ggml update to 0.11.0， llama-cpp update to 9030 #51551
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,71 @@ | ||
| diff --git a/src/ggml-vulkan/ggml-vulkan.cpp b/src/ggml-vulkan/ggml-vulkan.cpp | ||
| index 423e01d..3cd5488 100644 | ||
| --- a/src/ggml-vulkan/ggml-vulkan.cpp | ||
| +++ b/src/ggml-vulkan/ggml-vulkan.cpp | ||
| @@ -2036,7 +2036,7 @@ void vk_memory_logger::log_allocation(vk_buffer_ref buf_ref, size_t size) { | ||
| allocations[buf->buffer] = size; | ||
| total_device += device ? size : 0; | ||
| total_host += device ? 0 : size; | ||
| - VK_LOG_MEMORY(buf->device->name << ": +" << format_size(size) << " " << type << " at " << buf->buffer << ". Total device: " << format_size(total_device) << ", total host: " << format_size(total_host)); | ||
| + VK_LOG_MEMORY(buf->device->name << ": +" << format_size(size) << " " << type << " at " << static_cast<VkBuffer>(buf->buffer) << ". Total device: " << format_size(total_device) << ", total host: " << format_size(total_host)); | ||
| } | ||
|
|
||
| void vk_memory_logger::log_deallocation(vk_buffer_ref buf_ref) { | ||
| @@ -2052,10 +2052,10 @@ void vk_memory_logger::log_deallocation(vk_buffer_ref buf_ref) { | ||
| total_device -= device ? it->second : 0; | ||
| total_host -= device ? 0 : it->second; | ||
| if (it != allocations.end()) { | ||
| - VK_LOG_MEMORY(buf->device->name << ": -" << format_size(it->second) << " " << type << " at " << buf->buffer << ". Total device: " << format_size(total_device) << ", total host: " << format_size(total_host)); | ||
| + VK_LOG_MEMORY(buf->device->name << ": -" << format_size(it->second) << " " << type << " at " << static_cast<VkBuffer>(buf->buffer) << ". Total device: " << format_size(total_device) << ", total host: " << format_size(total_host)); | ||
| allocations.erase(it); | ||
| } else { | ||
| - VK_LOG_MEMORY("ERROR " << buf->device->name << ": Attempted to deallocate unknown " << type << " memory at " << buf->buffer); | ||
| + VK_LOG_MEMORY("ERROR " << buf->device->name << ": Attempted to deallocate unknown " << type << " memory at " << static_cast<VkBuffer>(buf->buffer)); | ||
| } | ||
| } | ||
|
|
||
| @@ -6951,7 +6951,7 @@ static bool ggml_vk_buffer_write_2d_async(vk_context subctx, vk_buffer& dst, siz | ||
| } | ||
|
|
||
| ggml_vk_sync_buffers(nullptr, subctx); | ||
| - subctx->s->buffer->buf.copyBuffer((VkBuffer)staging_buffer->buffer, (VkBuffer)dst->buffer, slices); | ||
| + subctx->s->buffer->buf.copyBuffer(staging_buffer->buffer, dst->buffer, slices); | ||
|
|
||
| if (width == spitch) { | ||
| deferred_memcpy((uint8_t *)staging_buffer->ptr, src, staging_size, &subctx->in_memcpys); | ||
| @@ -7086,7 +7086,7 @@ static bool ggml_vk_buffer_read_async(vk_context subctx, vk_buffer& src, size_t | ||
| } | ||
|
|
||
| static void ggml_vk_buffer_read_2d(vk_buffer& src, size_t offset, void * dst, size_t spitch, size_t dpitch, size_t width, size_t height) { | ||
| - VK_LOG_DEBUG("ggml_vk_buffer_read_2d(" << src->buffer << ", " << offset << ", " << width << ", " << height << ")"); | ||
| + VK_LOG_DEBUG("ggml_vk_buffer_read_2d(" << static_cast<VkBuffer>(src->buffer) << ", " << offset << ", " << width << ", " << height << ")"); | ||
|
|
||
| // If the device is not an UMA device the memory is host-accessible through rebar. While writing | ||
| // through PCIe is sufficient fast reading back data from PCIe is slower than going through | ||
| @@ -7118,7 +7118,7 @@ static void ggml_vk_buffer_read_2d(vk_buffer& src, size_t offset, void * dst, si | ||
| } | ||
|
|
||
| static void ggml_vk_buffer_read(vk_buffer& src, size_t offset, void * dst, size_t size) { | ||
| - VK_LOG_DEBUG("ggml_vk_buffer_read(" << src->buffer << ", " << offset << ", " << size << ")"); | ||
| + VK_LOG_DEBUG("ggml_vk_buffer_read(" << static_cast<VkBuffer>(src->buffer) << ", " << offset << ", " << size << ")"); | ||
| ggml_vk_buffer_read_2d(src, offset, dst, size, size, size, 1); | ||
| } | ||
|
|
||
| @@ -7286,7 +7286,7 @@ static void ggml_vk_matmul( | ||
| uint32_t batch_stride_a, uint32_t batch_stride_b, uint32_t batch_stride_d, | ||
| uint32_t split_k, uint32_t batch, uint32_t ne02, uint32_t ne12, uint32_t broadcast2, uint32_t broadcast3, | ||
| uint32_t padded_n) { | ||
| - VK_LOG_DEBUG("ggml_vk_matmul(a: (" << a.buffer->buffer << ", " << a.offset << ", " << a.size << "), b: (" << b.buffer->buffer << ", " << b.offset << ", " << b.size << "), d: (" << d.buffer->buffer << ", " << d.offset << ", " << d.size << "), split_k: (" << (split_k_buffer.buffer != nullptr ? split_k_buffer.buffer->buffer : VK_NULL_HANDLE) << ", " << split_k_buffer.offset << ", " << split_k_buffer.size << "), m: " << m << ", n: " << n << ", k: " << k << ", stride_a: " << stride_a << ", stride_b: " << stride_b << ", stride_d: " << stride_d << ", batch_stride_a: " << batch_stride_a << ", batch_stride_b: " << batch_stride_b << ", batch_stride_d: " << batch_stride_d << ", split_k: " << split_k << ", batch: " << batch << ", ne02: " << ne02 << ", ne12: " << ne12 << ", broadcast2: " << broadcast2 << ", broadcast3: " << broadcast3 << ", padded_n: " << padded_n << ")"); | ||
| + VK_LOG_DEBUG("ggml_vk_matmul(a: (" << static_cast<VkBuffer>(a.buffer->buffer) << ", " << a.offset << ", " << a.size << "), b: (" << static_cast<VkBuffer>(b.buffer->buffer) << ", " << b.offset << ", " << b.size << "), d: (" << static_cast<VkBuffer>(d.buffer->buffer) << ", " << d.offset << ", " << d.size << "), split_k: (" << (split_k_buffer.buffer != nullptr ? static_cast<VkBuffer>(split_k_buffer.buffer->buffer) : VkBuffer{}) << ", " << split_k_buffer.offset << ", " << split_k_buffer.size << "), m: " << m << ", n: " << n << ", k: " << k << ", stride_a: " << stride_a << ", stride_b: " << stride_b << ", stride_d: " << stride_d << ", batch_stride_a: " << batch_stride_a << ", batch_stride_b: " << batch_stride_b << ", batch_stride_d: " << batch_stride_d << ", split_k: " << split_k << ", batch: " << batch << ", ne02: " << ne02 << ", ne12: " << ne12 << ", broadcast2: " << broadcast2 << ", broadcast3: " << broadcast3 << ", padded_n: " << padded_n << ")"); | ||
| if (split_k == 1) { | ||
| ggml_pipeline_request_descriptor_sets(ctx, pipeline, CEIL_DIV(batch, ctx->device->properties.limits.maxComputeWorkGroupCount[2])); | ||
|
|
||
| @@ -7366,7 +7366,7 @@ static void ggml_vk_matmul_id( | ||
| uint32_t batch_stride_a, uint32_t batch_stride_b, uint32_t batch_stride_d, | ||
| uint32_t n_as, uint32_t nei0, uint32_t nei1, uint32_t nbi1, uint32_t ne11, | ||
| uint32_t padded_n) { | ||
| - VK_LOG_DEBUG("ggml_vk_matmul_id(a: (" << a.buffer->buffer << ", " << a.offset << ", " << a.size << "), b: (" << b.buffer->buffer << ", " << b.offset << ", " << b.size << "), d: (" << d.buffer->buffer << ", " << d.offset << ", " << d.size << "), ids: (" << ids.buffer->buffer << ", " << ids.offset << ", " << ids.size << "), expert_count: (" << expert_count_buf.buffer->buffer << ", " << expert_count_buf.offset << ", " << expert_count_buf.size << "), " << | ||
| + VK_LOG_DEBUG("ggml_vk_matmul_id(a: (" << static_cast<VkBuffer>(a.buffer->buffer) << ", " << a.offset << ", " << a.size << "), b: (" << static_cast<VkBuffer>(b.buffer->buffer) << ", " << b.offset << ", " << b.size << "), d: (" << static_cast<VkBuffer>(d.buffer->buffer) << ", " << d.offset << ", " << d.size << "), ids: (" << static_cast<VkBuffer>(ids.buffer->buffer) << ", " << ids.offset << ", " << ids.size << "), expert_count: (" << static_cast<VkBuffer>(expert_count_buf.buffer->buffer) << ", " << expert_count_buf.offset << ", " << expert_count_buf.size << "), " << | ||
| "m: " << m << ", n: " << n << ", k: " << k << ", stride_a: " << stride_a << ", stride_b: " << stride_b << ", stride_d: " << stride_d << ", " << | ||
| "batch_stride_a: " << batch_stride_a << ", batch_stride_b: " << batch_stride_b << ", batch_stride_d: " << batch_stride_d << ", " << | ||
| "n_as: " << n_as << ", nei0: " << nei0 << ", nei1: " << nei1 << ", nbi1: " << nbi1 << ", ne11: " << ne11 << ")"); |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,66 @@ | ||
| diff --git a/src/ggml-vulkan/ggml-vulkan.cpp b/src/ggml-vulkan/ggml-vulkan.cpp | ||
| index 423e01dbff1..0a7931002ab 100644 | ||
| --- a/src/ggml-vulkan/ggml-vulkan.cpp | ||
| +++ b/src/ggml-vulkan/ggml-vulkan.cpp | ||
| @@ -2149,11 +2149,11 @@ static void ggml_vk_create_pipeline_func(vk_device& device, vk_pipeline& pipelin | ||
|
|
||
| // Patch SPIR-V to enable RTE rounding for FP16, avoiding the need for | ||
| // separate shader variants compiled with -DRTE16. | ||
| - std::vector<uint32_t> spv; | ||
| + std::vector<uint32_t> spirv; | ||
| if (device->float_controls_rte_fp16) { | ||
| const uint32_t* spv_words = reinterpret_cast<const uint32_t *>(spv_data); | ||
| size_t word_count = spv_size / sizeof(uint32_t); | ||
| - spv.assign(spv_words, spv_words + word_count); | ||
| + spirv.assign(spv_words, spv_words + word_count); | ||
|
|
||
| // Find insertion points respecting SPIR-V layout order: | ||
| // Header(5) -> OpCapability -> OpExtension -> ... -> OpEntryPoint -> OpExecutionMode -> ... | ||
| @@ -2163,9 +2163,9 @@ static void ggml_vk_create_pipeline_func(vk_device& device, vk_pipeline& pipelin | ||
| size_t exec_insert_pos = pos; | ||
| uint32_t entry_point_id = 0; | ||
|
|
||
| - while (pos < spv.size()) { | ||
| - uint32_t opcode = spv[pos] & spv::OpCodeMask; | ||
| - uint32_t len = spv[pos] >> spv::WordCountShift; | ||
| + while (pos < spirv.size()) { | ||
| + uint32_t opcode = spirv[pos] & spv::OpCodeMask; | ||
| + uint32_t len = spirv[pos] >> spv::WordCountShift; | ||
| if (len == 0) break; | ||
|
|
||
| if (opcode == spv::OpCapability) { | ||
| @@ -2174,7 +2174,7 @@ static void ggml_vk_create_pipeline_func(vk_device& device, vk_pipeline& pipelin | ||
| } else if (opcode == spv::OpExtension) { | ||
| ext_insert_pos = pos + len; | ||
| } else if (opcode == spv::OpEntryPoint) { | ||
| - entry_point_id = spv[pos + 2]; | ||
| + entry_point_id = spirv[pos + 2]; | ||
| exec_insert_pos = pos + len; | ||
| } else if (opcode == spv::OpExecutionMode || opcode == spv::OpExecutionModeId) { | ||
| exec_insert_pos = pos + len; | ||
| @@ -2189,7 +2189,7 @@ static void ggml_vk_create_pipeline_func(vk_device& device, vk_pipeline& pipelin | ||
|
|
||
| // OpExecutionMode %entrypoint RoundingModeRTE 16 | ||
| uint32_t exec_mode[] = { (4u << spv::WordCountShift) | spv::OpExecutionMode, entry_point_id, spv::ExecutionModeRoundingModeRTE, 16 }; | ||
| - spv.insert(spv.begin() + exec_insert_pos, std::begin(exec_mode), std::end(exec_mode)); | ||
| + spirv.insert(spirv.begin() + exec_insert_pos, std::begin(exec_mode), std::end(exec_mode)); | ||
|
|
||
| // OpExtension "SPV_KHR_float_controls" | ||
| const char ext_str[] = "SPV_KHR_float_controls"; | ||
| @@ -2197,13 +2197,13 @@ static void ggml_vk_create_pipeline_func(vk_device& device, vk_pipeline& pipelin | ||
| std::vector<uint32_t> extension(1 + ext_str_words, 0); | ||
| extension[0] = (uint32_t)((1 + ext_str_words) << spv::WordCountShift) | spv::OpExtension; | ||
| memcpy(&extension[1], ext_str, sizeof(ext_str)); | ||
| - spv.insert(spv.begin() + ext_insert_pos, extension.begin(), extension.end()); | ||
| + spirv.insert(spirv.begin() + ext_insert_pos, extension.begin(), extension.end()); | ||
|
|
||
| // OpCapability RoundingModeRTE | ||
| uint32_t capability[] = { (2u << spv::WordCountShift) | spv::OpCapability, spv::CapabilityRoundingModeRTE }; | ||
| - spv.insert(spv.begin() + cap_insert_pos, std::begin(capability), std::end(capability)); | ||
| + spirv.insert(spirv.begin() + cap_insert_pos, std::begin(capability), std::end(capability)); | ||
|
|
||
| - shader_module_create_info = vk::ShaderModuleCreateInfo({}, spv.size() * sizeof(uint32_t), spv.data()); | ||
| + shader_module_create_info = vk::ShaderModuleCreateInfo({}, spirv.size() * sizeof(uint32_t), spirv.data()); | ||
| } | ||
|
|
||
| pipeline->shader_module = device->device.createShaderModule(shader_module_create_info); |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Uh oh!
There was an error while loading. Please reload this page.