Fuse write back via VPM

[nomaddo]

In testing/bugs/68_remove_moves_of_r4.cl, fusion of write-back via VPM is partially successful but failed mostly.

$ ./build/VC4C --asm -o /tmp/hoge.s testing/bugs/68_remove_moves_of_r4.cl

// Module with 1 kernels, global data with 0 words (64-bit each), starting at offset 1 words and 0 words of stack-frame
// Kernel 'test' with 147 instructions, offset 2, with following parameters: __global in out float* a (4 B, 1 items), int n (4 B, 1 items)
// label: %start_of_function
or ra0, unif, unif
or -, unif, unif
or -, mutex_acq, mutex_acq
ldi vpr_setup, vdr_setup(rows: 1, elements: 1 words, address: h32(0))
ldi vpr_setup, vdr_setup(stride: 0)
or vpr_addr, ra0, ra0                                      // address + 0
or r0, ra0, ra0
add r2, r0, 4 (4)
add r1, r0, 8 (8)
or -, vpr_wait, vpr_wait
ldi vpr_setup, vpm_setup(num: 1, size: 16 words, stride: 1 rows, address: h32(0))
or r0, vpm, vpm
or mutex_rel, 1 (1), 1 (1)
or -, mutex_acq, mutex_acq
ldi vpw_setup, vpm_setup(size: 16 words, stride: 1 rows, address: h32(0))
fadd vpm, r0, 1.000000 (32)
ldi vpw_setup, vdw_setup(rows: 1, elements: 1 words, address: h32(0))
ldi vpw_setup, vdw_setup(stride: 0)
or vpw_addr, ra0, ra0
or -, vpw_wait, vpw_wait
or mutex_rel, 1 (1), 1 (1)
or -, mutex_acq, mutex_acq
ldi vpr_setup, vdr_setup(rows: 1, elements: 1 words, address: h32(0))
ldi vpr_setup, vdr_setup(stride: 0)
or vpr_addr, r2, r2                                          // address + 4
or r0, ra0, ra0
add r3, r0, 12 (12)
or -, vpr_wait, vpr_wait
ldi vpr_setup, vpm_setup(num: 1, size: 16 words, stride: 1 rows, address: h32(0))
or r0, vpm, vpm
or mutex_rel, 1 (1), 1 (1)
or -, mutex_acq, mutex_acq
ldi vpw_setup, vpm_setup(size: 16 words, stride: 1 rows, address: h32(0))
fadd vpm, r0, 1.000000 (32)
ldi vpw_setup, vdw_setup(rows: 1, elements: 1 words, address: h32(0))
ldi vpw_setup, vdw_setup(stride: 0)
or vpw_addr, r2, r2
or -, vpw_wait, vpw_wait
or mutex_rel, 1 (1), 1 (1)
or -, mutex_acq, mutex_acq
ldi vpr_setup, vdr_setup(rows: 1, elements: 1 words, address: h32(0))
ldi vpr_setup, vdr_setup(stride: 0)
or vpr_addr, r1, r1                                           // address + 8
v8adds r0, 8 (8), 8 (8)
add r2, ra0, r0
or -, vpr_wait, vpr_wait
ldi vpr_setup, vpm_setup(num: 1, size: 16 words, stride: 1 rows, address: h32(0))
or r0, vpm, vpm
or mutex_rel, 1 (1), 1 (1)
or -, mutex_acq, mutex_acq
ldi vpw_setup, vpm_setup(size: 16 words, stride: 1 rows, address: h32(0))
fadd vpm, r0, 1.000000 (32)
ldi vpw_setup, vdw_setup(rows: 1, elements: 1 words, address: h32(0))
ldi vpw_setup, vdw_setup(stride: 0)
or vpw_addr, r1, r1
or -, vpw_wait, vpw_wait
or mutex_rel, 1 (1), 1 (1)
or -, mutex_acq, mutex_acq
ldi vpr_setup, vdr_setup(rows: 1, elements: 1 words, address: h32(0))
ldi vpr_setup, vdr_setup(stride: 0)
or vpr_addr, r3, r3                                            // address + 12
v8adds r0, 10 (10), 10 (10)
add r1, ra0, r0
or -, vpr_wait, vpr_wait
ldi vpr_setup, vpm_setup(num: 1, size: 16 words, stride: 1 rows, address: h32(0))
or r0, vpm, vpm
or mutex_rel, 1 (1), 1 (1)
or -, mutex_acq, mutex_acq
ldi vpw_setup, vpm_setup(size: 16 words, stride: 1 rows, address: h32(0))
fadd vpm, r0, 1.000000 (32)
ldi vpw_setup, vdw_setup(rows: 1, elements: 1 words, address: h32(0))
ldi vpw_setup, vdw_setup(stride: 0)
or vpw_addr, r3, r3
or -, vpw_wait, vpw_wait
or mutex_rel, 1 (1), 1 (1)
or -, mutex_acq, mutex_acq
ldi vpr_setup, vdr_setup(rows: 1, elements: 1 words, address: h32(0))
ldi vpr_setup, vdr_setup(stride: 0)
or vpr_addr, r2, r2                                              // address + 16
v8adds r0, 12 (12), 12 (12)
add r3, ra0, r0
or -, vpr_wait, vpr_wait
ldi vpr_setup, vpm_setup(num: 1, size: 16 words, stride: 1 rows, address: h32(0))
or r0, vpm, vpm
or mutex_rel, 1 (1), 1 (1)
or -, mutex_acq, mutex_acq
ldi vpw_setup, vpm_setup(size: 16 words, stride: 1 rows, address: h32(0))
fadd vpm, r0, 1.000000 (32)
ldi vpw_setup, vdw_setup(rows: 1, elements: 1 words, address: h32(0))
ldi vpw_setup, vdw_setup(stride: 0)
or vpw_addr, r2, r2
or -, vpw_wait, vpw_wait
or mutex_rel, 1 (1), 1 (1)
or -, mutex_acq, mutex_acq
ldi vpr_setup, vdr_setup(rows: 1, elements: 1 words, address: h32(0))
ldi vpr_setup, vdr_setup(stride: 0)
or vpr_addr, r1, r1                                               // address + 20
or -, vpr_wait, vpr_wait
ldi vpr_setup, vpm_setup(num: 1, size: 16 words, stride: 1 rows, address: h32(0))
or r0, vpm, vpm
or mutex_rel, 1 (1), 1 (1)
or -, mutex_acq, mutex_acq
ldi vpw_setup, vpm_setup(size: 16 words, stride: 1 rows, address: h32(0))
fadd vpm, r0, 1.000000 (32)
ldi vpw_setup, vdw_setup(rows: 1, elements: 1 words, address: h32(0))
ldi vpw_setup, vdw_setup(stride: 0)
or vpw_addr, r1, r1
or -, vpw_wait, vpw_wait
or mutex_rel, 1 (1), 1 (1)
or -, mutex_acq, mutex_acq
ldi vpr_setup, vdr_setup(rows: 1, elements: 1 words, address: h32(0))
ldi vpr_setup, vdr_setup(stride: 0)
or vpr_addr, r3, r3                                                // address + 24
or -, vpr_wait, vpr_wait
ldi vpr_setup, vpm_setup(num: 1, size: 16 words, stride: 1 rows, address: h32(0))
or r0, vpm, vpm
or mutex_rel, 1 (1), 1 (1)
or -, mutex_acq, mutex_acq
ldi vpw_setup, vpm_setup(size: 16 words, stride: 1 rows, address: h32(0))
fadd vpm, r0, 1.000000 (32)
ldi vpw_setup, vdw_setup(rows: 4, elements: 1 words, address: h32(0))  // only this write-back is 4-rows
ldi vpw_setup, vdw_setup(stride: 0)
v8adds r0, 14 (14), 14 (14)
add tmu0s, ra0, r0
nop.load_tmu0.never
fadd vpm, r4, 1.000000 (32)
ftoi r0, 32.000000 (37)
add tmu0s, ra0, r0
nop.load_tmu0.never
fadd vpm, r4, 1.000000 (32)
mul24 r0, 6 (6), 6 (6)
add tmu0s, ra0, r0
nop.load_tmu0.never
fadd vpm, r4, 1.000000 (32)
or vpw_addr, r3, r3
or -, vpw_wait, vpw_wait
or mutex_rel, 1 (1), 1 (1)
// label: %end_of_function
or r0, unif, unif
or.setf -, elem_num, r0
brr.ifallzc (pc+4) + -143 // to %start_of_function
nop.never
nop.never
nop.never
not irq, qpu_num
nop.thrend.never
nop.never
nop.never