cmake .. -G "Ninja" -DCMAKE_BUILD_TYPE=Release -DLLAMA_HIPBLAS=ON -DLLAMA_CUDA_DMMV_X=32 -DLLAMA_CUDA_MMV_Y=2 -DLLAMA_CUDA_KQUANTS_ITER=2 -DCMAKE_C_COMPILER="C:/Program Files/AMD/ROCm/5.7/bin/clang.exe" -DCMAKE_CXX_COMPILER="C:/Program Files/AMD/ROCm/5.7/bin/clang++.exe" -DAMDGPU_TARGETS="gfx1100"



cmake --build . --config Release
[1/54] Generating build details from Git
-- Found Git: C:/Program Files/Git/cmd/git.exe (found version "2.43.0.windows.1")
_: line 1: C:/Program: No such file or directory
[3/54] Building CXX object CMakeFiles/ggml.dir/ggml-cuda.cu.obj
FAILED: CMakeFiles/ggml.dir/ggml-cuda.cu.obj
ccache C:\PROGRA~1\AMD\ROCm\5.7\bin\CLANG_~1.EXE -DGGML_CUDA_DMMV_X=32 -DGGML_CUDA_MMV_Y=2 -DGGML_USE_CUBLAS -DGGML_USE_HIPBLAS -DK_QUANTS_PER_ITERATION=2 -D_CRT_SECURE_NO_WARNINGS -D_XOPEN_SOURCE=600 -D__HIP_PLATFORM_AMD__=1 -D__HIP_PLATFORM_HCC__=1 -IW:/git/test/llama.cpp/. -isystem "C:/Program Files/AMD/ROCm/5.7/include" -O3 -DNDEBUG -D_DLL -D_MT -Xclang --dependent-lib=msvcrt -std=gnu++14 -Wmissing-declarations -Wmissing-noreturn -Wall -Wextra -Wpedantic -Wcast-qual -Wno-unused-function -Wunreachable-code-break -Wunreachable-code-return -Wmissing-prototypes -Wextra-semi -march=native -mllvm -amdgpu-early-inline-all=true -mllvm -amdgpu-function-calls=false -x hip --offload-arch=gfx1100 -MD -MT CMakeFiles/ggml.dir/ggml-cuda.cu.obj -MF CMakeFiles\ggml.dir\ggml-cuda.cu.obj.d -o CMakeFiles/ggml.dir/ggml-cuda.cu.obj -c W:/git/test/llama.cpp/ggml-cuda.cu
W:/git/test/llama.cpp/ggml-cuda.cu:175:17: error: typedef redefinition with different types ('uint8_t __attribute__((ext_vector_type(4)))' (vector of 4 'uint8_t' values) vs 'int8_t __attribute__((ext_vector_type(4)))' (vector of 4 'int8_t' values))
typedef uint8_t int8x4_t __attribute__((ext_vector_type(4)));
                ^
W:/git/test/llama.cpp/ggml-cuda.cu:174:16: note: previous definition is here
typedef int8_t int8x4_t __attribute__((ext_vector_type(4)));
               ^
W:/git/test/llama.cpp/ggml-cuda.cu:201:11: error: unknown type name 'uint8x4_t'; did you mean 'int8x4_t'?
    const uint8x4_t& va = reinterpret_cast<const uint8x4_t&>(a);
          ^~~~~~~~~
          int8x4_t
W:/git/test/llama.cpp/ggml-cuda.cu:174:16: note: 'int8x4_t' declared here
typedef int8_t int8x4_t __attribute__((ext_vector_type(4)));
               ^
W:/git/test/llama.cpp/ggml-cuda.cu:201:50: error: unknown type name 'uint8x4_t'; did you mean 'int8x4_t'?
    const uint8x4_t& va = reinterpret_cast<const uint8x4_t&>(a);
                                                 ^~~~~~~~~
                                                 int8x4_t
W:/git/test/llama.cpp/ggml-cuda.cu:174:16: note: 'int8x4_t' declared here
typedef int8_t int8x4_t __attribute__((ext_vector_type(4)));
               ^
W:/git/test/llama.cpp/ggml-cuda.cu:202:11: error: unknown type name 'uint8x4_t'; did you mean 'int8x4_t'?
    const uint8x4_t& vb = reinterpret_cast<const uint8x4_t&>(b);
          ^~~~~~~~~
          int8x4_t
W:/git/test/llama.cpp/ggml-cuda.cu:174:16: note: 'int8x4_t' declared here
typedef int8_t int8x4_t __attribute__((ext_vector_type(4)));
               ^
W:/git/test/llama.cpp/ggml-cuda.cu:202:50: error: unknown type name 'uint8x4_t'; did you mean 'int8x4_t'?
    const uint8x4_t& vb = reinterpret_cast<const uint8x4_t&>(b);
                                                 ^~~~~~~~~
                                                 int8x4_t
W:/git/test/llama.cpp/ggml-cuda.cu:174:16: note: 'int8x4_t' declared here
typedef int8_t int8x4_t __attribute__((ext_vector_type(4)));
               ^
W:/git/test/llama.cpp/ggml-cuda.cu:204:5: error: unknown type name 'uint8x4_t'; did you mean 'int8x4_t'?
    uint8x4_t& vc = reinterpret_cast<uint8x4_t&>(c);
    ^~~~~~~~~
    int8x4_t
W:/git/test/llama.cpp/ggml-cuda.cu:174:16: note: 'int8x4_t' declared here
typedef int8_t int8x4_t __attribute__((ext_vector_type(4)));
               ^
W:/git/test/llama.cpp/ggml-cuda.cu:204:38: error: unknown type name 'uint8x4_t'; did you mean 'int8x4_t'?
    uint8x4_t& vc = reinterpret_cast<uint8x4_t&>(c);
                                     ^~~~~~~~~
                                     int8x4_t
W:/git/test/llama.cpp/ggml-cuda.cu:174:16: note: 'int8x4_t' declared here
typedef int8_t int8x4_t __attribute__((ext_vector_type(4)));
               ^
W:/git/test/llama.cpp/ggml-cuda.cu:953:20: warning: comparison of integers of different signs: 'R' (aka 'unsigned int') and 'const int' [-Wsign-compare]
    if (blockIdx.z < ne02) { // src0
        ~~~~~~~~~~ ^ ~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:999:56: warning: comparison of integers of different signs: 'R' (aka 'unsigned int') and 'const int' [-Wsign-compare]
    if (nidx < ne00 && blockIdx.y < ne01 && blockIdx.z < ne02) {
                                            ~~~~~~~~~~ ^ ~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:999:35: warning: comparison of integers of different signs: 'R' (aka 'unsigned int') and 'const int' [-Wsign-compare]
    if (nidx < ne00 && blockIdx.y < ne01 && blockIdx.z < ne02) {
                       ~~~~~~~~~~ ^ ~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:4821:40: warning: cast from 'const signed char *' to 'int *' drops const qualifier [-Wcast-qual]
        sumi = __dp4a(grid_l, *((int *)q8+0), sumi);
                                       ^
W:/git/test/llama.cpp/ggml-cuda.cu:4822:40: warning: cast from 'const signed char *' to 'int *' drops const qualifier [-Wcast-qual]
        sumi = __dp4a(grid_h, *((int *)q8+1), sumi);
                                       ^
W:/git/test/llama.cpp/ggml-cuda.cu:4856:40: warning: cast from 'const signed char *' to 'int *' drops const qualifier [-Wcast-qual]
        sumi = __dp4a(grid_l, *((int *)q8+0), sumi);
                                       ^
W:/git/test/llama.cpp/ggml-cuda.cu:4857:40: warning: cast from 'const signed char *' to 'int *' drops const qualifier [-Wcast-qual]
        sumi = __dp4a(grid_h, *((int *)q8+1), sumi);
                                       ^
W:/git/test/llama.cpp/ggml-cuda.cu:10266:10: warning: variable 'any_pascal_with_slow_fp16' set but not used [-Wunused-but-set-variable]
    bool any_pascal_with_slow_fp16 = false;
         ^
W:/git/test/llama.cpp/ggml-cuda.cu:2858:31: warning: cast from 'const void *' to 'int *' drops const qualifier [-Wcast-qual]
    const int * x0 = ((int *) vx) + blockIdx.x * nint;
                              ^
W:/git/test/llama.cpp/ggml-cuda.cu:2858:31: warning: cast from 'const void *' to 'int *' drops const qualifier [-Wcast-qual]
W:/git/test/llama.cpp/ggml-cuda.cu:6920:9: note: in instantiation of function template specialization 'dequantize_block_q8_0_f16<false>' requested here
        dequantize_block_q8_0_f16<need_check><<<num_blocks, WARP_SIZE, 0, stream>>>(vx, y, k);
        ^
W:/git/test/llama.cpp/ggml-cuda.cu:2875:51: warning: comparison of integers of different signs: 'R' (aka 'unsigned int') and 'const int' [-Wsign-compare]
        if (need_check && i0 + iy + 2*threadIdx.x >= k) {
                          ~~~~~~~~~~~~~~~~~~~~~~~ ^  ~
W:/git/test/llama.cpp/ggml-cuda.cu:2858:31: warning: cast from 'const void *' to 'int *' drops const qualifier [-Wcast-qual]
    const int * x0 = ((int *) vx) + blockIdx.x * nint;
                              ^
W:/git/test/llama.cpp/ggml-cuda.cu:6923:9: note: in instantiation of function template specialization 'dequantize_block_q8_0_f16<true>' requested here
        dequantize_block_q8_0_f16<need_check><<<num_blocks, WARP_SIZE, 0, stream>>>(vx, y, k);
        ^
W:/git/test/llama.cpp/ggml-cuda.cu:2875:51: warning: comparison of integers of different signs: 'R' (aka 'unsigned int') and 'const int' [-Wsign-compare]
        if (need_check && i0 + iy + 2*threadIdx.x >= k) {
                          ~~~~~~~~~~~~~~~~~~~~~~~ ^  ~
W:/git/test/llama.cpp/ggml-cuda.cu:2847:33: warning: cast from 'const void *__restrict' to 'type-parameter-0-0 *' drops const qualifier [-Wcast-qual]
    const src_t * x = (src_t *) vx;
                                ^
W:/git/test/llama.cpp/ggml-cuda.cu:7069:20: note: in instantiation of function template specialization 'convert_unary_cuda<float, __half>' requested here
            return convert_unary_cuda<float>;
                   ^
W:/git/test/llama.cpp/ggml-cuda.cu:2847:33: warning: cast from 'const void *' to 'float *' drops const qualifier [-Wcast-qual]
    const src_t * x = (src_t *) vx;
                                ^
W:/git/test/llama.cpp/ggml-cuda.cu:7026:5: note: in instantiation of function template specialization 'convert_unary<float, __half>' requested here
    convert_unary<src_t><<<num_blocks, CUDA_DEQUANTIZE_BLOCK_SIZE, 0, stream>>>(vx, y, k);
    ^
W:/git/test/llama.cpp/ggml-cuda.cu:7069:20: note: in instantiation of function template specialization 'convert_unary_cuda<float, __half>' requested here
            return convert_unary_cuda<float>;
                   ^
W:/git/test/llama.cpp/ggml-cuda.cu:2847:33: warning: cast from 'const void *' to '__half *' drops const qualifier [-Wcast-qual]
    const src_t * x = (src_t *) vx;
                                ^
W:/git/test/llama.cpp/ggml-cuda.cu:7026:5: note: in instantiation of function template specialization 'convert_unary<__half, float>' requested here
    convert_unary<src_t><<<num_blocks, CUDA_DEQUANTIZE_BLOCK_SIZE, 0, stream>>>(vx, y, k);
    ^
W:/git/test/llama.cpp/ggml-cuda.cu:7110:20: note: in instantiation of function template specialization 'convert_unary_cuda<__half, float>' requested here
            return convert_unary_cuda<half>;
                   ^
W:/git/test/llama.cpp/ggml-cuda.cu:6419:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const float base = h < n_head_log2 ? m0 : m1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:6420:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const int   exp  = h < n_head_log2 ? h + 1 : 2*(h - n_head_log2) + 1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:6419:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const float base = h < n_head_log2 ? m0 : m1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:7963:17: note: in instantiation of function template specialization 'soft_max_f32<true, 32, 32>' requested here
                soft_max_f32<true, 32, 32><<<block_nums, block_dims, shmem, stream>>>(x, mask, pos, dst, ncols_x, nrows_y, scale, max_bias, m0, m1, n_head_log2);
                ^
W:/git/test/llama.cpp/ggml-cuda.cu:6420:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const int   exp  = h < n_head_log2 ? h + 1 : 2*(h - n_head_log2) + 1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:6419:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const float base = h < n_head_log2 ? m0 : m1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:7966:17: note: in instantiation of function template specialization 'soft_max_f32<true, 64, 64>' requested here
                soft_max_f32<true, 64, 64><<<block_nums, block_dims, shmem, stream>>>(x, mask, pos, dst, ncols_x, nrows_y, scale, max_bias, m0, m1, n_head_log2);
                ^
W:/git/test/llama.cpp/ggml-cuda.cu:6420:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const int   exp  = h < n_head_log2 ? h + 1 : 2*(h - n_head_log2) + 1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:6419:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const float base = h < n_head_log2 ? m0 : m1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:7969:17: note: in instantiation of function template specialization 'soft_max_f32<true, 128, 128>' requested here
                soft_max_f32<true, 128, 128><<<block_nums, block_dims, shmem, stream>>>(x, mask, pos, dst, ncols_x, nrows_y, scale, max_bias, m0, m1, n_head_log2);
                ^
W:/git/test/llama.cpp/ggml-cuda.cu:6420:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const int   exp  = h < n_head_log2 ? h + 1 : 2*(h - n_head_log2) + 1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:6419:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const float base = h < n_head_log2 ? m0 : m1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:7972:17: note: in instantiation of function template specialization 'soft_max_f32<true, 256, 256>' requested here
                soft_max_f32<true, 256, 256><<<block_nums, block_dims, shmem, stream>>>(x, mask, pos, dst, ncols_x, nrows_y, scale, max_bias, m0, m1, n_head_log2);
                ^
W:/git/test/llama.cpp/ggml-cuda.cu:6420:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const int   exp  = h < n_head_log2 ? h + 1 : 2*(h - n_head_log2) + 1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:6419:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const float base = h < n_head_log2 ? m0 : m1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:7975:17: note: in instantiation of function template specialization 'soft_max_f32<true, 512, 512>' requested here
                soft_max_f32<true, 512, 512><<<block_nums, block_dims, shmem, stream>>>(x, mask, pos, dst, ncols_x, nrows_y, scale, max_bias, m0, m1, n_head_log2);
                ^
W:/git/test/llama.cpp/ggml-cuda.cu:6420:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const int   exp  = h < n_head_log2 ? h + 1 : 2*(h - n_head_log2) + 1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:6419:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const float base = h < n_head_log2 ? m0 : m1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:7978:17: note: in instantiation of function template specialization 'soft_max_f32<true, 1024, 1024>' requested here
                soft_max_f32<true, 1024, 1024><<<block_nums, block_dims, shmem, stream>>>(x, mask, pos, dst, ncols_x, nrows_y, scale, max_bias, m0, m1, n_head_log2);
                ^
W:/git/test/llama.cpp/ggml-cuda.cu:6420:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const int   exp  = h < n_head_log2 ? h + 1 : 2*(h - n_head_log2) + 1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:6419:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const float base = h < n_head_log2 ? m0 : m1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:7981:17: note: in instantiation of function template specialization 'soft_max_f32<true, 2048, 1024>' requested here
                soft_max_f32<true, 2048, 1024><<<block_nums, block_dims, shmem, stream>>>(x, mask, pos, dst, ncols_x, nrows_y, scale, max_bias, m0, m1, n_head_log2);
                ^
W:/git/test/llama.cpp/ggml-cuda.cu:6420:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const int   exp  = h < n_head_log2 ? h + 1 : 2*(h - n_head_log2) + 1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:6419:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const float base = h < n_head_log2 ? m0 : m1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:7984:17: note: in instantiation of function template specialization 'soft_max_f32<true, 4096, 1024>' requested here
                soft_max_f32<true, 4096, 1024><<<block_nums, block_dims, shmem, stream>>>(x, mask, pos, dst, ncols_x, nrows_y, scale, max_bias, m0, m1, n_head_log2);
                ^
W:/git/test/llama.cpp/ggml-cuda.cu:6420:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const int   exp  = h < n_head_log2 ? h + 1 : 2*(h - n_head_log2) + 1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:6419:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const float base = h < n_head_log2 ? m0 : m1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:7987:17: note: in instantiation of function template specialization 'soft_max_f32<true, 0, 0>' requested here
                soft_max_f32<true, 0, 0><<<block_nums, block_dims, shmem, stream>>>(x, mask, pos, dst, ncols_x, nrows_y, scale, max_bias, m0, m1, n_head_log2);
                ^
W:/git/test/llama.cpp/ggml-cuda.cu:6420:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const int   exp  = h < n_head_log2 ? h + 1 : 2*(h - n_head_log2) + 1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:6419:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const float base = h < n_head_log2 ? m0 : m1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:7992:9: note: in instantiation of function template specialization 'soft_max_f32<false, 0, 0>' requested here
        soft_max_f32<false, 0, 0><<<block_nums, block_dims, shmem_low, stream>>>(x, mask, pos, dst, ncols_x, nrows_y, scale, max_bias, m0, m1, n_head_log2);
        ^
W:/git/test/llama.cpp/ggml-cuda.cu:6420:30: warning: comparison of integers of different signs: 'const int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
        const int   exp  = h < n_head_log2 ? h + 1 : 2*(h - n_head_log2) + 1;
                           ~ ^ ~~~~~~~~~~~
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7257:13: note: in instantiation of function template specialization 'mul_mat_vec_q<1, 32, 4, block_q4_0, 2, &vec_dot_q4_0_q8_1>' requested here
            mul_mat_vec_q<1, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8927:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q4_0, 2, &vec_dot_q4_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_0, QI4_0, block_q4_0, VDR_Q4_0_Q8_1_MMVQ, vec_dot_q4_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7261:13: note: in instantiation of function template specialization 'mul_mat_vec_q<2, 32, 4, block_q4_0, 2, &vec_dot_q4_0_q8_1>' requested here
            mul_mat_vec_q<2, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8927:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q4_0, 2, &vec_dot_q4_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_0, QI4_0, block_q4_0, VDR_Q4_0_Q8_1_MMVQ, vec_dot_q4_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7265:13: note: in instantiation of function template specialization 'mul_mat_vec_q<3, 32, 4, block_q4_0, 2, &vec_dot_q4_0_q8_1>' requested here
            mul_mat_vec_q<3, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8927:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q4_0, 2, &vec_dot_q4_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_0, QI4_0, block_q4_0, VDR_Q4_0_Q8_1_MMVQ, vec_dot_q4_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7269:13: note: in instantiation of function template specialization 'mul_mat_vec_q<4, 32, 4, block_q4_0, 2, &vec_dot_q4_0_q8_1>' requested here
            mul_mat_vec_q<4, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8927:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q4_0, 2, &vec_dot_q4_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_0, QI4_0, block_q4_0, VDR_Q4_0_Q8_1_MMVQ, vec_dot_q4_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7273:13: note: in instantiation of function template specialization 'mul_mat_vec_q<5, 32, 4, block_q4_0, 2, &vec_dot_q4_0_q8_1>' requested here
            mul_mat_vec_q<5, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8927:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q4_0, 2, &vec_dot_q4_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_0, QI4_0, block_q4_0, VDR_Q4_0_Q8_1_MMVQ, vec_dot_q4_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7277:13: note: in instantiation of function template specialization 'mul_mat_vec_q<6, 32, 4, block_q4_0, 2, &vec_dot_q4_0_q8_1>' requested here
            mul_mat_vec_q<6, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8927:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q4_0, 2, &vec_dot_q4_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_0, QI4_0, block_q4_0, VDR_Q4_0_Q8_1_MMVQ, vec_dot_q4_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7281:13: note: in instantiation of function template specialization 'mul_mat_vec_q<7, 32, 4, block_q4_0, 2, &vec_dot_q4_0_q8_1>' requested here
            mul_mat_vec_q<7, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8927:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q4_0, 2, &vec_dot_q4_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_0, QI4_0, block_q4_0, VDR_Q4_0_Q8_1_MMVQ, vec_dot_q4_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7285:13: note: in instantiation of function template specialization 'mul_mat_vec_q<8, 32, 4, block_q4_0, 2, &vec_dot_q4_0_q8_1>' requested here
            mul_mat_vec_q<8, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8927:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q4_0, 2, &vec_dot_q4_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_0, QI4_0, block_q4_0, VDR_Q4_0_Q8_1_MMVQ, vec_dot_q4_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7257:13: note: in instantiation of function template specialization 'mul_mat_vec_q<1, 32, 4, block_q4_1, 2, &vec_dot_q4_1_q8_1>' requested here
            mul_mat_vec_q<1, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8931:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q4_1, 2, &vec_dot_q4_1_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_1, QI4_1, block_q4_1, VDR_Q4_1_Q8_1_MMVQ, vec_dot_q4_1_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7261:13: note: in instantiation of function template specialization 'mul_mat_vec_q<2, 32, 4, block_q4_1, 2, &vec_dot_q4_1_q8_1>' requested here
            mul_mat_vec_q<2, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8931:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q4_1, 2, &vec_dot_q4_1_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_1, QI4_1, block_q4_1, VDR_Q4_1_Q8_1_MMVQ, vec_dot_q4_1_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7265:13: note: in instantiation of function template specialization 'mul_mat_vec_q<3, 32, 4, block_q4_1, 2, &vec_dot_q4_1_q8_1>' requested here
            mul_mat_vec_q<3, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8931:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q4_1, 2, &vec_dot_q4_1_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_1, QI4_1, block_q4_1, VDR_Q4_1_Q8_1_MMVQ, vec_dot_q4_1_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7269:13: note: in instantiation of function template specialization 'mul_mat_vec_q<4, 32, 4, block_q4_1, 2, &vec_dot_q4_1_q8_1>' requested here
            mul_mat_vec_q<4, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8931:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q4_1, 2, &vec_dot_q4_1_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_1, QI4_1, block_q4_1, VDR_Q4_1_Q8_1_MMVQ, vec_dot_q4_1_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7273:13: note: in instantiation of function template specialization 'mul_mat_vec_q<5, 32, 4, block_q4_1, 2, &vec_dot_q4_1_q8_1>' requested here
            mul_mat_vec_q<5, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8931:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q4_1, 2, &vec_dot_q4_1_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_1, QI4_1, block_q4_1, VDR_Q4_1_Q8_1_MMVQ, vec_dot_q4_1_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7277:13: note: in instantiation of function template specialization 'mul_mat_vec_q<6, 32, 4, block_q4_1, 2, &vec_dot_q4_1_q8_1>' requested here
            mul_mat_vec_q<6, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8931:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q4_1, 2, &vec_dot_q4_1_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_1, QI4_1, block_q4_1, VDR_Q4_1_Q8_1_MMVQ, vec_dot_q4_1_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7281:13: note: in instantiation of function template specialization 'mul_mat_vec_q<7, 32, 4, block_q4_1, 2, &vec_dot_q4_1_q8_1>' requested here
            mul_mat_vec_q<7, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8931:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q4_1, 2, &vec_dot_q4_1_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_1, QI4_1, block_q4_1, VDR_Q4_1_Q8_1_MMVQ, vec_dot_q4_1_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7285:13: note: in instantiation of function template specialization 'mul_mat_vec_q<8, 32, 4, block_q4_1, 2, &vec_dot_q4_1_q8_1>' requested here
            mul_mat_vec_q<8, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8931:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q4_1, 2, &vec_dot_q4_1_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_1, QI4_1, block_q4_1, VDR_Q4_1_Q8_1_MMVQ, vec_dot_q4_1_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7257:13: note: in instantiation of function template specialization 'mul_mat_vec_q<1, 32, 4, block_q5_0, 2, &vec_dot_q5_0_q8_1>' requested here
            mul_mat_vec_q<1, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8935:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q5_0, 2, &vec_dot_q5_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK5_0, QI5_0, block_q5_0, VDR_Q5_0_Q8_1_MMVQ, vec_dot_q5_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7261:13: note: in instantiation of function template specialization 'mul_mat_vec_q<2, 32, 4, block_q5_0, 2, &vec_dot_q5_0_q8_1>' requested here
            mul_mat_vec_q<2, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8935:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q5_0, 2, &vec_dot_q5_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK5_0, QI5_0, block_q5_0, VDR_Q5_0_Q8_1_MMVQ, vec_dot_q5_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7265:13: note: in instantiation of function template specialization 'mul_mat_vec_q<3, 32, 4, block_q5_0, 2, &vec_dot_q5_0_q8_1>' requested here
            mul_mat_vec_q<3, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8935:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q5_0, 2, &vec_dot_q5_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK5_0, QI5_0, block_q5_0, VDR_Q5_0_Q8_1_MMVQ, vec_dot_q5_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7269:13: note: in instantiation of function template specialization 'mul_mat_vec_q<4, 32, 4, block_q5_0, 2, &vec_dot_q5_0_q8_1>' requested here
            mul_mat_vec_q<4, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8935:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q5_0, 2, &vec_dot_q5_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK5_0, QI5_0, block_q5_0, VDR_Q5_0_Q8_1_MMVQ, vec_dot_q5_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7273:13: note: in instantiation of function template specialization 'mul_mat_vec_q<5, 32, 4, block_q5_0, 2, &vec_dot_q5_0_q8_1>' requested here
            mul_mat_vec_q<5, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8935:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q5_0, 2, &vec_dot_q5_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK5_0, QI5_0, block_q5_0, VDR_Q5_0_Q8_1_MMVQ, vec_dot_q5_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7277:13: note: in instantiation of function template specialization 'mul_mat_vec_q<6, 32, 4, block_q5_0, 2, &vec_dot_q5_0_q8_1>' requested here
            mul_mat_vec_q<6, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8935:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q5_0, 2, &vec_dot_q5_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK5_0, QI5_0, block_q5_0, VDR_Q5_0_Q8_1_MMVQ, vec_dot_q5_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7281:13: note: in instantiation of function template specialization 'mul_mat_vec_q<7, 32, 4, block_q5_0, 2, &vec_dot_q5_0_q8_1>' requested here
            mul_mat_vec_q<7, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8935:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q5_0, 2, &vec_dot_q5_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK5_0, QI5_0, block_q5_0, VDR_Q5_0_Q8_1_MMVQ, vec_dot_q5_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7285:13: note: in instantiation of function template specialization 'mul_mat_vec_q<8, 32, 4, block_q5_0, 2, &vec_dot_q5_0_q8_1>' requested here
            mul_mat_vec_q<8, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8935:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q5_0, 2, &vec_dot_q5_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK5_0, QI5_0, block_q5_0, VDR_Q5_0_Q8_1_MMVQ, vec_dot_q5_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7257:13: note: in instantiation of function template specialization 'mul_mat_vec_q<1, 32, 4, block_q5_1, 2, &vec_dot_q5_1_q8_1>' requested here
            mul_mat_vec_q<1, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8939:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q5_1, 2, &vec_dot_q5_1_q8_1>' requested here
            mul_mat_vec_q_cuda<QK5_1, QI5_1, block_q5_1, VDR_Q5_1_Q8_1_MMVQ, vec_dot_q5_1_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7261:13: note: in instantiation of function template specialization 'mul_mat_vec_q<2, 32, 4, block_q5_1, 2, &vec_dot_q5_1_q8_1>' requested here
            mul_mat_vec_q<2, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8939:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q5_1, 2, &vec_dot_q5_1_q8_1>' requested here
            mul_mat_vec_q_cuda<QK5_1, QI5_1, block_q5_1, VDR_Q5_1_Q8_1_MMVQ, vec_dot_q5_1_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7265:13: note: in instantiation of function template specialization 'mul_mat_vec_q<3, 32, 4, block_q5_1, 2, &vec_dot_q5_1_q8_1>' requested here
            mul_mat_vec_q<3, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8939:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q5_1, 2, &vec_dot_q5_1_q8_1>' requested here
            mul_mat_vec_q_cuda<QK5_1, QI5_1, block_q5_1, VDR_Q5_1_Q8_1_MMVQ, vec_dot_q5_1_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7269:13: note: in instantiation of function template specialization 'mul_mat_vec_q<4, 32, 4, block_q5_1, 2, &vec_dot_q5_1_q8_1>' requested here
            mul_mat_vec_q<4, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8939:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q5_1, 2, &vec_dot_q5_1_q8_1>' requested here
            mul_mat_vec_q_cuda<QK5_1, QI5_1, block_q5_1, VDR_Q5_1_Q8_1_MMVQ, vec_dot_q5_1_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7273:13: note: in instantiation of function template specialization 'mul_mat_vec_q<5, 32, 4, block_q5_1, 2, &vec_dot_q5_1_q8_1>' requested here
            mul_mat_vec_q<5, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8939:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q5_1, 2, &vec_dot_q5_1_q8_1>' requested here
            mul_mat_vec_q_cuda<QK5_1, QI5_1, block_q5_1, VDR_Q5_1_Q8_1_MMVQ, vec_dot_q5_1_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7277:13: note: in instantiation of function template specialization 'mul_mat_vec_q<6, 32, 4, block_q5_1, 2, &vec_dot_q5_1_q8_1>' requested here
            mul_mat_vec_q<6, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8939:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q5_1, 2, &vec_dot_q5_1_q8_1>' requested here
            mul_mat_vec_q_cuda<QK5_1, QI5_1, block_q5_1, VDR_Q5_1_Q8_1_MMVQ, vec_dot_q5_1_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7281:13: note: in instantiation of function template specialization 'mul_mat_vec_q<7, 32, 4, block_q5_1, 2, &vec_dot_q5_1_q8_1>' requested here
            mul_mat_vec_q<7, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8939:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q5_1, 2, &vec_dot_q5_1_q8_1>' requested here
            mul_mat_vec_q_cuda<QK5_1, QI5_1, block_q5_1, VDR_Q5_1_Q8_1_MMVQ, vec_dot_q5_1_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7285:13: note: in instantiation of function template specialization 'mul_mat_vec_q<8, 32, 4, block_q5_1, 2, &vec_dot_q5_1_q8_1>' requested here
            mul_mat_vec_q<8, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8939:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_q5_1, 2, &vec_dot_q5_1_q8_1>' requested here
            mul_mat_vec_q_cuda<QK5_1, QI5_1, block_q5_1, VDR_Q5_1_Q8_1_MMVQ, vec_dot_q5_1_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7257:13: note: in instantiation of function template specialization 'mul_mat_vec_q<1, 32, 8, block_q8_0, 2, &vec_dot_q8_0_q8_1>' requested here
            mul_mat_vec_q<1, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8943:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 8, block_q8_0, 2, &vec_dot_q8_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK8_0, QI8_0, block_q8_0, VDR_Q8_0_Q8_1_MMVQ, vec_dot_q8_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7261:13: note: in instantiation of function template specialization 'mul_mat_vec_q<2, 32, 8, block_q8_0, 2, &vec_dot_q8_0_q8_1>' requested here
            mul_mat_vec_q<2, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8943:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 8, block_q8_0, 2, &vec_dot_q8_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK8_0, QI8_0, block_q8_0, VDR_Q8_0_Q8_1_MMVQ, vec_dot_q8_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7265:13: note: in instantiation of function template specialization 'mul_mat_vec_q<3, 32, 8, block_q8_0, 2, &vec_dot_q8_0_q8_1>' requested here
            mul_mat_vec_q<3, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8943:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 8, block_q8_0, 2, &vec_dot_q8_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK8_0, QI8_0, block_q8_0, VDR_Q8_0_Q8_1_MMVQ, vec_dot_q8_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7269:13: note: in instantiation of function template specialization 'mul_mat_vec_q<4, 32, 8, block_q8_0, 2, &vec_dot_q8_0_q8_1>' requested here
            mul_mat_vec_q<4, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8943:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 8, block_q8_0, 2, &vec_dot_q8_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK8_0, QI8_0, block_q8_0, VDR_Q8_0_Q8_1_MMVQ, vec_dot_q8_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7273:13: note: in instantiation of function template specialization 'mul_mat_vec_q<5, 32, 8, block_q8_0, 2, &vec_dot_q8_0_q8_1>' requested here
            mul_mat_vec_q<5, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8943:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 8, block_q8_0, 2, &vec_dot_q8_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK8_0, QI8_0, block_q8_0, VDR_Q8_0_Q8_1_MMVQ, vec_dot_q8_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7277:13: note: in instantiation of function template specialization 'mul_mat_vec_q<6, 32, 8, block_q8_0, 2, &vec_dot_q8_0_q8_1>' requested here
            mul_mat_vec_q<6, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8943:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 8, block_q8_0, 2, &vec_dot_q8_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK8_0, QI8_0, block_q8_0, VDR_Q8_0_Q8_1_MMVQ, vec_dot_q8_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7281:13: note: in instantiation of function template specialization 'mul_mat_vec_q<7, 32, 8, block_q8_0, 2, &vec_dot_q8_0_q8_1>' requested here
            mul_mat_vec_q<7, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8943:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 8, block_q8_0, 2, &vec_dot_q8_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK8_0, QI8_0, block_q8_0, VDR_Q8_0_Q8_1_MMVQ, vec_dot_q8_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7285:13: note: in instantiation of function template specialization 'mul_mat_vec_q<8, 32, 8, block_q8_0, 2, &vec_dot_q8_0_q8_1>' requested here
            mul_mat_vec_q<8, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8943:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 8, block_q8_0, 2, &vec_dot_q8_0_q8_1>' requested here
            mul_mat_vec_q_cuda<QK8_0, QI8_0, block_q8_0, VDR_Q8_0_Q8_1_MMVQ, vec_dot_q8_0_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7257:13: note: in instantiation of function template specialization 'mul_mat_vec_q<1, 256, 16, block_q2_K, 1, &vec_dot_q2_K_q8_1>' requested here
            mul_mat_vec_q<1, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8947:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 16, block_q2_K, 1, &vec_dot_q2_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_K, block_q2_K, VDR_Q2_K_Q8_1_MMVQ, vec_dot_q2_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7261:13: note: in instantiation of function template specialization 'mul_mat_vec_q<2, 256, 16, block_q2_K, 1, &vec_dot_q2_K_q8_1>' requested here
            mul_mat_vec_q<2, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8947:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 16, block_q2_K, 1, &vec_dot_q2_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_K, block_q2_K, VDR_Q2_K_Q8_1_MMVQ, vec_dot_q2_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7265:13: note: in instantiation of function template specialization 'mul_mat_vec_q<3, 256, 16, block_q2_K, 1, &vec_dot_q2_K_q8_1>' requested here
            mul_mat_vec_q<3, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8947:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 16, block_q2_K, 1, &vec_dot_q2_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_K, block_q2_K, VDR_Q2_K_Q8_1_MMVQ, vec_dot_q2_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7269:13: note: in instantiation of function template specialization 'mul_mat_vec_q<4, 256, 16, block_q2_K, 1, &vec_dot_q2_K_q8_1>' requested here
            mul_mat_vec_q<4, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8947:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 16, block_q2_K, 1, &vec_dot_q2_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_K, block_q2_K, VDR_Q2_K_Q8_1_MMVQ, vec_dot_q2_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7273:13: note: in instantiation of function template specialization 'mul_mat_vec_q<5, 256, 16, block_q2_K, 1, &vec_dot_q2_K_q8_1>' requested here
            mul_mat_vec_q<5, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8947:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 16, block_q2_K, 1, &vec_dot_q2_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_K, block_q2_K, VDR_Q2_K_Q8_1_MMVQ, vec_dot_q2_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7277:13: note: in instantiation of function template specialization 'mul_mat_vec_q<6, 256, 16, block_q2_K, 1, &vec_dot_q2_K_q8_1>' requested here
            mul_mat_vec_q<6, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8947:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 16, block_q2_K, 1, &vec_dot_q2_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_K, block_q2_K, VDR_Q2_K_Q8_1_MMVQ, vec_dot_q2_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7281:13: note: in instantiation of function template specialization 'mul_mat_vec_q<7, 256, 16, block_q2_K, 1, &vec_dot_q2_K_q8_1>' requested here
            mul_mat_vec_q<7, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8947:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 16, block_q2_K, 1, &vec_dot_q2_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_K, block_q2_K, VDR_Q2_K_Q8_1_MMVQ, vec_dot_q2_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7285:13: note: in instantiation of function template specialization 'mul_mat_vec_q<8, 256, 16, block_q2_K, 1, &vec_dot_q2_K_q8_1>' requested here
            mul_mat_vec_q<8, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8947:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 16, block_q2_K, 1, &vec_dot_q2_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_K, block_q2_K, VDR_Q2_K_Q8_1_MMVQ, vec_dot_q2_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7257:13: note: in instantiation of function template specialization 'mul_mat_vec_q<1, 256, 16, block_q3_K, 1, &vec_dot_q3_K_q8_1>' requested here
            mul_mat_vec_q<1, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8951:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 16, block_q3_K, 1, &vec_dot_q3_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_K, block_q3_K, VDR_Q3_K_Q8_1_MMVQ, vec_dot_q3_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7261:13: note: in instantiation of function template specialization 'mul_mat_vec_q<2, 256, 16, block_q3_K, 1, &vec_dot_q3_K_q8_1>' requested here
            mul_mat_vec_q<2, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8951:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 16, block_q3_K, 1, &vec_dot_q3_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_K, block_q3_K, VDR_Q3_K_Q8_1_MMVQ, vec_dot_q3_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7265:13: note: in instantiation of function template specialization 'mul_mat_vec_q<3, 256, 16, block_q3_K, 1, &vec_dot_q3_K_q8_1>' requested here
            mul_mat_vec_q<3, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8951:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 16, block_q3_K, 1, &vec_dot_q3_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_K, block_q3_K, VDR_Q3_K_Q8_1_MMVQ, vec_dot_q3_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7269:13: note: in instantiation of function template specialization 'mul_mat_vec_q<4, 256, 16, block_q3_K, 1, &vec_dot_q3_K_q8_1>' requested here
            mul_mat_vec_q<4, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8951:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 16, block_q3_K, 1, &vec_dot_q3_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_K, block_q3_K, VDR_Q3_K_Q8_1_MMVQ, vec_dot_q3_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7273:13: note: in instantiation of function template specialization 'mul_mat_vec_q<5, 256, 16, block_q3_K, 1, &vec_dot_q3_K_q8_1>' requested here
            mul_mat_vec_q<5, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8951:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 16, block_q3_K, 1, &vec_dot_q3_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_K, block_q3_K, VDR_Q3_K_Q8_1_MMVQ, vec_dot_q3_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7277:13: note: in instantiation of function template specialization 'mul_mat_vec_q<6, 256, 16, block_q3_K, 1, &vec_dot_q3_K_q8_1>' requested here
            mul_mat_vec_q<6, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8951:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 16, block_q3_K, 1, &vec_dot_q3_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_K, block_q3_K, VDR_Q3_K_Q8_1_MMVQ, vec_dot_q3_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7281:13: note: in instantiation of function template specialization 'mul_mat_vec_q<7, 256, 16, block_q3_K, 1, &vec_dot_q3_K_q8_1>' requested here
            mul_mat_vec_q<7, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8951:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 16, block_q3_K, 1, &vec_dot_q3_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_K, block_q3_K, VDR_Q3_K_Q8_1_MMVQ, vec_dot_q3_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7285:13: note: in instantiation of function template specialization 'mul_mat_vec_q<8, 256, 16, block_q3_K, 1, &vec_dot_q3_K_q8_1>' requested here
            mul_mat_vec_q<8, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8951:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 16, block_q3_K, 1, &vec_dot_q3_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_K, block_q3_K, VDR_Q3_K_Q8_1_MMVQ, vec_dot_q3_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7257:13: note: in instantiation of function template specialization 'mul_mat_vec_q<1, 256, 32, block_q4_K, 2, &vec_dot_q4_K_q8_1>' requested here
            mul_mat_vec_q<1, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8955:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q4_K, 2, &vec_dot_q4_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI4_K, block_q4_K, VDR_Q4_K_Q8_1_MMVQ, vec_dot_q4_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7261:13: note: in instantiation of function template specialization 'mul_mat_vec_q<2, 256, 32, block_q4_K, 2, &vec_dot_q4_K_q8_1>' requested here
            mul_mat_vec_q<2, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8955:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q4_K, 2, &vec_dot_q4_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI4_K, block_q4_K, VDR_Q4_K_Q8_1_MMVQ, vec_dot_q4_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7265:13: note: in instantiation of function template specialization 'mul_mat_vec_q<3, 256, 32, block_q4_K, 2, &vec_dot_q4_K_q8_1>' requested here
            mul_mat_vec_q<3, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8955:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q4_K, 2, &vec_dot_q4_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI4_K, block_q4_K, VDR_Q4_K_Q8_1_MMVQ, vec_dot_q4_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7269:13: note: in instantiation of function template specialization 'mul_mat_vec_q<4, 256, 32, block_q4_K, 2, &vec_dot_q4_K_q8_1>' requested here
            mul_mat_vec_q<4, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8955:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q4_K, 2, &vec_dot_q4_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI4_K, block_q4_K, VDR_Q4_K_Q8_1_MMVQ, vec_dot_q4_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7273:13: note: in instantiation of function template specialization 'mul_mat_vec_q<5, 256, 32, block_q4_K, 2, &vec_dot_q4_K_q8_1>' requested here
            mul_mat_vec_q<5, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8955:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q4_K, 2, &vec_dot_q4_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI4_K, block_q4_K, VDR_Q4_K_Q8_1_MMVQ, vec_dot_q4_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7277:13: note: in instantiation of function template specialization 'mul_mat_vec_q<6, 256, 32, block_q4_K, 2, &vec_dot_q4_K_q8_1>' requested here
            mul_mat_vec_q<6, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8955:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q4_K, 2, &vec_dot_q4_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI4_K, block_q4_K, VDR_Q4_K_Q8_1_MMVQ, vec_dot_q4_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7281:13: note: in instantiation of function template specialization 'mul_mat_vec_q<7, 256, 32, block_q4_K, 2, &vec_dot_q4_K_q8_1>' requested here
            mul_mat_vec_q<7, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8955:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q4_K, 2, &vec_dot_q4_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI4_K, block_q4_K, VDR_Q4_K_Q8_1_MMVQ, vec_dot_q4_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7285:13: note: in instantiation of function template specialization 'mul_mat_vec_q<8, 256, 32, block_q4_K, 2, &vec_dot_q4_K_q8_1>' requested here
            mul_mat_vec_q<8, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8955:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q4_K, 2, &vec_dot_q4_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI4_K, block_q4_K, VDR_Q4_K_Q8_1_MMVQ, vec_dot_q4_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7257:13: note: in instantiation of function template specialization 'mul_mat_vec_q<1, 256, 32, block_q5_K, 2, &vec_dot_q5_K_q8_1>' requested here
            mul_mat_vec_q<1, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8959:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q5_K, 2, &vec_dot_q5_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI5_K, block_q5_K, VDR_Q5_K_Q8_1_MMVQ, vec_dot_q5_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7261:13: note: in instantiation of function template specialization 'mul_mat_vec_q<2, 256, 32, block_q5_K, 2, &vec_dot_q5_K_q8_1>' requested here
            mul_mat_vec_q<2, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8959:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q5_K, 2, &vec_dot_q5_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI5_K, block_q5_K, VDR_Q5_K_Q8_1_MMVQ, vec_dot_q5_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7265:13: note: in instantiation of function template specialization 'mul_mat_vec_q<3, 256, 32, block_q5_K, 2, &vec_dot_q5_K_q8_1>' requested here
            mul_mat_vec_q<3, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8959:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q5_K, 2, &vec_dot_q5_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI5_K, block_q5_K, VDR_Q5_K_Q8_1_MMVQ, vec_dot_q5_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7269:13: note: in instantiation of function template specialization 'mul_mat_vec_q<4, 256, 32, block_q5_K, 2, &vec_dot_q5_K_q8_1>' requested here
            mul_mat_vec_q<4, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8959:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q5_K, 2, &vec_dot_q5_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI5_K, block_q5_K, VDR_Q5_K_Q8_1_MMVQ, vec_dot_q5_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7273:13: note: in instantiation of function template specialization 'mul_mat_vec_q<5, 256, 32, block_q5_K, 2, &vec_dot_q5_K_q8_1>' requested here
            mul_mat_vec_q<5, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8959:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q5_K, 2, &vec_dot_q5_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI5_K, block_q5_K, VDR_Q5_K_Q8_1_MMVQ, vec_dot_q5_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7277:13: note: in instantiation of function template specialization 'mul_mat_vec_q<6, 256, 32, block_q5_K, 2, &vec_dot_q5_K_q8_1>' requested here
            mul_mat_vec_q<6, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8959:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q5_K, 2, &vec_dot_q5_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI5_K, block_q5_K, VDR_Q5_K_Q8_1_MMVQ, vec_dot_q5_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7281:13: note: in instantiation of function template specialization 'mul_mat_vec_q<7, 256, 32, block_q5_K, 2, &vec_dot_q5_K_q8_1>' requested here
            mul_mat_vec_q<7, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8959:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q5_K, 2, &vec_dot_q5_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI5_K, block_q5_K, VDR_Q5_K_Q8_1_MMVQ, vec_dot_q5_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7285:13: note: in instantiation of function template specialization 'mul_mat_vec_q<8, 256, 32, block_q5_K, 2, &vec_dot_q5_K_q8_1>' requested here
            mul_mat_vec_q<8, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8959:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q5_K, 2, &vec_dot_q5_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI5_K, block_q5_K, VDR_Q5_K_Q8_1_MMVQ, vec_dot_q5_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7257:13: note: in instantiation of function template specialization 'mul_mat_vec_q<1, 256, 32, block_q6_K, 1, &vec_dot_q6_K_q8_1>' requested here
            mul_mat_vec_q<1, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8963:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q6_K, 1, &vec_dot_q6_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI6_K, block_q6_K, VDR_Q6_K_Q8_1_MMVQ, vec_dot_q6_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7261:13: note: in instantiation of function template specialization 'mul_mat_vec_q<2, 256, 32, block_q6_K, 1, &vec_dot_q6_K_q8_1>' requested here
            mul_mat_vec_q<2, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8963:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q6_K, 1, &vec_dot_q6_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI6_K, block_q6_K, VDR_Q6_K_Q8_1_MMVQ, vec_dot_q6_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7265:13: note: in instantiation of function template specialization 'mul_mat_vec_q<3, 256, 32, block_q6_K, 1, &vec_dot_q6_K_q8_1>' requested here
            mul_mat_vec_q<3, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8963:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q6_K, 1, &vec_dot_q6_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI6_K, block_q6_K, VDR_Q6_K_Q8_1_MMVQ, vec_dot_q6_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7269:13: note: in instantiation of function template specialization 'mul_mat_vec_q<4, 256, 32, block_q6_K, 1, &vec_dot_q6_K_q8_1>' requested here
            mul_mat_vec_q<4, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8963:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q6_K, 1, &vec_dot_q6_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI6_K, block_q6_K, VDR_Q6_K_Q8_1_MMVQ, vec_dot_q6_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7273:13: note: in instantiation of function template specialization 'mul_mat_vec_q<5, 256, 32, block_q6_K, 1, &vec_dot_q6_K_q8_1>' requested here
            mul_mat_vec_q<5, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8963:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q6_K, 1, &vec_dot_q6_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI6_K, block_q6_K, VDR_Q6_K_Q8_1_MMVQ, vec_dot_q6_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7277:13: note: in instantiation of function template specialization 'mul_mat_vec_q<6, 256, 32, block_q6_K, 1, &vec_dot_q6_K_q8_1>' requested here
            mul_mat_vec_q<6, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8963:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q6_K, 1, &vec_dot_q6_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI6_K, block_q6_K, VDR_Q6_K_Q8_1_MMVQ, vec_dot_q6_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7281:13: note: in instantiation of function template specialization 'mul_mat_vec_q<7, 256, 32, block_q6_K, 1, &vec_dot_q6_K_q8_1>' requested here
            mul_mat_vec_q<7, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8963:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q6_K, 1, &vec_dot_q6_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI6_K, block_q6_K, VDR_Q6_K_Q8_1_MMVQ, vec_dot_q6_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7285:13: note: in instantiation of function template specialization 'mul_mat_vec_q<8, 256, 32, block_q6_K, 1, &vec_dot_q6_K_q8_1>' requested here
            mul_mat_vec_q<8, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8963:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 32, block_q6_K, 1, &vec_dot_q6_K_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI6_K, block_q6_K, VDR_Q6_K_Q8_1_MMVQ, vec_dot_q6_K_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7257:13: note: in instantiation of function template specialization 'mul_mat_vec_q<1, 256, 8, block_iq2_xxs, 1, &vec_dot_iq2_xxs_q8_1>' requested here
            mul_mat_vec_q<1, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8967:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq2_xxs, 1, &vec_dot_iq2_xxs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_XXS, block_iq2_xxs, 1, vec_dot_iq2_xxs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7261:13: note: in instantiation of function template specialization 'mul_mat_vec_q<2, 256, 8, block_iq2_xxs, 1, &vec_dot_iq2_xxs_q8_1>' requested here
            mul_mat_vec_q<2, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8967:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq2_xxs, 1, &vec_dot_iq2_xxs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_XXS, block_iq2_xxs, 1, vec_dot_iq2_xxs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7265:13: note: in instantiation of function template specialization 'mul_mat_vec_q<3, 256, 8, block_iq2_xxs, 1, &vec_dot_iq2_xxs_q8_1>' requested here
            mul_mat_vec_q<3, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8967:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq2_xxs, 1, &vec_dot_iq2_xxs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_XXS, block_iq2_xxs, 1, vec_dot_iq2_xxs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7269:13: note: in instantiation of function template specialization 'mul_mat_vec_q<4, 256, 8, block_iq2_xxs, 1, &vec_dot_iq2_xxs_q8_1>' requested here
            mul_mat_vec_q<4, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8967:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq2_xxs, 1, &vec_dot_iq2_xxs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_XXS, block_iq2_xxs, 1, vec_dot_iq2_xxs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7273:13: note: in instantiation of function template specialization 'mul_mat_vec_q<5, 256, 8, block_iq2_xxs, 1, &vec_dot_iq2_xxs_q8_1>' requested here
            mul_mat_vec_q<5, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8967:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq2_xxs, 1, &vec_dot_iq2_xxs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_XXS, block_iq2_xxs, 1, vec_dot_iq2_xxs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7277:13: note: in instantiation of function template specialization 'mul_mat_vec_q<6, 256, 8, block_iq2_xxs, 1, &vec_dot_iq2_xxs_q8_1>' requested here
            mul_mat_vec_q<6, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8967:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq2_xxs, 1, &vec_dot_iq2_xxs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_XXS, block_iq2_xxs, 1, vec_dot_iq2_xxs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7281:13: note: in instantiation of function template specialization 'mul_mat_vec_q<7, 256, 8, block_iq2_xxs, 1, &vec_dot_iq2_xxs_q8_1>' requested here
            mul_mat_vec_q<7, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8967:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq2_xxs, 1, &vec_dot_iq2_xxs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_XXS, block_iq2_xxs, 1, vec_dot_iq2_xxs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7285:13: note: in instantiation of function template specialization 'mul_mat_vec_q<8, 256, 8, block_iq2_xxs, 1, &vec_dot_iq2_xxs_q8_1>' requested here
            mul_mat_vec_q<8, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8967:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq2_xxs, 1, &vec_dot_iq2_xxs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_XXS, block_iq2_xxs, 1, vec_dot_iq2_xxs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7257:13: note: in instantiation of function template specialization 'mul_mat_vec_q<1, 256, 8, block_iq2_xs, 1, &vec_dot_iq2_xs_q8_1>' requested here
            mul_mat_vec_q<1, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8971:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq2_xs, 1, &vec_dot_iq2_xs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_XS, block_iq2_xs, 1, vec_dot_iq2_xs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7261:13: note: in instantiation of function template specialization 'mul_mat_vec_q<2, 256, 8, block_iq2_xs, 1, &vec_dot_iq2_xs_q8_1>' requested here
            mul_mat_vec_q<2, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8971:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq2_xs, 1, &vec_dot_iq2_xs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_XS, block_iq2_xs, 1, vec_dot_iq2_xs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7265:13: note: in instantiation of function template specialization 'mul_mat_vec_q<3, 256, 8, block_iq2_xs, 1, &vec_dot_iq2_xs_q8_1>' requested here
            mul_mat_vec_q<3, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8971:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq2_xs, 1, &vec_dot_iq2_xs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_XS, block_iq2_xs, 1, vec_dot_iq2_xs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7269:13: note: in instantiation of function template specialization 'mul_mat_vec_q<4, 256, 8, block_iq2_xs, 1, &vec_dot_iq2_xs_q8_1>' requested here
            mul_mat_vec_q<4, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8971:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq2_xs, 1, &vec_dot_iq2_xs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_XS, block_iq2_xs, 1, vec_dot_iq2_xs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7273:13: note: in instantiation of function template specialization 'mul_mat_vec_q<5, 256, 8, block_iq2_xs, 1, &vec_dot_iq2_xs_q8_1>' requested here
            mul_mat_vec_q<5, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8971:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq2_xs, 1, &vec_dot_iq2_xs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_XS, block_iq2_xs, 1, vec_dot_iq2_xs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7277:13: note: in instantiation of function template specialization 'mul_mat_vec_q<6, 256, 8, block_iq2_xs, 1, &vec_dot_iq2_xs_q8_1>' requested here
            mul_mat_vec_q<6, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8971:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq2_xs, 1, &vec_dot_iq2_xs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_XS, block_iq2_xs, 1, vec_dot_iq2_xs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7281:13: note: in instantiation of function template specialization 'mul_mat_vec_q<7, 256, 8, block_iq2_xs, 1, &vec_dot_iq2_xs_q8_1>' requested here
            mul_mat_vec_q<7, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8971:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq2_xs, 1, &vec_dot_iq2_xs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_XS, block_iq2_xs, 1, vec_dot_iq2_xs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7285:13: note: in instantiation of function template specialization 'mul_mat_vec_q<8, 256, 8, block_iq2_xs, 1, &vec_dot_iq2_xs_q8_1>' requested here
            mul_mat_vec_q<8, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8971:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq2_xs, 1, &vec_dot_iq2_xs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI2_XS, block_iq2_xs, 1, vec_dot_iq2_xs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7257:13: note: in instantiation of function template specialization 'mul_mat_vec_q<1, 256, 8, block_iq3_xxs, 1, &vec_dot_iq3_xxs_q8_1>' requested here
            mul_mat_vec_q<1, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8975:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq3_xxs, 1, &vec_dot_iq3_xxs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_XXS, block_iq3_xxs, 1, vec_dot_iq3_xxs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7261:13: note: in instantiation of function template specialization 'mul_mat_vec_q<2, 256, 8, block_iq3_xxs, 1, &vec_dot_iq3_xxs_q8_1>' requested here
            mul_mat_vec_q<2, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8975:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq3_xxs, 1, &vec_dot_iq3_xxs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_XXS, block_iq3_xxs, 1, vec_dot_iq3_xxs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7265:13: note: in instantiation of function template specialization 'mul_mat_vec_q<3, 256, 8, block_iq3_xxs, 1, &vec_dot_iq3_xxs_q8_1>' requested here
            mul_mat_vec_q<3, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8975:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq3_xxs, 1, &vec_dot_iq3_xxs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_XXS, block_iq3_xxs, 1, vec_dot_iq3_xxs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7269:13: note: in instantiation of function template specialization 'mul_mat_vec_q<4, 256, 8, block_iq3_xxs, 1, &vec_dot_iq3_xxs_q8_1>' requested here
            mul_mat_vec_q<4, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8975:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq3_xxs, 1, &vec_dot_iq3_xxs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_XXS, block_iq3_xxs, 1, vec_dot_iq3_xxs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7273:13: note: in instantiation of function template specialization 'mul_mat_vec_q<5, 256, 8, block_iq3_xxs, 1, &vec_dot_iq3_xxs_q8_1>' requested here
            mul_mat_vec_q<5, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8975:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq3_xxs, 1, &vec_dot_iq3_xxs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_XXS, block_iq3_xxs, 1, vec_dot_iq3_xxs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7277:13: note: in instantiation of function template specialization 'mul_mat_vec_q<6, 256, 8, block_iq3_xxs, 1, &vec_dot_iq3_xxs_q8_1>' requested here
            mul_mat_vec_q<6, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8975:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq3_xxs, 1, &vec_dot_iq3_xxs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_XXS, block_iq3_xxs, 1, vec_dot_iq3_xxs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7281:13: note: in instantiation of function template specialization 'mul_mat_vec_q<7, 256, 8, block_iq3_xxs, 1, &vec_dot_iq3_xxs_q8_1>' requested here
            mul_mat_vec_q<7, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8975:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq3_xxs, 1, &vec_dot_iq3_xxs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_XXS, block_iq3_xxs, 1, vec_dot_iq3_xxs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7285:13: note: in instantiation of function template specialization 'mul_mat_vec_q<8, 256, 8, block_iq3_xxs, 1, &vec_dot_iq3_xxs_q8_1>' requested here
            mul_mat_vec_q<8, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8975:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq3_xxs, 1, &vec_dot_iq3_xxs_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_XXS, block_iq3_xxs, 1, vec_dot_iq3_xxs_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7257:13: note: in instantiation of function template specialization 'mul_mat_vec_q<1, 256, 8, block_iq1_s, 1, &vec_dot_iq1_s_q8_1>' requested here
            mul_mat_vec_q<1, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8979:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq1_s, 1, &vec_dot_iq1_s_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI1_S, block_iq1_s, 1, vec_dot_iq1_s_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7261:13: note: in instantiation of function template specialization 'mul_mat_vec_q<2, 256, 8, block_iq1_s, 1, &vec_dot_iq1_s_q8_1>' requested here
            mul_mat_vec_q<2, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8979:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq1_s, 1, &vec_dot_iq1_s_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI1_S, block_iq1_s, 1, vec_dot_iq1_s_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7265:13: note: in instantiation of function template specialization 'mul_mat_vec_q<3, 256, 8, block_iq1_s, 1, &vec_dot_iq1_s_q8_1>' requested here
            mul_mat_vec_q<3, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8979:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq1_s, 1, &vec_dot_iq1_s_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI1_S, block_iq1_s, 1, vec_dot_iq1_s_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7269:13: note: in instantiation of function template specialization 'mul_mat_vec_q<4, 256, 8, block_iq1_s, 1, &vec_dot_iq1_s_q8_1>' requested here
            mul_mat_vec_q<4, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8979:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq1_s, 1, &vec_dot_iq1_s_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI1_S, block_iq1_s, 1, vec_dot_iq1_s_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7273:13: note: in instantiation of function template specialization 'mul_mat_vec_q<5, 256, 8, block_iq1_s, 1, &vec_dot_iq1_s_q8_1>' requested here
            mul_mat_vec_q<5, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8979:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq1_s, 1, &vec_dot_iq1_s_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI1_S, block_iq1_s, 1, vec_dot_iq1_s_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7277:13: note: in instantiation of function template specialization 'mul_mat_vec_q<6, 256, 8, block_iq1_s, 1, &vec_dot_iq1_s_q8_1>' requested here
            mul_mat_vec_q<6, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8979:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq1_s, 1, &vec_dot_iq1_s_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI1_S, block_iq1_s, 1, vec_dot_iq1_s_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7281:13: note: in instantiation of function template specialization 'mul_mat_vec_q<7, 256, 8, block_iq1_s, 1, &vec_dot_iq1_s_q8_1>' requested here
            mul_mat_vec_q<7, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8979:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq1_s, 1, &vec_dot_iq1_s_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI1_S, block_iq1_s, 1, vec_dot_iq1_s_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7285:13: note: in instantiation of function template specialization 'mul_mat_vec_q<8, 256, 8, block_iq1_s, 1, &vec_dot_iq1_s_q8_1>' requested here
            mul_mat_vec_q<8, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8979:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq1_s, 1, &vec_dot_iq1_s_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI1_S, block_iq1_s, 1, vec_dot_iq1_s_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7257:13: note: in instantiation of function template specialization 'mul_mat_vec_q<1, 32, 4, block_iq4_nl, 2, &vec_dot_iq4_nl_q8_1>' requested here
            mul_mat_vec_q<1, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8983:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_iq4_nl, 2, &vec_dot_iq4_nl_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_NL, QI4_NL, block_iq4_nl, VDR_Q4_0_Q8_1_MMVQ, vec_dot_iq4_nl_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7261:13: note: in instantiation of function template specialization 'mul_mat_vec_q<2, 32, 4, block_iq4_nl, 2, &vec_dot_iq4_nl_q8_1>' requested here
            mul_mat_vec_q<2, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8983:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_iq4_nl, 2, &vec_dot_iq4_nl_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_NL, QI4_NL, block_iq4_nl, VDR_Q4_0_Q8_1_MMVQ, vec_dot_iq4_nl_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7265:13: note: in instantiation of function template specialization 'mul_mat_vec_q<3, 32, 4, block_iq4_nl, 2, &vec_dot_iq4_nl_q8_1>' requested here
            mul_mat_vec_q<3, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8983:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_iq4_nl, 2, &vec_dot_iq4_nl_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_NL, QI4_NL, block_iq4_nl, VDR_Q4_0_Q8_1_MMVQ, vec_dot_iq4_nl_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7269:13: note: in instantiation of function template specialization 'mul_mat_vec_q<4, 32, 4, block_iq4_nl, 2, &vec_dot_iq4_nl_q8_1>' requested here
            mul_mat_vec_q<4, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8983:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_iq4_nl, 2, &vec_dot_iq4_nl_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_NL, QI4_NL, block_iq4_nl, VDR_Q4_0_Q8_1_MMVQ, vec_dot_iq4_nl_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7273:13: note: in instantiation of function template specialization 'mul_mat_vec_q<5, 32, 4, block_iq4_nl, 2, &vec_dot_iq4_nl_q8_1>' requested here
            mul_mat_vec_q<5, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8983:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_iq4_nl, 2, &vec_dot_iq4_nl_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_NL, QI4_NL, block_iq4_nl, VDR_Q4_0_Q8_1_MMVQ, vec_dot_iq4_nl_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7277:13: note: in instantiation of function template specialization 'mul_mat_vec_q<6, 32, 4, block_iq4_nl, 2, &vec_dot_iq4_nl_q8_1>' requested here
            mul_mat_vec_q<6, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8983:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_iq4_nl, 2, &vec_dot_iq4_nl_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_NL, QI4_NL, block_iq4_nl, VDR_Q4_0_Q8_1_MMVQ, vec_dot_iq4_nl_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7281:13: note: in instantiation of function template specialization 'mul_mat_vec_q<7, 32, 4, block_iq4_nl, 2, &vec_dot_iq4_nl_q8_1>' requested here
            mul_mat_vec_q<7, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8983:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_iq4_nl, 2, &vec_dot_iq4_nl_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_NL, QI4_NL, block_iq4_nl, VDR_Q4_0_Q8_1_MMVQ, vec_dot_iq4_nl_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7285:13: note: in instantiation of function template specialization 'mul_mat_vec_q<8, 32, 4, block_iq4_nl, 2, &vec_dot_iq4_nl_q8_1>' requested here
            mul_mat_vec_q<8, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8983:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<32, 4, block_iq4_nl, 2, &vec_dot_iq4_nl_q8_1>' requested here
            mul_mat_vec_q_cuda<QK4_NL, QI4_NL, block_iq4_nl, VDR_Q4_0_Q8_1_MMVQ, vec_dot_iq4_nl_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7257:13: note: in instantiation of function template specialization 'mul_mat_vec_q<1, 256, 8, block_iq3_s, 1, &vec_dot_iq3_s_q8_1>' requested here
            mul_mat_vec_q<1, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8987:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq3_s, 1, &vec_dot_iq3_s_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_XS, block_iq3_s, 1, vec_dot_iq3_s_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7261:13: note: in instantiation of function template specialization 'mul_mat_vec_q<2, 256, 8, block_iq3_s, 1, &vec_dot_iq3_s_q8_1>' requested here
            mul_mat_vec_q<2, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8987:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq3_s, 1, &vec_dot_iq3_s_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_XS, block_iq3_s, 1, vec_dot_iq3_s_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7265:13: note: in instantiation of function template specialization 'mul_mat_vec_q<3, 256, 8, block_iq3_s, 1, &vec_dot_iq3_s_q8_1>' requested here
            mul_mat_vec_q<3, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8987:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq3_s, 1, &vec_dot_iq3_s_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_XS, block_iq3_s, 1, vec_dot_iq3_s_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7269:13: note: in instantiation of function template specialization 'mul_mat_vec_q<4, 256, 8, block_iq3_s, 1, &vec_dot_iq3_s_q8_1>' requested here
            mul_mat_vec_q<4, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8987:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq3_s, 1, &vec_dot_iq3_s_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_XS, block_iq3_s, 1, vec_dot_iq3_s_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7273:13: note: in instantiation of function template specialization 'mul_mat_vec_q<5, 256, 8, block_iq3_s, 1, &vec_dot_iq3_s_q8_1>' requested here
            mul_mat_vec_q<5, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8987:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq3_s, 1, &vec_dot_iq3_s_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_XS, block_iq3_s, 1, vec_dot_iq3_s_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7277:13: note: in instantiation of function template specialization 'mul_mat_vec_q<6, 256, 8, block_iq3_s, 1, &vec_dot_iq3_s_q8_1>' requested here
            mul_mat_vec_q<6, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8987:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq3_s, 1, &vec_dot_iq3_s_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_XS, block_iq3_s, 1, vec_dot_iq3_s_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7281:13: note: in instantiation of function template specialization 'mul_mat_vec_q<7, 256, 8, block_iq3_s, 1, &vec_dot_iq3_s_q8_1>' requested here
            mul_mat_vec_q<7, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8987:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq3_s, 1, &vec_dot_iq3_s_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_XS, block_iq3_s, 1, vec_dot_iq3_s_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:5778:48: warning: suggest braces around initialization of subobject [-Wmissing-braces]
    float tmp[ncols_y][rows_per_cuda_block] = {0.0f};
                                               ^~~~
                                               {   }
W:/git/test/llama.cpp/ggml-cuda.cu:7285:13: note: in instantiation of function template specialization 'mul_mat_vec_q<8, 256, 8, block_iq3_s, 1, &vec_dot_iq3_s_q8_1>' requested here
            mul_mat_vec_q<8, qk, qi, block_q_t, vdr, vec_dot>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:8987:13: note: in instantiation of function template specialization 'mul_mat_vec_q_cuda<256, 8, block_iq3_s, 1, &vec_dot_iq3_s_q8_1>' requested here
            mul_mat_vec_q_cuda<QK_K, QI3_XS, block_iq3_s, 1, vec_dot_iq3_s_q8_1>
            ^
W:/git/test/llama.cpp/ggml-cuda.cu:6595:17: warning: enumeration value 'GGML_OP_POOL_COUNT' not handled in switch [-Wswitch]
        switch (op) {
                ^~
W:/git/test/llama.cpp/ggml-cuda.cu:6607:25: warning: enumeration value 'GGML_OP_POOL_COUNT' not handled in switch [-Wswitch]
                switch (op) {
                        ^~
W:/git/test/llama.cpp/ggml-cuda.cu:6595:17: warning: enumeration value 'GGML_OP_POOL_COUNT' not handled in switch [-Wswitch]
        switch (op) {
                ^~
W:/git/test/llama.cpp/ggml-cuda.cu:9317:5: note: in instantiation of function template specialization 'pool2d_nchw_kernel<float, float>' requested here
    pool2d_nchw_kernel<<<block_nums, CUDA_IM2COL_BLOCK_SIZE, 0, main_stream>>>(IH, IW, OH, OW, k1, k0, s1, s0, p1, p0, parallel_elements, src0_dd, dst_dd, op);
    ^
W:/git/test/llama.cpp/ggml-cuda.cu:6607:25: warning: enumeration value 'GGML_OP_POOL_COUNT' not handled in switch [-Wswitch]
                switch (op) {
                        ^~
170 warnings and 7 errors generated when compiling for gfx1100.
[5/54] Building C object CMakeFiles/ggml.dir/ggml.c.obj
ninja: build stopped: subcommand failed.