nvc0: add support for TGSI FMA ops

This will allow the nouveau backend to not try and split up ops that are
fused in GLSL.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

diff --git a/src/gallium/drivers/nouveau/codegen/nv50_ir_from_tgsi.cpp b/src/gallium/drivers/nouveau/codegen/nv50_ir_from_tgsi.cpp
index 868372202f955d66a83d131ee26bd73ea9adf454..b06d86ad7607cc3934fc83decbe819254a380dc6 100644 (file)
--- a/src/gallium/drivers/nouveau/codegen/nv50_ir_from_tgsi.cpp
+++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_from_tgsi.cpp
@@ -525,6 +525,7 @@ nv50_ir::DataType Instruction::inferSrcType() const
    case TGSI_OPCODE_DRCP:
    case TGSI_OPCODE_DSQRT:
    case TGSI_OPCODE_DMAD:
+   case TGSI_OPCODE_DFMA:
    case TGSI_OPCODE_DFRAC:
    case TGSI_OPCODE_DRSQ:
    case TGSI_OPCODE_DTRUNC:
@@ -624,6 +625,7 @@ static nv50_ir::operation translateOpcode(uint opcode)
    NV50_IR_OPCODE_CASE(SLT, SET);
    NV50_IR_OPCODE_CASE(SGE, SET);
    NV50_IR_OPCODE_CASE(MAD, MAD);
+   NV50_IR_OPCODE_CASE(FMA, FMA);
    NV50_IR_OPCODE_CASE(SUB, SUB);
 
    NV50_IR_OPCODE_CASE(FLR, FLOOR);
@@ -723,6 +725,7 @@ static nv50_ir::operation translateOpcode(uint opcode)
    NV50_IR_OPCODE_CASE(DRCP, RCP);
    NV50_IR_OPCODE_CASE(DSQRT, SQRT);
    NV50_IR_OPCODE_CASE(DMAD, MAD);
+   NV50_IR_OPCODE_CASE(DFMA, FMA);
    NV50_IR_OPCODE_CASE(D2I, CVT);
    NV50_IR_OPCODE_CASE(D2U, CVT);
    NV50_IR_OPCODE_CASE(I2D, CVT);
@@ -2672,6 +2675,7 @@ Converter::handleInstruction(const struct tgsi_full_instruction *insn)
    case TGSI_OPCODE_MAD:
    case TGSI_OPCODE_UMAD:
    case TGSI_OPCODE_SAD:
+   case TGSI_OPCODE_FMA:
       FOR_EACH_DST_ENABLED_CHANNEL(0, c, tgsi) {
          src0 = fetchSrc(0, c);
          src1 = fetchSrc(1, c);
@@ -3395,6 +3399,7 @@ Converter::handleInstruction(const struct tgsi_full_instruction *insn)
       }
       break;
    case TGSI_OPCODE_DMAD:
+   case TGSI_OPCODE_DFMA:
       FOR_EACH_DST_ENABLED_CHANNEL(0, c, tgsi) {
          src0 = getSSA(8);
          src1 = getSSA(8);

diff --git a/src/gallium/drivers/nouveau/nvc0/nvc0_screen.c b/src/gallium/drivers/nouveau/nvc0/nvc0_screen.c
index 114d868e919dd69f7336b909a6806bac7523ca6b..30afdf2f03c9f7300ec4c9b0995034fcdd03cd87 100644 (file)
--- a/src/gallium/drivers/nouveau/nvc0/nvc0_screen.c
+++ b/src/gallium/drivers/nouveau/nvc0/nvc0_screen.c
@@ -337,8 +337,9 @@ nvc0_screen_get_shader_param(struct pipe_screen *pscreen, unsigned shader,
       return 1;
    case PIPE_SHADER_CAP_TGSI_DROUND_SUPPORTED:
       return 1;
-   case PIPE_SHADER_CAP_TGSI_DFRACEXP_DLDEXP_SUPPORTED:
    case PIPE_SHADER_CAP_TGSI_FMA_SUPPORTED:
+      return 1;
+   case PIPE_SHADER_CAP_TGSI_DFRACEXP_DLDEXP_SUPPORTED:
    case PIPE_SHADER_CAP_TGSI_ANY_INOUT_DECL_RANGE:
       return 0;
    case PIPE_SHADER_CAP_MAX_SHADER_BUFFERS: