[RISC-V] some vector schedules cause LLVM error

[miles-rusch-berkeley]

matrix_multiplication.cpp.zip

the following vectorization schedule, defined in line 47, causes an error when compiling to RISCv Vector but not when compiling to ARM SVE:

//ISSUE: no llvm error if we replace y with x in the following line:
out.vectorize(y, vlen/arch_bits);

[miles-rusch-berkeley]

The error message is:

Intrinsic name not mangled correctly for type arguments! Should be: llvm.experimental.vp.strided.store.nxv4f32.p0.i64 ptr @llvm.experimental.vp.strided.store.nxv4f32.p0.v32781f64 Internal Error at /Users/milesrusch/Documents/risc-v/rvv/llvm15/Halide/src/CodeGen_LLVM.cpp:632 triggered by user code at : Condition failed: !verifyFunction(*function, &llvm::errs()):

matrix_multiplication.cpp.zip

[zvookin]

Fixed in #7212 .

[miles-rusch-berkeley]

After pulling main with the patch, I am now getting a similar error:

% g++ gaussian_blur_gen.cpp -g -std=c++17 -I includeHalide -L libHalide -lHalide -o gaussian_blur_gen
% DYLD_LIBRARY_PATH=libHalide ./gaussian_blur_gen
...
Intrinsic name not mangled correctly for type arguments! Should be: llvm.vector.insert.nxv8i8.nxv4i8
ptr @llvm.vector.insert.nxv8i8.v32i8
...
Intrinsic name not mangled correctly for type arguments! Should be: llvm.vector.insert.nxv8i8.nxv4i8
ptr @llvm.vector.insert.nxv8i8.v32i8
Internal Error at /Users/milesrusch/Documents/risc-v/rvv/llvm15/Halide/src/CodeGen_LLVM.cpp:632 triggered by user code at :
Condition failed: !verifyFunction(*function, 
[gaussian_blur_gen.cpp.zip](https://github.com/halide/Halide/files/10192566/gaussian_blur_gen.cpp.zip)
&llvm::errs()):
zsh: abort      DYLD_LIBRARY_PATH=libHalide ./$1_gen

gaussian_blur_gen.cpp.zip