From 8d5004b760c4829ddda1c07be1dd9e35a9c4adbd Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Rickard=20Hallerb=C3=A4ck?= Date: Wed, 13 Sep 2023 16:23:03 +0200 Subject: [PATCH] metal : setting function like logging macro to capital letters --- ggml-metal.m | 90 ++++++++++++++++++++++++++-------------------------- 1 file changed, 45 insertions(+), 45 deletions(-) diff --git a/ggml-metal.m b/ggml-metal.m index c6e3e6fd2..7ad59db33 100644 --- a/ggml-metal.m +++ b/ggml-metal.m @@ -12,13 +12,13 @@ #define MAX(a, b) ((a) > (b) ? (a) : (b)) #ifdef GGML_METAL_NDEBUG -#define ggml_metal_log_info(...) -#define ggml_metal_log_warn(...) -#define ggml_metal_log_error(...) +#define GGML_METAL_LOG_INFO(...) +#define GGML_METAL_LOG_WARN(...) +#define GGML_METAL_LOG_ERROR(...) #else -#define ggml_metal_log_info(...) ggml_metal_log(LLAMA_LOG_LEVEL_INFO, __VA_ARGS__) -#define ggml_metal_log_warn(...) ggml_metal_log(LLAMA_LOG_LEVEL_WARN, __VA_ARGS__) -#define ggml_metal_log_error(...) ggml_metal_log(LLAMA_LOG_LEVEL_ERROR, __VA_ARGS__) +#define GGML_METAL_LOG_INFO(...) ggml_metal_log(LLAMA_LOG_LEVEL_INFO, __VA_ARGS__) +#define GGML_METAL_LOG_WARN(...) ggml_metal_log(LLAMA_LOG_LEVEL_WARN, __VA_ARGS__) +#define GGML_METAL_LOG_ERROR(...) ggml_metal_log(LLAMA_LOG_LEVEL_ERROR, __VA_ARGS__) #endif #define UNUSED(x) (void)(x) @@ -150,7 +150,7 @@ static void ggml_metal_log(enum llama_log_level level, const char* format, ...){ struct ggml_metal_context * ggml_metal_init(int n_cb) { - ggml_metal_log_info("%s: allocating\n", __func__); + GGML_METAL_LOG_INFO("%s: allocating\n", __func__); id device; NSString * s; @@ -160,14 +160,14 @@ struct ggml_metal_context * ggml_metal_init(int n_cb) { NSArray * devices = MTLCopyAllDevices(); for (device in devices) { s = [device name]; - ggml_metal_log_info("%s: found device: %s\n", __func__, [s UTF8String]); + GGML_METAL_LOG_INFO("%s: found device: %s\n", __func__, [s UTF8String]); } #endif // Pick and show default Metal device device = MTLCreateSystemDefaultDevice(); s = [device name]; - ggml_metal_log_info("%s: picking default device: %s\n", __func__, [s UTF8String]); + GGML_METAL_LOG_INFO("%s: picking default device: %s\n", __func__, [s UTF8String]); // Configure context struct ggml_metal_context * ctx = malloc(sizeof(struct ggml_metal_context)); @@ -194,7 +194,7 @@ struct ggml_metal_context * ggml_metal_init(int n_cb) { ctx->library = [ctx->device newLibraryWithURL:libURL error:&error]; if (error) { - ggml_metal_log_error("%s: error: %s\n", __func__, [[error description] UTF8String]); + GGML_METAL_LOG_ERROR("%s: error: %s\n", __func__, [[error description] UTF8String]); return NULL; } } @@ -208,11 +208,11 @@ struct ggml_metal_context * ggml_metal_init(int n_cb) { //NSString * path = [[NSBundle mainBundle] pathForResource:@"../../examples/metal/metal" ofType:@"metal"]; NSBundle * bundle = [NSBundle bundleForClass:[GGMLMetalClass class]]; NSString * path = [bundle pathForResource:@"ggml-metal" ofType:@"metal"]; - ggml_metal_log_info("%s: loading '%s'\n", __func__, [path UTF8String]); + GGML_METAL_LOG_INFO("%s: loading '%s'\n", __func__, [path UTF8String]); NSString * src = [NSString stringWithContentsOfFile:path encoding:NSUTF8StringEncoding error:&error]; if (error) { - ggml_metal_log_error("%s: error: %s\n", __func__, [[error description] UTF8String]); + GGML_METAL_LOG_ERROR("%s: error: %s\n", __func__, [[error description] UTF8String]); return NULL; } @@ -224,7 +224,7 @@ struct ggml_metal_context * ggml_metal_init(int n_cb) { ctx->library = [ctx->device newLibraryWithSource:src options:nil error:&error]; #endif if (error) { - ggml_metal_log_error("%s: error: %s\n", __func__, [[error description] UTF8String]); + GGML_METAL_LOG_ERROR("%s: error: %s\n", __func__, [[error description] UTF8String]); return NULL; } } @@ -236,11 +236,11 @@ struct ggml_metal_context * ggml_metal_init(int n_cb) { #define GGML_METAL_ADD_KERNEL(name) \ ctx->function_##name = [ctx->library newFunctionWithName:@"kernel_"#name]; \ ctx->pipeline_##name = [ctx->device newComputePipelineStateWithFunction:ctx->function_##name error:&error]; \ - ggml_metal_log_info("%s: loaded %-32s %16p | th_max = %4d | th_width = %4d\n", __func__, "kernel_"#name, (void *) ctx->pipeline_##name, \ + GGML_METAL_LOG_INFO("%s: loaded %-32s %16p | th_max = %4d | th_width = %4d\n", __func__, "kernel_"#name, (void *) ctx->pipeline_##name, \ (int) ctx->pipeline_##name.maxTotalThreadsPerThreadgroup, \ (int) ctx->pipeline_##name.threadExecutionWidth); \ if (error) { \ - ggml_metal_log_error("%s: error: load pipeline error: %s\n", __func__, [[error description] UTF8String]); \ + GGML_METAL_LOG_ERROR("%s: error: load pipeline error: %s\n", __func__, [[error description] UTF8String]); \ return NULL; \ } @@ -296,13 +296,13 @@ struct ggml_metal_context * ggml_metal_init(int n_cb) { #undef GGML_METAL_ADD_KERNEL } - 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 / 1024.0 / 1024.0); + GGML_METAL_LOG_INFO("%s: recommendedMaxWorkingSetSize = %8.2f MB\n", __func__, ctx->device.recommendedMaxWorkingSetSize / 1024.0 / 1024.0); if (ctx->device.maxTransferRate != 0) { - ggml_metal_log_info("%s: maxTransferRate = %8.2f MB/s\n", __func__, ctx->device.maxTransferRate / 1024.0 / 1024.0); + GGML_METAL_LOG_INFO("%s: maxTransferRate = %8.2f MB/s\n", __func__, ctx->device.maxTransferRate / 1024.0 / 1024.0); } else { - ggml_metal_log_info("%s: maxTransferRate = built-in GPU\n", __func__); + GGML_METAL_LOG_INFO("%s: maxTransferRate = built-in GPU\n", __func__); } #endif @@ -310,7 +310,7 @@ struct ggml_metal_context * ggml_metal_init(int n_cb) { } void ggml_metal_free(struct ggml_metal_context * ctx) { - ggml_metal_log_info("%s: deallocating\n", __func__); + GGML_METAL_LOG_INFO("%s: deallocating\n", __func__); #define GGML_METAL_DEL_KERNEL(name) \ [ctx->function_##name release]; \ [ctx->pipeline_##name release]; @@ -382,7 +382,7 @@ void * ggml_metal_host_malloc(size_t n) { void * data = NULL; const int result = posix_memalign((void **) &data, sysconf(_SC_PAGESIZE), n); if (result != 0) { - ggml_metal_log_error("%s: error: posix_memalign failed\n", __func__); + GGML_METAL_LOG_ERROR("%s: error: posix_memalign failed\n", __func__); return NULL; } @@ -410,7 +410,7 @@ int * ggml_metal_get_concur_list(struct ggml_metal_context * ctx) { // Metal buffer based on the host memory pointer // static id ggml_metal_get_buffer(struct ggml_metal_context * ctx, struct ggml_tensor * t, size_t * offs) { - //ggml_metal_log_info("%s: data tensor '%16s', offs_data = %8ld, offs_eval = %8ld, offs_cach = %8ld\n", __func__, t->name, offs_data, offs_eval, offs_cach); + //GGML_METAL_LOG_INFO("%s: data tensor '%16s', offs_data = %8ld, offs_eval = %8ld, offs_cach = %8ld\n", __func__, t->name, offs_data, offs_eval, offs_cach); const int64_t tsize = ggml_nbytes(t); @@ -421,13 +421,13 @@ static id ggml_metal_get_buffer(struct ggml_metal_context * ctx, stru if (ioffs >= 0 && ioffs + tsize <= (int64_t) ctx->buffers[i].size) { *offs = (size_t) ioffs; - //ggml_metal_log_info("%s: '%s' tensor '%16s', offs = %8ld\n", __func__, ctx->buffers[i].name, t->name, *offs); + //GGML_METAL_LOG_INFO("%s: '%s' tensor '%16s', offs = %8ld\n", __func__, ctx->buffers[i].name, t->name, *offs); return ctx->buffers[i].metal; } } - ggml_metal_log_error("%s: error: buffer is nil\n", __func__); + GGML_METAL_LOG_ERROR("%s: error: buffer is nil\n", __func__); return nil; } @@ -439,7 +439,7 @@ bool ggml_metal_add_buffer( size_t size, size_t max_size) { if (ctx->n_buffers >= GGML_METAL_MAX_BUFFERS) { - ggml_metal_log_error("%s: error: too many buffers\n", __func__); + GGML_METAL_LOG_ERROR("%s: error: too many buffers\n", __func__); return false; } @@ -449,7 +449,7 @@ bool ggml_metal_add_buffer( const int64_t ioffs = (int64_t) data - (int64_t) ctx->buffers[i].data; if (ioffs >= 0 && ioffs < (int64_t) ctx->buffers[i].size) { - ggml_metal_log_error("%s: error: buffer '%s' overlaps with '%s'\n", __func__, name, ctx->buffers[i].name); + GGML_METAL_LOG_ERROR("%s: error: buffer '%s' overlaps with '%s'\n", __func__, name, ctx->buffers[i].name); return false; } } @@ -470,11 +470,11 @@ bool ggml_metal_add_buffer( ctx->buffers[ctx->n_buffers].metal = [ctx->device newBufferWithBytesNoCopy:data length:size_aligned options:MTLResourceStorageModeShared deallocator:nil]; if (ctx->buffers[ctx->n_buffers].metal == nil) { - ggml_metal_log_error("%s: error: failed to allocate '%-16s' buffer, size = %8.2f MB\n", __func__, name, size_aligned / 1024.0 / 1024.0); + GGML_METAL_LOG_ERROR("%s: error: failed to allocate '%-16s' buffer, size = %8.2f MB\n", __func__, name, size_aligned / 1024.0 / 1024.0); return false; } - ggml_metal_log_info("%s: allocated '%-16s' buffer, size = %8.2f MB", __func__, name, size_aligned / 1024.0 / 1024.0); + GGML_METAL_LOG_INFO("%s: allocated '%-16s' buffer, size = %8.2f MB", __func__, name, size_aligned / 1024.0 / 1024.0); ++ctx->n_buffers; } else { @@ -494,13 +494,13 @@ bool ggml_metal_add_buffer( ctx->buffers[ctx->n_buffers].metal = [ctx->device newBufferWithBytesNoCopy:(void *) ((uint8_t *) data + i) length:size_step_aligned options:MTLResourceStorageModeShared deallocator:nil]; if (ctx->buffers[ctx->n_buffers].metal == nil) { - ggml_metal_log_error("%s: error: failed to allocate '%-16s' buffer, size = %8.2f MB\n", __func__, name, size_step_aligned / 1024.0 / 1024.0); + GGML_METAL_LOG_ERROR("%s: error: failed to allocate '%-16s' buffer, size = %8.2f MB\n", __func__, name, size_step_aligned / 1024.0 / 1024.0); return false; } - ggml_metal_log_info("%s: allocated '%-16s' buffer, size = %8.2f MB, offs = %12ld", __func__, name, size_step_aligned / 1024.0 / 1024.0, i); + GGML_METAL_LOG_INFO("%s: allocated '%-16s' buffer, size = %8.2f MB, offs = %12ld", __func__, name, size_step_aligned / 1024.0 / 1024.0, i); if (i + size_step < size) { - ggml_metal_log_info("\n"); + GGML_METAL_LOG_INFO("\n"); } ++ctx->n_buffers; @@ -508,17 +508,17 @@ bool ggml_metal_add_buffer( } #if TARGET_OS_OSX - ggml_metal_log_info(", (%8.2f / %8.2f)", + GGML_METAL_LOG_INFO(", (%8.2f / %8.2f)", ctx->device.currentAllocatedSize / 1024.0 / 1024.0, ctx->device.recommendedMaxWorkingSetSize / 1024.0 / 1024.0); if (ctx->device.currentAllocatedSize > ctx->device.recommendedMaxWorkingSetSize) { - ggml_metal_log_warn("%s: warning: current allocated size is greater than the recommended max working set size\n", __func__); + 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"); + GGML_METAL_LOG_INFO("\n"); } #else - ggml_metal_log_info(", (%8.2f)\n", ctx->device.currentAllocatedSize / 1024.0 / 1024.0); + GGML_METAL_LOG_INFO(", (%8.2f)\n", ctx->device.currentAllocatedSize / 1024.0 / 1024.0); #endif } @@ -631,7 +631,7 @@ void ggml_metal_graph_find_concurrency( } if (ctx->concur_list_len > GGML_MAX_CONCUR) { - ggml_metal_log_warn("%s: too many elements for metal ctx->concur_list!\n", __func__); + GGML_METAL_LOG_WARN("%s: too many elements for metal ctx->concur_list!\n", __func__); } } @@ -685,7 +685,7 @@ void ggml_metal_graph_compute( continue; } - //ggml_metal_log_info("%s: encoding node %3d, op = %8s\n", __func__, i, ggml_op_name(gf->nodes[i]->op)); + //GGML_METAL_LOG_INFO("%s: encoding node %3d, op = %8s\n", __func__, i, ggml_op_name(gf->nodes[i]->op)); struct ggml_tensor * src0 = gf->nodes[i]->src[0]; struct ggml_tensor * src1 = gf->nodes[i]->src[1]; @@ -729,17 +729,17 @@ void ggml_metal_graph_compute( id id_src1 = src1 ? ggml_metal_get_buffer(ctx, src1, &offs_src1) : nil; id id_dst = dst ? ggml_metal_get_buffer(ctx, dst, &offs_dst) : nil; - //ggml_metal_log_info("%s: op - %s\n", __func__, ggml_op_name(dst->op)); + //GGML_METAL_LOG_INFO("%s: op - %s\n", __func__, ggml_op_name(dst->op)); //if (src0) { - // ggml_metal_log_info("%s: src0 - %4s [%5lld, %5lld, %5lld], %d, %s\n", __func__, ggml_type_name(src0t), ne00, ne01, ne02, + // GGML_METAL_LOG_INFO("%s: src0 - %4s [%5lld, %5lld, %5lld], %d, %s\n", __func__, ggml_type_name(src0t), ne00, ne01, ne02, // ggml_is_contiguous(src0), src0->name); //} //if (src1) { - // ggml_metal_log_info("%s: src1 - %4s [%5lld, %5lld, %5lld], %d, %s\n", __func__, ggml_type_name(src1t), ne10, ne11, ne12, + // GGML_METAL_LOG_INFO("%s: src1 - %4s [%5lld, %5lld, %5lld], %d, %s\n", __func__, ggml_type_name(src1t), ne10, ne11, ne12, // ggml_is_contiguous(src1), src1->name); //} //if (dst) { - // ggml_metal_log_info("%s: dst - %4s [%5lld, %5lld, %5lld], 1, %s\n", __func__, ggml_type_name(dstt), ne0, ne1, ne2, + // GGML_METAL_LOG_INFO("%s: dst - %4s [%5lld, %5lld, %5lld], 1, %s\n", __func__, ggml_type_name(dstt), ne0, ne1, ne2, // dst->name); //} @@ -845,7 +845,7 @@ void ggml_metal_graph_compute( } break; default: { - ggml_metal_log_warn("%s: node %3d, op = %8s not implemented\n", __func__, i, ggml_op_name(dst->op)); + GGML_METAL_LOG_WARN("%s: node %3d, op = %8s not implemented\n", __func__, i, ggml_op_name(dst->op)); GGML_ASSERT(false); } } break; @@ -1025,7 +1025,7 @@ void ggml_metal_graph_compute( } break; default: { - ggml_metal_log_error("Asserting on type %d\n",(int)src0t); + GGML_METAL_LOG_ERROR("Asserting on type %d\n",(int)src0t); GGML_ASSERT(false && "not implemented"); } }; @@ -1266,7 +1266,7 @@ void ggml_metal_graph_compute( } break; default: { - ggml_metal_log_error("%s: error: node %3d, op = %8s not implemented\n", __func__, i, ggml_op_name(dst->op)); + GGML_METAL_LOG_ERROR("%s: error: node %3d, op = %8s not implemented\n", __func__, i, ggml_op_name(dst->op)); GGML_ASSERT(false); } } @@ -1291,7 +1291,7 @@ void ggml_metal_graph_compute( MTLCommandBufferStatus status = (MTLCommandBufferStatus) [ctx->command_buffers[i] status]; if (status != MTLCommandBufferStatusCompleted) { - ggml_metal_log_info("%s: command buffer %d failed with status %lu\n", __func__, i, status); + GGML_METAL_LOG_INFO("%s: command buffer %d failed with status %lu\n", __func__, i, status); GGML_ASSERT(false); } }