RISC-V Vector Automatic Tests Generator

Prerequisite

RVV Compiler
1. Set gcc,objdump, and riscof directory's path variables in file: scripts/constants.py
RISCV-ISAC RVV Support
1. git clone https://github.com/hushenwei2000/riscv-isac-rvv
2. cd riscv-isac-rvv
3. git checkout vetcor
4. pip install . --prefix=~/.local # Anywhere you want
5. Add ~/.local to your PATH
If the terminal can find riscv_isac command then it's successful
Spike
1. Install Latest https://github.com/riscv-software-src/riscv-isa-sim.
$ apt-get install device-tree-compiler
$ mkdir build
$ cd build
$ ../configure --prefix=$RISCV --enable-commitlog
$ make
$ [sudo] make install

⚠️ Need to add --enable-commitlog
2. Add <spike_path>/build to your PATH

If the terminal can find spike command then it's successful

Usage

Generate One Instruction

python run.py -i <instruction> -t <type> [--vlen VLEN] [--vsew VSEW]

The type shall be consistent with the instruction: i (integer), f (floating point), m (mask), p (permute), x (fix point), l (load store)
Supported instruction and type can be seen in cgfs/<type>/<instruction>.yaml
vlen VLEN Vector Register Length: 32, 64, 128(default), 256, 512, 1024
vsew VSEW Selected Element Width: 8, 16, 32(default), 64
Other options seen in python run.py -h

Generate All Tests

python generate_all.py

This will use default parameter configuration to generate all integer instructions tests.
Modify runcommand_<type> function to run different parameter.
Modyfy main function to run different type of instructions.
Test file will generate in generate_all folder.
⚠️ Because we still have bugs, those tests are not correct (i.e. CAN'T be used as test file) will generate spike.log in generate_all folder. You can see FAIL in the ending of these logs.

Check All Generated Tests

python check_all.py

Check all generated folders. Find out if test file FAIL and the statistics of COVERAGE.
Output will in 'check_all.out'.

Support Configuration

Parameter	Numbers	Current Support	Note
vlen	128 ~ 2^16	128 ~ 1024	Spike now support largest 4096
vsew	8, 16, 32, 64	All
lmul	1/8, 1/4, 1/2, 1, 2, 4, 8	All	lmul=8 except load/store， 1/8 except eew>=32 loadstore
vta	0, 1	0
vma	0, 1	0

Notice of Failure Tests

elen = 64 for default
- vsew should <= elen * lmul
- So, when lmul = 0.125, load/store eew=32+ cannot test.
If there are failure tests, most probably is because that configuration can not be tested on that instruction. For example test widen instruction when vsew=64, or test floating point when vsew=8/16, or the elen related restriction above.
- At present, users need to determine which configurations can be produced to avoid such failure. FutureSupport

Develop

Add a instruction

Give a instr such as vadd, this will include vadd.vv/x/i tests. If it is difficult to gather them, it can also be separated.
Put a yaml formatted CGF to cgfs/<type>/ directory. The name should be <instr>.yaml .
Add a file to scripts/create_test_<type>/<instr>.yaml (can copy vadd first).You should mainly modify generate_macros and generate_tests.There are two functions:
Add a file to scripts/create_test_<type>/<instr>.yaml (can copy vadd first).You should mainly modify generate_macros and generate_tests.There are two functions:
1. create_empty_test_vadd: create test which only contains one test, this file is used to generate coverage report.
2. create_first_test_vadd: create test which has all operands from coverage report, but the expected answers are wrong. Used to run sail model to generate true results.
Add import info in scripts/lib.py.

Add configuration

Add vsew<vsew>_lmul<lmul> and vsew<vsew>_lmul<lmul>_nofail folders in env/macros just like the current directory. You can copy .h from any other configuration folder.
If you modify vsew:
1. Replace old vsew to new vsew globally. For example, search and replace 16 to 32 globally.
2. Replace floating load instruction globally. For example vsew=64, search and replace 1) flw to fld, 2) .word to .dword, 3) feq.s to feq.d 4) offset of fld like fld f2, 8(a0); to fld f2, 16(a0); globally.
3. Modify first part(below) carefully.

// VSEW temporarily hard-coded to 16 bits
#define RVTEST_VSET vsetivli x31, 1, e16, tu, mu;
#define __riscv_vsew 16
......
#define VSET_DOUBLE_VSEW vsetivli x31, 1, e32, tu, mu;
#define VSET_CONST_VSEW(eew_num) vsetivli x31, ##eew_num, m4, tu, mu;

If you modify lmul:
1. Replace old m<lmul> to new m<lmul>. For example, search and replace m1 to m8 globally.

CheckList

"-" : Not test/support yet
"P" : Test file is correct(pass Spike), but not fully cover val_comb
"P P" : Test file is correct and fully cover val_comb (rd, rs may not fully cover)
- *Now cover over 95% coverage points is regarded as pass, remaining need some fine tune
"/" : Configuration not support for this instruction(no need to test)
Not listed instruction are not tested yet

Mask

vmand, vmandnot, vmnand, vmor, vmornot, vmxnor, vmxor; vmsbf; vpopc, vfirst; vid, viota

Config	Status	Config	Status	Config	Status
vlen128 vsew8 lmul1	P P	vlen128 vsew8 lmul2-8	P P	vlen128 vsew8 lmul0.125-0.5	P P
vlen128 vsew16 lmul1	P P	vlen128 vsew16 lmul2-8	P P	vlen128 vsew16 lmul0.125-0.5	P P
vlen128 vsew32 lmul1(default)	P P	vlen128 vsew32 lmul2-8	P P	vlen128 vsew32 lmul0.125-0.5	P P
vlen128 vsew64 lmul1	P P	vlen128 vsew64 lmul2-8	P P	vlen128 vsew64 lmul0.125-0.5	P P
vlen256 vsew8 lmul1	P P	vlen256 vsew8 lmul2-8	P P	vlen256 vsew8 lmul0.125-0.5	P P
vlen256 vsew16 lmul1	P P	vlen256 vsew16 lmul2-8	P P	vlen256 vsew16 lmul0.125-0.5	P P
vlen256 vsew32 lmul1	P P	vlen256 vsew32 lmul2-8	P P	vlen256 vsew32 lmul0.125-0.5	P P
vlen256 vsew64 lmul1	P P	vlen256 vsew64 lmul2-8	P P	vlen256 vsew64 lmul0.125-0.5	P P
vlen512 vsew8 lmul1	P P	vlen512 vsew8 lmul2-8	P P	vlen512 vsew8 lmul0.125-0.5	P P
vlen512 vsew16 lmul1	P P	vlen512 vsew16 lmul2-8	P P	vlen512 vsew16 lmul0.125-0.5	P P
vlen512 vsew32 lmul1	P P	vlen512 vsew32 lmul2-8	P P	vlen512 vsew32 lmul0.125-0.5	P P
vlen512 vsew64 lmul1	P P	vlen512 vsew64 lmul2-8	P P	vlen512 vsew64 lmul0.125-0.5	P P
vlen1024 vsew8 lmul1	P P	vlen1024 vsew8 lmul2-8	P P	vlen1024 vsew8 lmul0.125-0.5	P P
vlen1024 vsew16 lmul1	P P	vlen1024 vsew16 lmul2-8	P P	vlen1024 vsew16 lmul0.125-0.5	P P
vlen1024 vsew32 lmul1	P P	vlen1024 vsew32 lmul2-8	P P	vlen1024 vsew32 lmul0.125-0.5	P P
vlen1024 vsew64 lmul1	P P	vlen1024 vsew64 lmul2-8	P P	vlen1024 vsew64 lmul0.125-0.5	P P
vlen2048 vsew8 lmul1	P P	vlen2048 vsew8 lmul2-8	P P	vlen2048 vsew8 lmul0.125-0.5	P P
vlen2048 vsew16 lmul1	P P	vlen2048 vsew16 lmul2-8	P P	vlen2048 vsew16 lmul0.125-0.5	P P
vlen2048 vsew32 lmul1	P P	vlen2048 vsew32 lmul2-8	P P	vlen2048 vsew32 lmul0.125-0.5	P P
vlen2048 vsew64 lmul1	P P	vlen2048 vsew64 lmul2-8	P P	vlen2048 vsew64 lmul0.125-0.5	P P
⚠️ When lmul is equal to 8, val_comb may not fully cover because isac counts the result of only one register at present. (RISCV-ISAC-RVV bug)

Permute

⚠️ vslide, vcompress will generate too many tests if num_elems (vlen*lmul/vsew) is too big. Test them will take a long time (longer than 20min)

vcompress, vmre, vmv, vfmv, vrgather, vrgatherei16, vslide, vslide1, vfslide

Config	vcompress	vmre	vmv	vfmv	vrgather	vrgatherei16	vslide	vslide1	vfslide
vlen128 vsew8 lmul1	P	P	P	/	P	P	P	P	/
vlen128 vsew16 lmul1	P	P	P	/	P	P P	P	P	/
vlen128 vsew32 lmul1(default)	P P	P	P P	P P	P P	P P	P P	P	P
vlen128 vsew64 lmul1	P P	P	P P	P P	P P	P P	P P	P P	P
vlen256 vsew8 lmul1	P	P	P	/	P	P	P	P	/
vlen256 vsew16 lmul1	P	P	P	/	P	P P	P	P	/
vlen256 vsew32 lmul1	P P	P	P P	P P	P P	P P	P	P	P
vlen256 vsew64 lmul1	P P	P	P P	P P	P P	P P	P P	P	P
vlen512 vsew8 lmul1	P	P	P	/	P	P	P	P	/
vlen512 vsew16 lmul1	P	P	P	/	P	P	P	P	/
vlen512 vsew32 lmul1	P P	P	P P	P P	P P	P P	P	P	P
vlen512 vsew64 lmul1	P P	P	P P	P P	P P	P P	P	P	P
vlen1024 vsew8 lmul1	P	P	P P	/	P	P	P	P	/
vlen1024 vsew16 lmul1	P	P	P	/	P	P P	P P	P	/
vlen1024 vsew32 lmul1	P	P	P P	P P	P P	P P	P	P	P
vlen1024 vsew64 lmul1	P P	P	P P	P P	P P	P P	P P	P	P
-----------------------------	---------	----	---	----	--------	------------	------	-------	-------
vlen128-1024 vsew8 lmul2-8	P P	P	P P	P P	P P	P	P	P	P
vlen128-1024 vsew16 lmul2-8	P P	P	P P	P P	P P	P	P	P	P
vlen128-1024 vsew32 lmul2-8	P P	P	P P	P P	P P	P	P	P	P
vlen128-1024 vsew64 lmul2-8	P P	P	P P	P P	P P	P	P	P	P
-----------------------------	---------	----	---	----	--------	------------	------	-------	-------
vlen128-1024 vsew8 lmul0.125-0.5	P P	P	P P	P P	P P	P	P	P	P
vlen128-1024 vsew16 lmul0.125-0.5	P P	P	P P	P P	P P	P	P	P	P
vlen128-1024 vsew32 lmul0.125-0.5	P P	P	P P	P P	P P	P	P	P	P
vlen128-1024 vsew64 lmul0.125-0.5	P P	P	P P	P P	P P	P	P	P	P
note:

vmv

vsew32 requires rs1val_walking_vector_unsgn
vsew64 requires rs1val_walking_vector

Integer

Simple Arithmetic: vadc, vsbc; vadd, vand, vdiv, vdivu, vmul, vmulh, vmulhsu, vmulhu, vsll, vsra, vsrl, vsub, vxor, vrem, vremu, vrsub, vsadd, vsaddu, vssub, vssubu, vmax, vmaxu, vmin, vminu,

Config	Status	Config	Status	Config	Status
vlen128 vsew8 lmul1	P P	vlen128 vsew8 lmul2-8	P P	vlen128 vsew8 lmul0.125-0.5	P P
vlen128 vsew16 lmul1	P P	vlen128 vsew16 lmul2-8	P P	vlen128 vsew16 lmul0.125-0.5	P P
vlen128 vsew32 lmul1(default)	P P	vlen128 vsew32 lmul2-8	P P	vlen128 vsew32 lmul0.125-0.5	P P
vlen128 vsew64 lmul1	P P	vlen128 vsew64 lmul2-8	P P	vlen128 vsew64 lmul0.125-0.5	P P
vlen256 vsew8 lmul1	P P	vlen256 vsew8 lmul2-8	P P	vlen256 vsew8 lmul0.125-0.5	P P
vlen256 vsew16 lmul1	P P	vlen256 vsew16 lmul2-8	P P	vlen256 vsew16 lmul0.125-0.5	P P
vlen256 vsew32 lmul1	P P	vlen256 vsew32 lmul2-8	P P	vlen256 vsew32 lmul0.125-0.5	P P
vlen256 vsew64 lmul1	P P	vlen256 vsew64 lmul2-8	P P	vlen256 vsew64 lmul0.125-0.5	P P
vlen512 vsew8 lmul1	P P	vlen512 vsew8 lmul2-8	P P	vlen512 vsew8 lmul0.125-0.5	P P
vlen512 vsew16 lmul1	P P	vlen512 vsew16 lmul2-8	P P	vlen512 vsew16 lmul0.125-0.5	P P
vlen512 vsew32 lmul1	P P	vlen512 vsew32 lmul2-8	P P	vlen512 vsew32 lmul0.125-0.5	P P
vlen512 vsew64 lmul1	P P	vlen512 vsew64 lmul2-8	P P	vlen512 vsew64 lmul0.125-0.5	P P
vlen1024 vsew8 lmul1	P P	vlen1024 vsew8 lmul2-8	P P	vlen1024 vsew8 lmul0.125-0.5	P P
vlen1024 vsew16 lmul1	P P	vlen1024 vsew16 lmul2-8	P P	vlen1024 vsew16 lmul0.125-0.5	P P
vlen1024 vsew32 lmul1	P P	vlen1024 vsew32 lmuL2-8	P P	vlen1024 vsew32 lmul0.125-0.5	P P
vlen1024 vsew64 lmul1	P P	vlen1024 vsew64 lmul2-8	P P	vlen1024 vsew64 lmul0.125-0.5	P P

Multiply-Add & Add-with-Carry Subtract-with-Borrow & Comparison: vmacc, vmadd, vnmsac, vnmsub & vmadc, vmsbc & vmseq, vmsgt, vmsgtu, vmsle, vmsleu, vmslt, vmsltu, vmsne

Config	Status	Config	Status	Config	Status
vlen128 vsew8 lmul1	P P	vlen128 vsew8 lmul2-8	P P	vlen128 vsew8 lmul0.125-0.5	P P
vlen128 vsew16 lmul1	P P	vlen128 vsew16 lmul2-8	P P	vlen128 vsew16 lmul0.125-0.5	P P
vlen128 vsew32 lmul1(default)	P P	vlen128 vsew32 lmul2-8	P P	vlen128 vsew32 lmul0.125-0.5	P P
vlen128 vsew64 lmul1	P P	vlen128 vsew64 lmul2-8	P P	vlen128 vsew64 lmul0.125-0.5	P P
vlen256 vsew8 lmul1	P P	vlen256 vsew8 lmul2-8	P P	vlen256 vsew8 lmul0.125-0.5	P P
vlen256 vsew16 lmul1	P P	vlen256 vsew16 lmul2-8	P P	vlen256 vsew16 lmul0.125-0.5	P P
vlen256 vsew32 lmul1	P P	vlen256 vsew32 lmul2-8	P P	vlen256 vsew32 lmul0.125-0.5	P P
vlen256 vsew64 lmul1	P P	vlen256 vsew64 lmul2-8	P P	vlen256 vsew64 lmul0.125-0.5	P P
vlen512 vsew8 lmul1	P P	vlen512 vsew8 lmul2-8	P P	vlen512 vsew8 lmul0.125-0.5	P P
vlen512 vsew16 lmul1	P P	vlen512 vsew16 lmul2-8	P P	vlen512 vsew16 lmul0.125-0.5	P P
vlen512 vsew32 lmul1	P P	vlen512 vsew32 lmul2-8	P P	vlen512 vsew32 lmul0.125-0.5	P P
vlen512 vsew64 lmul1	P P	vlen512 vsew64 lmul2-8	P P	vlen512 vsew64 lmul0.125-0.5	P P
vlen1024 vsew8 lmul1	P P	vlen1024 vsew8 lmul2-8	P P	vlen1024 vsew8 lmul0.125-0.5	P P
vlen1024 vsew16 lmul1	P P	vlen1024 vsew16 lmul2-8	P P	vlen1024 vsew16 lmul0.125-0.5	P P
vlen1024 vsew32 lmul1	P P	vlen1024 vsew32 lmul2-8	P P	vlen1024 vsew32 lmul0.125-0.5	P P
vlen1024 vsew64 lmul1	P P	vlen1024 vsew64 lmul2-8	P P	vlen1024 vsew64 lmul0.125-0.5	P P

Reduction Arithmetic: vor; vredand, vredmax, vredmaxu, vredmin, vredminu, vredor, vredsum, vredxor;

Config	Status	Config	Status	Config	Status
vlen128 vsew8 lmul1	P P	vlen128 vsew8 lmul2-8	P P	vlen128 vsew8 lmul0.125-0.5	P P
vlen128 vsew16 lmul1	P P	vlen128 vsew16 lmul2-8	P P	vlen128 vsew16 lmul0.125-0.5	P P
vlen128 vsew32 lmul1(default)	P P	vlen128 vsew32 lmul2-8	P P	vlen128 vsew32 lmul0.125-0.5	P P
vlen128 vsew64 lmul1	P P	vlen128 vsew64 lmul2-8	P P	vlen128 vsew64 lmul0.125-0.5	P P
vlen256 vsew8 lmul1	P P	vlen256 vsew8 lmul2-8	P P	vlen256 vsew8 lmul0.125-0.5	P P
vlen256 vsew16 lmul1	P P	vlen256 vsew16 lmul2-8	P P	vlen256 vsew16 lmul0.125-0.5	P P
vlen256 vsew32 lmul1	P P	vlen256 vsew32 lmul2-8	P P	vlen256 vsew32 lmul0.125-0.5	P P
vlen256 vsew64 lmul1	P P	vlen256 vsew64 lmul2-8	P P	vlen256 vsew64 lmul0.125-0.5	P P
vlen512 vsew8 lmul1	P P	vlen512 vsew8 lmul2-8	P P	vlen512 vsew8 lmul0.125-0.5	P P
vlen512 vsew16 lmul1	P P	vlen512 vsew16 lmul2-8	P P	vlen512 vsew16 lmul0.125-0.5	P P
vlen512 vsew32 lmul1	P P	vlen512 vsew32 lmul2-8	P P	vlen512 vsew32 lmul0.125-0.5	P P
vlen512 vsew64 lmul1	P P	vlen512 vsew64 lmul2-8	P P	vlen512 vsew64 lmul0.125-0.5	P P
vlen1024 vsew8 lmul1	P P	vlen1024 vsew8 lmul2-8	P P	vlen1024 vsew8 lmul0.125-0.5	P P
vlen1024 vsew16 lmul1	P P	vlen1024 vsew16 lmul2-8	P P	vlen1024 vsew16 lmul0.125-0.5	P P
vlen1024 vsew32 lmul1	P P	vlen1024 vsew32 lmul2-8	P P	vlen1024 vsew32 lmul0.125-0.5	P P
vlen1024 vsew64 lmul1	P P	vlen1024 vsew64 lmul2-8	P P	vlen1024 vsew64 lmul0.125-0.5	P P

Widen Arithmetic: vwadd, vwaddu, vwmacc, vwmaccsu, vwmaccu, vwmaccus; vwmul, vwmulsu, vwmulu, vwredsum, vwredsumu, vwsub, vwsubu; vnsra, vnsrl;

widen and narrow instruction not support VSEW=64 (vsew should be <= 64)
widen instruction(except VWRED*) not support LMUL=8 (lmul should be <= 4), lmul8 in table only means VWRED* pass

Config	Status	Config	Status	Config	Status
vlen128 vsew8 lmul1	P P	vlen128 vsew8 lmul2-8	P P	vlen128 vsew8 lmul0.125-0.5	P P
vlen128 vsew16 lmul1	P P	vlen128 vsew16 lmul2-8	P P	vlen128 vsew16 lmul0.125-0.5	P P
vlen128 vsew32 lmul1(default)	P P	vlen128 vsew32 lmul2-8	P P	vlen128 vsew32 lmul0.125-0.5	P P
vlen256 vsew8 lmul1	P P	vlen256 vsew8 lmul2-8	P P	vlen256 vsew8 lmul0.125-0.5	P P
vlen256 vsew16 lmul1	P P	vlen256 vsew16 lmul2-8	P P	vlen256 vsew16 lmul0.125-0.5	P P
vlen256 vsew32 lmul1	P P	vlen256 vsew32 lmul2-8	P P	vlen256 vsew32 lmul0.125-0.5	P P
vlen512 vsew8 lmul1	P P	vlen512 vsew8 lmul2-8	P P	vlen512 vsew8 lmul0.125-0.5	P P
vlen512 vsew16 lmul1	P P	vlen512 vsew16 lmul2-8	P P	vlen512 vsew16 lmul0.125-0.5	P P
vlen512 vsew32 lmul1	P P	vlen512 vsew32 lmul2-8	P P	vlen512 vsew32 lmul0.125-0.5	P P
vlen1024 vsew8 lmul1	P P	vlen1024 vsew8 lmul2-8	P P	vlen1024 vsew8 lmul0.125-0.5	P P
vlen1024 vsew16 lmul1	P P	vlen1024 vsew16 lmul2-8	P P	vlen1024 vsew16 lmul0.125-0.5	P P
vlen1024 vsew32 lmul1	P P	vlen1024 vsew32 lmul2-8	P P	vlen1024 vsew32 lmul0.125-0.5	P P

Floating Points

vfadd, vfclass, vfcvt, vfdiv, vfmacc, vfmax, vfmin, vfmsac, vfmsub, vfmul, `vfncvt`, vfnmacc, vfnmadd, vfnmsac, vfnmsub, vfrdiv, vfrec7, vfredmax, vfredmin

Config	vfadd	vfclass	vfcvt	vfdiv	vfmacc	vfmadd	vfmax	vfmin	vfmsac	vfmsub	vfmul	`vfncvt`	vfnmacc	vfnmadd	vfnmsac	vfnmsub	vfrdiv	vfrec7	vfredmax	vfredmin
vlen128 vsew32 lmul1(default)	P P	P P	-	P P	P P	P P	P P	P P	P P	P P	P P	-	P P	P P	P P	P P	P P	P P	P P	P P
vlen128 vsew64 lmul1	P P	P P	-	P P	P P	P P	P P	P P	P P	P P	P P	-	P P	P P	P P	P P	P P	P P	P P	P P
vlen256 vsew32 lmul1	P P	P P	-	P P	P P	P P	P P	P P	P P	P P	P P	-	P P	P P	P P	P P	P P	P P	P P	P P
vlen256 vsew64 lmul1	P P	P P	-	P P	P P	P P	P P	P P	P P	P P	P P	-	P P	P P	P P	P P	P P	P P	P P	P P
vlen512 vsew32 lmul1	P P	P P	-	P P	P P	P P	P P	P P	P P	P P	P P	-	P P	P P	P P	P P	P P	P P	P P	P P
vlen512 vsew64 lmul1	P P	P P	-	P P	P P	P P	P P	P P	P P	P P	P P	-	P P	P P	P P	P P	P P	P P	P P	P P
vlen1024 vsew32 lmul1	P P	P P	-	P P	P P	P P	P P	P P	P P	P P	P P	-	P P	P P	P P	P P	P P	P P	P P	P P
vlen1024 vsew64 lmul1	P P	P P	-	P P	P P	P P	P P	P P	P P	P P	P P	-	P P	P P	P P	P P	P P	P P	P P	P P
-----------------------------	-----	------	---------	-----	------	------	-----	-----	------	------	-----	----------	-------	-------	-------	-------	------	------	--------	--------
vlen128-1024 vsew32 lmul2-8	P	P	-	P	P	P	P	P	P	P	P	-	P	P	P	P	P	P	P	P
vlen128-1024 vsew64 lmul2-8	P	P	-	P	P	P	P	P	P	P	P	-	P	P	P	P	P	P	P	P
-----------------------------	-----	------	---------	-----	------	------	-----	-----	------	------	-----	----------	-------	-------	-------	-------	------	------	--------	--------
vlen128-1024 vsew32 lmul0.125-0.5	P	P	-	P	P	P	P	P	P	P	P	-	P	P	P	P	P	P	P	P
vlen128-1024 vsew64 lmul0.125-0.5	P	P	-	P	P	P	P	P	P	P	P	-	P	P	P	P	P	P	P	P
-----------------------------	-----	------	---------	-----	------	------	-----	-----	------	------	-----	----------	-------	-------	-------	-------	------	------	--------	--------
vlen1024-4096 vsew32 lmul0.125-8	P P	P P	-	P P	P P	P P	P P	P P	P P	P P	P P	-	P P	P P	P P	P P	P P	P P	P P	P P
vlen1024-4096 vsew64 lmul0.125-8	P P	P P	-	P P	P P	P P	P P	P P	P P	P P	P P	-	P P	P P	P P	P P	P P	P P	P P	P P

vfcvt, vfcvt.x.f.v, vfcvt.rtz.xu.f.v, etc. not support in Spike.

vfredosum, vfredusum, vfrsqrt7, vfrsub, vfsgnj, vfsgnjn, vfsgnjx, vfsqrt, vfsub

Config	vfredosum	vfredusum	vfrsqrt7	vfrsub	vfsgnj	vfsgnjn	vfsgnjx	vfsqrt	vfsub
vlen128 vsew32 lmul1(default)	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen128 vsew64 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen256 vsew32 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen256 vsew64 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen512 vsew32 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen512 vsew64 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen1024 vsew32 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen1024 vsew64 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
-----------------------------	---------	---------	--------	------	------	-------	-------	------	-----
vlen128-1024 vsew32 lmul2-8	P	P	P	P	P	P	P	P	P
vlen128-1024 vsew64 lmul2-8	P	P	P	P	P	P	P	P	P
-----------------------------	---------	---------	--------	------	------	-------	-------	------	-----
vlen1024-4096 vsew32 lmul0.125-8	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen1024-4096 vsew64 lmul0.125-8	P P	P P	P P	P P	P P	P P	P P	P P	P P

vfwadd, vfwcvt, vfwmacc, vfwmsac, vfwmul, vfwnmacc, vfwnmsac, vfwredsum, vfwsub

Config	vfwadd	vfwcvt	vfwmacc	vfwmsac	vfwmul	vfwnmacc	vfwnmsac	vfwredsum	vfwsub
vlen128 vsew16 lmul1	/	/	/	/	/	/	/	/	/
vlen128 vsew32 lmul1(default)	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen256 vsew16 lmul1	/	/	/	/	/	/	/	/	/
vlen256 vsew32 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen512 vsew16 lmul1	/	/	/	/	/	/	/	/	/
vlen512 vsew32 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen1024 vsew16 lmul1	/	/	/	/	/	/	/	/	/
vlen1024 vsew32 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
-----------------------------	------	------	-------	-------	------	--------	--------	---------	------
vlen128 vsew16 lmul2-8	/	/	/	/	/	/	/	/	/
vlen128 vsew32 lmul2-8	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen256 vsew16 lmul2-8	/	/	/	/	/	/	/	/	/
vlen256 vsew32 lmul2-8	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen512 vsew16 lmul2-8	/	/	/	/	/	/	/	/	/
vlen512 vsew32 lmul2-8	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen1024 vsew16 lmul2-8	/	/	/	/	/	/	/	/	/
vlen1024 vsew32 lmul2-8	P P	P P	P P	P P	P P	P P	P P	P P	P P
-----------------------------	------	------	-------	-------	------	--------	--------	---------	------
vlen128 vsew16 lmul0.125-0.5	/	/	/	/	/	/	/	/	/
vlen128 vsew32 lmul0.125-0.5	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen256 vsew16 lmul0.125-0.5	/	/	/	/	/	/	/	/	/
vlen256 vsew32 lmul0.125-0.5	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen512 vsew16 lmul0.125-0.5	/	/	/	/	/	/	/	/	/
vlen512 vsew32 lmul0.125-0.5	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen1024 vsew16 lmul0.125-0.5	/	/	/	/	/	/	/	/	/
vlen1024 vsew32 lmul0.125-0.5	P P	P P	P P	P P	P P	P P	P P	P P	P P
note:

we do not have 16-bit floating point dataset
VFWiden instructions require lmul <= 4

Fix Points

	vaadd	vaaddu	vasub	vasubu	vnclip	vnclipu	vsmul	vssra	vssrl
vlen128 vsew8 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen128 vsew16 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen128 vsew32 lmul1(default)	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen128 vsew64 lmul1	P P	P P	P P	P P	/	/	P P	P P	P P
vlen256 vsew8 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen256 vsew16 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen256 vsew32 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen256 vsew64 lmul1	P P	P P	P P	P P	/	/	P P	P P	P P
vlen512 vsew8 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen512 vsew16 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen512 vsew32 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen512 vsew64 lmul1	P P	P P	P P	P P	/	/	P P	P P	P P
vlen1024 vsew8 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen1024 vsew16 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen1024 vsew32 lmul1	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen1024 vsew64 lmul1	P P	P P	P P	P P	/	/	P P	P P	P P
-----------------------------	-----	------	-----	------	------	-------	-----	-----	-----
vlen128-1024 vsew8 lmul2-8	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew16 lmul2-8	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew32 lmul2-8	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew64 lmul2-8	P P	P P	P P	P P	P P	P P	P P	P P	P P
-----------------------------	-----	------	-----	------	------	-------	-----	-----	-----
vlen128-1024 vsew8 lmul0.125-0.5	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew16 lmul0.125-0.5	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew32 lmul0.125-0.5	P P	P P	P P	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew64 lmul0.125-0.5	P P	P P	P P	P P	P P	P P	P P	P P	P P

*vnclip and vnclipu not support vsew=64 and lmul=8

LOAD

Unable to pass the lmul==8 parameter test： vlsege32 vlssege32 vluxsegei8 vluxsegei16 vluxsegei32
If the EMUL would be out of range (EMUL>8 or EMUL<1/8), the instruction encoding is reserved. Use / to represent.
except <eew> means this configuration not satisfy vemul >= 0.125 && vemul <= 8 or vsew <= elen * lmul

	vle8/16/32/64	vlre8/16/32	vlse8/16/32/64	vls(s)ege8/16/32	vluxei8/16/32	vluxeiseg8/16/32	note
vlen128-1024 vsew8 lmul1	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew16 lmul1	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew32 lmul1(default)	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew64 lmul1	P P	P P	P P	P P	P P	P P
-----------------------------	-------------	-----------	-----------	----------------	-------------	----------------
vlen128-1024 vsew8 lmul0.125	P P	P P	P P	P P	P P	P P	vle, vlse, vls(s)ege, vlre except32; vle, vlse except 64;
vlen128-1024 vsew16 lmul0.125	P P	P P	P P	P P	P P	P P	vle, vlse, vls(s)ege, vlre except32; vle, vlse except 64; vluxseg except 8
vlen128-1024 vsew32 lmul0.125	/	/	/	/	/	/
vlen128-1024 vsew64 lmul0.125	/	/	/	/	/	/
-----------------------------	-------------	-----------	-----------	----------------	-------------	----------------
vlen128-1024 vsew8 lmul0.25	P P	P P	P P	P P	P P	P P	vle,vlse except 64
vlen128-1024 vsew16 lmul0.25	P P	P P	P P	P P	P P	P P	vle,vlse except 64
vlen128-1024 vsew32 lmul0.25	P P	P P	P P	P P	P P	P P	vle,vlse except 64; vluxseg except 8
vlen128-1024 vsew64 lmul0.25	/	/	/	/	/	/
-----------------------------	-------------	-----------	-----------	----------------	-------------	----------------
vlen128-1024 vsew8 lmul0.5	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew16 lmul0.5	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew32 lmul0.5	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew64 lmul0.5	P P	P P	P P	P P	P P	P P	vluxseg except 8
-----------------------------	-------------	-----------	-----------	----------------	-------------	----------------
vlen128-1024 vsew8 lmul2	P P	P P	P P	P P	P P	x
vlen128-1024 vsew16 lmul2	P P	P P	P P	P P	P P	x
vlen128-1024 vsew32 lmul2	P P	P P	P P	P P	P P	x
vlen128-1024 vsew64 lmul2	P P	P P	P P	P P	P P	x
-----------------------------	-------------	-----------	-----------	----------------	-------------	----------------
vlen128-1024 vsew8 lmul4	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew16 lmul4	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew32 lmul4	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew64 lmul4	P P	P P	P P	P P	P P	P P
-----------------------------	-------------	-----------	-----------	----------------	-------------	----------------
vlen128-1024 vsew8 lmul8	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew16 lmul8	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew32 lmul8	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew64 lmul8	P P	P P	P P	P P	P P	P P

Store

	vse8/16/32/64	vs1/2/4/8r	vsse8/16/32	vss(s)ege8/16/32	vsuxei8/16/32	vsuxeiseg8/16/32	note
vlen128-1024 vsew8 lmul1	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew16 lmul1	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew32 lmul1(default)	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew64 lmul1	P P	P P	P P	P P	P P	P P
-----------------------------	-------------	-----------	-----------	----------------	-------------	----------------
vlen128-1024 vsew8 lmul0.125	P P	/	P P	P P	P P	P P	vse, vss(s)ege except32;
vlen128-1024 vsew16 lmul0.125	P P	/	P P	P P	P P	P P	vse, vss(s)ege except32;
vlen128-1024 vsew32 lmul0.125	/	/	/	/	/	/
vlen128-1024 vsew64 lmul0.125	/	/	/	/	/	/
-----------------------------	-------------	-----------	-----------	----------------	-------------	----------------
vlen128-1024 vsew8 lmul0.25	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew16 lmul0.25	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew32 lmul0.25	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew64 lmul0.25	/	/	/	/	/	/
-----------------------------	-------------	-----------	-----------	----------------	-------------	----------------
vlen128-1024 vsew8 lmul0.5	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew16 lmul0.5	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew32 lmul0.5	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew64 lmul0.5	P P	P P	P P	P P	P P	P P
-----------------------------	-------------	-----------	-----------	----------------	-------------	----------------
vlen128-1024 vsew8 lmul2	P P	P P	P P	P P	P P	x
vlen128-1024 vsew16 lmul2	P P	P P	P P	P P	P P	x
vlen128-1024 vsew32 lmul2	P P	P P	P P	P P	P P	x
vlen128-1024 vsew64 lmul2	P P	P P	P P	P P	P P	x
-----------------------------	-------------	-----------	-----------	----------------	-------------	----------------
vlen128-1024 vsew8 lmul4	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew16 lmul4	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew32 lmul4	P P	P P	P P	P P	P P	P P
vlen128-1024 vsew64 lmul4	P P	P P	P P	P P	P P	P P
-----------------------------	-------------	-----------	-----------	----------------	-------------	----------------
vlen128-1024 vsew8 lmul8	P P	x	P P	P P	P P	P P
vlen128-1024 vsew16 lmul8	P P	x	P P	P P	P P	P P
vlen128-1024 vsew32 lmul8	P P	x	P P	P P	P P	P P
vlen128-1024 vsew64 lmul8	P P	x	P P	P P	P P	P P

Code Explain

Register Alignment

For example, require_noover(insn.rd(), rd_lmul, insn.rs1(), rs1_lmul) require rs1 and rd will not overwrite each other
- no overlap condition is either rightest rs1 is smaller than rd or rightest rd is smaller than rs1
- ((rs1 + rs1_lmul - 1 < rd) or (rd + rd_lmul - 1 < rs1))
- There are many conditions like this in generating macros for different registers.

Fill Empty Tests

def print_ending(f):
    print("  RVTEST_SIGBASE( x20,signature_x20_2)\n\
    \n\
    TEST_VV_OP(32766, vadd.vv, 2, 1, 1)\n\
    TEST_PASSFAIL\n\
    #endif\n\

The TEST_VV_OP(32766, vadd.vv, 2, 1, 1) is in order to fill those tests that don't have tests(For example, when emul > 8, load/store will not have tests) with one vadd test.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

RISC-V Vector Automatic Tests Generator

Prerequisite

Usage

Generate One Instruction

Generate All Tests

Check All Generated Tests

Support Configuration

Notice of Failure Tests

Develop

Add a instruction

Add configuration

CheckList

Mask

vmand, vmandnot, vmnand, vmor, vmornot, vmxnor, vmxor; vmsbf; vpopc, vfirst; vid, viota

Permute

vcompress, vmre, vmv, vfmv, vrgather, vrgatherei16, vslide, vslide1, vfslide

Integer

Simple Arithmetic: vadc, vsbc; vadd, vand, vdiv, vdivu, vmul, vmulh, vmulhsu, vmulhu, vsll, vsra, vsrl, vsub, vxor, vrem, vremu, vrsub, vsadd, vsaddu, vssub, vssubu, vmax, vmaxu, vmin, vminu,

Multiply-Add & Add-with-Carry Subtract-with-Borrow & Comparison: vmacc, vmadd, vnmsac, vnmsub & vmadc, vmsbc & vmseq, vmsgt, vmsgtu, vmsle, vmsleu, vmslt, vmsltu, vmsne

Reduction Arithmetic: vor; vredand, vredmax, vredmaxu, vredmin, vredminu, vredor, vredsum, vredxor;

Widen Arithmetic: vwadd, vwaddu, vwmacc, vwmaccsu, vwmaccu, vwmaccus; vwmul, vwmulsu, vwmulu, vwredsum, vwredsumu, vwsub, vwsubu; vnsra, vnsrl;

Floating Points

vfadd, vfclass, vfcvt, vfdiv, vfmacc, vfmax, vfmin, vfmsac, vfmsub, vfmul, `vfncvt`, vfnmacc, vfnmadd, vfnmsac, vfnmsub, vfrdiv, vfrec7, vfredmax, vfredmin

vfredosum, vfredusum, vfrsqrt7, vfrsub, vfsgnj, vfsgnjn, vfsgnjx, vfsqrt, vfsub

vfwadd, vfwcvt, vfwmacc, vfwmsac, vfwmul, vfwnmacc, vfwnmsac, vfwredsum, vfwsub

Fix Points

LOAD

Store

Code Explain

Register Alignment

Fill Empty Tests

About

Releases

Packages

Contributors 8

Languages

Name		Name	Last commit message	Last commit date
Latest commit History 126 Commits
cgfs		cgfs
env		env
scripts		scripts
.gitignore		.gitignore
README.md		README.md
__init__.py		__init__.py
check_all.py		check_all.py
generate_all.py		generate_all.py
run.py		run.py

B1ACK917/rvv-atg

Folders and files

Latest commit

History

Repository files navigation

RISC-V Vector Automatic Tests Generator

Prerequisite

Usage

Generate One Instruction

Generate All Tests

Check All Generated Tests

Support Configuration

Notice of Failure Tests

Develop

Add a instruction

Add configuration

CheckList

Mask

vmand, vmandnot, vmnand, vmor, vmornot, vmxnor, vmxor; vmsbf; vpopc, vfirst; vid, viota

Permute

vcompress, vmre, vmv, vfmv, vrgather, vrgatherei16, vslide, vslide1, vfslide

Integer

Simple Arithmetic: vadc, vsbc; vadd, vand, vdiv, vdivu, vmul, vmulh, vmulhsu, vmulhu, vsll, vsra, vsrl, vsub, vxor, vrem, vremu, vrsub, vsadd, vsaddu, vssub, vssubu, vmax, vmaxu, vmin, vminu,

Multiply-Add & Add-with-Carry Subtract-with-Borrow & Comparison: vmacc, vmadd, vnmsac, vnmsub & vmadc, vmsbc & vmseq, vmsgt, vmsgtu, vmsle, vmsleu, vmslt, vmsltu, vmsne

Reduction Arithmetic: vor; vredand, vredmax, vredmaxu, vredmin, vredminu, vredor, vredsum, vredxor;

Widen Arithmetic: vwadd, vwaddu, vwmacc, vwmaccsu, vwmaccu, vwmaccus; vwmul, vwmulsu, vwmulu, vwredsum, vwredsumu, vwsub, vwsubu; vnsra, vnsrl;

Floating Points

vfadd, vfclass, vfcvt, vfdiv, vfmacc, vfmax, vfmin, vfmsac, vfmsub, vfmul, vfncvt, vfnmacc, vfnmadd, vfnmsac, vfnmsub, vfrdiv, vfrec7, vfredmax, vfredmin

vfredosum, vfredusum, vfrsqrt7, vfrsub, vfsgnj, vfsgnjn, vfsgnjx, vfsqrt, vfsub

vfwadd, vfwcvt, vfwmacc, vfwmsac, vfwmul, vfwnmacc, vfwnmsac, vfwredsum, vfwsub

Fix Points

LOAD

Store

Code Explain

Register Alignment

Fill Empty Tests

About

Resources

Stars

Watchers

Forks

Releases

Packages 0

Contributors 8

Languages

vfadd, vfclass, vfcvt, vfdiv, vfmacc, vfmax, vfmin, vfmsac, vfmsub, vfmul, `vfncvt`, vfnmacc, vfnmadd, vfnmsac, vfnmsub, vfrdiv, vfrec7, vfredmax, vfredmin

Packages