X-Git-Url: https://git.distorted.org.uk/~mdw/catacomb/blobdiff_plain/ac82eac807a9818e676c316f8afeab5cff2572cd..d2679863cf27b0812a4b88397be1ebf0b1319305:/base/asm-common.h

diff --git a/base/asm-common.h b/base/asm-common.h
index 44c223da..cfa056f7 100644
--- a/base/asm-common.h
+++ b/base/asm-common.h
@@ -311,6 +311,7 @@ name:
 // R_r(decor) applies decoration decor to register r, which is an internal
 // register name.  The internal register names are: `ip', `a', `b', `c', `d',
 // `si', `di', `bp', `sp', `r8'--`r15'.
+#define R_nil(decor) nil
 #define R_ip(decor) _DECOR(ip, decor, ip)
 #define R_a(decor) _DECOR(abcd, decor, a)
 #define R_b(decor) _DECOR(abcd, decor, b)
@@ -345,6 +346,8 @@ name:
 // assembler-level register name, in place of any decoration that register
 // name has already.
 
+#define _REGFORM_nil(decor) R_nil(decor)
+
 #define _REGFORM_ip(decor) R_ip(decor)
 #define _REGFORM_eip(decor) R_ip(decor)
 
@@ -452,10 +455,20 @@ name:
 #endif
 #define WHOLE(reg) _REGFORM(reg, r)
 
+// Macros for some common registers.
+#define AX R_a(r)
+#define BX R_b(r)
+#define CX R_c(r)
+#define DX R_d(r)
+#define SI R_si(r)
+#define DI R_di(r)
+#define BP R_bp(r)
+#define SP R_sp(r)
+
 // Stack management and unwinding.
-.macro	setfp	fp=R_bp(r), offset=0
+.macro	setfp	fp=BP, offset=0
   .if \offset == 0
-	mov	\fp, R_sp(r)
+	mov	\fp, SP
 #if __ELF__
 	  .cfi_def_cfa_register \fp
 #endif
@@ -463,7 +476,7 @@ name:
 	  .seh_setframe \fp, 0
 #endif
   .else
-	lea	\fp, [R_sp(r) + \offset]
+	lea	\fp, [SP + \offset]
 #if __ELF__
 	  .cfi_def_cfa_register \fp
 	  .cfi_adjust_cfa_offset -\offset
@@ -478,14 +491,14 @@ name:
 
 .macro	_dropfp	fp, offset=0
   .if \offset == 0
-	mov	R_sp(r), \fp
+	mov	SP, \fp
 #if __ELF__
-	  .cfi_def_cfa_register R_sp(r)
+	  .cfi_def_cfa_register SP
 #endif
   .else
-	lea	R_sp(r), [\fp - \offset]
+	lea	SP, [\fp - \offset]
 #if __ELF__
-	  .cfi_def_cfa_register R_sp(r)
+	  .cfi_def_cfa_register SP
 	  .cfi_adjust_cfa_offset +\offset
 #endif
   .endif
@@ -494,7 +507,7 @@ name:
 .endm
 
 .macro	stalloc	n
-	sub	R_sp(r), \n
+	sub	SP, \n
 #if __ELF__
 	  .cfi_adjust_cfa_offset +\n
 #endif
@@ -504,7 +517,7 @@ name:
 .endm
 
 .macro	stfree	n
-	add	R_sp(r), \n
+	add	SP, \n
 #if __ELF__
 	  .cfi_adjust_cfa_offset -\n
 #endif
@@ -530,14 +543,14 @@ name:
 .endm
 
 .macro	savexmm	r, offset
-	movdqa	[R_sp(r) + \offset], \r
+	movdqa	[SP + \offset], \r
 #if ABI_WIN && CPUFAM_AMD64
 	  .seh_savexmm \r, \offset
 #endif
 .endm
 
 .macro	rstrxmm	r, offset
-	movdqa	\r, [R_sp(r) + \offset]
+	movdqa	\r, [SP + \offset]
 .endm
 
 .macro	endprologue
@@ -657,6 +670,8 @@ name:
 // Internal macros: `_REGFORM_r(decor)' applies decoration decor to register
 // name r.
 
+#define _REGFORM_nil(decor) nil
+
 #define _REGFORM_s0(decor) _DECOR(s, decor, 0)
 #define _REGFORM_s1(decor) _DECOR(s, decor, 1)
 #define _REGFORM_s2(decor) _DECOR(s, decor, 2)
@@ -1008,6 +1023,319 @@ name:
 	ext	\vd\().16b, \vn\().16b, \vz\().16b, #\nbit >> 3
 .endm
 
+// Register class conversions.
+#define _GPNUM_w0   0
+#define _GPNUM_w1   1
+#define _GPNUM_w2   2
+#define _GPNUM_w3   3
+#define _GPNUM_w4   4
+#define _GPNUM_w5   5
+#define _GPNUM_w6   6
+#define _GPNUM_w7   7
+#define _GPNUM_w8   8
+#define _GPNUM_w9   9
+#define _GPNUM_w10 10
+#define _GPNUM_w11 11
+#define _GPNUM_w12 12
+#define _GPNUM_w13 13
+#define _GPNUM_w14 14
+#define _GPNUM_w15 15
+#define _GPNUM_w16 16
+#define _GPNUM_w17 17
+#define _GPNUM_w18 18
+#define _GPNUM_w19 19
+#define _GPNUM_w20 20
+#define _GPNUM_w21 21
+#define _GPNUM_w22 22
+#define _GPNUM_w23 23
+#define _GPNUM_w24 24
+#define _GPNUM_w25 25
+#define _GPNUM_w26 26
+#define _GPNUM_w27 27
+#define _GPNUM_w28 28
+#define _GPNUM_w29 29
+#define _GPNUM_w30 30
+#define _GPNUM_wzr zr
+#define _GPNUM_wsp sp
+
+#define _GPNUM_x0   0
+#define _GPNUM_x1   1
+#define _GPNUM_x2   2
+#define _GPNUM_x3   3
+#define _GPNUM_x4   4
+#define _GPNUM_x5   5
+#define _GPNUM_x6   6
+#define _GPNUM_x7   7
+#define _GPNUM_x8   8
+#define _GPNUM_x9   9
+#define _GPNUM_x10 10
+#define _GPNUM_x11 11
+#define _GPNUM_x12 12
+#define _GPNUM_x13 13
+#define _GPNUM_x14 14
+#define _GPNUM_x15 15
+#define _GPNUM_x16 16
+#define _GPNUM_x17 17
+#define _GPNUM_x18 18
+#define _GPNUM_x19 19
+#define _GPNUM_x20 20
+#define _GPNUM_x21 21
+#define _GPNUM_x22 22
+#define _GPNUM_x23 23
+#define _GPNUM_x24 24
+#define _GPNUM_x25 25
+#define _GPNUM_x26 26
+#define _GPNUM_x27 27
+#define _GPNUM_x28 28
+#define _GPNUM_x29 29
+#define _GPNUM_x30 30
+#define _GPNUM_xzr zr
+#define _GPNUM_sp  sp
+#define _GPNUM_xsp sp
+#define xsp sp
+
+#define _VNUM_b0   0
+#define _VNUM_b1   1
+#define _VNUM_b2   2
+#define _VNUM_b3   3
+#define _VNUM_b4   4
+#define _VNUM_b5   5
+#define _VNUM_b6   6
+#define _VNUM_b7   7
+#define _VNUM_b8   8
+#define _VNUM_b9   9
+#define _VNUM_b10 10
+#define _VNUM_b11 11
+#define _VNUM_b12 12
+#define _VNUM_b13 13
+#define _VNUM_b14 14
+#define _VNUM_b15 15
+#define _VNUM_b16 16
+#define _VNUM_b17 17
+#define _VNUM_b18 18
+#define _VNUM_b19 19
+#define _VNUM_b20 20
+#define _VNUM_b21 21
+#define _VNUM_b22 22
+#define _VNUM_b23 23
+#define _VNUM_b24 24
+#define _VNUM_b25 25
+#define _VNUM_b26 26
+#define _VNUM_b27 27
+#define _VNUM_b28 28
+#define _VNUM_b29 29
+#define _VNUM_b30 30
+#define _VNUM_b31 31
+
+#define _VNUM_h0   0
+#define _VNUM_h1   1
+#define _VNUM_h2   2
+#define _VNUM_h3   3
+#define _VNUM_h4   4
+#define _VNUM_h5   5
+#define _VNUM_h6   6
+#define _VNUM_h7   7
+#define _VNUM_h8   8
+#define _VNUM_h9   9
+#define _VNUM_h10 10
+#define _VNUM_h11 11
+#define _VNUM_h12 12
+#define _VNUM_h13 13
+#define _VNUM_h14 14
+#define _VNUM_h15 15
+#define _VNUM_h16 16
+#define _VNUM_h17 17
+#define _VNUM_h18 18
+#define _VNUM_h19 19
+#define _VNUM_h20 20
+#define _VNUM_h21 21
+#define _VNUM_h22 22
+#define _VNUM_h23 23
+#define _VNUM_h24 24
+#define _VNUM_h25 25
+#define _VNUM_h26 26
+#define _VNUM_h27 27
+#define _VNUM_h28 28
+#define _VNUM_h29 29
+#define _VNUM_h30 30
+#define _VNUM_h31 31
+
+#define _VNUM_s0   0
+#define _VNUM_s1   1
+#define _VNUM_s2   2
+#define _VNUM_s3   3
+#define _VNUM_s4   4
+#define _VNUM_s5   5
+#define _VNUM_s6   6
+#define _VNUM_s7   7
+#define _VNUM_s8   8
+#define _VNUM_s9   9
+#define _VNUM_s10 10
+#define _VNUM_s11 11
+#define _VNUM_s12 12
+#define _VNUM_s13 13
+#define _VNUM_s14 14
+#define _VNUM_s15 15
+#define _VNUM_s16 16
+#define _VNUM_s17 17
+#define _VNUM_s18 18
+#define _VNUM_s19 19
+#define _VNUM_s20 20
+#define _VNUM_s21 21
+#define _VNUM_s22 22
+#define _VNUM_s23 23
+#define _VNUM_s24 24
+#define _VNUM_s25 25
+#define _VNUM_s26 26
+#define _VNUM_s27 27
+#define _VNUM_s28 28
+#define _VNUM_s29 29
+#define _VNUM_s30 30
+#define _VNUM_s31 31
+
+#define _VNUM_d0   0
+#define _VNUM_d1   1
+#define _VNUM_d2   2
+#define _VNUM_d3   3
+#define _VNUM_d4   4
+#define _VNUM_d5   5
+#define _VNUM_d6   6
+#define _VNUM_d7   7
+#define _VNUM_d8   8
+#define _VNUM_d9   9
+#define _VNUM_d10 10
+#define _VNUM_d11 11
+#define _VNUM_d12 12
+#define _VNUM_d13 13
+#define _VNUM_d14 14
+#define _VNUM_d15 15
+#define _VNUM_d16 16
+#define _VNUM_d17 17
+#define _VNUM_d18 18
+#define _VNUM_d19 19
+#define _VNUM_d20 20
+#define _VNUM_d21 21
+#define _VNUM_d22 22
+#define _VNUM_d23 23
+#define _VNUM_d24 24
+#define _VNUM_d25 25
+#define _VNUM_d26 26
+#define _VNUM_d27 27
+#define _VNUM_d28 28
+#define _VNUM_d29 29
+#define _VNUM_d30 30
+#define _VNUM_d31 31
+
+#define _VNUM_q0   0
+#define _VNUM_q1   1
+#define _VNUM_q2   2
+#define _VNUM_q3   3
+#define _VNUM_q4   4
+#define _VNUM_q5   5
+#define _VNUM_q6   6
+#define _VNUM_q7   7
+#define _VNUM_q8   8
+#define _VNUM_q9   9
+#define _VNUM_q10 10
+#define _VNUM_q11 11
+#define _VNUM_q12 12
+#define _VNUM_q13 13
+#define _VNUM_q14 14
+#define _VNUM_q15 15
+#define _VNUM_q16 16
+#define _VNUM_q17 17
+#define _VNUM_q18 18
+#define _VNUM_q19 19
+#define _VNUM_q20 20
+#define _VNUM_q21 21
+#define _VNUM_q22 22
+#define _VNUM_q23 23
+#define _VNUM_q24 24
+#define _VNUM_q25 25
+#define _VNUM_q26 26
+#define _VNUM_q27 27
+#define _VNUM_q28 28
+#define _VNUM_q29 29
+#define _VNUM_q30 30
+#define _VNUM_q31 31
+
+#define _VNUM_v0   0
+#define _VNUM_v1   1
+#define _VNUM_v2   2
+#define _VNUM_v3   3
+#define _VNUM_v4   4
+#define _VNUM_v5   5
+#define _VNUM_v6   6
+#define _VNUM_v7   7
+#define _VNUM_v8   8
+#define _VNUM_v9   9
+#define _VNUM_v10 10
+#define _VNUM_v11 11
+#define _VNUM_v12 12
+#define _VNUM_v13 13
+#define _VNUM_v14 14
+#define _VNUM_v15 15
+#define _VNUM_v16 16
+#define _VNUM_v17 17
+#define _VNUM_v18 18
+#define _VNUM_v19 19
+#define _VNUM_v20 20
+#define _VNUM_v21 21
+#define _VNUM_v22 22
+#define _VNUM_v23 23
+#define _VNUM_v24 24
+#define _VNUM_v25 25
+#define _VNUM_v26 26
+#define _VNUM_v27 27
+#define _VNUM_v28 28
+#define _VNUM_v29 29
+#define _VNUM_v30 30
+#define _VNUM_v31 31
+
+#define _VNUM_z0   0
+#define _VNUM_z1   1
+#define _VNUM_z2   2
+#define _VNUM_z3   3
+#define _VNUM_z4   4
+#define _VNUM_z5   5
+#define _VNUM_z6   6
+#define _VNUM_z7   7
+#define _VNUM_z8   8
+#define _VNUM_z9   9
+#define _VNUM_z10 10
+#define _VNUM_z11 11
+#define _VNUM_z12 12
+#define _VNUM_z13 13
+#define _VNUM_z14 14
+#define _VNUM_z15 15
+#define _VNUM_z16 16
+#define _VNUM_z17 17
+#define _VNUM_z18 18
+#define _VNUM_z19 19
+#define _VNUM_z20 20
+#define _VNUM_z21 21
+#define _VNUM_z22 22
+#define _VNUM_z23 23
+#define _VNUM_z24 24
+#define _VNUM_z25 25
+#define _VNUM_z26 26
+#define _VNUM_z27 27
+#define _VNUM_z28 28
+#define _VNUM_z29 29
+#define _VNUM_z30 30
+#define _VNUM_z31 31
+
+#define _RDECOR(cls, pre, r) GLUE(pre, _##cls##NUM_##r)
+#define W(r) _RDECOR(GP, w, r)
+#define X(r) _RDECOR(GP, x, r)
+#define B(r) _RDECOR(V, b, r)
+#define H(r) _RDECOR(V, h, r)
+#define S(r) _RDECOR(V, s, r)
+#define D(r) _RDECOR(V, d, r)
+#define Q(r) _RDECOR(V, q, r)
+#define V(r) _RDECOR(V, v, r)
+#define Z(r) _RDECOR(V, z, r)
+
 // Stack management and unwinding.
 .macro	setfp	fp=x29, offset=0
   // If you're just going through the motions with a fixed-size stack frame,
@@ -1099,6 +1427,59 @@ name:
 .macro	endprologue
 .endm
 
+// cmov RD, RN, CC: set RD to RN if CC is satisfied, otherwise do nothing
+.macro	cmov	rd, rn, cc
+	csel	\rd, \rn, \rd, \cc
+.endm
+
+// Notational improvement: write `csel.CC' etc., rather than `csel ..., CC'.
+#define _COND(_)							\
+	_(eq) _(ne) _(cs) _(cc) _(vs) _(vc) _(mi) _(pl)			\
+	_(ge) _(lt) _(gt) _(le) _(hi) _(ls) _(al) _(nv)			\
+	_(hs) _(lo)
+#define _INST(_)							\
+	_(ccmp) _(ccmn)							\
+	_(csel) _(cmov)							\
+	_(csinc) _(cinc) _(cset)					\
+	_(csneg) _(cneg)						\
+	_(csinv) _(cinv) _(csetm)
+#define _CONDVAR(cc) _definstvar cc;
+#define _INSTVARS(inst)							\
+	.macro _definstvar cc;						\
+	  .macro inst.\cc args:vararg; inst \args, \cc; .endm;		\
+	.endm;								\
+	_COND(_CONDVAR);						\
+	.purgem _definstvar;
+	_INST(_INSTVARS)
+#undef _COND
+#undef _INST
+#undef _CONDVAR
+#undef _INSTVARS
+
+// Flag bits for `ccmp' and friends.
+#define CCMP_N 8
+#define CCMP_Z 4
+#define CCMP_C 2
+#define CCMP_V 1
+
+// Flag settings for satisfying conditions.
+#define CCMP_MI CCMP_N
+#define CCMP_PL 0
+#define CCMP_EQ CCMP_Z
+#define CCMP_NE 0
+#define CCMP_CS CCMP_C
+#define CCMP_HS CCMP_C
+#define CCMP_CC 0
+#define CCMP_LO 0
+#define CCMP_VS CCMP_V
+#define CCMP_VC 0
+#define CCMP_HI CCMP_C
+#define CCMP_LS 0
+#define CCMP_LT CCMP_N
+#define CCMP_GE 0
+#define CCMP_LE CCMP_N
+#define CCMP_GT 0
+
 #endif
 
 ///--------------------------------------------------------------------------