aarch64: Use VCodeConstant for f64/v128 constants (#5997)

* aarch64: Translate float and splat lowering to ISLE

I was looking into `constant_f128` and its fallback lowering into memory
and to get familiar with the code I figured it'd be good to port some
Rust logic to ISLE. This commit ports the `constant_{f128,f64,f32}`
helpers into ISLE from Rust as well as the `splat_const` helper which
ended up being closely related.

Tests reflect a number of regalloc changes that happened but also namely
one major difference is that in the lowering of `f32` a 32-bit immediate
is created now instead of a 64-bit immediate (in a GP register before
it's moved into a FP register). This semantically has no change but the
generated code is slightly different in a few minor cases.

* aarch64: Load f64/v128 constants from a pool

This commit removes the `LoadFpuConst64` and `LoadFpuConst128`
pseudo-instructions from the AArch64 backend which internally loaded a
nearby constant and then jumped over it. Constants now go through the
`VCodeConstant` infrastructure which gets placed at the end of the
function similar to how x64 works. Some minor support was added in as
well to add a new addressing mode for a `MachLabel`-relative load.

cranelift/filetests/filetests/isa/aarch64/floating-point.clif

@@ -531,9 +531,9 @@ block0(v0: f32):
 ;   fmov s4, #-1
 ;   fcmp s0, s4
 ;   b.gt 8 ; udf
-;   movz x9, #20352, LSL #16
-;   fmov s17, w9
-;   fcmp s0, s17
+;   movz w8, #20352, LSL #16
+;   fmov s18, w8
+;   fcmp s0, s18
 ;   b.lt 8 ; udf
 ;   fcvtzu w0, s0
 ;   ret
@@ -547,9 +547,9 @@ block0(v0: f32):
 ;   fcmp s0, s4
 ;   b.gt #0x1c
 ;   .byte 0x1f, 0xc1, 0x00, 0x00 ; trap: int_ovf
-;   mov x9, #0x4f800000
-;   fmov s17, w9
-;   fcmp s0, s17
+;   mov w8, #0x4f800000
+;   fmov s18, w8
+;   fcmp s0, s18
 ;   b.lt #0x30
 ;   .byte 0x1f, 0xc1, 0x00, 0x00 ; trap: int_ovf
 ;   fcvtzu w0, s0
@@ -565,13 +565,13 @@ block0(v0: f32):
 ; block0:
 ;   fcmp s0, s0
 ;   b.vc 8 ; udf
-;   movz x5, #52992, LSL #16
-;   fmov s5, w5
-;   fcmp s0, s5
+;   movz w4, #52992, LSL #16
+;   fmov s6, w4
+;   fcmp s0, s6
 ;   b.ge 8 ; udf
-;   movz x11, #20224, LSL #16
-;   fmov s19, w11
-;   fcmp s0, s19
+;   movz w10, #20224, LSL #16
+;   fmov s20, w10
+;   fcmp s0, s20
 ;   b.lt 8 ; udf
 ;   fcvtzs w0, s0
 ;   ret
@@ -581,14 +581,14 @@ block0(v0: f32):
 ;   fcmp s0, s0
 ;   b.vc #0xc
 ;   .byte 0x1f, 0xc1, 0x00, 0x00 ; trap: bad_toint
-;   mov x5, #0xcf000000
-;   fmov s5, w5
-;   fcmp s0, s5
+;   mov w4, #-0x31000000
+;   fmov s6, w4
+;   fcmp s0, s6
 ;   b.ge #0x20
 ;   .byte 0x1f, 0xc1, 0x00, 0x00 ; trap: int_ovf
-;   mov x11, #0x4f000000
-;   fmov s19, w11
-;   fcmp s0, s19
+;   mov w10, #0x4f000000
+;   fmov s20, w10
+;   fcmp s0, s20
 ;   b.lt #0x34
 ;   .byte 0x1f, 0xc1, 0x00, 0x00 ; trap: int_ovf
 ;   fcvtzs w0, s0
@@ -607,9 +607,9 @@ block0(v0: f32):
 ;   fmov s4, #-1
 ;   fcmp s0, s4
 ;   b.gt 8 ; udf
-;   movz x9, #24448, LSL #16
-;   fmov s17, w9
-;   fcmp s0, s17
+;   movz w8, #24448, LSL #16
+;   fmov s18, w8
+;   fcmp s0, s18
 ;   b.lt 8 ; udf
 ;   fcvtzu x0, s0
 ;   ret
@@ -623,9 +623,9 @@ block0(v0: f32):
 ;   fcmp s0, s4
 ;   b.gt #0x1c
 ;   .byte 0x1f, 0xc1, 0x00, 0x00 ; trap: int_ovf
-;   mov x9, #0x5f800000
-;   fmov s17, w9
-;   fcmp s0, s17
+;   mov w8, #0x5f800000
+;   fmov s18, w8
+;   fcmp s0, s18
 ;   b.lt #0x30
 ;   .byte 0x1f, 0xc1, 0x00, 0x00 ; trap: int_ovf
 ;   fcvtzu x0, s0
@@ -641,13 +641,13 @@ block0(v0: f32):
 ; block0:
 ;   fcmp s0, s0
 ;   b.vc 8 ; udf
-;   movz x5, #57088, LSL #16
-;   fmov s5, w5
-;   fcmp s0, s5
+;   movz w4, #57088, LSL #16
+;   fmov s6, w4
+;   fcmp s0, s6
 ;   b.ge 8 ; udf
-;   movz x11, #24320, LSL #16
-;   fmov s19, w11
-;   fcmp s0, s19
+;   movz w10, #24320, LSL #16
+;   fmov s20, w10
+;   fcmp s0, s20
 ;   b.lt 8 ; udf
 ;   fcvtzs x0, s0
 ;   ret
@@ -657,14 +657,14 @@ block0(v0: f32):
 ;   fcmp s0, s0
 ;   b.vc #0xc
 ;   .byte 0x1f, 0xc1, 0x00, 0x00 ; trap: bad_toint
-;   mov x5, #0xdf000000
-;   fmov s5, w5
-;   fcmp s0, s5
+;   mov w4, #-0x21000000
+;   fmov s6, w4
+;   fcmp s0, s6
 ;   b.ge #0x20
 ;   .byte 0x1f, 0xc1, 0x00, 0x00 ; trap: int_ovf
-;   mov x11, #0x5f000000
-;   fmov s19, w11
-;   fcmp s0, s19
+;   mov w10, #0x5f000000
+;   fmov s20, w10
+;   fcmp s0, s20
 ;   b.lt #0x34
 ;   .byte 0x1f, 0xc1, 0x00, 0x00 ; trap: int_ovf
 ;   fcvtzs x0, s0
@@ -683,9 +683,9 @@ block0(v0: f64):
 ;   fmov d4, #-1
 ;   fcmp d0, d4
 ;   b.gt 8 ; udf
-;   movz x9, #16880, LSL #48
-;   fmov d17, x9
-;   fcmp d0, d17
+;   movz x8, #16880, LSL #48
+;   fmov d18, x8
+;   fcmp d0, d18
 ;   b.lt 8 ; udf
 ;   fcvtzu w0, d0
 ;   ret
@@ -699,9 +699,9 @@ block0(v0: f64):
 ;   fcmp d0, d4
 ;   b.gt #0x1c
 ;   .byte 0x1f, 0xc1, 0x00, 0x00 ; trap: int_ovf
-;   mov x9, #0x41f0000000000000
-;   fmov d17, x9
-;   fcmp d0, d17
+;   mov x8, #0x41f0000000000000
+;   fmov d18, x8
+;   fcmp d0, d18
 ;   b.lt #0x30
 ;   .byte 0x1f, 0xc1, 0x00, 0x00 ; trap: int_ovf
 ;   fcvtzu w0, d0
@@ -717,12 +717,12 @@ block0(v0: f64):
 ; block0:
 ;   fcmp d0, d0
 ;   b.vc 8 ; udf
-;   ldr d4, pc+8 ; b 12 ; data.f64 -2147483649
+;   ldr d4, [const(0)]
 ;   fcmp d0, d4
 ;   b.gt 8 ; udf
-;   movz x9, #16864, LSL #48
-;   fmov d17, x9
-;   fcmp d0, d17
+;   movz x8, #16864, LSL #48
+;   fmov d18, x8
+;   fcmp d0, d18
 ;   b.lt 8 ; udf
 ;   fcvtzs w0, d0
 ;   ret
@@ -732,20 +732,19 @@ block0(v0: f64):
 ;   fcmp d0, d0
 ;   b.vc #0xc
 ;   .byte 0x1f, 0xc1, 0x00, 0x00 ; trap: bad_toint
-;   ldr d4, #0x14
-;   b #0x1c
-;   .byte 0x00, 0x00, 0x20, 0x00
-;   .byte 0x00, 0x00, 0xe0, 0xc1
+;   ldr d4, #0x38
 ;   fcmp d0, d4
-;   b.gt #0x28
+;   b.gt #0x1c
 ;   .byte 0x1f, 0xc1, 0x00, 0x00 ; trap: int_ovf
-;   mov x9, #0x41e0000000000000
-;   fmov d17, x9
-;   fcmp d0, d17
-;   b.lt #0x3c
+;   mov x8, #0x41e0000000000000
+;   fmov d18, x8
+;   fcmp d0, d18
+;   b.lt #0x30
 ;   .byte 0x1f, 0xc1, 0x00, 0x00 ; trap: int_ovf
 ;   fcvtzs w0, d0
 ;   ret
+;   .byte 0x00, 0x00, 0x20, 0x00
+;   .byte 0x00, 0x00, 0xe0, 0xc1
 
 function %f39(f64) -> i64 {
 block0(v0: f64):
@@ -760,9 +759,9 @@ block0(v0: f64):
 ;   fmov d4, #-1
 ;   fcmp d0, d4
 ;   b.gt 8 ; udf
-;   movz x9, #17392, LSL #48
-;   fmov d17, x9
-;   fcmp d0, d17
+;   movz x8, #17392, LSL #48
+;   fmov d18, x8
+;   fcmp d0, d18
 ;   b.lt 8 ; udf
 ;   fcvtzu x0, d0
 ;   ret
@@ -776,9 +775,9 @@ block0(v0: f64):
 ;   fcmp d0, d4
 ;   b.gt #0x1c
 ;   .byte 0x1f, 0xc1, 0x00, 0x00 ; trap: int_ovf
-;   mov x9, #0x43f0000000000000
-;   fmov d17, x9
-;   fcmp d0, d17
+;   mov x8, #0x43f0000000000000
+;   fmov d18, x8
+;   fcmp d0, d18
 ;   b.lt #0x30
 ;   .byte 0x1f, 0xc1, 0x00, 0x00 ; trap: int_ovf
 ;   fcvtzu x0, d0
@@ -794,13 +793,13 @@ block0(v0: f64):
 ; block0:
 ;   fcmp d0, d0
 ;   b.vc 8 ; udf
-;   movz x5, #50144, LSL #48
-;   fmov d5, x5
-;   fcmp d0, d5
+;   movz x4, #50144, LSL #48
+;   fmov d6, x4
+;   fcmp d0, d6
 ;   b.ge 8 ; udf
-;   movz x11, #17376, LSL #48
-;   fmov d19, x11
-;   fcmp d0, d19
+;   movz x10, #17376, LSL #48
+;   fmov d20, x10
+;   fcmp d0, d20
 ;   b.lt 8 ; udf
 ;   fcvtzs x0, d0
 ;   ret
@@ -810,14 +809,14 @@ block0(v0: f64):
 ;   fcmp d0, d0
 ;   b.vc #0xc
 ;   .byte 0x1f, 0xc1, 0x00, 0x00 ; trap: bad_toint
-;   mov x5, #-0x3c20000000000000
-;   fmov d5, x5
-;   fcmp d0, d5
+;   mov x4, #-0x3c20000000000000
+;   fmov d6, x4
+;   fcmp d0, d6
 ;   b.ge #0x20
 ;   .byte 0x1f, 0xc1, 0x00, 0x00 ; trap: int_ovf
-;   mov x11, #0x43e0000000000000
-;   fmov d19, x11
-;   fcmp d0, d19
+;   mov x10, #0x43e0000000000000
+;   fmov d20, x10
+;   fcmp d0, d20
 ;   b.lt #0x34
 ;   .byte 0x1f, 0xc1, 0x00, 0x00 ; trap: int_ovf
 ;   fcvtzs x0, d0