runtime(cinn): update cinn jit instruction to support dynamic shape

[6clc]

PR types

Others

PR changes

Others

Description

cinn jit instruction 支持动态shape。

• 1. cinn jit instruction 由调用 cuda kernel ptr 改为 host kernel ptr

• 2. CINNKernelInfo额外包含非tensor arg信息。数据结构详见代码注释。

以调用cos算子为例子，
host kernel cinn ir 如下

Module Pir_host {
     function fn_cos (kernel_args, kernel_args_num, kernel_stream)
     cinn_call_cuda_kernel(fn_cos_kernel, kernel_args, kernel_args_num, 1, 1, 1, 4, 1, 1, kernel_stream)
}

cuda kernel cinn ir 如下

function fn_cos_kernel (_var_0, _var_3)
 { 
   if ((threadIdx.x < 4)) {
      var_3[(threadIdx.x / 2), (threadIdx.x % 2)] = cinn_nvgpu_cos_fp32(var_0[(threadIdx.x / 2), (threadIdx.x % 2)])
   }
}

Pcard-78120

[paddle-bot]

你的PR提交成功，感谢你对开源项目的贡献!
请关注后续CI自动化测试结果，详情请参考Paddle-CI手册。
Your PR has been submitted. Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

[paddle-bot]

✅ This PR's description meets the template requirements!
Please wait for other CI results.

[zhhsplendid]

LGTM

[zhangbo9674]

LGTM for cinn_jit_instruction