streaming tensors for GPU

streaming tensors for GPU (#6)

* first milestone of t2x

* [Add nvidia_libdevice_bitcode]

* enable fifo and AOT (still have bugs in AOT)

* solve a bug in channel promotion

* rewrite channel promotion

* fix bugs

* fix a bug

* pass correctness tests

* [Cosmetic Revision]

* [Turn on WITH_EXCEPTIONS=1 for make by default. Revised README]

* [Add an intro video]

* [Revised README for GEMM]

* [Change all test.sh files to be executable]

* [Revised fft funct.h]

* [Make bash files executable]

* [Temporarily delete the intro video for being too large]

* [Add intro video again]

* [Add first slide image of the intro video]

* [Use URL for the intro video]

* [Remove embedding of video, which does not work]

* revise code and README

* Update README.md

* [Revised README for GEMM. Not fully done yet. To further revise]

* [Revise SGEMM README. Keep only AOT mode for simplicity. To verify on DevCloud]

* [Verified TINY emulation on DevCloud A10 1.2.1]

* [Recover gemm-run.cpp in aot]

* [Fix for security]

* [Revise GEMM README for more accurate instructions on DevCloud]

* fix a bug in insert_fpga_reg

* [Shared utilis for JIT runtime, AOT runtime, and roofline drawing. SGEMM perf test works]

* [Fix opencl.cpp: not to use SharedUtilsInC, which allocates and frees
 memory and sees to introduce issues with Halide]

* [Fix test/correctness/roofline]

* [Allow plotting rooflines without xserver]

* [Minor fixes to gemm perf test commands, etc.]

* [Misc tiny fixes]

* perf test on GEN9

* Update README.md

* conv perf test on a10/gen9

* fix a bug in conv-run

* Update README.md

* capsule perf test on A10/GEN9

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* setenv for GPUs

* GEMM.cpp: remove kkk from space loop

* fix a bug

* SGEMM perf test: Move host-files/gemm-run.cpp to be gemm-run-gpu.cpp

* Fix SGEMM perf test environment.

* Revised Conv and Capsule perf test

* Fix sgemm perf table

* SEMM perf test: fix gpu env

* SGEMM perf test: Add a design diagram to README

* Stensor: remove bankwidth(). Combine bankwidth() and banks() into out()

* setenv: Remove MARCH variable

* Perf tests: simplify commands about architectures.

* Perf test READMEs: tiny format fix

* README: add throughput numbers

* Minor fix

* GEMM perf test: notes on perf numbers

* Renaming vars in conv.

* 2d conv perf test: Rename Funcs

* conv perf test: more renaming of vars

* Minor fix

* Minor revision to readmes

* Add equation to 2dconv

* SGEMM perf test: revised README

* Security fix.

* fix bugs with trivial loops

* Fix a typo

* conv perf test: Fix a compile error for FPGA

* Revised README of GEMM and Conv perf test

* Update README.md

* update conv perf test

* Update README.md

* Fix typos

* change banks to outs in output network

* Add test scripts for perf tests

* Revise README for perf test instructions.

* Revise testing scripts for perf

* Revise perf test scripts

* Disable building libsanatizer, which failed on a GEN9 machine.

* Revise perf test scripts

* Break install-tools to two files

* Revise README

* update perf test for GPUs

* revise perf test

* Fix ITER for gpu perf tests

* Tiny fx to capsule-run-fpga.cpp

* slightly improve capsule performance

* Simplify test perf readmes.

* Update README.md

* Update README.md

* Update README.md

* Revise README

* Revise README

* Update README.md

* Revise conv perf design figure

* Revise README of capsule

* modify capsule interface

* Revise capseule figure

* Rename vars in capsule

* Fix a typo

* Minor update to test scripts

* Minor revision

* Update README.md

* Update tests.sh

* Update const-parameters.h

* Update README.md

* Update conv perf test figure

* Update conv perf test figure

* Update README.md

* Update const-parameters.h

* Update capsule figure

* Update const-parameters.h

* Revise capsule number based on latest test

* Revise READMEs

* Print tensor sizes

* Minor fix

* Fix readmes

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* Update README.md

Co-authored-by: xiaochen.hao <xiaochen.hao@stu.pku.edu.cn>
Co-authored-by: haoxiaochen <71343933+haoxiaochen@users.noreply.github.com>
Co-authored-by: xiaochen.hao <xiaochen.hao@intel.com>

v1.1

First milestone of T2X: streaming tensors for FPGAs, perf test of SGEMM on A10

v1.0

[Code base for programming FPGAs]