It's a fix for Bug 33838 - Incorrect operand sizes for MMX instructions: I changed the description of the instrs in td file. When I was working with this issue I seriously changed the debug logging for asm-matcher but I decided to introduce those changes in a separate patch which will be published soon.
Details
Details
Diff Detail
Diff Detail
- Repository
- rL LLVM
Event Timeline
lib/Target/X86/X86InstrMMX.td | ||
---|---|---|
112 ↗ | (On Diff #128227) | Can we add a X86MemOperand parameter to MMXI_binop_rm_int that defaults to i64mem. Then just make MMXI_binop_rm32_int inherit from MMXI_binop_rm_int and pass the i32mem to that parameter. |
test/MC/X86/punpack-intel.s | ||
---|---|---|
1 ↗ | (On Diff #128227) | Just add these to intel-syntax.s and intel-syntax-error.s I don't think its worth a new test file. |
Comment Actions
I changed the implementation accodingly Craig's requiremetns: now it's shorter and more effective.
Comment Actions
Please remember to prefix phab subjects, in this case "[X86][MMX]" - and tag it with PR33838 as well:
"[X86][MMX] Incorrect operand sizes for some MMX instructions: punpcklwd, punpcklbw and punpckldq (PR33838)"
Repeat on the commit message as well - it makes email filtering a lot easier - cheers!