root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-25 10:54:36 +00:00
Code Issues Actions 9 Packages Projects Releases Wiki Activity

metal : log recommendedMaxWorkingSetSize on iOS 16+ (#4936)

Browse Source 
* metal: Log `recommendedMaxWorkingSetSize` on iOS 16+

* Only log on iOS and macOS, ignoring tvOS and other platforms

* Check for Xcode version before using recommendedMaxWorkingSetSize

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
Alex Azarov 2024-01-16 14:33:02 +01:00 committed by GitHub
parent 122ed4840c
commit 7c8d3abd1a
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 27 additions and 31 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

58
ggml-metal.m
View File

@ -369,8 +369,12 @@ static struct ggml_metal_context * ggml_metal_init(int n_cb) {
GGML_METAL_LOG_INFO("%s: simdgroup reduction support = %s\n", __func__, ctx->support_simdgroup_reduction ? "true" : "false"); GGML_METAL_LOG_INFO("%s: simdgroup reduction support = %s\n", __func__, ctx->support_simdgroup_reduction ? "true" : "false");
GGML_METAL_LOG_INFO("%s: simdgroup matrix mul. support = %s\n", __func__, ctx->support_simdgroup_mm ? "true" : "false"); GGML_METAL_LOG_INFO("%s: simdgroup matrix mul. support = %s\n", __func__, ctx->support_simdgroup_mm ? "true" : "false");
GGML_METAL_LOG_INFO("%s: hasUnifiedMemory = %s\n", __func__, ctx->device.hasUnifiedMemory ? "true" : "false"); GGML_METAL_LOG_INFO("%s: hasUnifiedMemory = %s\n", __func__, ctx->device.hasUnifiedMemory ? "true" : "false");
#if TARGET_OS_OSX
GGML_METAL_LOG_INFO("%s: recommendedMaxWorkingSetSize = %8.2f MB\n", __func__, ctx->device.recommendedMaxWorkingSetSize / 1e6); #if TARGET_OS_OSX || (TARGET_OS_IOS && __clang_major__ >= 15)
if (@available(macOS 10.12, iOS 16.0, *)) {
GGML_METAL_LOG_INFO("%s: recommendedMaxWorkingSetSize = %8.2f MB\n", __func__, ctx->device.recommendedMaxWorkingSetSize / 1e6);
}
#elif TARGET_OS_OSX
if (ctx->device.maxTransferRate != 0) { if (ctx->device.maxTransferRate != 0) {
GGML_METAL_LOG_INFO("%s: maxTransferRate = %8.2f MB/s\n", __func__, ctx->device.maxTransferRate / 1e6); GGML_METAL_LOG_INFO("%s: maxTransferRate = %8.2f MB/s\n", __func__, ctx->device.maxTransferRate / 1e6);
} else { } else {
@ -2369,6 +2373,25 @@ GGML_CALL static const char * ggml_backend_metal_buffer_type_get_name(ggml_backe
UNUSED(buft); UNUSED(buft);
} }
static void ggml_backend_metal_log_allocated_size(id<MTLDevice> device) {
#if TARGET_OS_OSX || (TARGET_OS_IOS && __clang_major__ >= 15)
if (@available(macOS 10.12, iOS 16.0, *)) {
GGML_METAL_LOG_INFO(", (%8.2f / %8.2f)",
device.currentAllocatedSize / 1024.0 / 1024.0,
device.recommendedMaxWorkingSetSize / 1024.0 / 1024.0);
if (device.currentAllocatedSize > device.recommendedMaxWorkingSetSize) {
GGML_METAL_LOG_WARN("%s: warning: current allocated size is greater than the recommended max working set size\n", __func__);
} else {
GGML_METAL_LOG_INFO("\n");
}
} else {
GGML_METAL_LOG_INFO(", (%8.2f)\n", device.currentAllocatedSize / 1024.0 / 1024.0);
}
#endif
UNUSED(device);
}
GGML_CALL static ggml_backend_buffer_t ggml_backend_metal_buffer_type_alloc_buffer(ggml_backend_buffer_type_t buft, size_t size) { GGML_CALL static ggml_backend_buffer_t ggml_backend_metal_buffer_type_alloc_buffer(ggml_backend_buffer_type_t buft, size_t size) {
struct ggml_backend_metal_buffer_context * ctx = malloc(sizeof(struct ggml_backend_metal_buffer_context)); struct ggml_backend_metal_buffer_context * ctx = malloc(sizeof(struct ggml_backend_metal_buffer_context));
@ -2401,22 +2424,7 @@ GGML_CALL static ggml_backend_buffer_t ggml_backend_metal_buffer_type_alloc_buff
} }
GGML_METAL_LOG_INFO("%s: allocated buffer, size = %8.2f MiB", __func__, size_aligned / 1024.0 / 1024.0); GGML_METAL_LOG_INFO("%s: allocated buffer, size = %8.2f MiB", __func__, size_aligned / 1024.0 / 1024.0);
ggml_backend_metal_log_allocated_size(device);
#if TARGET_OS_OSX
GGML_METAL_LOG_INFO(", (%8.2f / %8.2f)",
device.currentAllocatedSize / 1024.0 / 1024.0,
device.recommendedMaxWorkingSetSize / 1024.0 / 1024.0);
if (device.currentAllocatedSize > device.recommendedMaxWorkingSetSize) {
GGML_METAL_LOG_WARN("%s: warning: current allocated size is greater than the recommended max working set size\n", __func__);
} else {
GGML_METAL_LOG_INFO("\n");
}
#else
GGML_METAL_LOG_INFO(", (%8.2f)\n", device.currentAllocatedSize / 1024.0 / 1024.0);
#endif
return ggml_backend_buffer_init(buft, ggml_backend_metal_buffer_i, ctx, size); return ggml_backend_buffer_init(buft, ggml_backend_metal_buffer_i, ctx, size);
} }
@ -2524,19 +2532,7 @@ GGML_CALL ggml_backend_buffer_t ggml_backend_metal_buffer_from_ptr(void * data,
} }
} }
#if TARGET_OS_OSX ggml_backend_metal_log_allocated_size(device);
GGML_METAL_LOG_INFO(", (%8.2f / %8.2f)",
device.currentAllocatedSize / 1024.0 / 1024.0,
device.recommendedMaxWorkingSetSize / 1024.0 / 1024.0);
if (device.currentAllocatedSize > device.recommendedMaxWorkingSetSize) {
GGML_METAL_LOG_WARN("%s: warning: current allocated size is greater than the recommended max working set size\n", __func__);
} else {
GGML_METAL_LOG_INFO("\n");
}
#else
GGML_METAL_LOG_INFO(", (%8.2f)\n", device.currentAllocatedSize / 1024.0 / 1024.0);
#endif
return ggml_backend_buffer_init(ggml_backend_metal_buffer_type(), ggml_backend_metal_buffer_i, ctx, size); return ggml_backend_buffer_init(ggml_backend_metal_buffer_type(), ggml_backend_metal_buffer_i, ctx, size);
} }