handle universal real div integer evaluation,

more optimizations added,
multi-thread ready grt,
bug fixes

1 files changed, 179 insertions, 66 deletions

diff --git a/translate/grt/config/ia64.S b/translate/grt/config/ia64.S
index cd77d49..d7fb2d1 100644
--- a/translate/grt/config/ia64.S
+++ b/translate/grt/config/ia64.S
@@ -13,6 +13,8 @@ grt_stack_loop:
 	;;
 	br 1b
 	.endp
+
+	frame_size = 480
 	
 	.global grt_stack_switch#
 	.proc grt_stack_switch#
@@ -23,118 +25,209 @@ grt_stack_loop:
 	//  f2-f5, f16-f31                         [20]
 	//  p1-p5, p16-p63                         [1] ???
 	//  b1-b5                                  [5]
+	//  f2-f5, f16-f31			   [20*16]
 grt_stack_switch:
 	.prologue 2, 2
 	.vframe r2
+	{
 	alloc r31=ar.pfs, 2, 0, 0, 0
-	adds r12 = -160, r12
+	mov r14 = ar.rsc
+	adds r12 = -(frame_size + 16), r12
 	.body
 	;;
+	}
 	// Save ar.rsc, ar.bsp, ar.pfs
 	{
-	mov r14 = ar.rsc
+	st8 [r12] = r14      		// sp + 0  <- ar.rsc
 	mov r15 = ar.bsp
 	adds r22 = (5*8), r12
-	}
 	;;
+	}
 	{
-	st8 [r12] = r14      		// sp + 0  <- ar.rsc
 	st8.spill [r22] = r1, 8		// sp + 40 <- r1
+	;; 
+	st8.spill [r22] = r4, 8		// sp + 48 <- r4
 	adds r20 = 8, r12
-	}
 	;;
+	}
 	st8 [r20] = r15, 8		// sp + 8  <- ar.bsp
-	st8.spill [r22] = r4, 8		// sp + 48 <- r4
-	;;
-	mov r15 = ar.lc
-	st8 [r20] = r31, 8		// sp + 16 <- ar.pfs
 	st8.spill [r22] = r5, 8		// sp + 56 <- r5
+	mov r15 = ar.lc
 	;;
-	mov r14 = b0
-	st8 [r20] = r15, 8		// sp + 24 <- ar.lc
+	{
+	st8 [r20] = r31, 8		// sp + 16 <- ar.pfs
 	//  Flush dirty registers to the backing store
 	flushrs
+	mov r14 = b0
 	;;
+	}
+	{
+	st8 [r20] = r15, 8		// sp + 24 <- ar.lc
 	//  Set the RSE in enforced lazy mode.
 	mov ar.rsc = 0
-	;; 
-	mov r15 = ar.rnat	
-	st8.spill [r22] = r6, 8		// sp + 64 <- r6
 	;;
+	}
+	{
+	//  Save sp.
+	st8 [r33] = r12
+	mov r15 = ar.rnat
 	mov r16 = b1
-	st8.spill [r22] = r7, 16	// sp + 72 <- r7
+	;;
+	}
+	{
+	st8.spill [r22] = r6, 8		// sp + 64 <- r6
 	st8 [r20] = r15, 64		// sp + 32 <- ar.rnat
 	;;
+	}
+	{
+	st8.spill [r22] = r7, 16	// sp + 72 <- r7
+	st8 [r20] = r14, 8		// sp + 96 <- b0
 	mov r15 = b2
+	;;
+	}
+	{
 	mov r17 = ar.unat
-	st8 [r20] = r14, 8		// sp + 96 <- b0
 	;;
-	mov r14 = b3
 	st8 [r22] = r17, 24		// sp + 88 <- ar.unat
+	mov r14 = b3
+	;; 
+	}
+	{
 	st8 [r20] = r16, 16		// sp + 104 <- b1
-	;;
 	st8 [r22] = r15, 16		// sp + 112 <- b2
-	st8 [r20] = r14, 16		// sp + 120 <- b3
-	mov r16 = b4
+	mov r17 = b4
 	;;
-	st8 [r22] = r16, 16		// sp + 128 <- b4
+	}
+	{
+	st8 [r20] = r14, 16		// sp + 120 <- b3
+	st8 [r22] = r17, 16		// sp + 128 <- b4
+	mov r15 = b5
+	;; 
+	}
+	{
 	//  Read new sp.
 	ld8 r21 = [r32]
-	mov r15 = b5
 	;; 
+	st8 [r20] = r15, 24		// sp + 136 <- b5
 	mov r14 = pr
-	st8 [r20] = r15			// sp + 136 <- b5
+	;; 
+	}
 	;;
-	st8 [r22] = r14			// sp + 144 <- pr
-	adds r20 = 8, r21
+	st8 [r22] = r14, 32		// sp + 144 <- pr
+	stf.spill [r20] = f2, 32	// sp + 160 <- f2
+	;;
+	stf.spill [r22] = f3, 32	// sp + 176 <- f3
+	stf.spill [r20] = f4, 32	// sp + 192 <- f4
+	;;
+	stf.spill [r22] = f5, 32	// sp + 208 <- f5
+	stf.spill [r20] = f16, 32	// sp + 224 <- f16
+	;;
+	stf.spill [r22] = f17, 32	// sp + 240 <- f17
+	stf.spill [r20] = f18, 32	// sp + 256 <- f18
 	;;
+	stf.spill [r22] = f19, 32	// sp + 272 <- f19
+	stf.spill [r20] = f20, 32	// sp + 288 <- f20
+	;;
+	stf.spill [r22] = f21, 32	// sp + 304 <- f21
+	stf.spill [r20] = f22, 32	// sp + 320 <- f22
+	;;
+	stf.spill [r22] = f23, 32	// sp + 336 <- f23
+	stf.spill [r20] = f24, 32	// sp + 352 <- f24
+	;;
+	stf.spill [r22] = f25, 32	// sp + 368 <- f25
+	stf.spill [r20] = f26, 32	// sp + 384 <- f26
+	;;
+	stf.spill [r22] = f27, 32	// sp + 400 <- f27
+	stf.spill [r20] = f28, 32	// sp + 416 <- f28
+	;;
+	stf.spill [r22] = f29, 32	// sp + 432 <- f29
+	stf.spill [r20] = f30, 32	// sp + 448 <- f30
+	;;
+	{
+	stf.spill [r22] = f31, 32	// sp + 464 <- f31
 	invala	
-	//  Save sp.
-	st8 [r33] = r12
+	adds r20 = 8, r21
+	;;
+	}
 	ld8 r14 = [r21], 88		// sp + 0 (ar.rsc)
+	ld8 r16 = [r20], 8		// sp + 8 (ar.bsp)
 	;; 
 	ld8 r15 = [r21], -56		// sp + 88 (ar.unat)
-	ld8 r16 = [r20], 8		// sp + 8 (ar.bsp)
-	;;
-	ld8 r17 = [r21], 8		// sp + 32 (ar.rnat)
+	;; 
 	ld8 r18 = [r20], 8		// sp + 16 (ar.pfs)
 	mov ar.unat = r15
+	ld8 r17 = [r21], 8		// sp + 32 (ar.rnat)
 	;;
 	ld8 r15 = [r20], 72		// sp + 24 (ar.lc)
 	ld8.fill r1 = [r21], 8		// sp + 40 (r1)
 	mov ar.bspstore = r16
 	;; 
-	mov ar.rnat = r17
-	mov ar.pfs = r18
 	ld8.fill r4 = [r21], 8		// sp + 48 (r4)
+	mov ar.pfs = r18
+	mov ar.rnat = r17
 	;;
 	mov ar.rsc = r14
 	mov ar.lc = r15
 	ld8 r17 = [r20], 8		// sp + 96 (b0)
 	;;
-	mov b0 = r17
+	{
 	ld8.fill r5 = [r21], 8		// sp + 56 (r5)
 	ld8 r14 = [r20], 8		// sp + 104 (b1)
+	mov b0 = r17
 	;;
-	mov b1 = r14
+	}
+	{
 	ld8.fill r6 = [r21], 8		// sp + 64 (r6)
 	ld8 r15 = [r20], 8		// sp + 112 (b2)
+	mov b1 = r14
 	;;
-	mov b2 = r15
-	ld8.fill r7 = [r21], 8		// sp + 72 (r7)
+	}
+	ld8.fill r7 = [r21], 64		// sp + 72 (r7)
 	ld8 r14 = [r20], 8		// sp + 120 (b3)
+	mov b2 = r15
 	;;
+	ld8 r15 = [r20], 16		// sp + 128 (b4)
+	ld8 r16 = [r21], 40		// sp + 136 (b5)
 	mov b3 = r14
-	ld8 r15 = [r20], 8		// sp + 128 (b4)
 	;;
+	{
+	ld8 r14 = [r20], 16		// sp + 144 (pr)
+	;;
+	ldf.fill f2 = [r20], 32		// sp + 160 (f2)
 	mov b4 = r15
-	ld8 r14 = [r20], 8		// sp + 136 (b5)
 	;;
-	mov b5 = r14
-	ld8 r15 = [r20], 8		// sp + 144 (pr)
-	mov r12 = r21
+	}
+	ldf.fill f3 = [r21], 32		// sp + 176 (f3)
+	ldf.fill f4 = [r20], 32		// sp + 192 (f4)
+	mov b5 = r16
+	;;
+	ldf.fill f5 = [r21], 32		// sp + 208 (f5)
+	ldf.fill f16 = [r20], 32	// sp + 224 (f16)
+	mov pr = r14, -1
+	;;
+	ldf.fill f17 = [r21], 32	// sp + 240 (f17)
+	ldf.fill f18 = [r20], 32	// sp + 256 (f18)
 	;;
-	mov pr = r15, -1
+	ldf.fill f19 = [r21], 32	// sp + 272 (f19)
+	ldf.fill f20 = [r20], 32	// sp + 288 (f20)
+	;;
+	ldf.fill f21 = [r21], 32	// sp + 304 (f21)
+	ldf.fill f22 = [r20], 32	// sp + 320 (f22)
+	;;
+	ldf.fill f23 = [r21], 32	// sp + 336 (f23)
+	ldf.fill f24 = [r20], 32	// sp + 352 (f24)
+	;;
+	ldf.fill f25 = [r21], 32	// sp + 368 (f25)
+	ldf.fill f26 = [r20], 32	// sp + 384 (f26)
+	;;
+	ldf.fill f27 = [r21], 32	// sp + 400 (f27)
+	ldf.fill f28 = [r20], 32	// sp + 416 (f28)
+	;;
+	ldf.fill f29 = [r21], 32	// sp + 432 (f29)
+	ldf.fill f30 = [r20], 32	// sp + 448 (f30)
+	;;
+	ldf.fill f31 = [r21], 32	// sp + 464 (f31)
+	adds r12 = 16, r20
 	br.ret.sptk.many b0
 	;;
 	.endp grt_stack_switch#
@@ -146,48 +239,68 @@ grt_stack_switch:
 grt_stack_create:
 	.prologue 14, 34
 	.save ar.pfs, r35
-	alloc r35 = ar.pfs, 2, 4, 0, 0
+	alloc r35 = ar.pfs, 2, 3, 0, 0
 	.save rp, r34
-	mov r34 = b0
+	//  Compute backing store.
+	movl r14 = stack_max_size
 	;; 
 	.body
+	{
+	ld4 r36 = [r14]		// r14: bsp
+	mov r34 = b0
 	br.call.sptk.many b0 = grt_stack_allocate#
 	;;
-	//  Compute backing store.
-	movl r14=stack_max_size
+	}
+	{
 	ld8 r22 = [r32], 8	// read ip (-> b1)
-	adds r20 = -(160 + 16), r8
-	adds r21 = -(160 + 16) + 32, r8
 	;;
-	mov r18 = 0x0f		// ar.rsc: LE, PL=3, Eager
-	ld4 r14 = [r14]		// r16: bsp
-	st8 [r21] = r0, 8	// sp + 32 (ar.rnat = 0)
 	ld8 r23 = [r32]		// read r1 from func
-	st8 [r8] = r20		// Save cur_sp
+	adds r21 = -(frame_size + 16) + 32, r8
+	;;
+	}
+	{
+	st8 [r21] = r0, -32	// sp + 32 (ar.rnat = 0)
+	;; 
+	st8 [r8] = r21		// Save cur_sp
+	mov r18 = 0x0f		// ar.rsc: LE, PL=3, Eager
+	;;
+	}
+	{
+	st8 [r21] = r18, 40	// sp + 0 (ar.rsc)
 	;;
-	st8 [r20] = r18, 8	// sp + 0 (ar.rsc)
 	st8 [r21] = r23, 64	// sp + 40 (r1 = func.r1)
-	sub r14 = r8, r14	// Backing store base
+	mov b0 = r34
 	;;
-	adds r14 = 16, r14	// Add sizeof (stack_context)
-	st8 [r21] = r22, -8	// sp + 104 (b1 = func.ip)
-	;; 
+	}
+	{
+	st8 [r21] = r22, -96	// sp + 104 (b1 = func.ip)
 	movl r15 = grt_stack_loop
-	mov r16 = (0 << 7) | 1	// CFM:	sol=0, sof=1
-	st8 [r20] = r14, 8	// sp + 8 (ar.bsp)
 	;; 
-	st8 [r21] = r15, -48	// sp + 96 (b0 = grt_stack_loop)
-	st8 [r20] = r16, 8	// sp + 16 (ar.pfs)
+	}	
+	sub r14 = r8, r36	// Backing store base
 	;;
-	st8 [r20] = r0, 8	// sp + 24 (ar.lc)
-	st8 [r21] = r33		// sp + 48 (r4 = arg)
+	adds r14 = 16, r14	// Add sizeof (stack_context)
+	adds r20 = 40, r21
 	;;
-	st8 [r20] = r0, 8	// sp + 32 (ar.rnat)
+	{
+	st8 [r21] = r14, 88	// sp + 8 (ar.bsp)
 	;; 
-	
+	st8 [r21] = r15, -80	// sp + 96 (b0 = grt_stack_loop)
+	mov r16 = (0 << 7) | 1	// CFM:	sol=0, sof=1
+	;;
+	}
+	{
+	st8 [r21] = r16, 8	// sp + 16 (ar.pfs)
+	;; 
+	st8 [r21] = r0, 24	// sp + 24 (ar.lc)
 	mov ar.pfs = r35
-	mov b0 = r34
+	;;
+	}
+	{
+	st8 [r20] = r0, 8	// sp + 32 (ar.rnat)
+	st8 [r21] = r33		// sp + 48 (r4 = arg)
 	br.ret.sptk.many b0
 	;;
+	}
 	.endp grt_stack_create#
 	.ident	"GCC: (GNU) 4.0.2"