This removes the stupid strict-conformance fallback code I broke when adding ARB_sampler_objects.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=36572 --- src/mesa/drivers/dri/i965/brw_draw.c | 36 -------------------- src/mesa/drivers/dri/i965/brw_fs.cpp | 17 +++++++--- src/mesa/drivers/dri/i965/brw_fs.h | 6 ++- src/mesa/drivers/dri/i965/brw_wm.c | 10 ++++++ src/mesa/drivers/dri/i965/brw_wm.h | 2 +- src/mesa/drivers/dri/i965/brw_wm_emit.c | 5 +++ src/mesa/drivers/dri/i965/brw_wm_sampler_state.c | 38 +++++++++++++++------- 7 files changed, 58 insertions(+), 56 deletions(-) diff --git a/src/mesa/drivers/dri/i965/brw_draw.c b/src/mesa/drivers/dri/i965/brw_draw.c index 9ab5331..2c3c373 100644 --- a/src/mesa/drivers/dri/i965/brw_draw.c +++ b/src/mesa/drivers/dri/i965/brw_draw.c @@ -270,42 +270,6 @@ static GLboolean check_fallbacks( struct brw_context *brw, return GL_TRUE; } - /* BRW hardware doesn't handle GL_CLAMP texturing correctly; - * brw_wm_sampler_state:translate_wrap_mode() treats GL_CLAMP - * as GL_CLAMP_TO_EDGE instead. If we're using GL_CLAMP, and - * we want strict conformance, force the fallback. - * Right now, we only do this for 2D textures. - */ - { - int u; - for (u = 0; u < ctx->Const.MaxTextureCoordUnits; u++) { - struct gl_texture_unit *texUnit = &ctx->Texture.Unit[u]; - - if (texUnit->Enabled) { - struct gl_sampler_object *sampler = _mesa_get_samplerobj(ctx, u); - - if (texUnit->Enabled & TEXTURE_1D_BIT) { - if (sampler->WrapS == GL_CLAMP) { - return GL_TRUE; - } - } - if (texUnit->Enabled & TEXTURE_2D_BIT) { - if (sampler->WrapS == GL_CLAMP || - sampler->WrapT == GL_CLAMP) { - return GL_TRUE; - } - } - if (texUnit->Enabled & TEXTURE_3D_BIT) { - if (sampler->WrapS == GL_CLAMP || - sampler->WrapT == GL_CLAMP || - sampler->WrapR == GL_CLAMP) { - return GL_TRUE; - } - } - } - } - } - /* Nothing stopping us from the fast path now */ return GL_FALSE; } diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp b/src/mesa/drivers/dri/i965/brw_fs.cpp index 4ea9f2b..0d520e2 100644 --- a/src/mesa/drivers/dri/i965/brw_fs.cpp +++ b/src/mesa/drivers/dri/i965/brw_fs.cpp @@ -1172,7 +1172,8 @@ fs_visitor::visit(ir_assignment *ir) } fs_inst * -fs_visitor::emit_texture_gen4(ir_texture *ir, fs_reg dst, fs_reg coordinate) +fs_visitor::emit_texture_gen4(ir_texture *ir, fs_reg dst, fs_reg coordinate, + int sampler) { int mlen; int base_mrf = 1; @@ -1184,7 +1185,12 @@ fs_visitor::emit_texture_gen4(ir_texture *ir, fs_reg dst, fs_reg coordinate) if (ir->shadow_comparitor) { for (int i = 0; i < ir->coordinate->type->vector_elements; i++) { - emit(BRW_OPCODE_MOV, fs_reg(MRF, base_mrf + mlen + i), coordinate); + fs_inst *inst = emit(BRW_OPCODE_MOV, + fs_reg(MRF, base_mrf + mlen + i), coordinate); + + if (i < 3 && c->key.gl_clamp_mask[i] & (1 << sampler)) + inst->saturate = true; + coordinate.reg_offset++; } /* gen4's SIMD8 sampler always has the slots for u,v,r present. */ @@ -1298,7 +1304,8 @@ fs_visitor::emit_texture_gen4(ir_texture *ir, fs_reg dst, fs_reg coordinate) * surprising in the disassembly. */ fs_inst * -fs_visitor::emit_texture_gen5(ir_texture *ir, fs_reg dst, fs_reg coordinate) +fs_visitor::emit_texture_gen5(ir_texture *ir, fs_reg dst, fs_reg coordinate, + int sampler) { int mlen = 1; /* g0 header always present. */ int base_mrf = 1; @@ -1457,9 +1464,9 @@ fs_visitor::visit(ir_texture *ir) fs_reg dst = fs_reg(this, glsl_type::vec4_type); if (intel->gen < 5) { - inst = emit_texture_gen4(ir, dst, coordinate); + inst = emit_texture_gen4(ir, dst, coordinate, sampler); } else { - inst = emit_texture_gen5(ir, dst, coordinate); + inst = emit_texture_gen5(ir, dst, coordinate, sampler); } /* If there's an offset, we already set up m1. To avoid the implied move, diff --git a/src/mesa/drivers/dri/i965/brw_fs.h b/src/mesa/drivers/dri/i965/brw_fs.h index 3901207..dd63777 100644 --- a/src/mesa/drivers/dri/i965/brw_fs.h +++ b/src/mesa/drivers/dri/i965/brw_fs.h @@ -512,8 +512,10 @@ public: fs_reg *emit_general_interpolation(ir_variable *ir); void emit_interpolation_setup_gen4(); void emit_interpolation_setup_gen6(); - fs_inst *emit_texture_gen4(ir_texture *ir, fs_reg dst, fs_reg coordinate); - fs_inst *emit_texture_gen5(ir_texture *ir, fs_reg dst, fs_reg coordinate); + fs_inst *emit_texture_gen4(ir_texture *ir, fs_reg dst, fs_reg coordinate, + int sampler); + fs_inst *emit_texture_gen5(ir_texture *ir, fs_reg dst, fs_reg coordinate, + int sampler); fs_inst *emit_math(fs_opcodes op, fs_reg dst, fs_reg src0); fs_inst *emit_math(fs_opcodes op, fs_reg dst, fs_reg src0, fs_reg src1); bool try_emit_saturate(ir_expression *ir); diff --git a/src/mesa/drivers/dri/i965/brw_wm.c b/src/mesa/drivers/dri/i965/brw_wm.c index 06512de..6054918 100644 --- a/src/mesa/drivers/dri/i965/brw_wm.c +++ b/src/mesa/drivers/dri/i965/brw_wm.c @@ -409,6 +409,16 @@ static void brw_wm_populate_key( struct brw_context *brw, swizzles[GET_SWZ(t->_Swizzle, 1)], swizzles[GET_SWZ(t->_Swizzle, 2)], swizzles[GET_SWZ(t->_Swizzle, 3)]); + + if (sampler->MinFilter != GL_NEAREST && + sampler->MagFilter != GL_NEAREST) { + if (sampler->WrapS == GL_CLAMP) + key->gl_clamp_mask[0] |= 1 << i; + if (sampler->WrapT == GL_CLAMP) + key->gl_clamp_mask[1] |= 1 << i; + if (sampler->WrapR == GL_CLAMP) + key->gl_clamp_mask[2] |= 1 << i; + } } else { key->tex_swizzles[i] = SWIZZLE_NOOP; diff --git a/src/mesa/drivers/dri/i965/brw_wm.h b/src/mesa/drivers/dri/i965/brw_wm.h index a5f99a0..8ab531b 100644 --- a/src/mesa/drivers/dri/i965/brw_wm.h +++ b/src/mesa/drivers/dri/i965/brw_wm.h @@ -71,9 +71,9 @@ struct brw_wm_prog_key { GLuint shadowtex_mask:16; GLuint yuvtex_mask:16; GLuint yuvtex_swap_mask:16; /* UV swaped */ + uint16_t gl_clamp_mask[3]; GLushort tex_swizzles[BRW_MAX_TEX_UNIT]; - GLushort drawable_height; GLbitfield64 vp_outputs_written; GLuint iz_lookup; diff --git a/src/mesa/drivers/dri/i965/brw_wm_emit.c b/src/mesa/drivers/dri/i965/brw_wm_emit.c index fd4cd89..f61757a 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_emit.c +++ b/src/mesa/drivers/dri/i965/brw_wm_emit.c @@ -1100,11 +1100,16 @@ void emit_tex(struct brw_wm_compile *c, /* Emit the texcoords. */ for (i = 0; i < nr_texcoords; i++) { + if (c->key.gl_clamp_mask[i] & (1 << sampler)) + brw_set_saturate(p, true); + if (emit & (1<<i)) brw_MOV(p, brw_message_reg(cur_mrf), arg[i]); else brw_MOV(p, brw_message_reg(cur_mrf), brw_imm_f(0)); cur_mrf += mrf_per_channel; + + brw_set_saturate(p, false); } /* Fill in the shadow comparison reference value. */ diff --git a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c index 7b93bf9..6ac02d2 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c @@ -29,7 +29,7 @@ * Keith Whitwell <ke...@tungstengraphics.com> */ - +#include <stdbool.h> #include "brw_context.h" #include "brw_state.h" #include "brw_defines.h" @@ -44,19 +44,27 @@ -/* The brw (and related graphics cores) do not support GL_CLAMP. The - * Intel drivers for "other operating systems" implement GL_CLAMP as - * GL_CLAMP_TO_EDGE, so the same is done here. - */ -static GLuint translate_wrap_mode( GLenum wrap ) +static GLuint translate_wrap_mode(GLenum wrap, bool using_nearest) { switch( wrap ) { case GL_REPEAT: return BRW_TEXCOORDMODE_WRAP; - case GL_CLAMP: - return BRW_TEXCOORDMODE_CLAMP; + case GL_CLAMP: + /* GL_CLAMP is the weird mode where coordinates are clamped to + * [0.0, 1.0], so linear filtering of coordinates outside of + * [0.0, 1.0] give you half edge texel value and half border + * color. The fragment shader will clamp the coordinates, and + * we set clamp_border here, which gets the result desired. We + * just use clamp(_to_edge) for nearest, because for nearest + * clamping to 1.0 gives border color instead of the desired + * edge texels. + */ + if (using_nearest) + return BRW_TEXCOORDMODE_CLAMP; + else + return BRW_TEXCOORDMODE_CLAMP_BORDER; case GL_CLAMP_TO_EDGE: - return BRW_TEXCOORDMODE_CLAMP; /* conform likes it this way */ + return BRW_TEXCOORDMODE_CLAMP; case GL_CLAMP_TO_BORDER: return BRW_TEXCOORDMODE_CLAMP_BORDER; case GL_MIRRORED_REPEAT: @@ -151,11 +159,13 @@ static void brw_update_sampler_state(struct brw_context *brw, struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; struct gl_texture_object *texObj = texUnit->_Current; struct gl_sampler_object *gl_sampler = _mesa_get_samplerobj(ctx, unit); + bool using_nearest = false; switch (gl_sampler->MinFilter) { case GL_NEAREST: sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST; sampler->ss0.mip_filter = BRW_MIPFILTER_NONE; + using_nearest = true; break; case GL_LINEAR: sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR; @@ -196,6 +206,7 @@ static void brw_update_sampler_state(struct brw_context *brw, switch (gl_sampler->MagFilter) { case GL_NEAREST: sampler->ss0.mag_filter = BRW_MAPFILTER_NEAREST; + using_nearest = true; break; case GL_LINEAR: sampler->ss0.mag_filter = BRW_MAPFILTER_LINEAR; @@ -205,9 +216,12 @@ static void brw_update_sampler_state(struct brw_context *brw, } } - sampler->ss1.r_wrap_mode = translate_wrap_mode(gl_sampler->WrapR); - sampler->ss1.s_wrap_mode = translate_wrap_mode(gl_sampler->WrapS); - sampler->ss1.t_wrap_mode = translate_wrap_mode(gl_sampler->WrapT); + sampler->ss1.r_wrap_mode = translate_wrap_mode(gl_sampler->WrapR, + using_nearest); + sampler->ss1.s_wrap_mode = translate_wrap_mode(gl_sampler->WrapS, + using_nearest); + sampler->ss1.t_wrap_mode = translate_wrap_mode(gl_sampler->WrapT, + using_nearest); if (intel->gen >= 6 && sampler->ss0.min_filter != sampler->ss0.mag_filter) -- 1.7.5.1 _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
