Merge "msm: vidc: Fix bus voting overflow issues on 32 bit platform" (cb7ab9f5) · Commits · e / devices / android_kernel_fairphone_FP3

drivers/media/platform/msm/vidc/governors/fixedpoint.h

+4 −4

Original line number	Diff line number	Diff line
		/* Copyright (c) 2015-2017, The Linux Foundation. All rights reserved.
		/* Copyright (c) 2015-2019, The Linux Foundation. All rights reserved.
		*
		* This program is free software; you can redistribute it and/or modify
		* it under the terms of the GNU General Public License version 2 and
		@@ -22,14 +22,14 @@
		* Also should be normally typedef'ed to intmax_t but that doesn't seem to be
		* available in the kernel
		*/
		#define fp_t size_t
		#define fp_t uint64_t

		/* (Arbitrarily) make the first 25% of the bits to be the fractional bits */
		#define FP_FRACTIONAL_BITS ((sizeof(fp_t) * 8) / 4)

		#define FP(__i, __f_n, __f_d) \
		((((fp_t)(__i)) << FP_FRACTIONAL_BITS) + \
		(((__f_n) << FP_FRACTIONAL_BITS) / (__f_d)))
		div64_u64(((__f_n) << FP_FRACTIONAL_BITS), (__f_d)))

		#define FP_INT(__i) FP(__i, 0, 1)
		#define FP_ONE FP_INT(1)
		@@ -66,7 +66,7 @@ static inline fp_t fp_mult(fp_t a, fp_t b)

		static inline fp_t fp_div(fp_t a, fp_t b)
		{
		return (a << FP_FRACTIONAL_BITS) / b;
		return div64_u64(a << FP_FRACTIONAL_BITS, b);
		}

		#endif

drivers/media/platform/msm/vidc/governors/msm_vidc_dyn_gov.c

+66 −42

Original line number	Diff line number	Diff line
		/* Copyright (c) 2015-2018, The Linux Foundation. All rights reserved.
		/* Copyright (c) 2015-2019, The Linux Foundation. All rights reserved.
		*
		* This program is free software; you can redistribute it and/or modify
		* it under the terms of the GNU General Public License version 2 and
		@@ -37,7 +37,7 @@ struct governor {
		* bandwidth(BASELINE_DIMENSIONS.width, BASELINE_DIMENSIONS.height)
		*/
		const struct {
		int height, width;
		uint32_t height, width;
		} BASELINE_DIMENSIONS = {
		.width = 1280,
		.height = 720,
		@@ -76,7 +76,7 @@ const unsigned long NOMINAL_BW_MBPS = 6000 /* ideally 320 Mhz */,
		static struct lut {
		int frame_size; /* width x height */
		int frame_rate;
		unsigned long bitrate;
		fp_t bitrate;
		struct {
		int bpp;
		fp_t ratio;
		@@ -87,10 +87,12 @@ static struct lut {
		.frame_rate = 30,
		.bitrate = 14,
		.compression_ratio = {
		/* FP(1, 28, 100) */
		GENERATE_COMPRESSION_PROFILE(8,
		FP(1, 28, 100)),
		83886),
		/* FP(1, 23, 100) */
		GENERATE_COMPRESSION_PROFILE(10,
		FP(1, 23, 100)),
		80609),
		}
		},
		{
		@@ -98,10 +100,12 @@ static struct lut {
		.frame_rate = 60,
		.bitrate = 22,
		.compression_ratio = {
		/* FP(1, 28, 100) */
		GENERATE_COMPRESSION_PROFILE(8,
		FP(1, 28, 100)),
		83886),
		/* FP(1, 23, 100) */
		GENERATE_COMPRESSION_PROFILE(10,
		FP(1, 23, 100)),
		80609),
		}
		},
		{
		@@ -109,10 +113,12 @@ static struct lut {
		.frame_rate = 30,
		.bitrate = 40,
		.compression_ratio = {
		/* FP(1, 28, 100) */
		GENERATE_COMPRESSION_PROFILE(8,
		FP(1, 28, 100)),
		83886),
		/* FP(1, 23, 100) */
		GENERATE_COMPRESSION_PROFILE(10,
		FP(1, 23, 100)),
		80609),
		}
		},
		{
		@@ -120,10 +126,12 @@ static struct lut {
		.frame_rate = 60,
		.bitrate = 64,
		.compression_ratio = {
		/* FP(1, 28, 100) */
		GENERATE_COMPRESSION_PROFILE(8,
		FP(1, 28, 100)),
		83886),
		/* FP(1, 23, 100) */
		GENERATE_COMPRESSION_PROFILE(10,
		FP(1, 23, 100)),
		80609),
		}
		},
		{
		@@ -131,10 +139,12 @@ static struct lut {
		.frame_rate = 30,
		.bitrate = 64,
		.compression_ratio = {
		/* FP(1, 26, 100) */
		GENERATE_COMPRESSION_PROFILE(8,
		FP(1, 26, 100)),
		82575),
		/* FP(1, 22, 100) */
		GENERATE_COMPRESSION_PROFILE(10,
		FP(1, 22, 100)),
		79953),
		}
		},
		{
		@@ -142,10 +152,12 @@ static struct lut {
		.frame_rate = 60,
		.bitrate = 102,
		.compression_ratio = {
		/* FP(1, 26, 100) */
		GENERATE_COMPRESSION_PROFILE(8,
		FP(1, 26, 100)),
		82575),
		/* FP(1, 22, 100) */
		GENERATE_COMPRESSION_PROFILE(10,
		FP(1, 22, 100)),
		79953),
		}
		},
		{
		@@ -153,10 +165,12 @@ static struct lut {
		.frame_rate = 30,
		.bitrate = 84,
		.compression_ratio = {
		/* FP(1, 26, 100) */
		GENERATE_COMPRESSION_PROFILE(8,
		FP(1, 26, 100)),
		82575),
		/* FP(1, 22, 100) */
		GENERATE_COMPRESSION_PROFILE(10,
		FP(1, 22, 100)),
		79953),
		}
		},
		{
		@@ -164,10 +178,12 @@ static struct lut {
		.frame_rate = 60,
		.bitrate = 134,
		.compression_ratio = {
		/* FP(1, 26, 100) */
		GENERATE_COMPRESSION_PROFILE(8,
		FP(1, 26, 100)),
		82575),
		/* FP(1, 22, 100) */
		GENERATE_COMPRESSION_PROFILE(10,
		FP(1, 22, 100)),
		79953),
		}
		},
		{
		@@ -175,10 +191,12 @@ static struct lut {
		.frame_rate = 30,
		.bitrate = 88,
		.compression_ratio = {
		/* FP(1, 26, 100) */
		GENERATE_COMPRESSION_PROFILE(8,
		FP(1, 26, 100)),
		82575),
		/* FP(1, 22, 100) */
		GENERATE_COMPRESSION_PROFILE(10,
		FP(1, 22, 100)),
		79953),
		}
		},
		{
		@@ -186,10 +204,12 @@ static struct lut {
		.frame_rate = 60,
		.bitrate = 141,
		.compression_ratio = {
		/* FP(1, 26, 100) */
		GENERATE_COMPRESSION_PROFILE(8,
		FP(1, 26, 100)),
		82575),
		/* FP(1, 22, 100) */
		GENERATE_COMPRESSION_PROFILE(10,
		FP(1, 22, 100)),
		79953),
		}
		},
		{
		@@ -197,10 +217,12 @@ static struct lut {
		.frame_rate = 30,
		.bitrate = 96,
		.compression_ratio = {
		/* FP(1, 26, 100) */
		GENERATE_COMPRESSION_PROFILE(8,
		FP(1, 26, 100)),
		82575),
		/* FP(1, 22, 100) */
		GENERATE_COMPRESSION_PROFILE(10,
		FP(1, 22, 100)),
		79953),
		}
		},
		{
		@@ -208,17 +230,19 @@ static struct lut {
		.frame_rate = 60,
		.bitrate = 154,
		.compression_ratio = {
		/* FP(1, 26, 100) */
		GENERATE_COMPRESSION_PROFILE(8,
		FP(1, 26, 100)),
		82575),
		/* FP(1, 22, 100) */
		GENERATE_COMPRESSION_PROFILE(10,
		FP(1, 22, 100)),
		79953),
		}
		},
		};

		static struct lut const *__lut(int width, int height, int fps)
		static struct lut const *__lut(uint32_t width, uint32_t height, uint32_t fps)
		{
		int frame_size = height * width, c = 0;
		uint32_t frame_size = height * width, c = 0;

		do {
		if (LUT[c].frame_size >= frame_size && LUT[c].frame_rate >= fps)
		@@ -330,17 +354,17 @@ static unsigned long __calculate_decoder(struct vidc_bus_vote_data *d,
		* measured heuristics and hardcoded numbers taken from the firmware.
		*/
		/* Decoder parameters */
		int width, height, lcu_size, dpb_bpp, opb_bpp, fps, opb_factor;
		uint32_t width, height, lcu_size, dpb_bpp, opb_bpp, fps;
		bool unified_dpb_opb, dpb_compression_enabled, opb_compression_enabled,
		llc_ref_read_l2_cache_enabled = false,
		llc_vpss_ds_line_buf_enabled = false;
		fp_t dpb_opb_scaling_ratio, dpb_read_compression_factor,
		dpb_write_compression_factor, opb_compression_factor,
		qsmmu_bw_overhead_factor, height_ratio;
		qsmmu_bw_overhead_factor, height_ratio, opb_factor;

		/* Derived parameters */
		int lcu_per_frame, tnbr_per_lcu, colocated_bytes_per_lcu;
		unsigned long bitrate;
		fp_t lcu_per_frame, tnbr_per_lcu, colocated_bytes_per_lcu;
		fp_t bitrate;

		fp_t bins_to_bit_factor, dpb_write_factor, ten_bpc_packing_factor,
		ten_bpc_bpp_factor, vsp_read_factor, vsp_write_factor,
		@@ -360,8 +384,8 @@ static unsigned long __calculate_decoder(struct vidc_bus_vote_data *d,
		fp_t dpb_read, opb_read, total;
		} llc = {0};

		unsigned long ret = 0;
		unsigned int integer_part, frac_part;
		fp_t ret = 0;
		fp_t integer_part, frac_part;

		width = max(d->input_width, BASELINE_DIMENSIONS.width);
		height = max(d->input_height, BASELINE_DIMENSIONS.height);
		@@ -446,13 +470,13 @@ static unsigned long __calculate_decoder(struct vidc_bus_vote_data *d,
		ddr.vsp_write = fp_div(fp_mult(FP_INT(bitrate),
		vsp_write_factor), FP_INT(8));

		ddr.collocated_read = FP_INT(lcu_per_frame *
		colocated_bytes_per_lcu * fps / bps(1));
		ddr.collocated_write = FP_INT(lcu_per_frame *
		colocated_bytes_per_lcu * fps / bps(1));
		ddr.collocated_read = FP_INT(div64_u64(lcu_per_frame *
		colocated_bytes_per_lcu * fps, bps(1)));
		ddr.collocated_write = FP_INT(div64_u64(lcu_per_frame *
		colocated_bytes_per_lcu * fps, bps(1)));

		ddr.line_buffer_read = FP_INT(tnbr_per_lcu *
		lcu_per_frame * fps / bps(1));
		ddr.line_buffer_read = FP_INT(div64_u64(tnbr_per_lcu *
		lcu_per_frame * fps, bps(1)));
		ddr.line_buffer_write = ddr.line_buffer_read;

		bw_for_1x_8bpc = fp_div(FP_INT(width * height), FP_INT(32 * 8));
		@@ -590,7 +614,7 @@ static unsigned long __calculate_encoder(struct vidc_bus_vote_data *d,
		*/
		/* Encoder Parameters */

		int width, height, fps, dpb_bpp, lcu_per_frame, lcu_size,
		u32 width, height, fps, dpb_bpp, lcu_per_frame, lcu_size,
		vertical_tile_width, colocated_bytes_per_lcu, bitrate,
		ref_overlap_bw_factor;
		enum hal_uncompressed_format dpb_color_format, original_color_format;

drivers/media/platform/msm/vidc/msm_vidc_clocks.c

+1 −1

Original line number	Diff line number	Diff line
		@@ -163,7 +163,7 @@ static int fill_dynamic_stats(struct msm_vidc_inst *inst,
		}

		dprintk(VIDC_PROF,
		"Input CR = %d Recon CR = %d Complexity Factor = %d\n",
		"Input CR = %d Recon CR = %llu Complexity Factor = %llu\n",
		vote_data->input_cr, vote_data->compression_ratio,
		vote_data->complexity_factor);

drivers/media/platform/msm/vidc/vidc_hfi_api.h

+5 −5

Original line number	Diff line number	Diff line
		@@ -1407,11 +1407,11 @@ struct vidc_bus_vote_data {
		enum hal_video_codec codec;
		enum hal_uncompressed_format color_formats[2];
		int num_formats; /* 1 = DPB-OPB unified; 2 = split */
		int input_height, input_width, fps, bitrate;
		int output_height, output_width;
		int compression_ratio;
		int complexity_factor;
		int input_cr;
		u32 input_height, input_width, fps, bitrate;
		u32 output_height, output_width;
		uint64_t compression_ratio;
		uint64_t complexity_factor;
		u32 input_cr;
		bool use_dpb_read;
		unsigned int lcu_size;
		enum msm_vidc_power_mode power_mode;