Rework br_table to use BlockCall (#5731)

Rework br_table to use BlockCall, allowing us to avoid adding new nodes during ssa construction to hold block arguments. Additionally, many places where we previously matched on InstructionData to extract branch destinations can be replaced with a use of branch_destination or branch_destination_mut.

cranelift/filetests/filetests/isa/x64/branches.clif

@@ -727,3 +727,185 @@ block2:
 ;   popq %rbp
 ;   retq
 
+function %br_table_i32(i32) -> i32 {
+block0(v0: i32):
+  br_table v0, block4, [block1, block2, block2, block3]
+
+block1:
+  v1 = iconst.i32 1
+  jump block5(v1)
+
+block2:
+  v2 = iconst.i32 2
+  jump block5(v2)
+
+block3:
+  v3 = iconst.i32 3
+  jump block5(v3)
+
+block4:
+  v4 = iconst.i32 4
+  jump block5(v4)
+
+block5(v5: i32):
+  v6 = iadd.i32 v0, v5
+  return v6
+}
+
+; VCode:
+;   pushq   %rbp
+;   movq    %rsp, %rbp
+; block0:
+;   cmpl    $4, %edi
+;   br_table %rdi, %rcx, %rdx
+; block1:
+;   movl    $4, %edx
+;   jmp     label2
+; block2:
+;   jmp     label10
+; block3:
+;   movl    $1, %edx
+;   jmp     label4
+; block4:
+;   jmp     label10
+; block5:
+;   jmp     label7
+; block6:
+;   jmp     label7
+; block7:
+;   movl    $2, %edx
+;   jmp     label10
+; block8:
+;   movl    $3, %edx
+;   jmp     label9
+; block9:
+;   jmp     label10
+; block10:
+;   movq    %rdi, %rax
+;   addl    %eax, %edx, %eax
+;   movq    %rbp, %rsp
+;   popq    %rbp
+;   ret
+; 
+; Disassembled:
+; block0: ; offset 0x0
+;   pushq %rbp
+;   movq %rsp, %rbp
+; block1: ; offset 0x4
+;   cmpl $4, %edi
+;   jae 0x39
+;   movl %edi, %edx
+;   movl $0, %ecx
+;   cmovaeq %rcx, %rdx
+;   leaq 0xa(%rip), %rcx
+;   movslq (%rcx, %rdx, 4), %rdx
+;   addq %rdx, %rcx
+;   jmpq *%rcx
+;   sbbb (%rax), %al
+;   addb %al, (%rax)
+;   andb $0, %al
+;   addb %al, (%rax)
+;   andb $0, %al
+;   addb %al, (%rax)
+;   addb %al, %cs:(%rax)
+; block2: ; offset 0x39
+;   movl $4, %edx
+; block3: ; offset 0x3e
+;   jmp 0x5c
+; block4: ; offset 0x43
+;   movl $1, %edx
+; block5: ; offset 0x48
+;   jmp 0x5c
+; block6: ; offset 0x4d
+;   movl $2, %edx
+;   jmp 0x5c
+; block7: ; offset 0x57
+;   movl $3, %edx
+; block8: ; offset 0x5c
+;   movq %rdi, %rax
+;   addl %edx, %eax
+;   movq %rbp, %rsp
+;   popq %rbp
+;   retq
+
+function %br_table_i32_inline(i32) -> i32 {
+block0(v0: i32):
+  v1 = iconst.i32 1
+  v2 = iconst.i32 2
+  v3 = iconst.i32 3
+  v4 = iconst.i32 4
+  br_table v0, block1(v4), [block1(v1), block1(v2), block1(v2), block1(v3)]
+
+block1(v5: i32):
+  return v5
+}
+
+; VCode:
+;   pushq   %rbp
+;   movq    %rsp, %rbp
+; block0:
+;   movl    $1, %edx
+;   movl    $2, %r8d
+;   movl    $3, %r9d
+;   movl    $4, %eax
+;   cmpl    $4, %edi
+;   br_table %rdi, %r11, %r10
+; block1:
+;   jmp     label6
+; block2:
+;   movq    %rdx, %rax
+;   jmp     label6
+; block3:
+;   movq    %r8, %rax
+;   jmp     label6
+; block4:
+;   movq    %r8, %rax
+;   jmp     label6
+; block5:
+;   movq    %r9, %rax
+;   jmp     label6
+; block6:
+;   movq    %rbp, %rsp
+;   popq    %rbp
+;   ret
+; 
+; Disassembled:
+; block0: ; offset 0x0
+;   pushq %rbp
+;   movq %rsp, %rbp
+; block1: ; offset 0x4
+;   movl $1, %edx
+;   movl $2, %r8d
+;   movl $3, %r9d
+;   movl $4, %eax
+;   cmpl $4, %edi
+;   jae 0x72
+;   movl %edi, %r10d
+;   movl $0, %r11d
+;   cmovaeq %r11, %r10
+;   leaq 0xb(%rip), %r11
+;   movslq (%r11, %r10, 4), %r10
+;   addq %r10, %r11
+;   jmpq *%r11
+;   adcl $0x1d000000, %eax
+;   addb %al, (%rax)
+;   addb %ah, 0x2d000000(%rip)
+;   addb %al, (%rax)
+; block2: ; offset 0x52
+;   jmp 0x72
+; block3: ; offset 0x57
+;   movq %rdx, %rax
+;   jmp 0x72
+; block4: ; offset 0x5f
+;   movq %r8, %rax
+;   jmp 0x72
+; block5: ; offset 0x67
+;   movq %r8, %rax
+;   jmp 0x72
+; block6: ; offset 0x6f
+;   movq %r9, %rax
+; block7: ; offset 0x72
+;   movq %rbp, %rsp
+;   popq %rbp
+;   retq
+

cranelift/filetests/filetests/runtests/br_table.clif

@@ -56,3 +56,47 @@ block2:
 }
 ; run: %br_table_cold_block(0) == 0
 ; run: %br_table_cold_block(1) == 0
+
+function %br_table_i32_inline(i32) -> i32 {
+block0(v0: i32):
+  v1 = iconst.i32 1
+  v2 = iconst.i32 2
+  v3 = iconst.i32 3
+  v4 = iconst.i32 4
+  br_table v0, block1(v4), [block1(v1), block1(v2), block1(v2), block1(v3)]
+
+block1(v5: i32):
+  return v5
+}
+
+; run: %br_table_i32_inline(0) == 1
+; run: %br_table_i32_inline(1) == 2
+; run: %br_table_i32_inline(2) == 2
+; run: %br_table_i32_inline(3) == 3
+; run: %br_table_i32_inline(4) == 4
+; run: %br_table_i32_inline(297) == 4
+; run: %br_table_i32_inline(65535) == 4
+
+function %br_table_i32_inline_varied(i32) -> i32 {
+block0(v0: i32):
+  v1 = iconst.i32 1
+  v2 = iconst.i32 2
+  v3 = iconst.i32 3
+  v4 = iconst.i32 4
+  br_table v0, block1(v4), [block1(v1), block2(v2, v4), block2(v4, v3), block1(v3)]
+
+block2(v6: i32, v7: i32):
+  v8 = iadd v6, v7
+  jump block1(v8)
+
+block1(v5: i32):
+  return v5
+}
+
+; run: %br_table_i32_inline_varied(0) == 1
+; run: %br_table_i32_inline_varied(1) == 6
+; run: %br_table_i32_inline_varied(2) == 7
+; run: %br_table_i32_inline_varied(3) == 3
+; run: %br_table_i32_inline_varied(4) == 4
+; run: %br_table_i32_inline_varied(297) == 4
+; run: %br_table_i32_inline_varied(65535) == 4

cranelift/filetests/filetests/verifier/type_check.clif

@@ -63,12 +63,10 @@ function %fn_call_incorrect_arg_type(i64) {
         return
 }
 
-; TODO: Should we instead just verify that jump tables contain no blocks that take arguments? This
-; error doesn't occur if no instruction uses the jump table.
 function %jump_table_args() {
     block0:
         v0 = iconst.i32 0
-        br_table v0, block2, [block1] ; error: takes no arguments, but had target block1 with 1 arguments
+        br_table v0, block2, [block1] ; error: mismatched argument count
 
     block1(v5: i32):
         return