Update patch set 1

Patch Set 1:

(9 comments)

Addressed all review comments in the latest patchset and added Apple M6 benchmark numbers to the CL description.

Patch-set: 1
This commit is contained in:
Gerrit User 1115898 2026-10-27 14:44:20 -07:00 committed by Gerrit Code Review
parent 3802406e7b
commit 94acd60f98

View File

@ -23,6 +23,30 @@
"revId": "80158557b1987cec9f6c4bf5904d87bbe518baa7",
"serverId": "3ce6091f-6c88-37e8-8c75-72f92ae8dfba"
},
{
"unresolved": false,
"key": {
"uuid": "787de970_eab27fb3",
"filename": "source/row_neon64.cc",
"patchSetId": 1
},
"lineNbr": 5797,
"author": {
"id": 1115898
},
"writtenOn": "2026-10-27T21:44:20Z",
"side": 1,
"message": "Done. Removed the 3 ins instructions and used v4.h[0]..v7.h[0] directly in umull/umull2.",
"parentUuid": "ee5a89e6_662caf99",
"range": {
"startLine": 5795,
"startChar": 0,
"endLine": 5797,
"endChar": 52
},
"revId": "80158557b1987cec9f6c4bf5904d87bbe518baa7",
"serverId": "3ce6091f-6c88-37e8-8c75-72f92ae8dfba"
},
{
"unresolved": true,
"key": {
@ -46,6 +70,30 @@
"revId": "80158557b1987cec9f6c4bf5904d87bbe518baa7",
"serverId": "3ce6091f-6c88-37e8-8c75-72f92ae8dfba"
},
{
"unresolved": false,
"key": {
"uuid": "ac603bac_e8a27125",
"filename": "source/row_neon64.cc",
"patchSetId": 1
},
"lineNbr": 5807,
"author": {
"id": 1115898
},
"writtenOn": "2026-10-27T21:44:20Z",
"side": 1,
"message": "Done. Replaced the 4 shrn/shrn2 instructions with 2 uzp2 .8h instructions.",
"parentUuid": "0f5cbbbc_e47452b5",
"range": {
"startLine": 5804,
"startChar": 7,
"endLine": 5807,
"endChar": 52
},
"revId": "80158557b1987cec9f6c4bf5904d87bbe518baa7",
"serverId": "3ce6091f-6c88-37e8-8c75-72f92ae8dfba"
},
{
"unresolved": true,
"key": {
@ -69,6 +117,30 @@
"revId": "80158557b1987cec9f6c4bf5904d87bbe518baa7",
"serverId": "3ce6091f-6c88-37e8-8c75-72f92ae8dfba"
},
{
"unresolved": false,
"key": {
"uuid": "49ab9cb2_4b4c5c4e",
"filename": "source/row_neon64.cc",
"patchSetId": 1
},
"lineNbr": 5836,
"author": {
"id": 1115898
},
"writtenOn": "2026-10-27T21:44:20Z",
"side": 1,
"message": "Done. Replaced mov + fmla with fmul + fadd in both the 4-pixel loop and 2-pixel tail.",
"parentUuid": "6cbc24bd_fead1b5d",
"range": {
"startLine": 5833,
"startChar": 0,
"endLine": 5836,
"endChar": 59
},
"revId": "80158557b1987cec9f6c4bf5904d87bbe518baa7",
"serverId": "3ce6091f-6c88-37e8-8c75-72f92ae8dfba"
},
{
"unresolved": true,
"key": {
@ -92,6 +164,30 @@
"revId": "80158557b1987cec9f6c4bf5904d87bbe518baa7",
"serverId": "3ce6091f-6c88-37e8-8c75-72f92ae8dfba"
},
{
"unresolved": false,
"key": {
"uuid": "800b4395_c905c3ef",
"filename": "source/row_neon64.cc",
"patchSetId": 1
},
"lineNbr": 5933,
"author": {
"id": 1115898
},
"writtenOn": "2026-10-27T21:44:20Z",
"side": 1,
"message": "Done. Converted both ARGBColorTableRow_NEON and RGBColorTableRow_NEON to use scalar GP register loads, bfi/lsr, and strh/strb.",
"parentUuid": "c6a82245_94abd1e9",
"range": {
"startLine": 5926,
"startChar": 3,
"endLine": 5933,
"endChar": 52
},
"revId": "80158557b1987cec9f6c4bf5904d87bbe518baa7",
"serverId": "3ce6091f-6c88-37e8-8c75-72f92ae8dfba"
},
{
"unresolved": true,
"key": {
@ -115,6 +211,30 @@
"revId": "80158557b1987cec9f6c4bf5904d87bbe518baa7",
"serverId": "3ce6091f-6c88-37e8-8c75-72f92ae8dfba"
},
{
"unresolved": false,
"key": {
"uuid": "26df9eed_9746249e",
"filename": "source/row_neon64.cc",
"patchSetId": 1
},
"lineNbr": 5980,
"author": {
"id": 1115898
},
"writtenOn": "2026-10-27T21:44:20Z",
"side": 1,
"message": "Yes, lumacoeff packs the 4 uint8_t coefficients [bc, gc, rc, 0] for one pixel into a uint32_t, so dup v4.4s broadcasts the 4-byte pattern across all 4 pixels in v4.16b. Added a clarifying comment.",
"parentUuid": "017bcaac_dbda25d5",
"range": {
"startLine": 5980,
"startChar": 0,
"endLine": 5980,
"endChar": 66
},
"revId": "80158557b1987cec9f6c4bf5904d87bbe518baa7",
"serverId": "3ce6091f-6c88-37e8-8c75-72f92ae8dfba"
},
{
"unresolved": true,
"key": {
@ -138,6 +258,30 @@
"revId": "80158557b1987cec9f6c4bf5904d87bbe518baa7",
"serverId": "3ce6091f-6c88-37e8-8c75-72f92ae8dfba"
},
{
"unresolved": false,
"key": {
"uuid": "701fe42e_75143d2c",
"filename": "source/row_neon64.cc",
"patchSetId": 1
},
"lineNbr": 5991,
"author": {
"id": 1115898
},
"writtenOn": "2026-10-27T21:44:20Z",
"side": 1,
"message": "Done. Replaced shrn + zip1 + zip1 with tbl using kLumaTblIdx.",
"parentUuid": "f4eb7572_69b09485",
"range": {
"startLine": 5989,
"startChar": 7,
"endLine": 5991,
"endChar": 52
},
"revId": "80158557b1987cec9f6c4bf5904d87bbe518baa7",
"serverId": "3ce6091f-6c88-37e8-8c75-72f92ae8dfba"
},
{
"unresolved": true,
"key": {
@ -161,6 +305,30 @@
"revId": "80158557b1987cec9f6c4bf5904d87bbe518baa7",
"serverId": "3ce6091f-6c88-37e8-8c75-72f92ae8dfba"
},
{
"unresolved": false,
"key": {
"uuid": "0d2bccf9_c0ff6cf4",
"filename": "source/row_neon64.cc",
"patchSetId": 1
},
"lineNbr": 5998,
"author": {
"id": 1115898
},
"writtenOn": "2026-10-27T21:44:20Z",
"side": 1,
"message": "Done. Used ldr b3, [%5] for the first byte load into v3 (followed by ld1 {v3.b}[...]) and merged the original alpha bytes from v0 via bit.",
"parentUuid": "a861e773_5905d7eb",
"range": {
"startLine": 5998,
"startChar": 0,
"endLine": 5998,
"endChar": 52
},
"revId": "80158557b1987cec9f6c4bf5904d87bbe518baa7",
"serverId": "3ce6091f-6c88-37e8-8c75-72f92ae8dfba"
},
{
"unresolved": true,
"key": {
@ -184,6 +352,30 @@
"revId": "80158557b1987cec9f6c4bf5904d87bbe518baa7",
"serverId": "3ce6091f-6c88-37e8-8c75-72f92ae8dfba"
},
{
"unresolved": false,
"key": {
"uuid": "7e2c3262_a7a278ae",
"filename": "source/row_neon64.cc",
"patchSetId": 1
},
"lineNbr": 6057,
"author": {
"id": 1115898
},
"writtenOn": "2026-10-27T21:44:20Z",
"side": 1,
"message": "Done. Replaced ldp + dup with a pair of ld1r {v2.2d} and ld1r {v6.2d}.",
"parentUuid": "e0740c7f_699c8bc4",
"range": {
"startLine": 6057,
"startChar": 0,
"endLine": 6057,
"endChar": 73
},
"revId": "80158557b1987cec9f6c4bf5904d87bbe518baa7",
"serverId": "3ce6091f-6c88-37e8-8c75-72f92ae8dfba"
},
{
"unresolved": true,
"key": {
@ -206,6 +398,30 @@
},
"revId": "80158557b1987cec9f6c4bf5904d87bbe518baa7",
"serverId": "3ce6091f-6c88-37e8-8c75-72f92ae8dfba"
},
{
"unresolved": false,
"key": {
"uuid": "8b64bbfd_ef3d547e",
"filename": "source/row_neon64.cc",
"patchSetId": 1
},
"lineNbr": 6083,
"author": {
"id": 1115898
},
"writtenOn": "2026-10-27T21:44:20Z",
"side": 1,
"message": "Done. Replaced the vector ldr + ins + str q0 sequence with scalar GP loads and stp %w3, %w4, [%0], #8.",
"parentUuid": "b586ab32_03848ea3",
"range": {
"startLine": 6082,
"startChar": 0,
"endLine": 6083,
"endChar": 52
},
"revId": "80158557b1987cec9f6c4bf5904d87bbe518baa7",
"serverId": "3ce6091f-6c88-37e8-8c75-72f92ae8dfba"
}
]
}