_0);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_F16, GGML_TYPE_Q5_0);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_F16, GGML_TYPE_Q5_0);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q5_1.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q5_1.cpp
index 7aabb6ff6e4f35a225a2634a442118c4d76a4f44..753bae09f838255dc2bd110cd18cc6934caa2968 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q5_1.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q5_1.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_F16, GGML_TYPE_Q5_1);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_F16, GGML_TYPE_Q5_1);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_F16, GGML_TYPE_Q5_1);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_F16, GGML_TYPE_Q5_1);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q8_0.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q8_0.cpp
index 148ea217f6231673b3c81859f1498f132210ce85..546a93b257070039a736e061aff37a6c3e3759a4 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q8_0.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q8_0.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_F16, GGML_TYPE_Q8_0);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_F16, GGML_TYPE_Q8_0);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_F16, GGML_TYPE_Q8_0);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_F16, GGML_TYPE_Q8_0);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-f16.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-f16.cpp
index 4b169dbcdbc021bb3e35299c9c311b7fbe0024fd..53c8c2f2654bc1e08d91ee8801995960b348e452 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-f16.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-f16.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q4_0, GGML_TYPE_F16);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_F16);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_0, GGML_TYPE_F16);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q4_0, GGML_TYPE_F16);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q4_0.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q4_0.cpp
index 79f530b1815945f047f34b4f0f9fc8249d771fcc..5b409c55f217d480ddd92f3aac6a0e772727dbce 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q4_0.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q4_0.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q4_0, GGML_TYPE_Q4_0);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q4_0);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_0, GGML_TYPE_Q4_0);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q4_0, GGML_TYPE_Q4_0);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q4_1.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q4_1.cpp
index 2f7db51ce826e802e77524c14c094eac33e08fa0..8c4ef588d634c9e5327c5ce3142dd307cff7d837 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q4_1.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q4_1.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q4_0, GGML_TYPE_Q4_1);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q4_1);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_0, GGML_TYPE_Q4_1);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q4_0, GGML_TYPE_Q4_1);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q5_0.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q5_0.cpp
index 9e3bf0b14a1063c781adb85938ea342d5a424c7e..83f0a07552e54638986a5a2588be2df0744142d5 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q5_0.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q5_0.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q4_0, GGML_TYPE_Q5_0);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q5_0);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_0, GGML_TYPE_Q5_0);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q4_0, GGML_TYPE_Q5_0);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q5_1.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q5_1.cpp
index 18081879cecca3b0de4e9fb257d4744625c3d2ca..9df9b03bba43784cd242409289d6bd16345804fa 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q5_1.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q5_1.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q4_0, GGML_TYPE_Q5_1);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q5_1);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_0, GGML_TYPE_Q5_1);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q4_0, GGML_TYPE_Q5_1);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q8_0.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q8_0.cpp
index 1c387b0d87c4747bf307738aa70e41451e0cc663..6980c2a65bb0d16fc6c1931844d92caace58917a 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q8_0.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q8_0.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q4_0, GGML_TYPE_Q8_0);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_0, GGML_TYPE_Q8_0);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_0, GGML_TYPE_Q8_0);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q4_0, GGML_TYPE_Q8_0);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-f16.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-f16.cpp
index f005b3762ccc1ea59360c3580490ad163f7d53b4..bd61bc1dc2b2c00a1dff5969a9bcc183c23a767f 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-f16.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-f16.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q4_1, GGML_TYPE_F16);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_F16);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_1, GGML_TYPE_F16);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q4_1, GGML_TYPE_F16);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q4_0.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q4_0.cpp
index 3553b1cdd163e55bf5afa1f261862fc5eb5ef36d..492e229a58e804e08c320768266e293d27dd36c0 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q4_0.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q4_0.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q4_1, GGML_TYPE_Q4_0);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_Q4_0);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_1, GGML_TYPE_Q4_0);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q4_1, GGML_TYPE_Q4_0);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q4_1.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q4_1.cpp
index 687ec5671152aa453662973bf990733215b9c5e0..30f88a2ebd5ff0a136a082c5a60a3607e28c25f3 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q4_1.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q4_1.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q4_1, GGML_TYPE_Q4_1);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_Q4_1);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_1, GGML_TYPE_Q4_1);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q4_1, GGML_TYPE_Q4_1);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q5_0.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q5_0.cpp
index 2663bfe7466c1d0878f3c98b465f15fa765e1678..db76663604ed7538897208523db134cba6cea189 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q5_0.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q5_0.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q4_1, GGML_TYPE_Q5_0);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_Q5_0);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_1, GGML_TYPE_Q5_0);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q4_1, GGML_TYPE_Q5_0);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q5_1.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q5_1.cpp
index 641b7c7ae2a94686549ec9acd4e2ecc3037fde41..1dbcc8a85a8fb2b2813d7770582a15b7ef0b47a2 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q5_1.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q5_1.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q4_1, GGML_TYPE_Q5_1);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_Q5_1);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_1, GGML_TYPE_Q5_1);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q4_1, GGML_TYPE_Q5_1);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q8_0.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q8_0.cpp
index 3d3181d47191137961c49c06d25807085f45dcdb..d30996a62598d27982731dab78420fd263714167 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q8_0.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q8_0.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q4_1, GGML_TYPE_Q8_0);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q4_1, GGML_TYPE_Q8_0);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q4_1, GGML_TYPE_Q8_0);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q4_1, GGML_TYPE_Q8_0);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-f16.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-f16.cpp
index 85d5026ad4f9ac9f0de5b74794c297ab8899060a..bc0f635d922d4bf87d2a286bf9cf2795ac82df0b 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-f16.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-f16.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_0, GGML_TYPE_F16);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_F16);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_0, GGML_TYPE_F16);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q5_0, GGML_TYPE_F16);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q4_0.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q4_0.cpp
index 1e81401a2c90330f72e2febd695ee6284b4edf5f..9e0378107cb13c6ca56224e71655b2b978e12e84 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q4_0.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q4_0.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_0, GGML_TYPE_Q4_0);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_Q4_0);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_0, GGML_TYPE_Q4_0);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q5_0, GGML_TYPE_Q4_0);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q4_1.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q4_1.cpp
index 54251473f97adac5b1f5f0adb9c6f75708f1668c..a8535ac9156d6f127afc1c6d224ed2824ed7220e 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q4_1.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q4_1.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_0, GGML_TYPE_Q4_1);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_Q4_1);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_0, GGML_TYPE_Q4_1);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q5_0, GGML_TYPE_Q4_1);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q5_0.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q5_0.cpp
index d418c1fb21eda237177060ebd887b5781a82613f..43d4fae9a619e926cb53963ee37ebd254c6daac7 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q5_0.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q5_0.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_0, GGML_TYPE_Q5_0);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_Q5_0);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_0, GGML_TYPE_Q5_0);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q5_0, GGML_TYPE_Q5_0);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q5_1.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q5_1.cpp
index 0f26cfabd090a6156f2054eda9839b10707ecc46..23335a41640b2da565ed23aa4f26bd1f570cb49b 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q5_1.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q5_1.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_0, GGML_TYPE_Q5_1);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_Q5_1);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_0, GGML_TYPE_Q5_1);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q5_0, GGML_TYPE_Q5_1);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q8_0.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q8_0.cpp
index 4fb98723519ada6c0358ab9de07b21d8816a72a2..52550a33757bfe00ccab91166d1de474f201b473 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q8_0.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q8_0.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_0, GGML_TYPE_Q8_0);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_0, GGML_TYPE_Q8_0);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_0, GGML_TYPE_Q8_0);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q5_0, GGML_TYPE_Q8_0);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-f16.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-f16.cpp
index 85b79cd19761d8eb61199dc0479b0516a0e37c88..4651f14c05050026d57bdc5e2e10c3135679c8f3 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-f16.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-f16.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_1, GGML_TYPE_F16);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_F16);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_1, GGML_TYPE_F16);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q5_1, GGML_TYPE_F16);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q4_0.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q4_0.cpp
index 7348323b28bb62101bb20844870da0dc0be01af1..2310fd8792c353d0810bad67a7835a9455f3585a 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q4_0.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q4_0.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_1, GGML_TYPE_Q4_0);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_Q4_0);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_1, GGML_TYPE_Q4_0);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q5_1, GGML_TYPE_Q4_0);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q4_1.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q4_1.cpp
index f19af2aa0bac517ba0aec787fa0738fd7a0bf650..d2494048bc1491028e38fdd3880169ca5e3c45ac 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q4_1.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q4_1.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_1, GGML_TYPE_Q4_1);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_Q4_1);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_1, GGML_TYPE_Q4_1);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q5_1, GGML_TYPE_Q4_1);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q5_0.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q5_0.cpp
index d7075bac600fe360d3078ba74f1e033ef65ce3fa..be3a1fe97f54b5cb7fb0ac7e0f80872b11b5d8d9 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q5_0.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q5_0.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_1, GGML_TYPE_Q5_0);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_Q5_0);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_1, GGML_TYPE_Q5_0);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q5_1, GGML_TYPE_Q5_0);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q5_1.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q5_1.cpp
index 627f9a577555c320264b99793df5a95bdf234d13..be0a89409cafc8bf0ec8ed5f1aedf0a11c742912 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q5_1.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q5_1.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_1, GGML_TYPE_Q5_1);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_Q5_1);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_1, GGML_TYPE_Q5_1);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q5_1, GGML_TYPE_Q5_1);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q8_0.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q8_0.cpp
index 23304eecd355a87a4bb8b9ced023f019501f7120..6781efcb0d22c52d59b837e5ef338489ca767267 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q8_0.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q8_0.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q5_1, GGML_TYPE_Q8_0);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q5_1, GGML_TYPE_Q8_0);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q5_1, GGML_TYPE_Q8_0);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q5_1, GGML_TYPE_Q8_0);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-f16.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-f16.cpp
index 95acb5d4fbf0399a7fa98943cf0cad5c4d961aa1..43a70ae3543f87a7673e75fc54ac775399a65c9c 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-f16.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-f16.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q8_0, GGML_TYPE_F16);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_F16);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q8_0, GGML_TYPE_F16);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q8_0, GGML_TYPE_F16);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q4_0.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q4_0.cpp
index 5e88f4bab8a9a35ce8f1467cd40845ee8a91f778..fa7eb8163cac0717e30ce18303f910618512dcd1 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q4_0.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q4_0.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q8_0, GGML_TYPE_Q4_0);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q4_0);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q8_0, GGML_TYPE_Q4_0);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q8_0, GGML_TYPE_Q4_0);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q4_1.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q4_1.cpp
index 69f297feb0c9b48cc53df8594a9ed0a2cfe967a6..79d9cfbee96584160c219f586bb51b5699e754fc 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q4_1.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q4_1.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q8_0, GGML_TYPE_Q4_1);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q4_1);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q8_0, GGML_TYPE_Q4_1);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q8_0, GGML_TYPE_Q4_1);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q5_0.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q5_0.cpp
index 455842a9421281a772fa4d33bb252942c977dc2a..86befd5d3273b46f5e70c47324ad86d3dabadd2a 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q5_0.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q5_0.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q8_0, GGML_TYPE_Q5_0);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q5_0);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q8_0, GGML_TYPE_Q5_0);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q8_0, GGML_TYPE_Q5_0);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q5_1.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q5_1.cpp
index f7ef73915712671c2d356224a1a02477a0621e84..c2f619b0b166e43d6e69bbcba6ea1d27610c6318 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q5_1.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q5_1.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q8_0, GGML_TYPE_Q5_1);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q5_1);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q8_0, GGML_TYPE_Q5_1);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q8_0, GGML_TYPE_Q5_1);

diff --git a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q8_0.cpp b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q8_0.cpp
index 1c633bdf2fa98244f1f0eacd69deb926fd795303..7cf31f8b8a135ffc664938b65b46123db31109f6 100644 (file)
--- a/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q8_0.cpp
+++ b/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q8_0.cpp
@@ -5,3 +5,4 @@
 DECL_FATTN_VEC_CASE( 64, GGML_TYPE_Q8_0, GGML_TYPE_Q8_0);
 DECL_FATTN_VEC_CASE(128, GGML_TYPE_Q8_0, GGML_TYPE_Q8_0);
 DECL_FATTN_VEC_CASE(256, GGML_TYPE_Q8_0, GGML_TYPE_Q8_0);
+DECL_FATTN_VEC_CASE(512, GGML_TYPE_Q8_0, GGML_TYPE_Q8_0);