From 55d6088581e04176539e0e7c6771fb107a125cd3 Mon Sep 17 00:00:00 2001 From: Carrie Fallows Date: Tue, 19 Nov 2024 11:19:46 -0500 Subject: [PATCH] Minor modifications removing hardcoded variables in roofline files. Signed-off-by: Carrie Fallows --- src/roofline.py | 2 +- src/utils/roofline_calc.py | 39 +++++++++++++++++++------------------- 2 files changed, 20 insertions(+), 21 deletions(-) diff --git a/src/roofline.py b/src/roofline.py index 77f491b59..86c15b6e8 100644 --- a/src/roofline.py +++ b/src/roofline.py @@ -110,7 +110,7 @@ def empirical_roofline( """Generate a set of empirical roofline plots given a directory containing required profiling and benchmarking data""" # Create arithmetic intensity data that will populate the roofline model console_debug("roofline", "Path: %s" % self.__run_parameters["workload_dir"]) - self.__ai_data = calc_ai(self.__run_parameters["sort_type"], ret_df) + self.__ai_data = calc_ai(self.__mspec, self.__run_parameters["sort_type"], ret_df) msg = "AI at each mem level:" for i in self.__ai_data: diff --git a/src/utils/roofline_calc.py b/src/utils/roofline_calc.py index 22dc7a71f..f06bea5dc 100644 --- a/src/utils/roofline_calc.py +++ b/src/utils/roofline_calc.py @@ -34,8 +34,6 @@ IMGNAME = "empirRoof" -L2_BANKS = 32 # default assuming mi200, mi300 - XMIN = 0.01 XMAX = 1000 @@ -188,7 +186,7 @@ def calc_ceilings(roofline_parameters, dtype, benchmark_data): # Overlay application performance # ------------------------------------------------------------------------------------- # Calculate relevant metrics for ai calculation -def calc_ai(sort_type, ret_df): +def calc_ai(mspec, sort_type, ret_df): """Given counter data, calculate arithmetic intensity for each kernel in the application.""" df = ret_df["pmc_perf"] # Sort by top kernels or top dispatches? @@ -306,8 +304,8 @@ def calc_ai(sort_type, ret_df): lds_data += ( (df["SQ_LDS_IDX_ACTIVE"][idx] - df["SQ_LDS_BANK_CONFLICT"][idx]) * 4 - * L2_BANKS - ) # L2_BANKS = 32 (since assuming mi200 or mi300) + * (mspec.lds_banks_per_cu) + ) except KeyError: console_debug( "roofline", @@ -338,37 +336,38 @@ def calc_ai(sort_type, ret_df): ) pass try: - if df.keys().str.contains("TCC_BUBBLE").sum() > 0: - # MI300 uses TCC_BUBBLE_sum to calculate hbm_data + if mspec.gpu_model == "MI200": hbm_data += ( - (df["TCC_BUBBLE_sum"][idx] * 128) - + (df["TCC_EA0_RDREQ_32B_sum"][idx] * 32) + (df["TCC_EA_RDREQ_32B_sum"][idx] * 32) + ( - ( - df["TCC_EA0_RDREQ_sum"][idx] - - df["TCC_BUBBLE_sum"][idx] - - df["TCC_EA0_RDREQ_32B_sum"][idx] - ) + (df["TCC_EA_RDREQ_sum"][idx] - df["TCC_EA_RDREQ_32B_sum"][idx]) * 64 ) + + (df["TCC_EA_WRREQ_64B_sum"][idx] * 64) + ( - (df["TCC_EA0_WRREQ_sum"][idx] - df["TCC_EA0_WRREQ_64B_sum"][idx]) + (df["TCC_EA_WRREQ_sum"][idx] - df["TCC_EA_WRREQ_64B_sum"][idx]) * 32 ) - + (df["TCC_EA0_WRREQ_64B_sum"][idx] * 64) ) + else: + # Use TCC_BUBBLE_sum to calculate hbm_data hbm_data += ( - (df["TCC_EA_RDREQ_32B_sum"][idx] * 32) + (df["TCC_BUBBLE_sum"][idx] * 128) + + (df["TCC_EA0_RDREQ_32B_sum"][idx] * 32) + ( - (df["TCC_EA_RDREQ_sum"][idx] - df["TCC_EA_RDREQ_32B_sum"][idx]) + ( + df["TCC_EA0_RDREQ_sum"][idx] + - df["TCC_BUBBLE_sum"][idx] + - df["TCC_EA0_RDREQ_32B_sum"][idx] + ) * 64 ) - + (df["TCC_EA_WRREQ_64B_sum"][idx] * 64) + ( - (df["TCC_EA_WRREQ_sum"][idx] - df["TCC_EA_WRREQ_64B_sum"][idx]) + (df["TCC_EA0_WRREQ_sum"][idx] - df["TCC_EA0_WRREQ_64B_sum"][idx]) * 32 ) + + (df["TCC_EA0_WRREQ_64B_sum"][idx] * 64) ) except KeyError: console_debug(