[type] [cuda] Support bit-level pointer on cuda backend #2065
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Hanke98
commented
Nov 27, 2020
taichi/codegen/codegen_llvm.cpp
Outdated
Comment on lines
1200
to
1225
| // Compute float(digits) * scale | ||
| llvm::Value *cast = nullptr; | ||
| auto compute_type = cft->get_compute_type()->as<PrimitiveType>(); | ||
| if (cft->get_digits_type()->cast<CustomIntType>()->get_is_signed()) { | ||
| cast = builder->CreateSIToFP(digits, llvm_type(compute_type)); | ||
| } else { | ||
| cast = builder->CreateUIToFP(digits, llvm_type(compute_type)); | ||
| } | ||
| llvm::Value *s = | ||
| llvm::ConstantFP::get(*llvm_context, llvm::APFloat(cft->get_scale())); | ||
| s = builder->CreateFPCast(s, llvm_type(compute_type)); | ||
| auto scaled = builder->CreateFMul(cast, s); | ||
| llvm_val[stmt] = scaled; | ||
| llvm_val[stmt] = restore_custom_float(digits, val_type); |
There was a problem hiding this comment.
Moved to restore_custom_float() to reuse.
There was a problem hiding this comment.
reconstruct_custom_float seems a better name
Comment on lines
+1170
to
1187
| llvm::Value *CodeGenLLVM::extract_custom_int(llvm::Value *physical_value, | ||
| llvm::Value *bit_offset, | ||
| Type *load_type) { | ||
| // bit shifting | ||
| // first left shift `physical_type - (offset + num_bits)` | ||
| // then right shift `physical_type - num_bits` | ||
| auto cit = load_type->as<CustomIntType>(); | ||
| auto bit_end = | ||
| builder->CreateAdd(bit_offset, tlctx->get_constant(cit->get_num_bits())); | ||
| auto left = builder->CreateSub( | ||
| tlctx->get_constant(data_type_bits(cit->get_physical_type())), bit_end); | ||
| auto right = builder->CreateSub( | ||
| tlctx->get_constant(data_type_bits(cit->get_physical_type())), | ||
| tlctx->get_constant(cit->get_num_bits())); | ||
| left = builder->CreateIntCast(left, bit_level_container->getType(), false); | ||
| right = builder->CreateIntCast(right, bit_level_container->getType(), false); | ||
| auto step1 = builder->CreateShl(bit_level_container, left); | ||
| left = builder->CreateIntCast(left, physical_value->getType(), false); | ||
| right = builder->CreateIntCast(right, physical_value->getType(), false); | ||
| auto step1 = builder->CreateShl(physical_value, left); | ||
| llvm::Value *step2 = nullptr; |
There was a problem hiding this comment.
This function is separated from load_as_custom_int with slight changes.
Hanke98
changed the title
yuanming-hu
reviewed
Nov 28, 2020
taichi/codegen/codegen_llvm.cpp
Outdated
Comment on lines
1200
to
1225
| // Compute float(digits) * scale | ||
| llvm::Value *cast = nullptr; | ||
| auto compute_type = cft->get_compute_type()->as<PrimitiveType>(); | ||
| if (cft->get_digits_type()->cast<CustomIntType>()->get_is_signed()) { | ||
| cast = builder->CreateSIToFP(digits, llvm_type(compute_type)); | ||
| } else { | ||
| cast = builder->CreateUIToFP(digits, llvm_type(compute_type)); | ||
| } | ||
| llvm::Value *s = | ||
| llvm::ConstantFP::get(*llvm_context, llvm::APFloat(cft->get_scale())); | ||
| s = builder->CreateFPCast(s, llvm_type(compute_type)); | ||
| auto scaled = builder->CreateFMul(cast, s); | ||
| llvm_val[stmt] = scaled; | ||
| llvm_val[stmt] = restore_custom_float(digits, val_type); |
There was a problem hiding this comment.
reconstruct_custom_float seems a better name
yuanming-hu
approved these changes
Nov 30, 2020
Co-authored-by: Yuanming Hu <[email protected]>
Merged
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Related issue = #1905
In this pr, we want to support bit-level pointer on coda backend.
For now, most test cases passed, but the CustomInt/Float type with 16-bits or 8-bits physical type would fail. The error occurs in JIT module, which I know very little about. @yuanming-hu could you please have a look? I will left the error message and reproducing code in JIRA. Thanks a lot!
[Click here for the format server]