vmalloc, kmalloc - 2.4.x
A message that is not in the archive (sent off-list or not yet received): 1044284924.2402.12.camel@gregs
A message that is not in the archive (sent off-list or not yet received): [email protected]
Quoted 3 lines · Alan Cox
Firstly vmalloc isnt permitted in interrupt context (use kmalloc with
GFP_KERNEL), although for such small chunks you might want to vmalloc a bigger buffer once at startup.
i've allso tried kmalloc with the same result.
Also, in this example it is timer - module isn't cleanly wroted becouse it supose to be only an example.
Signature · 2 lines
Grzegorz Jaskiewicz <[email protected]> K4 Labs
Quoted 3 lines · Alan Cox
Firstly vmalloc isnt permitted in interrupt context (use kmalloc with
GFP_KERNEL), although for such small chunks you might want to vmalloc a bigger buffer once at startup.
i've allso tried kmalloc with the same result.
Also, in this example it is timer - module isn't cleanly wroted becouse it supose to be only an example.
Signature · 2 lines
Grzegorz Jaskiewicz <[email protected]> K4 Labs
Quoted 8 lines · Grzegorz Jaskiewicz
Quoted 3 lines · Alan Cox
Firstly vmalloc isnt permitted in interrupt context (use kmalloc with
GFP_KERNEL), although for such small chunks you might want to vmalloc a bigger buffer once at startup.i've allso tried kmalloc with the same result.
Also, in this example it is timer - module isn't cleanly wroted becouse it supose to be only an example.
If I build the example using the cleanups I suggested it works for me. The FPU one btw seems to be a red herring, my gcc is outputting a precomputed integer constant
Alan
Quoted 2 lines · Alan Cox
If I build the example using the cleanups I suggested it works for me. The FPU one btw seems to be a red herring, my gcc is outputting a precomputed integer constant
This is purely coincidence. GCC is perfectly entitled to use integer arithmetic even though you used floats in the source.
GCC is likewise perfectly entitled to use floating point even if you only used integers in the source. There's a good reason why the SH port builds with '-mno-implicit-fp' and why all other ports should have this _before_ it becomes a problem rather than afterwards.
Signature · 1 line
dwmw2
Quoted 4 lines · David Woodhouse
GCC is likewise perfectly entitled to use floating point even if you only used integers in the source. There's a good reason why the SH port builds with '-mno-implicit-fp' and why all other ports should have this _before_ it becomes a problem rather than afterwards.
I'm sorry, but it seems you forgot to use the -u option in your diff.
Quoted 4 lines · David Woodhouse
GCC is likewise perfectly entitled to use floating point even if you only used integers in the source. There's a good reason why the SH port builds with '-mno-implicit-fp' and why all other ports should have this _before_ it becomes a problem rather than afterwards.
I was wondering about this yesterday when toying with the -march options we now pass. With (for eg) -march=c3, we now tell gcc it can emit 3dnow instructions if it wants, likewise SSE/SSE2 in other -march options.
Dave
Signature · 2 lines
| Dave Jones. http://www.codemonkey.org.uk | SuSE Labs
Quoted 8 lines · Grzegorz Jaskiewicz
Quoted 3 lines · Alan Cox
Firstly vmalloc isnt permitted in interrupt context (use kmalloc with
GFP_KERNEL), although for such small chunks you might want to vmalloc a bigger buffer once at startup.i've allso tried kmalloc with the same result.
Also, in this example it is timer - module isn't cleanly wroted becouse it supose to be only an example.
I suppose you meant kmalloc with GFP_ATOMIC...
Signature · 1 line
Benjamin Herrenschmidt <[email protected]>