Consolidation of asm code in 3.5

1 files changed, 252 insertions, 0 deletions

diff --git a/src/mesa/x86/sse_normal.S b/src/mesa/x86/sse_normal.S
new file mode 100644
index 00000000000..8dd84967154
--- /dev/null
+++ b/src/mesa/x86/sse_normal.S
@@ -0,0 +1,252 @@
+/* $Id: sse_normal.S,v 1.1 2001/03/29 06:46:16 gareth Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2001  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/** TODO:
+  * - insert PREFETCH instructions to avoid cache-misses !
+  * - some more optimizations are possible...
+  * - for 40-50% more performance in the SSE-functions, the
+  *   data (trans-matrix, src_vert, dst_vert) needs to be 16byte aligned !
+  */
+
+#include "matypes.h"
+#include "norm_args.h"
+
+   SEG_TEXT
+
+#define M(i)    REGOFF(i * 4, EDX)
+#define S(i)	REGOFF(i * 4, ESI)
+#define D(i)	REGOFF(i * 4, EDI)
+#define STRIDE  REGOFF(12, ESI)
+
+
+ALIGNTEXT16
+GLOBL GLNAME(_mesa_sse_transform_rescale_normals_no_rot)
+GLNAME(_mesa_sse_transform_rescale_normals_no_rot):
+
+#define FRAME_OFFSET 8
+	PUSH_L  ( ESI )
+	PUSH_L  ( EDI )
+
+	MOV_L	( ARG_IN, ESI )				/* ptr to source GLvector3f */
+	MOV_L	( ARG_DEST, EDI )			/* ptr to dest GLvector3f */
+
+	MOV_L	( ARG_MAT, EDX )			/* ptr to matrix */
+	MOV_L	( REGOFF(MATRIX_INV, EDX), EDX)		/* matrix->inv */
+
+	MOV_L	( REGOFF(V3F_COUNT, ESI), ECX ) 	/* source count */
+
+	TEST_L	( ECX, ECX )
+	JZ( LLBL(K_G3TRNNRR_finish) )			/* count was zero; go to finish */
+
+	MOV_L	( STRIDE, EAX )				/* stride */
+	MOV_L	( ECX, REGOFF(V3F_COUNT, EDI) )		/* set dest-count */
+
+	IMUL_L( CONST(12), ECX )			/* count *= 12 */
+	MOV_L( REGOFF(V3F_START, ESI), ESI )		/* ptr to first source vertex */
+
+	MOV_L( REGOFF(V3F_START, EDI), EDI )		/* ptr to first dest vertex */
+	ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+ALIGNTEXT32
+	MOVSS	( M(0), XMM1 )				/* m0 */
+	MOVSS	( M(5), XMM2 )				/* m5 */
+	UNPCKLPS( XMM2, XMM1 )				/* m5 | m0 */
+	MOVSS	( ARG_SCALE, XMM0 )			/* scale */
+	SHUFPS	( CONST(0x0), XMM0, XMM0 )		/* scale | scale */
+	MULPS	( XMM0, XMM1 )				/* m5*scale | m0*scale */
+	MULSS	( M(10), XMM0 )				/* m10*scale */
+
+ALIGNTEXT32
+LLBL(K_G3TRNNRR_top):
+	MOVLPS	( S(0), XMM2 )				/* uy | ux */
+	MULPS	( XMM1, XMM2 )				/* uy*m5*scale | ux*m0*scale */
+	MOVLPS	( XMM2, D(0) )				/* ->D(1) | D(0) */
+
+	MOVSS	( S(2), XMM2 )				/* uz */
+	MULSS	( XMM0, XMM2 )				/* uz*m10*scale */
+	MOVSS	( XMM2, D(2) )				/* ->D(2) */
+
+LLBL(K_G3TRNNRR_skip):
+	ADD_L	( CONST(12), EDI )
+	ADD_L	( EAX, ESI )
+	CMP_L	( ECX, EDI )
+	JNE	( LLBL(K_G3TRNNRR_top) )
+
+LLBL(K_G3TRNNRR_finish):
+	POP_L	( EDI )
+	POP_L	( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+
+ALIGNTEXT16
+GLOBL GLNAME(_mesa_sse_transform_rescale_normals)
+GLNAME(_mesa_sse_transform_rescale_normals):
+
+#define FRAME_OFFSET 8
+	PUSH_L  ( ESI )
+	PUSH_L  ( EDI )
+
+	MOV_L	( ARG_IN, ESI )				/* ptr to source GLvector3f */
+	MOV_L	( ARG_DEST, EDI )			/* ptr to dest GLvector3f */
+
+	MOV_L	( ARG_MAT, EDX )			/* ptr to matrix */
+	MOV_L	( REGOFF(MATRIX_INV, EDX), EDX)		/* matrix->inv */
+
+	MOV_L	( REGOFF(V3F_COUNT, ESI), ECX ) 	/* source count */
+
+	TEST_L	( ECX, ECX )
+	JZ( LLBL(K_G3TRNR_finish) )			/* count was zero; go to finish */
+
+	MOV_L	( STRIDE, EAX )				/* stride */
+	MOV_L	( ECX, REGOFF(V3F_COUNT, EDI) )		/* set dest-count */
+
+	IMUL_L( CONST(12), ECX )			/* count *= 12 */
+	MOV_L( REGOFF(V3F_START, ESI), ESI )		/* ptr to first source vertex */
+
+	MOV_L( REGOFF(V3F_START, EDI), EDI )		/* ptr to first dest vertex */
+	ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+ALIGNTEXT32
+	MOVSS	( M(0), XMM0 )				/* m0 */
+	MOVSS	( M(4), XMM1 )				/* m4 */
+	UNPCKLPS( XMM1, XMM0 )				/* m4 | m0 */
+
+	MOVSS	( ARG_SCALE, XMM4 )			/* scale */
+	SHUFPS	( CONST(0x0), XMM4, XMM4 )		/* scale | scale */
+
+	MULPS	( XMM4, XMM0 )				/* m4*scale | m0*scale */
+	MOVSS	( M(1), XMM1 )				/* m1 */
+	MOVSS	( M(5), XMM2 )				/* m5 */
+	UNPCKLPS( XMM2, XMM1 )				/* m5 | m1 */
+	MULPS	( XMM4, XMM1 )				/* m5*scale | m1*scale */
+	MOVSS	( M(2), XMM2 )				/* m2 */
+	MOVSS	( M(6), XMM3 )				/* m6 */
+	UNPCKLPS( XMM3, XMM2 )				/* m6 | m2 */
+	MULPS	( XMM4, XMM2 )				/* m6*scale | m2*scale */
+
+	MOVSS	( M(8), XMM6 )				/* m8 */
+	MULSS	( ARG_SCALE, XMM6 )			/* m8*scale */
+	MOVSS	( M(9), XMM7 )				/* m9 */
+	MULSS	( ARG_SCALE, XMM7 )			/* m9*scale */
+
+ALIGNTEXT32
+LLBL(K_G3TRNR_top):
+	MOVSS	( S(0), XMM3 )				/* ux */
+	SHUFPS	( CONST(0x0), XMM3, XMM3 )		/* ux | ux */
+	MULPS	( XMM0, XMM3 )				/* ux*m4 | ux*m0 */
+	MOVSS	( S(1), XMM4 )				/* uy */
+	SHUFPS	( CONST(0x0), XMM4, XMM4 )		/* uy | uy */
+	MULPS	( XMM1, XMM4 )				/* uy*m5 | uy*m1 */
+	MOVSS	( S(2), XMM5 )				/* uz */
+	SHUFPS	( CONST(0x0), XMM5, XMM5 )		/* uz | uz */
+	MULPS	( XMM2, XMM5 )				/* uz*m6 | uz*m2 */
+
+	ADDPS	( XMM4, XMM3 )
+	ADDPS	( XMM5, XMM3 )
+	MOVLPS	( XMM3, D(0) )
+
+	MOVSS	( M(10), XMM3 )				/* m10 */
+	MULSS	( ARG_SCALE, XMM3 )			/* m10*scale */
+	MULSS	( S(2), XMM3 )				/* m10*scale*uz */
+	MOVSS	( S(1), XMM4 )				/* uy */
+	MULSS	( XMM7, XMM4 )				/* uy*m9*scale */
+	MOVSS	( S(0), XMM5 )				/* ux */
+	MULSS	( XMM6, XMM5 )				/* ux*m8*scale */
+
+	ADDSS	( XMM4, XMM3 )
+	ADDSS	( XMM5, XMM3 )
+	MOVSS	( XMM3, D(2) )
+
+LLBL(K_G3TRNR_skip):
+	ADD_L	( CONST(12), EDI )
+	ADD_L	( EAX, ESI )
+	CMP_L	( ECX, EDI )
+	JNE	( LLBL(K_G3TRNR_top) )
+
+LLBL(K_G3TRNR_finish):
+	POP_L	( EDI )
+	POP_L	( ESI )
+	RET
+#undef FRAME_OFFSET
+
+
+ALIGNTEXT16
+GLOBL GLNAME(_mesa_sse_transform_normals_no_rot)
+GLNAME(_mesa_sse_transform_normals_no_rot):
+
+#define FRAME_OFFSET 8
+	PUSH_L  ( ESI )
+	PUSH_L  ( EDI )
+
+	MOV_L	( ARG_IN, ESI )				/* ptr to source GLvector3f */
+	MOV_L	( ARG_DEST, EDI )			/* ptr to dest GLvector3f */
+
+	MOV_L	( ARG_MAT, EDX )			/* ptr to matrix */
+	MOV_L	( REGOFF(MATRIX_INV, EDX), EDX)		/* matrix->inv */
+
+	MOV_L	( REGOFF(V3F_COUNT, ESI), ECX ) 	/* source count */
+
+	TEST_L	( ECX, ECX )
+	JZ( LLBL(K_G3TNNRR_finish) )			/* count was zero; go to finish */
+
+	MOV_L	( STRIDE, EAX )				/* stride */
+	MOV_L	( ECX, REGOFF(V3F_COUNT, EDI) )		/* set dest-count */
+
+	IMUL_L( CONST(12), ECX )			/* count *= 12 */
+	MOV_L( REGOFF(V3F_START, ESI), ESI )		/* ptr to first source vertex */
+
+	MOV_L( REGOFF(V3F_START, EDI), EDI )		/* ptr to first dest vertex */
+	ADD_L( EDI, ECX ) 				/* count += dest ptr */
+
+ALIGNTEXT32
+	MOVSS( M(0), XMM0 )				/* m0 */
+	MOVSS( M(5), XMM1 )				/* m5 */
+	UNPCKLPS( XMM1, XMM0 )				/* m5 | m0 */
+	MOVSS( M(10), XMM1 )				/* m10 */
+
+ALIGNTEXT32
+LLBL(K_G3TNNRR_top):
+	MOVLPS( S(0), XMM2 )				/* uy | ux */
+	MULPS( XMM0, XMM2 )				/* uy*m5 | ux*m0 */
+	MOVLPS( XMM2, D(0) )
+
+	MOVSS( S(2), XMM2 )				/* uz */
+	MULSS( XMM1, XMM2 )				/* uz*m10 */
+	MOVSS( XMM2, D(2) )
+
+LLBL(K_G3TNNRR_skip):
+	ADD_L	( CONST(12), EDI )
+	ADD_L	( EAX, ESI )
+	CMP_L	( ECX, EDI )
+	JNE	( LLBL(K_G3TNNRR_top) )
+
+LLBL(K_G3TNNRR_finish):
+	POP_L	( EDI )
+	POP_L	( ESI )
+	RET
+#undef FRAME_OFFSET