nvir: introduce OP_LOP3_LUT

Will be required to support SM70, but is also available on earlier GPUs.

v2:
- add convenience macro suggested by Karol

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5377>

diff --git a/src/gallium/drivers/nouveau/codegen/nv50_ir.h b/src/gallium/drivers/nouveau/codegen/nv50_ir.h
index 2718975b98a09e3293b5024ab0025947c5c5b848..a4ade030b008251a2df76337e98df0d2511ce48d 100644 (file)
--- a/src/gallium/drivers/nouveau/codegen/nv50_ir.h
+++ b/src/gallium/drivers/nouveau/codegen/nv50_ir.h
@@ -67,6 +67,7 @@ enum operation
    OP_AND,
    OP_OR,
    OP_XOR,
+   OP_LOP3_LUT,
    OP_SHL,
    OP_SHR,
    OP_MAX,
@@ -254,6 +255,15 @@ enum operation
 #define NV50_IR_SUBOP_VOTE_ALL 0
 #define NV50_IR_SUBOP_VOTE_ANY 1
 #define NV50_IR_SUBOP_VOTE_UNI 2
+#define NV50_IR_SUBOP_LOP3_LUT_SRC0 0xf0
+#define NV50_IR_SUBOP_LOP3_LUT_SRC1 0xcc
+#define NV50_IR_SUBOP_LOP3_LUT_SRC2 0xaa
+#define NV50_IR_SUBOP_LOP3_LUT(exp) ({         \
+      uint8_t a = NV50_IR_SUBOP_LOP3_LUT_SRC0; \
+      uint8_t b = NV50_IR_SUBOP_LOP3_LUT_SRC1; \
+      uint8_t c = NV50_IR_SUBOP_LOP3_LUT_SRC2; \
+      (uint8_t)(exp);                          \
+})
 
 #define NV50_IR_SUBOP_MINMAX_LOW  1
 #define NV50_IR_SUBOP_MINMAX_MED  2

diff --git a/src/gallium/drivers/nouveau/codegen/nv50_ir_peephole.cpp b/src/gallium/drivers/nouveau/codegen/nv50_ir_peephole.cpp
index e24e74a64637467524e7aec65ddb916209c2c080..5c3d15968cf4891b54ada546b75582ee58d64a1c 100644 (file)
--- a/src/gallium/drivers/nouveau/codegen/nv50_ir_peephole.cpp
+++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_peephole.cpp
@@ -780,6 +780,14 @@ ConstantFolding::expr(Instruction *i,
    memset(&res.data, 0, sizeof(res.data));
 
    switch (i->op) {
+   case OP_LOP3_LUT:
+      for (int n = 0; n < 32; n++) {
+         uint8_t lut = ((a->data.u32 >> n) & 1) << 2 |
+                       ((b->data.u32 >> n) & 1) << 1 |
+                       ((c->data.u32 >> n) & 1);
+         res.data.u32 |= !!(i->subOp & (1 << lut)) << n;
+      }
+      break;
    case OP_INSBF: {
       int offset = b->data.u32 & 0xff;
       int width = (b->data.u32 >> 8) & 0xff;

diff --git a/src/gallium/drivers/nouveau/codegen/nv50_ir_print.cpp b/src/gallium/drivers/nouveau/codegen/nv50_ir_print.cpp
index 5dcbf3c3e0c2449682dabb3489db5076ae0408cb..747382210807f1e103eeb1352705f41e1d558363 100644 (file)
--- a/src/gallium/drivers/nouveau/codegen/nv50_ir_print.cpp
+++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_print.cpp
@@ -93,6 +93,7 @@ const char *operationStr[OP_LAST + 1] =
    "and",
    "or",
    "xor",
+   "lop3 lut",
    "shl",
    "shr",
    "max",

diff --git a/src/gallium/drivers/nouveau/codegen/nv50_ir_target.cpp b/src/gallium/drivers/nouveau/codegen/nv50_ir_target.cpp
index 5c6d0570ae2b84516f6c609c3682949dbd5e1da0..92ac7101cfc2c9c4792399f3d48cd04e7b14ebdc 100644 (file)
--- a/src/gallium/drivers/nouveau/codegen/nv50_ir_target.cpp
+++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_target.cpp
@@ -33,7 +33,7 @@ const uint8_t Target::operationSrcNr[] =
    2, 2, 2, 2, 2, 3, 3, 3, // ADD, SUB, MUL, DIV, MOD, MAD, FMA, SAD
    3, 3,                   // SHLADD, XMAD
    1, 1, 1,                // ABS, NEG, NOT
-   2, 2, 2, 2, 2,          // AND, OR, XOR, SHL, SHR
+   2, 2, 2, 3, 2, 2,       // AND, OR, XOR, LOP3_LUT, SHL, SHR
    2, 2, 1,                // MAX, MIN, SAT
    1, 1, 1, 1,             // CEIL, FLOOR, TRUNC, CVT
    3, 3, 3, 2, 3, 3,       // SET_AND,OR,XOR, SET, SELP, SLCT
@@ -75,9 +75,9 @@ const OpClass Target::operationClass[] =
    OPCLASS_ARITH, OPCLASS_ARITH, OPCLASS_ARITH,
    OPCLASS_ARITH, OPCLASS_ARITH,
    OPCLASS_ARITH, OPCLASS_ARITH, OPCLASS_ARITH, OPCLASS_ARITH, OPCLASS_ARITH,
-   // ABS, NEG; NOT, AND, OR, XOR; SHL, SHR
+   // ABS, NEG; NOT, AND, OR, XOR, LOP3_LUT; SHL, SHR
    OPCLASS_CONVERT, OPCLASS_CONVERT,
-   OPCLASS_LOGIC, OPCLASS_LOGIC, OPCLASS_LOGIC, OPCLASS_LOGIC,
+   OPCLASS_LOGIC, OPCLASS_LOGIC, OPCLASS_LOGIC, OPCLASS_LOGIC, OPCLASS_LOGIC,
    OPCLASS_SHIFT, OPCLASS_SHIFT,
    // MAX, MIN
    OPCLASS_COMPARE, OPCLASS_COMPARE,