[Mesa-dev] [PATCH] panfrost/midgard: Add fround(_even), ftrunc, ffma

[Alyssa Rosenzweig]

These ops were discovered by invoking the correspondingly names GLSL
functions. The rounding ops here behave exact as expected and are mapped
to their corresponding NIR ops where applicable. The ffma behaves as a
LUT instruction and requires some special argument packing (since
Midgard normally only allows for 2 arguments); this quirk will be
addressed in the future, but for now FMA is still lowered.

Signed-off-by: Alyssa Rosenzweig <alyssa at rosenzweig.io>
---
 src/gallium/drivers/panfrost/midgard/helpers.h         | 4 ++++
 src/gallium/drivers/panfrost/midgard/midgard.h         | 8 ++++++++
 src/gallium/drivers/panfrost/midgard/midgard_compile.c | 2 ++
 3 files changed, 14 insertions(+)

diff --git a/src/gallium/drivers/panfrost/midgard/helpers.h b/src/gallium/drivers/panfrost/midgard/helpers.h
index 6940f27b4ab..12235f8f9ff 100644
--- a/src/gallium/drivers/panfrost/midgard/helpers.h
+++ b/src/gallium/drivers/panfrost/midgard/helpers.h
@@ -195,8 +195,12 @@ static unsigned alu_opcode_props[256] = {
         [midgard_alu_op_imin]		 = UNITS_MOST,
         [midgard_alu_op_imax]		 = UNITS_MOST,
         [midgard_alu_op_fmov]		 = UNITS_ALL | QUIRK_FLIPPED_R24,
+        [midgard_alu_op_fround]          = UNITS_ADD,
+        [midgard_alu_op_froundeven]      = UNITS_ADD,
+        [midgard_alu_op_ftrunc]          = UNITS_ADD,
         [midgard_alu_op_ffloor]		 = UNITS_ADD,
         [midgard_alu_op_fceil]		 = UNITS_ADD,
+        [midgard_alu_op_ffma]		 = UNIT_VLUT,
 
         /* Though they output a scalar, they need to run on a vector unit
          * since they process vectors */
diff --git a/src/gallium/drivers/panfrost/midgard/midgard.h b/src/gallium/drivers/panfrost/midgard/midgard.h
index 04e195a635d..03ac2f5597e 100644
--- a/src/gallium/drivers/panfrost/midgard/midgard.h
+++ b/src/gallium/drivers/panfrost/midgard/midgard.h
@@ -55,8 +55,11 @@ typedef enum {
         midgard_alu_op_fmin       = 0x28,
         midgard_alu_op_fmax       = 0x2C,
         midgard_alu_op_fmov       = 0x30,
+        midgard_alu_op_froundeven = 0x34,
+        midgard_alu_op_ftrunc     = 0x35,
         midgard_alu_op_ffloor     = 0x36,
         midgard_alu_op_fceil      = 0x37,
+        midgard_alu_op_ffma       = 0x38,
         midgard_alu_op_fdot3      = 0x3C,
         midgard_alu_op_fdot3r     = 0x3D,
         midgard_alu_op_fdot4      = 0x3E,
@@ -98,6 +101,7 @@ typedef enum {
         midgard_alu_op_u2f        = 0xBC,
         midgard_alu_op_icsel      = 0xC1,
         midgard_alu_op_fcsel      = 0xC5,
+        midgard_alu_op_fround     = 0xC6,
         midgard_alu_op_fatan_pt2  = 0xE8,
         midgard_alu_op_frcp       = 0xF0,
         midgard_alu_op_frsqrt     = 0xF2,
@@ -402,8 +406,11 @@ static char *alu_opcode_names[256] = {
         [midgard_alu_op_fmin]       = "fmin",
         [midgard_alu_op_fmax]       = "fmax",
         [midgard_alu_op_fmov]       = "fmov",
+        [midgard_alu_op_froundeven] = "froundeven",
+        [midgard_alu_op_ftrunc]     = "ftrunc",
         [midgard_alu_op_ffloor]     = "ffloor",
         [midgard_alu_op_fceil]      = "fceil",
+        [midgard_alu_op_ffma]       = "ffma",
         [midgard_alu_op_fdot3]      = "fdot3",
         [midgard_alu_op_fdot3r]     = "fdot3r",
         [midgard_alu_op_fdot4]      = "fdot4",
@@ -445,6 +452,7 @@ static char *alu_opcode_names[256] = {
         [midgard_alu_op_u2f]        = "u2f",
         [midgard_alu_op_icsel]      = "icsel",
         [midgard_alu_op_fcsel]      = "fcsel",
+        [midgard_alu_op_fround]     = "fround",
         [midgard_alu_op_fatan_pt2]  = "fatan_pt2",
         [midgard_alu_op_frcp]       = "frcp",
         [midgard_alu_op_frsqrt]     = "frsqrt",
diff --git a/src/gallium/drivers/panfrost/midgard/midgard_compile.c b/src/gallium/drivers/panfrost/midgard/midgard_compile.c
index a7da54c00f9..65822d3b28f 100644
--- a/src/gallium/drivers/panfrost/midgard/midgard_compile.c
+++ b/src/gallium/drivers/panfrost/midgard/midgard_compile.c
@@ -950,6 +950,8 @@ emit_alu(compiler_context *ctx, nir_alu_instr *instr)
                 ALU_CASE(imax, imax);
                 ALU_CASE(fmov, fmov);
                 ALU_CASE(ffloor, ffloor);
+                ALU_CASE(fround_even, froundeven);
+                ALU_CASE(ftrunc, ftrunc);
                 ALU_CASE(fceil, fceil);
                 ALU_CASE(fdot3, fdot3);
                 ALU_CASE(fdot4, fdot4);
-- 
2.20.1