diff options
| author | Brian Paul <brianp@vmware.com> | 2010-01-22 12:17:02 -0700 |
|---|---|---|
| committer | Brian Paul <brianp@vmware.com> | 2010-01-22 12:17:02 -0700 |
| commit | cd8614b0287dc5a69725ec4ee0208fad61f7789e (patch) | |
| tree | 3ee089b8384e7a60c5c3a3cc87f2a633bd724bbe /src/gallium/drivers | |
| parent | 2b20b604277e3cdf7afb2431b50dbb05da12ff1c (diff) | |
| parent | 64871747bb7b611ffe429fbf1724bd98ee25dd84 (diff) | |
Merge branch 'mesa_7_7_branch'
Conflicts:
src/gallium/auxiliary/draw/draw_context.c
src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
src/gallium/auxiliary/pipebuffer/Makefile
src/gallium/auxiliary/pipebuffer/SConscript
src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
src/gallium/auxiliary/tgsi/tgsi_scan.c
src/gallium/drivers/i915/i915_surface.c
src/gallium/drivers/i915/i915_texture.c
src/gallium/drivers/llvmpipe/lp_setup.c
src/gallium/drivers/llvmpipe/lp_tex_sample_c.c
src/gallium/drivers/llvmpipe/lp_texture.c
src/gallium/drivers/softpipe/sp_prim_vbuf.c
src/gallium/state_trackers/xorg/xorg_dri2.c
src/gallium/winsys/drm/intel/gem/intel_drm_api.c
src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
src/gallium/winsys/drm/radeon/core/radeon_drm.c
src/gallium/winsys/drm/vmware/core/vmw_screen_dri.c
src/mesa/state_tracker/st_cb_clear.c
Diffstat (limited to 'src/gallium/drivers')
67 files changed, 1830 insertions, 211 deletions
diff --git a/src/gallium/drivers/i915/i915_clear.c b/src/gallium/drivers/i915/i915_clear.c index 90530f2826..0d0859f8f3 100644 --- a/src/gallium/drivers/i915/i915_clear.c +++ b/src/gallium/drivers/i915/i915_clear.c @@ -32,7 +32,6 @@ #include "util/u_clear.h" #include "i915_context.h" -#include "i915_state.h" /** diff --git a/src/gallium/drivers/i915/i915_context.c b/src/gallium/drivers/i915/i915_context.c index 89feeade75..4c5ff37ca8 100644 --- a/src/gallium/drivers/i915/i915_context.c +++ b/src/gallium/drivers/i915/i915_context.c @@ -29,12 +29,9 @@ #include "i915_state.h" #include "i915_screen.h" #include "i915_batch.h" -#include "i915_texture.h" -#include "i915_reg.h" #include "draw/draw_context.h" #include "pipe/p_defines.h" -#include "pipe/internal/p_winsys_screen.h" #include "pipe/p_inlines.h" #include "util/u_memory.h" #include "pipe/p_screen.h" diff --git a/src/gallium/drivers/i915/i915_debug.c b/src/gallium/drivers/i915/i915_debug.c index c6e6d6fd31..237654d26b 100644 --- a/src/gallium/drivers/i915/i915_debug.c +++ b/src/gallium/drivers/i915/i915_debug.c @@ -29,7 +29,6 @@ #include "i915_context.h" #include "i915_debug.h" #include "i915_batch.h" -#include "pipe/internal/p_winsys_screen.h" #include "util/u_debug.h" diff --git a/src/gallium/drivers/i915/i915_debug_fp.c b/src/gallium/drivers/i915/i915_debug_fp.c index 9c5b117b6d..f9c40d8a11 100644 --- a/src/gallium/drivers/i915/i915_debug_fp.c +++ b/src/gallium/drivers/i915/i915_debug_fp.c @@ -29,7 +29,6 @@ #include "i915_reg.h" #include "i915_debug.h" #include "pipe/internal/p_winsys_screen.h" -#include "util/u_memory.h" static void diff --git a/src/gallium/drivers/i915/i915_state.c b/src/gallium/drivers/i915/i915_state.c index 0fab6e1bc3..06949c15f8 100644 --- a/src/gallium/drivers/i915/i915_state.c +++ b/src/gallium/drivers/i915/i915_state.c @@ -30,7 +30,6 @@ #include "draw/draw_context.h" -#include "pipe/internal/p_winsys_screen.h" #include "pipe/p_inlines.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -38,7 +37,6 @@ #include "i915_context.h" #include "i915_reg.h" -#include "i915_state.h" #include "i915_state_inlines.h" #include "i915_fpc.h" diff --git a/src/gallium/drivers/i915/i915_state_derived.c b/src/gallium/drivers/i915/i915_state_derived.c index 03dd5091a6..f5b0e9f011 100644 --- a/src/gallium/drivers/i915/i915_state_derived.c +++ b/src/gallium/drivers/i915/i915_state_derived.c @@ -33,7 +33,6 @@ #include "i915_context.h" #include "i915_state.h" #include "i915_reg.h" -#include "i915_fpc.h" diff --git a/src/gallium/drivers/i915/i915_state_sampler.c b/src/gallium/drivers/i915/i915_state_sampler.c index cbac4175c8..e5c6d87215 100644 --- a/src/gallium/drivers/i915/i915_state_sampler.c +++ b/src/gallium/drivers/i915/i915_state_sampler.c @@ -27,7 +27,6 @@ #include "pipe/p_context.h" #include "pipe/p_state.h" -#include "util/u_memory.h" #include "i915_state_inlines.h" #include "i915_context.h" diff --git a/src/gallium/drivers/i915/i915_surface.c b/src/gallium/drivers/i915/i915_surface.c index c693eb30e8..13684aa59c 100644 --- a/src/gallium/drivers/i915/i915_surface.c +++ b/src/gallium/drivers/i915/i915_surface.c @@ -27,14 +27,10 @@ #include "i915_context.h" #include "i915_blit.h" -#include "i915_state.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "pipe/p_inlines.h" #include "pipe/internal/p_winsys_screen.h" #include "util/u_format.h" -#include "util/u_tile.h" -#include "util/u_rect.h" /* Assumes all values are within bounds -- no checking at this level - diff --git a/src/gallium/drivers/i915/i915_texture.c b/src/gallium/drivers/i915/i915_texture.c index 50a9e19094..441bc4f193 100644 --- a/src/gallium/drivers/i915/i915_texture.c +++ b/src/gallium/drivers/i915/i915_texture.c @@ -41,7 +41,6 @@ #include "i915_context.h" #include "i915_texture.h" -#include "i915_debug.h" #include "i915_screen.h" #include "intel_winsys.h" diff --git a/src/gallium/drivers/llvmpipe/lp_bld_alpha.c b/src/gallium/drivers/llvmpipe/lp_bld_alpha.c index 2b4bc5c819..7245730350 100644 --- a/src/gallium/drivers/llvmpipe/lp_bld_alpha.c +++ b/src/gallium/drivers/llvmpipe/lp_bld_alpha.c @@ -35,7 +35,6 @@ #include "lp_bld_type.h" #include "lp_bld_const.h" -#include "lp_bld_arit.h" #include "lp_bld_logic.h" #include "lp_bld_flow.h" #include "lp_bld_debug.h" diff --git a/src/gallium/drivers/llvmpipe/lp_bld_arit.c b/src/gallium/drivers/llvmpipe/lp_bld_arit.c index eea6b5d6a5..1aee9b35f3 100644 --- a/src/gallium/drivers/llvmpipe/lp_bld_arit.c +++ b/src/gallium/drivers/llvmpipe/lp_bld_arit.c @@ -56,7 +56,6 @@ #include "lp_bld_intr.h" #include "lp_bld_logic.h" #include "lp_bld_pack.h" -#include "lp_bld_debug.h" #include "lp_bld_arit.h" diff --git a/src/gallium/drivers/llvmpipe/lp_bld_blend_soa.c b/src/gallium/drivers/llvmpipe/lp_bld_blend_soa.c index 9511299d55..5d5ca7a5d4 100644 --- a/src/gallium/drivers/llvmpipe/lp_bld_blend_soa.c +++ b/src/gallium/drivers/llvmpipe/lp_bld_blend_soa.c @@ -71,7 +71,6 @@ #include "pipe/p_state.h" #include "lp_bld_type.h" -#include "lp_bld_const.h" #include "lp_bld_arit.h" #include "lp_bld_blend.h" diff --git a/src/gallium/drivers/llvmpipe/lp_bld_conv.c b/src/gallium/drivers/llvmpipe/lp_bld_conv.c index 9935209437..ebf554cd04 100644 --- a/src/gallium/drivers/llvmpipe/lp_bld_conv.c +++ b/src/gallium/drivers/llvmpipe/lp_bld_conv.c @@ -63,11 +63,9 @@ #include "util/u_debug.h" #include "util/u_math.h" -#include "util/u_cpu_detect.h" #include "lp_bld_type.h" #include "lp_bld_const.h" -#include "lp_bld_intr.h" #include "lp_bld_arit.h" #include "lp_bld_pack.h" #include "lp_bld_conv.h" diff --git a/src/gallium/drivers/llvmpipe/lp_bld_format_aos.c b/src/gallium/drivers/llvmpipe/lp_bld_format_aos.c index 10e82f120b..dfa080b853 100644 --- a/src/gallium/drivers/llvmpipe/lp_bld_format_aos.c +++ b/src/gallium/drivers/llvmpipe/lp_bld_format_aos.c @@ -38,7 +38,6 @@ #include "lp_bld_type.h" #include "lp_bld_const.h" -#include "lp_bld_logic.h" #include "lp_bld_swizzle.h" #include "lp_bld_format.h" diff --git a/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c b/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c index fb1eda4423..85e3b1bdd4 100644 --- a/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c +++ b/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c @@ -47,13 +47,11 @@ #include "tgsi/tgsi_exec.h" #include "lp_bld_type.h" #include "lp_bld_const.h" -#include "lp_bld_intr.h" #include "lp_bld_arit.h" #include "lp_bld_logic.h" #include "lp_bld_swizzle.h" #include "lp_bld_flow.h" #include "lp_bld_tgsi.h" -#include "lp_bld_debug.h" #define LP_MAX_TEMPS 256 diff --git a/src/gallium/drivers/llvmpipe/lp_draw_arrays.c b/src/gallium/drivers/llvmpipe/lp_draw_arrays.c index c152b4413f..a0316194c6 100644 --- a/src/gallium/drivers/llvmpipe/lp_draw_arrays.c +++ b/src/gallium/drivers/llvmpipe/lp_draw_arrays.c @@ -33,8 +33,6 @@ #include "pipe/p_defines.h" #include "pipe/p_context.h" -#include "pipe/internal/p_winsys_screen.h" -#include "pipe/p_inlines.h" #include "util/u_prim.h" #include "lp_buffer.h" diff --git a/src/gallium/drivers/llvmpipe/lp_flush.c b/src/gallium/drivers/llvmpipe/lp_flush.c index cd8381fe30..6c81012e84 100644 --- a/src/gallium/drivers/llvmpipe/lp_flush.c +++ b/src/gallium/drivers/llvmpipe/lp_flush.c @@ -37,8 +37,6 @@ #include "lp_surface.h" #include "lp_state.h" #include "lp_tile_cache.h" -#include "lp_tex_cache.h" -#include "lp_winsys.h" void diff --git a/src/gallium/drivers/llvmpipe/lp_jit.c b/src/gallium/drivers/llvmpipe/lp_jit.c index 4ef0783f3e..9e0118c55b 100644 --- a/src/gallium/drivers/llvmpipe/lp_jit.c +++ b/src/gallium/drivers/llvmpipe/lp_jit.c @@ -39,7 +39,6 @@ #include "util/u_cpu_detect.h" #include "lp_screen.h" #include "lp_bld_intr.h" -#include "lp_bld_misc.h" #include "lp_jit.h" diff --git a/src/gallium/drivers/llvmpipe/lp_state_fs.c b/src/gallium/drivers/llvmpipe/lp_state_fs.c index 9f4bbef73f..70e2a082cc 100644 --- a/src/gallium/drivers/llvmpipe/lp_state_fs.c +++ b/src/gallium/drivers/llvmpipe/lp_state_fs.c @@ -62,7 +62,6 @@ #include "util/u_memory.h" #include "util/u_format.h" #include "util/u_debug_dump.h" -#include "pipe/internal/p_winsys_screen.h" #include "pipe/p_shader_tokens.h" #include "draw/draw_context.h" #include "tgsi/tgsi_dump.h" @@ -85,7 +84,6 @@ #include "lp_context.h" #include "lp_buffer.h" #include "lp_state.h" -#include "lp_quad.h" #include "lp_tex_sample.h" #include "lp_debug.h" diff --git a/src/gallium/drivers/llvmpipe/lp_state_surface.c b/src/gallium/drivers/llvmpipe/lp_state_surface.c index e37ff04f3d..3b08b0d1d7 100644 --- a/src/gallium/drivers/llvmpipe/lp_state_surface.c +++ b/src/gallium/drivers/llvmpipe/lp_state_surface.c @@ -30,7 +30,6 @@ #include "lp_context.h" #include "lp_state.h" -#include "lp_surface.h" #include "lp_tile_cache.h" #include "draw/draw_context.h" diff --git a/src/gallium/drivers/llvmpipe/lp_state_vertex.c b/src/gallium/drivers/llvmpipe/lp_state_vertex.c index 1a17631a4c..57ac25ea0c 100644 --- a/src/gallium/drivers/llvmpipe/lp_state_vertex.c +++ b/src/gallium/drivers/llvmpipe/lp_state_vertex.c @@ -31,7 +31,6 @@ #include "lp_context.h" #include "lp_state.h" -#include "lp_surface.h" #include "draw/draw_context.h" diff --git a/src/gallium/drivers/llvmpipe/lp_test_blend.c b/src/gallium/drivers/llvmpipe/lp_test_blend.c index 6c29e8d8ac..7b65bab43a 100644 --- a/src/gallium/drivers/llvmpipe/lp_test_blend.c +++ b/src/gallium/drivers/llvmpipe/lp_test_blend.c @@ -38,7 +38,6 @@ #include "lp_bld_type.h" -#include "lp_bld_arit.h" #include "lp_bld_blend.h" #include "lp_bld_debug.h" #include "lp_test.h" diff --git a/src/gallium/drivers/llvmpipe/lp_tex_cache.c b/src/gallium/drivers/llvmpipe/lp_tex_cache.c index a6d9a2c1ac..8094625d74 100644 --- a/src/gallium/drivers/llvmpipe/lp_tex_cache.c +++ b/src/gallium/drivers/llvmpipe/lp_tex_cache.c @@ -38,7 +38,6 @@ #include "util/u_format.h" #include "util/u_math.h" #include "lp_context.h" -#include "lp_surface.h" #include "lp_texture.h" #include "lp_tex_cache.h" diff --git a/src/gallium/drivers/llvmpipe/lp_tex_sample_c.c b/src/gallium/drivers/llvmpipe/lp_tex_sample_c.c new file mode 100644 index 0000000000..d1f5d9505d --- /dev/null +++ b/src/gallium/drivers/llvmpipe/lp_tex_sample_c.c @@ -0,0 +1,1712 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * Copyright 2008 VMware, Inc. All rights reserved. + * + * 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, sub license, 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 NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS 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. + * + **************************************************************************/ + +/** + * Texture sampling + * + * Authors: + * Brian Paul + */ + +#include "lp_context.h" +#include "lp_quad.h" +#include "lp_texture.h" +#include "lp_tex_sample.h" +#include "lp_tex_cache.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" +#include "util/u_math.h" +#include "util/u_memory.h" + + + +/* + * Note, the FRAC macro has to work perfectly. Otherwise you'll sometimes + * see 1-pixel bands of improperly weighted linear-filtered textures. + * The tests/texwrap.c demo is a good test. + * Also note, FRAC(x) doesn't truly return the fractional part of x for x < 0. + * Instead, if x < 0 then FRAC(x) = 1 - true_frac(x). + */ +#define FRAC(f) ((f) - util_ifloor(f)) + + +/** + * Linear interpolation macro + */ +static INLINE float +lerp(float a, float v0, float v1) +{ + return v0 + a * (v1 - v0); +} + + +/** + * Do 2D/biliner interpolation of float values. + * v00, v10, v01 and v11 are typically four texture samples in a square/box. + * a and b are the horizontal and vertical interpolants. + * It's important that this function is inlined when compiled with + * optimization! If we find that's not true on some systems, convert + * to a macro. + */ +static INLINE float +lerp_2d(float a, float b, + float v00, float v10, float v01, float v11) +{ + const float temp0 = lerp(a, v00, v10); + const float temp1 = lerp(a, v01, v11); + return lerp(b, temp0, temp1); +} + + +/** + * As above, but 3D interpolation of 8 values. + */ +static INLINE float +lerp_3d(float a, float b, float c, + float v000, float v100, float v010, float v110, + float v001, float v101, float v011, float v111) +{ + const float temp0 = lerp_2d(a, b, v000, v100, v010, v110); + const float temp1 = lerp_2d(a, b, v001, v101, v011, v111); + return lerp(c, temp0, temp1); +} + + + +/** + * If A is a signed integer, A % B doesn't give the right value for A < 0 + * (in terms of texture repeat). Just casting to unsigned fixes that. + */ +#define REMAINDER(A, B) ((unsigned) (A) % (unsigned) (B)) + + +/** + * Apply texture coord wrapping mode and return integer texture indexes + * for a vector of four texcoords (S or T or P). + * \param wrapMode PIPE_TEX_WRAP_x + * \param s the incoming texcoords + * \param size the texture image size + * \param icoord returns the integer texcoords + * \return integer texture index + */ +static INLINE void +nearest_texcoord_4(unsigned wrapMode, const float s[4], unsigned size, + int icoord[4]) +{ + uint ch; + switch (wrapMode) { + case PIPE_TEX_WRAP_REPEAT: + /* s limited to [0,1) */ + /* i limited to [0,size-1] */ + for (ch = 0; ch < 4; ch++) { + int i = util_ifloor(s[ch] * size); + icoord[ch] = REMAINDER(i, size); + } + return; + case PIPE_TEX_WRAP_CLAMP: + /* s limited to [0,1] */ + /* i limited to [0,size-1] */ + for (ch = 0; ch < 4; ch++) { + if (s[ch] <= 0.0F) + icoord[ch] = 0; + else if (s[ch] >= 1.0F) + icoord[ch] = size - 1; + else + icoord[ch] = util_ifloor(s[ch] * size); + } + return; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + { + /* s limited to [min,max] */ + /* i limited to [0, size-1] */ + const float min = 1.0F / (2.0F * size); + const float max = 1.0F - min; + for (ch = 0; ch < 4; ch++) { + if (s[ch] < min) + icoord[ch] = 0; + else if (s[ch] > max) + icoord[ch] = size - 1; + else + icoord[ch] = util_ifloor(s[ch] * size); + } + } + return; + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + { + /* s limited to [min,max] */ + /* i limited to [-1, size] */ + const float min = -1.0F / (2.0F * size); + const float max = 1.0F - min; + for (ch = 0; ch < 4; ch++) { + if (s[ch] <= min) + icoord[ch] = -1; + else if (s[ch] >= max) + icoord[ch] = size; + else + icoord[ch] = util_ifloor(s[ch] * size); + } + } + return; + case PIPE_TEX_WRAP_MIRROR_REPEAT: + { + const float min = 1.0F / (2.0F * size); + const float max = 1.0F - min; + for (ch = 0; ch < 4; ch++) { + const int flr = util_ifloor(s[ch]); + float u; + if (flr & 1) + u = 1.0F - (s[ch] - (float) flr); + else + u = s[ch] - (float) flr; + if (u < min) + icoord[ch] = 0; + else if (u > max) + icoord[ch] = size - 1; + else + icoord[ch] = util_ifloor(u * size); + } + } + return; + case PIPE_TEX_WRAP_MIRROR_CLAMP: + for (ch = 0; ch < 4; ch++) { + /* s limited to [0,1] */ + /* i limited to [0,size-1] */ + const float u = fabsf(s[ch]); + if (u <= 0.0F) + icoord[ch] = 0; + else if (u >= 1.0F) + icoord[ch] = size - 1; + else + icoord[ch] = util_ifloor(u * size); + } + return; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: + { + /* s limited to [min,max] */ + /* i limited to [0, size-1] */ + const float min = 1.0F / (2.0F * size); + const float max = 1.0F - min; + for (ch = 0; ch < 4; ch++) { + const float u = fabsf(s[ch]); + if (u < min) + icoord[ch] = 0; + else if (u > max) + icoord[ch] = size - 1; + else + icoord[ch] = util_ifloor(u * size); + } + } + return; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: + { + /* s limited to [min,max] */ + /* i limited to [0, size-1] */ + const float min = -1.0F / (2.0F * size); + const float max = 1.0F - min; + for (ch = 0; ch < 4; ch++) { + const float u = fabsf(s[ch]); + if (u < min) + icoord[ch] = -1; + else if (u > max) + icoord[ch] = size; + else + icoord[ch] = util_ifloor(u * size); + } + } + return; + default: + assert(0); + } +} + + +/** + * Used to compute texel locations for linear sampling for four texcoords. + * \param wrapMode PIPE_TEX_WRAP_x + * \param s the texcoords + * \param size the texture image size + * \param icoord0 returns first texture indexes + * \param icoord1 returns second texture indexes (usually icoord0 + 1) + * \param w returns blend factor/weight between texture indexes + * \param icoord returns the computed integer texture coords + */ +static INLINE void +linear_texcoord_4(unsigned wrapMode, const float s[4], unsigned size, + int icoord0[4], int icoord1[4], float w[4]) +{ + uint ch; + + switch (wrapMode) { + case PIPE_TEX_WRAP_REPEAT: + for (ch = 0; ch < 4; ch++) { + float u = s[ch] * size - 0.5F; + icoord0[ch] = REMAINDER(util_ifloor(u), size); + icoord1[ch] = REMAINDER(icoord0[ch] + 1, size); + w[ch] = FRAC(u); + } + break;; + case PIPE_TEX_WRAP_CLAMP: + for (ch = 0; ch < 4; ch++) { + float u = CLAMP(s[ch], 0.0F, 1.0F); + u = u * size - 0.5f; + icoord0[ch] = util_ifloor(u); + icoord1[ch] = icoord0[ch] + 1; + w[ch] = FRAC(u); + } + break;; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + for (ch = 0; ch < 4; ch++) { + float u = CLAMP(s[ch], 0.0F, 1.0F); + u = u * size - 0.5f; + icoord0[ch] = util_ifloor(u); + icoord1[ch] = icoord0[ch] + 1; + if (icoord0[ch] < 0) + icoord0[ch] = 0; + if (icoord1[ch] >= (int) size) + icoord1[ch] = size - 1; + w[ch] = FRAC(u); + } + break;; + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + { + const float min = -1.0F / (2.0F * size); + const float max = 1.0F - min; + for (ch = 0; ch < 4; ch++) { + float u = CLAMP(s[ch], min, max); + u = u * size - 0.5f; + icoord0[ch] = util_ifloor(u); + icoord1[ch] = icoord0[ch] + 1; + w[ch] = FRAC(u); + } + } + break;; + case PIPE_TEX_WRAP_MIRROR_REPEAT: + for (ch = 0; ch < 4; ch++) { + const int flr = util_ifloor(s[ch]); + float u; + if (flr & 1) + u = 1.0F - (s[ch] - (float) flr); + else + u = s[ch] - (float) flr; + u = u * size - 0.5F; + icoord0[ch] = util_ifloor(u); + icoord1[ch] = icoord0[ch] + 1; + if (icoord0[ch] < 0) + icoord0[ch] = 0; + if (icoord1[ch] >= (int) size) + icoord1[ch] = size - 1; + w[ch] = FRAC(u); + } + break;; + case PIPE_TEX_WRAP_MIRROR_CLAMP: + for (ch = 0; ch < 4; ch++) { + float u = fabsf(s[ch]); + if (u >= 1.0F) + u = (float) size; + else + u *= size; + u -= 0.5F; + icoord0[ch] = util_ifloor(u); + icoord1[ch] = icoord0[ch] + 1; + w[ch] = FRAC(u); + } + break;; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: + for (ch = 0; ch < 4; ch++) { + float u = fabsf(s[ch]); + if (u >= 1.0F) + u = (float) size; + else + u *= size; + u -= 0.5F; + icoord0[ch] = util_ifloor(u); + icoord1[ch] = icoord0[ch] + 1; + if (icoord0[ch] < 0) + icoord0[ch] = 0; + if (icoord1[ch] >= (int) size) + icoord1[ch] = size - 1; + w[ch] = FRAC(u); + } + break;; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: + { + const float min = -1.0F / (2.0F * size); + const float max = 1.0F - min; + for (ch = 0; ch < 4; ch++) { + float u = fabsf(s[ch]); + if (u <= min) + u = min * size; + else if (u >= max) + u = max * size; + else + u *= size; + u -= 0.5F; + icoord0[ch] = util_ifloor(u); + icoord1[ch] = icoord0[ch] + 1; + w[ch] = FRAC(u); + } + } + break;; + default: + assert(0); + } +} + + +/** + * For RECT textures / unnormalized texcoords + * Only a subset of wrap modes supported. + */ +static INLINE void +nearest_texcoord_unnorm_4(unsigned wrapMode, const float s[4], unsigned size, + int icoord[4]) +{ + uint ch; + switch (wrapMode) { + case PIPE_TEX_WRAP_CLAMP: + for (ch = 0; ch < 4; ch++) { + int i = util_ifloor(s[ch]); + icoord[ch]= CLAMP(i, 0, (int) size-1); + } + return; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + /* fall-through */ + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + for (ch = 0; ch < 4; ch++) { + icoord[ch]= util_ifloor( CLAMP(s[ch], 0.5F, (float) size - 0.5F) ); + } + return; + default: + assert(0); + } +} + + +/** + * For RECT textures / unnormalized texcoords. + * Only a subset of wrap modes supported. + */ +static INLINE void +linear_texcoord_unnorm_4(unsigned wrapMode, const float s[4], unsigned size, + int icoord0[4], int icoord1[4], float w[4]) +{ + uint ch; + switch (wrapMode) { + case PIPE_TEX_WRAP_CLAMP: + for (ch = 0; ch < 4; ch++) { + /* Not exactly what the spec says, but it matches NVIDIA output */ + float u = CLAMP(s[ch] - 0.5F, 0.0f, (float) size - 1.0f); + icoord0[ch] = util_ifloor(u); + icoord1[ch] = icoord0[ch] + 1; + w[ch] = FRAC(u); + } + return; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + /* fall-through */ + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + for (ch = 0; ch < 4; ch++) { + float u = CLAMP(s[ch], 0.5F, (float) size - 0.5F); + u -= 0.5F; + icoord0[ch] = util_ifloor(u); + icoord1[ch] = icoord0[ch] + 1; + if (icoord1[ch] > (int) size - 1) + icoord1[ch] = size - 1; + w[ch] = FRAC(u); + } + break; + default: + assert(0); + } +} + + +static unsigned +choose_cube_face(float rx, float ry, float rz, float *newS, float *newT) +{ + /* + major axis + direction target sc tc ma + ---------- ------------------------------- --- --- --- + +rx TEXTURE_CUBE_MAP_POSITIVE_X_EXT -rz -ry rx + -rx TEXTURE_CUBE_MAP_NEGATIVE_X_EXT +rz -ry rx + +ry TEXTURE_CUBE_MAP_POSITIVE_Y_EXT +rx +rz ry + -ry TEXTURE_CUBE_MAP_NEGATIVE_Y_EXT +rx -rz ry + +rz TEXTURE_CUBE_MAP_POSITIVE_Z_EXT +rx -ry rz + -rz TEXTURE_CUBE_MAP_NEGATIVE_Z_EXT -rx -ry rz + */ + const float arx = fabsf(rx), ary = fabsf(ry), arz = fabsf(rz); + unsigned face; + float sc, tc, ma; + + if (arx > ary && arx > arz) { + if (rx >= 0.0F) { + face = PIPE_TEX_FACE_POS_X; + sc = -rz; + tc = -ry; + ma = arx; + } + else { + face = PIPE_TEX_FACE_NEG_X; + sc = rz; + tc = -ry; + ma = arx; + } + } + else if (ary > arx && ary > arz) { + if (ry >= 0.0F) { + face = PIPE_TEX_FACE_POS_Y; + sc = rx; + tc = rz; + ma = ary; + } + else { + face = PIPE_TEX_FACE_NEG_Y; + sc = rx; + tc = -rz; + ma = ary; + } + } + else { + if (rz > 0.0F) { + face = PIPE_TEX_FACE_POS_Z; + sc = rx; + tc = -ry; + ma = arz; + } + else { + face = PIPE_TEX_FACE_NEG_Z; + sc = -rx; + tc = -ry; + ma = arz; + } + } + + *newS = ( sc / ma + 1.0F ) * 0.5F; + *newT = ( tc / ma + 1.0F ) * 0.5F; + + return face; +} + + +/** + * Examine the quad's texture coordinates to compute the partial + * derivatives w.r.t X and Y, then compute lambda (level of detail). + * + * This is only done for fragment shaders, not vertex shaders. + */ +static float +compute_lambda(struct tgsi_sampler *tgsi_sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias) +{ + const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler); + const struct pipe_texture *texture = samp->texture; + const struct pipe_sampler_state *sampler = samp->sampler; + float rho, lambda; + + if (samp->processor == TGSI_PROCESSOR_VERTEX) + return lodbias; + + assert(sampler->normalized_coords); + + assert(s); + { + float dsdx = s[QUAD_BOTTOM_RIGHT] - s[QUAD_BOTTOM_LEFT]; + float dsdy = s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT]; + dsdx = fabsf(dsdx); + dsdy = fabsf(dsdy); + rho = MAX2(dsdx, dsdy) * texture->width[0]; + } + if (t) { + float dtdx = t[QUAD_BOTTOM_RIGHT] - t[QUAD_BOTTOM_LEFT]; + float dtdy = t[QUAD_TOP_LEFT] - t[QUAD_BOTTOM_LEFT]; + float max; + dtdx = fabsf(dtdx); + dtdy = fabsf(dtdy); + max = MAX2(dtdx, dtdy) * texture->height[0]; + rho = MAX2(rho, max); + } + if (p) { + float dpdx = p[QUAD_BOTTOM_RIGHT] - p[QUAD_BOTTOM_LEFT]; + float dpdy = p[QUAD_TOP_LEFT] - p[QUAD_BOTTOM_LEFT]; + float max; + dpdx = fabsf(dpdx); + dpdy = fabsf(dpdy); + max = MAX2(dpdx, dpdy) * texture->depth[0]; + rho = MAX2(rho, max); + } + + lambda = util_fast_log2(rho); + lambda += lodbias + sampler->lod_bias; + lambda = CLAMP(lambda, sampler->min_lod, sampler->max_lod); + + return lambda; +} + + +/** + * Do several things here: + * 1. Compute lambda from the texcoords, if needed + * 2. Determine if we're minifying or magnifying + * 3. If minifying, choose mipmap levels + * 4. Return image filter to use within mipmap images + * \param level0 Returns first mipmap level to sample from + * \param level1 Returns second mipmap level to sample from + * \param levelBlend Returns blend factor between levels, in [0,1] + * \param imgFilter Returns either the min or mag filter, depending on lambda + */ +static void +choose_mipmap_levels(struct tgsi_sampler *tgsi_sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + unsigned *level0, unsigned *level1, float *levelBlend, + unsigned *imgFilter) +{ + const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler); + const struct pipe_texture *texture = samp->texture; + const struct pipe_sampler_state *sampler = samp->sampler; + + if (sampler->min_mip_filter == PIPE_TEX_MIPFILTER_NONE) { + /* no mipmap selection needed */ + *level0 = *level1 = CLAMP((int) sampler->min_lod, + 0, (int) texture->last_level); + + if (sampler->min_img_filter != sampler->mag_img_filter) { + /* non-mipmapped texture, but still need to determine if doing + * minification or magnification. + */ + float lambda = compute_lambda(tgsi_sampler, s, t, p, lodbias); + if (lambda <= 0.0) { + *imgFilter = sampler->mag_img_filter; + } + else { + *imgFilter = sampler->min_img_filter; + } + } + else { + *imgFilter = sampler->mag_img_filter; + } + } + else { + float lambda = compute_lambda(tgsi_sampler, s, t, p, lodbias); + + if (lambda <= 0.0) { /* XXX threshold depends on the filter */ + /* magnifying */ + *imgFilter = sampler->mag_img_filter; + *level0 = *level1 = 0; + } + else { + /* minifying */ + *imgFilter = sampler->min_img_filter; + + /* choose mipmap level(s) and compute the blend factor between them */ + if (sampler->min_mip_filter == PIPE_TEX_MIPFILTER_NEAREST) { + /* Nearest mipmap level */ + const int lvl = (int) (lambda + 0.5); + *level0 = + *level1 = CLAMP(lvl, 0, (int) texture->last_level); + } + else { + /* Linear interpolation between mipmap levels */ + const int lvl = (int) lambda; + *level0 = CLAMP(lvl, 0, (int) texture->last_level); + *level1 = CLAMP(lvl + 1, 0, (int) texture->last_level); + *levelBlend = FRAC(lambda); /* blending weight between levels */ + } + } + } +} + + +/** + * Get a texel from a texture, using the texture tile cache. + * + * \param face the cube face in 0..5 + * \param level the mipmap level + * \param x the x coord of texel within 2D image + * \param y the y coord of texel within 2D image + * \param z which slice of a 3D texture + * \param rgba the quad to put the texel/color into + * \param j which element of the rgba quad to write to + * + * XXX maybe move this into lp_tile_cache.c and merge with the + * lp_get_cached_tile_tex() function. Also, get 4 texels instead of 1... + */ +static void +get_texel_quad_2d(const struct tgsi_sampler *tgsi_sampler, + unsigned face, unsigned level, int x, int y, + const uint8_t *out[4]) +{ + const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler); + + const struct llvmpipe_cached_tex_tile *tile + = lp_get_cached_tex_tile(samp->cache, + tex_tile_address(x, y, 0, face, level)); + + y %= TEX_TILE_SIZE; + x %= TEX_TILE_SIZE; + + out[0] = &tile->color[y ][x ][0]; + out[1] = &tile->color[y ][x+1][0]; + out[2] = &tile->color[y+1][x ][0]; + out[3] = &tile->color[y+1][x+1][0]; +} + +static INLINE const uint8_t * +get_texel_2d_ptr(const struct tgsi_sampler *tgsi_sampler, + unsigned face, unsigned level, int x, int y) +{ + const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler); + + const struct llvmpipe_cached_tex_tile *tile + = lp_get_cached_tex_tile(samp->cache, + tex_tile_address(x, y, 0, face, level)); + + y %= TEX_TILE_SIZE; + x %= TEX_TILE_SIZE; + + return &tile->color[y][x][0]; +} + + +static void +get_texel_quad_2d_mt(const struct tgsi_sampler *tgsi_sampler, + unsigned face, unsigned level, + int x0, int y0, + int x1, int y1, + const uint8_t *out[4]) +{ + unsigned i; + + for (i = 0; i < 4; i++) { + unsigned tx = (i & 1) ? x1 : x0; + unsigned ty = (i >> 1) ? y1 : y0; + + out[i] = get_texel_2d_ptr( tgsi_sampler, face, level, tx, ty ); + } +} + +static void +get_texel(const struct tgsi_sampler *tgsi_sampler, + unsigned face, unsigned level, int x, int y, int z, + float rgba[NUM_CHANNELS][QUAD_SIZE], unsigned j) +{ + const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler); + const struct pipe_texture *texture = samp->texture; + const struct pipe_sampler_state *sampler = samp->sampler; + + if (x < 0 || x >= (int) texture->width[level] || + y < 0 || y >= (int) texture->height[level] || + z < 0 || z >= (int) texture->depth[level]) { + rgba[0][j] = sampler->border_color[0]; + rgba[1][j] = sampler->border_color[1]; + rgba[2][j] = sampler->border_color[2]; + rgba[3][j] = sampler->border_color[3]; + } + else { + const unsigned tx = x % TEX_TILE_SIZE; + const unsigned ty = y % TEX_TILE_SIZE; + const struct llvmpipe_cached_tex_tile *tile; + + tile = lp_get_cached_tex_tile(samp->cache, + tex_tile_address(x, y, z, face, level)); + + rgba[0][j] = ubyte_to_float(tile->color[ty][tx][0]); + rgba[1][j] = ubyte_to_float(tile->color[ty][tx][1]); + rgba[2][j] = ubyte_to_float(tile->color[ty][tx][2]); + rgba[3][j] = ubyte_to_float(tile->color[ty][tx][3]); + if (0) + { + debug_printf("Get texel %f %f %f %f from %s\n", + rgba[0][j], rgba[1][j], rgba[2][j], rgba[3][j], + pf_name(texture->format)); + } + } +} + + +/** + * Compare texcoord 'p' (aka R) against texture value 'rgba[0]' + * When we sampled the depth texture, the depth value was put into all + * RGBA channels. We look at the red channel here. + * \param rgba quad of (depth) texel values + * \param p texture 'P' components for four pixels in quad + * \param j which pixel in the quad to test [0..3] + */ +static INLINE void +shadow_compare(const struct pipe_sampler_state *sampler, + float rgba[NUM_CHANNELS][QUAD_SIZE], + const float p[QUAD_SIZE], + uint j) +{ + int k; + switch (sampler->compare_func) { + case PIPE_FUNC_LESS: + k = p[j] < rgba[0][j]; + break; + case PIPE_FUNC_LEQUAL: + k = p[j] <= rgba[0][j]; + break; + case PIPE_FUNC_GREATER: + k = p[j] > rgba[0][j]; + break; + case PIPE_FUNC_GEQUAL: + k = p[j] >= rgba[0][j]; + break; + case PIPE_FUNC_EQUAL: + k = p[j] == rgba[0][j]; + break; + case PIPE_FUNC_NOTEQUAL: + k = p[j] != rgba[0][j]; + break; + case PIPE_FUNC_ALWAYS: + k = 1; + break; + case PIPE_FUNC_NEVER: + k = 0; + break; + default: + k = 0; + assert(0); + break; + } + + /* XXX returning result for default GL_DEPTH_TEXTURE_MODE = GL_LUMINANCE */ + rgba[0][j] = rgba[1][j] = rgba[2][j] = (float) k; + rgba[3][j] = 1.0F; +} + + +/** + * As above, but do four z/texture comparisons. + */ +static INLINE void +shadow_compare4(const struct pipe_sampler_state *sampler, + float rgba[NUM_CHANNELS][QUAD_SIZE], + const float p[QUAD_SIZE]) +{ + int j, k0, k1, k2, k3; + float val; + + /* compare four texcoords vs. four texture samples */ + switch (sampler->compare_func) { + case PIPE_FUNC_LESS: + k0 = p[0] < rgba[0][0]; + k1 = p[1] < rgba[0][1]; + k2 = p[2] < rgba[0][2]; + k3 = p[3] < rgba[0][3]; + break; + case PIPE_FUNC_LEQUAL: + k0 = p[0] <= rgba[0][0]; + k1 = p[1] <= rgba[0][1]; + k2 = p[2] <= rgba[0][2]; + k3 = p[3] <= rgba[0][3]; + break; + case PIPE_FUNC_GREATER: + k0 = p[0] > rgba[0][0]; + k1 = p[1] > rgba[0][1]; + k2 = p[2] > rgba[0][2]; + k3 = p[3] > rgba[0][3]; + break; + case PIPE_FUNC_GEQUAL: + k0 = p[0] >= rgba[0][0]; + k1 = p[1] >= rgba[0][1]; + k2 = p[2] >= rgba[0][2]; + k3 = p[3] >= rgba[0][3]; + break; + case PIPE_FUNC_EQUAL: + k0 = p[0] == rgba[0][0]; + k1 = p[1] == rgba[0][1]; + k2 = p[2] == rgba[0][2]; + k3 = p[3] == rgba[0][3]; + break; + case PIPE_FUNC_NOTEQUAL: + k0 = p[0] != rgba[0][0]; + k1 = p[1] != rgba[0][1]; + k2 = p[2] != rgba[0][2]; + k3 = p[3] != rgba[0][3]; + break; + case PIPE_FUNC_ALWAYS: + k0 = k1 = k2 = k3 = 1; + break; + case PIPE_FUNC_NEVER: + k0 = k1 = k2 = k3 = 0; + break; + default: + k0 = k1 = k2 = k3 = 0; + assert(0); + break; + } + + /* convert four pass/fail values to an intensity in [0,1] */ + val = 0.25F * (k0 + k1 + k2 + k3); + + /* XXX returning result for default GL_DEPTH_TEXTURE_MODE = GL_LUMINANCE */ + for (j = 0; j < 4; j++) { + rgba[0][j] = rgba[1][j] = rgba[2][j] = val; + rgba[3][j] = 1.0F; + } +} + + + +static void +lp_get_samples_2d_linear_repeat_POT(struct tgsi_sampler *tgsi_sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler); + unsigned j; + unsigned level = samp->level; + unsigned xpot = 1 << (samp->xpot - level); + unsigned ypot = 1 << (samp->ypot - level); + unsigned xmax = (xpot - 1) & (TEX_TILE_SIZE - 1); /* MIN2(TEX_TILE_SIZE, xpot) - 1; */ + unsigned ymax = (ypot - 1) & (TEX_TILE_SIZE - 1); /* MIN2(TEX_TILE_SIZE, ypot) - 1; */ + + for (j = 0; j < QUAD_SIZE; j++) { + int c; + + float u = s[j] * xpot - 0.5F; + float v = t[j] * ypot - 0.5F; + + int uflr = util_ifloor(u); + int vflr = util_ifloor(v); + + float xw = u - (float)uflr; + float yw = v - (float)vflr; + + int x0 = uflr & (xpot - 1); + int y0 = vflr & (ypot - 1); + + const uint8_t *tx[4]; + + + /* Can we fetch all four at once: + */ + if (x0 < xmax && y0 < ymax) + { + get_texel_quad_2d(tgsi_sampler, 0, level, x0, y0, tx); + } + else + { + unsigned x1 = (x0 + 1) & (xpot - 1); + unsigned y1 = (y0 + 1) & (ypot - 1); + get_texel_quad_2d_mt(tgsi_sampler, 0, level, + x0, y0, x1, y1, tx); + } + + + /* interpolate R, G, B, A */ + for (c = 0; c < 4; c++) { + rgba[c][j] = lerp_2d(xw, yw, + ubyte_to_float(tx[0][c]), ubyte_to_float(tx[1][c]), + ubyte_to_float(tx[2][c]), ubyte_to_float(tx[3][c])); + } + } +} + + +static void +lp_get_samples_2d_nearest_repeat_POT(struct tgsi_sampler *tgsi_sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler); + unsigned j; + unsigned level = samp->level; + unsigned xpot = 1 << (samp->xpot - level); + unsigned ypot = 1 << (samp->ypot - level); + + for (j = 0; j < QUAD_SIZE; j++) { + int c; + + float u = s[j] * xpot; + float v = t[j] * ypot; + + int uflr = util_ifloor(u); + int vflr = util_ifloor(v); + + int x0 = uflr & (xpot - 1); + int y0 = vflr & (ypot - 1); + + const uint8_t *out = get_texel_2d_ptr(tgsi_sampler, 0, level, x0, y0); + + for (c = 0; c < 4; c++) { + rgba[c][j] = ubyte_to_float(out[c]); + } + } +} + + +static void +lp_get_samples_2d_nearest_clamp_POT(struct tgsi_sampler *tgsi_sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler); + unsigned j; + unsigned level = samp->level; + unsigned xpot = 1 << (samp->xpot - level); + unsigned ypot = 1 << (samp->ypot - level); + + for (j = 0; j < QUAD_SIZE; j++) { + int c; + + float u = s[j] * xpot; + float v = t[j] * ypot; + + int x0, y0; + const uint8_t *out; + + x0 = util_ifloor(u); + if (x0 < 0) + x0 = 0; + else if (x0 > xpot - 1) + x0 = xpot - 1; + + y0 = util_ifloor(v); + if (y0 < 0) + y0 = 0; + else if (y0 > ypot - 1) + y0 = ypot - 1; + + out = get_texel_2d_ptr(tgsi_sampler, 0, level, x0, y0); + + for (c = 0; c < 4; c++) { + rgba[c][j] = ubyte_to_float(out[c]); + } + } +} + + +static void +lp_get_samples_2d_linear_mip_linear_repeat_POT(struct tgsi_sampler *tgsi_sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler); + const struct pipe_texture *texture = samp->texture; + int level0; + float lambda; + + lambda = compute_lambda(tgsi_sampler, s, t, p, lodbias); + level0 = (int)lambda; + + if (lambda < 0.0) { + samp->level = 0; + lp_get_samples_2d_linear_repeat_POT( tgsi_sampler, + s, t, p, 0, rgba ); + } + else if (level0 >= texture->last_level) { + samp->level = texture->last_level; + lp_get_samples_2d_linear_repeat_POT( tgsi_sampler, + s, t, p, 0, rgba ); + } + else { + float levelBlend = lambda - level0; + float rgba0[4][4]; + float rgba1[4][4]; + int c,j; + + samp->level = level0; + lp_get_samples_2d_linear_repeat_POT( tgsi_sampler, + s, t, p, 0, rgba0 ); + + samp->level = level0+1; + lp_get_samples_2d_linear_repeat_POT( tgsi_sampler, + s, t, p, 0, rgba1 ); + + for (j = 0; j < QUAD_SIZE; j++) { + for (c = 0; c < 4; c++) { + rgba[c][j] = lerp(levelBlend, rgba0[c][j], rgba1[c][j]); + } + } + } +} + +/** + * Common code for sampling 1D/2D/cube textures. + * Could probably extend for 3D... + */ +static void +lp_get_samples_2d_common(struct tgsi_sampler *tgsi_sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE], + const unsigned faces[4]) +{ + const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler); + const struct pipe_texture *texture = samp->texture; + const struct pipe_sampler_state *sampler = samp->sampler; + unsigned level0, level1, j, imgFilter; + int width, height; + float levelBlend = 0.0F; + + choose_mipmap_levels(tgsi_sampler, s, t, p, + lodbias, + &level0, &level1, &levelBlend, &imgFilter); + + assert(sampler->normalized_coords); + + width = texture->width[level0]; + height = texture->height[level0]; + + assert(width > 0); + + switch (imgFilter) { + case PIPE_TEX_FILTER_NEAREST: + { + int x[4], y[4]; + nearest_texcoord_4(sampler->wrap_s, s, width, x); + nearest_texcoord_4(sampler->wrap_t, t, height, y); + + for (j = 0; j < QUAD_SIZE; j++) { + get_texel(tgsi_sampler, faces[j], level0, x[j], y[j], 0, rgba, j); + if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + shadow_compare(sampler, rgba, p, j); + } + + if (level0 != level1) { + /* get texels from second mipmap level and blend */ + float rgba2[4][4]; + unsigned c; + x[j] /= 2; + y[j] /= 2; + get_texel(tgsi_sampler, faces[j], level1, x[j], y[j], 0, + rgba2, j); + if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){ + shadow_compare(sampler, rgba2, p, j); + } + + for (c = 0; c < NUM_CHANNELS; c++) { + rgba[c][j] = lerp(levelBlend, rgba[c][j], rgba2[c][j]); + } + } + } + } + break; + case PIPE_TEX_FILTER_LINEAR: + case PIPE_TEX_FILTER_ANISO: + { + int x0[4], y0[4], x1[4], y1[4]; + float xw[4], yw[4]; /* weights */ + + linear_texcoord_4(sampler->wrap_s, s, width, x0, x1, xw); + linear_texcoord_4(sampler->wrap_t, t, height, y0, y1, yw); + + for (j = 0; j < QUAD_SIZE; j++) { + float tx[4][4]; /* texels */ + int c; + get_texel(tgsi_sampler, faces[j], level0, x0[j], y0[j], 0, tx, 0); + get_texel(tgsi_sampler, faces[j], level0, x1[j], y0[j], 0, tx, 1); + get_texel(tgsi_sampler, faces[j], level0, x0[j], y1[j], 0, tx, 2); + get_texel(tgsi_sampler, faces[j], level0, x1[j], y1[j], 0, tx, 3); + if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + shadow_compare4(sampler, tx, p); + } + + /* interpolate R, G, B, A */ + for (c = 0; c < 4; c++) { + rgba[c][j] = lerp_2d(xw[j], yw[j], + tx[c][0], tx[c][1], + tx[c][2], tx[c][3]); + } + + if (level0 != level1) { + /* get texels from second mipmap level and blend */ + float rgba2[4][4]; + + /* XXX: This is incorrect -- will often end up with (x0 + * == x1 && y0 == y1), meaning that we fetch the same + * texel four times and linearly interpolate between + * identical values. The correct approach would be to + * call linear_texcoord again for the second level. + */ + x0[j] /= 2; + y0[j] /= 2; + x1[j] /= 2; + y1[j] /= 2; + get_texel(tgsi_sampler, faces[j], level1, x0[j], y0[j], 0, tx, 0); + get_texel(tgsi_sampler, faces[j], level1, x1[j], y0[j], 0, tx, 1); + get_texel(tgsi_sampler, faces[j], level1, x0[j], y1[j], 0, tx, 2); + get_texel(tgsi_sampler, faces[j], level1, x1[j], y1[j], 0, tx, 3); + if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){ + shadow_compare4(sampler, tx, p); + } + + /* interpolate R, G, B, A */ + for (c = 0; c < 4; c++) { + rgba2[c][j] = lerp_2d(xw[j], yw[j], + tx[c][0], tx[c][1], tx[c][2], tx[c][3]); + } + + for (c = 0; c < NUM_CHANNELS; c++) { + rgba[c][j] = lerp(levelBlend, rgba[c][j], rgba2[c][j]); + } + } + } + } + break; + default: + assert(0); + } +} + + +static INLINE void +lp_get_samples_1d(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + static const unsigned faces[4] = {0, 0, 0, 0}; + static const float tzero[4] = {0, 0, 0, 0}; + lp_get_samples_2d_common(sampler, s, tzero, NULL, + lodbias, rgba, faces); +} + + +static INLINE void +lp_get_samples_2d(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + static const unsigned faces[4] = {0, 0, 0, 0}; + lp_get_samples_2d_common(sampler, s, t, p, + lodbias, rgba, faces); +} + + +static INLINE void +lp_get_samples_3d(struct tgsi_sampler *tgsi_sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler); + const struct pipe_texture *texture = samp->texture; + const struct pipe_sampler_state *sampler = samp->sampler; + /* get/map pipe_surfaces corresponding to 3D tex slices */ + unsigned level0, level1, j, imgFilter; + int width, height, depth; + float levelBlend; + const uint face = 0; + + choose_mipmap_levels(tgsi_sampler, s, t, p, + lodbias, + &level0, &level1, &levelBlend, &imgFilter); + + assert(sampler->normalized_coords); + + width = texture->width[level0]; + height = texture->height[level0]; + depth = texture->depth[level0]; + + assert(width > 0); + assert(height > 0); + assert(depth > 0); + + switch (imgFilter) { + case PIPE_TEX_FILTER_NEAREST: + { + int x[4], y[4], z[4]; + nearest_texcoord_4(sampler->wrap_s, s, width, x); + nearest_texcoord_4(sampler->wrap_t, t, height, y); + nearest_texcoord_4(sampler->wrap_r, p, depth, z); + for (j = 0; j < QUAD_SIZE; j++) { + get_texel(tgsi_sampler, face, level0, x[j], y[j], z[j], rgba, j); + if (level0 != level1) { + /* get texels from second mipmap level and blend */ + float rgba2[4][4]; + unsigned c; + x[j] /= 2; + y[j] /= 2; + z[j] /= 2; + get_texel(tgsi_sampler, face, level1, x[j], y[j], z[j], rgba2, j); + for (c = 0; c < NUM_CHANNELS; c++) { + rgba[c][j] = lerp(levelBlend, rgba2[c][j], rgba[c][j]); + } + } + } + } + break; + case PIPE_TEX_FILTER_LINEAR: + case PIPE_TEX_FILTER_ANISO: + { + int x0[4], x1[4], y0[4], y1[4], z0[4], z1[4]; + float xw[4], yw[4], zw[4]; /* interpolation weights */ + linear_texcoord_4(sampler->wrap_s, s, width, x0, x1, xw); + linear_texcoord_4(sampler->wrap_t, t, height, y0, y1, yw); + linear_texcoord_4(sampler->wrap_r, p, depth, z0, z1, zw); + + for (j = 0; j < QUAD_SIZE; j++) { + int c; + float tx0[4][4], tx1[4][4]; + get_texel(tgsi_sampler, face, level0, x0[j], y0[j], z0[j], tx0, 0); + get_texel(tgsi_sampler, face, level0, x1[j], y0[j], z0[j], tx0, 1); + get_texel(tgsi_sampler, face, level0, x0[j], y1[j], z0[j], tx0, 2); + get_texel(tgsi_sampler, face, level0, x1[j], y1[j], z0[j], tx0, 3); + get_texel(tgsi_sampler, face, level0, x0[j], y0[j], z1[j], tx1, 0); + get_texel(tgsi_sampler, face, level0, x1[j], y0[j], z1[j], tx1, 1); + get_texel(tgsi_sampler, face, level0, x0[j], y1[j], z1[j], tx1, 2); + get_texel(tgsi_sampler, face, level0, x1[j], y1[j], z1[j], tx1, 3); + + /* interpolate R, G, B, A */ + for (c = 0; c < 4; c++) { + rgba[c][j] = lerp_3d(xw[j], yw[j], zw[j], + tx0[c][0], tx0[c][1], + tx0[c][2], tx0[c][3], + tx1[c][0], tx1[c][1], + tx1[c][2], tx1[c][3]); + } + + if (level0 != level1) { + /* get texels from second mipmap level and blend */ + float rgba2[4][4]; + x0[j] /= 2; + y0[j] /= 2; + z0[j] /= 2; + x1[j] /= 2; + y1[j] /= 2; + z1[j] /= 2; + get_texel(tgsi_sampler, face, level1, x0[j], y0[j], z0[j], tx0, 0); + get_texel(tgsi_sampler, face, level1, x1[j], y0[j], z0[j], tx0, 1); + get_texel(tgsi_sampler, face, level1, x0[j], y1[j], z0[j], tx0, 2); + get_texel(tgsi_sampler, face, level1, x1[j], y1[j], z0[j], tx0, 3); + get_texel(tgsi_sampler, face, level1, x0[j], y0[j], z1[j], tx1, 0); + get_texel(tgsi_sampler, face, level1, x1[j], y0[j], z1[j], tx1, 1); + get_texel(tgsi_sampler, face, level1, x0[j], y1[j], z1[j], tx1, 2); + get_texel(tgsi_sampler, face, level1, x1[j], y1[j], z1[j], tx1, 3); + + /* interpolate R, G, B, A */ + for (c = 0; c < 4; c++) { + rgba2[c][j] = lerp_3d(xw[j], yw[j], zw[j], + tx0[c][0], tx0[c][1], + tx0[c][2], tx0[c][3], + tx1[c][0], tx1[c][1], + tx1[c][2], tx1[c][3]); + } + + /* blend mipmap levels */ + for (c = 0; c < NUM_CHANNELS; c++) { + rgba[c][j] = lerp(levelBlend, rgba[c][j], rgba2[c][j]); + } + } + } + } + break; + default: + assert(0); + } +} + + +static void +lp_get_samples_cube(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + unsigned faces[QUAD_SIZE], j; + float ssss[4], tttt[4]; + for (j = 0; j < QUAD_SIZE; j++) { + faces[j] = choose_cube_face(s[j], t[j], p[j], ssss + j, tttt + j); + } + lp_get_samples_2d_common(sampler, ssss, tttt, NULL, + lodbias, rgba, faces); +} + + +static void +lp_get_samples_rect(struct tgsi_sampler *tgsi_sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler); + const struct pipe_texture *texture = samp->texture; + const struct pipe_sampler_state *sampler = samp->sampler; + const uint face = 0; + unsigned level0, level1, j, imgFilter; + int width, height; + float levelBlend; + + choose_mipmap_levels(tgsi_sampler, s, t, p, + lodbias, + &level0, &level1, &levelBlend, &imgFilter); + + /* texture RECTS cannot be mipmapped */ + assert(level0 == level1); + + width = texture->width[level0]; + height = texture->height[level0]; + + assert(width > 0); + + switch (imgFilter) { + case PIPE_TEX_FILTER_NEAREST: + { + int x[4], y[4]; + nearest_texcoord_unnorm_4(sampler->wrap_s, s, width, x); + nearest_texcoord_unnorm_4(sampler->wrap_t, t, height, y); + for (j = 0; j < QUAD_SIZE; j++) { + get_texel(tgsi_sampler, face, level0, x[j], y[j], 0, rgba, j); + if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + shadow_compare(sampler, rgba, p, j); + } + } + } + break; + case PIPE_TEX_FILTER_LINEAR: + case PIPE_TEX_FILTER_ANISO: + { + int x0[4], y0[4], x1[4], y1[4]; + float xw[4], yw[4]; /* weights */ + linear_texcoord_unnorm_4(sampler->wrap_s, s, width, x0, x1, xw); + linear_texcoord_unnorm_4(sampler->wrap_t, t, height, y0, y1, yw); + for (j = 0; j < QUAD_SIZE; j++) { + float tx[4][4]; /* texels */ + int c; + get_texel(tgsi_sampler, face, level0, x0[j], y0[j], 0, tx, 0); + get_texel(tgsi_sampler, face, level0, x1[j], y0[j], 0, tx, 1); + get_texel(tgsi_sampler, face, level0, x0[j], y1[j], 0, tx, 2); + get_texel(tgsi_sampler, face, level0, x1[j], y1[j], 0, tx, 3); + if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + shadow_compare4(sampler, tx, p); + } + for (c = 0; c < 4; c++) { + rgba[c][j] = lerp_2d(xw[j], yw[j], + tx[c][0], tx[c][1], tx[c][2], tx[c][3]); + } + } + } + break; + default: + assert(0); + } +} + + +/** + * Error condition handler + */ +static INLINE void +lp_get_samples_null(struct tgsi_sampler *tgsi_sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + int i,j; + + for (i = 0; i < 4; i++) + for (j = 0; j < 4; j++) + rgba[i][j] = 1.0; +} + +/** + * Called via tgsi_sampler::get_samples() when using a sampler for the + * first time. Determine the actual sampler function, link it in and + * call it. + */ +void +lp_get_samples(struct tgsi_sampler *tgsi_sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler); + const struct pipe_texture *texture = samp->texture; + const struct pipe_sampler_state *sampler = samp->sampler; + + /* Default to the 'undefined' case: + */ + tgsi_sampler->get_samples = lp_get_samples_null; + + if (!texture) { + assert(0); /* is this legal?? */ + goto out; + } + + if (!sampler->normalized_coords) { + assert (texture->target == PIPE_TEXTURE_2D); + tgsi_sampler->get_samples = lp_get_samples_rect; + goto out; + } + + switch (texture->target) { + case PIPE_TEXTURE_1D: + tgsi_sampler->get_samples = lp_get_samples_1d; + break; + case PIPE_TEXTURE_2D: + tgsi_sampler->get_samples = lp_get_samples_2d; + break; + case PIPE_TEXTURE_3D: + tgsi_sampler->get_samples = lp_get_samples_3d; + break; + case PIPE_TEXTURE_CUBE: + tgsi_sampler->get_samples = lp_get_samples_cube; + break; + default: + assert(0); + break; + } + + /* Do this elsewhere: + */ + samp->xpot = util_unsigned_logbase2( samp->texture->width[0] ); + samp->ypot = util_unsigned_logbase2( samp->texture->height[0] ); + + /* Try to hook in a faster sampler. Ultimately we'll have to + * code-generate these. Luckily most of this looks like it is + * orthogonal state within the sampler. + */ + if (texture->target == PIPE_TEXTURE_2D && + sampler->min_img_filter == sampler->mag_img_filter && + sampler->wrap_s == sampler->wrap_t && + sampler->compare_mode == FALSE && + sampler->normalized_coords) + { + if (sampler->min_mip_filter == PIPE_TEX_MIPFILTER_NONE) { + samp->level = CLAMP((int) sampler->min_lod, + 0, (int) texture->last_level); + + if (sampler->wrap_s == PIPE_TEX_WRAP_REPEAT) { + switch (sampler->min_img_filter) { + case PIPE_TEX_FILTER_NEAREST: + tgsi_sampler->get_samples = lp_get_samples_2d_nearest_repeat_POT; + break; + case PIPE_TEX_FILTER_LINEAR: + tgsi_sampler->get_samples = lp_get_samples_2d_linear_repeat_POT; + break; + default: + break; + } + } + else if (sampler->wrap_s == PIPE_TEX_WRAP_CLAMP) { + switch (sampler->min_img_filter) { + case PIPE_TEX_FILTER_NEAREST: + tgsi_sampler->get_samples = lp_get_samples_2d_nearest_clamp_POT; + break; + default: + break; + } + } + } + else if (sampler->min_mip_filter == PIPE_TEX_MIPFILTER_LINEAR) { + if (sampler->wrap_s == PIPE_TEX_WRAP_REPEAT) { + switch (sampler->min_img_filter) { + case PIPE_TEX_FILTER_LINEAR: + tgsi_sampler->get_samples = lp_get_samples_2d_linear_mip_linear_repeat_POT; + break; + default: + break; + } + } + } + } + else if (0) { + _debug_printf("target %d/%d min_mip %d/%d min_img %d/%d wrap %d/%d compare %d/%d norm %d/%d\n", + texture->target, PIPE_TEXTURE_2D, + sampler->min_mip_filter, PIPE_TEX_MIPFILTER_NONE, + sampler->min_img_filter, sampler->mag_img_filter, + sampler->wrap_s, sampler->wrap_t, + sampler->compare_mode, FALSE, + sampler->normalized_coords, TRUE); + } + +out: + tgsi_sampler->get_samples( tgsi_sampler, s, t, p, lodbias, rgba ); +} + + +void PIPE_CDECL +lp_fetch_texel_soa( struct tgsi_sampler **samplers, + uint32_t unit, + float *store ) +{ + struct tgsi_sampler *sampler = samplers[unit]; + +#if 0 + uint j; + + debug_printf("%s sampler: %p (%p) store: %p\n", + __FUNCTION__, + sampler, *sampler, + store ); + + debug_printf("lodbias %f\n", store[12]); + + for (j = 0; j < 4; j++) + debug_printf("sample %d texcoord %f %f\n", + j, + store[0+j], + store[4+j]); +#endif + + { + float rgba[NUM_CHANNELS][QUAD_SIZE]; + sampler->get_samples(sampler, + &store[0], + &store[4], + &store[8], + 0.0f, /*store[12], lodbias */ + rgba); + memcpy(store, rgba, sizeof rgba); + } + +#if 0 + for (j = 0; j < 4; j++) + debug_printf("sample %d result %f %f %f %f\n", + j, + store[0+j], + store[4+j], + store[8+j], + store[12+j]); +#endif +} + + +#include "lp_bld_type.h" +#include "lp_bld_intr.h" +#include "lp_bld_tgsi.h" + + +struct lp_c_sampler_soa +{ + struct lp_build_sampler_soa base; + + LLVMValueRef context_ptr; + + LLVMValueRef samplers_ptr; + + /** Coords/texels store */ + LLVMValueRef store_ptr; +}; + + +static void +lp_c_sampler_soa_destroy(struct lp_build_sampler_soa *sampler) +{ + FREE(sampler); +} + + +static void +lp_c_sampler_soa_emit_fetch_texel(struct lp_build_sampler_soa *_sampler, + LLVMBuilderRef builder, + struct lp_type type, + unsigned unit, + unsigned num_coords, + const LLVMValueRef *coords, + LLVMValueRef lodbias, + LLVMValueRef *texel) +{ + struct lp_c_sampler_soa *sampler = (struct lp_c_sampler_soa *)_sampler; + LLVMTypeRef vec_type = LLVMTypeOf(coords[0]); + LLVMValueRef args[3]; + unsigned i; + + if(!sampler->samplers_ptr) + sampler->samplers_ptr = lp_jit_context_samplers(builder, sampler->context_ptr); + + if(!sampler->store_ptr) + sampler->store_ptr = LLVMBuildArrayAlloca(builder, + vec_type, + LLVMConstInt(LLVMInt32Type(), 4, 0), + "texel_store"); + + for (i = 0; i < num_coords; i++) { + LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), i, 0); + LLVMValueRef coord_ptr = LLVMBuildGEP(builder, sampler->store_ptr, &index, 1, ""); + LLVMBuildStore(builder, coords[i], coord_ptr); + } + + args[0] = sampler->samplers_ptr; + args[1] = LLVMConstInt(LLVMInt32Type(), unit, 0); + args[2] = sampler->store_ptr; + + lp_build_intrinsic(builder, "fetch_texel", LLVMVoidType(), args, 3); + + for (i = 0; i < NUM_CHANNELS; ++i) { + LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), i, 0); + LLVMValueRef texel_ptr = LLVMBuildGEP(builder, sampler->store_ptr, &index, 1, ""); + texel[i] = LLVMBuildLoad(builder, texel_ptr, ""); + } +} + + +struct lp_build_sampler_soa * +lp_c_sampler_soa_create(LLVMValueRef context_ptr) +{ + struct lp_c_sampler_soa *sampler; + + sampler = CALLOC_STRUCT(lp_c_sampler_soa); + if(!sampler) + return NULL; + + sampler->base.destroy = lp_c_sampler_soa_destroy; + sampler->base.emit_fetch_texel = lp_c_sampler_soa_emit_fetch_texel; + sampler->context_ptr = context_ptr; + + return &sampler->base; +} + diff --git a/src/gallium/drivers/llvmpipe/lp_tex_sample_llvm.c b/src/gallium/drivers/llvmpipe/lp_tex_sample_llvm.c index d2a6ae21f5..5138ccf7c9 100644 --- a/src/gallium/drivers/llvmpipe/lp_tex_sample_llvm.c +++ b/src/gallium/drivers/llvmpipe/lp_tex_sample_llvm.c @@ -44,7 +44,6 @@ #include "pipe/p_shader_tokens.h" #include "lp_bld_debug.h" #include "lp_bld_type.h" -#include "lp_bld_intr.h" #include "lp_bld_sample.h" #include "lp_bld_tgsi.h" #include "lp_state.h" diff --git a/src/gallium/drivers/llvmpipe/lp_texture.c b/src/gallium/drivers/llvmpipe/lp_texture.c index 2c135029ea..039539d682 100644 --- a/src/gallium/drivers/llvmpipe/lp_texture.c +++ b/src/gallium/drivers/llvmpipe/lp_texture.c @@ -42,7 +42,6 @@ #include "lp_context.h" #include "lp_state.h" #include "lp_texture.h" -#include "lp_tex_cache.h" #include "lp_screen.h" #include "lp_winsys.h" diff --git a/src/gallium/drivers/llvmpipe/lp_tile_cache.c b/src/gallium/drivers/llvmpipe/lp_tile_cache.c index 7a1ecf5107..971d933333 100644 --- a/src/gallium/drivers/llvmpipe/lp_tile_cache.c +++ b/src/gallium/drivers/llvmpipe/lp_tile_cache.c @@ -38,8 +38,6 @@ #include "util/u_tile.h" #include "util/u_rect.h" #include "lp_context.h" -#include "lp_surface.h" -#include "lp_texture.h" #include "lp_tile_soa.h" #include "lp_tile_cache.h" diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 8e01793940..73e075f0d8 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -43,7 +43,6 @@ #include "sp_surface.h" #include "sp_tile_cache.h" #include "sp_tex_tile_cache.h" -#include "sp_texture.h" #include "sp_winsys.h" #include "sp_query.h" diff --git a/src/gallium/drivers/softpipe/sp_flush.c b/src/gallium/drivers/softpipe/sp_flush.c index 75dac810a1..e8952bf4fb 100644 --- a/src/gallium/drivers/softpipe/sp_flush.c +++ b/src/gallium/drivers/softpipe/sp_flush.c @@ -34,11 +34,9 @@ #include "draw/draw_context.h" #include "sp_flush.h" #include "sp_context.h" -#include "sp_surface.h" #include "sp_state.h" #include "sp_tile_cache.h" #include "sp_tex_tile_cache.h" -#include "sp_winsys.h" void diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c index 5812d1eefe..98c08eaffa 100644 --- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c @@ -526,7 +526,8 @@ static void sp_vbuf_destroy(struct vbuf_render *vbr) { struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); - align_free(cvbr->vertex_buffer); + if(cvbr->vertex_buffer) + align_free(cvbr->vertex_buffer); sp_setup_destroy_context(cvbr->setup); FREE(cvbr); } diff --git a/src/gallium/drivers/softpipe/sp_quad_blend.c b/src/gallium/drivers/softpipe/sp_quad_blend.c index d9babe81da..3b8c2d5789 100644 --- a/src/gallium/drivers/softpipe/sp_quad_blend.c +++ b/src/gallium/drivers/softpipe/sp_quad_blend.c @@ -35,7 +35,6 @@ #include "util/u_memory.h" #include "sp_context.h" #include "sp_quad.h" -#include "sp_surface.h" #include "sp_tile_cache.h" #include "sp_quad_pipe.h" diff --git a/src/gallium/drivers/softpipe/sp_quad_depth_test.c b/src/gallium/drivers/softpipe/sp_quad_depth_test.c index 0ca86c4e1c..a981775cbd 100644 --- a/src/gallium/drivers/softpipe/sp_quad_depth_test.c +++ b/src/gallium/drivers/softpipe/sp_quad_depth_test.c @@ -30,11 +30,11 @@ */ #include "pipe/p_defines.h" +#include "util/u_format.h" #include "util/u_memory.h" #include "tgsi/tgsi_scan.h" #include "sp_context.h" #include "sp_quad.h" -#include "sp_surface.h" #include "sp_quad_pipe.h" #include "sp_tile_cache.h" #include "sp_state.h" /* for sp_fragment_shader */ @@ -651,6 +651,20 @@ static unsigned mask_count[16] = +/** helper to get number of Z buffer bits */ +static unsigned +get_depth_bits(struct quad_stage *qs) +{ + struct pipe_surface *zsurf = qs->softpipe->framebuffer.zsbuf; + if (zsurf) + return util_format_get_component_bits(zsurf->format, + UTIL_FORMAT_COLORSPACE_ZS, 0); + else + return 0; +} + + + static void depth_test_quads_fallback(struct quad_stage *qs, struct quad_header *quads[], @@ -666,7 +680,7 @@ depth_test_quads_fallback(struct quad_stage *qs, nr = alpha_test_quads(qs, quads, nr); } - if (qs->softpipe->framebuffer.zsbuf && + if (get_depth_bits(qs) > 0 && (qs->softpipe->depth_stencil->depth.enabled || qs->softpipe->depth_stencil->stencil[0].enabled)) { @@ -884,7 +898,7 @@ choose_depth_test(struct quad_stage *qs, boolean alpha = qs->softpipe->depth_stencil->alpha.enabled; - boolean depth = (qs->softpipe->framebuffer.zsbuf && + boolean depth = (get_depth_bits(qs) > 0 && qs->softpipe->depth_stencil->depth.enabled); unsigned depthfunc = qs->softpipe->depth_stencil->depth.func; @@ -895,7 +909,6 @@ choose_depth_test(struct quad_stage *qs, boolean occlusion = qs->softpipe->active_query_count; - if (!alpha && !depth && !stencil) { diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c index 1e7533d0f9..e799df136e 100644 --- a/src/gallium/drivers/softpipe/sp_quad_fs.c +++ b/src/gallium/drivers/softpipe/sp_quad_fs.c @@ -45,8 +45,6 @@ #include "sp_state.h" #include "sp_quad.h" #include "sp_quad_pipe.h" -#include "sp_texture.h" -#include "sp_tex_sample.h" struct quad_shade_stage diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index 3da75364c5..f6c3a2b38b 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -41,7 +41,6 @@ #include "draw/draw_private.h" #include "draw/draw_vertex.h" #include "pipe/p_shader_tokens.h" -#include "pipe/p_thread.h" #include "util/u_math.h" #include "util/u_memory.h" diff --git a/src/gallium/drivers/softpipe/sp_state_surface.c b/src/gallium/drivers/softpipe/sp_state_surface.c index f6154109ea..3946678219 100644 --- a/src/gallium/drivers/softpipe/sp_state_surface.c +++ b/src/gallium/drivers/softpipe/sp_state_surface.c @@ -30,7 +30,6 @@ #include "sp_context.h" #include "sp_state.h" -#include "sp_surface.h" #include "sp_tile_cache.h" #include "draw/draw_context.h" diff --git a/src/gallium/drivers/softpipe/sp_state_vertex.c b/src/gallium/drivers/softpipe/sp_state_vertex.c index 46b6991195..b491d92ed1 100644 --- a/src/gallium/drivers/softpipe/sp_state_vertex.c +++ b/src/gallium/drivers/softpipe/sp_state_vertex.c @@ -31,7 +31,6 @@ #include "sp_context.h" #include "sp_state.h" -#include "sp_surface.h" #include "draw/draw_context.h" diff --git a/src/gallium/drivers/softpipe/sp_tex_tile_cache.c b/src/gallium/drivers/softpipe/sp_tex_tile_cache.c index e50a76a73b..50242d5bd6 100644 --- a/src/gallium/drivers/softpipe/sp_tex_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tex_tile_cache.c @@ -37,7 +37,6 @@ #include "util/u_tile.h" #include "util/u_math.h" #include "sp_context.h" -#include "sp_surface.h" #include "sp_texture.h" #include "sp_tex_tile_cache.h" diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 45e80c5447..a5fff91507 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -38,7 +38,6 @@ #include "util/u_memory.h" #include "sp_context.h" -#include "sp_state.h" #include "sp_texture.h" #include "sp_screen.h" #include "sp_winsys.h" diff --git a/src/gallium/drivers/svga/svga_draw_arrays.c b/src/gallium/drivers/svga/svga_draw_arrays.c index 75492dffca..6b6ebc9b58 100644 --- a/src/gallium/drivers/svga/svga_draw_arrays.c +++ b/src/gallium/drivers/svga/svga_draw_arrays.c @@ -26,7 +26,6 @@ #include "svga_cmd.h" #include "pipe/p_inlines.h" -#include "util/u_prim.h" #include "indices/u_indices.h" #include "svga_hw_reg.h" diff --git a/src/gallium/drivers/svga/svga_draw_elements.c b/src/gallium/drivers/svga/svga_draw_elements.c index 167d817831..022b444eb9 100644 --- a/src/gallium/drivers/svga/svga_draw_elements.c +++ b/src/gallium/drivers/svga/svga_draw_elements.c @@ -24,7 +24,6 @@ **********************************************************/ #include "pipe/p_inlines.h" -#include "util/u_prim.h" #include "util/u_upload_mgr.h" #include "indices/u_indices.h" diff --git a/src/gallium/drivers/svga/svga_pipe_blend.c b/src/gallium/drivers/svga/svga_pipe_blend.c index 855d228755..3ad3f97816 100644 --- a/src/gallium/drivers/svga/svga_pipe_blend.c +++ b/src/gallium/drivers/svga/svga_pipe_blend.c @@ -29,7 +29,6 @@ #include "util/u_memory.h" #include "svga_context.h" -#include "svga_state.h" #include "svga_hw_reg.h" diff --git a/src/gallium/drivers/svga/svga_pipe_constants.c b/src/gallium/drivers/svga/svga_pipe_constants.c index ca2c7c49d7..93022f3cc5 100644 --- a/src/gallium/drivers/svga/svga_pipe_constants.c +++ b/src/gallium/drivers/svga/svga_pipe_constants.c @@ -30,9 +30,6 @@ #include "tgsi/tgsi_parse.h" #include "svga_context.h" -#include "svga_state.h" -#include "svga_hw_reg.h" -#include "svga_cmd.h" /*********************************************************************** * Constant buffers diff --git a/src/gallium/drivers/svga/svga_pipe_depthstencil.c b/src/gallium/drivers/svga/svga_pipe_depthstencil.c index df636c08a0..34e60cb341 100644 --- a/src/gallium/drivers/svga/svga_pipe_depthstencil.c +++ b/src/gallium/drivers/svga/svga_pipe_depthstencil.c @@ -29,7 +29,6 @@ #include "util/u_memory.h" #include "svga_context.h" -#include "svga_state.h" #include "svga_hw_reg.h" diff --git a/src/gallium/drivers/svga/svga_pipe_draw.c b/src/gallium/drivers/svga/svga_pipe_draw.c index 0f24ef4ee8..4e0c499dc3 100644 --- a/src/gallium/drivers/svga/svga_pipe_draw.c +++ b/src/gallium/drivers/svga/svga_pipe_draw.c @@ -33,7 +33,6 @@ #include "svga_hw_reg.h" #include "svga_context.h" #include "svga_screen.h" -#include "svga_winsys.h" #include "svga_draw.h" #include "svga_state.h" #include "svga_swtnl.h" diff --git a/src/gallium/drivers/svga/svga_pipe_flush.c b/src/gallium/drivers/svga/svga_pipe_flush.c index 0becb0765a..3eb10336c4 100644 --- a/src/gallium/drivers/svga/svga_pipe_flush.c +++ b/src/gallium/drivers/svga/svga_pipe_flush.c @@ -28,13 +28,8 @@ #include "svga_screen_texture.h" #include "svga_context.h" #include "svga_winsys.h" -#include "svga_draw.h" #include "svga_debug.h" -#include "svga_hw_reg.h" - - - static void svga_flush( struct pipe_context *pipe, unsigned flags, diff --git a/src/gallium/drivers/svga/svga_pipe_fs.c b/src/gallium/drivers/svga/svga_pipe_fs.c index 5f1213e46a..32f07fb261 100644 --- a/src/gallium/drivers/svga/svga_pipe_fs.c +++ b/src/gallium/drivers/svga/svga_pipe_fs.c @@ -32,11 +32,9 @@ #include "svga_screen.h" #include "svga_context.h" -#include "svga_state.h" #include "svga_tgsi.h" #include "svga_hw_reg.h" #include "svga_cmd.h" -#include "svga_draw.h" #include "svga_debug.h" diff --git a/src/gallium/drivers/svga/svga_pipe_misc.c b/src/gallium/drivers/svga/svga_pipe_misc.c index 58cb1e6e23..8cf1f2e083 100644 --- a/src/gallium/drivers/svga/svga_pipe_misc.c +++ b/src/gallium/drivers/svga/svga_pipe_misc.c @@ -27,12 +27,6 @@ #include "svga_context.h" #include "svga_screen_texture.h" -#include "svga_state.h" -#include "svga_winsys.h" - -#include "svga_hw_reg.h" - - static void svga_set_scissor_state( struct pipe_context *pipe, diff --git a/src/gallium/drivers/svga/svga_pipe_query.c b/src/gallium/drivers/svga/svga_pipe_query.c index 01336b0a2c..08283e3731 100644 --- a/src/gallium/drivers/svga/svga_pipe_query.c +++ b/src/gallium/drivers/svga/svga_pipe_query.c @@ -32,7 +32,6 @@ #include "svga_screen.h" #include "svga_screen_buffer.h" #include "svga_winsys.h" -#include "svga_draw.h" #include "svga_debug.h" diff --git a/src/gallium/drivers/svga/svga_pipe_rasterizer.c b/src/gallium/drivers/svga/svga_pipe_rasterizer.c index b03f8eb9cf..9ea11aad9a 100644 --- a/src/gallium/drivers/svga/svga_pipe_rasterizer.c +++ b/src/gallium/drivers/svga/svga_pipe_rasterizer.c @@ -30,7 +30,6 @@ #include "util/u_memory.h" #include "svga_context.h" -#include "svga_state.h" #include "svga_hw_reg.h" diff --git a/src/gallium/drivers/svga/svga_pipe_sampler.c b/src/gallium/drivers/svga/svga_pipe_sampler.c index 460a101f8c..161c66dd4f 100644 --- a/src/gallium/drivers/svga/svga_pipe_sampler.c +++ b/src/gallium/drivers/svga/svga_pipe_sampler.c @@ -32,9 +32,6 @@ #include "svga_context.h" #include "svga_screen_texture.h" -#include "svga_state.h" - -#include "svga_hw_reg.h" #include "svga_debug.h" diff --git a/src/gallium/drivers/svga/svga_pipe_vertex.c b/src/gallium/drivers/svga/svga_pipe_vertex.c index 42f290d162..0bf43fa900 100644 --- a/src/gallium/drivers/svga/svga_pipe_vertex.c +++ b/src/gallium/drivers/svga/svga_pipe_vertex.c @@ -32,10 +32,6 @@ #include "svga_screen.h" #include "svga_screen_buffer.h" #include "svga_context.h" -#include "svga_state.h" -#include "svga_winsys.h" - -#include "svga_hw_reg.h" static void svga_set_vertex_buffers(struct pipe_context *pipe, diff --git a/src/gallium/drivers/svga/svga_pipe_vs.c b/src/gallium/drivers/svga/svga_pipe_vs.c index 7e6ab576ad..c4ac5304ac 100644 --- a/src/gallium/drivers/svga/svga_pipe_vs.c +++ b/src/gallium/drivers/svga/svga_pipe_vs.c @@ -33,7 +33,6 @@ #include "svga_screen.h" #include "svga_context.h" -#include "svga_state.h" #include "svga_tgsi.h" #include "svga_hw_reg.h" #include "svga_cmd.h" diff --git a/src/gallium/drivers/svga/svga_screen.c b/src/gallium/drivers/svga/svga_screen.c index fc1b3c980e..cd1ed7bac8 100644 --- a/src/gallium/drivers/svga/svga_screen.c +++ b/src/gallium/drivers/svga/svga_screen.c @@ -33,10 +33,8 @@ #include "svga_screen.h" #include "svga_screen_texture.h" #include "svga_screen_buffer.h" -#include "svga_cmd.h" #include "svga_debug.h" -#include "svga_hw_reg.h" #include "svga3d_shaderdefs.h" @@ -393,8 +391,6 @@ svga_screen_create(struct svga_winsys_screen *sws) pipe_mutex_init(svgascreen->tex_mutex); pipe_mutex_init(svgascreen->swc_mutex); - LIST_INITHEAD(&svgascreen->cached_buffers); - svga_screen_cache_init(svgascreen); return screen; diff --git a/src/gallium/drivers/svga/svga_screen.h b/src/gallium/drivers/svga/svga_screen.h index b94ca7fc1c..a009b60720 100644 --- a/src/gallium/drivers/svga/svga_screen.h +++ b/src/gallium/drivers/svga/svga_screen.h @@ -68,12 +68,6 @@ struct svga_screen pipe_mutex tex_mutex; pipe_mutex swc_mutex; /* Protects the use of swc and dirty_buffers */ - /** - * List of buffers with cached GMR. Ordered from the most recently used to - * the least recently used - */ - struct list_head cached_buffers; - struct svga_host_surface_cache cache; }; diff --git a/src/gallium/drivers/svga/svga_screen_buffer.c b/src/gallium/drivers/svga/svga_screen_buffer.c index 58a1aba464..430a6970fd 100644 --- a/src/gallium/drivers/svga/svga_screen_buffer.c +++ b/src/gallium/drivers/svga/svga_screen_buffer.c @@ -113,68 +113,9 @@ svga_buffer_destroy_hw_storage(struct svga_screen *ss, struct svga_buffer *sbuf) if(sbuf->hw.buf) { sws->buffer_destroy(sws, sbuf->hw.buf); sbuf->hw.buf = NULL; - assert(sbuf->head.prev && sbuf->head.next); - LIST_DEL(&sbuf->head); -#ifdef DEBUG - sbuf->head.next = sbuf->head.prev = NULL; -#endif } } -static INLINE enum pipe_error -svga_buffer_backup(struct svga_screen *ss, struct svga_buffer *sbuf) -{ - if (sbuf->hw.buf && sbuf->hw.num_ranges) { - void *src; - - if (!sbuf->swbuf) - sbuf->swbuf = align_malloc(sbuf->base.size, sbuf->base.alignment); - if (!sbuf->swbuf) - return PIPE_ERROR_OUT_OF_MEMORY; - - src = ss->sws->buffer_map(ss->sws, sbuf->hw.buf, - PIPE_BUFFER_USAGE_CPU_READ); - if (!src) - return PIPE_ERROR; - - memcpy(sbuf->swbuf, src, sbuf->base.size); - ss->sws->buffer_unmap(ss->sws, sbuf->hw.buf); - } - - return PIPE_OK; -} - -/** - * Try to make GMR space available by freeing the hardware storage of - * unmapped - */ -boolean -svga_buffer_free_cached_hw_storage(struct svga_screen *ss) -{ - struct list_head *curr; - struct svga_buffer *sbuf; - enum pipe_error ret = PIPE_OK; - - curr = ss->cached_buffers.prev; - - /* free the least recently used buffer's hw storage which is not mapped */ - do { - if(curr == &ss->cached_buffers) - return FALSE; - - sbuf = LIST_ENTRY(struct svga_buffer, curr, head); - - curr = curr->prev; - if (sbuf->map.count == 0) - ret = svga_buffer_backup(ss, sbuf); - - } while(sbuf->map.count != 0 || ret != PIPE_OK); - - svga_buffer_destroy_hw_storage(ss, sbuf); - - return TRUE; -} - struct svga_winsys_buffer * svga_winsys_buffer_create( struct svga_screen *ss, unsigned alignment, @@ -195,12 +136,6 @@ svga_winsys_buffer_create( struct svga_screen *ss, svga_screen_flush(ss, NULL); buf = sws->buffer_create(sws, alignment, usage, size); - SVGA_DBG(DEBUG_DMA|DEBUG_PERF, "evicting buffers to find %d bytes GMR\n", - size); - - /* Try evicing all buffer storage */ - while(!buf && svga_buffer_free_cached_hw_storage(ss)) - buf = sws->buffer_create(sws, alignment, usage, size); } return buf; @@ -226,8 +161,6 @@ svga_buffer_create_hw_storage(struct svga_screen *ss, return PIPE_ERROR_OUT_OF_MEMORY; assert(!sbuf->needs_flush); - assert(!sbuf->head.prev && !sbuf->head.next); - LIST_ADD(&sbuf->head, &ss->cached_buffers); } return PIPE_OK; @@ -311,7 +244,6 @@ static void svga_buffer_upload_flush(struct svga_context *svga, struct svga_buffer *sbuf) { - struct svga_screen *ss = svga_screen(svga->pipe.screen); SVGA3dCopyBox *boxes; unsigned i; @@ -348,13 +280,16 @@ svga_buffer_upload_flush(struct svga_context *svga, assert(sbuf->head.prev && sbuf->head.next); LIST_DEL(&sbuf->head); +#ifdef DEBUG + sbuf->head.next = sbuf->head.prev = NULL; +#endif sbuf->needs_flush = FALSE; - /* XXX: do we care about cached_buffers any more ?*/ - LIST_ADD(&sbuf->head, &ss->cached_buffers); sbuf->hw.svga = NULL; sbuf->hw.boxes = NULL; + sbuf->host_written = TRUE; + /* Decrement reference count */ pipe_reference(&(sbuf->base.reference), NULL); sbuf = NULL; @@ -437,17 +372,17 @@ svga_buffer_map_range( struct pipe_screen *screen, } else { if(!sbuf->hw.buf) { - struct svga_winsys_surface *handle = sbuf->handle; - if(svga_buffer_create_hw_storage(ss, sbuf) != PIPE_OK) return NULL; /* Populate the hardware storage if the host surface pre-existed */ - if((usage & PIPE_BUFFER_USAGE_CPU_READ) && handle) { + if(sbuf->host_written) { SVGA3dSurfaceDMAFlags flags; enum pipe_error ret; struct pipe_fence_handle *fence = NULL; + assert(sbuf->handle); + SVGA_DBG(DEBUG_DMA|DEBUG_PERF, "dma from sid %p (buffer), bytes %u - %u\n", sbuf->handle, 0, sbuf->base.size); @@ -478,17 +413,6 @@ svga_buffer_map_range( struct pipe_screen *screen, sws->fence_reference(sws, &fence, NULL); } } - else { - if((usage & PIPE_BUFFER_USAGE_CPU_READ) && !sbuf->needs_flush) { - /* We already had the hardware storage but we would have to issue - * a download if we hadn't, so move the buffer to the begginning - * of the LRU list. - */ - assert(sbuf->head.prev && sbuf->head.next); - LIST_DEL(&sbuf->head); - LIST_ADD(&sbuf->head, &ss->cached_buffers); - } - } map = sws->buffer_map(sws, sbuf->hw.buf, usage); } @@ -572,10 +496,8 @@ svga_buffer_destroy( struct pipe_buffer *buf ) assert(!sbuf->needs_flush); - if(sbuf->handle) { - SVGA_DBG(DEBUG_DMA, "release sid %p sz %d\n", sbuf->handle, sbuf->base.size); - svga_screen_surface_destroy(ss, &sbuf->key, &sbuf->handle); - } + if(sbuf->handle) + svga_buffer_destroy_host_surface(ss, sbuf); if(sbuf->hw.buf) svga_buffer_destroy_hw_storage(ss, sbuf); @@ -595,6 +517,9 @@ svga_buffer_create(struct pipe_screen *screen, struct svga_screen *ss = svga_screen(screen); struct svga_buffer *sbuf; + assert(size); + assert(alignment); + sbuf = CALLOC_STRUCT(svga_buffer); if(!sbuf) goto error1; @@ -755,8 +680,7 @@ svga_buffer_handle(struct svga_context *svga, assert(sbuf->hw.svga == svga); sbuf->needs_flush = TRUE; - assert(sbuf->head.prev && sbuf->head.next); - LIST_DEL(&sbuf->head); + assert(!sbuf->head.prev && !sbuf->head.next); LIST_ADDTAIL(&sbuf->head, &svga->dirty_buffers); } diff --git a/src/gallium/drivers/svga/svga_screen_buffer.h b/src/gallium/drivers/svga/svga_screen_buffer.h index 5d7af5a7c5..448ac107c7 100644 --- a/src/gallium/drivers/svga/svga_screen_buffer.h +++ b/src/gallium/drivers/svga/svga_screen_buffer.h @@ -135,6 +135,11 @@ struct svga_buffer */ struct svga_winsys_surface *handle; + /** + * Whether the host has been ever written. + */ + boolean host_written; + struct { unsigned count; boolean writing; @@ -178,9 +183,6 @@ svga_buffer_handle(struct svga_context *svga, void svga_context_flush_buffers(struct svga_context *svga); -boolean -svga_buffer_free_cached_hw_storage(struct svga_screen *ss); - struct svga_winsys_buffer * svga_winsys_buffer_create(struct svga_screen *ss, unsigned alignment, diff --git a/src/gallium/drivers/svga/svga_screen_texture.c b/src/gallium/drivers/svga/svga_screen_texture.c index 2224c2d394..0d69007fd8 100644 --- a/src/gallium/drivers/svga/svga_screen_texture.c +++ b/src/gallium/drivers/svga/svga_screen_texture.c @@ -306,11 +306,19 @@ svga_texture_create(struct pipe_screen *screen, tex->key.numFaces = 1; } + tex->key.cachable = 1; + if(templat->tex_usage & PIPE_TEXTURE_USAGE_SAMPLER) tex->key.flags |= SVGA3D_SURFACE_HINT_TEXTURE; - if(templat->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY) + if(templat->tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) { + tex->key.cachable = 0; + } + + if(templat->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY) { tex->key.flags |= SVGA3D_SURFACE_HINT_SCANOUT; + tex->key.cachable = 0; + } /* * XXX: Never pass the SVGA3D_SURFACE_HINT_RENDERTARGET hint. Mesa cannot @@ -333,8 +341,6 @@ svga_texture_create(struct pipe_screen *screen, if(tex->key.format == SVGA3D_FORMAT_INVALID) goto error2; - tex->key.cachable = 1; - SVGA_DBG(DEBUG_DMA, "surface_create for texture\n", tex->handle); tex->handle = svga_screen_surface_create(svgascreen, &tex->key); if (tex->handle) @@ -416,6 +422,62 @@ svga_texture_blanket(struct pipe_screen * screen, } +struct pipe_texture * +svga_screen_texture_wrap_surface(struct pipe_screen *screen, + struct pipe_texture *base, + enum SVGA3dSurfaceFormat format, + struct svga_winsys_surface *srf) +{ + struct svga_texture *tex; + assert(screen); + + /* Only supports one type */ + if (base->target != PIPE_TEXTURE_2D || + base->last_level != 0 || + base->depth0 != 1) { + return NULL; + } + + if (!srf) + return NULL; + + if (svga_translate_format(base->format) != format) { + unsigned f1 = svga_translate_format(base->format); + unsigned f2 = format; + + /* It's okay for XRGB and ARGB or depth with/out stencil to get mixed up */ + if ( !( (f1 == SVGA3D_X8R8G8B8 && f2 == SVGA3D_A8R8G8B8) || + (f1 == SVGA3D_A8R8G8B8 && f2 == SVGA3D_X8R8G8B8) || + (f1 == SVGA3D_Z_D24X8 && f2 == SVGA3D_Z_D24S8) ) ) { + debug_printf("%s wrong format %u != %u\n", __FUNCTION__, f1, f2); + return NULL; + } + } + + tex = CALLOC_STRUCT(svga_texture); + if (!tex) + return NULL; + + tex->base = *base; + + + if (format == 1) + tex->base.format = PIPE_FORMAT_X8R8G8B8_UNORM; + else if (format == 2) + tex->base.format = PIPE_FORMAT_A8R8G8B8_UNORM; + + pipe_reference_init(&tex->base.reference, 1); + tex->base.screen = screen; + + SVGA_DBG(DEBUG_DMA, "wrap surface sid %p\n", srf); + + tex->key.cachable = 0; + tex->handle = srf; + + return &tex->base; +} + + static void svga_texture_destroy(struct pipe_texture *pt) { diff --git a/src/gallium/drivers/svga/svga_state_framebuffer.c b/src/gallium/drivers/svga/svga_state_framebuffer.c index cfdcae4ee4..eda1aefd67 100644 --- a/src/gallium/drivers/svga/svga_state_framebuffer.c +++ b/src/gallium/drivers/svga/svga_state_framebuffer.c @@ -32,8 +32,6 @@ #include "svga_cmd.h" #include "svga_debug.h" -#include "svga_hw_reg.h" - /*********************************************************************** * Hardware state update diff --git a/src/gallium/drivers/svga/svga_state_rss.c b/src/gallium/drivers/svga/svga_state_rss.c index 8b6803a285..2f9adaeb56 100644 --- a/src/gallium/drivers/svga/svga_state_rss.c +++ b/src/gallium/drivers/svga/svga_state_rss.c @@ -31,9 +31,6 @@ #include "svga_state.h" #include "svga_cmd.h" -#include "svga_hw_reg.h" - - struct rs_queue { unsigned rs_count; diff --git a/src/gallium/drivers/svga/svga_state_tss.c b/src/gallium/drivers/svga/svga_state_tss.c index b313794520..b3c9687b1a 100644 --- a/src/gallium/drivers/svga/svga_state_tss.c +++ b/src/gallium/drivers/svga/svga_state_tss.c @@ -33,8 +33,6 @@ #include "svga_state.h" #include "svga_cmd.h" -#include "svga_hw_reg.h" - void svga_cleanup_tss_binding(struct svga_context *svga) { diff --git a/src/gallium/drivers/svga/svga_swtnl_backend.c b/src/gallium/drivers/svga/svga_swtnl_backend.c index b4f757a47a..aafb3e26b0 100644 --- a/src/gallium/drivers/svga/svga_swtnl_backend.c +++ b/src/gallium/drivers/svga/svga_swtnl_backend.c @@ -31,7 +31,6 @@ #include "pipe/p_inlines.h" #include "util/u_math.h" #include "util/u_memory.h" -#include "util/u_simple_shaders.h" #include "svga_context.h" #include "svga_state.h" @@ -87,13 +86,13 @@ svga_vbuf_render_allocate_vertices( struct vbuf_render *render, if (!svga_render->vbuf) { svga_render->vbuf_size = MAX2(size, svga_render->vbuf_alloc_size); svga_render->vbuf = pipe_buffer_create(screen, - 0, + 16, PIPE_BUFFER_USAGE_VERTEX, svga_render->vbuf_size); if(!svga_render->vbuf) { svga_context_flush(svga, NULL); svga_render->vbuf = pipe_buffer_create(screen, - 0, + 16, PIPE_BUFFER_USAGE_VERTEX, svga_render->vbuf_size); assert(svga_render->vbuf); @@ -123,7 +122,9 @@ svga_vbuf_render_map_vertices( struct vbuf_render *render ) char *ptr = (char*)pipe_buffer_map(screen, svga_render->vbuf, PIPE_BUFFER_USAGE_CPU_WRITE | - PIPE_BUFFER_USAGE_FLUSH_EXPLICIT); + PIPE_BUFFER_USAGE_FLUSH_EXPLICIT | + PIPE_BUFFER_USAGE_DISCARD | + PIPE_BUFFER_USAGE_UNSYNCHRONIZED); return ptr + svga_render->vbuf_offset; } @@ -259,14 +260,14 @@ svga_vbuf_render_draw( struct vbuf_render *render, if (!svga_render->ibuf) { svga_render->ibuf_size = MAX2(size, svga_render->ibuf_alloc_size); svga_render->ibuf = pipe_buffer_create(screen, - 0, + 2, PIPE_BUFFER_USAGE_VERTEX, svga_render->ibuf_size); svga_render->ibuf_offset = 0; } - pipe_buffer_write(screen, svga_render->ibuf, - svga_render->ibuf_offset, 2 * nr_indices, indices); + pipe_buffer_write_nooverlap(screen, svga_render->ibuf, + svga_render->ibuf_offset, 2 * nr_indices, indices); /* off to hardware */ diff --git a/src/gallium/drivers/svga/svga_swtnl_draw.c b/src/gallium/drivers/svga/svga_swtnl_draw.c index 7655121bec..0ae58c7d9d 100644 --- a/src/gallium/drivers/svga/svga_swtnl_draw.c +++ b/src/gallium/drivers/svga/svga_swtnl_draw.c @@ -27,7 +27,6 @@ #include "draw/draw_vbuf.h" #include "pipe/p_inlines.h" #include "pipe/p_state.h" -#include "util/u_memory.h" #include "svga_context.h" #include "svga_swtnl.h" diff --git a/src/gallium/drivers/svga/svga_swtnl_state.c b/src/gallium/drivers/svga/svga_swtnl_state.c index 94b6ccc62d..fe03e207ff 100644 --- a/src/gallium/drivers/svga/svga_swtnl_state.c +++ b/src/gallium/drivers/svga/svga_swtnl_state.c @@ -27,7 +27,6 @@ #include "draw/draw_vbuf.h" #include "pipe/p_inlines.h" #include "pipe/p_state.h" -#include "util/u_memory.h" #include "svga_context.h" #include "svga_swtnl.h" diff --git a/src/gallium/drivers/svga/svga_tgsi_decl_sm20.c b/src/gallium/drivers/svga/svga_tgsi_decl_sm20.c index 23b3ace7f3..1ae9906761 100644 --- a/src/gallium/drivers/svga/svga_tgsi_decl_sm20.c +++ b/src/gallium/drivers/svga/svga_tgsi_decl_sm20.c @@ -29,9 +29,6 @@ #include "util/u_memory.h" #include "svga_tgsi_emit.h" -#include "svga_context.h" - - static boolean ps20_input( struct svga_shader_emitter *emit, diff --git a/src/gallium/drivers/svga/svga_tgsi_decl_sm30.c b/src/gallium/drivers/svga/svga_tgsi_decl_sm30.c index d1c7336dec..43fc0d3235 100644 --- a/src/gallium/drivers/svga/svga_tgsi_decl_sm30.c +++ b/src/gallium/drivers/svga/svga_tgsi_decl_sm30.c @@ -29,7 +29,6 @@ #include "util/u_memory.h" #include "svga_tgsi_emit.h" -#include "svga_context.h" static boolean translate_vs_ps_semantic( struct tgsi_declaration_semantic semantic, unsigned *usage, diff --git a/src/gallium/drivers/svga/svga_winsys.h b/src/gallium/drivers/svga/svga_winsys.h index 59f299c185..27b99fe4c1 100644 --- a/src/gallium/drivers/svga/svga_winsys.h +++ b/src/gallium/drivers/svga/svga_winsys.h @@ -296,4 +296,10 @@ svga_screen_buffer_from_texture(struct pipe_texture *texture, struct pipe_buffer **buffer, unsigned *stride); +struct pipe_texture * +svga_screen_texture_wrap_surface(struct pipe_screen *screen, + struct pipe_texture *base, + enum SVGA3dSurfaceFormat format, + struct svga_winsys_surface *srf); + #endif /* SVGA_WINSYS_H_ */ diff --git a/src/gallium/drivers/trace/tr_drm.c b/src/gallium/drivers/trace/tr_drm.c index 48d1c4051c..e7ca3a86ea 100644 --- a/src/gallium/drivers/trace/tr_drm.c +++ b/src/gallium/drivers/trace/tr_drm.c @@ -173,6 +173,7 @@ trace_drm_create(struct drm_api *api) if (!tr_api) goto error; + tr_api->base.driver_name = api->driver_name; tr_api->base.create_screen = trace_drm_create_screen; tr_api->base.create_context = trace_drm_create_context; tr_api->base.texture_from_shared_handle = trace_drm_texture_from_shared_handle; |