eessppeelllloo/ffmpeg
1
0
Fork 0
mirror of https://git.ffmpeg.org/ffmpeg.git synced 2026-06-16 04:32:47 +02:00
Code Issues Packages Projects Releases Wiki Activity

lavc/vulkan/common: allow configurable bitstream caching in shared memory

Browse Source
This commit is contained in:
averne
2025-12-14 23:01:45 +01:00
committed by Lynne
parent 4ac3b3a6da
commit a2475d16ed
1 changed files with 18 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files
libavcodec/vulkan/common.comp
+18 -15
View File
@@ -229,13 +229,15 @@ struct GetBitContext {
gb.bits_valid += 32; \
}
#else /* GET_BITS_SMEM */
shared u32vec4 gb_storage[gl_WorkGroupSize.x*gl_WorkGroupSize.y*gl_WorkGroupSize.z];
shared u32vec4 gb_storage[gl_WorkGroupSize.x*gl_WorkGroupSize.y*gl_WorkGroupSize.z*GET_BITS_SMEM];
#define FILL_SMEM() \
{ \
u32vec4buf ptr = u32vec4buf(gb.buf); \
gb_storage[gl_LocalInvocationIndex] = ptr[0].v; \
gb.cur_smem_pos = 0; \
#define FILL_SMEM() \
{ \
u32vec4buf ptr = u32vec4buf(gb.buf); \
[[unroll]] \
for (uint i = 0; i < GET_BITS_SMEM; ++i) \
gb_storage[gl_LocalInvocationIndex * GET_BITS_SMEM + i] = ptr[i].v; \
gb.cur_smem_pos = 0; \
}
#define LOAD64() \
@@ -251,15 +253,16 @@ shared u32vec4 gb_storage[gl_WorkGroupSize.x*gl_WorkGroupSize.y*gl_WorkGroupSize
FILL_SMEM(); \
}
#define RELOAD32() \
{ \
if (gb.cur_smem_pos >= 4) \
FILL_SMEM(); \
uint v = gb_storage[gl_LocalInvocationIndex][gb.cur_smem_pos]; \
gb.buf += 4; \
gb.bits = uint64_t(reverse4(v)) << (32 - gb.bits_valid) | gb.bits; \
gb.bits_valid += 32; \
gb.cur_smem_pos += 1; \
#define RELOAD32() \
{ \
if (gb.cur_smem_pos >= 4*GET_BITS_SMEM) \
FILL_SMEM(); \
u32vec4 vec = gb_storage[gl_LocalInvocationIndex * GET_BITS_SMEM + (gb.cur_smem_pos >> 2)]; \
uint v = vec[gb.cur_smem_pos & 3]; \
gb.buf += 4; \
gb.bits = uint64_t(reverse4(v)) << (32 - gb.bits_valid) | gb.bits; \
gb.bits_valid += 32; \
gb.cur_smem_pos += 1; \
}
#endif /* GET_BITS_SMEM */