adreno系列gpu dts配置与gpu列表

Posted bubbleben

tags:

篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了adreno系列gpu dts配置与gpu列表相关的知识,希望对你有一定的参考价值。

以Find x3 pro设备树文件为例分析:
https://github.com/oppo-source/android_kernel_modules_and_devicetree_oppo_sm8350/blob/oppo/sm8350_s_12.1_find_x3_pro/vendor/qcom/proprietary/devicetree/qcom/sm8150-gpu.dtsi

1. sm8150-gpu.dtsi

#define MHZ_TO_KBPS(mhz, w) ((mhz * 1000000 * w) / (1024))

&soc 
	pil_gpu: qcom,kgsl-hyp 
		compatible = "qcom,pil-tz-generic";
		qcom,pas-id = <13>;
		qcom,firmware-name = "a640_zap";
		memory-region = <&pil_gpu_mem>;
	;

    // kgsl-3d0配置
	msm_gpu: qcom,kgsl-3d0@2c00000 
		label = "kgsl-3d0";
		compatible = "qcom,kgsl-3d0", "qcom,kgsl-3d";
		status = "ok";
		reg = <0x2c00000 0x40000>, <0x2c61000 0x800>,
				<0x6900000 0x44000>, <0x780000 0x6fff>;
		reg-names = "kgsl_3d0_reg_memory", "cx_dbgc",
					"qdss_gfx", "qfprom_memory";
		interrupts = <0 300 IRQ_TYPE_LEVEL_HIGH>;
        // 中断名称
		interrupt-names = "kgsl_3d0_irq";
		qcom,id = <0>;

        // 主版本号:6
        // 副版本号:4
        // patchid: 0
		qcom,chipid = <0x06040000>;

        // 默认频率
		qcom,initial-pwrlevel = <5>;

		qcom,gpu-quirk-secvid-set-once;
		qcom,gpu-quirk-cx-gdsc;

		qcom,idle-timeout = <80>; //msecs
		qcom,no-nap;

		qcom,highest-bank-bit = <15>;

		qcom,min-access-length = <32>;

		qcom,ubwc-mode = <3>;

		qcom,gpu-qdss-stm = <0x161c0000 0x40000>; // base addr, size

		#cooling-cells = <2>;

        // 时钟
		clocks = <&gcc GCC_GPU_CFG_AHB_CLK>,
			<&gpucc GPU_CC_CXO_CLK>,
			<&gcc GCC_DDRSS_GPU_AXI_CLK>,
			<&gcc GCC_GPU_MEMNOC_GFX_CLK>,
			<&gpucc GPU_CC_CX_GMU_CLK>,
			<&gpucc GPU_CC_AHB_CLK>;

		clock-names = "gcc_gpu_ahb", "rbbmtimer_clk",
					"gcc_gpu_axi_clk", "gcc_gpu_memnoc_gfx",
					"gmu_clk", "gpu_cc_ahb";

		qcom,isense-clk-on-level = <1>;

		interconnects = <&gem_noc MASTER_GFX3D &mc_virt SLAVE_EBI1>;
		interconnect-names = "gpu_icc_path";

		qcom,bus-table-ddr =
			<MHZ_TO_KBPS(0, 4)>,    /* index=0  */
			<MHZ_TO_KBPS(100, 4)>,  /* index=1  */
			<MHZ_TO_KBPS(150, 4)>,  /* index=2  */
			<MHZ_TO_KBPS(200, 4)>,  /* index=3  */
			<MHZ_TO_KBPS(300, 4)>,  /* index=4  */
			<MHZ_TO_KBPS(412, 4)>,  /* index=5  */
			<MHZ_TO_KBPS(547, 4)>, /* index=6  */
			<MHZ_TO_KBPS(681, 4)>, /* index=7  */
			<MHZ_TO_KBPS(768, 4)>, /* index=8  */
			<MHZ_TO_KBPS(1017, 4)>, /* index=9  */
			<MHZ_TO_KBPS(1296, 4)>, /* index=10  */
			<MHZ_TO_KBPS(1555, 4)>, /* index=11  */
			<MHZ_TO_KBPS(1804, 4)>; /* index=12  */

		qcom,bus-table-cnoc =
			<0>,   /* Off */
			<100>; /* On */


		/* GDSC regulator names */
		regulator-names = "vddcx", "vdd";
		/* GDSC oxili regulators */
		vddcx-supply = <&gpu_cx_gdsc>;
		vdd-supply = <&gpu_gx_gdsc>;

		qcom,l3-pwrlevels 
			#address-cells = <1>;
			#size-cells = <0>;

			compatible = "qcom,l3-pwrlevels";

			qcom,l3-pwrlevel@0 
				reg = <0>;
				qcom,l3-freq = <0>;
			;

			qcom,l3-pwrlevel@1 
				reg = <1>;
				qcom,l3-freq = <864000000>;
			;

			qcom,l3-pwrlevel@2 
				reg = <2>;
				qcom,l3-freq = <1344000000>;
			;
		;

		/* GPU Mempools */
        // kgsl内存池配置
		qcom,gpu-mempools 
			#address-cells = <1>;
			#size-cells = <0>;
			compatible = "qcom,gpu-mempools";

			/* 4K Page Pool configuration */
			qcom,gpu-mempool@0 
				reg = <0>;
				qcom,mempool-page-size = <4096>;
				qcom,mempool-reserved = <2048>;
				qcom,mempool-allocate;
			;
			/* 8K Page Pool configuration */
			qcom,gpu-mempool@1 
				reg = <1>;
				qcom,mempool-page-size = <8192>;
				qcom,mempool-reserved = <1024>;
				qcom,mempool-allocate;
			;
			/* 64K Page Pool configuration */
			qcom,gpu-mempool@2 
				reg = <2>;
				qcom,mempool-page-size = <65536>;
				qcom,mempool-reserved = <256>;
			;
			/* 1M Page Pool configuration */
			qcom,gpu-mempool@3 
				reg = <3>;
				qcom,mempool-page-size = <1048576>;
				qcom,mempool-reserved = <32>;
			;
		;

		/* Power levels */
        // kgsl频率配置
		qcom,gpu-pwrlevels 
			#address-cells = <1>;
			#size-cells = <0>;

			compatible = "qcom,gpu-pwrlevels";

			qcom,gpu-pwrlevel@0 
				reg = <0>;
				qcom,gpu-freq = <600000000>;
				qcom,level = <RPMH_REGULATOR_LEVEL_TURBO>;

				qcom,bus-freq = <12>;
				qcom,bus-min = <10>;
				qcom,bus-max = <12>;
			;

			qcom,gpu-pwrlevel@1 
				reg = <1>;
				qcom,gpu-freq = <553850000>;
				qcom,level = <RPMH_REGULATOR_LEVEL_NOM_L1>;

				qcom,bus-freq = <10>;
				qcom,bus-min = <9>;
				qcom,bus-max = <11>;
			;

			qcom,gpu-pwrlevel@2 
				reg = <2>;
				qcom,gpu-freq = <486460000>;
				qcom,level = <RPMH_REGULATOR_LEVEL_NOM>;

				qcom,bus-freq = <9>;
				qcom,bus-min = <8>;
				qcom,bus-max = <10>;
			;

			qcom,gpu-pwrlevel@3 
				reg = <3>;
				qcom,gpu-freq = <379650000>;
				qcom,level = <RPMH_REGULATOR_LEVEL_SVS_L1>;

				qcom,bus-freq = <8>;
				qcom,bus-min = <7>;
				qcom,bus-max = <9>;
			;

			qcom,gpu-pwrlevel@4 
				reg = <4>;
				qcom,gpu-freq = <309110000>;
				qcom,level = <RPMH_REGULATOR_LEVEL_SVS>;

				qcom,bus-freq = <5>;
				qcom,bus-min = <5>;
				qcom,bus-max = <7>;
			;

			qcom,gpu-pwrlevel@5 
				reg = <5>;
				qcom,gpu-freq = <215000000>;
				qcom,level = <RPMH_REGULATOR_LEVEL_LOW_SVS>;

				qcom,bus-freq = <4>;
				qcom,bus-min = <3>;
				qcom,bus-max = <5>;
			;

			qcom,gpu-pwrlevel@6 
				reg = <6>;
				qcom,gpu-freq = <0>;
				qcom,level = <RPMH_REGULATOR_LEVEL_RETENTION>;

				qcom,bus-freq = <0>;
				qcom,bus-min = <0>;
				qcom,bus-max = <0>;
			;
		;

	;

    // iommu配置
	kgsl_msm_iommu: qcom,kgsl-iommu@2ca0000 
		compatible = "qcom,kgsl-smmu-v2";

		reg = <0x2ca0000 0x10000>;

		gfx3d_user: gfx3d_user 
			compatible = "qcom,smmu-kgsl-cb";
			label = "gfx3d_user";
			iommus = <&kgsl_smmu 0x0 0x401>;
			qcom,iommu-dma = "disabled";
		;

		gfx3d_secure: gfx3d_secure 
			compatible = "qcom,smmu-kgsl-cb";
			label = "gfx3d_secure";
			iommus = <&kgsl_smmu 0x2 0x400>;
			qcom,iommu-dma = "disabled";
		;
	;

    // gmu配置
	gmu: qcom,gmu@2c6a000 
		compatible = "qcom,gpu-gmu";

		reg = <0x2c6a000 0x30000>,
			<0xb280000 0x10000>,
			<0xb480000 0x10000>;
		reg-names = "kgsl_gmu_reg",
			"kgsl_gmu_pdc_cfg",
			"kgsl_gmu_pdc_seq";

		interrupts = <0 304 IRQ_TYPE_LEVEL_HIGH>,
			<0 305 IRQ_TYPE_LEVEL_HIGH>;
		interrupt-names = "kgsl_hfi_irq", "kgsl_gmu_irq";

		regulator-names = "vddcx", "vdd";

		iommus = <&kgsl_smmu 0x5 0x400>;
		qcom,iommu-dma = "disabled";

		vddcx-supply = <&gpu_cx_gdsc>;
		vdd-supply = <&gpu_gx_gdsc>;

		clocks = <&gpucc GPU_CC_CX_GMU_CLK>,
			<&gpucc GPU_CC_CXO_CLK>,
			<&gcc GCC_DDRSS_GPU_AXI_CLK>,
			<&gcc GCC_GPU_MEMNOC_GFX_CLK>,
			<&gpucc GPU_CC_AHB_CLK>;

		clock-names = "gmu_clk", "cxo_clk", "axi_clk",
				    "gcc_gpu_memnoc_gfx_clk", "gpu_cc_ahb";

		/* AOP mailbox for sending ACD enable and disable messages */
		mboxes = <&qmp_aop 0>;
		mbox-names = "aop";

	;
;

2. adreno_gpu_core

/**
 * struct adreno_gpu_core - A specific GPU core definition
 */
struct adreno_gpu_core 
	enum adreno_gpurev gpurev;
	unsigned int core, major, minor, patchid;
	/**
	 * @compatible: If specified, use the compatible string to match the
	 * device
	 */
	const char *compatible;
	unsigned long features;
    // Pointer to the GPU family specific functions for this core[见2.1节]
	const struct adreno_gpudev *gpudev;
    // 性能计数器
	const struct adreno_perfcounters *perfcounters;
	size_t gmem_size;
	u32 bus_width;
	/** @snapshot_size: Size of the static snapshot region in bytes */
	u32 snapshot_size;
;

2.1 adreno_gpudev

struct adreno_gpudev 
	/*
	 * These registers are in a different location on different devices,
	 * so define them in the structure and use them as variables.
	 */
	unsigned int *const reg_offsets;
	const struct adreno_ft_perf_counters *ft_perf_counters;
	unsigned int ft_perf_counters_count;

	struct adreno_coresight *coresight[2];

	/* GPU specific function hooks */
	int (*probe)(struct platform_device *pdev, u32 chipid,
		const struct adreno_gpu_core *gpucore);
	void (*snapshot)(struct adreno_device *adreno_dev,
				struct kgsl_snapshot *snapshot);
	irqreturn_t (*irq_handler)(struct adreno_device *adreno_dev);
	int (*init)(struct adreno_device *adreno_dev);
	void (*remove)(struct adreno_device *adreno_dev);
	int (*rb_start)(struct adreno_device *adreno_dev);
	int (*microcode_read)(struct adreno_device *adreno_dev);
	void (*start)(struct adreno_device *adreno_dev);
	bool (*is_sptp_idle)(struct adreno_device *adreno_dev);
	int (*regulator_enable)(struct adreno_device *adreno_dev);
	void (*regulator_disable)(struct adreno_device *adreno_dev);
	void (*pwrlevel_change_settings)(struct adreno_device *adreno_dev,
				unsigned int prelevel, unsigned int postlevel,
				bool post);
	int64_t (*read_throttling_counters)(struct adreno_device *adreno_dev);
	void (*count_throttles)(struct adreno_device *adreno_dev,
					uint64_t adj);
	unsigned int (*preemption_pre_ibsubmit)(
				struct adreno_device *adreno_dev,
				struct adreno_ringbuffer *rb,
				unsigned int *cmds,
				struct kgsl_context *context);
	int (*preemption_yield_enable)(unsigned int *cmds);
	unsigned int (*set_marker)(unsigned int *cmds,
				enum adreno_cp_marker_type type);
	unsigned int (*preemption_post_ibsubmit)(
				struct adreno_device *adreno_dev,
				unsigned int *c

以上是关于adreno系列gpu dts配置与gpu列表的主要内容,如果未能解决你的问题,请参考以下文章

adreno源码系列dts配置与gpu列表

adreno系列dts配置与gpu列表

Adreno GPU Profiler工具使用总结

Android GPU Inspector

GPU结构与CUDA系列1GPU与CPU比较:GPU介绍设计差异计算流程

GPU结构与CUDA系列3GPU软件抽象:Grid,Block,Thread,Warp定义说明与硬件的映射执行细节