GPGPU processor supporting RISCV-V extension, developed with Verilog.
Copyright (c) 2023-2024 C*Core Technology Co.,Ltd,Suzhou.
这是“乘影”的Verilog版本,原版(Chisel HDL)链接在这里
乘影开源GPGPU项目网站:opengpgpu.org.cn
目前乘影在硬件设计上还有很多不足,如果您有意愿参与到“乘影”的开发中,欢迎在github上pull request
乘影的硬件架构文档在这里
承影的硬件结构框图:
SM核的硬件结构框图:
以gassian用例为例,进入testcase/test_gpgpu_axi_top/tc_gaussian
:
tc.v
,选择case的warp数和thread数在
modules/define/define.v
目录下,修改NUM_THREAD
,可以更改warp内的thread数量
make run-vcs-4w4t
PASSED
或FAILED
:make verdi
testcase/test_gpgpu_top/tc_gaussian
,步骤同上测试集 | warp/thread数 | 是否通过 | 说明 |
---|---|---|---|
vecadd | 4w16t | pass | 向量加 |
4w8t | pass | ||
4w32t | pass | ||
8w4t | pass | ||
matadd | 1w16t | pass | 矩阵加 |
2w8t | pass | ||
4w4t | pass | ||
nn | 2w16t | pass | 最近邻内插法 |
4w8t | pass | ||
4w16t | pass | ||
8w4t | pass | ||
8w8t | pass | ||
gassian | 1w16t | pass | 高斯消元 |
2w8t | pass | ||
4w4t | pass | ||
4w8t | pass | ||
bfs | 2w16t | pass | 宽度优先搜索算法 |
4w8t | pass | ||
4w32t | pass | ||
8w4t | pass |
We refer to some open-source design when developing Ventus GPGPU.
Sub module | Source | Detail |
---|---|---|
CTA scheduler | MIAOW | Our CTA scheduler module is based on MiaoW ultra-threads dispatcher |
L2Cache | block-inclusivecache-sifive | Our L2Cache design is inspired by Sifive's block-inclusivecache |
FPU | XiangShan | We reused Array Multiplier in XiangShan. FPU design is also inspired by XiangShan |
SFU | openhwgroup | Our SFU module is based on pulp-platform |
Config, ... | rocket-chip | Some modules are sourced from RocketChip |