Use the faster span read/write template for 16bpp
[mesa.git] / src / mesa / drivers / dri / unichrome / via_span.c
index fcb977187587e26ec61c046db0c80ed1b63fb97e..7a4eab6b85df0b155de9adc0c9b4baff5eff4761 100644 (file)
 #include "swrast/swrast.h"
 
 #define DBG 0
-#define LOCAL_VARS                                      \
-    __DRIdrawablePrivate *dPriv = vmesa->driDrawable;   \
-    viaScreenPrivate *viaScreen = vmesa->viaScreen;     \
-    GLuint pitch = vmesa->drawPitch;                    \
-    GLuint height = dPriv->h;                           \
-    GLushort p;                                         \
-    char *buf = (char *)(vmesa->drawMap +               \
-       dPriv->x * viaScreen->bytesPerPixel +           \
-                         dPriv->y * pitch);             \
-    char *read_buf = (char *)(vmesa->readMap +          \
-        dPriv->x * viaScreen->bytesPerPixel +           \
-                              dPriv->y * pitch);        \
-    (void)read_buf; (void)buf; (void)p
-
-#define LOCAL_DEPTH_VARS                                \
-    __DRIdrawablePrivate *dPriv = vmesa->driDrawable;   \
-    viaScreenPrivate *viaScreen = vmesa->viaScreen;     \
-    GLuint pitch = viaScreen->backPitch;                \
-    GLuint height = dPriv->h;                           \
-    char *buf = (char *)(vmesa->depth.map +             \
-                         dPriv->x * 2 +                 \
-                         dPriv->y * pitch)
 
 #define CLIPPIXEL(_x,_y) (_x >= minx && _x < maxx &&    \
                           _y >= miny && _y < maxy)
 
 #define Y_FLIP(_y) (height - _y - 1)
 
-#define HW_LOCK()                               \
-    viaContextPtr vmesa = VIA_CONTEXT(ctx);     \
-    LOCK_HARDWARE_QUIESCENT(vmesa);
-
-/*=* [DBG] csmash saam : bitmap option menu can't be drawn in saam *=*/        
-/*#define HW_CLIPLOOP()                                             \
-    do {                                                            \
-        __DRIdrawablePrivate *dPriv = vmesa->driDrawable;           \
-        int _nc = dPriv->numClipRects;                              \
-        while (_nc--) {                                             \
-            int minx = dPriv->pClipRects[_nc].x1 - dPriv->x;        \
-            int miny = dPriv->pClipRects[_nc].y1 - dPriv->y;        \
-            int maxx = dPriv->pClipRects[_nc].x2 - dPriv->x;        \
-            int maxy = dPriv->pClipRects[_nc].y2 - dPriv->y;*/
-#define HW_CLIPLOOP()                                               \
-    do {                                                            \
-        __DRIdrawablePrivate *dPriv = vmesa->driDrawable;           \
-        int _nc = dPriv->numClipRects;                              \
-       GLuint scrn = vmesa->saam & S_MASK;                         \
-       if(scrn == S1) _nc = 1;                                     \
-        while (_nc--) {                                             \
-           int minx;                                               \
-           int miny;                                               \
-           int maxx;                                               \
-           int maxy;                                               \
-           if (!vmesa->saam) {                                     \
-               minx = dPriv->pClipRects[_nc].x1 - dPriv->x;        \
-               miny = dPriv->pClipRects[_nc].y1 - dPriv->y;        \
-               maxx = dPriv->pClipRects[_nc].x2 - dPriv->x;        \
-               maxy = dPriv->pClipRects[_nc].y2 - dPriv->y;        \
-           }                                                       \
-           else {                                                  \
-               minx = -10000;                                      \
-               miny = -10000;                                      \
-               maxx = 10000;                                       \
-               maxy = 10000;                                       \
-           }
-           
-           /*else if (scrn == S0) {                                \
-               minx = dPriv->pClipRects[_nc].x1 - dPriv->x;        \
-               miny = dPriv->pClipRects[_nc].y1 - dPriv->y;        \
-               maxx = dPriv->pClipRects[_nc].x2 - dPriv->x;        \
-               maxy = dPriv->pClipRects[_nc].y2 - dPriv->y;        \
-           }                                                       \
-           else if (scrn == S1) {                                  \
-               drm_clip_rect_t *b = vmesa->sarea->boxes;           \
-               minx = b->x1;                                       \
-               miny = b->y1;                                       \
-               maxx = b->x2;                                       \
-               maxy = b->y2;                                       \
-           }                                                       \
-           else {                                                  \
-               drm_clip_rect_t *b = vmesa->sarea->boxes + vmesa->numClipRects;\
-               minx = b->x1;        \
-               miny = b->y1;        \
-               maxx = b->x2;        \
-               maxy = b->y2;        \
-           }*/
+#define HW_LOCK() 
+#define HW_CLIPLOOP()                                                          \
+    do {                                                                       \
+        __DRIdrawablePrivate *dPriv = vmesa->driDrawable;                      \
+        int _nc = dPriv->numClipRects;                                         \
+        while (_nc--) {                                                                \
+               int minx = dPriv->pClipRects[_nc].x1 - dPriv->x;                \
+               int miny = dPriv->pClipRects[_nc].y1 - dPriv->y;                \
+               int maxx = dPriv->pClipRects[_nc].x2 - dPriv->x;                \
+               int maxy = dPriv->pClipRects[_nc].y2 - dPriv->y;        
+
 
 #define HW_ENDCLIPLOOP()                                            \
         }                                                           \
     } while (0)
 
-#define HW_UNLOCK()                                    \
-    UNLOCK_HARDWARE(vmesa);
-
-
-/* 16 bit, 565 rgb color spanline and pixel functions
- */
-/*=* [DBG] csmash : fix options worng position *=*/
-/*#define LOCAL_VARS                                    \
-    __DRIdrawablePrivate *dPriv = vmesa->driDrawable;   \
-    GLuint pitch = vmesa->drawPitch;                    \
-    GLuint height = dPriv->h;                           \
-    GLushort p;                                         \
-    char *buf = (char *)(vmesa->drawMap +               \
-                         dPriv->x * 2 +                 \
-                         dPriv->y * pitch);             \
-    char *read_buf = (char *)(vmesa->readMap +          \
-                              dPriv->x * 2 +            \
-                              dPriv->y * pitch);        \
-    (void)read_buf; (void)buf; (void)p*/
+#define HW_UNLOCK()
 
 #undef LOCAL_VARS
 #define LOCAL_VARS                                                     \
+    viaContextPtr vmesa = VIA_CONTEXT(ctx);                            \
     __DRIdrawablePrivate *dPriv = vmesa->driDrawable;                  \
-    GLuint pitch = vmesa->drawPitch;                                   \
+    GLuint draw_pitch = vmesa->drawBuffer->pitch;                       \
+    GLuint read_pitch = vmesa->readBuffer->pitch;                       \
     GLuint height = dPriv->h;                                          \
-    GLushort p;                                                        \
-    char *buf, *read_buf;                                              \
-    p = 0;                                                             \
-    if (vmesa->glCtx->Color._DrawDestMask == __GL_BACK_BUFFER_MASK) {  \
-       buf = (char *)(vmesa->drawMap);                                 \
-       read_buf = (char *)(vmesa->readMap);                            \
-    }                                                                  \
-    else {                                                             \
-       buf = (char *)(vmesa->drawMap +                                 \
-                         dPriv->x * 2 +                                \
-                         dPriv->y * pitch);                            \
-       read_buf = (char *)(vmesa->readMap +                            \
-                              dPriv->x * 2 +                           \
-                              dPriv->y * pitch);                       \
-    }
+    GLint p = 0;                                                       \
+    char *buf = (char *)(vmesa->drawBuffer->origMap + vmesa->drawXoff * vmesa->viaScreen->bytesPerPixel); \
+    char *read_buf = (char *)(vmesa->readBuffer->origMap + vmesa->drawXoff * vmesa->viaScreen->bytesPerPixel); \
+    (void) (read_pitch && draw_pitch && buf && read_buf && p);
 
-#define INIT_MONO_PIXEL(p, color)                       \
-    p = PACK_COLOR_565(color[0], color[1], color[2])
-    
-#define WRITE_RGBA(_x, _y, r, g, b, a)                                      \
-    *(GLushort *)(buf + _x * 2 + _y * pitch) = ((((int)r & 0xf8) << 8) |    \
-                                                (((int)g & 0xfc) << 3) |    \
-                                                (((int)b & 0xf8) >> 3))
-
-#define WRITE_PIXEL(_x, _y, p)                      \
-    *(GLushort *)(buf + _x * 2 + _y * pitch) = p
-
-#define READ_RGBA(rgba, _x, _y)                                             \
-    do {                                                                    \
-        GLushort p = *(GLushort *)(read_buf + _x * 2 + _y * pitch);         \
-        rgba[0] = ((p >> 8) & 0xf8) * 255 / 0xf8;                           \
-        rgba[1] = ((p >> 3) & 0xfc) * 255 / 0xfc;                           \
-        rgba[2] = ((p << 3) & 0xf8) * 255 / 0xf8;                           \
-        rgba[3] = 255;                                                      \
-    } while (0)
-
-#define TAG(x) via##x##_565
-#include "spantmp.h"
+/* ================================================================
+ * Color buffer
+ */
 
-/* 32 bit, 8888 argb color spanline and pixel functions
+/* 16 bit, RGB565 color spanline and pixel functions
  */
-#undef LOCAL_VARS
-#undef LOCAL_DEPTH_VARS
-/*=* [DBG] csmash : fix options worng position *=*/
-/*#define LOCAL_VARS                                    \
-    __DRIdrawablePrivate *dPriv = vmesa->driDrawable;   \
-    GLuint pitch = vmesa->drawPitch;                    \
-    GLuint height = dPriv->h;                           \
-    GLuint p;                                           \
-    char *buf = (char *)(vmesa->drawMap +               \
-                         dPriv->x * 4 +                 \
-                         dPriv->y * pitch);             \
-    char *read_buf = (char *)(vmesa->readMap +          \
-                              dPriv->x * 4 +            \
-                              dPriv->y * pitch);        \
-    (void)read_buf; (void)buf; (void)p*/
-#define LOCAL_VARS                                                     \
-    __DRIdrawablePrivate *dPriv = vmesa->driDrawable;                  \
-    GLuint pitch = vmesa->drawPitch;                                   \
-    GLuint height = dPriv->h;                                          \
-    GLuint p;                                                          \
-    char *buf, *read_buf;                                              \
-    p = 0;                                                             \
-    if (vmesa->glCtx->Color._DrawDestMask == __GL_BACK_BUFFER_MASK) {  \
-       buf = (char *)(vmesa->drawMap);                                 \
-       read_buf = (char *)(vmesa->readMap);                            \
-    }                                                                  \
-    else {                                                             \
-       buf = (char *)(vmesa->drawMap +                                 \
-                         dPriv->x * 4 +                                \
-                         dPriv->y * pitch);                            \
-       read_buf = (char *)(vmesa->readMap +                            \
-                              dPriv->x * 4 +                           \
-                              dPriv->y * pitch);                       \
-    }
+#define GET_SRC_PTR(_x, _y) (read_buf + _x * 2 + _y * read_pitch)
+#define GET_DST_PTR(_x, _y) (     buf + _x * 2 + _y * draw_pitch)
+#define SPANTMP_PIXEL_FMT GL_RGB
+#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5
 
+#define TAG(x)    via##x##_565
+#define TAG2(x,y) via##x##_565##y
+#include "spantmp2.h"
 
-#undef INIT_MONO_PIXEL
-#define INIT_MONO_PIXEL(p, color)                       \
-    p = PACK_COLOR_8888(color[3], color[0], color[1], color[2]) 
-    
-#define WRITE_RGBA(_x, _y, r, g, b, a)                                  \
-    *(GLuint *)(buf + _x * 4 + _y * pitch) = ((r << 16) |              \
-                                              (g << 8) |               \
-                                              (b << 0) |               \
-                                              (a << 24));
-                                              
-
-#define WRITE_PIXEL(_x, _y, p)                      \
-    *(GLuint *)(buf + _x * 4 + _y * pitch) = p
-
-#define READ_RGBA(rgba, _x, _y)                                         \
-    do {                                                                \
-        GLuint p = *(GLuint *)(read_buf + _x * 4 + _y * pitch);         \
-        rgba[0] = (p >> 16) & 0xff;                                    \
-        rgba[1] = (p >> 8) & 0xff;                                     \
-        rgba[2] = (p >> 0) & 0xff;                                     \
-        rgba[3] = 255;                                                  \
-    } while (0)
 
-#define TAG(x) via##x##_8888
-#include "spantmp.h"
-/*#include "via_spantmp.h"*/
+/* 32 bit, ARGB8888 color spanline and pixel functions
+ */
+#define GET_SRC_PTR(_x, _y) (read_buf + _x * 4 + _y * read_pitch)
+#define GET_DST_PTR(_x, _y) (     buf + _x * 4 + _y * draw_pitch)
+#define SPANTMP_PIXEL_FMT GL_BGRA
+#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
+
+#define TAG(x)    via##x##_8888
+#define TAG2(x,y) via##x##_8888##y
+#include "spantmp2.h"
+
 
 /* 16 bit depthbuffer functions.
  */
-/*=* John Sheng [2003.6.16] fix exy press 'i' dirty screen *=*/
-/*#define LOCAL_DEPTH_VARS                                \
-    __DRIdrawablePrivate *dPriv = vmesa->driDrawable;   \
-    GLuint pitch = vmesa->depth.pitch;                  \
-    GLuint height = dPriv->h;                           \
-    char *buf = (char *)(vmesa->depth.map +             \
-                         dPriv->x * 2 +                 \
-                         dPriv->y * pitch)   */
 #define LOCAL_DEPTH_VARS                                \
+    viaContextPtr vmesa = VIA_CONTEXT(ctx);             \
     __DRIdrawablePrivate *dPriv = vmesa->driDrawable;   \
-    /*viaScreenPrivate *viaScreen = vmesa->viaScreen;*/ \
-    GLuint pitch = vmesa->depth.pitch;                  \
+    GLuint depth_pitch = vmesa->depth.pitch;                  \
     GLuint height = dPriv->h;                           \
-    char *buf = (char *)(vmesa->depth.map)   
+    char *buf = (char *)(vmesa->depth.map + (vmesa->drawXoff * vmesa->depth.bpp/8))
+
+#define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS 
 
 
 #define WRITE_DEPTH(_x, _y, d)                      \
-    *(GLushort *)(buf + _x * 2 + _y * pitch) = d;
+    *(GLushort *)(buf + _x * 2 + _y * depth_pitch) = d;
 
 #define READ_DEPTH(d, _x, _y)                       \
-    d = *(GLushort *)(buf + _x * 2 + _y * pitch);
+    d = *(GLushort *)(buf + _x * 2 + _y * depth_pitch);
 
 #define TAG(x) via##x##_16
 #include "depthtmp.h"
 /* 32 bit depthbuffer functions.
  */
 #define WRITE_DEPTH(_x, _y, d)                      \
-    *(GLuint *)(buf + _x * 4 + _y * pitch) = d;
+    *(GLuint *)(buf + _x * 4 + _y * depth_pitch) = d;
 
 #define READ_DEPTH(d, _x, _y)                       \
-    d = *(GLuint *)(buf + _x * 4 + _y * pitch);
+    d = *(GLuint *)(buf + _x * 4 + _y * depth_pitch);
 
 #define TAG(x) via##x##_32
 #include "depthtmp.h"
 
-/* 24/8 bit depthbuffer functions.
+
+
+/* 24/8 bit interleaved depth/stencil functions
  */
-/* 
-#define WRITE_DEPTH(_x, _y, d) {                               \
-    GLuint tmp = *(GLuint *)(buf + _x * 4 + y * pitch);                \
-    tmp &= 0xff;                                               \
-    tmp |= (d) & 0xffffff00;                                   \
-    *(GLuint *)(buf + _x * 4 + _y * pitch) = tmp;              \
+#define WRITE_DEPTH( _x, _y, d ) {                     \
+   GLuint tmp = *(GLuint *)(buf + (_x)*4 + (_y)*depth_pitch);  \
+   tmp &= 0x000000ff;                                  \
+   tmp |= ((d)<<8);                            \
+   *(GLuint *)(buf + (_x)*4 + (_y)*depth_pitch) = tmp;         \
+}
+
+#define READ_DEPTH( d, _x, _y )                \
+   d = (*(GLuint *)(buf + (_x)*4 + (_y)*depth_pitch)) >> 8;
 
-#define READ_DEPTH(d, _x, _y)                                  \
-    d = (*(GLuint *)(buf + _x * 4 + _y * pitch) & ~0xff) >> 8; 
 
 #define TAG(x) via##x##_24_8
 #include "depthtmp.h"
-*/
+
+#define WRITE_STENCIL( _x, _y, d ) {                   \
+   GLuint tmp = *(GLuint *)(buf + (_x)*4 + (_y)*depth_pitch);  \
+   tmp &= 0xffffff00;                                  \
+   tmp |= (d);                                 \
+   *(GLuint *)(buf + (_x)*4 + (_y)*depth_pitch) = tmp;         \
+}
+
+#define READ_STENCIL( d, _x, _y )                      \
+   d = *(GLuint *)(buf + (_x)*4 + (_y)*depth_pitch) & 0xff;
+
+#define TAG(x) via##x##_24_8
+#include "stenciltmp.h"
+
 
 
 static void viaSetBuffer(GLcontext *ctx, GLframebuffer *colorBuffer,
                       GLuint bufferBit)
 {
     viaContextPtr vmesa = VIA_CONTEXT(ctx);
-#ifdef DEBUG
-    if (VIA_DEBUG) fprintf(stderr, "%s in\n", __FUNCTION__);
-#endif
+
     if (bufferBit == DD_FRONT_LEFT_BIT) {
-       vmesa->drawMap = (char *)vmesa->driScreen->pFB;
-        vmesa->readMap = (char *)vmesa->driScreen->pFB;
-       vmesa->drawPitch = vmesa->front.pitch;
-       vmesa->readPitch = vmesa->front.pitch;
+       vmesa->drawBuffer = vmesa->readBuffer = &vmesa->front;
     }
     else if (bufferBit == DD_BACK_LEFT_BIT) {
-       vmesa->drawMap = vmesa->back.map;
-        vmesa->readMap = vmesa->back.map;
-       vmesa->drawPitch = vmesa->back.pitch;
-       vmesa->readPitch = vmesa->back.pitch;   
+       vmesa->drawBuffer = vmesa->readBuffer = &vmesa->back;
     }
     else {
         ASSERT(0);
     }
-#ifdef DEBUG
-    if (VIA_DEBUG) fprintf(stderr, "%s out\n", __FUNCTION__);
-#endif
 }
 
+/* Move locking out to get reasonable span performance.
+ */
+void viaSpanRenderStart( GLcontext *ctx )
+{
+   viaContextPtr vmesa = VIA_CONTEXT(ctx);     
+   VIA_FINISH_PRIM(vmesa);
+   LOCK_HARDWARE(vmesa);
+   viaFlushDmaLocked(vmesa, 0);
+   WAIT_IDLE(vmesa);
+}
+
+void viaSpanRenderFinish( GLcontext *ctx )
+{
+   viaContextPtr vmesa = VIA_CONTEXT(ctx);
+   _swrast_flush( ctx );
+   UNLOCK_HARDWARE( vmesa );
+}
 
 void viaInitSpanFuncs(GLcontext *ctx)
 {
@@ -354,42 +213,45 @@ void viaInitSpanFuncs(GLcontext *ctx)
     struct swrast_device_driver *swdd = _swrast_GetDeviceDriverReference(ctx);
 
     swdd->SetBuffer = viaSetBuffer;
-#ifdef DEBUG
-    if (VIA_DEBUG) fprintf(stderr, "%s in\n", __FUNCTION__);
-#endif
-    if (vmesa->viaScreen->bitsPerPixel == 0x10) {
-       swdd->WriteRGBASpan = viaWriteRGBASpan_565;
-       swdd->WriteRGBSpan = viaWriteRGBSpan_565;
-       swdd->WriteMonoRGBASpan = viaWriteMonoRGBASpan_565;
-       swdd->WriteRGBAPixels = viaWriteRGBAPixels_565;
-       swdd->WriteMonoRGBAPixels = viaWriteMonoRGBAPixels_565;
-       swdd->ReadRGBASpan = viaReadRGBASpan_565;
-       swdd->ReadRGBAPixels = viaReadRGBAPixels_565;
+    if (vmesa->viaScreen->bitsPerPixel == 16) {
+       viaInitPointers_565( swdd );
+    }
+    else if (vmesa->viaScreen->bitsPerPixel == 32) {
+       viaInitPointers_8888( swdd );
     }
-    else if (vmesa->viaScreen->bitsPerPixel == 0x20) {
-       swdd->WriteRGBASpan = viaWriteRGBASpan_8888;
-       swdd->WriteRGBSpan = viaWriteRGBSpan_8888;
-       swdd->WriteMonoRGBASpan = viaWriteMonoRGBASpan_8888;
-       swdd->WriteRGBAPixels = viaWriteRGBAPixels_8888;
-       swdd->WriteMonoRGBAPixels = viaWriteMonoRGBAPixels_8888;
-       swdd->ReadRGBASpan = viaReadRGBASpan_8888;
-       swdd->ReadRGBAPixels = viaReadRGBAPixels_8888;
+    else {
+       assert(0);
     }
-    else 
-       ASSERT(0);
        
-    if (vmesa->glCtx->Visual.depthBits == 0x10) {
+    if (vmesa->glCtx->Visual.depthBits == 16) {
        swdd->ReadDepthSpan = viaReadDepthSpan_16;
        swdd->WriteDepthSpan = viaWriteDepthSpan_16;
+       swdd->WriteMonoDepthSpan = viaWriteMonoDepthSpan_16;
        swdd->ReadDepthPixels = viaReadDepthPixels_16;
        swdd->WriteDepthPixels = viaWriteDepthPixels_16;
     }  
-    else if (vmesa->glCtx->Visual.depthBits == 0x20) {
+    else if (vmesa->glCtx->Visual.depthBits == 24) {
+        swdd->ReadDepthSpan = viaReadDepthSpan_24_8;
+       swdd->WriteDepthSpan = viaWriteDepthSpan_24_8;
+       swdd->ReadDepthPixels = viaReadDepthPixels_24_8;
+       swdd->WriteDepthPixels = viaWriteDepthPixels_24_8;
+
+       swdd->WriteStencilSpan = viaWriteStencilSpan_24_8;
+       swdd->ReadStencilSpan = viaReadStencilSpan_24_8;
+       swdd->WriteStencilPixels = viaWriteStencilPixels_24_8;
+       swdd->ReadStencilPixels = viaReadStencilPixels_24_8;
+    }
+    else if (vmesa->glCtx->Visual.depthBits == 32) {
        swdd->ReadDepthSpan = viaReadDepthSpan_32;
        swdd->WriteDepthSpan = viaWriteDepthSpan_32;
+       swdd->WriteMonoDepthSpan = viaWriteMonoDepthSpan_32;
        swdd->ReadDepthPixels = viaReadDepthPixels_32;
        swdd->WriteDepthPixels = viaWriteDepthPixels_32;
     }
+
+    swdd->SpanRenderStart = viaSpanRenderStart;
+    swdd->SpanRenderFinish = viaSpanRenderFinish; 
+
     
     swdd->WriteCI8Span = NULL;
     swdd->WriteCI32Span = NULL;
@@ -398,7 +260,4 @@ void viaInitSpanFuncs(GLcontext *ctx)
     swdd->WriteMonoCIPixels = NULL;
     swdd->ReadCI32Span = NULL;
     swdd->ReadCI32Pixels = NULL;       
-#ifdef DEBUG
-    if (VIA_DEBUG) fprintf(stderr, "%s out\n", __FUNCTION__);
-#endif
 }