master
Add attributes to functions pass (#539)
Add support for OpenCL 2.0 atomic functions (#657)
Support global type-mangled pod args in push constants (#588)
Embedded reflection (#612)
Use InsertSPIRVOp to create calls to barrier SPIR-V intrinsics (#389)
Lower kernel parameters (#694)
Handle sitofp with i1 arguments (#697)
Relax frontend 8- and 16-bit type check (#590)
Fix barrier and mem_fence scopes (#610)
Update LLVM (#751)
Fix loop merge and continue as same block (#485)
Add an LLVM IR library for more builtin functions (#715)
Builtins refactor - phase3 (#598)
Read write images (#643)
Remove the option to emit assembly (#320)
Automate pod arg implementation decision (#575)
Add support for llvm.{umax,umin} intrinsics (#747)
Support additional builtin functions (#732)
Normalize constant global variables (#366)
Fix how ArrayStride is generated for pointers (#640)
Fix crash in ReplacePointerBitcastPass (#707)
Test simplification (#434)
Remove the need for the type list (#745)
Update deps (#523)
Update llvm (#627)
Test simplification (#411)
Convert write_only images in CL2.0 and later for unspecialized images (#649)
SPIR-V 1.4 fixes (#755)
Add a pass to strip freeze instructions (#559)
Refactor checks about pointers in structs (#470)
Support atomic_work_item_fence (#703)
Handle more instcombine vector casts (#556)
SPIRVProducerPass refactor - phase2.2 (#586)
Lower OpenCL vloadN and vstoreN (#670)
Add support for warnings (#319)
Introduce stubs for long-vector support (#638)
Add support for llvm.{smax,smin} intrinsics (#735)
Don't buffer test output (#633)
Change default of -cluster-pod-kernel-args (#554)
Update LLVM (#736)
Fix casts from bool to half/double (#676)
Simplify tests to incfrease robustness (#398)
Make char support default on (#328)
Add convergent attribute in more cases (#658)
Prevent duplicate types and decorations (#654)
Support mixed reqd_work_group_size attributes (#630)
Update LLVM (#641)
Improve robustness of the IR emission test (#695)
Support fnegate (#443)
Disable PRE in GVN (#596)
More test check simplifications (#399)
Update LLVM (#474)
Specialize images types to support multiple sample types (#439)
Change selection merge generation (#378)
Add support for taking LLVM IR as input (#430)
Deprecate the sampler map (#453)
Support double-precision literals/constants (#329)
Add a tool to run individual clspv passes. (#283)
Add test output file for longvector-metadata.ll (#743)
Don't add int64 types due to literal operands (#498)
Add support for -cl-std=CL3.0 (#653)
Add a do-nothing implementation of the prefetch built-in function (#573)
Update LLVM (#713)
Add support for cl_arm_non_uniform_work_group_size (#720)
Initial support for non-uniform NDRanges (#572)
Only add used samplers to the descriptor map (#385)
More test simplification (#400)
Direct Resource Access optimization
Update LLVM (#737)
Update LLVM (#739)
Expand functionality to catch odd truncates (#585)