You can clone with
HTTPS or Subversion.
Fixed register passing for pinvoke.
Initial work on passing by register.
Fixed R4/R8 constant code for AOT compiling.
Fixed P/Invoke bug.
Changed ops that use static constants to load into scratch reg first,…
… since ALU ops expect 128-bit mem locations.
Last XMM reg is used for scratch space.
Updating fcall instruction lengths.
Fixed return value moves for R4 and R8.
Changed the MONO_ARCH_USE_SHARED_FP_SIMD_BANK symbol into a runtime f…
…lag that gets used in the appropriate methods to check for reg mirroring.
Marshaling FP return values from native/internal calls.
Updated get_call_info to take into account SSE registers.
Started working on method call support.
Fixed LCONV_TO_R4. All float tests now pass.
Adjusted SIMD ops to use SSE instead of FP stack ops when appropriate.
Stopped SIMD codepath from emitting specialized FCONV when FP-stack d…
Changed sqrt to use SSE instruction.
Implemented OP_FREM for SSE support.
Added mono_arch_compile_init for arch-specific initialization of Mono…
Updating STORER4 to go directly to register via cvtss2sd.
Fixed a bug in SSE FP-compares.
Implemented SSE sin/cos/sqrt.
Implemented LCONV_TO_R* opcodes.
Fixed more opcode lengths.
Implemented FNEG and ABS. SSE version no longer tries to special-case…
… Math Tan/Atan/Round methods.
Implemented CKFINITE, FPOP, and FXCH.
Fixed several opcode lengths in the MD file.
Added SSE arithmetic operations.
Implemented SSE float conversion ops.
Fixed a bug in movsd generation macro. Float tests all pass now.
Removed patching from FP constant opcodes.
Changed MONO_ARCH_USE_FPSTACK into a runtime flag and removed the pre…
Removed unused X86_FP_* ops and collect_fp_stack_space function.
Added SSE paths for FP compares and branches.