X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fmesa%2Ftnl%2Ft_vertex_sse.c;h=93189656f74733c4d92d7a2327fc4eef2663c3bf;hb=f0c8e7c32766cb78756de24a9ca5e8c28017a682;hp=a180441a5a67eb2c4e6f5e26305a68ad1f461da6;hpb=62767cf2dd1006621ecd6023b15d65b5cff41dfa;p=mesa.git

diff --git a/src/mesa/tnl/t_vertex_sse.c b/src/mesa/tnl/t_vertex_sse.c
index a180441a5a6..93189656f74 100644
--- a/src/mesa/tnl/t_vertex_sse.c
+++ b/src/mesa/tnl/t_vertex_sse.c
@@ -25,13 +25,13 @@
  *    Keith Whitwell <keithw@tungstengraphics.com>
  */
 
-#include "glheader.h"
-#include "context.h"
-#include "colormac.h"
+#include "main/glheader.h"
+#include "main/context.h"
+#include "main/colormac.h"
+#include "main/simple_list.h"
+#include "main/enums.h"
 #include "t_context.h"
 #include "t_vertex.h"
-#include "simple_list.h"
-#include "enums.h"
 
 #if defined(USE_SSE_ASM)
 
@@ -39,6 +39,12 @@
 #include "x86/common_x86_asm.h"
 
 
+/**
+ * Number of bytes to allocate for generated SSE functions
+ */
+#define MAX_SSE_CODE_SIZE 1024
+
+
 #define X    0
 #define Y    1
 #define Z    2
@@ -48,7 +54,7 @@
 struct x86_program {
    struct x86_function func;
 
-   GLcontext *ctx;
+   struct gl_context *ctx;
    GLboolean inputs_safe;
    GLboolean outputs_safe;
    GLboolean have_sse2;
@@ -140,7 +146,8 @@ static void emit_load3f_1( struct x86_program *p,
 			   struct x86_reg dest,
 			   struct x86_reg arg0 )
 {
-   emit_load4f_1(p, dest, arg0);
+   /* Loading from memory erases the upper bits. */
+   sse_movss(&p->func, dest, arg0);
 }
 
 static void emit_load2f_2( struct x86_program *p, 
@@ -154,7 +161,8 @@ static void emit_load2f_1( struct x86_program *p,
 			   struct x86_reg dest,
 			   struct x86_reg arg0 )
 {
-   emit_load4f_1(p, dest, arg0);
+   /* Loading from memory erases the upper bits. */
+   sse_movss(&p->func, dest, arg0);
 }
 
 static void emit_load1f_1( struct x86_program *p, 
@@ -334,7 +342,7 @@ static void update_src_ptr( struct x86_program *p,
  */
 static GLboolean build_vertex_emit( struct x86_program *p )
 {
-   GLcontext *ctx = p->ctx;
+   struct gl_context *ctx = p->ctx;
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
    GLuint j = 0;
@@ -346,6 +354,7 @@ static GLboolean build_vertex_emit( struct x86_program *p )
    struct x86_reg temp = x86_make_reg(file_XMM, 0);
    struct x86_reg vp0 = x86_make_reg(file_XMM, 1);
    struct x86_reg vp1 = x86_make_reg(file_XMM, 2);
+   struct x86_reg temp2 = x86_make_reg(file_XMM, 3);
    GLubyte *fixup, *label;
 
    /* Push a few regs?
@@ -487,7 +496,7 @@ static GLboolean build_vertex_emit( struct x86_program *p )
 	    update_src_ptr(p, srcECX, vtxESI, a);
 	 }
 	 else {
-	    _mesa_printf("Can't emit 1ub %x %x %d\n", a->vertoffset, a[-1].vertoffset, a[-1].vertattrsize );
+	    printf("Can't emit 1ub %x %x %d\n", a->vertoffset, a[-1].vertoffset, a[-1].vertattrsize );
 	    return GL_FALSE;
 	 }
 	 break;
@@ -518,7 +527,8 @@ static GLboolean build_vertex_emit( struct x86_program *p )
 	    sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z));
 
 	    get_src_ptr(p, srcECX, vtxESI, &a[1]);
-	    emit_load(p, temp, 1, x86_deref(srcECX), a[1].inputsize);
+	    emit_load(p, temp2, 1, x86_deref(srcECX), a[1].inputsize);
+	    sse_movss(&p->func, temp, temp2);
 	    update_src_ptr(p, srcECX, vtxESI, &a[1]);
 
 	    /* Rearrange and possibly do BGR conversion:
@@ -532,9 +542,9 @@ static GLboolean build_vertex_emit( struct x86_program *p )
 	    j++;		/* NOTE: two attrs consumed */
 	 }
 	 else {
-	    _mesa_printf("Can't emit 3ub\n");
+	    printf("Can't emit 3ub\n");
+	    return GL_FALSE;	/* add this later */
 	 }
-	 return GL_FALSE;	/* add this later */
 	 break;
 
       case EMIT_4UB_4F_RGBA:
@@ -580,12 +590,12 @@ static GLboolean build_vertex_emit( struct x86_program *p )
 	    break;
 	 case GL_UNSIGNED_SHORT:
 	 default:
-	    _mesa_printf("unknown CHAN_TYPE %s\n", _mesa_lookup_enum_by_nr(CHAN_TYPE));
+	    printf("unknown CHAN_TYPE %s\n", _mesa_lookup_enum_by_nr(CHAN_TYPE));
 	    return GL_FALSE;
 	 }
 	 break;
       default:
-	 _mesa_printf("unknown a[%d].format %d\n", j, a->format);
+	 printf("unknown a[%d].format %d\n", j, a->format);
 	 return GL_FALSE;	/* catch any new opcodes */
       }
       
@@ -619,13 +629,16 @@ static GLboolean build_vertex_emit( struct x86_program *p )
    x86_pop(&p->func, countEBP);
    x86_ret(&p->func);
 
+   assert(!vtx->emit);
    vtx->emit = (tnl_emit_func)x86_get_func(&p->func);
+
+   assert( (char *) p->func.csr - (char *) p->func.store <= MAX_SSE_CODE_SIZE );
    return GL_TRUE;
 }
 
 
 
-void _tnl_generate_sse_emit( GLcontext *ctx )
+void _tnl_generate_sse_emit( struct gl_context *ctx )
 {
    struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
    struct x86_program p;   
@@ -635,7 +648,7 @@ void _tnl_generate_sse_emit( GLcontext *ctx )
       return;
    }
 
-   _mesa_memset(&p, 0, sizeof(p));
+   memset(&p, 0, sizeof(p));
 
    p.ctx = ctx;
    p.inputs_safe = 0;		/* for now */
@@ -644,7 +657,10 @@ void _tnl_generate_sse_emit( GLcontext *ctx )
    p.identity = x86_make_reg(file_XMM, 6);
    p.chan0 = x86_make_reg(file_XMM, 7);
 
-   x86_init_func(&p.func);
+   if (!x86_init_func_size(&p.func, MAX_SSE_CODE_SIZE)) {
+      vtx->emit = NULL;
+      return;
+   }
 
    if (build_vertex_emit(&p)) {
       _tnl_register_fastpath( vtx, GL_TRUE );
@@ -660,7 +676,7 @@ void _tnl_generate_sse_emit( GLcontext *ctx )
 
 #else
 
-void _tnl_generate_sse_emit( GLcontext *ctx )
+void _tnl_generate_sse_emit( struct gl_context *ctx )
 {
    /* Dummy version for when USE_SSE_ASM not defined */
 }