Why doesn't `vortex` diff `scalar` and `vector` register files?

[FlashBarryAllen]

Here's the log of warp register files which only has vector RF but hasn't scalar RF.
scalar RF means only warp level and vector RF means the threads in a warp level.

The log shows that warp has 4 threads. Each thread has its own 32@RFs. However I can't find any info about warp level RF shared by all threads.
Can anyone tell me why?

DEBUG Fetch: code=0xeddff0ef, cid=0, wid=0, tmask=1111, PC=0x80000694 (#3063)
DEBUG Instr: JAL x1, 0xfffffedc, cid=0, wid=0, tmask=1111, PC=0x80000694(#3063)
DEBUG Dest Reg: x1={0x80000698, 0x80000698, 0x80000698, 0x80000698}
DEBUG *** Next PC=0x80000570
DEBUG Register state:
  %r00: 0000000000000000  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r01: 0000000080000698  0000000080000698  0000000080000698  0000000080000698 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r02: 00000001fffeffd0  00000001fffee000  00000001fffec000  00000001fffea000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r03: 0000000080004800  0000000080004800  0000000080004800  0000000080004800 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r04: 0000000080004270  0000000080004298  00000000800042c0  00000000800042e8 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r05: 0000000000000003  0000000000000001  0000000000000001  0000000000000001 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r06: 0000000000000000  0000000000000028  0000000000000028  0000000000000028 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r07: 0000000000000000  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r08: 0000000000000004  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r09: 0000000000000000  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r10: 000000000000000f  0000000000000001  0000000000000002  0000000000000003 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r11: ffffffffffffffff  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r12: 0000000000000000  0000000000000001  0000000000000001  0000000000000001 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r13: ffffffffffffffff  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r14: 00000001fffeff98  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r15: 0000000000009041  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r16: 0000000000000000  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r17: 0000000080004268  0000000000000000  0000000000000000  0000000000000000 | 000000000_0000000  0000000000000000  0000000000000000  0000000000000000 
  %r18: 0000000000000000  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r19: 0000000000000000  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r20: 0000000000000000  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r21: 0000000000000000  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r22: 0000000000000000  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r23: 0000000000000000  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r24: 0000000000000000  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r25: 0000000000000000  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r26: 0000000000000000  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r27: 0000000000000000  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r28: 0000000000000000  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r29: 0000000000000001  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r30: 0000000000000018  0000000000000018  0000000000000018  0000000000000018 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000 
  %r31: 0000000000000000  0000000000000000  0000000000000000  0000000000000000 | 0000000000000000  0000000000000000  0000000000000000  0000000000000000

[FlashBarryAllen]

Vortex relied heavily on a unified VRF for SIMT model, resulting in low efficiency. However, modern commercial GPGPU architectures, such as AMD's GCN/RDNA and NVIDIA's architectures introduced the SRF, dynamically removing scalar data from expensive VRFs and storing it only once in the SRF. This significantly improved register file area, power, and performance. Why vortex doesn't introduce SRF?

[gsvgit]

@FlashBarryAllen I have no answer to your question, but looks like design of register files in Vortex is a topic for discussion, because there are other nontrivial issues with it (e.g. #251, #205)