X-Git-Url: https://git.distorted.org.uk/~mdw/catacomb/blobdiff_plain/1a517bb3785891ff6940c73af7c5a136d0250ebf..0923a413958b0e778a3f059c76355ab58e5be414:/math/mpx-mul4-x86-sse2.S

diff --git a/math/mpx-mul4-x86-sse2.S b/math/mpx-mul4-x86-sse2.S
index 8f69a559..a6613ed0 100644
--- a/math/mpx-mul4-x86-sse2.S
+++ b/math/mpx-mul4-x86-sse2.S
@@ -315,6 +315,8 @@ INTFUNC(carryprop)
 	// form.  Store the low 128 bits of the represented carry to [EDI] as
 	// a packed 128-bit value, and leave the remaining 16 bits in the low
 	// 32 bits of XMM4.  On exit, XMM3, XMM5 and XMM6 are clobbered.
+  endprologue
+
 	propout	[edi +  0], xmm4, xmm5
 	propout	[edi +  4], xmm5, xmm6
 	propout	[edi +  8], xmm6, nil
@@ -333,6 +335,8 @@ INTFUNC(dmul4)
 	// [EDI], and update the carry registers with the carry out.  The
 	// registers XMM0, XMM1, XMM2, XMM3, and XMM7 are clobbered; the
 	// general-purpose registers are preserved.
+  endprologue
+
 	mulacc	[eax +  0], ecx, xmm4, xmm5, xmm6, xmm7, t
 	mulacc	[ebx +  0], edx, xmm4, xmm5, xmm6, xmm7, nil
 	propout	[edi +  0],	 xmm4, xmm5
@@ -365,6 +369,8 @@ INTFUNC(dmla4)
 	// [EDI], and update the carry registers with the carry out.  The
 	// registers XMM0, XMM1, XMM2, XMM3, and XMM7 are clobbered; the
 	// general-purpose registers are preserved.
+  endprologue
+
 	carryadd
 
 	mulacc	[eax +  0], ecx, xmm4, xmm5, xmm6, xmm7, nil
@@ -395,6 +401,8 @@ INTFUNC(mul4zc)
 	// and set the carry registers XMM4, XMM5, XMM6 to the carry out.
 	// The registers XMM0, XMM1, XMM2, XMM3, and XMM7 are clobbered; the
 	// general-purpose registers are preserved.
+  endprologue
+
 	mulcore	[ebx +  0], edx, xmm4, xmm5, xmm6, xmm7
 	propout	[edi +  0],	 xmm4, xmm5
 
@@ -421,6 +429,8 @@ INTFUNC(mul4)
 	// and update the carry registers with the carry out.  The registers
 	// XMM0, XMM1, XMM2, XMM3, and XMM7 are clobbered; the
 	// general-purpose registers are preserved.
+  endprologue
+
 	mulacc	[ebx +  0], edx, xmm4, xmm5, xmm6, xmm7, t
 	propout	[edi +  0],	 xmm4, xmm5
 
@@ -446,6 +456,8 @@ INTFUNC(mla4zc)
 	// and set the carry registers XMM4, XMM5, XMM6 to the carry out.
 	// The registers XMM0, XMM1, XMM2, XMM3, and XMM7 are clobbered; the
 	// general-purpose registers are preserved.
+  endprologue
+
 	movd	xmm4, [edi +  0]
 	movd	xmm5, [edi +  4]
 	movd	xmm6, [edi +  8]
@@ -478,6 +490,8 @@ INTFUNC(mla4)
 	// [EDI], and update the carry registers with the carry out.  The
 	// registers XMM0, XMM1, XMM2, XMM3, and XMM7 are clobbered; the
 	// general-purpose registers are preserved.
+  endprologue
+
 	carryadd
 
 	mulacc	[ebx +  0], edx, xmm4, xmm5, xmm6, xmm7, nil
@@ -508,7 +522,8 @@ INTFUNC(mmul4)
 	// of the sum U V + N Y to [EDI], leaving the remaining carry in
 	// XMM4, XMM5, and XMM6.  The registers XMM0, XMM1, XMM2, XMM3, and
 	// XMM7 are clobbered; the general-purpose registers are preserved.
-	sub	esp, 64			// space for the carries
+	stalloc	64			// space for the carries
+  endprologue
 
 	// Calculate W = U V, and leave it in the destination.  Stash the
 	// carry pieces for later.
@@ -532,7 +547,9 @@ INTFUNC(mmla4)
 	// carry in XMM4, XMM5, and XMM6.  The registers XMM0, XMM1, XMM2,
 	// XMM3, and XMM7 are clobbered; the general-purpose registers are
 	// preserved.
-	sub	esp, 64			// space for the carries
+	stalloc	64			// space for the carries
+  endprologue
+
 	movd	xmm4, [edi +  0]
 	movd	xmm5, [edi +  4]
 	movd	xmm6, [edi +  8]
@@ -599,7 +616,7 @@ INTFUNC(mmla4)
 	paddq	xmm6, [esp + 32]
 
 	// And, with that, we're done.
-	add	esp, 64
+	stfree	64
 	ret
 
 ENDFUNC
@@ -614,6 +631,7 @@ INTFUNC(mont4)
 	// of the sum W + N Y to [EDI], leaving the remaining carry in
 	// XMM4, XMM5, and XMM6.  The registers XMM0, XMM1, XMM2, XMM3, and
 	// XMM7 are clobbered; the general-purpose registers are preserved.
+  endprologue
 
 	// Calculate Y = W M.
 	mulcore	[edi +  0], esi, xmm4, xmm5, xmm6, xmm7
@@ -680,13 +698,14 @@ FUNC(mpx_umul4_x86_sse2)
 	//
 	//	esp +  0	expanded Y (32 bytes)
 	//	esp + 32	(top of locals)
-	push	ebp
-	push	ebx
-	push	esi
-	push	edi
-	mov	ebp, esp
+	pushreg	ebp
+	pushreg	ebx
+	pushreg	esi
+	pushreg	edi
+	setfp	ebp
 	and	esp, ~15
 	sub	esp, 32
+  endprologue
 
 	// Prepare for the first iteration.
 	mov	esi, [ebp + 32]		// -> bv[0]
@@ -753,7 +772,7 @@ FUNC(mpx_umul4_x86_sse2)
 	jb	1b
 
 	// All over.
-9:	mov	esp, ebp
+9:	dropfp
 	pop	edi
 	pop	esi
 	pop	ebx
@@ -787,13 +806,14 @@ FUNC(mpxmont_mul4_x86_sse2)
 	//	esp + 108	bv limit
 	//	esp + 112	(gap)
 	//	esp + 124	(top of locals)
-	push	ebp
-	push	ebx
-	push	esi
-	push	edi
-	mov	ebp, esp
+	pushreg	ebp
+	pushreg	ebx
+	pushreg	esi
+	pushreg	edi
+	setfp	ebp
 	and	esp, ~15
 	sub	esp, 124
+  endprologue
 
 	// Establish the expanded operands.
 	pxor	xmm7, xmm7
@@ -894,11 +914,11 @@ FUNC(mpxmont_mul4_x86_sse2)
 	movd	[edi + 16], xmm4
 
 	// All done.
-9:	mov	esp, ebp
-	pop	edi
-	pop	esi
-	pop	ebx
-	pop	ebp
+9:	dropfp
+	popreg	edi
+	popreg	esi
+	popreg	ebx
+	popreg	ebp
 	ret
 
 ENDFUNC
@@ -924,13 +944,14 @@ FUNC(mpxmont_redc4_x86_sse2)
 	//	esp + 12	expanded M (32 bytes)
 	//	esp + 44	expanded Y (32 bytes)
 	//	esp + 76	(top of locals)
-	push	ebp
-	push	ebx
-	push	esi
-	push	edi
-	mov	ebp, esp
+	pushreg	ebp
+	pushreg	ebx
+	pushreg	esi
+	pushreg	edi
+	setfp	ebp
 	and	esp, ~15
 	sub	esp, 76
+  endprologue
 
 	// Establish the expanded operands and the blocks-of-4 dv limit.
 	mov	edi, [ebp + 20]		// -> Z = dv[0]
@@ -1019,11 +1040,11 @@ FUNC(mpxmont_redc4_x86_sse2)
 	jmp	5b
 
 	// All over.
-9:	mov	esp, ebp
-	pop	edi
-	pop	esi
-	pop	ebx
-	pop	ebp
+9:	dropfp
+	popreg	edi
+	popreg	esi
+	popreg	ebx
+	popreg	ebp
 	ret
 
 ENDFUNC
@@ -1052,13 +1073,14 @@ ENDFUNC
 .endm
 
 .macro	testprologue
-	push	ebp
-	push	ebx
-	push	esi
-	push	edi
-	mov	ebp, esp
+	pushreg	ebp
+	pushreg	ebx
+	pushreg	esi
+	pushreg	edi
+	setfp	ebp
 	and	esp, ~15
 	sub	esp, 3*32 + 12
+  endprologue
 	// vars:
 	//	esp +  0 = cycles
 	//	esp + 12 = v expanded
@@ -1067,11 +1089,11 @@ ENDFUNC
 .endm
 
 .macro	testepilogue
-	mov	esp, ebp
-	pop	edi
-	pop	esi
-	pop	ebx
-	pop	ebp
+	dropfp
+	popreg	edi
+	popreg	esi
+	popreg	ebx
+	popreg	ebp
 	ret
 .endm
 
@@ -1133,8 +1155,7 @@ ENDFUNC
 	movdqu	[ecx + 32], xmm6
 .endm
 
-	.globl	test_dmul4
-test_dmul4:
+FUNC(test_dmul4)
 	testprologue
 	testldcarry [ebp + 24]
 	testexpand [ebp + 36], [ebp + 40]
@@ -1144,9 +1165,9 @@ test_dmul4:
 	testtail [ebp + 48], [ebp + 44]
 	testcarryout [ebp + 24]
 	testepilogue
+ENDFUNC
 
-	.globl	test_dmla4
-test_dmla4:
+FUNC(test_dmla4)
 	testprologue
 	testldcarry [ebp + 24]
 	testexpand [ebp + 36], [ebp + 40]
@@ -1156,9 +1177,9 @@ test_dmla4:
 	testtail [ebp + 48], [ebp + 44]
 	testcarryout [ebp + 24]
 	testepilogue
+ENDFUNC
 
-	.globl	test_mul4
-test_mul4:
+FUNC(test_mul4)
 	testprologue
 	testldcarry [ebp + 24]
 	testexpand nil, [ebp + 32]
@@ -1168,9 +1189,9 @@ test_mul4:
 	testtail [ebp + 40], [ebp + 36]
 	testcarryout [ebp + 24]
 	testepilogue
+ENDFUNC
 
-	.globl	test_mla4
-test_mla4:
+FUNC(test_mla4)
 	testprologue
 	testldcarry [ebp + 24]
 	testexpand nil, [ebp + 32]
@@ -1180,9 +1201,9 @@ test_mla4:
 	testtail [ebp + 40], [ebp + 36]
 	testcarryout [ebp + 24]
 	testepilogue
+ENDFUNC
 
-	.globl	test_mmul4
-test_mmul4:
+FUNC(test_mmul4)
 	testprologue
 	testexpand [ebp + 40], [ebp + 44]
 	mov	edi, [ebp + 20]
@@ -1196,9 +1217,9 @@ test_mmul4:
 	movdqu	[edi + 16], xmm1
 	testcarryout [ebp + 24]
 	testepilogue
+ENDFUNC
 
-	.globl	test_mmla4
-test_mmla4:
+FUNC(test_mmla4)
 	testprologue
 	testexpand [ebp + 40], [ebp + 44]
 	mov	edi, [ebp + 20]
@@ -1212,9 +1233,9 @@ test_mmla4:
 	movdqu	[edi + 16], xmm1
 	testcarryout [ebp + 24]
 	testepilogue
+ENDFUNC
 
-	.globl	test_mont4
-test_mont4:
+FUNC(test_mont4)
 	testprologue
 	testexpand nil, [ebp + 36]
 	mov	edi, [ebp + 20]
@@ -1228,6 +1249,7 @@ test_mont4:
 	movdqu	[edi + 16], xmm1
 	testcarryout [ebp + 24]
 	testepilogue
+ENDFUNC
 
 #endif