Mesa (main): aco: use ds_read_{u8,u16}_d16

[GitLab Mirror]

Module: Mesa
Branch: main
Commit: c129ede523820873e7d2ace73998e5842ec3e574
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=c129ede523820873e7d2ace73998e5842ec3e574

Author: Rhys Perry <pendingchaos02 at gmail.com>
Date:   Tue Jun  1 13:41:18 2021 +0100

aco: use ds_read_{u8,u16}_d16

This allows partial writes and writes to the upper half of the destination.

fossil-db (Sienna Cichlid):
Totals from 135 (0.09% of 149839) affected shaders:

Signed-off-by: Rhys Perry <pendingchaos02 at gmail.com>
Reviewed-by: Daniel Schürmann <daniel at schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11113>

---

 src/amd/compiler/aco_instruction_selection.cpp |  5 ++---
 src/amd/compiler/aco_opcodes.py                | 16 ++++++++--------
 2 files changed, 10 insertions(+), 11 deletions(-)

diff --git a/src/amd/compiler/aco_instruction_selection.cpp b/src/amd/compiler/aco_instruction_selection.cpp
index 0a2a66f4673..b985b5aad06 100644
--- a/src/amd/compiler/aco_instruction_selection.cpp
+++ b/src/amd/compiler/aco_instruction_selection.cpp
@@ -3668,7 +3668,6 @@ Temp lds_load_callback(Builder& bld, const LoadEmitInfo &info,
    bool read2 = false;
    unsigned size = 0;
    aco_opcode op;
-   //TODO: use ds_read_u8_d16_hi/ds_read_u16_d16_hi if beneficial
    if (bytes_needed >= 16 && align % 16 == 0 && large_ds_read) {
       size = 16;
       op = aco_opcode::ds_read_b128;
@@ -3691,10 +3690,10 @@ Temp lds_load_callback(Builder& bld, const LoadEmitInfo &info,
       op = aco_opcode::ds_read_b32;
    } else if (bytes_needed >= 2 && align % 2 == 0) {
       size = 2;
-      op = aco_opcode::ds_read_u16;
+      op = bld.program->chip_class >= GFX9 ? aco_opcode::ds_read_u16_d16 : aco_opcode::ds_read_u16;
    } else {
       size = 1;
-      op = aco_opcode::ds_read_u8;
+      op = bld.program->chip_class >= GFX9 ? aco_opcode::ds_read_u8_d16 : aco_opcode::ds_read_u8;
    }
 
    unsigned const_offset_unit = read2 ? size / 2u : 1u;
diff --git a/src/amd/compiler/aco_opcodes.py b/src/amd/compiler/aco_opcodes.py
index a4f2688fe89..7a1099d6909 100644
--- a/src/amd/compiler/aco_opcodes.py
+++ b/src/amd/compiler/aco_opcodes.py
@@ -1204,14 +1204,14 @@ DS = {
    (0x51, 0x51, 0x51, 0x51, 0x51, "ds_cmpst_f64"),
    (0x52, 0x52, 0x52, 0x52, 0x52, "ds_min_f64"),
    (0x53, 0x53, 0x53, 0x53, 0x53, "ds_max_f64"),
-   (  -1,   -1, 0x54, 0x54, 0xa0, "ds_write_b8_d16_hi"),
-   (  -1,   -1, 0x55, 0x55, 0xa1, "ds_write_b16_d16_hi"),
-   (  -1,   -1, 0x56, 0x56, 0xa2, "ds_read_u8_d16"),
-   (  -1,   -1, 0x57, 0x57, 0xa3, "ds_read_u8_d16_hi"),
-   (  -1,   -1, 0x58, 0x58, 0xa4, "ds_read_i8_d16"),
-   (  -1,   -1, 0x59, 0x59, 0xa5, "ds_read_i8_d16_hi"),
-   (  -1,   -1, 0x5a, 0x5a, 0xa6, "ds_read_u16_d16"),
-   (  -1,   -1, 0x5b, 0x5b, 0xa7, "ds_read_u16_d16_hi"),
+   (  -1,   -1,   -1, 0x54, 0xa0, "ds_write_b8_d16_hi"),
+   (  -1,   -1,   -1, 0x55, 0xa1, "ds_write_b16_d16_hi"),
+   (  -1,   -1,   -1, 0x56, 0xa2, "ds_read_u8_d16"),
+   (  -1,   -1,   -1, 0x57, 0xa3, "ds_read_u8_d16_hi"),
+   (  -1,   -1,   -1, 0x58, 0xa4, "ds_read_i8_d16"),
+   (  -1,   -1,   -1, 0x59, 0xa5, "ds_read_i8_d16_hi"),
+   (  -1,   -1,   -1, 0x5a, 0xa6, "ds_read_u16_d16"),
+   (  -1,   -1,   -1, 0x5b, 0xa7, "ds_read_u16_d16_hi"),
    (0x60, 0x60, 0x60, 0x60, 0x60, "ds_add_rtn_u64"),
    (0x61, 0x61, 0x61, 0x61, 0x61, "ds_sub_rtn_u64"),
    (0x62, 0x62, 0x62, 0x62, 0x62, "ds_rsub_rtn_u64"),