diff options
Diffstat (limited to 'src/mesa/drivers/dri/intel')
35 files changed, 1610 insertions, 1543 deletions
diff --git a/src/mesa/drivers/dri/intel/intel_batchbuffer.c b/src/mesa/drivers/dri/intel/intel_batchbuffer.c index 82a92a9f45..6aa36d10b1 100644 --- a/src/mesa/drivers/dri/intel/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/intel/intel_batchbuffer.c @@ -195,7 +195,11 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file, { struct intel_context *intel = batch->intel; GLuint used = batch->ptr - batch->map; - GLboolean was_locked = intel->locked; + + if (intel->first_post_swapbuffers_batch == NULL) { + intel->first_post_swapbuffers_batch = intel->batch->buf; + drm_intel_bo_reference(intel->first_post_swapbuffers_batch); + } if (intel->first_post_swapbuffers_batch == NULL) { intel->first_post_swapbuffers_batch = intel->batch->buf; @@ -248,13 +252,9 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file, /* TODO: Just pass the relocation list and dma buffer up to the * kernel. */ - if (!was_locked) - LOCK_HARDWARE(intel); - + LOCK_HARDWARE(intel); do_flush_locked(batch, used, GL_FALSE); - - if (!was_locked) - UNLOCK_HARDWARE(intel); + UNLOCK_HARDWARE(intel); if (INTEL_DEBUG & DEBUG_SYNC) { fprintf(stderr, "waiting for idle\n"); diff --git a/src/mesa/drivers/dri/intel/intel_blit.c b/src/mesa/drivers/dri/intel/intel_blit.c index 4919828131..0c5be4c798 100644 --- a/src/mesa/drivers/dri/intel/intel_blit.c +++ b/src/mesa/drivers/dri/intel/intel_blit.c @@ -108,6 +108,8 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv, CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB; } + assert(src->tiling != I915_TILING_Y); + assert(dst->tiling != I915_TILING_Y); #ifndef I915 if (src->tiling != I915_TILING_NONE) { CMD |= XY_SRC_TILED; @@ -175,66 +177,6 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv, UNLOCK_HARDWARE(intel); } - - - -void -intelEmitFillBlit(struct intel_context *intel, - GLuint cpp, - GLshort dst_pitch, - dri_bo *dst_buffer, - GLuint dst_offset, - uint32_t dst_tiling, - GLshort x, GLshort y, - GLshort w, GLshort h, - GLuint color) -{ - GLuint BR13, CMD; - BATCH_LOCALS; - - dst_pitch *= cpp; - - switch (cpp) { - case 1: - BR13 = (0xF0 << 16); - CMD = XY_COLOR_BLT_CMD; - break; - case 2: - BR13 = (0xF0 << 16) | BR13_565; - CMD = XY_COLOR_BLT_CMD; - break; - case 4: - BR13 = (0xF0 << 16) | BR13_8888; - CMD = XY_COLOR_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB; - break; - default: - return; - } -#ifndef I915 - if (dst_tiling != I915_TILING_NONE) { - CMD |= XY_DST_TILED; - dst_pitch /= 4; - } -#endif - - DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", - __FUNCTION__, dst_buffer, dst_pitch, dst_offset, x, y, w, h); - - assert(w > 0); - assert(h > 0); - - BEGIN_BATCH(6, NO_LOOP_CLIPRECTS); - OUT_BATCH(CMD); - OUT_BATCH(BR13 | dst_pitch); - OUT_BATCH((y << 16) | x); - OUT_BATCH(((y + h) << 16) | (x + w)); - OUT_RELOC(dst_buffer, - I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER, - dst_offset); - OUT_BATCH(color); - ADVANCE_BATCH(); -} - static GLuint translate_raster_op(GLenum logicop) { switch(logicop) { @@ -261,7 +203,7 @@ static GLuint translate_raster_op(GLenum logicop) /* Copy BitBlt */ -void +GLboolean intelEmitCopyBlit(struct intel_context *intel, GLuint cpp, GLshort src_pitch, @@ -283,6 +225,19 @@ intelEmitCopyBlit(struct intel_context *intel, dri_bo *aper_array[3]; BATCH_LOCALS; + if (dst_tiling != I915_TILING_NONE) { + if (dst_offset & 4095) + return GL_FALSE; + if (dst_tiling == I915_TILING_Y) + return GL_FALSE; + } + if (src_tiling != I915_TILING_NONE) { + if (src_offset & 4095) + return GL_FALSE; + if (src_tiling == I915_TILING_Y) + return GL_FALSE; + } + /* do space/cliprects check before going any further */ do { aper_array[0] = intel->batch->buf; @@ -297,12 +252,7 @@ intelEmitCopyBlit(struct intel_context *intel, } while (pass < 2); if (pass >= 2) { - GLboolean locked = GL_FALSE; - if (!intel->locked) { - LOCK_HARDWARE(intel); - locked = GL_TRUE; - } - + LOCK_HARDWARE(intel); dri_bo_map(dst_buffer, GL_TRUE); dri_bo_map(src_buffer, GL_FALSE); _mesa_copy_rect((GLubyte *)dst_buffer->virtual + dst_offset, @@ -316,11 +266,9 @@ intelEmitCopyBlit(struct intel_context *intel, dri_bo_unmap(src_buffer); dri_bo_unmap(dst_buffer); - - if (locked) - UNLOCK_HARDWARE(intel); + UNLOCK_HARDWARE(intel); - return; + return GL_TRUE; } intel_batchbuffer_require_space(intel->batch, 8 * 4, NO_LOOP_CLIPRECTS); @@ -347,7 +295,7 @@ intelEmitCopyBlit(struct intel_context *intel, CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB; break; default: - return; + return GL_FALSE; } #ifndef I915 @@ -362,7 +310,7 @@ intelEmitCopyBlit(struct intel_context *intel, #endif if (dst_y2 <= dst_y || dst_x2 <= dst_x) { - return; + return GL_TRUE; } assert(dst_x < dst_x2); @@ -384,6 +332,8 @@ intelEmitCopyBlit(struct intel_context *intel, ADVANCE_BATCH(); intel_batchbuffer_emit_mi_flush(intel->batch); + + return GL_TRUE; } @@ -527,6 +477,8 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) BR13 |= BR13_565; } + assert(irb->region->tiling != I915_TILING_Y); + #ifndef I915 if (irb->region->tiling != I915_TILING_NONE) { CMD |= XY_DST_TILED; @@ -596,7 +548,7 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) UNLOCK_HARDWARE(intel); } -void +GLboolean intelEmitImmediateColorExpandBlit(struct intel_context *intel, GLuint cpp, GLubyte *src_bits, GLuint src_size, @@ -612,11 +564,19 @@ intelEmitImmediateColorExpandBlit(struct intel_context *intel, int dwords = ALIGN(src_size, 8) / 4; uint32_t opcode, br13, blit_cmd; + if (dst_tiling != I915_TILING_NONE) { + if (dst_offset & 4095) + return GL_FALSE; + if (dst_tiling == I915_TILING_Y) + return GL_FALSE; + } + assert( logic_op - GL_CLEAR >= 0 ); assert( logic_op - GL_CLEAR < 0x10 ); + assert(dst_pitch > 0); if (w < 0 || h < 0) - return; + return GL_TRUE; dst_pitch *= cpp; @@ -673,4 +633,46 @@ intelEmitImmediateColorExpandBlit(struct intel_context *intel, REFERENCES_CLIPRECTS ); intel_batchbuffer_emit_mi_flush(intel->batch); + + return GL_TRUE; +} + +/* We don't have a memmove-type blit like some other hardware, so we'll do a + * rectangular blit covering a large space, then emit 1-scanline blit at the + * end to cover the last if we need. + */ +void +intel_emit_linear_blit(struct intel_context *intel, + drm_intel_bo *dst_bo, + unsigned int dst_offset, + drm_intel_bo *src_bo, + unsigned int src_offset, + unsigned int size) +{ + GLuint pitch, height; + + /* The pitch is a signed value. */ + pitch = MIN2(size, (1 << 15) - 1); + height = size / pitch; + intelEmitCopyBlit(intel, 1, + pitch, src_bo, src_offset, I915_TILING_NONE, + pitch, dst_bo, dst_offset, I915_TILING_NONE, + 0, 0, /* src x/y */ + 0, 0, /* dst x/y */ + pitch, height, /* w, h */ + GL_COPY); + + src_offset += pitch * height; + dst_offset += pitch * height; + size -= pitch * height; + assert (size < (1 << 15)); + if (size != 0) { + intelEmitCopyBlit(intel, 1, + size, src_bo, src_offset, I915_TILING_NONE, + size, dst_bo, dst_offset, I915_TILING_NONE, + 0, 0, /* src x/y */ + 0, 0, /* dst x/y */ + size, 1, /* w, h */ + GL_COPY); + } } diff --git a/src/mesa/drivers/dri/intel/intel_blit.h b/src/mesa/drivers/dri/intel/intel_blit.h index 52065b13ed..240cb7cd1b 100644 --- a/src/mesa/drivers/dri/intel/intel_blit.h +++ b/src/mesa/drivers/dri/intel/intel_blit.h @@ -35,7 +35,8 @@ extern void intelCopyBuffer(const __DRIdrawablePrivate * dpriv, extern void intelClearWithBlit(GLcontext * ctx, GLbitfield mask); -extern void intelEmitCopyBlit(struct intel_context *intel, +GLboolean +intelEmitCopyBlit(struct intel_context *intel, GLuint cpp, GLshort src_pitch, dri_bo *src_buffer, @@ -50,16 +51,7 @@ extern void intelEmitCopyBlit(struct intel_context *intel, GLshort w, GLshort h, GLenum logicop ); -extern void intelEmitFillBlit(struct intel_context *intel, - GLuint cpp, - GLshort dst_pitch, - dri_bo *dst_buffer, - GLuint dst_offset, - uint32_t dst_tiling, - GLshort x, GLshort y, - GLshort w, GLshort h, GLuint color); - -void +GLboolean intelEmitImmediateColorExpandBlit(struct intel_context *intel, GLuint cpp, GLubyte *src_bits, GLuint src_size, @@ -71,5 +63,11 @@ intelEmitImmediateColorExpandBlit(struct intel_context *intel, GLshort x, GLshort y, GLshort w, GLshort h, GLenum logic_op); +void intel_emit_linear_blit(struct intel_context *intel, + drm_intel_bo *dst_bo, + unsigned int dst_offset, + drm_intel_bo *src_bo, + unsigned int src_offset, + unsigned int size); #endif diff --git a/src/mesa/drivers/dri/intel/intel_buffer_objects.c b/src/mesa/drivers/dri/intel/intel_buffer_objects.c index 2e6b77824d..c55c5c426e 100644 --- a/src/mesa/drivers/dri/intel/intel_buffer_objects.c +++ b/src/mesa/drivers/dri/intel/intel_buffer_objects.c @@ -28,13 +28,18 @@ #include "main/imports.h" #include "main/mtypes.h" +#include "main/macros.h" #include "main/bufferobj.h" #include "intel_context.h" +#include "intel_blit.h" #include "intel_buffer_objects.h" #include "intel_batchbuffer.h" #include "intel_regions.h" +static GLboolean +intel_bufferobj_unmap(GLcontext * ctx, + GLenum target, struct gl_buffer_object *obj); /** Allocates a new dri_bo to store the data for the buffer object. */ static void @@ -100,7 +105,13 @@ intel_bufferobj_free(GLcontext * ctx, struct gl_buffer_object *obj) struct intel_buffer_object *intel_obj = intel_buffer_object(obj); assert(intel_obj); - assert(!obj->Pointer); /* Mesa should have unmapped it */ + + /* Buffer objects are automatically unmapped when deleting according + * to the spec, but Mesa doesn't do UnmapBuffer for us at context destroy + * (though it does if you call glDeleteBuffers) + */ + if (obj->Pointer) + intel_bufferobj_unmap(ctx, 0, obj); _mesa_free(intel_obj->sys_buffer); if (intel_obj->region) { @@ -119,9 +130,10 @@ intel_bufferobj_free(GLcontext * ctx, struct gl_buffer_object *obj) * Allocate space for and store data in a buffer object. Any data that was * previously stored in the buffer object is lost. If data is NULL, * memory will be allocated, but no copy will occur. - * Called via glBufferDataARB(). + * Called via ctx->Driver.BufferData(). + * \return GL_TRUE for success, GL_FALSE if out of memory */ -static void +static GLboolean intel_bufferobj_data(GLcontext * ctx, GLenum target, GLsizeiptrARB size, @@ -156,15 +168,19 @@ intel_bufferobj_data(GLcontext * ctx, if (intel_obj->sys_buffer != NULL) { if (data != NULL) memcpy(intel_obj->sys_buffer, data, size); - return; + return GL_TRUE; } } #endif intel_bufferobj_alloc_buffer(intel, intel_obj); + if (!intel_obj->buffer) + return GL_FALSE; if (data != NULL) dri_bo_subdata(intel_obj->buffer, 0, size, data); } + + return GL_TRUE; } @@ -234,6 +250,11 @@ intel_bufferobj_map(GLcontext * ctx, return obj->Pointer; } + /* Flush any existing batchbuffer that might have written to this + * buffer. + */ + intelFlush(ctx); + if (intel_obj->region) intel_bufferobj_cow(intel, intel_obj); @@ -251,32 +272,205 @@ intel_bufferobj_map(GLcontext * ctx, } obj->Pointer = intel_obj->buffer->virtual; + obj->Length = obj->Size; + obj->Offset = 0; + + return obj->Pointer; +} + +/** + * Called via glMapBufferRange(). + * + * The goal of this extension is to allow apps to accumulate their rendering + * at the same time as they accumulate their buffer object. Without it, + * you'd end up blocking on execution of rendering every time you mapped + * the buffer to put new data in. + * + * We support it in 3 ways: If unsynchronized, then don't bother + * flushing the batchbuffer before mapping the buffer, which can save blocking + * in many cases. If we would still block, and they allow the whole buffer + * to be invalidated, then just allocate a new buffer to replace the old one. + * If not, and we'd block, and they allow the subrange of the buffer to be + * invalidated, then we can make a new little BO, let them write into that, + * and blit it into the real BO at unmap time. + */ +static void * +intel_bufferobj_map_range(GLcontext * ctx, + GLenum target, GLintptr offset, GLsizeiptr length, + GLbitfield access, struct gl_buffer_object *obj) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_buffer_object *intel_obj = intel_buffer_object(obj); + + assert(intel_obj); + + /* _mesa_MapBufferRange (GL entrypoint) sets these, but the vbo module also + * internally uses our functions directly. + */ + obj->Offset = offset; + obj->Length = length; + obj->AccessFlags = access; + + if (intel_obj->sys_buffer) { + obj->Pointer = intel_obj->sys_buffer + offset; + return obj->Pointer; + } + + if (intel_obj->region) + intel_bufferobj_cow(intel, intel_obj); + + /* If the mapping is synchronized with other GL operations, flush + * the batchbuffer so that GEM knows about the buffer access for later + * syncing. + */ + if (!(access & GL_MAP_UNSYNCHRONIZED_BIT)) + intelFlush(ctx); + + if (intel_obj->buffer == NULL) { + obj->Pointer = NULL; + return NULL; + } + + /* If the user doesn't care about existing buffer contents and mapping + * would cause us to block, then throw out the old buffer. + */ + if (!(access & GL_MAP_UNSYNCHRONIZED_BIT) && + (access & GL_MAP_INVALIDATE_BUFFER_BIT) && + drm_intel_bo_busy(intel_obj->buffer)) { + drm_intel_bo_unreference(intel_obj->buffer); + intel_obj->buffer = dri_bo_alloc(intel->bufmgr, "bufferobj", + intel_obj->Base.Size, 64); + } + + /* If the user is mapping a range of an active buffer object but + * doesn't require the current contents of that range, make a new + * BO, and we'll copy what they put in there out at unmap or + * FlushRange time. + */ + if ((access & GL_MAP_INVALIDATE_RANGE_BIT) && + drm_intel_bo_busy(intel_obj->buffer)) { + if (access & GL_MAP_FLUSH_EXPLICIT_BIT) { + intel_obj->range_map_buffer = _mesa_malloc(length); + obj->Pointer = intel_obj->range_map_buffer; + } else { + intel_obj->range_map_bo = drm_intel_bo_alloc(intel->bufmgr, + "range map", + length, 64); + if (!(access & GL_MAP_READ_BIT) && + intel->intelScreen->kernel_exec_fencing) { + drm_intel_gem_bo_map_gtt(intel_obj->range_map_bo); + intel_obj->mapped_gtt = GL_TRUE; + } else { + drm_intel_bo_map(intel_obj->range_map_bo, + (access & GL_MAP_WRITE_BIT) != 0); + intel_obj->mapped_gtt = GL_FALSE; + } + obj->Pointer = intel_obj->range_map_bo->virtual; + } + return obj->Pointer; + } + + if (!(access & GL_MAP_READ_BIT) && + intel->intelScreen->kernel_exec_fencing) { + drm_intel_gem_bo_map_gtt(intel_obj->buffer); + intel_obj->mapped_gtt = GL_TRUE; + } else { + drm_intel_bo_map(intel_obj->buffer, (access & GL_MAP_WRITE_BIT) != 0); + intel_obj->mapped_gtt = GL_FALSE; + } + + obj->Pointer = intel_obj->buffer->virtual + offset; return obj->Pointer; } +/* Ideally we'd use a BO to avoid taking up cache space for the temporary + * data, but FlushMappedBufferRange may be followed by further writes to + * the pointer, so we would have to re-map after emitting our blit, which + * would defeat the point. + */ +static void +intel_bufferobj_flush_mapped_range(GLcontext *ctx, GLenum target, + GLintptr offset, GLsizeiptr length, + struct gl_buffer_object *obj) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_buffer_object *intel_obj = intel_buffer_object(obj); + drm_intel_bo *temp_bo; + + /* Unless we're in the range map using a temporary system buffer, + * there's no work to do. + */ + if (intel_obj->range_map_buffer == NULL) + return; + + temp_bo = drm_intel_bo_alloc(intel->bufmgr, "range map flush", length, 64); + + drm_intel_bo_subdata(temp_bo, 0, length, intel_obj->range_map_buffer); + + intel_emit_linear_blit(intel, + intel_obj->buffer, obj->Offset + offset, + temp_bo, 0, + length); + + drm_intel_bo_unreference(temp_bo); +} + /** - * Called via glMapBufferARB(). + * Called via glUnmapBuffer(). */ static GLboolean intel_bufferobj_unmap(GLcontext * ctx, GLenum target, struct gl_buffer_object *obj) { + struct intel_context *intel = intel_context(ctx); struct intel_buffer_object *intel_obj = intel_buffer_object(obj); assert(intel_obj); + assert(obj->Pointer); if (intel_obj->sys_buffer != NULL) { - assert(obj->Pointer); - obj->Pointer = NULL; + /* always keep the mapping around. */ + } else if (intel_obj->range_map_buffer != NULL) { + /* Since we've emitted some blits to buffers that will (likely) be used + * in rendering operations in other cache domains in this batch, emit a + * flush. Once again, we wish for a domain tracker in libdrm to cover + * usage inside of a batchbuffer. + */ + intel_batchbuffer_emit_mi_flush(intel->batch); + free(intel_obj->range_map_buffer); + intel_obj->range_map_buffer = NULL; + } else if (intel_obj->range_map_bo != NULL) { + if (intel_obj->mapped_gtt) { + drm_intel_gem_bo_unmap_gtt(intel_obj->range_map_bo); + } else { + drm_intel_bo_unmap(intel_obj->range_map_bo); + } + + intel_emit_linear_blit(intel, + intel_obj->buffer, obj->Offset, + intel_obj->range_map_bo, 0, + obj->Length); + + /* Since we've emitted some blits to buffers that will (likely) be used + * in rendering operations in other cache domains in this batch, emit a + * flush. Once again, we wish for a domain tracker in libdrm to cover + * usage inside of a batchbuffer. + */ + intel_batchbuffer_emit_mi_flush(intel->batch); + + drm_intel_bo_unreference(intel_obj->range_map_bo); + intel_obj->range_map_bo = NULL; } else if (intel_obj->buffer != NULL) { - assert(obj->Pointer); if (intel_obj->mapped_gtt) { drm_intel_gem_bo_unmap_gtt(intel_obj->buffer); } else { drm_intel_bo_unmap(intel_obj->buffer); } - obj->Pointer = NULL; } + obj->Pointer = NULL; + obj->Offset = 0; + obj->Length = 0; + return GL_TRUE; } @@ -294,30 +488,94 @@ intel_bufferobj_buffer(struct intel_context *intel, } if (intel_obj->buffer == NULL) { + void *sys_buffer = intel_obj->sys_buffer; + + /* only one of buffer and sys_buffer could be non-NULL */ intel_bufferobj_alloc_buffer(intel, intel_obj); + intel_obj->sys_buffer = NULL; + intel_bufferobj_subdata(&intel->ctx, GL_ARRAY_BUFFER_ARB, 0, intel_obj->Base.Size, - intel_obj->sys_buffer, + sys_buffer, &intel_obj->Base); - _mesa_free(intel_obj->sys_buffer); + _mesa_free(sys_buffer); intel_obj->sys_buffer = NULL; } return intel_obj->buffer; } +static void +intel_bufferobj_copy_subdata(GLcontext *ctx, + struct gl_buffer_object *src, + struct gl_buffer_object *dst, + GLintptr read_offset, GLintptr write_offset, + GLsizeiptr size) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_buffer_object *intel_src = intel_buffer_object(src); + struct intel_buffer_object *intel_dst = intel_buffer_object(dst); + drm_intel_bo *src_bo, *dst_bo; + + if (size == 0) + return; + + /* If we're in system memory, just map and memcpy. */ + if (intel_src->sys_buffer || intel_dst->sys_buffer) { + /* The same buffer may be used, but note that regions copied may + * not overlap. + */ + if (src == dst) { + char *ptr = intel_bufferobj_map(ctx, GL_COPY_WRITE_BUFFER, + GL_READ_WRITE, dst); + memcpy(ptr + write_offset, ptr + read_offset, size); + intel_bufferobj_unmap(ctx, GL_COPY_WRITE_BUFFER, dst); + } else { + const char *src_ptr; + char *dst_ptr; + + src_ptr = intel_bufferobj_map(ctx, GL_COPY_READ_BUFFER, + GL_READ_ONLY, src); + dst_ptr = intel_bufferobj_map(ctx, GL_COPY_WRITE_BUFFER, + GL_WRITE_ONLY, dst); + + memcpy(dst_ptr + write_offset, src_ptr + read_offset, size); + + intel_bufferobj_unmap(ctx, GL_COPY_READ_BUFFER, src); + intel_bufferobj_unmap(ctx, GL_COPY_WRITE_BUFFER, dst); + } + } + + /* Otherwise, we have real BOs, so blit them. */ + + dst_bo = intel_bufferobj_buffer(intel, intel_dst, INTEL_WRITE_PART); + src_bo = intel_bufferobj_buffer(intel, intel_src, INTEL_READ); + + intel_emit_linear_blit(intel, + dst_bo, write_offset, + src_bo, read_offset, size); + + /* Since we've emitted some blits to buffers that will (likely) be used + * in rendering operations in other cache domains in this batch, emit a + * flush. Once again, we wish for a domain tracker in libdrm to cover + * usage inside of a batchbuffer. + */ + intel_batchbuffer_emit_mi_flush(intel->batch); +} + void -intel_bufferobj_init(struct intel_context *intel) +intelInitBufferObjectFuncs(struct dd_function_table *functions) { - GLcontext *ctx = &intel->ctx; - - ctx->Driver.NewBufferObject = intel_bufferobj_alloc; - ctx->Driver.DeleteBuffer = intel_bufferobj_free; - ctx->Driver.BufferData = intel_bufferobj_data; - ctx->Driver.BufferSubData = intel_bufferobj_subdata; - ctx->Driver.GetBufferSubData = intel_bufferobj_get_subdata; - ctx->Driver.MapBuffer = intel_bufferobj_map; - ctx->Driver.UnmapBuffer = intel_bufferobj_unmap; + functions->NewBufferObject = intel_bufferobj_alloc; + functions->DeleteBuffer = intel_bufferobj_free; + functions->BufferData = intel_bufferobj_data; + functions->BufferSubData = intel_bufferobj_subdata; + functions->GetBufferSubData = intel_bufferobj_get_subdata; + functions->MapBuffer = intel_bufferobj_map; + functions->MapBufferRange = intel_bufferobj_map_range; + functions->FlushMappedBufferRange = intel_bufferobj_flush_mapped_range; + functions->UnmapBuffer = intel_bufferobj_unmap; + functions->CopyBufferSubData = intel_bufferobj_copy_subdata; } diff --git a/src/mesa/drivers/dri/intel/intel_buffer_objects.h b/src/mesa/drivers/dri/intel/intel_buffer_objects.h index 0431015631..bf3e08a320 100644 --- a/src/mesa/drivers/dri/intel/intel_buffer_objects.h +++ b/src/mesa/drivers/dri/intel/intel_buffer_objects.h @@ -48,6 +48,12 @@ struct intel_buffer_object struct intel_region *region; /* Is there a zero-copy texture associated with this (pixel) buffer object? */ + + drm_intel_bo *range_map_bo; + void *range_map_buffer; + unsigned int range_map_offset; + GLsizei range_map_size; + GLboolean mapped_gtt; }; @@ -60,7 +66,7 @@ dri_bo *intel_bufferobj_buffer(struct intel_context *intel, /* Hook the bufferobject implementation into mesa: */ -void intel_bufferobj_init(struct intel_context *intel); +void intelInitBufferObjectFuncs(struct dd_function_table *functions); @@ -72,10 +78,7 @@ void intel_bufferobj_init(struct intel_context *intel); static INLINE struct intel_buffer_object * intel_buffer_object(struct gl_buffer_object *obj) { - if (obj->Name) - return (struct intel_buffer_object *) obj; - else - return NULL; + return (struct intel_buffer_object *) obj; } /* Helpers for zerocopy image uploads. See also intel_regions.h: diff --git a/src/mesa/drivers/dri/intel/intel_buffers.c b/src/mesa/drivers/dri/intel/intel_buffers.c index 44e34330a6..e7357e78c5 100644 --- a/src/mesa/drivers/dri/intel/intel_buffers.c +++ b/src/mesa/drivers/dri/intel/intel_buffers.c @@ -157,7 +157,7 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) /* Do this here, not core Mesa, since this function is called from * many places within the driver. */ - if (ctx->NewState & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) { + if (ctx->NewState & _NEW_BUFFERS) { /* this updates the DrawBuffer->_NumColorDrawBuffers fields, etc */ _mesa_update_framebuffer(ctx); /* this updates the DrawBuffer's Width/Height if it's a FBO */ diff --git a/src/mesa/drivers/dri/intel/intel_chipset.h b/src/mesa/drivers/dri/intel/intel_chipset.h index a528d996dc..3dc8653a73 100644 --- a/src/mesa/drivers/dri/intel/intel_chipset.h +++ b/src/mesa/drivers/dri/intel/intel_chipset.h @@ -68,13 +68,18 @@ #define PCI_CHIP_G41_G 0x2E32 #define PCI_CHIP_B43_G 0x2E42 +#define PCI_CHIP_ILD_G 0x0042 +#define PCI_CHIP_ILM_G 0x0046 + #define IS_MOBILE(devid) (devid == PCI_CHIP_I855_GM || \ devid == PCI_CHIP_I915_GM || \ devid == PCI_CHIP_I945_GM || \ devid == PCI_CHIP_I945_GME || \ devid == PCI_CHIP_I965_GM || \ devid == PCI_CHIP_I965_GME || \ - devid == PCI_CHIP_GM45_GM || IS_IGD(devid)) + devid == PCI_CHIP_GM45_GM || \ + IS_IGD(devid) || \ + devid == PCI_CHIP_ILM_G) #define IS_G45(devid) (devid == PCI_CHIP_IGD_E_G || \ devid == PCI_CHIP_Q45_G || \ @@ -84,6 +89,10 @@ #define IS_GM45(devid) (devid == PCI_CHIP_GM45_GM) #define IS_G4X(devid) (IS_G45(devid) || IS_GM45(devid)) +#define IS_ILD(devid) (devid == PCI_CHIP_ILD_G) +#define IS_ILM(devid) (devid == PCI_CHIP_ILM_G) +#define IS_IGDNG(devid) (IS_ILD(devid) || IS_ILM(devid)) + #define IS_915(devid) (devid == PCI_CHIP_I915_G || \ devid == PCI_CHIP_E7221_G || \ devid == PCI_CHIP_I915_GM) @@ -101,7 +110,8 @@ devid == PCI_CHIP_I965_GM || \ devid == PCI_CHIP_I965_GME || \ devid == PCI_CHIP_I946_GZ || \ - IS_G4X(devid)) + IS_G4X(devid) || \ + IS_IGDNG(devid)) #define IS_9XX(devid) (IS_915(devid) || \ IS_945(devid) || \ diff --git a/src/mesa/drivers/dri/intel/intel_clear.c b/src/mesa/drivers/dri/intel/intel_clear.c index 19f47632ac..9010b910c7 100644 --- a/src/mesa/drivers/dri/intel/intel_clear.c +++ b/src/mesa/drivers/dri/intel/intel_clear.c @@ -27,25 +27,9 @@ **************************************************************************/ #include "main/glheader.h" -#include "main/enums.h" -#include "main/image.h" #include "main/mtypes.h" -#include "main/arrayobj.h" -#include "main/attrib.h" -#include "main/blend.h" -#include "main/bufferobj.h" -#include "main/buffers.h" -#include "main/depth.h" -#include "main/enable.h" -#include "main/macros.h" -#include "main/matrix.h" -#include "main/polygon.h" -#include "main/texstate.h" -#include "main/shaders.h" -#include "main/stencil.h" -#include "main/varray.h" -#include "glapi/dispatch.h" #include "swrast/swrast.h" +#include "drivers/common/meta.h" #include "intel_context.h" #include "intel_blit.h" @@ -53,240 +37,11 @@ #include "intel_clear.h" #include "intel_fbo.h" #include "intel_pixel.h" +#include "intel_regions.h" +#include "intel_batchbuffer.h" #define FILE_DEBUG_FLAG DEBUG_BLIT -#define TRI_CLEAR_COLOR_BITS (BUFFER_BIT_BACK_LEFT | \ - BUFFER_BIT_FRONT_LEFT | \ - BUFFER_BIT_COLOR0 | \ - BUFFER_BIT_COLOR1 | \ - BUFFER_BIT_COLOR2 | \ - BUFFER_BIT_COLOR3 | \ - BUFFER_BIT_COLOR4 | \ - BUFFER_BIT_COLOR5 | \ - BUFFER_BIT_COLOR6 | \ - BUFFER_BIT_COLOR7) - - -/** - * Per-context one-time init of things for intl_clear_tris(). - * Basically set up a private array object for vertex/color arrays. - */ -static void -init_clear(GLcontext *ctx) -{ - struct intel_context *intel = intel_context(ctx); - struct gl_array_object *arraySave = NULL; - const GLuint arrayBuffer = ctx->Array.ArrayBufferObj->Name; - const GLuint elementBuffer = ctx->Array.ElementArrayBufferObj->Name; - - /* create new array object */ - intel->clear.arrayObj = _mesa_new_array_object(ctx, ~0); - - /* save current array object, bind new one */ - _mesa_reference_array_object(ctx, &arraySave, ctx->Array.ArrayObj); - _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, intel->clear.arrayObj); - - /* one-time setup of vertex arrays (pos, color) */ - _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, 0); - _mesa_BindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0); - _mesa_ColorPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), intel->clear.color); - _mesa_VertexPointer(3, GL_FLOAT, 3 * sizeof(GLfloat), intel->clear.vertices); - _mesa_Enable(GL_COLOR_ARRAY); - _mesa_Enable(GL_VERTEX_ARRAY); - - /* restore original array object */ - _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, arraySave); - _mesa_reference_array_object(ctx, &arraySave, NULL); - - /* restore original buffer objects */ - _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, arrayBuffer); - _mesa_BindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, elementBuffer); -} - - - -/** - * Perform glClear where mask contains only color, depth, and/or stencil. - * - * The implementation is based on calling into Mesa to set GL state and - * performing normal triangle rendering. The intent of this path is to - * have as generic a path as possible, so that any driver could make use of - * it. - */ -void -intel_clear_tris(GLcontext *ctx, GLbitfield mask) -{ - struct intel_context *intel = intel_context(ctx); - GLfloat dst_z; - struct gl_framebuffer *fb = ctx->DrawBuffer; - int i; - GLboolean saved_fp_enable = GL_FALSE, saved_vp_enable = GL_FALSE; - GLuint saved_shader_program = 0; - unsigned int saved_active_texture; - struct gl_array_object *arraySave = NULL; - - if (!intel->clear.arrayObj) - init_clear(ctx); - - assert((mask & ~(TRI_CLEAR_COLOR_BITS | BUFFER_BIT_DEPTH | - BUFFER_BIT_STENCIL)) == 0); - - _mesa_PushAttrib(GL_COLOR_BUFFER_BIT | - GL_CURRENT_BIT | - GL_DEPTH_BUFFER_BIT | - GL_ENABLE_BIT | - GL_POLYGON_BIT | - GL_STENCIL_BUFFER_BIT | - GL_TRANSFORM_BIT | - GL_CURRENT_BIT); - saved_active_texture = ctx->Texture.CurrentUnit; - - /* Disable existing GL state we don't want to apply to a clear. */ - _mesa_Disable(GL_ALPHA_TEST); - _mesa_Disable(GL_BLEND); - _mesa_Disable(GL_CULL_FACE); - _mesa_Disable(GL_FOG); - _mesa_Disable(GL_POLYGON_SMOOTH); - _mesa_Disable(GL_POLYGON_STIPPLE); - _mesa_Disable(GL_POLYGON_OFFSET_FILL); - _mesa_Disable(GL_LIGHTING); - _mesa_Disable(GL_CLIP_PLANE0); - _mesa_Disable(GL_CLIP_PLANE1); - _mesa_Disable(GL_CLIP_PLANE2); - _mesa_Disable(GL_CLIP_PLANE3); - _mesa_Disable(GL_CLIP_PLANE4); - _mesa_Disable(GL_CLIP_PLANE5); - _mesa_PolygonMode(GL_FRONT_AND_BACK, GL_FILL); - if (ctx->Extensions.ARB_fragment_program && ctx->FragmentProgram.Enabled) { - saved_fp_enable = GL_TRUE; - _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB); - } - if (ctx->Extensions.ARB_vertex_program && ctx->VertexProgram.Enabled) { - saved_vp_enable = GL_TRUE; - _mesa_Disable(GL_VERTEX_PROGRAM_ARB); - } - if (ctx->Extensions.ARB_shader_objects && ctx->Shader.CurrentProgram) { - saved_shader_program = ctx->Shader.CurrentProgram->Name; - _mesa_UseProgramObjectARB(0); - } - - if (ctx->Texture._EnabledUnits != 0) { - int i; - - for (i = 0; i < ctx->Const.MaxTextureUnits; i++) { - _mesa_ActiveTextureARB(GL_TEXTURE0 + i); - _mesa_Disable(GL_TEXTURE_1D); - _mesa_Disable(GL_TEXTURE_2D); - _mesa_Disable(GL_TEXTURE_3D); - if (ctx->Extensions.ARB_texture_cube_map) - _mesa_Disable(GL_TEXTURE_CUBE_MAP_ARB); - if (ctx->Extensions.NV_texture_rectangle) - _mesa_Disable(GL_TEXTURE_RECTANGLE_NV); - if (ctx->Extensions.MESA_texture_array) { - _mesa_Disable(GL_TEXTURE_1D_ARRAY_EXT); - _mesa_Disable(GL_TEXTURE_2D_ARRAY_EXT); - } - } - } - - /* save current array object, bind our private one */ - _mesa_reference_array_object(ctx, &arraySave, ctx->Array.ArrayObj); - _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, intel->clear.arrayObj); - - intel_meta_set_passthrough_transform(intel); - - for (i = 0; i < 4; i++) { - COPY_4FV(intel->clear.color[i], ctx->Color.ClearColor); - } - - /* convert clear Z from [0,1] to NDC coord in [-1,1] */ - dst_z = -1.0 + 2.0 * ctx->Depth.Clear; - - /* Prepare the vertices, which are the same regardless of which buffer we're - * drawing to. - */ - intel->clear.vertices[0][0] = fb->_Xmin; - intel->clear.vertices[0][1] = fb->_Ymin; - intel->clear.vertices[0][2] = dst_z; - intel->clear.vertices[1][0] = fb->_Xmax; - intel->clear.vertices[1][1] = fb->_Ymin; - intel->clear.vertices[1][2] = dst_z; - intel->clear.vertices[2][0] = fb->_Xmax; - intel->clear.vertices[2][1] = fb->_Ymax; - intel->clear.vertices[2][2] = dst_z; - intel->clear.vertices[3][0] = fb->_Xmin; - intel->clear.vertices[3][1] = fb->_Ymax; - intel->clear.vertices[3][2] = dst_z; - - while (mask != 0) { - GLuint this_mask = 0; - GLuint color_bit; - - color_bit = _mesa_ffs(mask & TRI_CLEAR_COLOR_BITS); - if (color_bit != 0) - this_mask |= (1 << (color_bit - 1)); - - /* Clear depth/stencil in the same pass as color. */ - this_mask |= (mask & (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL)); - - /* Select the current color buffer and use the color write mask if - * we have one, otherwise don't write any color channels. - */ - if (this_mask & BUFFER_BIT_FRONT_LEFT) - _mesa_DrawBuffer(GL_FRONT_LEFT); - else if (this_mask & BUFFER_BIT_BACK_LEFT) - _mesa_DrawBuffer(GL_BACK_LEFT); - else if (color_bit != 0) - _mesa_DrawBuffer(GL_COLOR_ATTACHMENT0 + - (color_bit - BUFFER_COLOR0 - 1)); - else - _mesa_ColorMask(GL_FALSE, GL_FALSE, GL_FALSE, GL_FALSE); - - /* Control writing of the depth clear value to depth. */ - if (this_mask & BUFFER_BIT_DEPTH) { - _mesa_DepthFunc(GL_ALWAYS); - _mesa_Enable(GL_DEPTH_TEST); - } else { - _mesa_Disable(GL_DEPTH_TEST); - _mesa_DepthMask(GL_FALSE); - } - - /* Control writing of the stencil clear value to stencil. */ - if (this_mask & BUFFER_BIT_STENCIL) { - _mesa_Enable(GL_STENCIL_TEST); - _mesa_StencilOpSeparate(GL_FRONT_AND_BACK, - GL_REPLACE, GL_REPLACE, GL_REPLACE); - _mesa_StencilFuncSeparate(GL_FRONT_AND_BACK, GL_ALWAYS, - ctx->Stencil.Clear, - ctx->Stencil.WriteMask[0]); - } else { - _mesa_Disable(GL_STENCIL_TEST); - } - - _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4); - - mask &= ~this_mask; - } - - intel_meta_restore_transform(intel); - - _mesa_ActiveTextureARB(GL_TEXTURE0 + saved_active_texture); - if (saved_fp_enable) - _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB); - if (saved_vp_enable) - _mesa_Enable(GL_VERTEX_PROGRAM_ARB); - - if (saved_shader_program) - _mesa_UseProgramObjectARB(saved_shader_program); - - _mesa_PopAttrib(); - - /* restore current array object */ - _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, arraySave); - _mesa_reference_array_object(ctx, &arraySave, NULL); -} - static const char *buffer_names[] = { [BUFFER_FRONT_LEFT] = "front", [BUFFER_BACK_LEFT] = "back", @@ -340,7 +95,7 @@ intelClear(GLcontext *ctx, GLbitfield mask) = intel_get_rb_region(fb, BUFFER_STENCIL); if (stencilRegion) { /* have hw stencil */ - if (IS_965(intel->intelScreen->deviceID) || + if (stencilRegion->tiling == I915_TILING_Y || (ctx->Stencil.WriteMask[0] & 0xff) != 0xff) { /* We have to use the 3D engine if we're clearing a partial mask * of the stencil buffer, or if we're on a 965 which has a tiled @@ -357,9 +112,10 @@ intelClear(GLcontext *ctx, GLbitfield mask) /* HW depth */ if (mask & BUFFER_BIT_DEPTH) { + const struct intel_region *irb = intel_get_rb_region(fb, BUFFER_DEPTH); + /* clear depth with whatever method is used for stencil (see above) */ - if (IS_965(intel->intelScreen->deviceID) || - tri_mask & BUFFER_BIT_STENCIL) + if (irb->tiling == I915_TILING_Y || tri_mask & BUFFER_BIT_STENCIL) tri_mask |= BUFFER_BIT_DEPTH; else blit_mask |= BUFFER_BIT_DEPTH; @@ -369,7 +125,7 @@ intelClear(GLcontext *ctx, GLbitfield mask) * buffer with it. */ if (mask & (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL)) { - int color_bit = _mesa_ffs(mask & TRI_CLEAR_COLOR_BITS); + int color_bit = _mesa_ffs(mask & BUFFER_BITS_COLOR); if (color_bit != 0) { tri_mask |= blit_mask & (1 << (color_bit - 1)); blit_mask &= ~(1 << (color_bit - 1)); @@ -379,14 +135,18 @@ intelClear(GLcontext *ctx, GLbitfield mask) /* SW fallback clearing */ swrast_mask = mask & ~tri_mask & ~blit_mask; - for (i = 0; i < BUFFER_COUNT; i++) { - GLuint bufBit = 1 << i; - if ((blit_mask | tri_mask) & bufBit) { + { + /* look for non-Intel renderbuffers (clear them with swrast) */ + GLbitfield blit_or_tri = blit_mask | tri_mask; + while (blit_or_tri) { + GLuint i = _mesa_ffs(blit_or_tri) - 1; + GLbitfield bufBit = 1 << i; if (!fb->Attachment[i].Renderbuffer->ClassID) { blit_mask &= ~bufBit; tri_mask &= ~bufBit; swrast_mask |= bufBit; } + blit_or_tri ^= bufBit; } } @@ -411,7 +171,9 @@ intelClear(GLcontext *ctx, GLbitfield mask) } DBG("\n"); } - intel_clear_tris(ctx, tri_mask); + + _mesa_meta_clear(&intel->ctx, tri_mask); + intel_batchbuffer_flush(intel->batch); } if (swrast_mask) { diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c index e593b236a7..fce42e9c2d 100644 --- a/src/mesa/drivers/dri/intel/intel_context.c +++ b/src/mesa/drivers/dri/intel/intel_context.c @@ -38,6 +38,7 @@ #include "swrast_setup/swrast_setup.h" #include "tnl/tnl.h" #include "drivers/common/driverfuncs.h" +#include "drivers/common/meta.h" #include "i830_dri.h" @@ -164,6 +165,12 @@ intelGetString(GLcontext * ctx, GLenum name) case PCI_CHIP_B43_G: chipset = "Intel(R) B43"; break; + case PCI_CHIP_ILD_G: + chipset = "Intel(R) IGDNG_D"; + break; + case PCI_CHIP_ILM_G: + chipset = "Intel(R) IGDNG_M"; + break; default: chipset = "Unknown Intel Chipset"; break; @@ -401,7 +408,7 @@ intel_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h) if (!driContext->driScreenPriv->dri2.enabled) return; - if (!intel->internal_viewport_call && ctx->DrawBuffer->Name == 0) { + if (!intel->meta.internal_viewport_call && ctx->DrawBuffer->Name == 0) { /* If we're rendering to the fake front buffer, make sure all the pending * drawing has landed on the real front buffer. Otherwise when we * eventually get to DRI2GetBuffersWithFormat the stale real front @@ -587,10 +594,15 @@ intelInitDriverFunctions(struct dd_function_table *functions) functions->CopyConvolutionFilter2D = _swrast_CopyConvolutionFilter2D; intelInitTextureFuncs(functions); + intelInitTextureImageFuncs(functions); + intelInitTextureSubImageFuncs(functions); + intelInitTextureCopyImageFuncs(functions); intelInitStateFuncs(functions); intelInitClearFuncs(functions); intelInitBufferFuncs(functions); intelInitPixelFuncs(functions); + intelInitBufferObjectFuncs(functions); + intel_init_syncobj_functions(functions); } @@ -633,6 +645,10 @@ intelInitContext(struct intel_context *intel, intel->maxBatchSize = BATCH_SZ; intel->bufmgr = intelScreen->bufmgr; + + if (0) /* for debug */ + drm_intel_bufmgr_set_debug(intel->bufmgr, 1); + intel->ttm = intelScreen->ttm; if (intel->ttm) { int bo_reuse_mode; @@ -686,7 +702,15 @@ intelInitContext(struct intel_context *intel, */ _mesa_init_point(ctx); + meta_init_metaops(ctx, &intel->meta); ctx->Const.MaxColorAttachments = 4; /* XXX FBO: review this */ + if (IS_965(intelScreen->deviceID)) { + if (MAX_WIDTH > 8192) + ctx->Const.MaxRenderbufferSize = 8192; + } else { + if (MAX_WIDTH > 2048) + ctx->Const.MaxRenderbufferSize = 2048; + } /* Initialize the software rasterizer and helper modules. */ _swrast_CreateContext(ctx); @@ -698,6 +722,8 @@ intelInitContext(struct intel_context *intel, _swrast_allow_pixel_fog(ctx, GL_FALSE); _swrast_allow_vertex_fog(ctx, GL_TRUE); + _mesa_meta_init(ctx); + intel->hw_stencil = mesaVis->stencilBits && mesaVis->depthBits == 24; intel->hw_stipple = 1; @@ -744,7 +770,6 @@ intelInitContext(struct intel_context *intel, intel->batch = intel_batchbuffer_alloc(intel); - intel_bufferobj_init(intel); intel_fbo_init(intel); if (intel->ctx.Mesa_DXTn) { @@ -754,6 +779,15 @@ intelInitContext(struct intel_context *intel, else if (driQueryOptionb(&intel->optionCache, "force_s3tc_enable")) { _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); } + intel->use_texture_tiling = driQueryOptionb(&intel->optionCache, + "texture_tiling"); + if (intel->use_texture_tiling && + !intel->intelScreen->kernel_exec_fencing) { + fprintf(stderr, "No kernel support for execution fencing, " + "disabling texture tiling\n"); + intel->use_texture_tiling = GL_FALSE; + } + intel->use_early_z = driQueryOptionb(&intel->optionCache, "early_z"); intel->prim.primitive = ~0; @@ -793,8 +827,9 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) INTEL_FIREVERTICES(intel); - if (intel->clear.arrayObj) - _mesa_delete_array_object(&intel->ctx, intel->clear.arrayObj); + _mesa_meta_free(&intel->ctx); + + meta_destroy_metaops(&intel->meta); intel->vtbl.destroy(intel); @@ -817,13 +852,64 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) intel->first_post_swapbuffers_batch = NULL; if (release_texture_heaps) { - /* This share group is about to go away, free our private - * texture object data. + /* Nothing is currently done here to free texture heaps; + * but we're not using the texture heap utilities, so I + * rather think we shouldn't. I've taken a look, and can't + * find any private texture data hanging around anywhere, but + * I'm not yet certain there isn't any at all... */ - if (INTEL_DEBUG & DEBUG_TEXTURE) + /* if (INTEL_DEBUG & DEBUG_TEXTURE) fprintf(stderr, "do something to free texture heaps\n"); + */ } + /* XXX In intelMakeCurrent() below, the context's static regions are + * referenced inside the frame buffer; it's listed as a hack, + * with a comment of "XXX FBO temporary fix-ups!", but + * as long as it's there, we should release the regions here. + * The do/while loop around the block is used to allow the + * "continue" statements inside the block to exit the block, + * to avoid many layers of "if" constructs. + */ + do { + __DRIdrawablePrivate * driDrawPriv = intel->driDrawable; + struct intel_framebuffer *intel_fb; + struct intel_renderbuffer *irbDepth, *irbStencil; + if (!driDrawPriv) { + /* We're already detached from the drawable; exit this block. */ + continue; + } + intel_fb = (struct intel_framebuffer *) driDrawPriv->driverPrivate; + if (!intel_fb) { + /* The frame buffer is already gone; exit this block. */ + continue; + } + irbDepth = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH); + irbStencil = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL); + + /* If the regions of the frame buffer still match the regions + * of the context, release them. If they've changed somehow, + * leave them alone. + */ + if (intel_fb->color_rb[0] && intel_fb->color_rb[0]->region == intel->front_region) { + intel_renderbuffer_set_region(intel_fb->color_rb[0], NULL); + } + if (intel_fb->color_rb[1] && intel_fb->color_rb[1]->region == intel->back_region) { + intel_renderbuffer_set_region(intel_fb->color_rb[1], NULL); + } + + if (irbDepth && irbDepth->region == intel->depth_region) { + intel_renderbuffer_set_region(irbDepth, NULL); + } + /* Usually, the stencil buffer is the same as the depth buffer; + * but they're handled separately in MakeCurrent, so we'll + * handle them separately here. + */ + if (irbStencil && irbStencil->region == intel->depth_region) { + intel_renderbuffer_set_region(irbStencil, NULL); + } + } while (0); + intel_region_release(&intel->front_region); intel_region_release(&intel->back_region); intel_region_release(&intel->depth_region); @@ -871,7 +957,10 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, if (driDrawPriv != driReadPriv) intel_update_renderbuffers(driContextPriv, driReadPriv); } else { - /* XXX FBO temporary fix-ups! */ + /* XXX FBO temporary fix-ups! These are released in + * intelDextroyContext(), above. Changes here should be + * reflected there. + */ /* if the renderbuffers don't have regions, init them from the context */ struct intel_renderbuffer *irbDepth = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH); @@ -956,7 +1045,6 @@ intelContendedLock(struct intel_context *intel, GLuint flags) int me = intel->hHWContext; drmGetLock(intel->driFd, intel->hHWContext, flags); - intel->locked = 1; if (INTEL_DEBUG & DEBUG_LOCK) _mesa_printf("%s - got contended lock\n", __progname); @@ -1013,9 +1101,12 @@ void LOCK_HARDWARE( struct intel_context *intel ) struct intel_framebuffer *intel_fb = NULL; struct intel_renderbuffer *intel_rb = NULL; - _glthread_LOCK_MUTEX(lockMutex); - assert(!intel->locked); - intel->locked = 1; + intel->locked++; + if (intel->locked >= 2) + return; + + if (!sPriv->dri2.enabled) + _glthread_LOCK_MUTEX(lockMutex); if (intel->driDrawable) { intel_fb = intel->driDrawable->driverPrivate; @@ -1062,13 +1153,16 @@ void UNLOCK_HARDWARE( struct intel_context *intel ) { __DRIscreen *sPriv = intel->driScreen; - intel->vtbl.note_unlock( intel ); - intel->locked = 0; + intel->locked--; + if (intel->locked > 0) + return; - if (!sPriv->dri2.enabled) - DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext); + assert(intel->locked == 0); - _glthread_UNLOCK_MUTEX(lockMutex); + if (!sPriv->dri2.enabled) { + DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext); + _glthread_UNLOCK_MUTEX(lockMutex); + } if (INTEL_DEBUG & DEBUG_LOCK) _mesa_printf("%s - unlocked\n", __progname); diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h index e2b3943fb5..03e7cf39d6 100644 --- a/src/mesa/drivers/dri/intel/intel_context.h +++ b/src/mesa/drivers/dri/intel/intel_context.h @@ -33,6 +33,7 @@ #include "main/mtypes.h" #include "main/mm.h" #include "texmem.h" +#include "dri_metaops.h" #include "drm.h" #include "intel_bufmgr.h" @@ -79,9 +80,19 @@ extern void intelFallback(struct intel_context *intel, GLuint bit, #define INTEL_MAX_FIXUP 64 +struct intel_sync_object { + struct gl_sync_object Base; + + /** Batch associated with this sync object */ + drm_intel_bo *bo; +}; + +/** + * intel_context is derived from Mesa's context class: GLcontext. + */ struct intel_context { - GLcontext ctx; /* the parent class */ + GLcontext ctx; /**< base class, must be first field */ struct { @@ -91,7 +102,6 @@ struct intel_context void (*new_batch) (struct intel_context * intel); void (*emit_invarient_state) (struct intel_context * intel); void (*note_fence) (struct intel_context *intel, GLuint fence); - void (*note_unlock) (struct intel_context *intel); void (*update_texture_state) (struct intel_context * intel); void (*render_start) (struct intel_context * intel); @@ -158,19 +168,7 @@ struct intel_context void (*debug_batch)(struct intel_context *intel); } vtbl; - struct { - struct gl_fragment_program *bitmap_fp; - struct gl_vertex_program *passthrough_vp; - - struct gl_fragment_program *saved_fp; - GLboolean saved_fp_enable; - struct gl_vertex_program *saved_vp; - GLboolean saved_vp_enable; - - GLint saved_vp_x, saved_vp_y; - GLsizei saved_vp_width, saved_vp_height; - GLenum saved_matrix_mode; - } meta; + struct dri_metaops meta; GLint refcount; GLuint Fallback; @@ -182,7 +180,6 @@ struct intel_context struct intel_region *front_region; struct intel_region *back_region; struct intel_region *depth_region; - GLboolean internal_viewport_call; /** * This value indicates that the kernel memory manager is being used @@ -216,13 +213,6 @@ struct intel_context GLuint ClearColor565; GLuint ClearColor8888; - /* info for intel_clear_tris() */ - struct - { - struct gl_array_object *arrayObj; - GLfloat vertices[4][3]; - GLfloat color[4][4]; - } clear; /* Offsets of fields within the current vertex: */ @@ -304,6 +294,9 @@ struct intel_context */ GLboolean is_front_buffer_reading; + GLboolean use_texture_tiling; + GLboolean use_early_z; + drm_clip_rect_t fboRect; /**< cliprect for FBO rendering */ int perf_boxes; @@ -325,7 +318,7 @@ struct intel_context __DRIdrawablePrivate *driReadDrawable; __DRIscreenPrivate *driScreen; intelScreenPrivate *intelScreen; - volatile struct drm_i915_sarea *sarea; + volatile drm_i915_sarea_t *sarea; GLuint lastStamp; @@ -483,6 +476,8 @@ extern void intelFlush(GLcontext * ctx); extern void intelInitDriverFunctions(struct dd_function_table *functions); +void intel_init_syncobj_functions(struct dd_function_table *functions); + /* ================================================================ * intel_state.c: @@ -558,6 +553,9 @@ void intel_viewport(GLcontext * ctx, GLint x, GLint y, void intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable); +void i915_set_buf_info_for_region(uint32_t *state, struct intel_region *region, + uint32_t buffer_id); + /*====================================================================== * Inline conversion functions. * These are better-typed than the macros used previously: diff --git a/src/mesa/drivers/dri/intel/intel_extensions.c b/src/mesa/drivers/dri/intel/intel_extensions.c index 9ec1b4ec2f..2e61c556d8 100644 --- a/src/mesa/drivers/dri/intel/intel_extensions.c +++ b/src/mesa/drivers/dri/intel/intel_extensions.c @@ -30,10 +30,14 @@ #include "intel_extensions.h" +#define need_GL_ARB_copy_buffer #define need_GL_ARB_framebuffer_object +#define need_GL_ARB_map_buffer_range #define need_GL_ARB_occlusion_query #define need_GL_ARB_point_parameters #define need_GL_ARB_shader_objects +#define need_GL_ARB_sync +#define need_GL_ARB_vertex_array_object #define need_GL_ARB_vertex_program #define need_GL_ARB_vertex_shader #define need_GL_ARB_window_pos @@ -45,9 +49,12 @@ #define need_GL_EXT_fog_coord #define need_GL_EXT_framebuffer_object #define need_GL_EXT_framebuffer_blit +#define need_GL_EXT_gpu_program_parameters #define need_GL_EXT_point_parameters +#define need_GL_EXT_provoking_vertex #define need_GL_EXT_secondary_color #define need_GL_EXT_stencil_two_side +#define need_GL_APPLE_vertex_array_object #define need_GL_ATI_separate_stencil #define need_GL_ATI_envmap_bumpmap #define need_GL_NV_point_sprite @@ -65,8 +72,13 @@ * i965_dri. */ static const struct dri_extension card_extensions[] = { + { "GL_ARB_copy_buffer", GL_ARB_copy_buffer_functions }, + { "GL_ARB_half_float_pixel", NULL }, + { "GL_ARB_map_buffer_range", GL_ARB_map_buffer_range_functions }, { "GL_ARB_multitexture", NULL }, { "GL_ARB_point_parameters", GL_ARB_point_parameters_functions }, + { "GL_ARB_point_sprite", NULL }, + { "GL_ARB_sync", GL_ARB_sync_functions }, { "GL_ARB_texture_border_clamp", NULL }, { "GL_ARB_texture_cube_map", NULL }, { "GL_ARB_texture_env_add", NULL }, @@ -75,6 +87,7 @@ static const struct dri_extension card_extensions[] = { { "GL_ARB_texture_env_dot3", NULL }, { "GL_ARB_texture_mirrored_repeat", NULL }, { "GL_ARB_texture_rectangle", NULL }, + { "GL_ARB_vertex_array_object", GL_ARB_vertex_array_object_functions}, { "GL_ARB_vertex_program", GL_ARB_vertex_program_functions }, { "GL_ARB_window_pos", GL_ARB_window_pos_functions }, { "GL_EXT_blend_color", GL_EXT_blend_color_functions }, @@ -85,7 +98,9 @@ static const struct dri_extension card_extensions[] = { { "GL_EXT_blend_subtract", NULL }, { "GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions }, { "GL_EXT_fog_coord", GL_EXT_fog_coord_functions }, + { "GL_EXT_gpu_program_parameters", GL_EXT_gpu_program_parameters_functions }, { "GL_EXT_packed_depth_stencil", NULL }, + { "GL_EXT_provoking_vertex", GL_EXT_provoking_vertex_functions }, { "GL_EXT_secondary_color", GL_EXT_secondary_color_functions }, { "GL_EXT_stencil_wrap", NULL }, { "GL_EXT_texture_edge_clamp", NULL }, @@ -95,6 +110,7 @@ static const struct dri_extension card_extensions[] = { { "GL_EXT_texture_lod_bias", NULL }, { "GL_3DFX_texture_compression_FXT1", NULL }, { "GL_APPLE_client_storage", NULL }, + { "GL_APPLE_vertex_array_object", GL_APPLE_vertex_array_object_functions}, { "GL_MESA_pack_invert", NULL }, { "GL_MESA_ycbcr_texture", NULL }, { "GL_NV_blend_square", NULL }, @@ -112,8 +128,10 @@ static const struct dri_extension i915_extensions[] = { { "GL_ARB_fragment_program", NULL }, { "GL_ARB_shadow", NULL }, { "GL_ARB_texture_non_power_of_two", NULL }, + { "GL_ATI_separate_stencil", GL_ATI_separate_stencil_functions }, { "GL_ATI_texture_env_combine3", NULL }, { "GL_EXT_shadow_funcs", NULL }, + { "GL_EXT_stencil_two_side", GL_EXT_stencil_two_side_functions }, { "GL_NV_texture_env_combine4", NULL }, { NULL, NULL } }; @@ -128,6 +146,7 @@ static const struct dri_extension brw_extensions[] = { { "GL_ARB_framebuffer_object", GL_ARB_framebuffer_object_functions}, { "GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions }, { "GL_ARB_point_sprite", NULL }, + { "GL_ARB_seamless_cube_map", NULL }, { "GL_ARB_shader_objects", GL_ARB_shader_objects_functions }, { "GL_ARB_shading_language_100", GL_VERSION_2_0_functions }, { "GL_ARB_shading_language_120", GL_VERSION_2_1_functions }, diff --git a/src/mesa/drivers/dri/intel/intel_fbo.c b/src/mesa/drivers/dri/intel/intel_fbo.c index ed7c78e06c..804c034840 100644 --- a/src/mesa/drivers/dri/intel/intel_fbo.c +++ b/src/mesa/drivers/dri/intel/intel_fbo.c @@ -35,6 +35,7 @@ #include "main/context.h" #include "main/texformat.h" #include "main/texrender.h" +#include "drivers/common/meta.h" #include "intel_context.h" #include "intel_buffers.h" @@ -217,7 +218,8 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, DBG("Allocating %d x %d Intel RBO (pitch %d)\n", width, height, pitch); - irb->region = intel_region_alloc(intel, cpp, width, height, pitch, + irb->region = intel_region_alloc(intel, I915_TILING_NONE, + cpp, width, height, pitch, GL_TRUE); if (!irb->region) return GL_FALSE; /* out of memory? */ @@ -575,9 +577,10 @@ intel_render_texture(GLcontext * ctx, ASSERT(newImage); - if (newImage->Border != 0) { - /* Fallback on drawing to a texture with a border, which won't have a - * miptree. + intel_image = intel_texture_image(newImage); + if (!intel_image->mt) { + /* Fallback on drawing to a texture that doesn't have a miptree + * (has a border, width/height 0, etc.) */ _mesa_reference_renderbuffer(&att->Renderbuffer, NULL); _mesa_render_texture(ctx, fb, att); @@ -608,7 +611,6 @@ intel_render_texture(GLcontext * ctx, irb->Base.RefCount); /* point the renderbufer's region to the texture image region */ - intel_image = intel_texture_image(newImage); if (irb->region != intel_image->mt->region) { if (irb->region) intel_region_release(&irb->region); @@ -699,74 +701,6 @@ intel_validate_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb) /** - * Called from glBlitFramebuffer(). - * For now, we're doing an approximation with glCopyPixels(). - * XXX we need to bypass all the per-fragment operations, except scissor. - */ -static void -intel_blit_framebuffer(GLcontext *ctx, - GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, - GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, - GLbitfield mask, GLenum filter) -{ - const GLfloat xZoomSave = ctx->Pixel.ZoomX; - const GLfloat yZoomSave = ctx->Pixel.ZoomY; - GLsizei width, height; - GLfloat xFlip = 1.0F, yFlip = 1.0F; - - if (srcX1 < srcX0) { - GLint tmp = srcX1; - srcX1 = srcX0; - srcX0 = tmp; - xFlip = -1.0F; - } - - if (srcY1 < srcY0) { - GLint tmp = srcY1; - srcY1 = srcY0; - srcY0 = tmp; - yFlip = -1.0F; - } - - width = srcX1 - srcX0; - height = srcY1 - srcY0; - - ctx->Pixel.ZoomX = xFlip * (dstX1 - dstX0) / (srcX1 - srcY0); - ctx->Pixel.ZoomY = yFlip * (dstY1 - dstY0) / (srcY1 - srcY0); - - if (ctx->Pixel.ZoomX < 0.0F) { - dstX0 = MAX2(dstX0, dstX1); - } - else { - dstX0 = MIN2(dstX0, dstX1); - } - - if (ctx->Pixel.ZoomY < 0.0F) { - dstY0 = MAX2(dstY0, dstY1); - } - else { - dstY0 = MIN2(dstY0, dstY1); - } - - if (mask & GL_COLOR_BUFFER_BIT) { - ctx->Driver.CopyPixels(ctx, srcX0, srcY0, width, height, - dstX0, dstY0, GL_COLOR); - } - if (mask & GL_DEPTH_BUFFER_BIT) { - ctx->Driver.CopyPixels(ctx, srcX0, srcY0, width, height, - dstX0, dstY0, GL_DEPTH); - } - if (mask & GL_STENCIL_BUFFER_BIT) { - ctx->Driver.CopyPixels(ctx, srcX0, srcY0, width, height, - dstX0, dstY0, GL_STENCIL); - } - - ctx->Pixel.ZoomX = xZoomSave; - ctx->Pixel.ZoomY = yZoomSave; -} - - -/** * Do one-time context initializations related to GL_EXT_framebuffer_object. * Hook in device driver functions. */ @@ -781,5 +715,5 @@ intel_fbo_init(struct intel_context *intel) intel->ctx.Driver.FinishRenderTexture = intel_finish_render_texture; intel->ctx.Driver.ResizeBuffers = intel_resize_buffers; intel->ctx.Driver.ValidateFramebuffer = intel_validate_framebuffer; - intel->ctx.Driver.BlitFramebuffer = intel_blit_framebuffer; + intel->ctx.Driver.BlitFramebuffer = _mesa_meta_blit_framebuffer; } diff --git a/src/mesa/drivers/dri/intel/intel_generatemipmap.c b/src/mesa/drivers/dri/intel/intel_generatemipmap.c new file mode 100644 index 0000000000..12059e122c --- /dev/null +++ b/src/mesa/drivers/dri/intel/intel_generatemipmap.c @@ -0,0 +1,304 @@ +/* + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * Copyright © 2009 Intel Corporation + * + * 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 (including the next + * paragraph) 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 + * THE AUTHORS OR COPYRIGHT HOLDERS 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. + * + * Authors: + * Eric Anholt <eric@anholt.net> + * + */ + +#include "main/glheader.h" +#include "main/enums.h" +#include "main/image.h" +#include "main/mtypes.h" +#include "main/macros.h" +#include "main/bufferobj.h" +#include "main/teximage.h" +#include "main/texenv.h" +#include "main/texobj.h" +#include "main/texstate.h" +#include "main/texparam.h" +#include "main/varray.h" +#include "main/attrib.h" +#include "main/enable.h" +#include "main/buffers.h" +#include "main/fbobject.h" +#include "main/framebuffer.h" +#include "main/renderbuffer.h" +#include "main/depth.h" +#include "main/hash.h" +#include "main/mipmap.h" +#include "main/blend.h" +#include "glapi/dispatch.h" +#include "swrast/swrast.h" + +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_batchbuffer.h" +#include "intel_pixel.h" +#include "intel_tex.h" +#include "intel_mipmap_tree.h" + +static const char *intel_fp_tex2d = + "!!ARBfp1.0\n" + "TEX result.color, fragment.texcoord[0], texture[0], 2D;\n" + "END\n"; + +static GLboolean +intel_generate_mipmap_level(GLcontext *ctx, GLuint tex_name, + int level, int width, int height) +{ + struct intel_context *intel = intel_context(ctx); + GLfloat vertices[4][2]; + GLint status; + + /* Set to source from the previous level */ + _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_BASE_LEVEL, level - 1); + _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, level - 1); + + /* Set to draw into the current level */ + _mesa_FramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT, + GL_COLOR_ATTACHMENT0_EXT, + GL_TEXTURE_2D, + tex_name, + level); + /* Choose to render to the color attachment. */ + _mesa_DrawBuffer(GL_COLOR_ATTACHMENT0_EXT); + + status = _mesa_CheckFramebufferStatusEXT (GL_FRAMEBUFFER_EXT); + if (status != GL_FRAMEBUFFER_COMPLETE_EXT) + return GL_FALSE; + + meta_set_passthrough_transform(&intel->meta); + + /* XXX: Doing it right would involve setting up the transformation to do + * 0-1 mapping or something, and not changing the vertex data. + */ + vertices[0][0] = 0; + vertices[0][1] = 0; + vertices[1][0] = width; + vertices[1][1] = 0; + vertices[2][0] = width; + vertices[2][1] = height; + vertices[3][0] = 0; + vertices[3][1] = height; + + _mesa_VertexPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &vertices); + _mesa_Enable(GL_VERTEX_ARRAY); + meta_set_default_texrect(&intel->meta); + + _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4); + + meta_restore_texcoords(&intel->meta); + meta_restore_transform(&intel->meta); + + return GL_TRUE; +} + +static GLboolean +intel_generate_mipmap_2d(GLcontext *ctx, + GLenum target, + struct gl_texture_object *texObj) +{ + struct intel_context *intel = intel_context(ctx); + GLint old_active_texture; + int level, max_levels, start_level, end_level; + GLuint fb_name; + GLboolean success = GL_FALSE; + struct gl_framebuffer *saved_fbo = NULL; + struct gl_buffer_object *saved_array_buffer = NULL; + struct gl_buffer_object *saved_element_buffer = NULL; + + _mesa_PushAttrib(GL_ENABLE_BIT | GL_TEXTURE_BIT | + GL_CURRENT_BIT | GL_COLOR_BUFFER_BIT | + GL_DEPTH_BUFFER_BIT); + _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT); + old_active_texture = ctx->Texture.CurrentUnit; + _mesa_reference_framebuffer(&saved_fbo, ctx->DrawBuffer); + + /* use default array/index buffers */ + _mesa_reference_buffer_object(ctx, &saved_array_buffer, + ctx->Array.ArrayBufferObj); + _mesa_reference_buffer_object(ctx, &ctx->Array.ArrayBufferObj, + ctx->Shared->NullBufferObj); + _mesa_reference_buffer_object(ctx, &saved_element_buffer, + ctx->Array.ElementArrayBufferObj); + _mesa_reference_buffer_object(ctx, &ctx->Array.ElementArrayBufferObj, + ctx->Shared->NullBufferObj); + + _mesa_Disable(GL_POLYGON_STIPPLE); + _mesa_Disable(GL_DEPTH_TEST); + _mesa_Disable(GL_STENCIL_TEST); + _mesa_ColorMask(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE); + _mesa_DepthMask(GL_FALSE); + + /* Bind the given texture to GL_TEXTURE_2D with linear filtering for our + * minification. + */ + _mesa_ActiveTextureARB(GL_TEXTURE0_ARB); + _mesa_Enable(GL_TEXTURE_2D); + _mesa_BindTexture(GL_TEXTURE_2D, texObj->Name); + _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, + GL_LINEAR_MIPMAP_NEAREST); + _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR); + + /* Bind the new renderbuffer to the color attachment point. */ + _mesa_GenFramebuffersEXT(1, &fb_name); + _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, fb_name); + + meta_set_fragment_program(&intel->meta, &intel->meta.tex2d_fp, + intel_fp_tex2d); + meta_set_passthrough_vertex_program(&intel->meta); + + max_levels = _mesa_max_texture_levels(ctx, texObj->Target); + start_level = texObj->BaseLevel; + end_level = texObj->MaxLevel; + + /* Loop generating level+1 from level. */ + for (level = start_level; level < end_level && level < max_levels - 1; level++) { + const struct gl_texture_image *srcImage; + int width, height; + + srcImage = _mesa_select_tex_image(ctx, texObj, target, level); + if (srcImage->Border != 0) + goto fail; + + width = srcImage->Width / 2; + if (width < 1) + width = 1; + height = srcImage->Height / 2; + if (height < 1) + height = 1; + + if (width == srcImage->Width && + height == srcImage->Height) { + /* Neither _mesa_max_texture_levels nor texObj->MaxLevel are the + * maximum texture level for the object, so break out when we've gone + * over the edge. + */ + break; + } + + /* Make sure that there's space allocated for the target level. + * We could skip this if there's already space allocated and save some + * time. + */ + _mesa_TexImage2D(GL_TEXTURE_2D, level + 1, srcImage->InternalFormat, + width, height, 0, + GL_RGBA, GL_UNSIGNED_INT, NULL); + + if (!intel_generate_mipmap_level(ctx, texObj->Name, level + 1, + width, height)) + goto fail; + } + + success = GL_TRUE; + +fail: + meta_restore_fragment_program(&intel->meta); + meta_restore_vertex_program(&intel->meta); + + /* restore array/index buffers */ + _mesa_reference_buffer_object(ctx, &ctx->Array.ArrayBufferObj, + saved_array_buffer); + _mesa_reference_buffer_object(ctx, &saved_array_buffer, NULL); + _mesa_reference_buffer_object(ctx, &ctx->Array.ElementArrayBufferObj, + saved_element_buffer); + _mesa_reference_buffer_object(ctx, &saved_element_buffer, NULL); + + + _mesa_DeleteFramebuffersEXT(1, &fb_name); + _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture); + if (saved_fbo) + _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, saved_fbo->Name); + _mesa_reference_framebuffer(&saved_fbo, NULL); + _mesa_PopClientAttrib(); + _mesa_PopAttrib(); + + return success; +} + + +/** + * Generate new mipmap data from BASE+1 to BASE+p (the minimally-sized mipmap + * level). + * + * The texture object's miptree must be mapped. + * + * It would be really nice if this was just called by Mesa whenever mipmaps + * needed to be regenerated, rather than us having to remember to do so in + * each texture image modification path. + * + * This function should also include an accelerated path. + */ +void +intel_generate_mipmap(GLcontext *ctx, GLenum target, + struct gl_texture_object *texObj) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_texture_object *intelObj = intel_texture_object(texObj); + GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; + int face, i; + + /* HW path */ + if (target == GL_TEXTURE_2D && + ctx->Extensions.EXT_framebuffer_object && + ctx->Extensions.ARB_fragment_program && + ctx->Extensions.ARB_vertex_program) { + GLboolean success; + + /* We'll be accessing this texture using GL entrypoints, which should + * be resilient against other access to this texture. + */ + _mesa_unlock_texture(ctx, texObj); + success = intel_generate_mipmap_2d(ctx, target, texObj); + _mesa_lock_texture(ctx, texObj); + + if (success) + return; + } + + /* SW path */ + intel_tex_map_level_images(intel, intelObj, texObj->BaseLevel); + _mesa_generate_mipmap(ctx, target, texObj); + intel_tex_unmap_level_images(intel, intelObj, texObj->BaseLevel); + + /* Update the level information in our private data in the new images, since + * it didn't get set as part of a normal TexImage path. + */ + for (face = 0; face < nr_faces; face++) { + for (i = texObj->BaseLevel + 1; i < texObj->MaxLevel; i++) { + struct intel_texture_image *intelImage; + + intelImage = intel_texture_image(texObj->Image[face][i]); + if (intelImage == NULL) + break; + + intelImage->level = i; + intelImage->face = face; + /* Unreference the miptree to signal that the new Data is a bare + * pointer from mesa. + */ + intel_miptree_release(intel, &intelImage->mt); + } + } +} diff --git a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c index 6e1e034e53..c985da5aa2 100644 --- a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c +++ b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c @@ -57,14 +57,16 @@ intel_miptree_create_internal(struct intel_context *intel, GLuint last_level, GLuint width0, GLuint height0, - GLuint depth0, GLuint cpp, GLuint compress_byte) + GLuint depth0, GLuint cpp, GLuint compress_byte, + uint32_t tiling) { GLboolean ok; struct intel_mipmap_tree *mt = calloc(sizeof(*mt), 1); - DBG("%s target %s format %s level %d..%d\n", __FUNCTION__, + DBG("%s target %s format %s level %d..%d <-- %p\n", __FUNCTION__, _mesa_lookup_enum_by_nr(target), - _mesa_lookup_enum_by_nr(internal_format), first_level, last_level); + _mesa_lookup_enum_by_nr(internal_format), + first_level, last_level, mt); mt->target = target_to_target(target); mt->internal_format = internal_format; @@ -80,15 +82,16 @@ intel_miptree_create_internal(struct intel_context *intel, #ifdef I915 if (IS_945(intel->intelScreen->deviceID)) - ok = i945_miptree_layout(intel, mt); + ok = i945_miptree_layout(intel, mt, tiling); else - ok = i915_miptree_layout(intel, mt); + ok = i915_miptree_layout(intel, mt, tiling); #else - ok = brw_miptree_layout(intel, mt); + ok = brw_miptree_layout(intel, mt, tiling); #endif if (!ok) { free(mt); + DBG("%s not okay - returning NULL\n", __FUNCTION__); return NULL; } @@ -98,6 +101,7 @@ intel_miptree_create_internal(struct intel_context *intel, struct intel_mipmap_tree * intel_miptree_create(struct intel_context *intel, GLenum target, + GLenum base_format, GLenum internal_format, GLuint first_level, GLuint last_level, @@ -107,10 +111,23 @@ intel_miptree_create(struct intel_context *intel, GLboolean expect_accelerated_upload) { struct intel_mipmap_tree *mt; + uint32_t tiling; + + if (intel->use_texture_tiling && compress_byte == 0 && + intel->intelScreen->kernel_exec_fencing) { + if (IS_965(intel->intelScreen->deviceID) && + (base_format == GL_DEPTH_COMPONENT || + base_format == GL_DEPTH_STENCIL_EXT)) + tiling = I915_TILING_Y; + else + tiling = I915_TILING_X; + } else + tiling = I915_TILING_NONE; mt = intel_miptree_create_internal(intel, target, internal_format, first_level, last_level, width0, - height0, depth0, cpp, compress_byte); + height0, depth0, cpp, compress_byte, + tiling); /* * pitch == 0 || height == 0 indicates the null texture */ @@ -118,6 +135,7 @@ intel_miptree_create(struct intel_context *intel, return NULL; mt->region = intel_region_alloc(intel, + tiling, mt->cpp, mt->pitch, mt->total_height, @@ -147,7 +165,8 @@ intel_miptree_create_for_region(struct intel_context *intel, mt = intel_miptree_create_internal(intel, target, internal_format, first_level, last_level, region->width, region->height, 1, - region->cpp, compress_byte); + region->cpp, compress_byte, + I915_TILING_NONE); if (!mt) return mt; #if 0 @@ -185,6 +204,7 @@ intel_miptree_create_for_region(struct intel_context *intel, int intel_miptree_pitch_align (struct intel_context *intel, struct intel_mipmap_tree *mt, + uint32_t tiling, int pitch) { #ifdef I915 @@ -205,6 +225,11 @@ int intel_miptree_pitch_align (struct intel_context *intel, pitch_align = 4; } + if (tiling == I915_TILING_X) + pitch_align = 512; + else if (tiling == I915_TILING_Y) + pitch_align = 128; + pitch = ALIGN(pitch * mt->cpp, pitch_align); #ifdef I915 @@ -328,23 +353,31 @@ intel_miptree_set_level_info(struct intel_mipmap_tree *mt, } - void -intel_miptree_set_image_offset(struct intel_mipmap_tree *mt, - GLuint level, GLuint img, - GLuint x, GLuint y) +intel_miptree_set_image_offset_ex(struct intel_mipmap_tree *mt, + GLuint level, GLuint img, + GLuint x, GLuint y, + GLuint offset) { if (img == 0 && level == 0) assert(x == 0 && y == 0); assert(img < mt->level[level].nr_images); - mt->level[level].image_offset[img] = (x + y * mt->pitch) * mt->cpp; + mt->level[level].image_offset[img] = (x + y * mt->pitch) * mt->cpp + offset; DBG("%s level %d img %d pos %d,%d image_offset %x\n", __FUNCTION__, level, img, x, y, mt->level[level].image_offset[img]); } +void +intel_miptree_set_image_offset(struct intel_mipmap_tree *mt, + GLuint level, GLuint img, + GLuint x, GLuint y) +{ + intel_miptree_set_image_offset_ex(mt, level, img, x, y, 0); +} + /* Although we use the image_offset[] array to store relative offsets * to cube faces, Mesa doesn't know anything about this and expects @@ -454,11 +487,11 @@ intel_miptree_image_data(struct intel_context *intel, 0, 0, /* source x, y */ dst->level[level].width, height); /* width, height */ - src += src_image_pitch * dst->cpp; + src = (char *)src + src_image_pitch * dst->cpp; } } -extern GLuint intel_compressed_alignment(GLenum); +extern void intel_get_texture_alignment_unit(GLenum, GLuint *, GLuint *); /* Copy mipmap image between trees */ void @@ -475,20 +508,37 @@ intel_miptree_image_copy(struct intel_context *intel, const GLuint *dst_depth_offset = intel_miptree_depth_offsets(dst, level); const GLuint *src_depth_offset = intel_miptree_depth_offsets(src, level); GLuint i; + GLboolean success; if (dst->compressed) { - GLuint alignment = intel_compressed_alignment(dst->internal_format); + GLuint align_w, align_h; + + intel_get_texture_alignment_unit(dst->internal_format, &align_w, &align_h); height = (height + 3) / 4; - width = ((width + alignment - 1) & ~(alignment - 1)); + width = ALIGN(width, align_w); } for (i = 0; i < depth; i++) { - intel_region_copy(intel, - dst->region, dst_offset + dst_depth_offset[i], - 0, - 0, - src->region, src_offset + src_depth_offset[i], - 0, 0, width, height); + success = intel_region_copy(intel, + dst->region, dst_offset + dst_depth_offset[i], + 0, 0, + src->region, src_offset + src_depth_offset[i], + 0, 0, width, height, GL_COPY); + if (!success) { + GLubyte *src_ptr, *dst_ptr; + + src_ptr = intel_region_map(intel, src->region); + dst_ptr = intel_region_map(intel, dst->region); + + _mesa_copy_rect(dst_ptr + dst_offset + dst_depth_offset[i], + dst->cpp, + dst->pitch, + 0, 0, width, height, + src_ptr + src_offset + src_depth_offset[i], + src->pitch, + 0, 0); + intel_region_unmap(intel, src->region); + intel_region_unmap(intel, dst->region); + } } - } diff --git a/src/mesa/drivers/dri/intel/intel_mipmap_tree.h b/src/mesa/drivers/dri/intel/intel_mipmap_tree.h index 4060b9df78..c890b2a0d0 100644 --- a/src/mesa/drivers/dri/intel/intel_mipmap_tree.h +++ b/src/mesa/drivers/dri/intel/intel_mipmap_tree.h @@ -126,6 +126,7 @@ struct intel_mipmap_tree struct intel_mipmap_tree *intel_miptree_create(struct intel_context *intel, GLenum target, + GLenum base_format, GLenum internal_format, GLuint first_level, GLuint last_level, @@ -148,6 +149,7 @@ intel_miptree_create_for_region(struct intel_context *intel, int intel_miptree_pitch_align (struct intel_context *intel, struct intel_mipmap_tree *mt, + uint32_t tiling, int pitch); void intel_miptree_reference(struct intel_mipmap_tree **dst, @@ -194,6 +196,11 @@ void intel_miptree_set_level_info(struct intel_mipmap_tree *mt, GLuint x, GLuint y, GLuint w, GLuint h, GLuint d); +void intel_miptree_set_image_offset_ex(struct intel_mipmap_tree *mt, + GLuint level, + GLuint img, GLuint x, GLuint y, + GLuint offset); + void intel_miptree_set_image_offset(struct intel_mipmap_tree *mt, GLuint level, GLuint img, GLuint x, GLuint y); @@ -218,10 +225,13 @@ void intel_miptree_image_copy(struct intel_context *intel, /* i915_mipmap_tree.c: */ GLboolean i915_miptree_layout(struct intel_context *intel, - struct intel_mipmap_tree *mt); + struct intel_mipmap_tree *mt, + uint32_t tiling); GLboolean i945_miptree_layout(struct intel_context *intel, - struct intel_mipmap_tree *mt); + struct intel_mipmap_tree *mt, + uint32_t tiling); GLboolean brw_miptree_layout(struct intel_context *intel, - struct intel_mipmap_tree *mt); + struct intel_mipmap_tree *mt, + uint32_t tiling); #endif diff --git a/src/mesa/drivers/dri/intel/intel_pixel.c b/src/mesa/drivers/dri/intel/intel_pixel.c index defb80f85d..a300141655 100644 --- a/src/mesa/drivers/dri/intel/intel_pixel.c +++ b/src/mesa/drivers/dri/intel/intel_pixel.c @@ -27,9 +27,12 @@ #include "main/enums.h" #include "main/state.h" +#include "main/bufferobj.h" #include "main/context.h" #include "main/enable.h" #include "main/matrix.h" +#include "main/texstate.h" +#include "main/varray.h" #include "main/viewport.h" #include "swrast/swrast.h" #include "shader/arbprogram.h" @@ -174,167 +177,6 @@ intel_check_blit_format(struct intel_region * region, } void -intel_meta_set_passthrough_transform(struct intel_context *intel) -{ - GLcontext *ctx = &intel->ctx; - - intel->meta.saved_vp_x = ctx->Viewport.X; - intel->meta.saved_vp_y = ctx->Viewport.Y; - intel->meta.saved_vp_width = ctx->Viewport.Width; - intel->meta.saved_vp_height = ctx->Viewport.Height; - intel->meta.saved_matrix_mode = ctx->Transform.MatrixMode; - - intel->internal_viewport_call = GL_TRUE; - _mesa_Viewport(0, 0, ctx->DrawBuffer->Width, ctx->DrawBuffer->Height); - intel->internal_viewport_call = GL_FALSE; - - _mesa_MatrixMode(GL_PROJECTION); - _mesa_PushMatrix(); - _mesa_LoadIdentity(); - _mesa_Ortho(0, ctx->DrawBuffer->Width, 0, ctx->DrawBuffer->Height, 1, -1); - - _mesa_MatrixMode(GL_MODELVIEW); - _mesa_PushMatrix(); - _mesa_LoadIdentity(); -} - -void -intel_meta_restore_transform(struct intel_context *intel) -{ - _mesa_MatrixMode(GL_PROJECTION); - _mesa_PopMatrix(); - _mesa_MatrixMode(GL_MODELVIEW); - _mesa_PopMatrix(); - - _mesa_MatrixMode(intel->meta.saved_matrix_mode); - - intel->internal_viewport_call = GL_TRUE; - _mesa_Viewport(intel->meta.saved_vp_x, intel->meta.saved_vp_y, - intel->meta.saved_vp_width, intel->meta.saved_vp_height); - intel->internal_viewport_call = GL_FALSE; -} - -/** - * Set up a vertex program to pass through the position and first texcoord - * for pixel path. - */ -void -intel_meta_set_passthrough_vertex_program(struct intel_context *intel) -{ - GLcontext *ctx = &intel->ctx; - static const char *vp = - "!!ARBvp1.0\n" - "TEMP vertexClip;\n" - "DP4 vertexClip.x, state.matrix.mvp.row[0], vertex.position;\n" - "DP4 vertexClip.y, state.matrix.mvp.row[1], vertex.position;\n" - "DP4 vertexClip.z, state.matrix.mvp.row[2], vertex.position;\n" - "DP4 vertexClip.w, state.matrix.mvp.row[3], vertex.position;\n" - "MOV result.position, vertexClip;\n" - "MOV result.texcoord[0], vertex.texcoord[0];\n" - "MOV result.color, vertex.color;\n" - "END\n"; - - assert(intel->meta.saved_vp == NULL); - - _mesa_reference_vertprog(ctx, &intel->meta.saved_vp, - ctx->VertexProgram.Current); - if (intel->meta.passthrough_vp == NULL) { - GLuint prog_name; - _mesa_GenPrograms(1, &prog_name); - _mesa_BindProgram(GL_VERTEX_PROGRAM_ARB, prog_name); - _mesa_ProgramStringARB(GL_VERTEX_PROGRAM_ARB, - GL_PROGRAM_FORMAT_ASCII_ARB, - strlen(vp), (const GLubyte *)vp); - _mesa_reference_vertprog(ctx, &intel->meta.passthrough_vp, - ctx->VertexProgram.Current); - _mesa_DeletePrograms(1, &prog_name); - } - - FLUSH_VERTICES(ctx, _NEW_PROGRAM); - _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current, - intel->meta.passthrough_vp); - ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB, - &intel->meta.passthrough_vp->Base); - - intel->meta.saved_vp_enable = ctx->VertexProgram.Enabled; - _mesa_Enable(GL_VERTEX_PROGRAM_ARB); -} - -/** - * Restores the previous vertex program after - * intel_meta_set_passthrough_vertex_program() - */ -void -intel_meta_restore_vertex_program(struct intel_context *intel) -{ - GLcontext *ctx = &intel->ctx; - - FLUSH_VERTICES(ctx, _NEW_PROGRAM); - _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current, - intel->meta.saved_vp); - _mesa_reference_vertprog(ctx, &intel->meta.saved_vp, NULL); - ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB, - &ctx->VertexProgram.Current->Base); - - if (!intel->meta.saved_vp_enable) - _mesa_Disable(GL_VERTEX_PROGRAM_ARB); -} - -/** - * Binds the given program string to GL_FRAGMENT_PROGRAM_ARB, caching the - * program object. - */ -void -intel_meta_set_fragment_program(struct intel_context *intel, - struct gl_fragment_program **prog, - const char *prog_string) -{ - GLcontext *ctx = &intel->ctx; - assert(intel->meta.saved_fp == NULL); - - _mesa_reference_fragprog(ctx, &intel->meta.saved_fp, - ctx->FragmentProgram.Current); - if (*prog == NULL) { - GLuint prog_name; - _mesa_GenPrograms(1, &prog_name); - _mesa_BindProgram(GL_FRAGMENT_PROGRAM_ARB, prog_name); - _mesa_ProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, - GL_PROGRAM_FORMAT_ASCII_ARB, - strlen(prog_string), (const GLubyte *)prog_string); - _mesa_reference_fragprog(ctx, prog, ctx->FragmentProgram.Current); - /* Note that DeletePrograms unbinds the program on us */ - _mesa_DeletePrograms(1, &prog_name); - } - - FLUSH_VERTICES(ctx, _NEW_PROGRAM); - _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, *prog); - ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, &((*prog)->Base)); - - intel->meta.saved_fp_enable = ctx->FragmentProgram.Enabled; - _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB); -} - -/** - * Restores the previous fragment program after - * intel_meta_set_fragment_program() - */ -void -intel_meta_restore_fragment_program(struct intel_context *intel) -{ - GLcontext *ctx = &intel->ctx; - - FLUSH_VERTICES(ctx, _NEW_PROGRAM); - _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, - intel->meta.saved_fp); - _mesa_reference_fragprog(ctx, &intel->meta.saved_fp, NULL); - ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, - &ctx->FragmentProgram.Current->Base); - - if (!intel->meta.saved_fp_enable) - _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB); -} - -void intelInitPixelFuncs(struct dd_function_table *functions) { functions->Accum = _swrast_Accum; @@ -346,12 +188,3 @@ intelInitPixelFuncs(struct dd_function_table *functions) functions->ReadPixels = intelReadPixels; } -void -intel_free_pixel_state(struct intel_context *intel) -{ - GLcontext *ctx = &intel->ctx; - - _mesa_reference_vertprog(ctx, &intel->meta.passthrough_vp, NULL); - _mesa_reference_fragprog(ctx, &intel->meta.bitmap_fp, NULL); -} - diff --git a/src/mesa/drivers/dri/intel/intel_pixel.h b/src/mesa/drivers/dri/intel/intel_pixel.h index cb41fa182c..96a6dd17b2 100644 --- a/src/mesa/drivers/dri/intel/intel_pixel.h +++ b/src/mesa/drivers/dri/intel/intel_pixel.h @@ -31,16 +31,6 @@ #include "main/mtypes.h" void intelInitPixelFuncs(struct dd_function_table *functions); -void intel_meta_set_passthrough_transform(struct intel_context *intel); -void intel_meta_restore_transform(struct intel_context *intel); -void intel_meta_set_passthrough_vertex_program(struct intel_context *intel); -void intel_meta_restore_vertex_program(struct intel_context *intel); -void intel_meta_set_fragment_program(struct intel_context *intel, - struct gl_fragment_program **prog, - const char *prog_string); -void intel_meta_restore_fragment_program(struct intel_context *intel); -void intel_free_pixel_state(struct intel_context *intel); - GLboolean intel_check_blit_fragment_ops(GLcontext * ctx, GLboolean src_alpha_is_one); @@ -76,6 +66,4 @@ void intelBitmap(GLcontext * ctx, const struct gl_pixelstore_attrib *unpack, const GLubyte * pixels); -void intel_clear_tris(GLcontext *ctx, GLbitfield mask); - #endif diff --git a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c index d137aef13d..b543a0bbc3 100644 --- a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c +++ b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c @@ -42,6 +42,7 @@ #include "main/varray.h" #include "main/attrib.h" #include "main/enable.h" +#include "main/viewport.h" #include "shader/arbprogram.h" #include "glapi/dispatch.h" #include "swrast/swrast.h" @@ -92,19 +93,12 @@ static const GLubyte *map_pbo( GLcontext *ctx, return ADD_POINTERS(buf, bitmap); } -static GLboolean test_bit( const GLubyte *src, - GLuint bit ) +static GLboolean test_bit( const GLubyte *src, GLuint bit ) { return (src[bit/8] & (1<<(bit % 8))) ? 1 : 0; } -static GLboolean test_msb_bit(const GLubyte *src, GLuint bit) -{ - return (src[bit/8] & (1<<(7 - (bit % 8)))) ? 1 : 0; -} - -static void set_bit( GLubyte *dest, - GLuint bit ) +static void set_bit( GLubyte *dest, GLuint bit ) { dest[bit/8] |= 1 << (bit % 8); } @@ -194,7 +188,7 @@ do_blit_bitmap( GLcontext *ctx, struct gl_framebuffer *fb = ctx->DrawBuffer; GLfloat tmpColor[4]; GLubyte ubcolor[4]; - GLuint color8888, color565; + GLuint color; unsigned int num_cliprects; drm_clip_rect_t *cliprects; int x_off, y_off; @@ -232,8 +226,11 @@ do_blit_bitmap( GLcontext *ctx, UNCLAMPED_FLOAT_TO_UBYTE(ubcolor[2], tmpColor[2]); UNCLAMPED_FLOAT_TO_UBYTE(ubcolor[3], tmpColor[3]); - color8888 = INTEL_PACKCOLOR8888(ubcolor[0], ubcolor[1], ubcolor[2], ubcolor[3]); - color565 = INTEL_PACKCOLOR565(ubcolor[0], ubcolor[1], ubcolor[2]); + if (dst->cpp == 2) + color = INTEL_PACKCOLOR565(ubcolor[0], ubcolor[1], ubcolor[2]); + else + color = INTEL_PACKCOLOR8888(ubcolor[0], ubcolor[1], + ubcolor[2], ubcolor[3]); if (!intel_check_blit_fragment_ops(ctx, tmpColor[3] == 1.0F)) return GL_FALSE; @@ -307,21 +304,21 @@ do_blit_bitmap( GLcontext *ctx, fb->Name == 0 ? GL_TRUE : GL_FALSE) == 0) continue; - /* - */ - intelEmitImmediateColorExpandBlit( intel, - dst->cpp, - (GLubyte *)stipple, - sz, - (dst->cpp == 2) ? color565 : color8888, - dst->pitch, - dst->buffer, - 0, - dst->tiling, - box_x + px, - box_y + py, - w, h, - logic_op); + if (!intelEmitImmediateColorExpandBlit(intel, + dst->cpp, + (GLubyte *)stipple, + sz, + color, + dst->pitch, + dst->buffer, + 0, + dst->tiling, + box_x + px, + box_y + py, + w, h, + logic_op)) { + return GL_FALSE; + } } } } @@ -360,11 +357,8 @@ intel_texture_bitmap(GLcontext * ctx, "END\n"; GLuint texname; GLfloat vertices[4][4]; - GLfloat texcoords[4][2]; GLint old_active_texture; - GLubyte *unpacked_bitmap; GLubyte *a8_bitmap; - int x, y; GLfloat dst_z; /* We need a fragment program for the KIL effect */ @@ -424,22 +418,16 @@ intel_texture_bitmap(GLcontext * ctx, return GL_FALSE; } - /* Convert the A1 bitmap to an A8 format suitable for glTexImage */ if (unpack->BufferObj->Name) { bitmap = map_pbo(ctx, width, height, unpack, bitmap); if (bitmap == NULL) return GL_TRUE; /* even though this is an error, we're done */ } - unpacked_bitmap = _mesa_unpack_bitmap(width, height, bitmap, - unpack); + + /* Convert the A1 bitmap to an A8 format suitable for glTexImage */ a8_bitmap = _mesa_calloc(width * height); - for (y = 0; y < height; y++) { - for (x = 0; x < width; x++) { - if (test_msb_bit(unpacked_bitmap, ALIGN(width, 8) * y + x)) - a8_bitmap[y * width + x] = 0xff; - } - } - _mesa_free(unpacked_bitmap); + _mesa_expand_bitmap(width, height, unpack, bitmap, a8_bitmap, width, 0xff); + if (unpack->BufferObj->Name) { /* done with PBO so unmap it now */ ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, @@ -473,15 +461,18 @@ intel_texture_bitmap(GLcontext * ctx, GL_ALPHA, GL_UNSIGNED_BYTE, a8_bitmap); _mesa_free(a8_bitmap); - intel_meta_set_fragment_program(intel, &intel->meta.bitmap_fp, fp); + meta_set_fragment_program(&intel->meta, &intel->meta.bitmap_fp, fp); _mesa_ProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, 0, ctx->Current.RasterColor); - intel_meta_set_passthrough_vertex_program(intel); - intel_meta_set_passthrough_transform(intel); + meta_set_passthrough_vertex_program(&intel->meta); + meta_set_passthrough_transform(&intel->meta); /* convert rasterpos Z from [0,1] to NDC coord in [-1,1] */ dst_z = -1.0 + 2.0 * ctx->Current.RasterPos[2]; + /* RasterPos[2] already takes into account the DepthRange mapping. */ + _mesa_DepthRange(0.0, 1.0); + vertices[0][0] = dst_x; vertices[0][1] = dst_y; vertices[0][2] = dst_z; @@ -499,25 +490,15 @@ intel_texture_bitmap(GLcontext * ctx, vertices[3][2] = dst_z; vertices[3][3] = 1.0; - texcoords[0][0] = 0.0; - texcoords[0][1] = 0.0; - texcoords[1][0] = 1.0; - texcoords[1][1] = 0.0; - texcoords[2][0] = 1.0; - texcoords[2][1] = 1.0; - texcoords[3][0] = 0.0; - texcoords[3][1] = 1.0; - _mesa_VertexPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), &vertices); - _mesa_ClientActiveTextureARB(GL_TEXTURE0); - _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &texcoords); _mesa_Enable(GL_VERTEX_ARRAY); - _mesa_Enable(GL_TEXTURE_COORD_ARRAY); + meta_set_default_texrect(&intel->meta); _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4); - intel_meta_restore_transform(intel); - intel_meta_restore_fragment_program(intel); - intel_meta_restore_vertex_program(intel); + meta_restore_texcoords(&intel->meta); + meta_restore_transform(&intel->meta); + meta_restore_fragment_program(&intel->meta); + meta_restore_vertex_program(&intel->meta); _mesa_PopClientAttrib(); _mesa_Disable(GL_TEXTURE_2D); /* asserted that it was disabled at entry */ diff --git a/src/mesa/drivers/dri/intel/intel_pixel_copy.c b/src/mesa/drivers/dri/intel/intel_pixel_copy.c index d50dd68092..07ca8f7ddb 100644 --- a/src/mesa/drivers/dri/intel/intel_pixel_copy.c +++ b/src/mesa/drivers/dri/intel/intel_pixel_copy.c @@ -26,18 +26,13 @@ **************************************************************************/ #include "main/glheader.h" -#include "main/enums.h" #include "main/image.h" #include "main/state.h" #include "main/mtypes.h" -#include "main/macros.h" -#include "swrast/swrast.h" +#include "drivers/common/meta.h" -#include "intel_screen.h" #include "intel_context.h" -#include "intel_batchbuffer.h" #include "intel_buffers.h" -#include "intel_blit.h" #include "intel_regions.h" #include "intel_pixel.h" @@ -97,162 +92,6 @@ intel_check_copypixel_blit_fragment_ops(GLcontext * ctx) ctx->Color.BlendEnabled); } -#ifdef I915 -/* Doesn't work for overlapping regions. Could do a double copy or - * just fallback. - */ -static GLboolean -do_texture_copypixels(GLcontext * ctx, - GLint srcx, GLint srcy, - GLsizei width, GLsizei height, - GLint dstx, GLint dsty, GLenum type) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_region *dst = intel_drawbuf_region(intel); - struct intel_region *src = copypix_src_region(intel, type); - GLenum src_format; - GLenum src_type; - - DBG("%s %d,%d %dx%d --> %d,%d\n", __FUNCTION__, - srcx, srcy, width, height, dstx, dsty); - - if (!src || !dst || type != GL_COLOR) - return GL_FALSE; - - if (ctx->_ImageTransferState) { - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s: check_color failed\n", __FUNCTION__); - return GL_FALSE; - } - - /* Can't handle overlapping regions. Don't have sufficient control - * over rasterization to pull it off in-place. Punt on these for - * now. - * - * XXX: do a copy to a temporary. - */ - if (src->buffer == dst->buffer) { - drm_clip_rect_t srcbox; - drm_clip_rect_t dstbox; - drm_clip_rect_t tmp; - - srcbox.x1 = srcx; - srcbox.y1 = srcy; - srcbox.x2 = srcx + width; - srcbox.y2 = srcy + height; - - if (ctx->Pixel.ZoomX > 0) { - dstbox.x1 = dstx; - dstbox.x2 = dstx + width * ctx->Pixel.ZoomX; - } else { - dstbox.x1 = dstx + width * ctx->Pixel.ZoomX; - dstbox.x2 = dstx; - } - if (ctx->Pixel.ZoomY > 0) { - dstbox.y1 = dsty; - dstbox.y2 = dsty + height * ctx->Pixel.ZoomY; - } else { - dstbox.y1 = dsty + height * ctx->Pixel.ZoomY; - dstbox.y2 = dsty; - } - - DBG("src %d,%d %d,%d\n", srcbox.x1, srcbox.y1, srcbox.x2, srcbox.y2); - DBG("dst %d,%d %d,%d (%dx%d) (%f,%f)\n", dstbox.x1, dstbox.y1, dstbox.x2, dstbox.y2, - width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY); - - if (intel_intersect_cliprects(&tmp, &srcbox, &dstbox)) { - DBG("%s: regions overlap\n", __FUNCTION__); - return GL_FALSE; - } - } - - intelFlush(&intel->ctx); - - intel->vtbl.install_meta_state(intel); - - /* Is this true? Also will need to turn depth testing on according - * to state: - */ - intel->vtbl.meta_no_stencil_write(intel); - intel->vtbl.meta_no_depth_write(intel); - - /* Set the 3d engine to draw into the destination region: - */ - intel->vtbl.meta_draw_region(intel, dst, intel->depth_region); - - intel->vtbl.meta_import_pixel_state(intel); - - if (src->cpp == 2) { - src_format = GL_RGB; - src_type = GL_UNSIGNED_SHORT_5_6_5; - } - else { - src_format = GL_BGRA; - src_type = GL_UNSIGNED_BYTE; - } - - /* Set the frontbuffer up as a large rectangular texture. - */ - if (!intel->vtbl.meta_tex_rect_source(intel, src->buffer, 0, - src->pitch, - src->height, src_format, src_type)) { - intel->vtbl.leave_meta_state(intel); - return GL_FALSE; - } - - - intel->vtbl.meta_texture_blend_replace(intel); - - LOCK_HARDWARE(intel); - - if (intel->driDrawable->numClipRects) { - __DRIdrawablePrivate *dPriv = intel->driDrawable; - - - srcy = dPriv->h - srcy - height; /* convert from gl to hardware coords */ - - srcx += dPriv->x; - srcy += dPriv->y; - - /* Clip against the source region. This is the only source - * clipping we do. XXX: Just set the texcord wrap mode to clamp - * or similar. - * - */ - if (0) { - GLint orig_x = srcx; - GLint orig_y = srcy; - - if (!_mesa_clip_to_region(0, 0, src->pitch, src->height, - &srcx, &srcy, &width, &height)) - goto out; - - dstx += srcx - orig_x; - dsty += (srcy - orig_y) * ctx->Pixel.ZoomY; - } - - /* Just use the regular cliprect mechanism... Does this need to - * even hold the lock??? - */ - intel->vtbl.meta_draw_quad(intel, - dstx, - dstx + width * ctx->Pixel.ZoomX, - dPriv->h - (dsty + height * ctx->Pixel.ZoomY), - dPriv->h - (dsty), 0, /* XXX: what z value? */ - 0x00ff00ff, - srcx, srcx + width, srcy, srcy + height); - - out: - intel->vtbl.leave_meta_state(intel); - intel_batchbuffer_emit_mi_flush(intel->batch); - } - UNLOCK_HARDWARE(intel); - - DBG("%s: success\n", __FUNCTION__); - return GL_TRUE; -} -#endif /* I915 */ - /** * CopyPixels with the blitter. Don't support zooming, pixel transfer, etc. @@ -272,6 +111,12 @@ do_blit_copypixels(GLcontext * ctx, drm_clip_rect_t *cliprects; int x_off, y_off; + if (type == GL_DEPTH || type == GL_STENCIL) { + if (INTEL_DEBUG & DEBUG_FALLBACKS) + fprintf(stderr, "glCopyPixels() fallback: GL_DEPTH || GL_STENCIL\n"); + return GL_FALSE; + } + /* Update draw buffer bounds */ _mesa_update_state(ctx); @@ -362,14 +207,16 @@ do_blit_copypixels(GLcontext * ctx, &clip_x, &clip_y, &clip_w, &clip_h)) continue; - intelEmitCopyBlit(intel, dst->cpp, - src->pitch, src->buffer, 0, src->tiling, - dst->pitch, dst->buffer, 0, dst->tiling, - clip_x + delta_x, clip_y + delta_y, /* srcx, srcy */ - clip_x, clip_y, /* dstx, dsty */ - clip_w, clip_h, - ctx->Color.ColorLogicOpEnabled ? - ctx->Color.LogicOp : GL_COPY); + if (!intel_region_copy(intel, + dst, 0, clip_x, clip_y, + src, 0, clip_x + delta_x, clip_y + delta_y, + clip_w, clip_h, + ctx->Color.ColorLogicOpEnabled ? + ctx->Color.LogicOp : GL_COPY)) { + DBG("%s: blit failure\n", __FUNCTION__); + UNLOCK_HARDWARE(intel); + return GL_FALSE; + } } } out: @@ -392,12 +239,6 @@ intelCopyPixels(GLcontext * ctx, if (do_blit_copypixels(ctx, srcx, srcy, width, height, destx, desty, type)) return; -#ifdef I915 - if (do_texture_copypixels(ctx, srcx, srcy, width, height, destx, desty, type)) - return; -#endif - - DBG("fallback to _swrast_CopyPixels\n"); - - _swrast_CopyPixels(ctx, srcx, srcy, width, height, destx, desty, type); + /* this will use swrast if needed */ + _mesa_meta_copy_pixels(ctx, srcx, srcy, width, height, destx, desty, type); } diff --git a/src/mesa/drivers/dri/intel/intel_pixel_draw.c b/src/mesa/drivers/dri/intel/intel_pixel_draw.c index d80069dd58..7fbb89fd6a 100644 --- a/src/mesa/drivers/dri/intel/intel_pixel_draw.c +++ b/src/mesa/drivers/dri/intel/intel_pixel_draw.c @@ -29,8 +29,6 @@ #include "main/enums.h" #include "main/image.h" #include "main/mtypes.h" -#include "main/macros.h" -#include "main/bufferobj.h" #include "main/teximage.h" #include "main/texenv.h" #include "main/texobj.h" @@ -41,169 +39,22 @@ #include "main/enable.h" #include "main/buffers.h" #include "main/fbobject.h" -#include "main/renderbuffer.h" #include "main/depth.h" #include "main/hash.h" #include "main/blend.h" -#include "glapi/dispatch.h" #include "swrast/swrast.h" +#include "drivers/common/meta.h" -#include "intel_screen.h" #include "intel_context.h" #include "intel_batchbuffer.h" #include "intel_blit.h" #include "intel_buffers.h" #include "intel_regions.h" #include "intel_pixel.h" -#include "intel_buffer_objects.h" #include "intel_fbo.h" -static GLboolean -intel_texture_drawpixels(GLcontext * ctx, - GLint x, GLint y, - GLsizei width, GLsizei height, - GLenum format, - GLenum type, - const struct gl_pixelstore_attrib *unpack, - const GLvoid *pixels) -{ - struct intel_context *intel = intel_context(ctx); - GLuint texname; - GLfloat vertices[4][4]; - GLfloat texcoords[4][2]; - GLfloat z; - GLint old_active_texture; - GLenum internalFormat; - - /* We're going to mess with texturing with no regard to existing texture - * state, so if there is some set up we have to bail. - */ - if (ctx->Texture._EnabledUnits != 0) { - if (INTEL_DEBUG & DEBUG_FALLBACKS) - fprintf(stderr, "glDrawPixels() fallback: texturing enabled\n"); - return GL_FALSE; - } - - /* Can't do textured DrawPixels with a fragment program, unless we were - * to generate a new program that sampled our texture and put the results - * in the fragment color before the user's program started. - */ - if (ctx->FragmentProgram.Enabled) { - if (INTEL_DEBUG & DEBUG_FALLBACKS) - fprintf(stderr, "glDrawPixels() fallback: fragment program enabled\n"); - return GL_FALSE; - } - - /* We don't have a way to generate fragments with stencil values which - * will set the resulting stencil value. - */ - if (format == GL_STENCIL_INDEX || format == GL_DEPTH_STENCIL) - return GL_FALSE; - - /* Check that we can load in a texture this big. */ - if (width > (1 << (ctx->Const.MaxTextureLevels - 1)) || - height > (1 << (ctx->Const.MaxTextureLevels - 1))) { - if (INTEL_DEBUG & DEBUG_FALLBACKS) - fprintf(stderr, "glDrawPixels() fallback: bitmap too large (%dx%d)\n", - width, height); - return GL_FALSE; - } - - /* To do DEPTH_COMPONENT, we would need to change our setup to not draw to - * the color buffer, and sample the texture values into the fragment depth - * in a program. - */ - if (format == GL_DEPTH_COMPONENT) { - if (INTEL_DEBUG & DEBUG_FALLBACKS) - fprintf(stderr, - "glDrawPixels() fallback: format == GL_DEPTH_COMPONENT\n"); - return GL_FALSE; - } - - if (!ctx->Extensions.ARB_texture_non_power_of_two && - (!is_power_of_two(width) || !is_power_of_two(height))) { - if (INTEL_DEBUG & DEBUG_FALLBACKS) - fprintf(stderr, - "glDrawPixels() fallback: NPOT texture\n"); - return GL_FALSE; - } - - _mesa_PushAttrib(GL_ENABLE_BIT | GL_TEXTURE_BIT | - GL_CURRENT_BIT); - _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT); - - /* XXX: pixel store stuff */ - _mesa_Disable(GL_POLYGON_STIPPLE); - - old_active_texture = ctx->Texture.CurrentUnit; - _mesa_ActiveTextureARB(GL_TEXTURE0_ARB); - _mesa_Enable(GL_TEXTURE_2D); - _mesa_GenTextures(1, &texname); - _mesa_BindTexture(GL_TEXTURE_2D, texname); - _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST); - _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST); - _mesa_TexEnvf(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, GL_REPLACE); - if (type == GL_ALPHA) - internalFormat = GL_ALPHA; - else - internalFormat = GL_RGBA; - _mesa_TexImage2D(GL_TEXTURE_2D, 0, internalFormat, width, height, 0, format, - type, pixels); - - intel_meta_set_passthrough_transform(intel); - - /* convert rasterpos Z from [0,1] to NDC coord in [-1,1] */ - z = -1.0 + 2.0 * ctx->Current.RasterPos[2]; - - /* Create the vertex buffer based on the current raster pos. The x and y - * we're handed are ctx->Current.RasterPos[0,1] rounded to integers. - * We also apply the depth. However, the W component is already multiplied - * into ctx->Current.RasterPos[0,1,2] and we can ignore it at this point. - */ - vertices[0][0] = x; - vertices[0][1] = y; - vertices[0][2] = z; - vertices[0][3] = 1.0; - vertices[1][0] = x + width * ctx->Pixel.ZoomX; - vertices[1][1] = y; - vertices[1][2] = z; - vertices[1][3] = 1.0; - vertices[2][0] = x + width * ctx->Pixel.ZoomX; - vertices[2][1] = y + height * ctx->Pixel.ZoomY; - vertices[2][2] = z; - vertices[2][3] = 1.0; - vertices[3][0] = x; - vertices[3][1] = y + height * ctx->Pixel.ZoomY; - vertices[3][2] = z; - vertices[3][3] = 1.0; - - texcoords[0][0] = 0.0; - texcoords[0][1] = 0.0; - texcoords[1][0] = 1.0; - texcoords[1][1] = 0.0; - texcoords[2][0] = 1.0; - texcoords[2][1] = 1.0; - texcoords[3][0] = 0.0; - texcoords[3][1] = 1.0; - - _mesa_VertexPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), &vertices); - _mesa_ClientActiveTextureARB(GL_TEXTURE0); - _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &texcoords); - _mesa_Enable(GL_VERTEX_ARRAY); - _mesa_Enable(GL_TEXTURE_COORD_ARRAY); - _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4); - - intel_meta_restore_transform(intel); - - _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture); - _mesa_PopClientAttrib(); - _mesa_PopAttrib(); - - _mesa_DeleteTextures(1, &texname); - - return GL_TRUE; -} +/** XXX compare perf of this vs. _mesa_meta_draw_pixels(STENCIL) */ static GLboolean intel_stencil_drawpixels(GLcontext * ctx, GLint x, GLint y, @@ -216,14 +67,14 @@ intel_stencil_drawpixels(GLcontext * ctx, struct intel_context *intel = intel_context(ctx); GLuint texname, rb_name, fb_name, old_fb_name; GLfloat vertices[4][2]; - GLfloat texcoords[4][2]; struct intel_renderbuffer *irb; struct intel_renderbuffer *depth_irb; struct gl_renderbuffer *rb; struct gl_pixelstore_attrib old_unpack; GLstencil *stencil_pixels; - int row; + int row, y1, y2; GLint old_active_texture; + GLboolean rendering_to_fbo = ctx->DrawBuffer->Name != 0; if (format != GL_STENCIL_INDEX) return GL_FALSE; @@ -358,34 +209,35 @@ intel_stencil_drawpixels(GLcontext * ctx, ctx->Unpack = old_unpack; _mesa_free(stencil_pixels); - intel_meta_set_passthrough_transform(intel); + meta_set_passthrough_transform(&intel->meta); + /* Since we're rendering to the framebuffer as if it was an FBO, + * if it's the window system we have to flip the coordinates. + */ + if (rendering_to_fbo) { + y1 = y; + y2 = y + height * ctx->Pixel.ZoomY; + } else { + y1 = irb->Base.Height - (y + height * ctx->Pixel.ZoomY); + y2 = irb->Base.Height - y; + } vertices[0][0] = x; - vertices[0][1] = y; + vertices[0][1] = y1; vertices[1][0] = x + width * ctx->Pixel.ZoomX; - vertices[1][1] = y; + vertices[1][1] = y1; vertices[2][0] = x + width * ctx->Pixel.ZoomX; - vertices[2][1] = y + height * ctx->Pixel.ZoomY; + vertices[2][1] = y2; vertices[3][0] = x; - vertices[3][1] = y + height * ctx->Pixel.ZoomY; - - texcoords[0][0] = 0.0; - texcoords[0][1] = 0.0; - texcoords[1][0] = 1.0; - texcoords[1][1] = 0.0; - texcoords[2][0] = 1.0; - texcoords[2][1] = 1.0; - texcoords[3][0] = 0.0; - texcoords[3][1] = 1.0; + vertices[3][1] = y2; _mesa_VertexPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &vertices); - _mesa_ClientActiveTextureARB(GL_TEXTURE0); - _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &texcoords); _mesa_Enable(GL_VERTEX_ARRAY); - _mesa_Enable(GL_TEXTURE_COORD_ARRAY); + meta_set_default_texrect(&intel->meta); + _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4); - intel_meta_restore_transform(intel); + meta_restore_texcoords(&intel->meta); + meta_restore_transform(&intel->meta); _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture); _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, old_fb_name); @@ -409,17 +261,25 @@ intelDrawPixels(GLcontext * ctx, const struct gl_pixelstore_attrib *unpack, const GLvoid * pixels) { - if (intel_texture_drawpixels(ctx, x, y, width, height, format, type, - unpack, pixels)) - return; - +#if 0 + /* XXX this function doesn't seem to work reliably even when all + * the pre-requisite conditions are met. + * Note that this function is never hit with conform. + * Fall back to swrast because even the _mesa_meta_draw_pixels() approach + * isn't working because of an apparent stencil bug. + */ if (intel_stencil_drawpixels(ctx, x, y, width, height, format, type, unpack, pixels)) return; +#else + (void) intel_stencil_drawpixels; /* silence warning */ + if (format == GL_STENCIL_INDEX) { + _swrast_DrawPixels(ctx, x, y, width, height, format, type, + unpack, pixels); + return; + } +#endif - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s: fallback to swrast\n", __FUNCTION__); - - _swrast_DrawPixels(ctx, x, y, width, height, format, type, - unpack, pixels); + _mesa_meta_draw_pixels(ctx, x, y, width, height, format, type, + unpack, pixels); } diff --git a/src/mesa/drivers/dri/intel/intel_pixel_read.c b/src/mesa/drivers/dri/intel/intel_pixel_read.c index 0370255614..8713463ace 100644 --- a/src/mesa/drivers/dri/intel/intel_pixel_read.c +++ b/src/mesa/drivers/dri/intel/intel_pixel_read.c @@ -180,7 +180,7 @@ do_blit_readpixels(GLcontext * ctx, if (!src) return GL_FALSE; - if (dst) { + if (pack->BufferObj->Name) { /* XXX This validation should be done by core mesa: */ if (!_mesa_validate_pbo_access(2, pack, width, height, 1, @@ -261,16 +261,19 @@ do_blit_readpixels(GLcontext * ctx, if (!intel_intersect_cliprects(&rect, &src_rect, &box[i])) continue; - intelEmitCopyBlit(intel, - src->cpp, - src->pitch, src->buffer, 0, src->tiling, - rowLength, dst_buffer, dst_offset, GL_FALSE, - rect.x1, - rect.y1, - rect.x1 - src_rect.x1, - rect.y2 - src_rect.y2, - rect.x2 - rect.x1, rect.y2 - rect.y1, - GL_COPY); + if (!intelEmitCopyBlit(intel, + src->cpp, + src->pitch, src->buffer, 0, src->tiling, + rowLength, dst_buffer, dst_offset, GL_FALSE, + rect.x1, + rect.y1, + rect.x1 - src_rect.x1, + rect.y2 - src_rect.y2, + rect.x2 - rect.x1, rect.y2 - rect.y1, + GL_COPY)) { + UNLOCK_HARDWARE(intel); + return GL_FALSE; + } } } UNLOCK_HARDWARE(intel); diff --git a/src/mesa/drivers/dri/intel/intel_reg.h b/src/mesa/drivers/dri/intel/intel_reg.h index 57ac8f0cc1..d19f1bae34 100644 --- a/src/mesa/drivers/dri/intel/intel_reg.h +++ b/src/mesa/drivers/dri/intel/intel_reg.h @@ -189,6 +189,19 @@ #define S7_DEPTH_OFFSET_CONST_MASK ~0 +/* p143 */ +#define _3DSTATE_BUF_INFO_CMD (CMD_3D | (0x1d<<24) | (0x8e<<16) | 1) +/* Dword 1 */ +#define BUF_3D_ID_COLOR_BACK (0x3<<24) +#define BUF_3D_ID_DEPTH (0x7<<24) +#define BUF_3D_USE_FENCE (1<<23) +#define BUF_3D_TILED_SURFACE (1<<22) +#define BUF_3D_TILE_WALK_X 0 +#define BUF_3D_TILE_WALK_Y (1<<21) +#define BUF_3D_PITCH(x) (((x)/4)<<2) +/* Dword 2 */ +#define BUF_3D_ADDR(x) ((x) & ~0x3) + /* Primitive dispatch on 830-945 */ #define _3DPRIMITIVE (CMD_3D | (0x1f << 24)) #define PRIM_INDIRECT (1<<23) diff --git a/src/mesa/drivers/dri/intel/intel_regions.c b/src/mesa/drivers/dri/intel/intel_regions.c index b8d2dec4cb..a86c66a844 100644 --- a/src/mesa/drivers/dri/intel/intel_regions.c +++ b/src/mesa/drivers/dri/intel/intel_regions.c @@ -52,17 +52,77 @@ #define FILE_DEBUG_FLAG DEBUG_REGION +/* This should be set to the maximum backtrace size desired. + * Set it to 0 to disable backtrace debugging. + */ +#define DEBUG_BACKTRACE_SIZE 0 + +#if DEBUG_BACKTRACE_SIZE == 0 +/* Use the standard debug output */ +#define _DBG(...) DBG(__VA_ARGS__) +#else +/* Use backtracing debug output */ +#define _DBG(...) {debug_backtrace(); DBG(__VA_ARGS__);} + +/* Backtracing debug support */ +#include <execinfo.h> + +static void +debug_backtrace(void) +{ + void *trace[DEBUG_BACKTRACE_SIZE]; + char **strings = NULL; + int traceSize; + register int i; + + traceSize = backtrace(trace, DEBUG_BACKTRACE_SIZE); + strings = backtrace_symbols(trace, traceSize); + if (strings == NULL) { + DBG("no backtrace:"); + return; + } + + /* Spit out all the strings with a colon separator. Ignore + * the first, since we don't really care about the call + * to debug_backtrace() itself. Skip until the final "/" in + * the trace to avoid really long lines. + */ + for (i = 1; i < traceSize; i++) { + char *p = strings[i], *slash = strings[i]; + while (*p) { + if (*p++ == '/') { + slash = p; + } + } + + DBG("%s:", slash); + } + + /* Free up the memory, and we're done */ + free(strings); +} + +#endif + + + /* XXX: Thread safety? */ GLubyte * intel_region_map(struct intel_context *intel, struct intel_region *region) { - DBG("%s\n", __FUNCTION__); + intelFlush(&intel->ctx); + + _DBG("%s %p\n", __FUNCTION__, region); if (!region->map_refcount++) { if (region->pbo) intel_region_cow(intel, region); - dri_bo_map(region->buffer, GL_TRUE); + if (region->tiling != I915_TILING_NONE && + intel->intelScreen->kernel_exec_fencing) + drm_intel_gem_bo_map_gtt(region->buffer); + else + dri_bo_map(region->buffer, GL_TRUE); region->map = region->buffer->virtual; } @@ -72,9 +132,13 @@ intel_region_map(struct intel_context *intel, struct intel_region *region) void intel_region_unmap(struct intel_context *intel, struct intel_region *region) { - DBG("%s\n", __FUNCTION__); + _DBG("%s %p\n", __FUNCTION__, region); if (!--region->map_refcount) { - dri_bo_unmap(region->buffer); + if (region->tiling != I915_TILING_NONE && + intel->intelScreen->kernel_exec_fencing) + drm_intel_gem_bo_unmap_gtt(region->buffer); + else + dri_bo_unmap(region->buffer); region->map = NULL; } } @@ -87,10 +151,10 @@ intel_region_alloc_internal(struct intel_context *intel, { struct intel_region *region; - DBG("%s\n", __FUNCTION__); - - if (buffer == NULL) + if (buffer == NULL) { + _DBG("%s <-- NULL\n", __FUNCTION__); return NULL; + } region = calloc(sizeof(*region), 1); region->cpp = cpp; @@ -104,15 +168,33 @@ intel_region_alloc_internal(struct intel_context *intel, region->tiling = I915_TILING_NONE; region->bit_6_swizzle = I915_BIT_6_SWIZZLE_NONE; + _DBG("%s <-- %p\n", __FUNCTION__, region); return region; } struct intel_region * intel_region_alloc(struct intel_context *intel, + uint32_t tiling, GLuint cpp, GLuint width, GLuint height, GLuint pitch, GLboolean expect_accelerated_upload) { dri_bo *buffer; + struct intel_region *region; + + /* If we're tiled, our allocations are in 8 or 32-row blocks, so + * failure to align our height means that we won't allocate enough pages. + * + * If we're untiled, we still have to align to 2 rows high because the + * data port accesses 2x2 blocks even if the bottom row isn't to be + * rendered, so failure to align means we could walk off the end of the + * GTT and fault. + */ + if (tiling == I915_TILING_X) + height = ALIGN(height, 8); + else if (tiling == I915_TILING_Y) + height = ALIGN(height, 32); + else + height = ALIGN(height, 2); /* If we're untiled, we have to align to 2 rows high because the * data port accesses 2x2 blocks even if the bottom row isn't to be @@ -129,7 +211,16 @@ intel_region_alloc(struct intel_context *intel, pitch * cpp * height, 64); } - return intel_region_alloc_internal(intel, cpp, width, height, pitch, buffer); + region = intel_region_alloc_internal(intel, cpp, width, height, + pitch, buffer); + + if (tiling != I915_TILING_NONE) { + assert(((pitch * cpp) & 127) == 0); + drm_intel_bo_set_tiling(buffer, &tiling, pitch * cpp); + drm_intel_bo_get_tiling(buffer, ®ion->tiling, ®ion->bit_6_swizzle); + } + + return region; } struct intel_region * @@ -165,7 +256,7 @@ void intel_region_reference(struct intel_region **dst, struct intel_region *src) { if (src) - DBG("%s %p %d\n", __FUNCTION__, src, src->refcount); + _DBG("%s %p %d\n", __FUNCTION__, src, src->refcount); assert(*dst == NULL); if (src) { @@ -179,10 +270,12 @@ intel_region_release(struct intel_region **region_handle) { struct intel_region *region = *region_handle; - if (region == NULL) + if (region == NULL) { + _DBG("%s NULL\n", __FUNCTION__); return; + } - DBG("%s %p %d\n", __FUNCTION__, region, region->refcount - 1); + _DBG("%s %p %d\n", __FUNCTION__, region, region->refcount - 1); ASSERT(region->refcount > 0); region->refcount--; @@ -256,9 +349,7 @@ intel_region_data(struct intel_context *intel, const void *src, GLuint src_pitch, GLuint srcx, GLuint srcy, GLuint width, GLuint height) { - GLboolean locked = GL_FALSE; - - DBG("%s\n", __FUNCTION__); + _DBG("%s\n", __FUNCTION__); if (intel == NULL) return; @@ -271,39 +362,33 @@ intel_region_data(struct intel_context *intel, intel_region_cow(intel, dst); } - if (!intel->locked) { - LOCK_HARDWARE(intel); - locked = GL_TRUE; - } - + LOCK_HARDWARE(intel); _mesa_copy_rect(intel_region_map(intel, dst) + dst_offset, dst->cpp, dst->pitch, dstx, dsty, width, height, src, src_pitch, srcx, srcy); intel_region_unmap(intel, dst); - - if (locked) - UNLOCK_HARDWARE(intel); - + UNLOCK_HARDWARE(intel); } /* Copy rectangular sub-regions. Need better logic about when to * push buffers into AGP - will currently do so whenever possible. */ -void +GLboolean intel_region_copy(struct intel_context *intel, struct intel_region *dst, GLuint dst_offset, GLuint dstx, GLuint dsty, struct intel_region *src, GLuint src_offset, - GLuint srcx, GLuint srcy, GLuint width, GLuint height) + GLuint srcx, GLuint srcy, GLuint width, GLuint height, + GLenum logicop) { - DBG("%s\n", __FUNCTION__); + _DBG("%s\n", __FUNCTION__); if (intel == NULL) - return; + return GL_FALSE; if (dst->pbo) { if (dstx == 0 && @@ -315,41 +400,12 @@ intel_region_copy(struct intel_context *intel, assert(src->cpp == dst->cpp); - intelEmitCopyBlit(intel, - dst->cpp, - src->pitch, src->buffer, src_offset, src->tiling, - dst->pitch, dst->buffer, dst_offset, dst->tiling, - srcx, srcy, dstx, dsty, width, height, - GL_COPY); -} - -/* Fill a rectangular sub-region. Need better logic about when to - * push buffers into AGP - will currently do so whenever possible. - */ -void -intel_region_fill(struct intel_context *intel, - struct intel_region *dst, - GLuint dst_offset, - GLuint dstx, GLuint dsty, - GLuint width, GLuint height, GLuint color) -{ - DBG("%s\n", __FUNCTION__); - - if (intel == NULL) - return; - - if (dst->pbo) { - if (dstx == 0 && - dsty == 0 && width == dst->pitch && height == dst->height) - intel_region_release_pbo(intel, dst); - else - intel_region_cow(intel, dst); - } - - intelEmitFillBlit(intel, - dst->cpp, - dst->pitch, dst->buffer, dst_offset, dst->tiling, - dstx, dsty, width, height, color); + return intelEmitCopyBlit(intel, + dst->cpp, + src->pitch, src->buffer, src_offset, src->tiling, + dst->pitch, dst->buffer, dst_offset, dst->tiling, + srcx, srcy, dstx, dsty, width, height, + logicop); } /* Attach to a pbo, discarding our data. Effectively zero-copy upload @@ -360,9 +416,13 @@ intel_region_attach_pbo(struct intel_context *intel, struct intel_region *region, struct intel_buffer_object *pbo) { + dri_bo *buffer; + if (region->pbo == pbo) return; + _DBG("%s %p %p\n", __FUNCTION__, region, pbo); + /* If there is already a pbo attached, break the cow tie now. * Don't call intel_region_release_pbo() as that would * unnecessarily allocate a new buffer we would have to immediately @@ -378,10 +438,13 @@ intel_region_attach_pbo(struct intel_context *intel, region->buffer = NULL; } + /* make sure pbo has a buffer of its own */ + buffer = intel_bufferobj_buffer(intel, pbo, INTEL_WRITE_FULL); + region->pbo = pbo; region->pbo->region = region; - dri_bo_reference(pbo->buffer); - region->buffer = pbo->buffer; + dri_bo_reference(buffer); + region->buffer = buffer; } @@ -392,6 +455,7 @@ void intel_region_release_pbo(struct intel_context *intel, struct intel_region *region) { + _DBG("%s %p\n", __FUNCTION__, region); assert(region->buffer == region->pbo->buffer); region->pbo->region = NULL; region->pbo = NULL; @@ -410,34 +474,27 @@ void intel_region_cow(struct intel_context *intel, struct intel_region *region) { struct intel_buffer_object *pbo = region->pbo; - GLboolean was_locked = intel->locked; - - if (intel == NULL) - return; + GLboolean ok; intel_region_release_pbo(intel, region); assert(region->cpp * region->pitch * region->height == pbo->Base.Size); - DBG("%s (%d bytes)\n", __FUNCTION__, pbo->Base.Size); + _DBG("%s %p (%d bytes)\n", __FUNCTION__, region, pbo->Base.Size); /* Now blit from the texture buffer to the new buffer: */ - was_locked = intel->locked; - if (!was_locked) - LOCK_HARDWARE(intel); - - intelEmitCopyBlit(intel, - region->cpp, - region->pitch, region->buffer, 0, region->tiling, - region->pitch, pbo->buffer, 0, region->tiling, - 0, 0, 0, 0, - region->pitch, region->height, - GL_COPY); - - if (!was_locked) - UNLOCK_HARDWARE(intel); + LOCK_HARDWARE(intel); + ok = intelEmitCopyBlit(intel, + region->cpp, + region->pitch, pbo->buffer, 0, region->tiling, + region->pitch, region->buffer, 0, region->tiling, + 0, 0, 0, 0, + region->pitch, region->height, + GL_COPY); + assert(ok); + UNLOCK_HARDWARE(intel); } dri_bo * @@ -466,6 +523,10 @@ intel_recreate_static(struct intel_context *intel, if (region == NULL) { region = calloc(sizeof(*region), 1); region->refcount = 1; + _DBG("%s creating new region %p\n", __FUNCTION__, region); + } + else { + _DBG("%s %p\n", __FUNCTION__, region); } if (intel->ctx.Visual.rgbBits == 24) diff --git a/src/mesa/drivers/dri/intel/intel_regions.h b/src/mesa/drivers/dri/intel/intel_regions.h index 45e2bf4e77..0d379bdc6e 100644 --- a/src/mesa/drivers/dri/intel/intel_regions.h +++ b/src/mesa/drivers/dri/intel/intel_regions.h @@ -73,7 +73,8 @@ struct intel_region * copied by calling intel_reference_region(). */ struct intel_region *intel_region_alloc(struct intel_context *intel, - GLuint cpp, GLuint width, + uint32_t tiling, + GLuint cpp, GLuint width, GLuint height, GLuint pitch, GLboolean expect_accelerated_upload); @@ -109,21 +110,15 @@ void intel_region_data(struct intel_context *intel, /* Copy rectangular sub-regions */ -void intel_region_copy(struct intel_context *intel, - struct intel_region *dest, - GLuint dest_offset, - GLuint destx, GLuint desty, - struct intel_region *src, - GLuint src_offset, - GLuint srcx, GLuint srcy, GLuint width, GLuint height); - -/* Fill a rectangular sub-region - */ -void intel_region_fill(struct intel_context *intel, - struct intel_region *dest, - GLuint dest_offset, - GLuint destx, GLuint desty, - GLuint width, GLuint height, GLuint color); +GLboolean +intel_region_copy(struct intel_context *intel, + struct intel_region *dest, + GLuint dest_offset, + GLuint destx, GLuint desty, + struct intel_region *src, + GLuint src_offset, + GLuint srcx, GLuint srcy, GLuint width, GLuint height, + GLenum logicop); /* Helpers for zerocopy uploads, particularly texture image uploads: */ diff --git a/src/mesa/drivers/dri/intel/intel_screen.c b/src/mesa/drivers/dri/intel/intel_screen.c index f810850ef5..1b8c56e68d 100644 --- a/src/mesa/drivers/dri/intel/intel_screen.c +++ b/src/mesa/drivers/dri/intel/intel_screen.c @@ -49,6 +49,10 @@ #include "i915_drm.h" #include "i830_dri.h" +#define DRI_CONF_TEXTURE_TILING(def) \ + DRI_CONF_OPT_BEGIN(texture_tiling, bool, def) \ + DRI_CONF_DESC(en, "Enable texture tiling") \ + DRI_CONF_OPT_END \ PUBLIC const char __driConfigOptions[] = DRI_CONF_BEGIN @@ -64,6 +68,17 @@ PUBLIC const char __driConfigOptions[] = DRI_CONF_ENUM(1, "Enable reuse of all sizes of buffer objects") DRI_CONF_DESC_END DRI_CONF_OPT_END + +#ifdef I915 + DRI_CONF_TEXTURE_TILING(false) +#else + DRI_CONF_TEXTURE_TILING(true) +#endif + + DRI_CONF_OPT_BEGIN(early_z, bool, false) + DRI_CONF_DESC(en, "Enable early Z in classic mode (unstable, 945-only).") + DRI_CONF_OPT_END + DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY DRI_CONF_FORCE_S3TC_ENABLE(false) @@ -76,7 +91,7 @@ PUBLIC const char __driConfigOptions[] = DRI_CONF_SECTION_END DRI_CONF_END; -const GLuint __driNConfigOptions = 8; +const GLuint __driNConfigOptions = 10; #ifdef USE_NEW_INTERFACE static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; @@ -587,7 +602,7 @@ intel_init_bufmgr(intelScreenPrivate *intelScreen) GLboolean gem_supported; struct drm_i915_getparam gp; __DRIscreenPrivate *spriv = intelScreen->driScrnPriv; - int num_fences; + int num_fences = 0; intelScreen->no_hw = getenv("INTEL_NO_HW") != NULL; diff --git a/src/mesa/drivers/dri/intel/intel_span.c b/src/mesa/drivers/dri/intel/intel_span.c index 34b78ebc1a..8df4990880 100644 --- a/src/mesa/drivers/dri/intel/intel_span.c +++ b/src/mesa/drivers/dri/intel/intel_span.c @@ -501,7 +501,7 @@ intel_map_unmap_buffers(struct intel_context *intel, GLboolean map) /** - * Prepare for softare rendering. Map current read/draw framebuffers' + * Prepare for software rendering. Map current read/draw framebuffers' * renderbuffes and all currently bound texture objects. * * Old note: Moved locking out to get reasonable span performance. @@ -526,7 +526,7 @@ intelSpanRenderStart(GLcontext * ctx) } /** - * Called when done softare rendering. Unmap the buffers we mapped in + * Called when done software rendering. Unmap the buffers we mapped in * the above function. */ void diff --git a/src/mesa/drivers/dri/intel/intel_syncobj.c b/src/mesa/drivers/dri/intel/intel_syncobj.c new file mode 100644 index 0000000000..1286fe929b --- /dev/null +++ b/src/mesa/drivers/dri/intel/intel_syncobj.c @@ -0,0 +1,132 @@ +/* + * Copyright © 2008 Intel Corporation + * + * 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 (including the next + * paragraph) 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 + * THE AUTHORS OR COPYRIGHT HOLDERS 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. + * + * Authors: + * Eric Anholt <eric@anholt.net> + * + */ + +/** @file intel_syncobj.c + * + * Support for ARB_sync + * + * ARB_sync is implemented by flushing the current batchbuffer and keeping a + * reference on it. We can then check for completion or wait for compeltion + * using the normal buffer object mechanisms. This does mean that if an + * application is using many sync objects, it will emit small batchbuffers + * which may end up being a significant overhead. In other tests of removing + * gratuitous batchbuffer syncs in Mesa, it hasn't appeared to be a significant + * performance bottleneck, though. + */ + +#include "main/simple_list.h" +#include "main/imports.h" + +#include "intel_context.h" +#include "intel_batchbuffer.h" +#include "intel_reg.h" + +static struct gl_sync_object * +intel_new_sync_object(GLcontext *ctx, GLuint id) +{ + struct intel_sync_object *sync; + + sync = _mesa_calloc(sizeof(struct intel_sync_object)); + + return &sync->Base; +} + +static void +intel_delete_sync_object(GLcontext *ctx, struct gl_sync_object *s) +{ + struct intel_sync_object *sync = (struct intel_sync_object *)s; + + drm_intel_bo_unreference(sync->bo); + _mesa_free(sync); +} + +static void +intel_fence_sync(GLcontext *ctx, struct gl_sync_object *s, + GLenum condition, GLbitfield flags) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_sync_object *sync = (struct intel_sync_object *)s; + + assert(condition == GL_SYNC_GPU_COMMANDS_COMPLETE); + intel_batchbuffer_emit_mi_flush(intel->batch); + + sync->bo = intel->batch->buf; + drm_intel_bo_reference(sync->bo); + + intelFlush(ctx); +} + +/* We ignore the user-supplied timeout. This is weaselly -- we're allowed to + * round to an implementation-dependent accuracy, and right now our + * implementation "rounds" to the wait-forever value. + * + * The fix would be a new kernel function to do the GTT transition with a + * timeout. + */ +static void intel_client_wait_sync(GLcontext *ctx, struct gl_sync_object *s, + GLbitfield flags, GLuint64 timeout) +{ + struct intel_sync_object *sync = (struct intel_sync_object *)s; + + if (sync->bo) { + drm_intel_bo_wait_rendering(sync->bo); + s->StatusFlag = 1; + drm_intel_bo_unreference(sync->bo); + sync->bo = NULL; + } +} + +/* We have nothing to do for WaitSync. Our GL command stream is sequential, + * so given that the sync object has already flushed the batchbuffer, + * any batchbuffers coming after this waitsync will naturally not occur until + * the previous one is done. + */ +static void intel_server_wait_sync(GLcontext *ctx, struct gl_sync_object *s, + GLbitfield flags, GLuint64 timeout) +{ +} + +static void intel_check_sync(GLcontext *ctx, struct gl_sync_object *s) +{ + struct intel_sync_object *sync = (struct intel_sync_object *)s; + + if (sync->bo && drm_intel_bo_busy(sync->bo)) { + drm_intel_bo_unreference(sync->bo); + sync->bo = NULL; + s->StatusFlag = 1; + } +} + +void intel_init_syncobj_functions(struct dd_function_table *functions) +{ + functions->NewSyncObject = intel_new_sync_object; + functions->DeleteSyncObject = intel_delete_sync_object; + functions->FenceSync = intel_fence_sync; + functions->CheckSync = intel_check_sync; + functions->ClientWaitSync = intel_client_wait_sync; + functions->ServerWaitSync = intel_server_wait_sync; +} diff --git a/src/mesa/drivers/dri/intel/intel_tex.c b/src/mesa/drivers/dri/intel/intel_tex.c index ae0994b183..df63f29a42 100644 --- a/src/mesa/drivers/dri/intel/intel_tex.c +++ b/src/mesa/drivers/dri/intel/intel_tex.c @@ -158,81 +158,11 @@ timed_memcpy(void *dest, const void *src, size_t n) } #endif /* DO_DEBUG */ -/** - * Generate new mipmap data from BASE+1 to BASE+p (the minimally-sized mipmap - * level). - * - * The texture object's miptree must be mapped. - * - * It would be really nice if this was just called by Mesa whenever mipmaps - * needed to be regenerated, rather than us having to remember to do so in - * each texture image modification path. - * - * This function should also include an accelerated path. - */ -void -intel_generate_mipmap(GLcontext *ctx, GLenum target, - struct gl_texture_object *texObj) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_texture_object *intelObj = intel_texture_object(texObj); - GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; - int face, i; - - _mesa_generate_mipmap(ctx, target, texObj); - - /* Update the level information in our private data in the new images, since - * it didn't get set as part of a normal TexImage path. - */ - for (face = 0; face < nr_faces; face++) { - for (i = texObj->BaseLevel + 1; i < texObj->MaxLevel; i++) { - struct intel_texture_image *intelImage; - - intelImage = intel_texture_image(texObj->Image[face][i]); - if (intelImage == NULL) - break; - - intelImage->level = i; - intelImage->face = face; - /* Unreference the miptree to signal that the new Data is a bare - * pointer from mesa. - */ - intel_miptree_release(intel, &intelImage->mt); - } - } -} - -static void intelGenerateMipmap(GLcontext *ctx, GLenum target, struct gl_texture_object *texObj) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_texture_object *intelObj = intel_texture_object(texObj); - - intel_tex_map_level_images(intel, intelObj, texObj->BaseLevel); - intel_generate_mipmap(ctx, target, texObj); - intel_tex_unmap_level_images(intel, intelObj, texObj->BaseLevel); -} - void intelInitTextureFuncs(struct dd_function_table *functions) { functions->ChooseTextureFormat = intelChooseTextureFormat; - functions->TexImage1D = intelTexImage1D; - functions->TexImage2D = intelTexImage2D; - functions->TexImage3D = intelTexImage3D; - functions->TexSubImage1D = intelTexSubImage1D; - functions->TexSubImage2D = intelTexSubImage2D; - functions->TexSubImage3D = intelTexSubImage3D; - functions->CopyTexImage1D = intelCopyTexImage1D; - functions->CopyTexImage2D = intelCopyTexImage2D; - functions->CopyTexSubImage1D = intelCopyTexSubImage1D; - functions->CopyTexSubImage2D = intelCopyTexSubImage2D; - functions->GetTexImage = intelGetTexImage; - functions->GenerateMipmap = intelGenerateMipmap; - - /* compressed texture functions */ - functions->CompressedTexImage2D = intelCompressedTexImage2D; - functions->CompressedTexSubImage2D = intelCompressedTexSubImage2D; - functions->GetCompressedTexImage = intelGetCompressedTexImage; + functions->GenerateMipmap = intel_generate_mipmap; functions->NewTextureObject = intelNewTextureObject; functions->NewTextureImage = intelNewTextureImage; diff --git a/src/mesa/drivers/dri/intel/intel_tex.h b/src/mesa/drivers/dri/intel/intel_tex.h index f5372d82fb..471aa2a240 100644 --- a/src/mesa/drivers/dri/intel/intel_tex.h +++ b/src/mesa/drivers/dri/intel/intel_tex.h @@ -35,116 +35,17 @@ void intelInitTextureFuncs(struct dd_function_table *functions); +void intelInitTextureImageFuncs(struct dd_function_table *functions); + +void intelInitTextureSubImageFuncs(struct dd_function_table *functions); + +void intelInitTextureCopyImageFuncs(struct dd_function_table *functions); + const struct gl_texture_format *intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, GLenum format, GLenum type); - -void intelTexImage3D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint depth, - GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelTexSubImage3D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, GLint yoffset, GLint zoffset, - GLsizei width, GLsizei height, GLsizei depth, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelTexImage2D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelTexSubImage2D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, GLint yoffset, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelTexImage1D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelTexSubImage1D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, - GLsizei width, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, - GLenum internalFormat, - GLint x, GLint y, GLsizei width, GLint border); - -void intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, - GLenum internalFormat, - GLint x, GLint y, GLsizei width, GLsizei height, - GLint border); - -void intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, - GLint xoffset, GLint x, GLint y, GLsizei width); - -void intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, - GLint xoffset, GLint yoffset, - GLint x, GLint y, GLsizei width, GLsizei height); - -void intelGetTexImage(GLcontext * ctx, GLenum target, GLint level, - GLenum format, GLenum type, GLvoid * pixels, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLsizei imageSize, const GLvoid *data, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ); - -void intelCompressedTexSubImage2D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, GLint yoffset, - GLsizei width, GLsizei height, - GLenum format, GLsizei imageSize, - const GLvoid * pixels, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level, - GLvoid *pixels, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - void intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, unsigned long long offset, GLint depth, GLuint pitch); void intelSetTexBuffer(__DRIcontext *pDRICtx, diff --git a/src/mesa/drivers/dri/intel/intel_tex_copy.c b/src/mesa/drivers/dri/intel/intel_tex_copy.c index 90bbb8c6bb..028b49c14d 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_copy.c +++ b/src/mesa/drivers/dri/intel/intel_tex_copy.c @@ -73,11 +73,8 @@ get_teximage_source(struct intel_context *intel, GLenum internalFormat) return NULL; case GL_RGBA: case GL_RGBA8: - return intel_readbuf_region(intel); case GL_RGB: - if (intel->ctx.Visual.rgbBits == 16) - return intel_readbuf_region(intel); - return NULL; + return intel_readbuf_region(intel); default: return NULL; } @@ -99,14 +96,24 @@ do_copy_texsubimage(struct intel_context *intel, if (!intelImage->mt || !src) { if (INTEL_DEBUG & DEBUG_FALLBACKS) - fprintf(stderr, "%s fail %p %p\n", - __FUNCTION__, intelImage->mt, src); + fprintf(stderr, "%s fail %p %p (0x%08x)\n", + __FUNCTION__, intelImage->mt, src, internalFormat); + return GL_FALSE; + } + + if (intelImage->mt->cpp != src->cpp) { + if (INTEL_DEBUG & DEBUG_FALLBACKS) + fprintf(stderr, "%s fail %d vs %d cpp\n", + __FUNCTION__, intelImage->mt->cpp, src->cpp); return GL_FALSE; } intelFlush(ctx); LOCK_HARDWARE(intel); { + drm_intel_bo *dst_bo = intel_region_buffer(intel, + intelImage->mt->region, + INTEL_WRITE_PART); GLuint image_offset = intel_miptree_image_offset(intelImage->mt, intelImage->face, intelImage->level); @@ -118,8 +125,12 @@ do_copy_texsubimage(struct intel_context *intel, dstx += x - orig_x; dsty += y - orig_y; - /* image_offset may be non-page-aligned, but that's illegal for tiling. */ - assert(intelImage->mt->region->tiling == I915_TILING_NONE); + /* Can't blit to tiled buffers with non-tile-aligned offset. */ + if (intelImage->mt->region->tiling != I915_TILING_NONE && + (image_offset & 4095) != 0) { + UNLOCK_HARDWARE(intel); + return GL_FALSE; + } if (ctx->ReadBuffer->Name == 0) { /* reading from a window, adjust x, y */ @@ -140,35 +151,35 @@ do_copy_texsubimage(struct intel_context *intel, src_pitch = src->pitch; } - intelEmitCopyBlit(intel, - intelImage->mt->cpp, - src_pitch, - src->buffer, - 0, - src->tiling, - intelImage->mt->pitch, - intelImage->mt->region->buffer, - image_offset, - intelImage->mt->region->tiling, - x, y, dstx, dsty, width, height, - GL_COPY); + if (!intelEmitCopyBlit(intel, + intelImage->mt->cpp, + src_pitch, + src->buffer, + 0, + src->tiling, + intelImage->mt->pitch, + dst_bo, + image_offset, + intelImage->mt->region->tiling, + x, y, dstx, dsty, width, height, + GL_COPY)) { + UNLOCK_HARDWARE(intel); + return GL_FALSE; + } } UNLOCK_HARDWARE(intel); /* GL_SGIS_generate_mipmap */ if (intelImage->level == texObj->BaseLevel && texObj->GenerateMipmap) { - ctx->Driver.GenerateMipmap(ctx, target, texObj); + intel_generate_mipmap(ctx, target, texObj); } return GL_TRUE; } - - - -void +static void intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, GLenum internalFormat, GLint x, GLint y, GLsizei width, GLint border) @@ -214,7 +225,8 @@ intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, width, border); } -void + +static void intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, GLenum internalFormat, GLint x, GLint y, GLsizei width, GLsizei height, @@ -262,7 +274,7 @@ intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, } -void +static void intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, GLint xoffset, GLint x, GLint y, GLsizei width) { @@ -287,8 +299,7 @@ intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, } - -void +static void intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint x, GLint y, GLsizei width, GLsizei height) @@ -301,7 +312,6 @@ intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, _mesa_select_tex_image(ctx, texObj, target, level); GLenum internalFormat = texImage->InternalFormat; - /* Need to check texture is compatible with source format. */ @@ -316,3 +326,13 @@ intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, xoffset, yoffset, x, y, width, height); } } + + +void +intelInitTextureCopyImageFuncs(struct dd_function_table *functions) +{ + functions->CopyTexImage1D = intelCopyTexImage1D; + functions->CopyTexImage2D = intelCopyTexImage2D; + functions->CopyTexSubImage1D = intelCopyTexSubImage1D; + functions->CopyTexSubImage2D = intelCopyTexSubImage2D; +} diff --git a/src/mesa/drivers/dri/intel/intel_tex_image.c b/src/mesa/drivers/dri/intel/intel_tex_image.c index 5e61e9e95e..c5f5220837 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_image.c +++ b/src/mesa/drivers/dri/intel/intel_tex_image.c @@ -131,6 +131,7 @@ guess_and_alloc_mipmap_tree(struct intel_context *intel, comp_byte = intel_compressed_num_bytes(intelImage->base.TexFormat->MesaFormat); intelObj->mt = intel_miptree_create(intel, intelObj->base.Target, + intelImage->base._BaseFormat, intelImage->base.InternalFormat, firstLevel, lastLevel, @@ -205,7 +206,7 @@ try_pbo_upload(struct intel_context *intel, GLuint src_offset, src_stride; GLuint dst_offset, dst_stride; - if (!pbo || + if (unpack->BufferObj->Name == 0 || intel->ctx._ImageTransferState || unpack->SkipPixels || unpack->SkipRows) { DBG("%s: failure 1\n", __FUNCTION__); @@ -235,12 +236,15 @@ try_pbo_upload(struct intel_context *intel, INTEL_WRITE_FULL); - intelEmitCopyBlit(intel, - intelImage->mt->cpp, - src_stride, src_buffer, src_offset, GL_FALSE, - dst_stride, dst_buffer, dst_offset, GL_FALSE, - 0, 0, 0, 0, width, height, - GL_COPY); + if (!intelEmitCopyBlit(intel, + intelImage->mt->cpp, + src_stride, src_buffer, src_offset, GL_FALSE, + dst_stride, dst_buffer, dst_offset, GL_FALSE, + 0, 0, 0, 0, width, height, + GL_COPY)) { + UNLOCK_HARDWARE(intel); + return GL_FALSE; + } } UNLOCK_HARDWARE(intel); @@ -248,7 +252,6 @@ try_pbo_upload(struct intel_context *intel, } - static GLboolean try_pbo_zcopy(struct intel_context *intel, struct intel_texture_image *intelImage, @@ -261,7 +264,7 @@ try_pbo_zcopy(struct intel_context *intel, GLuint src_offset, src_stride; GLuint dst_offset, dst_stride; - if (!pbo || + if (unpack->BufferObj->Name == 0 || intel->ctx._ImageTransferState || unpack->SkipPixels || unpack->SkipRows) { DBG("%s: failure 1\n", __FUNCTION__); @@ -293,10 +296,6 @@ try_pbo_zcopy(struct intel_context *intel, } - - - - static void intelTexImage(GLcontext * ctx, GLint dims, @@ -307,7 +306,8 @@ intelTexImage(GLcontext * ctx, GLenum format, GLenum type, const void *pixels, const struct gl_pixelstore_attrib *unpack, struct gl_texture_object *texObj, - struct gl_texture_image *texImage, GLsizei imageSize, int compressed) + struct gl_texture_image *texImage, GLsizei imageSize, + GLboolean compressed) { struct intel_context *intel = intel_context(ctx); struct intel_texture_object *intelObj = intel_texture_object(texObj); @@ -316,7 +316,6 @@ intelTexImage(GLcontext * ctx, GLint postConvHeight = height; GLint texelBytes, sizeInBytes; GLuint dstRowStride = 0, srcRowStride = texImage->RowStride; - GLboolean needs_map; DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__, _mesa_lookup_enum_by_nr(target), level, width, height, depth, border); @@ -414,7 +413,9 @@ intelTexImage(GLcontext * ctx, * a miptree, so create one just for our level and store it in the image. * It'll get moved into the object miptree at validate time. */ - intelImage->mt = intel_miptree_create(intel, target, internalFormat, + intelImage->mt = intel_miptree_create(intel, target, + intelImage->base.TexFormat->BaseFormat, + internalFormat, level, level, width, height, depth, intelImage->base.TexFormat->TexelBytes, @@ -426,7 +427,7 @@ intelTexImage(GLcontext * ctx, */ if (dims <= 2 && intelImage->mt && - intel_buffer_object(unpack->BufferObj) && + unpack->BufferObj->Name != 0 && check_pbo_format(internalFormat, format, type, intelImage->base.TexFormat)) { @@ -464,8 +465,6 @@ intelTexImage(GLcontext * ctx, DBG("pbo upload failed\n"); } - - /* intelCopyTexImage calls this function with pixels == NULL, with * the expectation that the mipmap tree will be set up but nothing * more will be done. This is where those calls return: @@ -482,15 +481,8 @@ intelTexImage(GLcontext * ctx, LOCK_HARDWARE(intel); - /* Two cases where we need a mapping of the miptree: when the user supplied - * data is mapped as well (non-PBO, memcpy upload) or when we're going to do - * (software) mipmap generation. - */ - needs_map = (pixels != NULL) || (level == texObj->BaseLevel && - texObj->GenerateMipmap); - if (intelImage->mt) { - if (needs_map) + if (pixels != NULL) texImage->Data = intel_miptree_image_map(intel, intelImage->mt, intelImage->face, @@ -547,25 +539,26 @@ intelTexImage(GLcontext * ctx, format, type, pixels, unpack)) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); } - - /* GL_SGIS_generate_mipmap */ - if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - intel_generate_mipmap(ctx, target, texObj); - } } _mesa_unmap_teximage_pbo(ctx, unpack); if (intelImage->mt) { - if (needs_map) + if (pixels != NULL) intel_miptree_image_unmap(intel, intelImage->mt); texImage->Data = NULL; } UNLOCK_HARDWARE(intel); + + /* GL_SGIS_generate_mipmap */ + if (level == texObj->BaseLevel && texObj->GenerateMipmap) { + intel_generate_mipmap(ctx, target, texObj); + } } -void + +static void intelTexImage3D(GLcontext * ctx, GLenum target, GLint level, GLint internalFormat, @@ -578,11 +571,11 @@ intelTexImage3D(GLcontext * ctx, { intelTexImage(ctx, 3, target, level, internalFormat, width, height, depth, border, - format, type, pixels, unpack, texObj, texImage, 0, 0); + format, type, pixels, unpack, texObj, texImage, 0, GL_FALSE); } -void +static void intelTexImage2D(GLcontext * ctx, GLenum target, GLint level, GLint internalFormat, @@ -594,10 +587,11 @@ intelTexImage2D(GLcontext * ctx, { intelTexImage(ctx, 2, target, level, internalFormat, width, height, 1, border, - format, type, pixels, unpack, texObj, texImage, 0, 0); + format, type, pixels, unpack, texObj, texImage, 0, GL_FALSE); } -void + +static void intelTexImage1D(GLcontext * ctx, GLenum target, GLint level, GLint internalFormat, @@ -609,21 +603,24 @@ intelTexImage1D(GLcontext * ctx, { intelTexImage(ctx, 1, target, level, internalFormat, width, 1, 1, border, - format, type, pixels, unpack, texObj, texImage, 0, 0); + format, type, pixels, unpack, texObj, texImage, 0, GL_FALSE); } -void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLsizei imageSize, const GLvoid *data, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) + +static void +intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint border, + GLsizei imageSize, const GLvoid *data, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage ) { intelTexImage(ctx, 2, target, level, internalFormat, width, height, 1, border, - 0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, 1); + 0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, GL_TRUE); } + /** * Need to map texture image into memory before copying image data, * then unmap it. @@ -632,7 +629,7 @@ static void intel_get_tex_image(GLcontext * ctx, GLenum target, GLint level, GLenum format, GLenum type, GLvoid * pixels, struct gl_texture_object *texObj, - struct gl_texture_image *texImage, int compressed) + struct gl_texture_image *texImage, GLboolean compressed) { struct intel_context *intel = intel_context(ctx); struct intel_texture_image *intelImage = intel_texture_image(texImage); @@ -686,28 +683,29 @@ intel_get_tex_image(GLcontext * ctx, GLenum target, GLint level, } } -void + +static void intelGetTexImage(GLcontext * ctx, GLenum target, GLint level, GLenum format, GLenum type, GLvoid * pixels, struct gl_texture_object *texObj, struct gl_texture_image *texImage) { intel_get_tex_image(ctx, target, level, format, type, pixels, - texObj, texImage, 0); - - + texObj, texImage, GL_FALSE); } -void + +static void intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level, GLvoid *pixels, struct gl_texture_object *texObj, struct gl_texture_image *texImage) { intel_get_tex_image(ctx, target, level, 0, 0, pixels, - texObj, texImage, 1); + texObj, texImage, GL_TRUE); } + void intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, unsigned long long offset, GLint depth, GLuint pitch) @@ -816,3 +814,16 @@ intelSetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv) */ intelSetTexBuffer2(pDRICtx, target, GLX_TEXTURE_FORMAT_RGBA_EXT, dPriv); } + + +void +intelInitTextureImageFuncs(struct dd_function_table *functions) +{ + functions->TexImage1D = intelTexImage1D; + functions->TexImage2D = intelTexImage2D; + functions->TexImage3D = intelTexImage3D; + functions->GetTexImage = intelGetTexImage; + + functions->CompressedTexImage2D = intelCompressedTexImage2D; + functions->GetCompressedTexImage = intelGetCompressedTexImage; +} diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.c b/src/mesa/drivers/dri/intel/intel_tex_layout.c index e6f9a41779..7d69ea4484 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_layout.c +++ b/src/mesa/drivers/dri/intel/intel_tex_layout.c @@ -35,26 +35,39 @@ #include "intel_context.h" #include "main/macros.h" -GLuint intel_compressed_alignment(GLenum internalFormat) +void intel_get_texture_alignment_unit(GLenum internalFormat, GLuint *w, GLuint *h) { - GLuint alignment = 4; - switch (internalFormat) { case GL_COMPRESSED_RGB_FXT1_3DFX: case GL_COMPRESSED_RGBA_FXT1_3DFX: - alignment = 8; + *w = 8; + *h = 4; + break; + + case GL_RGB_S3TC: + case GL_RGB4_S3TC: + case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: + case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: + case GL_RGBA_S3TC: + case GL_RGBA4_S3TC: + case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: + case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: + *w = 4; + *h = 4; break; default: + *w = 4; + *h = 2; break; } - - return alignment; } -void i945_miptree_layout_2d( struct intel_context *intel, struct intel_mipmap_tree *mt ) +void i945_miptree_layout_2d( struct intel_context *intel, + struct intel_mipmap_tree *mt, + uint32_t tiling ) { - GLint align_h = 2, align_w = 4; + GLuint align_h = 2, align_w = 4; GLuint level; GLuint x = 0; GLuint y = 0; @@ -62,9 +75,9 @@ void i945_miptree_layout_2d( struct intel_context *intel, struct intel_mipmap_tr GLuint height = mt->height0; mt->pitch = mt->width0; + intel_get_texture_alignment_unit(mt->internal_format, &align_w, &align_h); if (mt->compressed) { - align_w = intel_compressed_alignment(mt->internal_format); mt->pitch = ALIGN(mt->width0, align_w); } @@ -92,7 +105,7 @@ void i945_miptree_layout_2d( struct intel_context *intel, struct intel_mipmap_tr /* Pitch must be a whole number of dwords, even though we * express it in texels. */ - mt->pitch = intel_miptree_pitch_align (intel, mt, mt->pitch); + mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, mt->pitch); mt->total_height = 0; for ( level = mt->first_level ; level <= mt->last_level ; level++ ) { diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.h b/src/mesa/drivers/dri/intel/intel_tex_layout.h index dbc90e6f9b..c9de9b5678 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_layout.h +++ b/src/mesa/drivers/dri/intel/intel_tex_layout.h @@ -38,5 +38,7 @@ static GLuint minify( GLuint d ) return MAX2(1, d>>1); } -extern void i945_miptree_layout_2d( struct intel_context *intel, struct intel_mipmap_tree *mt ); -extern GLuint intel_compressed_alignment(GLenum); +extern void i945_miptree_layout_2d(struct intel_context *intel, + struct intel_mipmap_tree *mt, + uint32_t tiling); +extern void intel_get_texture_alignment_unit(GLenum, GLuint *, GLuint *); diff --git a/src/mesa/drivers/dri/intel/intel_tex_subimage.c b/src/mesa/drivers/dri/intel/intel_tex_subimage.c index f86de56897..89037073f8 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_subimage.c +++ b/src/mesa/drivers/dri/intel/intel_tex_subimage.c @@ -44,10 +44,12 @@ intelTexSubimage(GLcontext * ctx, GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLint width, GLint height, GLint depth, + GLsizei imageSize, GLenum format, GLenum type, const void *pixels, const struct gl_pixelstore_attrib *packing, struct gl_texture_object *texObj, - struct gl_texture_image *texImage) + struct gl_texture_image *texImage, + GLboolean compressed) { struct intel_context *intel = intel_context(ctx); struct intel_texture_image *intelImage = intel_texture_image(texImage); @@ -59,9 +61,14 @@ intelTexSubimage(GLcontext * ctx, intelFlush(ctx); - pixels = - _mesa_validate_pbo_teximage(ctx, dims, width, height, depth, format, - type, pixels, packing, "glTexSubImage2D"); + if (compressed) + pixels = _mesa_validate_pbo_compressed_teximage(ctx, imageSize, + pixels, packing, + "glCompressedTexImage"); + else + pixels = _mesa_validate_pbo_teximage(ctx, dims, width, height, depth, + format, type, pixels, packing, + "glTexSubImage"); if (!pixels) return; @@ -90,20 +97,28 @@ intelTexSubimage(GLcontext * ctx, assert(dstRowStride); - if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat, - texImage->TexFormat, - texImage->Data, - xoffset, yoffset, zoffset, - dstRowStride, - texImage->ImageOffsets, - width, height, depth, - format, type, pixels, packing)) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "intelTexSubImage"); + if (compressed) { + if (intelImage->mt) { + struct intel_region *dst = intelImage->mt->region; + + _mesa_copy_rect(texImage->Data, dst->cpp, dst->pitch, + xoffset, yoffset / 4, + (width + 3) & ~3, (height + 3) / 4, + pixels, (width + 3) & ~3, 0, 0); + } else + memcpy(texImage->Data, pixels, imageSize); } - - /* GL_SGIS_generate_mipmap */ - if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - intel_generate_mipmap(ctx, target, texObj); + else { + if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat, + texImage->TexFormat, + texImage->Data, + xoffset, yoffset, zoffset, + dstRowStride, + texImage->ImageOffsets, + width, height, depth, + format, type, pixels, packing)) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "intelTexSubImage"); + } } _mesa_unmap_teximage_pbo(ctx, packing); @@ -114,13 +129,15 @@ intelTexSubimage(GLcontext * ctx, } UNLOCK_HARDWARE(intel); -} - - + /* GL_SGIS_generate_mipmap */ + if (level == texObj->BaseLevel && texObj->GenerateMipmap) { + intel_generate_mipmap(ctx, target, texObj); + } +} -void +static void intelTexSubImage3D(GLcontext * ctx, GLenum target, GLint level, @@ -132,18 +149,15 @@ intelTexSubImage3D(GLcontext * ctx, struct gl_texture_object *texObj, struct gl_texture_image *texImage) { - intelTexSubimage(ctx, 3, target, level, xoffset, yoffset, zoffset, - width, height, depth, - format, type, pixels, packing, texObj, texImage); - + width, height, depth, 0, + format, type, pixels, packing, texObj, texImage, GL_FALSE); } - -void +static void intelTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, @@ -155,17 +169,15 @@ intelTexSubImage2D(GLcontext * ctx, struct gl_texture_object *texObj, struct gl_texture_image *texImage) { - intelTexSubimage(ctx, 2, target, level, xoffset, yoffset, 0, - width, height, 1, - format, type, pixels, packing, texObj, texImage); - + width, height, 1, 0, + format, type, pixels, packing, texObj, texImage, GL_FALSE); } -void +static void intelTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, @@ -180,12 +192,11 @@ intelTexSubImage1D(GLcontext * ctx, intelTexSubimage(ctx, 1, target, level, xoffset, 0, 0, - width, 1, 1, - format, type, pixels, packing, texObj, texImage); - + width, 1, 1, 0, + format, type, pixels, packing, texObj, texImage, GL_FALSE); } -void +static void intelCompressedTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, @@ -196,6 +207,20 @@ intelCompressedTexSubImage2D(GLcontext * ctx, struct gl_texture_object *texObj, struct gl_texture_image *texImage) { - fprintf(stderr, "stubbed CompressedTexSubImage2D: %dx%d@%dx%d\n", - width, height, xoffset, yoffset); + intelTexSubimage(ctx, 2, + target, level, + xoffset, yoffset, 0, + width, height, 1, imageSize, + format, 0, pixels, &ctx->Unpack, texObj, texImage, GL_TRUE); +} + + + +void +intelInitTextureSubImageFuncs(struct dd_function_table *functions) +{ + functions->TexSubImage1D = intelTexSubImage1D; + functions->TexSubImage2D = intelTexSubImage2D; + functions->TexSubImage3D = intelTexSubImage3D; + functions->CompressedTexSubImage2D = intelCompressedTexSubImage2D; } diff --git a/src/mesa/drivers/dri/intel/intel_tex_validate.c b/src/mesa/drivers/dri/intel/intel_tex_validate.c index 05a375e1f3..a284d5475f 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_validate.c +++ b/src/mesa/drivers/dri/intel/intel_tex_validate.c @@ -199,6 +199,7 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) if (!intelObj->mt) { intelObj->mt = intel_miptree_create(intel, intelObj->base.Target, + firstImage->base._BaseFormat, firstImage->base.InternalFormat, intelObj->firstLevel, intelObj->lastLevel, @@ -241,7 +242,7 @@ intel_tex_map_level_images(struct intel_context *intel, struct intel_texture_image *intelImage = intel_texture_image(intelObj->base.Image[face][level]); - if (intelImage->mt) { + if (intelImage && intelImage->mt) { intelImage->base.Data = intel_miptree_image_map(intel, intelImage->mt, @@ -268,7 +269,7 @@ intel_tex_unmap_level_images(struct intel_context *intel, struct intel_texture_image *intelImage = intel_texture_image(intelObj->base.Image[face][level]); - if (intelImage->mt) { + if (intelImage && intelImage->mt) { intel_miptree_image_unmap(intel, intelImage->mt); intelImage->base.Data = NULL; } |