Enable USER_BUFFERS, RADEON_VTXFMT_A and HW_VBOS on big endian by forcing all elts

to 32 bits (it would be possible to do a half word swap for 16 bits if somebody is interested) and by using the input route mecanism to swap ubyte's properly in r300EmitArrays. Tested with cva, varray and quake3

4 files changed, 101 insertions, 43 deletions

diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index 21a474bf67..2dd6a6c335 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -384,6 +384,7 @@ static void r300FreeGartAllocations(r300ContextPtr r300)
 	resize_u_list(r300);
 #endif
 
+#ifdef USER_BUFFERS
     for (i = r300->rmm->u_last + 1; i > 0; i--) {
 	if (r300->rmm->u_list[i].ptr == NULL) {
 	    continue;
@@ -417,6 +418,7 @@ static void r300FreeGartAllocations(r300ContextPtr r300)
 	}
     }
     r300->rmm->u_head = i;
+#endif /* USER_BUFFERS */
 }
 
 /* Destroy the device specific context.
diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index 58f789eeae..65bf9964df 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -47,11 +47,13 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "colormac.h"
 #include "radeon_context.h"
 
-/* PPC doesnt support 16 bit elts ... */
-#ifndef MESA_BIG_ENDIAN
 #define USER_BUFFERS
 #define RADEON_VTXFMT_A
 #define HW_VBOS
+
+/* We don't handle 16 bits elts swapping yet */
+#ifdef MESA_BIG_ENDIAN
+#define FORCE_32BITS_ELTS
 #endif
 
 //#define OPTIMIZE_ELTS
diff --git a/src/mesa/drivers/dri/r300/r300_maos.c b/src/mesa/drivers/dri/r300/r300_maos.c
index 290ffb44ea..6b8365e6d9 100644
--- a/src/mesa/drivers/dri/r300/r300_maos.c
+++ b/src/mesa/drivers/dri/r300/r300_maos.c
@@ -252,6 +252,43 @@ void r300EmitElts(GLcontext * ctx, void *elts, unsigned long n_elts, int elt_siz
 	memcpy(out, elts, n_elts * elt_size);
 }
 
+	/* Mesa assumes that all missing components are from (0, 0, 0, 1) */
+#define ALL_COMPONENTS ((R300_INPUT_ROUTE_SELECT_X<<R300_INPUT_ROUTE_X_SHIFT) \
+		| (R300_INPUT_ROUTE_SELECT_Y<<R300_INPUT_ROUTE_Y_SHIFT) \
+		| (R300_INPUT_ROUTE_SELECT_Z<<R300_INPUT_ROUTE_Z_SHIFT) \
+		| (R300_INPUT_ROUTE_SELECT_W<<R300_INPUT_ROUTE_W_SHIFT))
+
+#define ALL_DEFAULT ((R300_INPUT_ROUTE_SELECT_ZERO<<R300_INPUT_ROUTE_X_SHIFT) \
+		| (R300_INPUT_ROUTE_SELECT_ZERO<<R300_INPUT_ROUTE_Y_SHIFT) \
+		| (R300_INPUT_ROUTE_SELECT_ZERO<<R300_INPUT_ROUTE_Z_SHIFT) \
+		| (R300_INPUT_ROUTE_SELECT_ONE<<R300_INPUT_ROUTE_W_SHIFT))
+
+
+static GLuint t_comps(GLuint aos_size)
+{
+	GLuint mask;
+	mask = (1 << (aos_size*3)) - 1;
+	return (ALL_COMPONENTS & mask) | (ALL_DEFAULT & ~mask);
+}
+
+static GLuint fix_comps(GLuint dw, int fmt)
+{	
+#ifdef MESA_BIG_ENDIAN
+	if (fmt == 2) {
+		GLuint dw_temp = 0;
+
+		dw_temp |= ((dw >> R300_INPUT_ROUTE_X_SHIFT) & R300_INPUT_ROUTE_SELECT_MASK) << R300_INPUT_ROUTE_W_SHIFT;
+		dw_temp |= ((dw >> R300_INPUT_ROUTE_Y_SHIFT) & R300_INPUT_ROUTE_SELECT_MASK) << R300_INPUT_ROUTE_Z_SHIFT;
+		dw_temp |= ((dw >> R300_INPUT_ROUTE_Z_SHIFT) & R300_INPUT_ROUTE_SELECT_MASK) << R300_INPUT_ROUTE_Y_SHIFT;
+		dw_temp |= ((dw >> R300_INPUT_ROUTE_W_SHIFT) & R300_INPUT_ROUTE_SELECT_MASK) << R300_INPUT_ROUTE_X_SHIFT;
+		
+		return dw_temp;
+	}
+#endif /* MESA_BIG_ENDIAN */
+	return dw;
+		
+}
+
 /* Emit vertex data to GART memory (unless immediate mode)
  * Route inputs to the vertex processor
  */
@@ -264,7 +301,7 @@ void r300EmitArrays(GLcontext * ctx, GLboolean immd)
 	//struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
 	GLuint nr = 0;
 	GLuint count = VB->Count;
-	GLuint dw,mask;
+	GLuint dw;
 	GLuint vic_1 = 0;	/* R300_VAP_INPUT_CNTL_1 */
 	GLuint aa_vap_reg = 0; /* VAP register assignment */
 	GLuint i;
@@ -469,17 +506,6 @@ void r300EmitArrays(GLcontext * ctx, GLboolean immd)
 	((drm_r300_cmd_header_t*)r300->hw.vir[0].cmd)->packet0.count = (nr+1)>>1;
 
 
-	/* Mesa assumes that all missing components are from (0, 0, 0, 1) */
-#define ALL_COMPONENTS ((R300_INPUT_ROUTE_SELECT_X<<R300_INPUT_ROUTE_X_SHIFT) \
-		| (R300_INPUT_ROUTE_SELECT_Y<<R300_INPUT_ROUTE_Y_SHIFT) \
-		| (R300_INPUT_ROUTE_SELECT_Z<<R300_INPUT_ROUTE_Z_SHIFT) \
-		| (R300_INPUT_ROUTE_SELECT_W<<R300_INPUT_ROUTE_W_SHIFT))
-
-#define ALL_DEFAULT ((R300_INPUT_ROUTE_SELECT_ZERO<<R300_INPUT_ROUTE_X_SHIFT) \
-		| (R300_INPUT_ROUTE_SELECT_ZERO<<R300_INPUT_ROUTE_Y_SHIFT) \
-		| (R300_INPUT_ROUTE_SELECT_ZERO<<R300_INPUT_ROUTE_Z_SHIFT) \
-		| (R300_INPUT_ROUTE_SELECT_ONE<<R300_INPUT_ROUTE_W_SHIFT))
-
 	R300_STATECHANGE(r300, vir[1]);
 
 	for(i=0; i < nr; i++)
@@ -488,33 +514,22 @@ void r300EmitArrays(GLcontext * ctx, GLboolean immd)
 			r300->state.aos[i].aos_size=/*3*/4; /* XXX */
 		}
 		
-		
-	for(i=0;i+1<nr;i+=2){
+	for (i=0;i+1<nr;i+=2) {
 		/* do i first.. */
-		mask=(1<<(r300->state.aos[i].aos_size*3))-1;
-		dw=(ALL_COMPONENTS & mask)
-		| (ALL_DEFAULT & ~mask)
-		| R300_INPUT_ROUTE_ENABLE;
-
+		dw = fix_comps(t_comps(r300->state.aos[i].aos_size), r300->state.aos[i].aos_format) | R300_INPUT_ROUTE_ENABLE;
 		/* i+1 */
-		mask=(1<<(r300->state.aos[i+1].aos_size*3))-1;
-		dw|=(
-		(ALL_COMPONENTS & mask)
-		| (ALL_DEFAULT & ~mask)
-		| R300_INPUT_ROUTE_ENABLE
-		)<<16;
-
+		dw |= (fix_comps(t_comps(r300->state.aos[i+1].aos_size), r300->state.aos[i+1].aos_format) | R300_INPUT_ROUTE_ENABLE) << 16;
+		
 		//fprintf(stderr, "vir1 dw=%08x\n", dw);
 		r300->hw.vir[1].cmd[R300_VIR_CNTL_0+(i>>1)]=dw;
-		}
-	if(nr & 1){
-		mask=(1<<(r300->state.aos[nr-1].aos_size*3))-1;
-		dw=(ALL_COMPONENTS & mask)
-		| (ALL_DEFAULT & ~mask)
-		| R300_INPUT_ROUTE_ENABLE;
-		r300->hw.vir[1].cmd[R300_VIR_CNTL_0+(nr>>1)]=dw;
+	}
+	if (nr & 1) {
+		dw = fix_comps(t_comps(r300->state.aos[nr-1].aos_size), r300->state.aos[nr-1].aos_format) | R300_INPUT_ROUTE_ENABLE;
+		
 		//fprintf(stderr, "vir1 dw=%08x\n", dw);
-		}
+		r300->hw.vir[1].cmd[R300_VIR_CNTL_0+(nr>>1)]=dw;
+	}
+
 	/* Set the rest of INPUT_ROUTE_1 to 0 */
 	//for(i=((count+1)>>1); i<8; i++)r300->hw.vir[1].cmd[R300_VIR_CNTL_0+i]=0x0;
 	((drm_r300_cmd_header_t*)r300->hw.vir[1].cmd)->packet0.count = (nr+1)>>1;
diff --git a/src/mesa/drivers/dri/r300/radeon_vtxfmt_a.c b/src/mesa/drivers/dri/r300/radeon_vtxfmt_a.c
index 8f7620482f..941ff0078e 100644
--- a/src/mesa/drivers/dri/r300/radeon_vtxfmt_a.c
+++ b/src/mesa/drivers/dri/r300/radeon_vtxfmt_a.c
@@ -216,14 +216,22 @@ static void radeonDrawElements( GLenum mode, GLsizei count, GLenum type, const G
 				max = ((unsigned char *)indices)[i];
 		}
 		
+#ifdef FORCE_32BITS_ELTS
+		elt_size = 4;
+#else
 		elt_size = 2;
-		
+#endif		
 		r300AllocDmaRegion(rmesa, &rvb, count * elt_size, elt_size);
 		rvb.aos_offset = GET_START(&rvb);
 		ptr = rvb.address + rvb.start;
 			
+#ifdef FORCE_32BITS_ELTS
+		for (i=0; i < count; i++)
+			((unsigned int *)ptr)[i] = ((unsigned char *)indices)[i] - min;
+#else
 		for (i=0; i < count; i++)
 			((unsigned short int *)ptr)[i] = ((unsigned char *)indices)[i] - min;
+#endif
 	break;
 		
 	case GL_UNSIGNED_SHORT:
@@ -234,14 +242,23 @@ static void radeonDrawElements( GLenum mode, GLsizei count, GLenum type, const G
 				max = ((unsigned short int *)indices)[i];
 		}
 		
+#ifdef FORCE_32BITS_ELTS
+		elt_size = 4;
+#else
 		elt_size = 2;
+#endif
 		
 		r300AllocDmaRegion(rmesa, &rvb, count * elt_size, elt_size);
 		rvb.aos_offset = GET_START(&rvb);
 		ptr = rvb.address + rvb.start;
 		
+#ifdef FORCE_32BITS_ELTS
+		for (i=0; i < count; i++)
+			((unsigned int *)ptr)[i] = ((unsigned short int *)indices)[i] - min;
+#else
 		for (i=0; i < count; i++)
 			((unsigned short int *)ptr)[i] = ((unsigned short int *)indices)[i] - min;
+#endif
 	break;
 	
 	case GL_UNSIGNED_INT:
@@ -252,17 +269,20 @@ static void radeonDrawElements( GLenum mode, GLsizei count, GLenum type, const G
 				max = ((unsigned int *)indices)[i];
 		}
 		
+#ifdef FORCE_32BITS_ELTS
+		elt_size = 4;
+#else
 		if (max - min <= 65535)
 			elt_size = 2;
 		else 
 			elt_size = 4;
-		
+#endif
 		r300AllocDmaRegion(rmesa, &rvb, count * elt_size, elt_size);
 		rvb.aos_offset = GET_START(&rvb);
 		ptr = rvb.address + rvb.start;
 		
 		
-		if (max - min <= 65535)
+		if (elt_size == 2)
 			for (i=0; i < count; i++)
 				((unsigned short int *)ptr)[i] = ((unsigned int *)indices)[i] - min;
 		else
@@ -388,19 +408,30 @@ static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei
 	memset(&rvb, 0, sizeof(rvb));
 	switch (type){
 	case GL_UNSIGNED_BYTE:
+#ifdef FORCE_32BITS_ELTS
+		elt_size = 4;
+#else
 		elt_size = 2;
-		
+#endif	
 		r300AllocDmaRegion(rmesa, &rvb, count * elt_size, elt_size);
 		rvb.aos_offset = GET_START(&rvb);
 		ptr = rvb.address + rvb.start;
 		
+#ifdef FORCE_32BITS_ELTS
+		for(i=0; i < count; i++)
+			((unsigned int *)ptr)[i] = ((unsigned char *)indices)[i] - min;
+#else
 		for(i=0; i < count; i++)
 			((unsigned short int *)ptr)[i] = ((unsigned char *)indices)[i] - min;
+#endif
 	break;
 	
 	case GL_UNSIGNED_SHORT:
+#ifdef FORCE_32BITS_ELTS
+		elt_size = 4;
+#else
 		elt_size = 2;
-		
+#endif	
 #ifdef OPTIMIZE_ELTS
 		if (min == 0 && ctx->Array.ElementArrayBufferObj->Name){
 			ptr = indices;
@@ -411,21 +442,29 @@ static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei
 		rvb.aos_offset = GET_START(&rvb);
 		ptr = rvb.address + rvb.start;
 
+#ifdef FORCE_32BITS_ELTS
+		for(i=0; i < count; i++)
+			((unsigned int *)ptr)[i] = ((unsigned short int *)indices)[i] - min;
+#else
 		for(i=0; i < count; i++)
 			((unsigned short int *)ptr)[i] = ((unsigned short int *)indices)[i] - min;
+#endif
 	break;
 	
 	case GL_UNSIGNED_INT:
+#ifdef FORCE_32BITS_ELTS
+		elt_size = 4;
+#else
 		if (max - min <= 65535)
 			elt_size = 2;
 		else 
 			elt_size = 4;
-		
+#endif	
 		r300AllocDmaRegion(rmesa, &rvb, count * elt_size, elt_size);
 		rvb.aos_offset = GET_START(&rvb);
 		ptr = rvb.address + rvb.start;
 		
-		if (max - min <= 65535)
+		if (elt_size == 2)
 			for (i=0; i < count; i++)
 				((unsigned short int *)ptr)[i] = ((unsigned int *)indices)[i] - min;
 		else