Merge remote-tracking branch 'origin/master' into pipe-video

[mesa.git] / src / gallium / drivers / r600 / r600_shader.c

diff --git a/src/gallium/drivers/r600/r600_shader.c b/src/gallium/drivers/r600/r600_shader.c
index e7285d624e3679a18e6900e4ad422e871f2038eb..845d41ace0278be86cca714b9eb62b41f125ec7d 100644 (file)
--- a/src/gallium/drivers/r600/r600_shader.c
+++ b/src/gallium/drivers/r600/r600_shader.c
@@ -21,6 +21,7 @@
  * USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
 #include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_info.h"
 #include "tgsi/tgsi_parse.h"
 #include "tgsi/tgsi_scan.h"
 #include "tgsi/tgsi_dump.h"
@@ -33,6 +34,7 @@
 #include "r600d.h"
 #include <stdio.h>
 #include <errno.h>
+#include <byteswap.h>
 
 int r600_find_vs_semantic_index(struct r600_shader *vs,
                                struct r600_shader *ps, int id)
@@ -52,7 +54,8 @@ static int r600_pipe_shader(struct pipe_context *ctx, struct r600_pipe_shader *s
 {
        struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx;
        struct r600_shader *rshader = &shader->shader;
-       void *ptr;
+       uint32_t *ptr;
+       int     i;
 
        /* copy new shader */
        if (shader->bo == NULL) {
@@ -60,8 +63,14 @@ static int r600_pipe_shader(struct pipe_context *ctx, struct r600_pipe_shader *s
                if (shader->bo == NULL) {
                        return -ENOMEM;
                }
-               ptr = r600_bo_map(rctx->radeon, shader->bo, 0, NULL);
-               memcpy(ptr, rshader->bc.bytecode, rshader->bc.ndw * 4);
+               ptr = (uint32_t*)r600_bo_map(rctx->radeon, shader->bo, 0, NULL);
+               if (R600_BIG_ENDIAN) {
+                       for (i = 0; i < rshader->bc.ndw; ++i) {
+                               ptr[i] = bswap_32(rshader->bc.bytecode[i]);
+                       }
+               } else {
+                       memcpy(ptr, rshader->bc.bytecode, rshader->bc.ndw * sizeof(*ptr));
+               }
                r600_bo_unmap(rctx->radeon, shader->bo);
        }
        /* build state */
@@ -467,6 +476,7 @@ static int tgsi_fetch_rel_const(struct r600_shader_ctx *ctx, unsigned int offset
        vtx.num_format_all = 2;         /* NUM_FORMAT_SCALED */
        vtx.format_comp_all = 1;        /* FORMAT_COMP_SIGNED */
        vtx.srf_mode_all = 1;           /* SRF_MODE_NO_ZERO */
+       vtx.endian = r600_endian_swap(32);
 
        if ((r = r600_bc_add_vtx(ctx->bc, &vtx)))
                return r;
@@ -815,7 +825,8 @@ out_err:
 
 static int tgsi_unsupported(struct r600_shader_ctx *ctx)
 {
-       R600_ERR("%d tgsi opcode unsupported\n", ctx->inst_info->tgsi_opcode);
+       R600_ERR("%s tgsi opcode unsupported\n",
+                tgsi_get_opcode_name(ctx->inst_info->tgsi_opcode));
        return -EINVAL;
 }
 
@@ -896,6 +907,8 @@ static int tgsi_op2_s(struct r600_shader_ctx *ctx, int swap)
                        break;
                case TGSI_OPCODE_ABS:
                        alu.src[0].abs = 1;
+                       if (alu.src[0].neg)
+                         alu.src[0].neg = 0;
                        break;
                default:
                        break;
@@ -1526,9 +1539,13 @@ static int tgsi_tex(struct r600_shader_ctx *ctx)
        unsigned src_gpr;
        int r, i;
        int opcode;
-       boolean src_not_temp =
-               inst->Src[0].Register.File != TGSI_FILE_TEMPORARY &&
-               inst->Src[0].Register.File != TGSI_FILE_INPUT;
+       /* Texture fetch instructions can only use gprs as source.
+        * Also they cannot negate the source or take the absolute value */
+       const boolean src_requires_loading =
+               (inst->Src[0].Register.File != TGSI_FILE_TEMPORARY &&
+               inst->Src[0].Register.File != TGSI_FILE_INPUT) ||
+               ctx->src[0].neg || ctx->src[0].abs;
+       boolean src_loaded = FALSE;
 
        src_gpr = ctx->file_offset[inst->Src[0].Register.File] + inst->Src[0].Register.Index;
 
@@ -1570,7 +1587,7 @@ static int tgsi_tex(struct r600_shader_ctx *ctx)
                r = r600_bc_add_alu(ctx->bc, &alu);
                if (r)
                        return r;
-               src_not_temp = FALSE;
+               src_loaded = TRUE;
                src_gpr = ctx->temp_reg;
        }
 
@@ -1655,11 +1672,11 @@ static int tgsi_tex(struct r600_shader_ctx *ctx)
                if (r)
                        return r;
 
-               src_not_temp = FALSE;
+               src_loaded = TRUE;
                src_gpr = ctx->temp_reg;
        }
 
-       if (src_not_temp) {
+       if (src_requires_loading && !src_loaded) {
                for (i = 0; i < 4; i++) {
                        memset(&alu, 0, sizeof(struct r600_bc_alu));
                        alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV);
@@ -1673,6 +1690,7 @@ static int tgsi_tex(struct r600_shader_ctx *ctx)
                        if (r)
                                return r;
                }
+               src_loaded = TRUE;
                src_gpr = ctx->temp_reg;
        }
 
@@ -1691,10 +1709,18 @@ static int tgsi_tex(struct r600_shader_ctx *ctx)
        tex.dst_sel_y = (inst->Dst[0].Register.WriteMask & 2) ? 1 : 7;
        tex.dst_sel_z = (inst->Dst[0].Register.WriteMask & 4) ? 2 : 7;
        tex.dst_sel_w = (inst->Dst[0].Register.WriteMask & 8) ? 3 : 7;
-       tex.src_sel_x = 0;
-       tex.src_sel_y = 1;
-       tex.src_sel_z = 2;
-       tex.src_sel_w = 3;
+       if (src_loaded) {
+               tex.src_sel_x = 0;
+               tex.src_sel_y = 1;
+               tex.src_sel_z = 2;
+               tex.src_sel_w = 3;
+       } else {
+               tex.src_sel_x = ctx->src[0].swizzle[0];
+               tex.src_sel_y = ctx->src[0].swizzle[1];
+               tex.src_sel_z = ctx->src[0].swizzle[2];
+               tex.src_sel_w = ctx->src[0].swizzle[3];
+               tex.src_rel = ctx->src[0].rel;
+       }
 
        if (inst->Texture.Texture == TGSI_TEXTURE_CUBE) {
                tex.src_sel_x = 1;
@@ -1712,12 +1738,12 @@ static int tgsi_tex(struct r600_shader_ctx *ctx)
 
        if (inst->Texture.Texture == TGSI_TEXTURE_1D_ARRAY) {
                tex.coord_type_z = 0;
-               tex.src_sel_z = 1;
+               tex.src_sel_z = tex.src_sel_y;
        } else if (inst->Texture.Texture == TGSI_TEXTURE_2D_ARRAY)
                tex.coord_type_z = 0;
 
        if (inst->Texture.Texture == TGSI_TEXTURE_SHADOW1D || inst->Texture.Texture == TGSI_TEXTURE_SHADOW2D)
-               tex.src_sel_w = 2;
+               tex.src_sel_w = tex.src_sel_z;
 
        r = r600_bc_add_tex(ctx->bc, &tex);
        if (r)
@@ -1969,9 +1995,11 @@ static int tgsi_exp(struct r600_shader_ctx *ctx)
                r600_bc_src(&alu.src[0], &ctx->src[0], 0);
 
                alu.dst.sel = ctx->temp_reg;
-//             r = tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst);
-//             if (r)
-//                     return r;
+#if 0
+               r = tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst);
+               if (r)
+                       return r;
+#endif
                alu.dst.write = 1;
                alu.dst.chan = 1;