Merge branch 'gallium-0.1' into gallium-tex-surfaces

25 files changed, 306 insertions, 269 deletions

diff --git a/src/gallium/README.portability b/src/gallium/README.portability
index ab0c1978476..18a97f449b4 100644
--- a/src/gallium/README.portability
+++ b/src/gallium/README.portability
@@ -24,6 +24,7 @@ headers in general, should stricly follow these guidelines to ensure
 * Don't use variable number of macro arguments. Use static inline functions
 instead.
 
+* Don't use C99 features.
 
 = Standard Library =
 
@@ -42,3 +43,67 @@ portable way.
 * Use the functions/macros in p_debug.h.
 
 * Don't include assert.h, call abort, printf, etc.
+
+
+= Code Style =
+
+== Inherantice in C ==
+
+The main thing we do is mimic inheritance by structure containment.
+
+Here's a silly made-up example:
+
+/* base class */
+struct buffer
+{
+  int size;
+  void (*validate)(struct buffer *buf);
+};
+
+/* sub-class of bufffer */
+struct texture_buffer
+{
+  struct buffer base;  /* the base class, MUST COME FIRST! */
+  int format;
+  int width, height;
+};
+
+
+Then, we'll typically have cast-wrapper functions to convert base-class 
+pointers to sub-class pointers where needed:
+
+static inline struct vertex_buffer *vertex_buffer(struct buffer *buf)
+{
+  return (struct vertex_buffer *) buf;
+}
+
+
+To create/init a sub-classed object:
+
+struct buffer *create_texture_buffer(int w, int h, int format)
+{
+  struct texture_buffer *t = malloc(sizeof(*t));
+  t->format = format;
+  t->width = w;
+  t->height = h;
+  t->base.size = w * h;
+  t->base.validate = tex_validate;
+  return &t->base;
+}
+
+Example sub-class method:
+
+void tex_validate(struct buffer *buf)
+{
+  struct texture_buffer *tb = texture_buffer(buf);
+  assert(tb->format);
+  assert(tb->width);
+  assert(tb->height);
+}
+
+
+Note that we typically do not use typedefs to make "class names"; we use
+'struct whatever' everywhere.
+
+Gallium's pipe_context and the subclassed psb_context, etc are prime examples 
+of this.  There's also many examples in Mesa and the Mesa state tracker.
diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c
index 3aa326acc7f..4bca92ff11d 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c
@@ -256,7 +256,7 @@ pstip_transform_inst(struct tgsi_transform_context *ctx,
          uint size = 4;
          immed = tgsi_default_full_immediate();
          immed.Immediate.Size = 1 + size; /* one for the token itself */
-         immed.u.ImmediateFloat32 = (struct tgsi_immediate_float32 *) value;
+         immed.u.Pointer = (void *) value;
          ctx->emit_immediate(ctx, &immed);
       }
 
diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c
index c85d8ded50c..355093f945e 100644
--- a/src/gallium/auxiliary/draw/draw_pt_varray.c
+++ b/src/gallium/auxiliary/draw/draw_pt_varray.c
@@ -200,7 +200,6 @@ static void varray_prepare(struct draw_pt_front_end *frontend,
                            unsigned opt)
 {
    struct varray_frontend *varray = (struct varray_frontend *)frontend;
-   const struct pipe_rasterizer_state *rasterizer = varray->draw->rasterizer;
 
    if (opt & PT_PIPELINE)
    {
diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c
index 2f9775814f5..6b3fb1406ba 100644
--- a/src/gallium/auxiliary/draw/draw_pt_vcache.c
+++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c
@@ -225,7 +225,6 @@ static void vcache_prepare( struct draw_pt_front_end *frontend,
 			    unsigned opt )
 {
    struct vcache_frontend *vcache = (struct vcache_frontend *)frontend;
-   const struct pipe_rasterizer_state *rasterizer = vcache->draw->rasterizer;
 
    if (opt & PT_PIPELINE)
    {
diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c
index 07f85bc448f..a57c938fbf1 100644
--- a/src/gallium/auxiliary/draw/draw_vs_sse.c
+++ b/src/gallium/auxiliary/draw/draw_vs_sse.c
@@ -51,17 +51,17 @@
 
 #if SSE_SWIZZLES
 typedef void (XSTDCALL *codegen_function) (
-   const struct tgsi_exec_vector *input,
-   struct tgsi_exec_vector *output,
-   float (*constant)[4],
-   struct tgsi_exec_vector *temporary,
-   float (*immediates)[4],
-   const float (*aos_input)[4],
-   uint num_inputs,
-   uint input_stride,
-   float (*aos_output)[4],
-   uint num_outputs,
-   uint output_stride );
+   const struct tgsi_exec_vector *input, /* 1 */
+   struct tgsi_exec_vector *output, /* 2 */
+   float (*constant)[4],        /* 3 */
+   struct tgsi_exec_vector *temporary, /* 4 */
+   float (*immediates)[4],      /* 5 */
+   const float (*aos_input)[4], /* 6 */
+   uint num_inputs,             /* 7 */
+   uint input_stride,           /* 8 */
+   float (*aos_output)[4],      /* 9 */
+   uint num_outputs,            /* 10 */
+   uint output_stride );        /* 11 */
 #else
 typedef void (XSTDCALL *codegen_function) (
    const struct tgsi_exec_vector *input,
diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c
index d7e22305573..4e036d9032c 100644
--- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c
+++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c
@@ -347,9 +347,9 @@ struct x86_reg x86_get_base_reg( struct x86_reg reg )
    return x86_make_reg( reg.file, reg.idx );
 }
 
-unsigned char *x86_get_label( struct x86_function *p )
+int x86_get_label( struct x86_function *p )
 {
-   return p->csr;
+   return p->csr - p->store;
 }
 
 
@@ -361,17 +361,22 @@ unsigned char *x86_get_label( struct x86_function *p )
 
 void x86_jcc( struct x86_function *p,
 	      enum x86_cc cc,
-	      unsigned char *label )
+	      int label )
 {
-   intptr_t offset = pointer_to_intptr( label ) - (pointer_to_intptr( x86_get_label(p) ) + 2);
+   int offset = label - (x86_get_label(p) + 2);
    DUMP_I(cc);
    
+   if (offset < 0) {
+      int amt = p->csr - p->store;
+      assert(amt > -offset);
+   }
+
    if (offset <= 127 && offset >= -128) {
       emit_1ub(p, 0x70 + cc);
       emit_1b(p, (char) offset);
    }
    else {
-      offset = pointer_to_intptr( label ) - (pointer_to_intptr( x86_get_label(p) ) + 6);
+      offset = label - (x86_get_label(p) + 6);
       emit_2ub(p, 0x0f, 0x80 + cc);
       emit_1i(p, offset);
    }
@@ -379,8 +384,8 @@ void x86_jcc( struct x86_function *p,
 
 /* Always use a 32bit offset for forward jumps:
  */
-unsigned char *x86_jcc_forward( struct x86_function *p,
-			  enum x86_cc cc )
+int x86_jcc_forward( struct x86_function *p,
+                     enum x86_cc cc )
 {
    DUMP_I(cc);
    emit_2ub(p, 0x0f, 0x80 + cc);
@@ -388,7 +393,7 @@ unsigned char *x86_jcc_forward( struct x86_function *p,
    return x86_get_label(p);
 }
 
-unsigned char *x86_jmp_forward( struct x86_function *p)
+int x86_jmp_forward( struct x86_function *p)
 {
    DUMP();
    emit_1ub(p, 0xe9);
@@ -396,7 +401,7 @@ unsigned char *x86_jmp_forward( struct x86_function *p)
    return x86_get_label(p);
 }
 
-unsigned char *x86_call_forward( struct x86_function *p)
+int x86_call_forward( struct x86_function *p)
 {
    DUMP();
 
@@ -408,42 +413,24 @@ unsigned char *x86_call_forward( struct x86_function *p)
 /* Fixup offset from forward jump:
  */
 void x86_fixup_fwd_jump( struct x86_function *p,
-			 unsigned char *fixup )
+			 int fixup )
 {
-   *(int *)(fixup - 4) = pointer_to_intptr( x86_get_label(p) ) - pointer_to_intptr( fixup );
+   *(int *)(p->store + fixup - 4) = x86_get_label(p) - fixup;
 }
 
-void x86_jmp( struct x86_function *p, unsigned char *label)
+void x86_jmp( struct x86_function *p, int label)
 {
    DUMP_I( label );
    emit_1ub(p, 0xe9);
-   emit_1i(p, pointer_to_intptr( label ) - pointer_to_intptr( x86_get_label(p) ) - 4);
-}
-
-#if 0
-static unsigned char *cptr( void (*label)() )
-{
-   return (unsigned char *) label;
+   emit_1i(p, label - x86_get_label(p) - 4);
 }
 
-/* This doesn't work once we start reallocating & copying the
- * generated code on buffer fills, because the call is relative to the
- * current pc.
- */
-void x86_call( struct x86_function *p, void (*label)())
-{
-   DUMP_I( label );
-   emit_1ub(p, 0xe8);
-   emit_1i(p, cptr(label) - x86_get_label(p) - 4);
-}
-#else
 void x86_call( struct x86_function *p, struct x86_reg reg)
 {
    DUMP_R( reg );
    emit_1ub(p, 0xff);
    emit_modrm_noreg(p, 2, reg);
 }
-#endif
 
 
 /* michal:
@@ -462,8 +449,15 @@ void x86_push( struct x86_function *p,
 	       struct x86_reg reg )
 {
    DUMP_R( reg );
-   assert(reg.mod == mod_REG);
-   emit_1ub(p, 0x50 + reg.idx);
+   if (reg.mod == mod_REG)
+      emit_1ub(p, 0x50 + reg.idx);
+   else 
+   {
+      emit_1ub(p, 0xff);
+      emit_modrm_noreg(p, 6, reg);
+   }
+
+
    p->stack_offset += 4;
 }
 
@@ -495,6 +489,7 @@ void x86_dec( struct x86_function *p,
 void x86_ret( struct x86_function *p )
 {
    DUMP();
+   assert(p->stack_offset == 0);
    emit_1ub(p, 0xc3);
 }
 
diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h
index ad79b1facf8..eacaeeaf6fe 100644
--- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h
+++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h
@@ -124,23 +124,23 @@ struct x86_reg x86_get_base_reg( struct x86_reg reg );
 
 /* Labels, jumps and fixup:
  */
-unsigned char *x86_get_label( struct x86_function *p );
+int x86_get_label( struct x86_function *p );
 
 void x86_jcc( struct x86_function *p,
 	      enum x86_cc cc,
-	      unsigned char *label );
+	      int label );
 
-unsigned char *x86_jcc_forward( struct x86_function *p,
+int x86_jcc_forward( struct x86_function *p,
 			  enum x86_cc cc );
 
-unsigned char *x86_jmp_forward( struct x86_function *p);
+int x86_jmp_forward( struct x86_function *p);
 
-unsigned char *x86_call_forward( struct x86_function *p);
+int x86_call_forward( struct x86_function *p);
 
 void x86_fixup_fwd_jump( struct x86_function *p,
-			 unsigned char *fixup );
+			 int fixup );
 
-void x86_jmp( struct x86_function *p, unsigned char *label );
+void x86_jmp( struct x86_function *p, int label );
 
 /* void x86_call( struct x86_function *p, void (*label)() ); */
 void x86_call( struct x86_function *p, struct x86_reg reg);
diff --git a/src/gallium/auxiliary/sct/sct.c b/src/gallium/auxiliary/sct/sct.c
index 97ee5882a16..5e4126e014d 100644
--- a/src/gallium/auxiliary/sct/sct.c
+++ b/src/gallium/auxiliary/sct/sct.c
@@ -209,6 +209,7 @@ remove_context_from_surface(struct sct_surface *si,
       }
       else {
          prev = curr;
+         next = curr->next;
       }
    }
 }
diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c
index d55f907c0d3..5d5125f7cbe 100644
--- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c
+++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c
@@ -1530,41 +1530,44 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_EXP:
-      debug_printf("TGSI: EXP opcode not implemented\n");
-      /* from ARB_v_p:
-      tmp = ScalarLoad(op0);
-      result.x = 2^floor(tmp);
-      result.y = tmp - floor(tmp);
-      result.z = RoughApprox2ToX(tmp);
-      result.w = 1.0;
-      */
-#if 0
-      /* something like this: */
       FETCH( &r[0], 0, CHAN_X );
-      micro_exp2( &r[0], &r[0] );
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-	 STORE( &r[0], 0, chan_index );
+      micro_flr( &r[1], &r[0] );  /* r1 = floor(r0) */
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) {
+         micro_exp2( &r[2], &r[1] );       /* r2 = 2 ^ r1 */
+	 STORE( &r[2], 0, CHAN_X );        /* store r2 */
+      }
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) {
+         micro_sub( &r[2], &r[0], &r[1] ); /* r2 = r0 - r1 */
+	 STORE( &r[2], 0, CHAN_Y );        /* store r2 */
+      }
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
+         micro_exp2( &r[2], &r[0] );       /* r2 = 2 ^ r0 */
+	 STORE( &r[2], 0, CHAN_Z );        /* store r2 */
+      }
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) {
+	 STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W );
       }
-#endif
       break;
 
    case TGSI_OPCODE_LOG:
-      debug_printf("TGSI: LOG opcode not implemented\n");
-      /* from ARB_v_p:
-      tmp = fabs(ScalarLoad(op0));
-      result.x = floor(log2(tmp));
-      result.y = tmp / 2^(floor(log2(tmp)));
-      result.z = RoughApproxLog2(tmp);
-      result.w = 1.0;
-      */
-#if 0
-      /* something like this: */
       FETCH( &r[0], 0, CHAN_X );
-      micro_lg2( &r[0], &r[0] );
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-	 STORE( &r[0], 0, chan_index );
+      micro_abs( &r[2], &r[0] );  /* r2 = abs(r0) */
+      micro_lg2( &r[1], &r[2] );  /* r1 = lg2(r2) */
+      micro_flr( &r[0], &r[1] );  /* r0 = floor(r1) */
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) {
+	 STORE( &r[0], 0, CHAN_X );
+      }
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) {
+         micro_exp2( &r[0], &r[0] );       /* r0 = 2 ^ r0 */
+         micro_div( &r[0], &r[2], &r[0] ); /* r0 = r2 / r0 */
+	 STORE( &r[0], 0, CHAN_Y );
+      }
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
+	 STORE( &r[1], 0, CHAN_Z );
+      }
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) {
+	 STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W );
       }
-#endif
       break;
 
    case TGSI_OPCODE_MUL:
diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c
index 1138f599972..2fd76a3072d 100755
--- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c
+++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c
@@ -103,15 +103,9 @@ get_output_base( void )
 static struct x86_reg
 get_temp_base( void )
 {
-#ifdef WIN32
    return x86_make_reg(
       file_REG32,
       reg_BX );
-#else
-   return x86_make_reg(
-      file_REG32,
-      reg_SI );
-#endif
 }
 
 static struct x86_reg
@@ -133,14 +127,6 @@ get_immediate_base( void )
  * Data access helpers.
  */
 
-static struct x86_reg
-get_argument(
-   unsigned index )
-{
-   return x86_make_disp(
-      x86_make_reg( file_REG32, reg_SP ),
-      (index + 1) * 4 );
-}
 
 static struct x86_reg
 get_immediate(
@@ -455,19 +441,13 @@ emit_push_gp(
 {
    x86_push(
       func,
-      get_const_base() );
+      x86_make_reg( file_REG32, reg_AX) );
    x86_push(
       func,
-      get_input_base() );
+      x86_make_reg( file_REG32, reg_CX) );
    x86_push(
       func,
-      get_output_base() );
-
-   /* It is important on non-win32 platforms that temp base is pushed last.
-    */
-   x86_push(
-      func,
-      get_temp_base() );
+      x86_make_reg( file_REG32, reg_DX) );
 }
 
 static void
@@ -478,16 +458,13 @@ x86_pop_gp(
     */
    x86_pop(
       func,
-      get_temp_base() );
-   x86_pop(
-      func,
-      get_output_base() );
+      x86_make_reg( file_REG32, reg_DX) );
    x86_pop(
       func,
-      get_input_base() );
+      x86_make_reg( file_REG32, reg_CX) );
    x86_pop(
       func,
-      get_const_base() );
+      x86_make_reg( file_REG32, reg_AX) );
 }
 
 static void
@@ -504,19 +481,23 @@ emit_func_call_dst(
    emit_push_gp(
       func );
 
-#ifdef WIN32
-   x86_push(
-      func,
-      get_temp( TEMP_R0, 0 ) );
-#endif
-
    {
       struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX );
 
+      x86_lea(
+         func,
+         ecx,
+         get_temp( TEMP_R0, 0 ) );
+
+      x86_push( func, ecx );
       x86_mov_reg_imm( func, ecx, (unsigned long) code );
       x86_call( func, ecx );
+#ifndef WIN32
+      x86_pop(func, ecx ); 
+#endif
    }
 
+
    x86_pop_gp(
       func );
 
@@ -577,11 +558,7 @@ static void XSTDCALL
 cos4f(
    float *store )
 {
-#ifdef WIN32
    const unsigned X = 0;
-#else
-   const unsigned X = TEMP_R0 * 16;
-#endif
 
    store[X + 0] = cosf( store[X + 0] );
    store[X + 1] = cosf( store[X + 1] );
@@ -604,11 +581,8 @@ static void XSTDCALL
 ex24f(
    float *store )
 {
-#ifdef WIN32
    const unsigned X = 0;
-#else
-   const unsigned X = TEMP_R0 * 16;
-#endif
+
    store[X + 0] = powf( 2.0f, store[X + 0] );
    store[X + 1] = powf( 2.0f, store[X + 1] );
    store[X + 2] = powf( 2.0f, store[X + 2] );
@@ -641,11 +615,8 @@ static void XSTDCALL
 flr4f(
    float *store )
 {
-#ifdef WIN32
    const unsigned X = 0;
-#else
-   const unsigned X = TEMP_R0 * 16;
-#endif
+
    store[X + 0] = floorf( store[X + 0] );
    store[X + 1] = floorf( store[X + 1] );
    store[X + 2] = floorf( store[X + 2] );
@@ -667,11 +638,8 @@ static void XSTDCALL
 frc4f(
    float *store )
 {
-#ifdef WIN32
    const unsigned X = 0;
-#else
-   const unsigned X = TEMP_R0 * 16;
-#endif
+
    store[X + 0] -= floorf( store[X + 0] );
    store[X + 1] -= floorf( store[X + 1] );
    store[X + 2] -= floorf( store[X + 2] );
@@ -693,11 +661,8 @@ static void XSTDCALL
 lg24f(
    float *store )
 {
-#ifdef WIN32
    const unsigned X = 0;
-#else
-   const unsigned X = TEMP_R0 * 16;
-#endif
+
    store[X + 0] = LOG2( store[X + 0] );
    store[X + 1] = LOG2( store[X + 1] );
    store[X + 2] = LOG2( store[X + 2] );
@@ -755,11 +720,8 @@ static void XSTDCALL
 pow4f(
    float *store )
 {
-#ifdef WIN32
    const unsigned X = 0;
-#else
-   const unsigned X = TEMP_R0 * 16;
-#endif
+
    store[X + 0] = powf( store[X + 0], store[X + 4] );
    store[X + 1] = powf( store[X + 1], store[X + 5] );
    store[X + 2] = powf( store[X + 2], store[X + 6] );
@@ -800,11 +762,8 @@ static void XSTDCALL
 rsqrt4f(
    float *store )
 {
-#ifdef WIN32
    const unsigned X = 0;
-#else
-   const unsigned X = TEMP_R0 * 16;
-#endif
+
    store[X + 0] = 1.0F / sqrtf( store[X + 0] );
    store[X + 1] = 1.0F / sqrtf( store[X + 1] );
    store[X + 2] = 1.0F / sqrtf( store[X + 2] );
@@ -878,11 +837,8 @@ static void XSTDCALL
 sin4f(
    float *store )
 {
-#ifdef WIN32
    const unsigned X = 0;
-#else
-   const unsigned X = TEMP_R0 * 16;
-#endif
+
    store[X + 0] = sinf( store[X + 0] );
    store[X + 1] = sinf( store[X + 1] );
    store[X + 2] = sinf( store[X + 2] );
@@ -1234,11 +1190,16 @@ emit_instruction(
 
    switch( inst->Instruction.Opcode ) {
    case TGSI_OPCODE_ARL:
+#if 0
+      /* XXX this isn't working properly (see glean vertProg1 test) */
       FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
          FETCH( func, *inst, 0, 0, chan_index );
          emit_f2it( func, 0 );
          STORE( func, *inst, 0, 0, chan_index );
       }
+#else
+      return 0;
+#endif
       break;
 
    case TGSI_OPCODE_MOV:
@@ -2029,40 +1990,40 @@ emit_declaration(
    }
 }
 
-static void aos_to_soa( struct x86_function *func, uint aos, uint soa, uint num, uint stride )
+static void aos_to_soa( struct x86_function *func, 
+                        uint arg_aos,
+                        uint arg_soa, 
+                        uint arg_num, 
+                        uint arg_stride )
 {
-   struct x86_reg soa_input;
-   struct x86_reg aos_input;
-   struct x86_reg num_inputs;
-   struct x86_reg temp;
-   unsigned char *inner_loop;
+   struct x86_reg soa_input = x86_make_reg( file_REG32, reg_AX );
+   struct x86_reg aos_input = x86_make_reg( file_REG32, reg_BX );
+   struct x86_reg num_inputs = x86_make_reg( file_REG32, reg_CX );
+   struct x86_reg stride = x86_make_reg( file_REG32, reg_DX );
+   int inner_loop;
 
-   soa_input = x86_make_reg( file_REG32, reg_AX );
-   aos_input = x86_make_reg( file_REG32, reg_BX );
-   num_inputs = x86_make_reg( file_REG32, reg_CX );
-   temp = x86_make_reg( file_REG32, reg_DX );
 
    /* Save EBX */
    x86_push( func, x86_make_reg( file_REG32, reg_BX ) );
 
-   x86_mov( func, soa_input, get_argument( soa + 1 ) );
-   x86_mov( func, aos_input, get_argument( aos + 1 ) );
-   x86_mov( func, num_inputs, get_argument( num + 1 ) );
+   x86_mov( func, aos_input,  x86_fn_arg( func, arg_aos ) );
+   x86_mov( func, soa_input,  x86_fn_arg( func, arg_soa ) );
+   x86_mov( func, num_inputs, x86_fn_arg( func, arg_num ) );
+   x86_mov( func, stride,     x86_fn_arg( func, arg_stride ) );
 
    /* do */
    inner_loop = x86_get_label( func );
    {
-      x86_mov( func, temp, get_argument( stride + 1 ) );
       x86_push( func, aos_input );
       sse_movlps( func, make_xmm( 0 ), x86_make_disp( aos_input, 0 ) );
       sse_movlps( func, make_xmm( 3 ), x86_make_disp( aos_input, 8 ) );
-      x86_add( func, aos_input, temp );
+      x86_add( func, aos_input, stride );
       sse_movhps( func, make_xmm( 0 ), x86_make_disp( aos_input, 0 ) );
       sse_movhps( func, make_xmm( 3 ), x86_make_disp( aos_input, 8 ) );
-      x86_add( func, aos_input, temp );
+      x86_add( func, aos_input, stride );
       sse_movlps( func, make_xmm( 1 ), x86_make_disp( aos_input, 0 ) );
       sse_movlps( func, make_xmm( 4 ), x86_make_disp( aos_input, 8 ) );
-      x86_add( func, aos_input, temp );
+      x86_add( func, aos_input, stride );
       sse_movhps( func, make_xmm( 1 ), x86_make_disp( aos_input, 0 ) );
       sse_movhps( func, make_xmm( 4 ), x86_make_disp( aos_input, 8 ) );
       x86_pop( func, aos_input );
@@ -2088,7 +2049,7 @@ static void aos_to_soa( struct x86_function *func, uint aos, uint soa, uint num,
    x86_jcc( func, cc_NE, inner_loop );
 
    /* Restore EBX */
-   x86_pop( func, x86_make_reg( file_REG32, reg_BX ) );
+   x86_pop( func, aos_input );
 }
 
 static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num, uint stride )
@@ -2097,7 +2058,7 @@ static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num,
    struct x86_reg aos_output;
    struct x86_reg num_outputs;
    struct x86_reg temp;
-   unsigned char *inner_loop;
+   int inner_loop;
 
    soa_output = x86_make_reg( file_REG32, reg_AX );
    aos_output = x86_make_reg( file_REG32, reg_BX );
@@ -2105,11 +2066,11 @@ static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num,
    temp = x86_make_reg( file_REG32, reg_DX );
 
    /* Save EBX */
-   x86_push( func, x86_make_reg( file_REG32, reg_BX ) );
+   x86_push( func, aos_output );
 
-   x86_mov( func, soa_output, get_argument( soa + 1 ) );
-   x86_mov( func, aos_output, get_argument( aos + 1 ) );
-   x86_mov( func, num_outputs, get_argument( num + 1 ) );
+   x86_mov( func, soa_output, x86_fn_arg( func, soa ) );
+   x86_mov( func, aos_output, x86_fn_arg( func, aos ) );
+   x86_mov( func, num_outputs, x86_fn_arg( func, num ) );
 
    /* do */
    inner_loop = x86_get_label( func );
@@ -2126,7 +2087,7 @@ static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num,
       sse_unpcklps( func, make_xmm( 3 ), make_xmm( 4 ) );
       sse_unpckhps( func, make_xmm( 5 ), make_xmm( 4 ) );
 
-      x86_mov( func, temp, get_argument( stride + 1 ) );
+      x86_mov( func, temp, x86_fn_arg( func, stride ) );
       x86_push( func, aos_output );
       sse_movlps( func, x86_make_disp( aos_output, 0 ), make_xmm( 0 ) );
       sse_movlps( func, x86_make_disp( aos_output, 8 ), make_xmm( 3 ) );
@@ -2150,7 +2111,7 @@ static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num,
    x86_jcc( func, cc_NE, inner_loop );
 
    /* Restore EBX */
-   x86_pop( func, x86_make_reg( file_REG32, reg_BX ) );
+   x86_pop( func, aos_output );
 }
 
 /**
@@ -2185,6 +2146,17 @@ tgsi_emit_sse2(
 
    tgsi_parse_init( &parse, tokens );
 
+   /* Can't just use EDI, EBX without save/restoring them:
+    */
+   x86_push(
+      func,
+      get_immediate_base() );
+
+   x86_push(
+      func,
+      get_temp_base() );
+
+
    /*
     * Different function args for vertex/fragment shaders:
     */
@@ -2193,51 +2165,55 @@ tgsi_emit_sse2(
       x86_mov(
          func,
          get_input_base(),
-         get_argument( 0 ) );
+         x86_fn_arg( func, 1 ) );
       /* skipping outputs argument here */
       x86_mov(
          func,
          get_const_base(),
-         get_argument( 2 ) );
+         x86_fn_arg( func, 3 ) );
       x86_mov(
          func,
          get_temp_base(),
-         get_argument( 3 ) );
+         x86_fn_arg( func, 4 ) );
       x86_mov(
          func,
          get_coef_base(),
-         get_argument( 4 ) );
+         x86_fn_arg( func, 5 ) );
       x86_mov(
          func,
          get_immediate_base(),
-         get_argument( 5 ) );
+         x86_fn_arg( func, 6 ) );
    }
    else {
       assert(parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX);
 
       if (do_swizzles)
-         aos_to_soa( func, 5, 0, 6, 7 );
+         aos_to_soa( func, 
+                     6,         /* aos_input */
+                     1,         /* machine->input */
+                     7,         /* num_inputs */
+                     8 );       /* input_stride */
 
       x86_mov(
          func,
          get_input_base(),
-         get_argument( 0 ) );
+         x86_fn_arg( func, 1 ) );
       x86_mov(
          func,
          get_output_base(),
-         get_argument( 1 ) );
+         x86_fn_arg( func, 2 ) );
       x86_mov(
          func,
          get_const_base(),
-         get_argument( 2 ) );
+         x86_fn_arg( func, 3 ) );
       x86_mov(
          func,
          get_temp_base(),
-         get_argument( 3 ) );
+         x86_fn_arg( func, 4 ) );
       x86_mov(
          func,
          get_immediate_base(),
-         get_argument( 4 ) );
+         x86_fn_arg( func, 5 ) );
    }
 
    while( !tgsi_parse_end_of_tokens( &parse ) && ok ) {
@@ -2260,7 +2236,7 @@ tgsi_emit_sse2(
                x86_mov(
                   func,
                   get_output_base(),
-                  get_argument( 1 ) );
+                  x86_fn_arg( func, 2 ) );
             }
          }
 
@@ -2307,9 +2283,19 @@ tgsi_emit_sse2(
 
    if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX) {
       if (do_swizzles)
-         soa_to_aos( func, 8, 1, 9, 10 );
+         soa_to_aos( func, 9, 2, 10, 11 );
    }
 
+   /* Can't just use EBX, EDI without save/restoring them:
+    */
+   x86_pop(
+      func,
+      get_temp_base() );
+
+   x86_pop(
+      func,
+      get_immediate_base() );
+
 #ifdef WIN32
    emit_retw( func, 16 );
 #else
diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c
index 26bfc2051f2..4c65ffd7807 100644
--- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c
+++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c
@@ -767,6 +767,31 @@ dump_instruction_short(
       SID( dst->DstRegister.Index );
       CHR( ']' );
 
+      switch (dst->DstRegisterExtModulate.Modulate) {
+      case TGSI_MODULATE_1X:
+         break;
+      case TGSI_MODULATE_2X:
+         TXT( "_2X" );
+         break;
+      case TGSI_MODULATE_4X:
+         TXT( "_4X" );
+         break;
+      case TGSI_MODULATE_8X:
+         TXT( "_8X" );
+         break;
+      case TGSI_MODULATE_HALF:
+         TXT( "_D2" );
+         break;
+      case TGSI_MODULATE_QUARTER:
+         TXT( "_D4" );
+         break;
+      case TGSI_MODULATE_EIGHTH:
+         TXT( "_D8" );
+         break;
+      default:
+         assert( 0 );
+      }
+
       if( dst->DstRegister.WriteMask != TGSI_WRITEMASK_XYZW ) {
          CHR( '.' );
          if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_X ) {
diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_parse.c b/src/gallium/auxiliary/tgsi/util/tgsi_parse.c
index c3526cb71ff..5bea7738401 100644
--- a/src/gallium/auxiliary/tgsi/util/tgsi_parse.c
+++ b/src/gallium/auxiliary/tgsi/util/tgsi_parse.c
@@ -43,7 +43,7 @@ tgsi_full_token_free(
    union tgsi_full_token *full_token )
 {
    if( full_token->Token.Type == TGSI_TOKEN_TYPE_IMMEDIATE ) {
-      FREE( full_token->FullImmediate.u.Pointer );
+      FREE( (void *) full_token->FullImmediate.u.Pointer );
    }
 }
 
@@ -156,7 +156,7 @@ tgsi_parse_token(
          imm->u.Pointer = MALLOC(
             sizeof( struct tgsi_immediate_float32 ) * (imm->Immediate.Size - 1) );
          for( i = 0; i < imm->Immediate.Size - 1; i++ ) {
-            next_token( ctx, &imm->u.ImmediateFloat32[i] );
+            next_token( ctx, (struct tgsi_immediate_float32 *) &imm->u.ImmediateFloat32[i] );
          }
          break;
 
diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_parse.h b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h
index da0121c482e..15e76feb7ca 100644
--- a/src/gallium/auxiliary/tgsi/util/tgsi_parse.h
+++ b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h
@@ -52,8 +52,8 @@ struct tgsi_full_immediate
    struct tgsi_immediate   Immediate;
    union
    {
-      void                          *Pointer;
-      struct tgsi_immediate_float32 *ImmediateFloat32;
+      const void                          *Pointer;
+      const struct tgsi_immediate_float32 *ImmediateFloat32;
    } u;
 };
 
diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c
index ea4a72967d6..65650ed22a4 100644
--- a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c
+++ b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c
@@ -103,18 +103,14 @@ tgsi_scan_shader(const struct tgsi_token *tokens,
                info->file_max[file] = MAX2(info->file_max[file], (int)i);
 
                if (file == TGSI_FILE_INPUT) {
-                  info->input_semantic_name[info->num_inputs]
-                     = (ubyte)fulldecl->Semantic.SemanticName;
-                  info->input_semantic_index[info->num_inputs]
-                     = (ubyte)fulldecl->Semantic.SemanticIndex;
+                  info->input_semantic_name[i] = (ubyte)fulldecl->Semantic.SemanticName;
+                  info->input_semantic_index[i] = (ubyte)fulldecl->Semantic.SemanticIndex;
                   info->num_inputs++;
                }
 
                if (file == TGSI_FILE_OUTPUT) {
-                  info->output_semantic_name[info->num_outputs]
-                     = (ubyte)fulldecl->Semantic.SemanticName;
-                  info->output_semantic_index[info->num_outputs]
-                     = (ubyte)fulldecl->Semantic.SemanticIndex;
+                  info->output_semantic_name[i] = (ubyte)fulldecl->Semantic.SemanticName;
+                  info->output_semantic_index[i] = (ubyte)fulldecl->Semantic.SemanticIndex;
                   info->num_outputs++;
                }
 
@@ -137,6 +133,9 @@ tgsi_scan_shader(const struct tgsi_token *tokens,
       }
    }
 
+   assert( info->file_max[TGSI_FILE_INPUT] + 1 == info->num_inputs );
+   assert( info->file_max[TGSI_FILE_OUTPUT] + 1 == info->num_outputs );
+
    info->uses_kill = (info->opcode_count[TGSI_OPCODE_KIL] ||
                       info->opcode_count[TGSI_OPCODE_KILP]);
 
diff --git a/src/gallium/auxiliary/translate/translate_sse.c b/src/gallium/auxiliary/translate/translate_sse.c
index f590d48b787..a54ac5a82ff 100644
--- a/src/gallium/auxiliary/translate/translate_sse.c
+++ b/src/gallium/auxiliary/translate/translate_sse.c
@@ -404,7 +404,7 @@ static boolean build_vertex_emit( struct translate_sse *p,
    struct x86_reg srcEAX       = x86_make_reg(file_REG32, reg_CX);
    struct x86_reg countEBP     = x86_make_reg(file_REG32, reg_BP);
    struct x86_reg translateESI = x86_make_reg(file_REG32, reg_SI);
-   uint8_t *fixup, *label;
+   int fixup, label;
    unsigned j;
 
    p->func = func;
diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c
index f1fb07bf5b9..4ec17466624 100644
--- a/src/gallium/auxiliary/util/p_debug.c
+++ b/src/gallium/auxiliary/util/p_debug.c
@@ -422,4 +422,4 @@ void debug_print_format(const char *msg, unsigned fmt )
 
    debug_printf("%s: %s\n", msg, fmtstr); 
 }
-#endif
\ No newline at end of file
+#endif
diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c
index 257473ab26f..b70bcbfa66d 100644
--- a/src/gallium/auxiliary/util/u_blit.c
+++ b/src/gallium/auxiliary/util/u_blit.c
@@ -296,6 +296,8 @@ util_blit_pixels(struct blit_state *ctx,
                       src, srcLeft, srcTop, /* src */
                       srcW, srcH);     /* size */
 
+   pipe->texture_update(pipe, tex, 0, 1 << 0);
+
    /* save state (restored below) */
    cso_save_blend(ctx->cso);
    cso_save_depth_stencil_alpha(ctx->cso);
diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c
index 6ed5503c9ae..056ae829aee 100644
--- a/src/gallium/auxiliary/util/u_gen_mipmap.c
+++ b/src/gallium/auxiliary/util/u_gen_mipmap.c
@@ -505,6 +505,9 @@ format_to_type_comps(enum pipe_format pformat,
       return;
    default:
       assert(0);
+      *datatype = UBYTE;
+      *comps = 0;
+      break;
    }
 }
 
diff --git a/src/gallium/drivers/i915simple/i915_debug_fp.c b/src/gallium/drivers/i915simple/i915_debug_fp.c
index 37a3508fe14..c024a051a58 100644
--- a/src/gallium/drivers/i915simple/i915_debug_fp.c
+++ b/src/gallium/drivers/i915simple/i915_debug_fp.c
@@ -333,12 +333,11 @@ void
 i915_disassemble_program(struct debug_stream *stream, 
 			 const unsigned * program, unsigned sz)
 {
-   unsigned size = program[0] & 0x1ff;
    unsigned i;
 
    PRINTF(stream, "\t\tBEGIN\n");
 
-   assert(size + 2 == sz);
+   assert((program[0] & 0x1ff) + 2 == sz);
 
    program++;
    for (i = 1; i < sz; i += 3, program += 3) {
diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c
index dcd349e478b..ba8f183bdf0 100644
--- a/src/gallium/drivers/i915simple/i915_screen.c
+++ b/src/gallium/drivers/i915simple/i915_screen.c
@@ -182,6 +182,7 @@ i915_is_format_supported( struct pipe_screen *screen,
       break;
    default:
       assert(0);
+      return FALSE;
    }
 
    for (i = 0; list[i] != PIPE_FORMAT_NONE; i++) {
diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c
index ee74826763c..2af0db37143 100644
--- a/src/gallium/drivers/softpipe/sp_context.c
+++ b/src/gallium/drivers/softpipe/sp_context.c
@@ -122,7 +122,7 @@ static void softpipe_destroy( struct pipe_context *pipe )
 struct pipe_context *
 softpipe_create( struct pipe_screen *screen,
                  struct pipe_winsys *pipe_winsys,
-                 struct softpipe_winsys *softpipe_winsys )
+                 void *unused )
 {
    struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context);
    uint i;
@@ -212,8 +212,6 @@ softpipe_create( struct pipe_screen *screen,
    softpipe->quad.colormask = sp_quad_colormask_stage(softpipe);
    softpipe->quad.output = sp_quad_output_stage(softpipe);
 
-   softpipe->winsys = softpipe_winsys;
-
    /*
     * Create drawing context and plug our rendering stage into it.
     */
diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h
index b3e2b2e4355..62eabfb30e8 100644
--- a/src/gallium/drivers/softpipe/sp_context.h
+++ b/src/gallium/drivers/softpipe/sp_context.h
@@ -57,8 +57,6 @@ struct sp_vertex_shader;
 
 struct softpipe_context {
    struct pipe_context pipe;  /**< base class */
-   struct softpipe_winsys *winsys;	/**< window system interface */
-
 
    /* The most recent drawing state as set by the driver:
     */
diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c
index 2921066ce36..9e77b7e91bc 100644
--- a/src/gallium/drivers/softpipe/sp_state_fs.c
+++ b/src/gallium/drivers/softpipe/sp_state_fs.c
@@ -82,10 +82,9 @@ softpipe_bind_fs_state(struct pipe_context *pipe, void *fs)
 void
 softpipe_delete_fs_state(struct pipe_context *pipe, void *fs)
 {
-   struct softpipe_context *softpipe = softpipe_context(pipe);
    struct sp_fragment_shader *state = fs;
 
-   assert(fs != softpipe->fs);
+   assert(fs != softpipe_context(pipe)->fs);
    
    state->delete( state );
 }
diff --git a/src/gallium/drivers/softpipe/sp_winsys.h b/src/gallium/drivers/softpipe/sp_winsys.h
index 291825dfe22..4ab666486c4 100644
--- a/src/gallium/drivers/softpipe/sp_winsys.h
+++ b/src/gallium/drivers/softpipe/sp_winsys.h
@@ -59,7 +59,7 @@ struct pipe_context;
 
 struct pipe_context *softpipe_create( struct pipe_screen *,
                                       struct pipe_winsys *,
-				      struct softpipe_winsys * );
+				      void *unused );
 
 
 struct pipe_screen *
diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c
index fd2f56eff24..8ca90ef4c6c 100644
--- a/src/gallium/winsys/xlib/xm_winsys.c
+++ b/src/gallium/winsys/xlib/xm_winsys.c
@@ -37,6 +37,7 @@
 #include "xmesaP.h"
 
 #undef ASSERT
+#undef Elements
 
 #include "pipe/p_winsys.h"
 #include "pipe/p_format.h"
@@ -88,18 +89,6 @@ struct xmesa_surface
 };
 
 
-/**
- * Derived from softpipe_winsys.
- * We just need one extra field which indicates the pixel format to use for
- * drawing surfaces so that we're compatible with the XVisual/window format.
- */
-struct xmesa_softpipe_winsys
-{
-   struct softpipe_winsys spws;
-   enum pipe_format pixelformat;
-};
-
-
 struct xmesa_pipe_winsys
 {
    struct pipe_winsys base;
@@ -119,12 +108,7 @@ xmesa_surface(struct pipe_surface *ps)
    return (struct xmesa_surface *) ps;
 }
 
-/** cast wrapper */
-static INLINE struct xmesa_softpipe_winsys *
-xmesa_softpipe_winsys(struct softpipe_winsys *spws)
-{
-   return (struct xmesa_softpipe_winsys *) spws;
-}
+
 
 /**
  * Turn the softpipe opaque buffer pointer into a dri_bufmgr opaque
@@ -526,8 +510,7 @@ xm_surface_alloc_storage(struct pipe_winsys *winsys,
 
 
 /**
- * Called via pipe->surface_alloc() to create new surfaces (textures,
- * renderbuffers, etc.
+ * Called via winsys->surface_alloc() to create new surfaces.
  */
 static struct pipe_surface *
 xm_surface_alloc(struct pipe_winsys *ws)
@@ -612,10 +595,19 @@ xmesa_get_pipe_winsys_aub(struct xmesa_visual *xm_vis)
 {
    static struct xmesa_pipe_winsys *ws = NULL;
 
-   if (!ws && getenv("XM_AUB")) {
+   if (!ws) {
       ws = (struct xmesa_pipe_winsys *) xmesa_create_pipe_winsys_aub();
    }
-   else if (!ws) {
+   return &ws->base;
+}
+
+
+struct pipe_winsys *
+xmesa_get_pipe_winsys(struct xmesa_visual *xm_vis)
+{
+   static struct xmesa_pipe_winsys *ws = NULL;
+
+   if (!ws) {
       ws = CALLOC_STRUCT(xmesa_pipe_winsys);
 
       ws->xm_visual = xm_vis;
@@ -646,45 +638,19 @@ xmesa_get_pipe_winsys_aub(struct xmesa_visual *xm_vis)
 }
 
 
-/**
- * Called via softpipe_winsys->is_format_supported().
- * This function is only called to test formats for front/back color surfaces.
- * The winsys being queried will have been created at glXCreateContext
- * time, with a pixel format corresponding to the context's visual.
- */
-static boolean
-xmesa_is_format_supported(struct softpipe_winsys *sws,
-                          enum pipe_format format)
-{
-   struct xmesa_softpipe_winsys *xmws = xmesa_softpipe_winsys(sws);
-   return (format == xmws->pixelformat);
-}
-
-
-/**
- * Return pointer to a softpipe_winsys object.
- */
-static struct softpipe_winsys *
-xmesa_get_softpipe_winsys(uint pixelformat)
-{
-   struct xmesa_softpipe_winsys *xmws
-      = CALLOC_STRUCT(xmesa_softpipe_winsys);
-   if (!xmws)
-      return NULL;
-
-   xmws->spws.is_format_supported = xmesa_is_format_supported;
-   xmws->pixelformat = pixelformat;
-
-   return &xmws->spws;
-}
-
-
 struct pipe_context *
 xmesa_create_pipe_context(XMesaContext xmesa, uint pixelformat)
 {
-   struct pipe_winsys *pws = xmesa_get_pipe_winsys_aub(xmesa->xm_visual);
+   struct pipe_winsys *pws;
    struct pipe_context *pipe;
    
+   if (getenv("XM_AUB")) {
+      pws = xmesa_get_pipe_winsys_aub(xmesa->xm_visual);
+   }
+   else {
+      pws = xmesa_get_pipe_winsys(xmesa->xm_visual);
+   }
+
 #ifdef GALLIUM_CELL
    if (!getenv("GALLIUM_NOCELL")) {
       struct cell_winsys *cws = cell_get_winsys(pixelformat);
@@ -695,10 +661,9 @@ xmesa_create_pipe_context(XMesaContext xmesa, uint pixelformat)
    else
 #endif
    {
-      struct softpipe_winsys *spws = xmesa_get_softpipe_winsys(pixelformat);
       struct pipe_screen *screen = softpipe_create_screen(pws);
 
-      pipe = softpipe_create(screen, pws, spws);
+      pipe = softpipe_create(screen, pws, NULL);
    }
 
    if (pipe)